Dcp2 C-terminal cis-binding elements control selective targeting of the decapping enzyme by forming distinct decapping complexes

  1. Feng He  Is a corresponding author
  2. Chan Wu
  3. Allan Jacobson  Is a corresponding author
  1. Department of Microbiology and Physiological Systems, University of Massachusetts Chan Medical School, United States

Abstract

A single Dcp1–Dcp2 decapping enzyme targets diverse classes of yeast mRNAs for decapping-dependent 5′ to 3′ decay, but the molecular mechanisms controlling mRNA selectivity by the enzyme remain elusive. Through extensive genetic analyses we reveal that Dcp2 C-terminal domain cis-regulatory elements control decapping enzyme target specificity by orchestrating formation of distinct decapping complexes. Two Upf1-binding motifs direct the decapping enzyme to nonsense-mediated mRNA decay substrates, a single Edc3-binding motif targets both Edc3 and Dhh1 substrates, and Pat1-binding leucine-rich motifs target Edc3 and Dhh1 substrates under selective conditions. Although it functions as a unique targeting component of specific complexes, Edc3 is a common component of multiple complexes. Scd6 and Xrn1 also have specific binding sites on Dcp2, allowing them to be directly recruited to decapping complexes. Collectively, our results demonstrate that Upf1, Edc3, Scd6, and Pat1 function as regulatory subunits of the holo-decapping enzyme, controlling both its substrate specificity and enzymatic activation.

Editor's evaluation

This paper is an important contribution to the fields of mRNA decapping and mRNA decay. Using a series of elegant genetic approaches and assays, this study tackles the difficult challenge of mapping binding interactions and subsequently assigning functions to binding partners involved in mRNA decapping. As a result, the findings reported in this work represent an important milestone towards deepening our mechanistic understanding of mRNA decapping and mRNA decay.

https://doi.org/10.7554/eLife.74410.sa0

Introduction

Decapping, the removal of mRNA 5′ cap structures, is a fundamental step in eukaryotic mRNA turnover that commits a transcript to complete 5′ to 3′ exoribonucleolytic digestion by Xrn1 (Grudzien-Nogalska and Kiledjian, 2017; Parker, 2012). Decapping plays a key role in general 5′ to 3′ mRNA decay (Decker and Parker, 1993), nonsense-mediated mRNA decay (NMD) (He and Jacobson, 2001), AU-rich element-mediated mRNA decay (Fenger-Grøn et al., 2005; Pedro-Segura et al., 2008), microRNA-mediated gene silencing (Behm-Ansmant et al., 2006; Rehwinkel et al., 2005), and transcript-specific degradation (Badis et al., 2004; Dong et al., 2007). In Saccharomyces cerevisiae, mRNA decapping is carried out by a single enzyme comprised of the Dcp1 regulatory subunit and the Dcp2 catalytic subunit. Dcp1 is a small EVH domain protein (Beelman et al., 1996; She et al., 2004) and Dcp2 is a 970-amino acid protein containing a highly conserved catalytic domain at its N-terminus and a largely disordered C-terminal domain embedded with multiple regulatory elements (Charenton et al., 2017; Dunckley and Parker, 1999; Gaudon et al., 1999; He and Jacobson, 2015).

In addition to the Dcp1–Dcp2 decapping enzyme, yeast mRNA decapping is also regulated by several proteins collectively known as decapping activators, including Upf1, Pat1, Lsm1, Dhh1, Scd6, Edc1, Edc2, and Edc3 (Parker, 2012). Upf1 is required for degradation of nonsense-containing mRNAs (He and Jacobson, 2001; He et al., 2003). Pat1, Lsm1, and Dhh1 were originally thought to be required for general mRNA decay (Bouveret et al., 2000; Coller et al., 2001; Fischer and Weis, 2002; Hatfield et al., 1996; Tharun et al., 2000), but recent evidence indicates that they each target a specific subset of yeast mRNAs (He et al., 2018). Scd6 targets a limited number of specific transcripts for decapping (Zeidan et al., 2018). Edc3 was also originally thought to be required for general mRNA decapping (Kshirsagar and Parker, 2004), but this factor exhibits exquisite substrate specificity and appears to target just two transcripts, YRA1 pre-mRNA and RPS28B mRNA (Badis et al., 2004; Dong et al., 2007). Edc1 and Edc2 were isolated as high-copy suppressors of dcp1 and dcp2 mutations (Dunckley et al., 2001), and these two factors can stimulate mRNA decapping in vitro (Borja et al., 2011; Steiger et al., 2003), but they do not appear to be required for mRNA decapping in vivo at least under normal growth conditions. Whether Edc1 and Edc2 also target-specific mRNAs have not been investigated. Several specific functions, including translation repression, decapping enzyme activation, and codon optimality sensing have been proposed for decapping activators (Coller and Parker, 2005; Nissan et al., 2010; Radhakrishnan et al., 2016), but each of the proposed functions is subject to sufficient controversy or uncertainty to render their actual roles largely unknown (Arribere et al., 2011; He and Jacobson, 2015; Sweet et al., 2012; Webster et al., 2018).

Mechanistic investigations of mRNA decapping over the last two decades have mostly used biochemical and structural approaches and focused on the catalytic mechanisms of the decapping enzyme (Charenton and Graille, 2018; Valkov et al., 2017). These studies provided significant insights into the structures of the Dcp1–Dcp2 decapping enzyme (She et al., 2006; She et al., 2004; She et al., 2008), its conformational dynamics (Floor et al., 2012; Wurm et al., 2017), cap- and RNA-binding properties (Deshmukh et al., 2008; Floor et al., 2010), and catalytic mechanisms (Aglietti et al., 2013). These experiments also revealed binding patterns of Edc1, Edc3, and Pat1 to the decapping enzyme and suggested potential functions or mechanisms of action of these factors in enzymatic activation (Charenton et al., 2017; Charenton et al., 2016; Fromm et al., 2012; Mugridge et al., 2018; Paquette et al., 2018; Valkov et al., 2016; Paquette et al., 2018; Valkov et al., 2016). However, these biochemical and structural studies all used C-terminally truncated Dcp2, isolated peptides or domains from the decapping activators, and generic mRNA substrates. Thus, some of the proposed mechanisms for decapping activation that emerged from these studies may need additional validation. Further, these experiments did not address how the decapping enzyme is targeted to different substrate mRNAs.

Our recent genetic experiments demonstrated that the Dcp2 C-terminal domain plays a crucial role in the control of mRNA decapping (He et al., 2018; He and Jacobson, 2015). This domain contains an inhibitory element (IE) and a set of specific linear binding elements for several decapping activators, including Edc3 (E3), Upf1 (U11, U12), and Pat1 (L1–L9) (Figure 1A). Here, we have generated specific DCP2 deletions that eliminated either a single element or combinations of different elements and analyzed the consequences of these deletions on Dcp2 interactions with specific decapping activators and decay of different decapping substrates. Our experiments uncover the molecular mechanisms that control the selective targeting of the yeast decapping enzyme and reveal the functional contributions of Upf1, Edc3, Scd6, Dhh1, and Pat1 in decapping of their respective mRNA targets.

Figure 1 with 2 supplements see all
Consequences of deleting Dcp2’s Edc3- and Upf1-binding motifs, and its inhibitory element.

(A) Dcp2 schematic depicting its distinct regulatory elements. D1, Dcp1-binding site; E3, Edc3-binding motif; U11 and U12, Upf1-binding motifs; L1–L9, leucine-rich Pat1-binding motifs; IE, the inhibitory element. (B) Two-hybrid assays evaluating the consequences of specific dcp2 deletions on Dcp2 interactions with Upf1, Edc3, and Pat1. Left, schematics of deletion alleles, with specific element deletions marked by triangles. The letter D in the allele names is used to indicate that a specific element has been deleted, such that E3D denotes deletion of the E3 element, U1D1 denotes deletion of the U11 element, etc. Right, two-hybrid colony color assays, duplicated horizontally, with blue color indicating interaction and white indicating no interaction. (C) Northern analyses of individual transcript levels in cells expressing specific dcp2 deletion alleles or harboring deletions of UPF1 or EDC3. Groupings depict transcripts subject to common regulation. (D) Northern analyses of Edc3 and Dhh1 substrates in cells expressing dcp2 alleles deleted for E3-1 or E3-2, or harboring deletions of EDC3 or DHH1. Further analyses of the transcripts considered in (C) and (D) are presented in Figure 1—figure supplement 1 and Figure 1—figure supplement 2. In all blots, lower case letters denote SCR1 blots duplicated for clarity of presentation.

Figure 1—source data 1

Northern analyses of Edc3 and Dhh1 substrates in cells harboring the E3-1 or E3-2 deletions of DCP2 or deletions of EDC3 or DHH1 (Figure 1D).

https://cdn.elifesciences.org/articles/74410/elife-74410-fig1-data1-v2.zip

Results

Loss of both Upf1-binding motifs eliminates Upf1’s binding to Dcp2 and causes selective partial stabilization of NMD substrates

Wild-type (WT) Dcp2 exhibited a strong two-hybrid interaction with Upf1. Deletion of either the first (U1D1) or the second (U1D2) Upf1-binding motif had no discernible effect on Upf1’s binding to Dcp2, but loss of both Upf1-binding motifs (U1D1–U1D2) eliminated Upf1 binding to Dcp2 (Figure 1B). In contrast, loss of both Upf1-binding motifs affected neither Edc3 nor Pat1 binding to Dcp2 (Figure 1B). These results indicate that the two Upf1-binding motifs have redundant functions and either can promote independent Upf1 binding to Dcp2.

WT yeast expressed low levels of NMD substrates, including the CYH2 pre-mRNA and the ade2-1, can1-100, and trp1-1 mRNAs. Loss of either the first or the second Upf1-binding motif had no significant effect on the levels of accumulation for each of these NMD substrates. However, loss of both Upf1-binding motifs caused approximately two- to threefold increases in the levels of these transcripts (Figure 1C, Figure 1—figure supplement 1A, B, and Figure 1—figure supplement 2, HA-dcp2 alleles containing U1D1–U1D2). These increases were much smaller in magnitude than those caused by deletion of UPF1, which usually led to >10-fold increases for these transcripts. Loss of both Upf1-binding motifs did not affect the levels of the Edc3, Dhh1, Pat1/Lsm1, and Pat1/Lsm1/Dhh1 substrates (Figure 1C, Figure 1—figure supplement 1A, B, and Figure 1—figure supplement 2). This selective stabilization of nonsense-containing mRNAs indicates that the two Upf1-binding motifs control targeting of the decapping enzyme to NMD substrates. In addition, the two motifs have independent activities and are functionally redundant in promoting NMD. The partial stabilization of NMD substrates caused by loss of both Upf1-binding motifs suggests that decapping is not a major rate-limiting step in the overall NMD pathway. It also appears that NMD substrates can be degraded by an alternative route in the absence of active recruitment of the decapping enzyme by Upf1.

Loss of the Edc3-binding motif eliminates Edc3 binding to Dcp2 and causes selective stabilization of both Edc3 and Dhh1 substrates

WT Dcp2 also exhibited strong two-hybrid interaction with Edc3. Loss of E3 (E3D) did not affect the binding of Upf1 or Pat1 to Dcp2 but eliminated Edc3 binding to Dcp2 (Figure 1B), indicating that the E3 motif promotes selective binding of Edc3 to the decapping enzyme. Loss of the E3 motif did not affect the levels of NMD and Pat1/Lsm1 substrates, but did lead to stabilization of the Edc3 substrates, RPS28B mRNA and YRA1 pre-mRNA, to different extents (Figure 1C, Figure 1—figure supplement 1A, B, and Figure 1—figure supplement 2, HA-dcp2-E3D). Compared to their fold increases in edc3Δ cells, loss of the Edc3-binding motif caused complete stabilization of RPS28B mRNA (~twofold increases in both E3D and edc3Δ cells), but only partial stabilization of YRA1 pre-mRNA (threefold increase in E3D vs. ninefold increase in edc3Δ cells). The different effects of loss of the E3 motif may be indicative of different roles of Edc3 in decapping of these transcripts. Decapping of RPS28B mRNA is likely rate limiting and totally dependent on Edc3-mediated recruitment of the decapping enzyme. The principal function of Edc3 in RPS28B mRNA decay would thus be to recruit the decapping enzyme and, when this recruitment is blocked by E3 deletion, RPS28B mRNA is not degraded efficiently by an alternative pathway. In contrast, Edc3-mediated recruitment of the decapping enzyme may contribute only partially to the overall decapping process of YRA1 pre-mRNA and Edc3 may play an additional role in the decay of this transcript. Also, when Edc3-mediated recruitment of the decapping enzyme is blocked by E3 deletion, YRA1 pre-mRNA appears to be degraded by an alternative route, albeit less efficiently. Both possibilities were validated by experiments described below.

Although deletion of EDC3 did not affect the levels of Dhh1 substrates, loss of the E3 motif caused partial stabilization of these mRNAs (Figure 1C, Figure 1—figure supplement 2, compare edc3Δ vs. HA-dcp2-E3D; He et al., 2018). This result indicates that, in addition to Edc3 substrates, E3 also controls targeting of the decapping enzyme to Dhh1-regulated mRNAs. Given the disparate mRNA decay phenotypes caused by edc3Δ and E3D for Dhh1 substrates, the E3-mediated targeting the decapping enzyme to Dhh1 substrates is unlikely to be carried out solely via interaction with Edc3. At least in the absence of Edc3, one additional factor may bind E3 (see below for Scd6 data) and target the decapping enzyme to Dhh1-regulated mRNAs. A role for E3 in the decapping of Dhh1 substrates may appear surprising, but physical interactions between Edc3 and Dhh1 have been reported (He and Jacobson, 2015; Sharif et al., 2013). The partial stabilization of Dhh1 substrates caused by loss of E3 can be explained similarly to that described above for YRA1 pre-mRNA, that is, recruitment of the decapping enzyme makes a small kinetic contribution to overall decay and degradation of these substrates can occur by an alternative route.

The dcp2 E3D allele contains a deletion of a conserved 37-codon segment that we identified previously as encoding the Edc3-binding element (He and Jacobson, 2015). Phylogenetic sequence comparisons suggested that this segment may encode composite-binding elements (Figure 1—figure supplement 1A). To map this region further, we generated two smaller deletions in the originally defined element. E3D1 eliminates the first part of the element (17 amino acids, E3-1) and E3D2 eliminates the second part (20 amino acids, E3-2). Loss of E3-2 did not affect Edc3 binding to Dcp2, but loss of E3-1 eliminated Edc3’s binding to Dcp2 (Figure 1B), indicating that Edc3 binds to the conserved 17-amino acid segment. Loss of E3-1 also caused complete stabilization of RPS28B mRNA and partial stabilization of YRA1 pre-mRNA and Dhh1-regulated mRNAs (Figure 1D, Figure 1—figure supplement 1C), suggesting that the E3-1 element controls selective targeting of the decapping enzyme to both Edc3 and Dhh1 substrates.

Loss of the Pat1-binding motifs eliminates Pat1 binding to Dcp2, but has no effect on levels of Pat1/Lsm1-regulated mRNAs

We next focused on the nine helical leucine-rich motifs L1–L9, each of which, except L8, was shown to bind Pat1 in yeast two-hybrid and GST-pulldown assays (Charenton et al., 2017; He and Jacobson, 2015). However, whether these eight motifs are all engaged in Pat1 binding and whether they promote independent or collaborative binding of Pat1 to Dcp2 in the context of full-length protein or decapping complex is unknown. WT Dcp2 exhibited strong two-hybrid interaction with Pat1 and loss of the first five leucine-rich motifs (L1–L5) eliminated this interaction (Figure 2A, alleles LD1-5 to LD1-9). In contrast, loss of the last four leucine-rich motifs (L9–L6) did not affect Pat1 binding to Dcp2 (Figure 2A, allele L9-6). Consecutive deletions between L1 and L5 from either the N-terminus or the C-terminus yielded graded responses for Pat1 binding. From the N-terminus, loss of L4 weakened and further loss of L5 eliminated Pat1 binding (Figure 2A, alleles L1–4 and L1–5). From the C-terminus, loss of L4 and L3 greatly weakened and further loss of L2 eliminated Pat1 binding (Figure 2A, alleles LD9-4, LD9-3, and LD9-2). None of the leucine-rich element deletions affected the binding of Edc3 or Upf1 to Dcp2 (Figure 2A). These results indicate that in the context of full-length Dcp2, leucine-rich motifs L1–L5 control the selective binding of Pat1 to Dcp2, most likely with a contribution from each motif. It is also possible that leucine-rich motifs L6 to L9 collaborate with L4 and L5 to promote the selective binding of Pat1 to Dcp2. Further, if Pat1 only uses its C-terminal binding domain to engage Dcp2, as demonstrated in a recent structural study (Charenton et al., 2017), then the multiple-motif requirement for Pat1’s binding to Dcp2 may indicate that Pat1 binds to full-length Dcp2 using allovalency (Klein et al., 2003).

Figure 2 with 1 supplement see all
Consequences of deleting Dcp2 leucine-rich Pat1-binding motifs.

(A) Two-hybrid assays evaluating the consequences of deleting the leucine-rich motifs (L1–L9) from the Dcp2 C-terminal domain. As in Figure 1B, schematics of the individual dcp2 alleles are shown on the left (with specific element deletions denoted by triangles) and duplicate two-hybrid assays are on the right. (B) Northern analyses of individual Pat1/Lsm1 and Pat1/Lsm1/Dhh1 substrate levels in cells harboring individual or combined deletions of Dcp2 leucine-rich motifs. (C) Bar graphs of average ± SEM for a subset of the northern analyses depicted in B. The relative levels of each decapping substrate in different strains were determined from three independent experiments, with one representative blot for each transcript shown in panel B. See also Figure 2—figure supplement 1.

Consistent with the selective targeting of the decapping enzyme by both the Edc3- and Upf1-binding motifs, none of the leucine-rich element deletions affected the levels of the Edc3, Dhh1, or NMD substrates (Figure 2B, Figure 2—figure supplement 1C). Surprisingly, none of the leucine-rich element deletions altered the levels of Pat1-regulated mRNAs, including both Pat1/Lsm1 and Pat1/Lsm1/Dhh1 substrates (Figure 2B, C). This indicates that Pat1-mediated targeting of the decapping enzyme does not make a significant contribution to the overall decay of Pat1-regulated transcripts, raising a question about the role of Pat1 in decapping. Because loss of Pat1 causes significant stabilization of both the Pat1/Lsm1 and Pat1/Lsm1/Dhh1 substrates (Figure 2B; He et al., 2018), one possible explanation for this surprising observation is that decapping is not rate limiting for Pat1-regulated mRNAs, and Pat1 performs an unidentified major function upstream, independent of its role in the recruitment and activation of the decapping enzyme (Lobel et al., 2019; Nissan et al., 2010).

In the absence of active recruitment of the decapping enzyme, different decapping substrates are still degraded by decapping-dependent 5′ to 3′ decay

As described above, the Upf1-, Edc3-, and Pat1-binding motifs in Dcp2 promote selective binding of these factors to the decapping enzyme. Yet, loss of both Upf1-binding motifs in Dcp2 led to only partial stabilization of NMD substrates and loss of the Edc3-binding motif resulted in partial stabilization of Edc3 and Dhh1 substrates except for the RPS28 mRNA. Further, loss of the Pat1-binding motifs did not cause significant stabilization of Pat1/Lsm1 or Pat1/Lsm1/Dhh1 substrates. The little or no effect of these deletions on different decapping substrates raised the question of decay mechanism for these mRNAs in the absence of active decapping enzyme recruitment. Two possibilities for their decay in the absence of active recruitment of the decapping enzyme seemed likely: degradation by a 3′ to 5′ exosome-dependent decay mechanism or decapping by an alternative route but overall degradation by decapping-dependent 5′ to 3′ decay. To distinguish between these possibilities, we generated double mutant cells that combine the dcp2 cis element deletions U1D1–U1D2, E3D or E3D1, and LD1-9 (Figure 3A) with deletions of key genes required for 5′ to 3′ decay (XRN1) or 3′ to 5′ decay (SKI2 or SKI7) and analyzed the mRNA decay phenotypes of different decapping substrates in the resulting ‘double’ mutant cells (Figure 3B, C).

Figure 3 with 1 supplement see all
Nonsense-mediated mRNA decay (NMD), Edc3, Dhh1, and Pat1 substrates are still degraded by decapping-dependent 5′ to 3′ decay in the absence of active decapping enzyme recruitment.

(A) Schematics of dcp2 alleles that eliminate Edc3, Upf1, or Pat1 binding to Dcp2. (B) Deletion of XRN1, but not deletions of SKI2 or SKI7, causes significant stabilization of NMD substrates in HA-dcp2-U1D1-U1D2 cells, and Edc3 and Dhh1 substrates in HA-dcp2-E3D or E3D1 cells. (C) Deletion of XRN1, but not deletions of SKI2 or SKI7, causes significant stabilization of Pat1 substrates in HA-dcp2-LD1-9 cells. Northern analyses in B and C as in Figures 1 and 2. Bar graphs in lower panels of B and C depict relative levels of decapping substrates in different strains determined from average ± SEM of three independent experiments. One representative northern blot for each transcript is shown in the upper panels. In the upper panel of B, lower case letters denote SCR1 blots duplicated for clarity of presentation. See also Figure 3—figure supplement 1.

Figure 3—source data 1

Northern analyses of different nonsense-mediated mRNA decay (NMD), Edc3, and Dhh1 substrates in cells harboring the HA-dcp2-U1D1-U1D2, E3D, or E3D1 alleles and deletions of XRN1, SKI2, or SKI7 (Figure 3B).

https://cdn.elifesciences.org/articles/74410/elife-74410-fig3-data1-v2.zip
Figure 3—source data 2

Northern analyses of different Pat1 substrates in cells harboring the HA-dcp2-LD1-9 allele and deletions of XRN1, SKI2, or SKI7 (Figure 3C).

https://cdn.elifesciences.org/articles/74410/elife-74410-fig3-data2-v2.zip

In dcp2-U1D1-U1D2 cells, deletion of SKI2 or SKI7 did not cause additional stabilization of the NMD substrates (CYH2 pre-mRNA and can1-100 mRNA) whereas deletion of XRN1 caused substantial stabilization of these transcripts (Figure 3B). Similarly, deletion of SKI2 or SKI7 did not cause additional stabilization of both Edc3 and Dhh1 substrates, for example, YRA1 pre-mRNA, EDC1, and SDS23 mRNAs, in dcp2-E3D or E3D1 cells, but deletion of XRN1 caused substantial stabilization of each of these transcripts (Figure 3B, Figure 3—figure supplement 1). Deletions of SKI2 or SKI7 also did not cause discernible stabilization of the Pat1/Lsm1 substrates, BUR6, DIF1, and LSM3 mRNAs, in dcp2-LD1-9 cells, but deletion of XRN1 stabilized these mRNAs substantially (Figure 3C). In each case, deletion of XRN1 in the respective dcp2 cis element mutant cells yielded similar fold increases in transcript levels as those caused by XRN1 deletion in DCP2 WT cells (Figure 3B, C). These results indicate that, in the absence of active recruitment of the decapping enzyme, the decapping substrates examined here, including the NMD, Edc3, Dhh1, and Pat1/Lsm1 substrates, are not degraded by exosome-dependent 3′ to 5′ decay, but are all still degraded by decapping-dependent 5′ to 3′ decay. Hence, these decapping substrates can all be decapped by an alternative route when the normal Dcp2 cis-element-mediated active recruitment of the decapping enzyme is blocked.

In the absence of active Edc3 recruitment of the decapping enzyme, Edc3 and Dhh1 substrates are degraded by Pat1-mediated decay

The experiments of Figure 3 demonstrated that different decapping substrates can be decapped by an alternative route when recruitment of the decapping enzyme to these mRNAs is blocked by specific dcp2 cis-element deletions. To determine whether the Edc3-, Upf1-, and Pat1-mediated decapping pathways have redundant activities or may function as backup systems for each other, we constructed a third set of dcp2 alleles harboring different combinations of element deletions that eliminate the binding of Edc3 (E3D or E3D1), Upf1 (U1D1–U1D2), and Pat1 (LD1-8, LD1-9, or LD9-2) to Dcp2 (Figure 4A and Figure 4—figure supplements 1A; 3A). In addition, we also included two inconsequential Upf1-binding element deletions (U1D1 or U1D2) and one partially functional Pat1-binding element deletion (LD9-3) in this set.

Figure 4 with 5 supplements see all
Genetic interactions between the Dcp2 Edc3-, Upf1-, and Pat1-binding motifs, or between the Edc3-binding motif and Edc3, that affect mRNA decapping.

(A) Two-hybrid assays examining the effects of different combinations of element deletions on Edc3, Upf1, and Pat1 binding to Dcp2. Allele schematics and two hybrid analyses are as in Figure 1B. (B) Northern analyses of the consequences of simultaneous loss of the Dcp2 Edc3-binding motif and leucine-rich motifs. See also Figure 4—figure supplement 1 and Figure 4—figure supplement 2. (C) Northern analyses of the consequences of simultaneous loss of the Dcp2 Upf1-binding motifs and leucine-rich motifs, and the roles of single Dcp2 Edc3 or Upf1-binding motifs. See also Figure 4—figure supplement 3 and Figure 4—figure supplement 4. (D) Northern analyses of the consequences for Edc3 and Dhh1 substrates caused by loss of the Dcp2 Edc3-binding motif and trans deletion of EDC3. See also Figure 4—figure supplement 5. Northern analyses as in Figures 1 and 2. Two-hybrid analyses as in Figure 1.

Figure 4—source data 1

Northern analyses of the consequences for Edc3 and Dhh1 substrates caused by loss of the Dcp2 Edc3-binding motif and trans deletion of EDC3 (Figure 4D).

https://cdn.elifesciences.org/articles/74410/elife-74410-fig4-data1-v2.zip

Each of these alleles yielded the expected binding patterns of Edc3, Upf1, and Pat1 to Dcp2 except the four alleles that contain single Upf1-binding motif deletions (Figure 4A, alleles E3D-U1D1-LD1-9, E3D-U1D2-LD1-9, E3D1-U1D1-LD1-9, E3D1-U1D2-LD1-9). The mutant Dcp2 proteins generated from these four alleles showed no two-hybrid interaction with Upf1 (Figure 4A), but appeared fully functional in NMD (see below). As both Upf1-binding motifs border the leucine-rich motifs, we suspect that the LD1-9 deletion may indirectly affect the conformation of these two binding motifs and diminish Upf1 binding to Dcp2 in the two-hybrid assay.

Although none of the leucine-rich element deletions affected the decay of Edc3 and Dhh1 substrates (Figure 2B), combining the leucine-rich element deletions LD1-8, LD1-9, LD9-3, and LD9-2 with the Edc3-binding element deletions E3D or E3D1 caused additional substantial stabilization of the Edc3 substrate YRA1 pre-mRNA and the Dhh1 substrates EDC1, SDS23, and HXT6 mRNAs (Figure 4B, Figure 4—figure supplement 2). None of the combinations of these element deletions affected the Pat1/Lsm1 substrates or the NMD substrates (Figure 4B, Figure 4—figure supplement 2). These results indicate that in the absence of active recruitment of the decapping enzyme by Edc3, both Edc3 and Dhh1 substrates are degraded by the Pat1-mediated decay pathway, suggesting that Pat1-mediated decapping serves as a backup or fail-safe system in the decay of Edc3 and Dhh1-regulated mRNAs. We also noticed a subtle functional difference for these dcp2 alleles. In otherwise the same deletion context, cells harboring E3D1 consistently had lower transcript levels than those harboring E3D for Edc3 and Dhh1 substrates (Figure 4B). These differences could be indicative of E3-2 element function in decay of these mRNAs, as E3D eliminates both E3-1 and E3-2 elements and E3D1 eliminates only the E3-1 element. Similarly, cells harboring LD9-3 also had consistently lower transcript levels than those harboring LD9-2 for both Edc3 and Dhh1 substrates, suggesting that LD9-3 deletion maintains more function of Dcp2 than that of LD9-2 in decay of these mRNAs. Consistent with this interpretation, LD9-3 weakens but LD9-2 eliminates Pat1 binding to Dcp2 (Figure 4A, compare alleles E3D-LD9-2 to E3D-LD9-3, and E3D1-LD9-2 to E3D1-LD9-3).

Combining the leucine-rich element deletion LD1-9 with the Upf1-binding element deletion U1D1–U1D2 did not cause additional stabilization of the NMD substrates CYH2 pre-mRNA, can1-100, and ade2-1 mRNAs (Figure 4C, Figure 4—figure supplement 4, allele U1D1-U1D2-LD1-9 compared to allele U1D1-U1D2 in Figure 1B). Further combining the Edc3-binding motif deletions E3D or E3D1 also did not cause additional stabilization of the CYH2 pre-mRNA and can1-100 mRNA, but did appear to cause small but significant increases in ade2-1 mRNA levels (Figure 4C, Figure 4—figure supplement 4, compare allele U1D1-U1D2-LD1-9 to alleles E3D-U1D1-U1D2-LD1-9 and E3D1-U1D1-U1D2-LD1-9). These results indicate that Pat1-mediated decapping either does not function as a backup system or does not make a significant contribution to the decay of NMD substrates in the absence of Upf1-mediated recruitment of the decapping enzyme. However, Edc3-mediated decapping may function as a backup system for some nonsense-containing mRNAs. Further, the combination of element deletions E3D-U1D1-U1D2-LD1-9 eliminates all the binding motifs of known decapping activators in Dcp2, yet this combination of deletions still did not have any significant effect on the levels of accumulation of the Pat1/Lsm1 substrates (Figure 4C, Figure 4—figure supplement 4). This result indicates that neither Edc3-mediated decapping nor Upf1-mediated decapping functions as a backup system in decay of Pat1/Lsm1 substrates, raising the possibility that Pat1/Lsm1 substrates can be decapped without the enhancement function of any decapping activators.

The combination of element deletions U1D1-U1D2-LD1-9 created a dcp2 allele harboring a lonely Edc3-binding motif. This dcp2 allele had the activity of WT DCP2 in promoting the decay of both Edc3 and Dhh1 substrates (Figure 4C, Figure 4—figure supplement 4, compare U1D1-U1D2-LD1-9 to WT). Similarly, the combination of deletions E3D-U1D2-LD1-9 and E3D1-U1D2-LD1-9 created two dcp2 alleles harboring a lonely Upf1-binding motif U11, and the combination of deletions E3D-U1D1-LD1-9 and E3D1-U1D1-LD1-9 created two dcp2 alleles harboring a lonely Upf1-binding motif U12. Each of these four dcp2 alleles had the activity of WT DCP2 in promoting NMD (Figure 4C, compare alleles E3D-U1D2-LD1-9, E3D1-U1D2-LD1-9, E3D-U1D1-LD1-9, and E3D1-U1D1-LD1-9 to WT). These results indicate that the Edc3-binding motif and each of the Upf1-binding motifs can function independently of other elements and promote specific mRNA decay activities.

Edc3 carries out one additional function upstream of recruitment of the decapping enzyme in decay of both Edc3 and Dhh1 substrates

Except for RPS28B mRNA, loss of the Dcp2 Edc3-binding motif, both Upf1-binding motifs, or the Pat1-binding motifs caused only partial stabilization, or no stabilization, of Edc3, NMD, and Pat1/Lsm1 substrates. In contrast, loss of the corresponding specific binding factors Edc3, Upf1, or Pat1 all resulted in substantial stabilization of their targeted mRNAs (see above). These results strongly suggest that Edc3-, Upf1-, or Pat1-mediated recruitment of the decapping enzyme is not rate limiting to the overall decay process and each of these decapping activators most likely carries out an additional major function upstream of the recruitment of the decapping enzyme in decay of the respective targeted mRNAs. To test this idea further, we constructed double mutant strains containing dcp2 cis deletions E3D or E3D1 and deletion of EDC3 and analyzed the decay phenotypes of Edc3 and Dhh1 substrates in the resulting strains. Because E3D and E3D1 both eliminate Edc3 binding to Dcp2, we reasoned that any additional stabilization caused by EDC3 deletion of specific mRNAs in E3D or E3D1 cells is likely due to loss of an extra function of Edc3.

Deletion of EDC3 caused additional seven- to eightfold stabilization of the Edc3 substrate YRA1 pre-mRNA in E3D or E3D1 cells (Figure 4D, Figure 4—figure supplement 5). Deletion of EDC3 also caused additional stabilization of the Dhh1 substrates EDC1 and SDS23 mRNAs and the Pat1/Lsm1/Dhh1 substrates HSP12 and HXT6 mRNAs (Figure 4D, Figure 4—figure supplement 5). These results provide evidence that Edc3 is directly involved in the decay of Dhh1-regulated mRNAs, and demonstrate that besides recruiting the decapping enzyme, Edc3 indeed carries out an additional function in decay of both Edc3 and Dhh1 substrates.

Dcp2 cis-binding elements promote the assembly of distinct decapping complexes in yeast cells

The experiments described above indicate that the Dcp2 C-terminal cis-binding elements promote independent binding of Edc3, Upf1, and Pat1 to Dcp2 and control the selective targeting of the decapping enzyme to different decapping substrates. These results and our earlier observation that different decapping activators each target a specific subset of yeast mRNAs (He et al., 2018) strongly suggest that different Dcp2 cis-binding elements promote the assembly of distinct decapping complexes. To test this idea, we explored two-hybrid interactions that are bridged through Dcp2 between Dcp1 and Edc3, Upf1, Pat1, or Dhh1 (Figure 5A; He and Jacobson, 2015) and dissected the molecular basis for each of these bridged interactions. We constructed two sets of yeast two-hybrid tester strains and analyzed each of the bridged interactions in these strains. One set of strains contains single deletions of EDC3, UPF1, PAT1, and SCD6 or a previously described dcp2-N245 truncation that eliminates the entire Dcp2 C-terminal domain (He and Jacobson, 2015). This set of strains also contains double mutants that combine EDC3 deletion either with deletions of UPF1, PAT1, and SCD6 or with the dcp2-N245 truncation. The second set of strains contains different C-terminal truncations of Dcp2 in both EDC3 and edc3Δ backgrounds. Each of these Dcp2 truncations eliminates a distinct set of its cis-binding elements. The first set of tester strains was designed to assess the factor requirements and the second set of tester strains was designed to evaluate the Dcp2 cis-binding element dependence for each of the observed interactions.

The Dcp1–Dcp2 decapping enzyme forms distinct decapping complexes in vivo with the decapping activators Edc3, Dhh1, Upf1, and Pat1.

Two-hybrid assays, as in Figure 1, were used to dissect the molecular basis of five Dcp2-bridged interactions between Dcp1 and Dhh1, Upf1, or Pat1, and between Edc3 and Upf1 or Pat1. (A) Dcp1:Dhh1, Dcp1:Upf1, Dcp1:Pat1, Edc3:Upf1, and Edc3:Pat1 interactions in tester strains harboring different gene deletions or truncations of decapping factors. (B) Dcp1:Dhh1, Dcp1:Upf1, and Dcp1:Pat1 interactions in tester strains harboring Dcp2 C-terminal truncations in EDC3 and edc3Δ cells. (C) Edc3:Upf1 and Edc3:Pat1 interactions in tester strains harboring Dcp2 C-terminal truncations in EDC3 and edc3Δ cells. (D) Three different decapping complexes inferred from two-hybrid analyses in A and B. (E) Two different decapping complexes inferred from two-hybrid analyses in A and C.

As shown in Figure 5A, Dhh1 exhibited strong two-hybrid interaction with Dcp1 in WT cells. Deletions of UPF1, PAT1, or SCD6 did not affect the Dhh1:Dcp1 interaction. In contrast, loss of the entire C-terminal domain of Dcp2 or deletion of EDC3 eliminated Dhh1:Dcp1 interaction. These results indicate that the Dhh1:Dcp1 interaction observed in WT cells requires both the Dcp2 C-terminal domain and Edc3, suggesting that this interaction occurs in a Dcp1–Dcp2–Edc3–Dhh1 complex. Further, in a WT EDC3 background, Dcp2 C-terminal truncations maintaining the E3 motif left the Dhh1:Dcp1 interaction intact. In contrast, the Dcp2 C-terminal truncation eliminating the E3 motif abolished the Dhh1:Dcp1 interaction (Figure 5B). These results show that the observed Dhh1:Dcp1 interaction also requires an intact Dcp2 E3 motif, indicating that the Dcp2 Edc3-binding motif promotes assembly of the Dcp1–Dcp2–Edc3–Dhh1 decapping complex in vivo (Figure 5D). In an edc3Δ background, Dhh1 exhibited no interaction with Dcp1 (the slight blue color of the transformants was due to self-activation of Gal4(DB)-Dhh1 fusion protein) (Figure 5B; He and Jacobson, 2015), indicating that Dhh1 assembles into the Dcp1–Dcp2–Edc3–Dhh1 via association with Edc3, consistent with the physical interaction between these two factors (He and Jacobson, 2015; Sharif et al., 2013).

In WT cells, Upf1 also exhibited strong two-hybrid interaction with Dcp1 (Figure 5A). Deletions of UPF1, PAT1, or SCD6 did not affect Upf1:Dcp1 interaction. In contrast, deletion of EDC3 diminished and loss of the entire C-terminal domain of Dcp2 eliminated Upf1:Dcp1 interaction. These results indicate that Upf1:Dcp1 interaction in WT cells requires the C-terminal domain of Dcp2 and is enhanced by the presence of Edc3, suggesting that this interaction occurs in a Dcp1–Dcp2–Edc3–Upf1 complex. In a WT EDC3 background, Dcp2 C-terminal truncations eliminating the second Upf1-binding motif U12 did not affect Upf1:Dcp1 interaction (Figure 5B, alleles N630 and N475); however, further eliminating the first Upf1-binding motif U11 abolished Upf1:Dcp1 interaction (Figure 5B, alleles N400, N300, and N245). In an edc3Δ background, Upf1 interacted weakly with Dcp1 and Dcp2 C-terminal truncations eliminating the second Upf1-binding motif U12 abolished Upf1:Dcp1 interaction (Figure 5B, compare alleles N771 to N630). These results indicate that the Upf1-binding motifs U11 and U12 can both promote the assembly of Upf1-containing decapping complexes and that Edc3 enhances the binding of Upf1 to the U11 motif to promote assembly of the Dcp1–Dcp2–Edc3–Upf1 decapping complex in vivo (Figure 5D).

In contrast to Dhh1 and Upf1, Pat1 exhibited only weak two-hybrid interaction with Dcp1 in WT cells (Figure 5A). Deletions of UPF1, PAT1, or SCD6 did not alter Pat1:Dcp1 interaction, but loss of the entire Dcp2 C-terminal domain eliminated Pat1:Dcp1 interaction (Figure 5A). Interestingly, deletion of EDC3 enhanced Pat1:Dcp1 interaction (Figure 5A). These results indicate that Pat1 can associate with the Dcp1–Dcp2 decapping enzyme both in the presence and absence of Edc3. Eliminating different leucine-rich motifs between L2 and L9 by progressive Dcp2 C-terminal truncations did not significantly alter the respective Pat1:Dcp1 interaction strength in EDC3 and edc3Δ backgrounds. However, further eliminating the leucine-rich motif L1 abolished Pat1:Dcp1 interaction in an EDC3 background and greatly diminished Pat1:Dcp1 interaction in an edc3Δ background (Figure 5B). These results show that Pat1:Dcp1 interaction requires the L1 motif in Dcp2 regardless of the cellular status of Edc3, suggesting that this motif may promote assembly of several different decapping complexes including a Dcp1–Dcp2–Edc3–Pat1 complex (Figure 5D).

Edc3 is a common component of multiple decapping complexes

To further validate our conclusion that Dcp2 cis-binding elements promote assembly of distinct decapping complexes and to explore the assembly dynamics of these complexes, we examined whether Dcp2 bridges interactions between Edc3 and Upf1 or Pat1. As shown in Figure 5A, in contrast to the Upf1:Dcp1 interaction, Upf1 did not show two-hybrid interaction with Edc3 in WT cells. Upf1 also did not show interaction with Edc3 in upf1Δ, pat1Δ, or scd6Δ cells. Surprisingly, Upf1 exhibited strong two-hybrid interaction with Edc3 in edc3Δ cells. Deletions of UPF1, PAT1, or SCD6 in edc3Δ cells did not alter this strong Upf1:Edc3 interaction; however, deletion of the entire Dcp2 C-terminal domain in edc3Δ cells eliminated Upf1:Edc3 interaction. These results indicate that Dcp2 bridges an interaction between Upf1 and Edc3 and that this bridged Upf1:Edc3 interaction is inhibited by endogenous Edc3. In an edc3Δ background, eliminating the second Upf1-binding motif U12 in Dcp2 by C-terminal truncation abolished the Upf1:Edc3 interaction (Figure 5C, compare alleles N771 and N630). This result shows that the Upf1:Edc3 interaction requires the U12 motif in Dcp2 and that this motif can promote assembly of a Dcp1–Dcp2–Edc3–Upf1 decapping complex in vivo, with Upf1 binding to the Dcp2 U12 motif (Figure 5E).

Similar to the Upf1:Edc3 interaction, Pat1 also did not show two-hybrid interaction with Edc3 in WT cells, deletion of EDC3 promoted Pat1:Edc3 interaction, and deletion of the entire C-terminal domain of Dcp2 eliminated Pat1:Edc3 interaction in edc3Δ cells (Figure 5A). These results indicate that Dcp2 also bridges an interaction between Pat1 and Edc3 and that this bridged interaction is inhibited by endogenous Edc3. In an edc3Δ background, eliminating different leucine-rich motifs between L3 and L9 did not affect Pat1:Edc3 interaction; however, further eliminating the L2 motif abolished Pat1:Edc3 interaction (Figure 5C). This result shows that Pat1:Edc3 interaction requires the L2 motif in Dcp2 and that this motif can promote assembly of a Dcp1–Dcp2–Edc3–Pat1 decapping complex in vivo, with Pat1 binding to the Dcp2 L2 motif (Figure 5E).

Our two-hybrid analyses of the bridged molecular interactions between Upf1 and Dcp1 or Edc3 revealed two distinct Upf1-containing decapping complexes with the same composition but different configurations. Similarly, analyses of the bridged molecular interactions between Pat1 and Dcp1 or Edc3 also revealed two distinct Pat1-containing decapping complexes of the same composition but different configurations (Figure 5D, E). These complexes share Dcp1, Dcp2, and Edc3 with Upf1 binding to either U11 or U12, or with Pat1 binding to either L1 or L2 in Dcp2. We suspect that the binding site preference for Upf1 and Pat1 in these decapping complexes is likely caused by the constraint of configurations of the Gal4 DNA-binding and activation domains fused to the respective binding partners. Nevertheless, these results indicate that Upf1 and Pat1 can both bind to two different binding motifs in Dcp2, raising the possibility that different Upf1- and Pat1-binding motifs in Dcp2 may control structural transitions of specific decapping complexes. It is also possible that the binding site preference for Upf1 and Pat1 is controlled by the number of Edc3 molecules (i.e., monomer or dimer) in the decapping complexes.

Our observation that endogenous Edc3 can completely inhibit the bridged molecular interactions between Edc3 and Upf1 or Pat1 is intriguing (Figure 5A). It indicates that free Edc3-binding sites from Dcp2 molecules are limiting and suggests that the entire pool of Dcp2 molecules are likely to be stably bound by endogenous Edc3. Consistent with this interpretation, as demonstrated above, Edc3 exists as a common component of multiple decapping complexes.

Xrn1 binds to an internal fragment of Dcp2 and is recruited to the decapping complex by Dcp2

To assess potential coupling between decapping and 5′ to 3′ exoribonucleolytic decay, we tested two hybrid interactions between Xrn1 and all known decapping factors. Xrn1 interacted with Dcp1, Edc3, Pat1, and Upf1 (Figure 6A). Xrn1 did not interact with full-length Dcp2, but interacted with a 726-amino acid C-terminal Dcp2 fragment Dcp2-ND244 (Figure 6A), suggesting that Xrn1 binding to Dcp2 may be dependent on a specific Dcp2 conformation. Deletions of EDC3, UPF1, or PAT1 and elimination of the entire C-terminal domain of Dcp2 did not affect binding of Xrn1 to Dcp2-ND244 or Pat1 (Figure 6C), suggesting that Xrn1 may bind to Dcp2 and Pat1 directly. To identify the Xrn1-binding region of Dcp2, we analyzed Xrn1 interactions with a panel of Dcp2 fragments (Figure 6—figure supplement 1A). In contrast to the binding of Edc3, Upf1, and Pat1 to Dcp2, Xrn1 binding to Dcp2 required a large Dcp2 fragment with the minimal Xrn1-binding region encompassing an internal fragment from aa 316–575 (Figure 6D, Figure 6—figure supplement 1B).

Figure 6 with 1 supplement see all
Xrn1 binds to Dcp2 and is directly recruited to different decapping complexes by Dcp2.

(A) Two-hybrid analyses of Xrn1 interactions with Dcp1, Edc3, Pat1, Upf1, and an N-terminally truncated Dcp2 fragment in WT cells. (B) Two-hybrid analyses of Dcp2-bridged interactions between Xrn1 and Dcp1, Edc3, or Upf1. (C) Two-hybrid analyses of interactions between Xrn1 and Pat1 or N-terminally truncated Dcp2. (D) Two-hybrid analyses of Xrn1 binding to an internal Dcp2 fragment. See also Figure 6—figure supplement 1. (E) Dcp1:Xrn1, Edc3:Xrn1, and Upf1:Xrn1 interactions in tester strains harboring Dcp2 C-terminal truncations and EDC3 or edc3Δ alleles. (F) Two different Xrn1-containing decapping complexes inferred from two-hybrid analyses in B and E. Allele schematics and two-hybrid analyses as in Figure 1.

In contrast to the Xrn1:Dcp2-ND244 and Xrn1:Pat1 interactions, the Xrn1:Dcp1, Xrn1:Edc3, and Xrn1:Upf1 interactions are all bridged by Dcp2. As shown in Figure 6B, deletions of UPF1, PAT1, or SCD6 did not significantly affect Xrn1:Dcp1 interaction. However, elimination of the Dcp2 C-terminal domain or deletion of EDC3 abolished Xrn1:Dcp1 interaction, suggesting that Xrn1:Dcp1 interaction is bridged by Dcp2 and enhanced by Edc3, and occurs in a Dcp1–Dcp2–Edc3–Xrn1 complex (Figure 6F). In support of the latter conclusion, Xrn1:Dcp1 interaction is also dependent on an intact Xrn1-binding site in Dcp2. In an EDC3 background, loss of partial Xrn1-binding region by Dcp2 C-terminal truncation abolished Xrn1:Dcp1 interaction (Figure 6E, compare alleles N630 and N475). Deletions of EDC3, UPF1, PAT1, or SCD6 did not affect Xrn1:Edc3 interaction, but elimination of the entire Dcp2 C-terminal domain abolished Xrn1:Edc3 interaction (Figure 6B), indicating that this interaction is bridged by Dcp2. As additional support for this conclusion, in both EDC3 and edc3Δ backgrounds, Xrn1:Edc3 interaction requires an intact Xrn1-binding site in Dcp2, as loss of partial Xrn1-binding region eliminates Xrn1:Edc3 interaction (Figure 6E, compare alleles N630 and N475). In contrast to Upf1:Edc3 and Pat1:Edc3 interactions, Xrn1:Edc3 interaction was not inhibited by endogenous Edc3. We suspect that the observed Xrn1:Edc3 interaction may involve dimerization of exogenous Edc3 with endogenous Edc3 bound to Dcp2 in a decapping complex. Elimination of the Dcp2 C-terminal domain and deletion of EDC3 both abolished Xrn1:Upf1 interaction (Figure 6B), indicating that this interaction is bridged by Dcp2 and occurs in a Dcp1–Dcp2–Edc3–Xrn1–Upf1 complex (Figure 6F). Consistent with being an interaction bridged by Dcp2, Xrn1:Upf1 interaction required the second Upf1-binding motif U12 in Dcp2, as loss of the U12 motif by Dcp2 C-terminal truncation eliminated Xrn1:Upf1 interaction (Figure 6E, compare alleles N771 and N630).

Collectively, our two-hybrid analyses show that Xrn1 binds to an internal fragment of Dcp2 and is recruited to the decapping complex by Dcp2. These results indicate that two important events in 5′ to 3′ mRNA decay, Dcp2-mediated decapping and Xrn1-mediated 5′ to 3′ exoribonucleolytic digestion, are physically coupled in vivo, suggesting that in addition to controlling the selective targeting of the decapping enzyme to different decapping substrates, the C-terminal domain of Dcp2 also controls efficient 5′ to 3′ exonucleolytic decay, ensuring that decapped mRNAs are degraded in a timely manner.

Scd6 interacts with multiple decapping factors and exists in both Edc3- and Pat1-containing decapping complexes

Our genetic analyses presented in Figure 1C, D indicated that the E3-1 motif binds not only Edc3 but an additional factor as well. As Scd6 and Edc3 share similar domain structures and appear to have redundant functions in mRNA decay (Albrecht and Lengauer, 2004; Decourty et al., 2008), we reasoned that Scd6 might bind to the Dcp2 E3-1 motif. To test this idea, we first examined two-hybrid interactions between Scd6 and each of the known yeast decapping factors. We found that Scd6 interacted with Dcp1, Dcp2, Edc3, and Pat1, but did not interact with all other decay factors tested, including Dhh1 (Figure 7A).

Figure 7 with 1 supplement see all
Scd6 binds to Dcp2 via multiple elements including the Edc3-binding and leucine-rich motifs and exists in both Edc3- and Pat1-containing decapping complexes.

(A) Two-hybrid analyses of Scd6 interactions with known yeast decapping factors in a WT tester strain. (B) Two-hybrid analyses of Dcp2-bridged interactions between Scd6 and Dcp1 or Edc3. (C) Two-hybrid analyses of interactions between Scd6 and Pat1 or N-terminally truncated Dcp2. (D) Two-hybrid analyses of Scd6 binding to different Dcp2 fragments. See also Figure 7—figure supplement 1. (E) Two-hybrid assays evaluating the consequences of deleting the Dcp2 inhibitory element and the Edc3- or Upf1-binding motifs on Dcp2 interactions with Scd6 and Edc3. (F) Two-hybrid assays evaluating the consequences of deleting the leucine-rich motifs (L1–L9) from the Dcp2 C-terminal domain on Dcp2 interactions with Scd6 and Pat1. (G) Two different decapping complexes inferred from two-hybrid analyses in B, E, and F. Allele schematics and two-hybrid analyses as in Figure 1.

To assess whether the observed interactions between Scd6 and Dcp1, Dcp2, Edc3, and Pat1 in WT cells are direct or mediated by interactions with other factors, we repeated the approach of Figure 5A, assaying each of the Scd6:Dcp1, Scd6:Dcp2, Scd6:Edc3, and Scd6:Pat1 interactions in a panel of tester strains harboring specific decay factor gene deletions or a large Dcp2 C-terminal truncation. As shown in Figure 7B, Scd6 interacted with Dcp1 strongly in WT cells and this interaction was not affected by deletions of UPF1, PAT1, or SCD6. However, Scd6:Dcp1 interaction was greatly diminished by deletion of EDC3, and was eliminated by the dcp2-N245 truncation. In an edc3Δ background, Scd6 interacted weakly with Dcp1. This weak Scd6:Dcp1 interaction was not affected by deletions of UPF1 or SCD6, but was eliminated by deletion of PAT1 or the dcp2-N245 truncation. Together, these results indicate that the observed Scd6:Dcp1 interaction in WT cells is bridged by the C-terminal domain of Dcp2 and that both Edc3 and Pat1 can promote the joining of Scd6 into decapping complexes. In the presence of Edc3, Scd6 exists in a Dcp1–Dcp2–Edc3–Scd6 complex and, in the absence of Edc3, Scd6 can exist in a Dcp1–Dcp2–Scd6–Pat1 complex.

Scd6 interacted weakly with Edc3 in WT cells (Figure 7B). Much like the Upf1:Edc3 and Pat1:Edc3 interactions shown in Figure 5A, the Scd6:Edc3 interaction was enhanced by deletion of EDC3 and was eliminated by the dcp2-N245 truncation in both EDC3 and edc3Δ backgrounds (Figure 7B). These results indicate that the observed Scd6:Edc3 interaction in WT cells is bridged by the C-terminal domain of Dcp2 and that endogenous Edc3 inhibits the Scd6:Edc3 two-hybrid interaction. Interestingly, in an EDC3 background, deletion of PAT1 also promoted the Scd6:Edc3 interaction (Figure 7B), suggesting that Pat1 may regulate the interaction of Scd6 with an Edc3-containing decapping complex.

The Scd6:Dcp2 and Scd6:Pat1 interactions observed in WT cells (Figure 7A) appear to be direct, as these two interactions were not affected by the dcp2-N245 truncation or the deletions of the EDC3 or PAT1 (Figure 7C). A direct Scd6:Dcp2 interaction is further supported by the fact that the observed Scd6:Dcp1 and Scd6:Edc3 interactions are both bridged through Dcp2 (Figure 7B). Collectively, these results indicate that Scd6 interacts with multiple decapping factors including Dcp1, Dcp2, Edc3, and Pat1 and that, depending on the Edc3 status in the cell, Scd6 can exist in Dcp1–Dcp2–Edc3–Scd6 or Dcp1–Dcp2–Scd6–Pat1 complexes.

Scd6 binds to Dcp2 through multiple elements including the Edc3-binding and leucine-rich motifs

To map the Scd6-binding site(s) on Dcp2, we tested the two-hybrid interaction of Scd6 with a panel of Dcp2 fragments (Figure 7—figure supplement 1). Scd6 interacted strongly with the ND244 fragment that contains the entire Dcp2 C-terminal domain (Figure 7D, Figure 7—figure supplement 1). Scd6 interacted moderately with two N-terminal Dcp2 fragments (N101-300 and N246-300) and moderately or weakly with six additional C-terminal fragments (ND315, ND415, F9-C10, F10-C10, F12-C10, and F13-C10) (Figure 7D, Figure 7—figure supplement 1). The N101-300 and N246-300 fragments and the ND244 fragment share a 55-amino acid overlap of the Edc3-binding region. The six C-terminal fragments share different overlaps with the N244 fragment, but have no overlap with the two N-terminal fragments. Thus, the pattern of Scd6 binding to different Dcp2 fragments indicates that Scd6 most likely binds multiple elements in Dcp2 and that one of the Scd6-binding elements coincides with the Edc3-binding region.

To further elucidate the mode of Scd6’s binding to Dcp2 and to identify the potential Scd6-binding elements in Dcp2, we analyzed the two-hybrid interactions of Scd6 with two panels of dcp2 element deletion alleles (Figure 7E, F). Scd6 interacted strongly with WT full-length Dcp2, and loss of the inhibitory element (ID), the first Upf1-binding element (U1D1), or the second Upf1-binding element (U1D2) either individually or in different combinations did not affect Scd6:Dcp2 binding (Figure 7E). Loss of the Edc3-binding motif E3-1 (E3D1) or the E3-2 motif (D3D2) partially weakened the Scd6:Dcp2 interaction; however, loss of both the E3-1 and E3-2 motifs (E3D) eliminated Scd6:Dcp2 interaction (Figure 7E, compare the E3D1 and E3D2 alleles to all the E3D-containing alleles). These results indicate that the E3-1 and E3-2 motifs can both promote Scd6 binding and function redundantly to recruit Sdc6 to Dcp2.

Loss of the leucine-rich motifs L1 and L2 (LD1-2) did not affect the binding of Scd6 to Dcp2; further loss of the L3 motif (LD1-3) or both L3 and L4 motifs (LD1-4) partially weakened Scd6:Dcp2 interaction to similar extents, and further loss of the L5 motif (LD1-5) eliminated Scd6:Dcp2 interaction (Figure 7F). These results indicate that leucine-rich motifs L3 and L5 are required for Scd6 binding to Dcp2. These two motifs each make a partial contribution to Scd6 binding and function redundantly to recruit Scd6 to Dcp2. Interestingly, loss of the last leucine-rich motif L9 (LD9) also weakened Scd6:Dcp2 interaction, and loss of the last two leucine-rich motifs L8 and L9 (LD9-8) eliminated Scd6:Dcp2 interaction (Figure 7F), indicating that leucine-rich motifs L8 and L9 are also required for Scd6 binding to Dcp2. The L8 and L9 motifs each make a partial contribution for Scd6 binding and appear to function redundantly in recruiting Scd6 to Dcp2.

In sum, we identified three pairs of functionally redundant binding motifs in Dcp2 (E3-1 and E3-2, L3 and L5, and L8 and L9) that promote Scd6 binding to Dcp2. As loss of any of these binding motif pairs eliminates Scd6 binding to Dcp2, Scd6 may use three different regions to engage these three pairs of Dcp2-binding motifs cooperatively. In this case, Scd6 may bind to full-length Dcp2 either as a monomer or as a binding-induced dimer with the dimer having a higher affinity, and each monomer engaging three different binding motifs, one from each of the three pairs. Alternatively, Scd6 may use its Lsm domain singlehandedly to engage Dcp2. In this case, Scd6 may bind to full-length Dcp2 using allovalency as described above for Pat1, requiring a minimum of five Dcp2-binding motifs.

Discussion

C-terminal cis-binding motifs promote independent binding of Edc3, Scd6, Upf1, and Pat1 to Dcp2 and control selective targeting of the decapping enzyme to distinct substrate mRNAs

The yeast decapping enzyme targets thousands of mRNAs for decapping-dependent decay (Celik et al., 2017; He et al., 2018) and the genetic experiments described here elucidate the mechanistic basis for the enzyme’s selective targeting of specific subgroups of those mRNAs. Dcp2’s large C-terminal domain encompasses multiple regulatory elements that serve as binding sites for proteins heretofore thought of as ‘decapping activators’ and our systematic deletion of those sites has defined their roles. Loss of both Upf1-binding motifs eliminated Upf1 binding to Dcp2 and caused selective partial stabilization of NMD substrates; loss of the Edc3-binding motif eliminated Edc3 binding to Dcp2 and caused selective partial or complete stabilization of Edc3 substrates, as well as selective partial stabilization of Dhh1 substrates (Figure 1C, D). Further, deletions of the leucine-rich Pat1-binding motifs eliminated Pat1 binding to Dcp2 and, when these deletions were combined with deletions of the Edc3-binding motif, caused selective additional stabilization of Dhh1 substrates (Figure 4B).

Our observation that loss of the Edc3-binding motif caused selective stabilization of Dhh1 substrates uncovers a new direct role of Edc3 in selective targeting of the decapping enzyme to Dhh1-regulated mRNAs. This role is not totally unexpected since Edc3 interacts directly with Dhh1 (He and Jacobson, 2015; Sharif et al., 2013), Dhh1 association with the decapping enzyme is dependent on Edc3 (Figure 5A), and deletion of EDC3 causes additional stabilization of Dhh1-regulated mRNAs in dcp2-E3D and dcp2-E3D1 cells (Figure 4D). In contrast to Dcp2 cis-element E3D or E3D1 deletions, trans deletion of EDC3 had no discernible effect on Dhh1 substrates (Figures 1B and 4B), a result suggesting that at least one additional factor binds to the Edc3-binding motif and can target the decapping enzyme to Dhh1-regulated mRNAs. Since Scd6 also binds to the Edc3-binding motif E3-1 and exists in both Edc3 and Pat1-containing decapping complexes (Figure 7B, E), we propose that Scd6 functions redundantly with Edc3 in targeting the decapping enzyme to Dhh1 substrates. Surprisingly, given the multiple proposed roles for Pat1 in mRNA decapping (Charenton et al., 2017; Lobel et al., 2019; Nissan et al., 2010), we found that loss of the Pat1-binding motifs had no effect on Pat1/Lsm1 substrates or any other decapping substrates (Figure 2B). We propose that Pat1 binding to the leucine-rich motifs in the Dcp2 C-terminal domain still controls the selective targeting of the decapping enzyme to Pat1/Lsm1 substrates and may even enhance the decapping rates of the targeted mRNAs as demonstrated in vitro (Lobel et al., 2019), but the Pat1-mediated decapping step is likely not rate limiting for decay of Pat1/Lsm1 substrates.

Dcp2 C-terminal cis-binding motifs promote assembly of distinct target-specific decapping complexes in vivo

Consistent with our observation that Dcp2 cis-binding elements control the selective targeting of the decapping enzyme to different decapping substrates, our two-hybrid experiments reveal that the same elements can promote in vivo assembly of distinct target-specific complexes. The Edc3-binding motif E3-1 can promote the assembly of a Dhh1-containing Dcp1–Dcp2–Edc3–Dhh1 complex, the two Upf1-binding motifs U11 and U12 can promote assembly of two distinct Upf1-containing Dcp1–Dcp2–Edc3–Upf1 complexes of the same composition but different configurations with Upf1 binding either to the U11 or U12 motifs, and the Pat1-binding motifs L1 and L2 can promote assembly of two distinct Pat1-containing Dcp1–Dcp2–Edc3–Pat1 complexes of the same composition but different configurations with Pat1 binding either to the L1 or L2 motifs (Figure 5D, E). Further, the motifs E3-1 and E3-2 in combination with motifs L3 and L5, and L8 and L9 can promote the assembly of a Dcp1–Dcp2–Edc3–Scd6 complex and a Dcp1–Dcp2–Scd6–Scd6–Pat1 complex (Figure 7G). Based on several observations, including: (1) Edc3 associates with itself (Decker et al., 2007; He and Jacobson, 2015), (2) Edc3 targets two specific transcripts RPS28B mRNA and YRA1 pre-mRNA for decapping-dependent decay (Badis et al., 2004; Dong et al., 2007), (3) the YRA1 intron contains two functionally redundant Edc3-binding elements (Dong et al., 2010), and (4) Edc3 binds to the 3′-UTR decay-inducing element of RPS28B mRNA as a dimer (He et al., 2014), we postulate that the Edc3-binding motif can promote assembly of at least one additional Edc3 substrate-specific decapping complex, a Dcp1–Dcp2–Edc3–Edc3 complex containing Edc3 as a homodimer (Figure 8). Because Edc3 is directly involved in decay of Dhh1 substrates (Figure 4D), and it also promotes Sdc6 joining to the decapping complex (Figure 7), we speculate that the Dcp1–Dcp2–Edc3–Dhh1 complex likely contains Edc3 either as an Edc3:Edc3 homodimer or as an Edc3:Scd6 heterodimer (Figure 8). Because the Scd6-binding motifs in Dcp2 exhibit functional redundancies, we suspect that the Dcp1–Dcp2–Scd6–Pat1 complex formed in the absence of Edc3 (Figure 7B) likely contain Scd6 as an Scd6:Scd6 homodimer (Figure 7G). Since Scd6 also binds to the Edc3-binding motif E3-1, the dimeric Scd6:Scd6-containing decapping complexes may form even in an EDC3 background. The dimeric forms of Edc3:Edc3, Edc3:Scd6, and Scd6:Scd6 can all occupy the E3-1 and E3-2 motifs and compete with each other for binding to Dcp2. Collectively, our results indicate that the cis-binding elements located in the Dcp2 C-terminal domain promote assembly of multiple decapping complexes with distinct composition and substrate specificities. Edc3 appears to be both a common core component of multiple decapping complexes and a unique targeting component of specific decapping complexes.

A new model: formation of target-specific decapping complexes controls mRNA decapping.

In brief, the Dcp1–Dcp2 decapping enzyme interacts with Edc3 and then Xrn1, forming a preassembled inactive Dcp1–Dcp2–Edc3–Xrn1 complex. The resulting complex binds to individual targeting components, forming target-specific decapping complexes. The final assembly and enzymatic activation of these complexes are postulated to occur on mRNPs destined for degradation.

Our observation that Edc3 is a common component of multiple decapping complexes raises the intriguing issue of Edc3 function in these complexes. Numerous biochemical and structural studies, all using C-terminally truncated Dcp2 fragments and largely the Lsm domain of Edc3, suggested that Edc3 promotes catalysis by the decapping enzyme by alleviating the autoinhibition imposed by Dcp2 inhibitory elements, thus enhancing substrate binding (Charenton et al., 2016; Fromm et al., 2012; Mugridge et al., 2018; Nissan et al., 2010; Paquette et al., 2018). Our genetic experiments reveal that the core Edc3 component must carry out two additional important functions in mRNA decapping. One function is to provide the decapping enzyme with a set of unique Edc3-binding surfaces or modules and thus endow the decapping enzyme with Edc3 targeting specificity. Consistent with this proposition, eliminating Edc3 binding to Dcp2 caused selective stabilization of both Edc3 and Dhh1 substrates (Figure 1C, D). Another Edc3 function is to promote the assembly of additional target-specific decapping complexes, such as the Dcp1–Dcp2–Edc3–Upf1 and Dcp1–Dcp2–Edc3–Scd6 complexes, by remodeling Dcp2 or providing weak but specific binding surfaces for specific targeting or coupling factors. In support of this notion deletion of EDC3 diminishes the Dcp2-bridged Dcp1:Upf1, Dcp1:Xrn1, and Dcp1:Scd6 interactions (Figures 5A7), and triggers a switch of Upf1 binding from the U11 to the U12 motif in Dcp2 (Figure 5B). Our proposition that Edc3 exists both as a common core component of multiple decapping complexes and as a unique targeting component of specific decapping complexes provides a unified theory for explaining the apparently contradictory proposed functions for Edc3, that is, that Edc3 functions as a general decapping activator (Kshirsagar and Parker, 2004) or as a transcript-specific decapping activator (Badis et al., 2004; Dong et al., 2007).

Edc3-, Upf1-, and Pat1-mediated active recruitment of the decapping enzyme makes distinct contributions to the overall decay of the respective targeted mRNAs

Our genetic analyses of Dcp2 cis element mutants revealed the relative mRNA decay contributions of specific factor-mediated decapping enzyme recruitment events and provided significant insights into the decay mechanisms for different decapping substrates. For the Edc3 substrate RPS28B mRNA, loss of the Edc3-binding motif by E3D or E3D1 in Dcp2 caused complete mRNA stabilization relative to deletion of EDC3 itself (Figure 1C, D). For the Edc3 substrate YRA1 pre-mRNA, and all the tested Dhh1 substrates, loss of the Edc3-binding site caused only partial stabilization relative to the respective EDC3 or DHH1 deletions (Figures 1C and 4B). Additional loss of the Pat1-binding motifs caused further stabilization of the latter Edc3 and Dhh1 substrates (Figure 4B). These results indicate that both Edc3 and Dhh1 substrates can be targeted by Pat1-mediated recruitment of the decapping enzyme. However, since loss of the Pat1-binding motifs alone had no effect on these Edc3 and Dhh1 substrates (Figure 2B), it appears that Pat1-mediated recruitment of the decapping enzyme merely functions as a backup system in the decay of Edc3- and Dhh1-regulated mRNAs. Because the combined deletions of the Edc3- and Pat1-binding motifs caused substantial stabilization of the Edc3 and Dhh1 substrates (Figure 4B), our results indicate that decapping is rate limiting for both Edc3 and Dhh1 substrates, suggesting that the major functions of Edc3 or Pat1 in decay of these mRNAs is recruiting the decapping enzyme.

For NMD substrates, loss of both Upf1-binding motifs caused only marginal stabilization (two- to threefold), but deletion of UPF1 caused substantial stabilization (>tenfold) (Figure 1C). These results indicate that Upf1-mediated recruitment of the decapping enzyme only makes a minor contribution to the overall decay of NMD substrates, suggesting that decapping is not a major rate-limiting step in NMD, active recruitment of the decapping enzyme to NMD substrates may be largely dispensable, and Upf1 must carry out one additional major function upstream of the recruitment of the decapping enzyme. Consistent with these ideas, in the absence of active recruitment of the decapping enzyme by Upf1, NMD substrates are still degraded by decapping-dependent pathway (Figure 3B). In addition, further deletions of the Edc3- and Pat1-binding motifs do not have significant effects on NMD substrates (Figure 4C), and in fact, even eliminating the entire C-terminal domain only caused marginal stabilization of NMD substrates (He and Jacobson, 2015), as observed for loss of both Dcp2 Upf1-binding motifs.

For Pat1/Lsm1 substrates, loss of the Pat1-binding motifs had no effect, but deletion of PAT1 caused substantial stabilization (Figure 2B). These results indicated that the Pat1-mediated recruitment of the decapping enzyme does not make a significant contribution to the overall decay of Pat1/Lsm1 substrates, suggesting that decapping is not rate limiting in decay of the Pat1/Lsm1 substrates, active recruitment of the decapping enzyme to the Pat1/Lsm1 substrates can be dispensable, and Pat1 must carry out one additional major function upstream of the recruitment of the decapping enzyme. In support of this conclusion, in the absence of active recruitment of the decapping enzyme by Pat1, Pat1/Lsm1 substrates are still degraded by the decapping-dependent pathway (Figure 3C). Importantly, eliminating all the known-binding motifs in the Dcp2 C-terminal domain also did not have any effect on the decay Pat1/Lsm1 substrates (Figure 4C).

Dcp2 directly recruits the 5′ to 3′ exoribonuclease Xrn1 to the decapping complexes

Our two-hybrid experiments revealed that the 5′ to 3′ exoribonuclease Xrn1 binds to Dcp2 and is directly recruited to decapping complexes by Dcp2. This conclusion is supported by several new observations. First, Xrn1 binds to a specific internal Dcp2 fragment (Figure 6D, Figure 6—figure supplement 1B). Second, Xrn1 also interacts with both Dcp1 and Edc3, two other core components of the decapping enzyme, and each of the interactions requires the Dcp2 C-terminal domain as well as an intact Xrn1-binding region in this domain (Figure 6B, E). Third, Xrn1 is a common component of both the Dcp1–Dcp2–Edc3–Edc3–Xrn1 and Dcp1–Dcp2–Edc3–Upf1–Xrn1 complexes (Figure 6F). Finally, Xrn1’s recruitment to the decapping complexes requires prior Edc3 binding to Dcp2, as deletion of EDC3 eliminated both Dcp1:Xrn1 and Upf1:Xrn1 interactions (Figure 6B).

Our observation that Xrn1 is directly recruited by Dcp2 to decapping complexes suggests that decapping and 5′ to 3′ exoribonucleolytic decay are physically and mechanistically linked. This coupling appears to be conserved over eukaryotic evolution, but the mechanism of this coupling may differ for different organisms, for example, D. melanogaster Xrn1 binds to Dcp1 (Braun et al., 2012) and human Xrn1 binds to Edc4 in its respective decapping complex (Chang et al., 2019). Xrn1 binding to decapping complexes likely ensures immediate 5′ to 3′ exoribonucleolytic degradation of decapped transcripts and may also serve to inhibit the catalytic activity or substrate binding of the decapping enzyme until the enzyme is targeted to substrate mRNAs. Consistent with this idea, Xrn1 binding to Dcp2 requires the inhibitory element (Figure 6D) and overexpression of Xrn1 in D. melanogaster cells can inhibit the decapping of different reporter mRNAs (Braun et al., 2012).

Based on the observation that the same Pat1 C-terminal extension binds to multiple helical leucine-rich motifs in Dcp2 and a single such motif in Xrn1, Pat1 was proposed to coordinate the decapping and exonucleolytic decay events in general 5′ to 3′ mRNA decay by first recruiting the decapping enzyme and then Xrn1 to targeted mRNAs (Charenton et al., 2017). This Pat1-mediated sequential coupling model of decapping and 5′ to 3′ exonucleolytic decay is at odds with our result that Xrn1 is directly recruited by Dcp2 to the decapping complexes. Since Pat1 targets only a subset of yeast transcripts (He et al., 2018), and loss of the Pat1-binding motifs had no effect on all tested decapping substrates (Figure 2B) yet loss of the Edc3- and Upf1-binding motifs each yielded specific effects (Figure 1C), we suggest that Pat1-mediated coupling may be limited to a small number of mRNAs and may not make a significant general contribution to the overall decay of most mRNAs.

A new model for control of mRNA decapping

Our genetic data suggest a new model for control of mRNA decapping in yeast (Figure 8) whose main features include: cis-regulatory elements located in the Dcp2 C-terminal domain control selective targeting of the decapping enzyme by forming distinct multicomponent decapping complexes; the Dcp1 and Dcp2 subunits are shared common components of all decapping complexes; Edc3 functions as a shared common component of multiple decapping complexes, but is also a unique targeting component of dimeric Edc3-containing complexes; Xrn1 associates with Dcp2 after the binding of Edc3 to Dcp2 and is a component of multiple or perhaps all targeting complexes; Upf1 and Pat1 each function as unique targeting components in their respective Upf1- or Pat1-containing complexes; Scd6 may function as a second targeting component of the Dhh1 complexes and likely also collaborates with Pat1 to target the Pat1/Lsm substrates; the targeting components Edc3, Scd6, Upf1, and Pat1 each have at least two separate binding modules, one for the Dcp1–Dcp2–Edc3 core complex and another for their targeted mRNPs; and the final assembly and decapping activation of the target-specific decapping complexes occur on a to-be-degraded mRNP and the decapping event is coupled with immediate 5′ to 3′ degradation by Xrn1.

We propose that monomeric Edc3 is present in the NMD and Pat1/Lsm1 substrate-specific decapping complexes because: (1) Dcp2 contains only one Edc3-binding motif mapped to a 17-amino acid fragment (E3-1) (Figure 1D) and this fragment can only engage one Edc3 monomer for binding in the K. lactis Dcp1-Dcp2-Edc3-m7-GDP structure (Charenton et al., 2016), and (2) Edc3 and Upf1 or Pat1 co-occupy the Dcp2 C-terminus (Figure 5A) and the joining of Upf1 and Pat1 to the decapping complexes is sensitive to Edc3 copy number, as endogenous Edc3 inhibits both Edc3–Upf1 and Edc3–Pat1 interactions in the two-hybrid assay (Figure 5C). We propose that an Edc3 dimer is present in the Edc3 and Dhh1 substrate-specific decapping complexes because our results suggest that one monomer functions as a core component and another monomer functions as a targeting component. In this scheme targeting of the Edc3- and Dhh1-specific decapping complexes to their respective substrates is mostly controlled by homodimerization of the targeting and core Edc3 components.

Target-specific decapping complexes other than those specific for NMD, Edc3, Dhh1, and Pat1 substrates must exist in yeast cells. Our observation that Scd6 binds to the sole Edc3-binding motif on Dcp2 also suggests the existence of non-Edc3-containing decapping complexes in yeast cells. In addition, the leucine-rich motifs L6 to L9 originally proposed for Pat1 binding in fact may not bind Pat1 in the context of full-length Dcp2 and the motifs L6 and L7 each still lack an assigned binding activity, suggesting that these motifs may bind other factors. Further, loss of all the known factor-binding motifs in Dcp2 does not have any discernible effect on decapping of Pat1/Lsm1 substrates, raising the possibility that these mRNAs may be decapped by the Dcp1/Dcp2 complex alone. Collectively, our results indicate that yeast cells likely contain many distinct decapping complexes. The notion of a single yeast decapping enzyme is too simplistic and should be replaced by the concept of a decapping enzyme family that shares the core components Dcp1 and Dcp2, similar to the PP1 and PP2A phosphatase families involved in protein dephosphorylation (Shi, 2009; Virshup and Shenolikar, 2009).

Materials and methods

Key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
Chemical compound, drug[α-32P]-dCTPPerkin ElmerBlu513Z
Chemical compound, drug5-Bromo-4-chloro-3-indolyl-beta-D-galactopyranoside (X-GAL)USB7240-90-640 µg/ml
Chemical compound, drugPhenolFisherA92-500
Chemical compound, drugPhenol:choroform:IAA (25:24:1)AmbionAM9732
Chemical compound, drugHerring Sperm DNAPromegaD1815
Chemical compound, drugGeneticin (G418 Sulfate)Gibco11811-023400 µg/ml
Chemical compound, drugHygromycin BRoche10843555001200 µg/ml
Peptide, recombinant proteinTaq DNA polymeraseRoche04-728-874-001
Commercial assay or kitQuikChange XL Site-Directed Mutagenesis KitAgilent Technologies200,519
Commercial assay or kitRandom Primed DNA labeling KitRoche11-004-760-001
Strain, strain background (Escherichia coli)DH5αInvitrogen11319-019Electrocompetent cells
Strain, strain background (Saccharomyces cerevisiae)Supplementary file 1This paperYeast strains used in this studySee Materials and methods
Genetic reagent (plasmid)Supplementary file 2This paperPlasmids used in this studySee Materials and methods
Genetic reagent (plasmid)Supplementary file 4This paperDNA probes used in this studySee Materials and methods
Sequence-based reagentSupplementary file 3This paperOligonucleotides used in this studySee Materials and methods
Software, algorithmMulti-Gauge softwareFujifilmScience lab 2005
Software, algorithmGraphPad Prism 9 for WindowsGraphPad Software, LLChttps://www.graphpad.com
OtherZeta-Probe Blotting MembranesBioRad1620159He et al., 2008
OtherMini Quick Spin RNA ColumnsRoche11-814-427-001He et al., 2008

Experimental model and cell growth conditions

Request a detailed protocol

The experiments described in this study used the yeast S. cerevisiae as a model system. Yeast strains used for phenotypic analyses of mRNA decay were all derived from the W303 background (MATa ade2-1 his3-11,15 leu2-3,112 trp1-1 ura3-1 can1-100). Yeast strains used for two-hybrid analyses were all derived from the GGY1::171 background (his3 leu2 URA3::GAL1-lacZ gal4Δ gal80Δ) (Fields and Song, 1989). All yeast strains used in this study are listed in Supplementary file 1.

Yeast cells used for RNA isolation and transformation were all grown at 30°C in YEPD media (1% yeast extract, 2% peptone, 2% D-glucose). For integrative yeast transformation, transformants were selected at 30°C on different media depending on the selection markers of the transforming DNA fragments. For the drug resistance genes KanMX and HygroR, G418 (Gibco #11811-023), and hygromycin (Roche #10843555001) were included in YEPD media (1% yeast extract, 2% peptone, 2% D-glucose, 30 g agar/l, 400 µg/ml G418, or 200 µg/ml hygromycin). For the auxotrophic markers, URA3 and ADE2, synthetic -ura or -ade drop-out media (6.7 g/l of yeast nitrogen base without amino acids, 2 g/l -ura or -ade drop-out mix, 30 g/l of agar, 100 ml/l of 20% D-glucose) were used. For yeast two-hybrid analyses, transformants were first selected on synthetic -leu-his drop-out media and then replica-plated on SSX media [6.7 g/l of yeast nitrogen base without amino acids, 2 g/l -leu-his drop-out mix, 30 g/l of agar, 100 ml/l of 20% sucrose, 100 ml/l of 1 M potassium phosphate buffer pH 7.0, 2 ml/l of 20 mg/ml of 5-bromo4-chloro-3-indolyl β-D-galactoside (X-gal) (USB #7240-90-6) in formamide] for color development.

Method details

Methodological overview: generation and functional analysis of Dcp2 element deletion mutants

Request a detailed protocol

To dissect the roles of different Dcp2 regulatory elements (Figure 1A) in mRNA decapping, we generated specific deletions in the C-terminal domain of DCP2 that eliminated either a single element or combinations of different elements and analyzed the consequences of these deletions on Dcp2 interactions with specific decapping activators and decay of different decapping substrates. In total, we constructed three sets of element deletions and generated 50 dcp2 mutant alleles. The first set of dcp2 alleles contains single deletions as well as double, triple, and quadruple deletions in all possible combinations of the inhibitory element and the Edc3- and Upf1-binding motifs. The second set of dcp2 alleles contains consecutive deletions of the nine leucine-rich Pat1-binding motifs, proceeding either from the N-terminal end of this set of motifs or from its C-terminus. The third set of dcp2 alleles contains deletions of different combinations of Edc3-, Upf1-, and Pat1-binding motifs.

To link the potential defect in mRNA decapping to a specific Dcp2 interaction, we analyzed each of the dcp2 element mutant alleles in two parallel assays. In the first assay, each dcp2 allele was fused to the GAL4 DNA-binding domain and the encoded fusion proteins were tested for interactions with those encoded by EDC3, UPF1, and PAT1 fused to the GAL4 activation domain in the yeast two-hybrid system. In the second assay, each dcp2 allele was N-terminally tagged with a triple HA epitope and integrated at the genomic locus of DCP2 for functional analysis. Using quantitative northern blotting analyses, we measured the steady-state levels of different decapping substrates in each dcp2 element mutant. Decapping substrates analyzed in this study include the NMD substrates CYH2 pre-mRNA and ade2-1, can1-100, and trp1-1 mRNAs; the Edc3 substrates YRA1 pre-mRNA and RPS28B mRNA; the Dhh1 substrates EDC1 and SDS23 mRNAs; the Pat1/Lsm1 substrates AGA1, BUR6, DIF1, and LSM3 mRNAs; and the Pat1/Lsm1/Dhh1 substrates CHA1, HSP12, and HXT6 mRNAs. It should be noted that the presence of the HA-tag at the N-terminus of Dcp2 had no effect on its function in mRNA decay (Figure 1C, compare HA-DCP2 to WT).

Plasmid construction

Request a detailed protocol

Plasmids used in this study are listed in Supplementary file 2. Plasmids containing the original WT DCP2 allele, HA-DCP2 allele, and dcp2 alleles with different C-terminal truncations were previously described (He and Jacobson, 2015). A plasmid containing the edc3::URA3 allele was described in Dong et al., 2007. Plasmids containing the xrn1::ADE2, ski2::URA3, ski7::URA3, and scd6::KanMX6 alleles were described in He et al., 2018. Plasmids containing full-length DCP1, UPF1, UPF2, UPF3, EDC3, PAT1, LSM1, and DHH1 alleles, or different DCP2 fragments fused to either GAL4(AD) or GAL4(DB) were described in He and Jacobson, 2015. Plasmids constructed in this study are described below.

Construction of the dcp2 element deletion alleles

Request a detailed protocol

dcp2 alleles harboring regulatory element deletions were all constructed in Bluescript by using the plasmid HFSE1645 either as an initial template or as a cloning vector (see Supplementary file 2). This plasmid contains the WT HA-DCP2 allele as a 3.7 kb SalI–BglII/NotI fragment, including 320 bp from the promoter/5′-UTR region and 401 bp from the 3′-UTR region. Specific element deletions were all generated by using the QuikChange Site-Directed Mutagenesis Kit from Agilent Technologies (cat# 200519) according to the manufacturer’s instructions. PCR primers used to generate specific Dcp2 element deletions are listed in Supplementary file 3. dcp2 alleles containing single element deletions were generated by using plasmid HFSE1645 as a template. dcp2 alleles containing two or more element deletions were generated by two different strategies. One strategy used sequential rounds of site-direct mutagenesis and the other used molecular cloning to combine pre-existing element deletions from different DCP2 regions. Details on construction of each dcp2 element deletion allele are provided in Supplementary file 2. Specific element deletions in individual dcp2 mutant alleles were all confirmed by DNA sequencing. Each of these dcp2 alleles can be isolated from the corresponding plasmids as a SalI–BglII DNA fragment. For clarity, we use the following matched pairs of abbreviations for each specific element and its deletion: the inhibitory element: IE/ID; the Edc3-binding motif: E3/E3D, E3-1/E3D1, and E3-2/E3D2; the Upf1-binding motifs U11/U1D1 and U12/U1D2; and the nine leucine-rich motifs L1 to L9/LD1 to LD9. Analysis of dcp2 alleles harboring different deletions of the inhibitory element and the Edc3- and Upf1-binding motifs revealed that loss of the inhibitory element affected neither the binding of Edc3, Upf1, and Pat1 to Dcp2, nor the steady-state levels of different decapping substrates (Figure 1B, C). Thus, dcp2 alleles harboring a deletion of the inhibitory element are not discussed further and we focused our analysis on deletions of the Edc3-, Upf1-, and Pat1-binding motifs.

Construction of C-terminally truncated dcp2 alleles

Request a detailed protocol

To facilitate the construction of the dcp2-KanMX6 knock-in alleles harboring different DCP2 C-terminal truncations, we generated modified versions of the dcp2-N925, N770, N635, N475, N400, and N300 alleles in pRS315. These modified dcp2 alleles are identical to those that were described in He and Jacobson, 2015 except that each has a shorter 3′-UTR fragment and two additional restriction sites BamHI and NotI added to the 3′-end. The modified dcp2 alleles were constructed in two steps. In the first step, a 5′ XbaI–NcoI fragment isolated from the original WT DCP2 allele and a 3′ NcoI–SalI fragment amplified from the DCP2 3′-UTR region were ligated to pRS315 digested by XbaI and XhoI, generating pRS315-DCP2-WT-M1(HFSE1632). In the second step, the individual SalI–NcoI dcp2 fragments were isolated from each of the original dcp2 truncation alleles and these DNA fragments were then ligated to HFSE1632 digested by SalI and NcoI. Each of the modified dcp2 truncation alleles can be isolated as either a XbaI–BamHI or a SalI–BamHI fragment.

Construction of the dcp2-KanMX6 knock-in alleles

Request a detailed protocol

All dcp2-KanMX6 knock-in alleles were constructed in Bluescript by using the plasmid HFSE1636 as the cloning vector (see Supplementary file 2). This plasmid contains the previously described dcp2-N245-KanMX6 allele as a 3.5 kb NotI–BamHI/NotI/SalI fragment, including 925 bp from the DCP2 promoter/5′-UTR region and 401 bp from the 3′-UTR region (He and Jacobson, 2015). In this dcp2 allele, the 1452 bp BglII–EcoRI KanMX6 selection cassette was inserted into the promoter region 588 bp upstream from the first base of the DCP2 initiation codon. Two classes of dcp2 knock-in alleles were constructed. One class contains different dcp2 element deletions and the other contains different dcp2 C-terminal truncations. For construction of the dcp2 element deletion knock-in alleles, individual SalI–BglII DNA fragments were isolated from each of the original dcp2 element deletion alleles in Bluescript and ligated to HFSE1636 previously digested by SalI and BamHI. For construction of the dcp2 truncation knock-in alleles, individual SalI–BamHI DNA fragments were isolated from each of the modified dcp2 truncation alleles in pRS315 and ligated to HFSE1636 previously digested by SalI and BamHI. Each of the dcp2 knock-in alleles can be isolated as a NotI–NotI fragment for integrative yeast transformation.

Construction of the Gal4(DB)-dcp2 fusion alleles

Request a detailed protocol

Coding sequences from each of the dcp2 element deletion alleles were fused to the GAL4 DNA-binding domain by using plasmid pMA424 as the cloning vector (see Supplementary file 2). To facilitate the construction of these fusion alleles, we first used the plasmid HFSE1718 as a cloning vector to generate an intermediate allele from each of the original dcp2 element deletion alleles. Plasmid HFSE1718 contains the entire WT DCP2 coding region and 222 bp from the 3′-UTR region as a BamHI–NcoI/SalI DNA fragment. To generate the intermediate dcp2 alleles, depending on specific cases, either a 5′ PCR-amplified BamHI–XhoI fragment, or a 3′ restriction XhoI–NcoI fragment, or both of these fragments were obtained from the dcp2 element deletion alleles in Bluescript and then ligated to plasmid HFSE1718 previously digested by BamH–XhoI, XhoI–NcoI, or BamH–NcoI, respectively. To generate the final Gal4(DB)-dcp2 fusion alleles, each of the dcp2 intermediate alleles was isolated from the corresponding plasmid as a BamHI–SalI DNA fragment and then ligated to pMA424 previously digested by BamHI–SalI. All Gal4(DB)-dcp2 fusion alleles were confirmed by DNA sequencing.

Construction of the XRN1 or SCD6 and Gal4(DB) or Gal4(AD) fusion alleles

Request a detailed protocol

The entire XRN1 coding sequence was fused to the GAL4 DNA-binding domain and activation domain by using plasmids pMA424 and pGAD-C2 as cloning vectors, respectively. Both fusion alleles were constructed through a three-piece ligation reaction by making use of the unique SacI restriction site located at nt 293–298 of the XRN1 coding region. The XRN1 coding sequences from nt 1–309 were amplified by PCR using oligonucleotide pair XRN1-5′-BamHI-F and XRN1-5′-SacI-R (Supplementary file 3). The resulting PCR product was digested by BamHI and SacI, yielding a 5′ BamHI–SacI XRN1 fragment encompassing the first 297 nts of its coding sequences. A 3′ SacI–SalI XRN1 fragment encompassing the coding sequences from nt 298 to 4584 was isolated from plasmid HFSE1532. The 5′ BamHI–SacI and 3′ SacI–SalI XRN1 fragments were then ligated to pMA4242 and pGAD-C2 digested by BamHI and SalI to generate the final XRN1 fusion alleles. Similarly, the entire 1050 bp coding region of was amplified by PCR using oligonucleotides SCD6-TH-F1 and SCD6-TH-C1 as primers. The resulting PCR product was cut by EcoRI and SalI and then ligated to pMA424 or pGAD-C2 previously digested by EcoRI and SalI. The XRN1 and SCD6 fusion alleles were confirmed by DNA sequencing.

Construction of the upf1::KanMX6 and edc3::Hygro alleles

Request a detailed protocol

The upf1::KanMX6 and edc3::Hygro alleles were constructed in Bluescript. To construct Bs-ks-upf1::KanMX6, a PCR-amplified 417 bp NotI–BglII/NcoI fragment from the UPF1 5′-UTR region and a PCR-amplified 416 bp NcoI/EcoRI–SalI fragment from the UPF1 3′-UTR region were ligated to Bluescript digested by NotI and SalI in a three-piece ligation reaction (NotI–NcoI/NcoI–SalI) to generate an intermediate plasmid containing a upf1 allele lacking its entire CDS. The resulting intermediate plasmid was digested by BglII and EcoRI and then ligated to a 1452 bp BglII–EcoRI KanMX6 selection cassette. To construct Bs-ks-edc3::Hygro, a 455 bp PCR-amplified NotI–XbaI/BamHI fragment from the EDC3 5′-UTR region and a 512 bp PCR-amplified BamHI/EcoRI–SalI fragment from the EDC3 3′-UTR region were ligated to Bluescript digested by NotI and SalI in a three-piece ligation reaction (NotI–BamHI/BamHI–SalI) to generate an intermediate plasmid containing an edc3 allele lacking its entire CDS. The resulting plasmid was digested by BamHI and SalI and then ligated to a 1666 bp BglII–SacI Hygromycin resistant gene and a 508 bp PCR-amplified SacI–SalI fragment from the EDC3 3′-UTR region. The upf1::KanMX6 and edc3::Hygro alleles can each be isolated as a NotI–SalI fragment for integrative yeast transformation.

Strain construction

Request a detailed protocol

Yeast strains used in this study are listed in Supplementary file 1. The WT strain (HFY114) and its isogenic derivative harboring a deletion of UPF1 were described previously (He et al., 1997), as were isogenic strains harboring deletions of EDC3 (CFY25), PAT1 (SYY2674), LSM1 (SYY2680), or DHH1 (SYY2686) (He and Jacobson, 2015). The yeast two-hybrid tester strain (GGY1::171) and its isogenic derivatives harboring deletions of EDC3 (SYY1774), PAT1 (SYY2451), and DHH1 (SYY2467), or the dcp2-N245 truncation of the Dcp2 C-terminal domain (SYY2390) were described previously (He and Jacobson, 2015). Yeast strains constructed in this study are described below.

Construction of yeast strains harboring dcp2 element deletions

Request a detailed protocol

To assess the consequences of deleting specific Dcp2 regulatory elements on mRNA decay, we constructed a set of yeast strains with the W303 background that harbor different dcp2 element deletion alleles. Each of these alleles was tagged by a triple-HA epitope at the 5′-end of its coding sequence and was integrated at the genomic locus of DCP2 by gene replacement (Guthrie and Fink, 1991). As a control, we also constructed a yeast strain harboring the HA-tagged WT DCP2 allele. Plasmids harboring different dcp2-KanMX6 alleles were digested by NotI to release the DNA fragments harboring specific dcp2 element deletion knock-in alleles. About 2 µg of each digested plasmid was transformed into the WT yeast strain HFY114 by the high-efficiency LiOAC method (Schiestl and Gietz, 1989). After transformation, cells were cultured in 1 ml YEPD media at room temperature for 90 min and then plated on G418-containing YEPD plates to select for integration events. Plates were incubated at 30°C for 3–4 days. Individual stable transformants were isolated from the plates and patched again on G418-containing YEPD plates. The patched transformants served as master cells for both genotyping and long-term storage. The correct integration and deletion of specific elements for each dcp2-kanMX knock-in allele were confirmed by genomic DNA PCR and sequencing. The primers used for genomic DNA PCR and sequencing are listed in Supplementary file 3. Approximately 10 transformants were screened for each integrative yeast transformation.

To assess the decay mechanisms of different decapping substrates, we also introduced an EDC3 deletion into yeast cells harboring dcp2-E3D or E3D1 alleles, as well as deletions of XRN1, SKI2, or SKI7 into yeast cells harboring dcp2-E3D, E3D1, U1D1-U1D2, and LD1-9 alleles. Plasmids harboring the edc3::URA3, xrn1::ADE2, ski2::URA3, and ski7::URA3 null alleles were used for yeast transformation. In each case, a NotI–SalI DNA fragment harboring the respective null allele was used for gene replacement. Each of the knock-out alleles was confirmed by genomic DNA PCR. In standard YEPD media at 30°C, yeast strains harboring individual dcp2 element mutant alleles grow at rates comparable to those of WT cells, with doubling times of about 90 min.

Construction of yeast strains expressing different Dcp2 C-terminal truncations

Request a detailed protocol

To assess the roles of the Dcp2 C-terminal domain in the formation of different decapping complexes, we constructed a set of two-hybrid tester strains harboring different C-terminal truncations of Dcp2 in both EDC3 and edc3Δ backgrounds. These strains were also constructed by gene replacement. The transformation and selection procedures were identical to those described for the construction of yeast strains harboring different dcp2 element deletion alleles except that: (1) plasmids harboring different dcp2 C-terminal truncation knock-in alleles were used, and (2) each digested plasmid was transformed into the GGY1::171 and GGY1::171 edc3::HygroR (SYY3064) stains. The correct integration and C-terminal truncation for each dcp2-kanMX knock-in allele were confirmed by genomic DNA PCR and sequencing.

To assess the roles of Edc3, Upf1, Pat1, and Scd6 in the formation of different decapping complexes, we constructed additional yeast two-hybrid tester strains harboring single gene deletions of EDC3, UPF1, or SCD6; double gene deletions of UPF1, PAT1, or SCD6 and EDC3; and the dcp2-N245 C-terminal truncation of Dcp2 and EDC3 deletion. Strains harboring the single EDC3, UPF1, or SCD6 deletions were constructed by transforming a DNA fragment containing either the edc3::HygroR or the upf1::KanMX6 or the scd6::KanMX6 null allele into the GGY1::171 strain. Strains harboring the double deletions of UPF1, PAT1, or SCD6 and EDC3, and strains harboring the dcp2-N245 C-terminal truncation and EDC3 deletion were constructed by transforming a DNA fragment containing the edc3::HygroR null allele into the single UPF1, PAT1, or SCD6 deletion strains upf1::KanMX6 (SYY2973), pat11::KanMX6 (SYY2451), scd6::KanMX6 (SYY2976) strains or the dcp2-N245 truncation strain dcp2-N245::KanMX6 (SYY2390), respectively. The edc3::HygroR and scd6::KanMX6 knock-out alleles in the respective strains were confirmed by genomic DNA PCR.

Yeast two-hybrid interaction assay

Request a detailed protocol

Two-hybrid assays employed previously described procedures (Fields and Song, 1989; He et al., 1997; He and Jacobson, 1995; He and Jacobson, 2015; He et al., 2014) and all tester strains used were in the GGY1::171 background. In each case, a GAL4(DB) fusion construct (1.5 µg) was cotransformed with a GAL4(AD) construct (1.5 µg) into a tester strain using the high-efficiency LiOAc method. Transformants were plated on standard synthetic -leu-his drop-out media, incubated for 3–5 days at 30°C, and then replica-plated on X-Gal-containing SSX plates to observe the color development of the transformant population. The color phenotypes of two independent transformants from each interaction assay are presented in the figures. To assess the potential for self-activation, each of the GAL4(DB) fusions was also cotransformed with a GAL4(AD) empty vector into the GGY1::171 strain. In this study, only GAL4(DB)-DHH1 exhibited weak self-activation (He and Jacobson, 2015).

Cell samples for RNA isolation

Request a detailed protocol

Cells used for RNA isolation were all grown in YEPD media at 30°C. In each case, 15 ml YEPD in a 50 ml tube was inoculated with 1.5 OD600 overnight culture and the resulting culture was grown in a shaking incubator (200 rpm) to an OD600 of 0.7. Yeast cells in the 50 ml tube were pelleted by centrifugation in a benchtop centrifuge at 5000 rpm for 5 min. Cell pellets were resuspended in 0.5 ml fresh YEPD liquid medium and the cell suspension was then transferred to a 2 ml microcentrifuge tube. Cells were pelleted by centrifugation in a microcentrifuge at 12,000 rpm for 1 min and the liquid medium was removed from the tube. Cell pellets were frozen on dry ice and then stored at −80°C until RNA isolation.

RNA isolation

Request a detailed protocol

Total RNA was isolated from yeast cells by using the hot phenol method described previously (Herrick et al., 1990). Briefly, each cell pellet from a 15-ml culture was resuspended in 500 µl buffer A (50 mM NaOAc pH5.2, 10 mM Ethylenediaminetetraacetic acid (EDTA), 1% Sodium dodecyl sulfate (SDS), 1% Diethyl pyrocarbonate (DEPC)) and mixed with 500 µl phenol presaturated with 50 mM NaOAc pH5.2, 10 mM EDTA (Fisher, Cat# A92-500) prewarmed to 65°C. RNA was extracted by six cycles of 10 s of vortexing followed by a 50 s water bath incubation at 65°C. Samples were centrifuged in a microcentrifuge at 12,000 rpm for 3 min. After centrifugation, the phenol layer from each sample was removed with a Pasteur pipette and 500 µl prewarmed buffer-saturated phenol was then added to each sample, followed by another six cycles of RNA extraction. After the final extraction cycle, samples were centrifuged in a microcentrifuge at 12,000 rpm for 10 min and the aqueous layer from each sample was recovered and transferred to a new microcentrifuge tube. Additional (400 µl) phenol/chloroform/isoamyl alcohol (25:24:1) (Ambion, Cat# AM9732) was added to the sample and the mixture was vortexed for 2 min followed by a 10-min centrifugation at 12,000 rpm. The aqueous layer was recovered and subjected to another round of phenol/chloroform extraction. The aqueous layer was recovered and 40 µl NaOAc (3 M, pH 5.2) and 1 ml ethanol were added to each sample. RNA was precipitated at −70°C for 1 hr. Samples were centrifuged at 12,000 rpm for 15 min and pellets were washed two times with 70% ethanol. RNA pellets were air-dried for 15 min and dissolved in 80–100 µl RNase free distilled water. The RNA concentration of each sample was determined by measuring the A260 value of a diluted sample.

Northern blotting analysis

Request a detailed protocol

Procedures for northern blotting were described previously (He and Jacobson, 1995). In brief, 15 µg total RNA from each sample was loaded onto a formaldehyde-containing 1% agarose gel that was electrophoresed in 1× 3-(N-morpholino)propanesulfonic acid (MOPS) buffer (40 mM MOPS, 10 mM NaoAc, 1 mM EDTA, pH 7.0) overnight (1 hr at 70 V and 16 hr at 23 V). RNA separated on the gel was transferred a cellulose membrane (BioRad, Zeta-probe #1620159) by vacuum blotting with a sequence of 5 min in 50 mM NaOH/100 mM NaCl, 5 min in 100 mM Tris–HCl pH 7.0, and 1 hr in 20× SSC buffer. After the transfer, the membrane was crosslinked with a UV Stratalinker 2400 and washed with a RPDW buffer (0.1× SSC, 0.1% SDS) at 58°C for 30 min. For random primed DNA probes, prehybridization of the membrane was carried out in Pre-Hyb buffer (50% formamide, 5× SSPE, 10× Denhardt’s solution, 1% SDS, 0.5 mg/ml sheared salmon sperm DNA) at 42°C for 2 hr, and hybridization was carried out in Hyb buffer (50% formamide, 5× SSPE, 2× Denhardt’s solution, 5% dextran sulfate, 1% SDS, 0.25 mg/ml sheared salmon sperm DNA) at 42°C overnight. The membrane was washed with RPDW buffer 2× at room temperature for 10 min each, and then 2× at 58°C for 30 min each. Transcript-specific hybridization signals on the membrane were detected and imaged with a FUJI BAS-2500 analyzer. The images of northern blots were analyzed with MultiGauge software.

Random primed DNA probes were generated by using the Random Primed DNA Labeling Kit from Roche (cat# 11-004-760-001). Generally, 25–50 ng of purified DNA fragment was used to make the probe. The reaction was carried out in a total volume of 20 µl, containing 10 µl denatured DNA fragment, 2 µl 10× concentrated reaction mixture, 1 µl 10 mM dATP, 1 µl 10 mM dGTP, 1 µl 10 mM dTTP, 3 µl α-32P-dCTP (6000 Ci/mmol, Perkin Elmer, Blu513Z), and 2 µl Klenow enzyme. The reaction was incubated at 37°C for 1–4 hr and stopped by heating at 65°C for 5 min. The reactions were then diluted to 80 µl with sterile distilled water and purified with a mini-Quick-Spin Column (Roche #11814427001) according to the manufacturer’s instructions.

DNA fragments used as template for making α-32P-dCTP-labeled probes were isolated from plasmids. Depending on the specific gene, the DNA fragment covered part of the coding region, the entire coding region, both exon and intron, the entire 3′-UTR, or the entire gene. DNA sequences for each of these template DNA fragments are listed in Supplementary file 4. DNA probes used in this study included those that are specific for NMD substrates (CYH2, CAN1, ADE2, TRP1, CPA1, and EST1), Edc3 substrates (YRA1 and RPS28B), Dhh1 substrates (EDC1 and SDS23), Pat1/Lsm1 substrates (LSM3, DIF1, BUR6, and AGA1), and Pat1/Lsm1/Dhh1 substrates (HXT6, HSP12, and CHA1). Each northern blot was also hybridized with an SCR1 probe to serve as a loading control. With a few exceptions, each experiment was repeated independently two or three times.

Quantification and statistical analysis

Request a detailed protocol

The transcript-specific image signals from each northern blot were quantified with MultiGauge software. The band intensity data were saved as txt files and then exported to and analyzed in Microsoft Excel. To determine the relative expression levels of a specific transcript in different strains, transcript-specific signals in these strains were first normalized to the corresponding SCR1 signals and the resulting normalized signals in different strains were then divided by those from the WT strain. With a few exceptions, each northern blotting experiment was repeated independently at least two times. Bar graphs in relevant figures were generated by GraphPad Prism 9, mostly using the average ± SEM data.

There are a large number of northern blots in our paper (157 blots in total), and 16 out of 157 (about 10%) of the blots do not have error bars. The 16 graphs that do not have error bars were all presented in our initial phenotypic analyses of dcp2 element mutants (Figures 1C and 2B). Eight blots (RPS28B, EDC1, SDS23, HXT6, HSP12, LSM3, BUR6, and DIF1 mRNAs) in Figure 1C did not have replicates (Figure 1—figure supplement 2). However, the phenotypic analyses for each of these substrates in the relevant dcp2 element mutants were independently repeated in our subsequent experiments shown in Figures 1D, 3B,, 4B, E. Likewise, eight blots (HSP12, can1-100, ade2-1, RPS28B, EDC1, and SDS23 mRNAs, and CYH2 and YRA1 pre-mRNAs) in Figure 2B did not have replicates (Figure 2—figure supplement 1). However, the phenotypic analyses for six out eight of these substrates in relevant dcp2 element mutants were independently repeated in our subsequent experiments (Figure 4B). In short, only two mRNAs in Figure 2B, HSP12 and ade2-1, did not have repeat experiments. Further, in the experiments of Figure 2B, deletion of the leucine-rich elements has no effect. From a genetics perspective each of the individual mRNA substrates comprises an independent experiment which thus has thirteen independent repeats. Our data for this experiment are thus extremely strong even without error bars. In sum, our analyses followed rigorous standards.

Data availability

Source data associated with figures of northern blotting analyses have been deposited in the Dryad repository (https://datadryad.org/stash) and within that site can be found at https://doi.org/10.5061/dryad.pc866t1px. The availability of source data files is indicated in the text.

The following data sets were generated
    1. He F
    2. Wu C
    3. Jacobson A
    (2021) Dryad Digital Repository
    Dcp2 C-terminal cis-binding elements control selective targeting of the decapping enzyme by forming distinct decapping complexes.
    https://doi.org/10.5061/dryad.pc866t1px

References

    1. Charenton C
    2. Graille M
    (2018) mRNA decapping: finding the right structures
    Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences 373:20180164.
    https://doi.org/10.1098/rstb.2018.0164
  1. Book
    1. Guthrie C
    2. Fink GR
    (1991)
    Methods in Enzymology: Molecular Biology of Saccharomyces cerevisiae
    Academic Press.

Decision letter

  1. Ruben L Gonzalez
    Reviewing Editor; Columbia University, United States
  2. James L Manley
    Senior Editor; Columbia University, United States

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Decision letter after peer review:

[Editors’ note: the authors submitted for reconsideration following the decision after peer review. What follows is the decision letter after the first round of review.]

Thank you for submitting the paper "Dcp2 C–terminal Cis–Binding Elements Control Selective Targeting of the Decapping Enzyme by Forming Distinct Decapping Complexes" for consideration by eLife. Your article has been reviewed by 2 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and a Senior Editor. The reviewers have opted to remain anonymous.

Comments to the Authors:

We are sorry to say that, based on the reviews (below) and after consultation between the reviewers and Reviewing Editor, we have decided that your manuscript, as it stands, cannot be considered further for publication by eLife.

Specifically, there were significant concerns regarding the statistical significance and strength of the data and whether the proposed model is well–enough supported by the data and takes other possible interpretations of the data into account. The reviewers thought that it would not be unlikely that the concerns could be addressed within the two months of revision that eLife thrives for, but if you are able to satisfactorily address all the issues raised, a revised manuscript could be considered as a new submission at a future date. We apologize for the unusually long time it took to reach this decision, but hope you find the reviewers' comments useful.

Reviewer #2:

The authors address how specific mRNA decapping components arrange into larger complexes and how this influences mRNA decapping/ degradation. To that end, the authors have used a tour–de–force experimental approach and have performed a very large number of Y2H experiments and northern blots.

I feel that the data is not always very strong (lack of error–bars). I fear that the hypothesis that is made (one element interacts with one binding partner; one mRNA is degraded in one manner) is not always correct and additional interactions that can take have not been considered.

The final model that the authors propose does not contain many new aspects and neglects known aspects (e.g. higher order complexes as found in P–bodies).

Unfortunately, I feel that the data is not always very strong. I fear that the hypothesis that is made (one element interacts with one binding partner; one mRNA is degraded in one manner) is not correct and additional interactions that can take have not been considered. In that light I have a large number of remarks that show I often disagree with the drawn conclusions or that I don't see much novelty. In that light I think that the paper is not suitable for publication in eLife, but would be a better fit for a more specialized journal (e.g. RNA).

Methods: "With a few exceptions, each northern blotting experiment was repeated independently at least two times. Bar graphs in relevant figures were generated by generated by GraphPad Prism 9, mostly using the average {plus minus} SEM data.". This description is not satisfactory at all. What are "few exceptions" (in a large number of the bar–graphs error–bars are lacking, which is not acceptable; a single blot does not seem to be an exception) and why are there exceptions at all (all experiments should be in triplicates)? And what was repeated, where those technical replicates or biological replicates? In summary, I would think that one needs to have at least 3 real replicates (independent growth of the cells, independent gel, blot and analysis). Two technical replicates (where the same blot was run twice with the same sample) is not very useful. Also, all gels should be available in full size in some supplement. The differences are small in many cases and this requires data and analysis that is much more rigorous as what is currently done.

The argumentation in the manuscript assumes that specific substrates are degraded by specific mechanism. I am not sure if that is really the case. A substrate is likely degraded by a combination of multiple mechanisms. For some substrates the decay rate might be slightly enhanced by one factor, for other substrates another factor might slightly shift the use of the different pathways. Often the authors just assume that substrates are only degraded in a specific manner, which is not true.

The authors also assume that the motifs in Dcp2 have a single function, either Edc3 binding, Upf1 binding or Pat1 binding. Based on previous studies, it is however clear that Edc3 can also bind to HLMs (that the authors refer to as Pat binding motifs only). Furthermore, the region around the motif that the authors call E3 has also been shown to have an inhibitory effect on the Dcp2 enzyme (e.g. 10.1093/nar/gky233). This effect has not been discussed in detail. Deletion of such elements can of course change mRNA levels, independent of any binding to activators. Deletion of a motif can thus have multiple effects (lower recruitment of a factor and increase/ decrease of activity through direct Dcp2 inhibition).

The authors should discuss any (growth) phenotypes of the strains they prepared. In case any of the deletions or truncations are functionally important I would expect that those strains have serious defects.

Pull–down experiments between the different components would be able to complement the Y2H experiments. This is important as I wonder to what degree do the endogenous proteins in the Y2H experiments interfere with the results? E.g. is the WT Dcp2 present in the cells that address the behavior of specific Dcp2 version.

Finally, I would like to mention that it is often hard to link the text to the figures. The figures are very large and finding the exact gels/ bars quickly is not always easy.

Page 3: "Dcp2 is a 970–amino acid protein" That depends strongly on the organism, this statement should be refined.

Page 3: "Edc3 was also originally thought …". Edc3 has also been reported to be important for the formation of processing bodies. This should be mentioned.

Page 4: "Edc1 and Edc2 were isolated as high–copy suppressors …. but they do not appear to be required for mRNA decapping in vivo. ". Deletion of Edc1 and Edc2 impairs mRNA decay (e.g. 10.1093/genetics/157.1.27). Based on that they are required for in vivo mRNA decapping. Please clarify.

Figure 5: Why is there an interaction between UPF1 and Dcp1 in the DelteEdc3 background in Figure 5A (row3, column2), but this interaction is not there in Figure 5B (row1, column4).

Page 5: "Wild–type (WT) Dcp2 exhibited a strong two–hybrid interaction with Upf1". This has been shown before: doi.org/10.1371/journal.pone.0026547. And that manuscript should thus be referenced.

Page 5: "Deletion of either the first (U1D1) or the second (U1D2) Upf1–binding motif had no discernible effect on Upf1's binding to Dcp2, but loss of both Upf1–binding motifs (U1D1–U1D2) eliminated Upf1 binding to Dcp2 (Figure 1B).". It appears that E3D–ID–U1D2 has an effect on the Dcp2–UPF interaction. Is that a real effect or not?

Page 6: "However, loss of both Upf1–binding motifs caused approximately two– to three–fold increases in the levels of these transcripts". Based on the bar–graphs the effect appears less than two–fold in some cases and rarely more than two–fold.

Page 6 "These increases were much smaller in magnitude than those caused by deletion of UPF1, which usually led to >10–fold increases for these transcripts.". This finding strongly suggests that additional UPF1 binding motifs must be present in Dcp2, or that Dcp2 does not play an important role in NMD. The authors should be more specific here, and not only mention that "decapping is maybe not rate–limiting" (what would then be rate limiting?) or by mentioning that "other pathways could be important" (which ones?). In the end, it raises the question if the Dcp2:Upf1 interaction is biologically important.

Page 7: The discussion of the Edc3 effects is very handwaving. For instance: "RPS28B mRNA is not likely degraded by an alternative pathway." Why not, there can be an alternative pathway that is completely independent of Edc3. Or: "…YRA1 pre–mRNA and Edc3 may play an additional role in the decay of this transcript." What roles of Edc3 would that be if it is not substrate recruitment, and why would that other role not be important for other mRNAs? Edc3 is known to enhance Dcp2 decapping efficiency in general.

Page 7: "…E3 also controls targeting of the decapping enzyme to Dhh1–regulated mRNAs". Is there any evidence that the E3 motif can directly interact with the Dhh1 helicase? Does E3 motif e.g. contain an FDF–like motif? Furthermore, it is known that Edc3 and Dhh1 directly interact. as the authors note and is it thus not most likely that Dhh1 is recruited to Dcp2 with the E3 motif and Edc3? This could be tested by changing the FDF motif in Edc3.

Page 8:"The partial stabilization of Dhh1 substrates caused by loss of E3 can be explained similarly as described above for YRA1 pre–mRNA.". I am not sure what the authors mean. They should be more explicit.

Page 8: "Edc3 binds to the conserved 17–amino acid segment." (E3–1). It is known that the Edc3 Lsm domain interacts with leucine rich motifs. Both E3–1 and E3–2 potentially contain such motifs (Figure S1). How do the authors explain then that only E3–1 is important? They could do some simple modeling to address this on a structural level.

Page 9: "These results indicate that in the context of full–length Dcp2, leucine–rich motifs L1 to L5 control the selective binding of Pat1 to Dcp2, most likely with a contribution from each motif, in contrast to the proposed mode for Pat1 binding to Dcp2 based on structural data (Charenton et al., 2017).". I disagree. First: based on the data the interaction between Dcp2 and Pat1 appears to require around 4 HLMs. Those can be either 1–4 (as in LD9–5) or 5–9 (as in LD1–4). The L1 to L5 are not more important than the others. Second, the Pat1:Dcp2 structure clearly shows that one HLM motif interacts with the C–terminal domain from Pat1, forming a 1:1 complex. Avidity effects can than result a more efficient recruitment of Pat1 to Dcp2 when multiple HLMs (any random ones) are present in Dcp2. The off–rates are just slower when more motifs are present. The data is thus fully consistent with the structure.

Figure 2—figure supplement 1: Why does deletion of Pat or Lsm1 stabilize the YRA1 and RPS28B mRNAs? These effects are on the same level as the effects that the authors discuss in the other figures and based on that I assume that they are real. However, the bar–graphs again have no error–bars.

Page 9: "…one possible explanation for this surprising observation is that decapping is not rate–limiting". Do these mRNAs contain strong secondary structure elements in the 5' end that would slow down Xrn1? Xrn1 is normally very processive and rapid and thus not rate limiting. What could be rate limiting if not decapping?

Page 9: "…Pat1 performs an unidentified major function upstream…". Please be more specific in what that could be. This is unsatisfactory.

Page 11: "This indicates that these decapping substrates can all be decapped by an alternative route when the normal Dcp2 cis–element–mediated active recruitment of the decapping enzyme is blocked.". This plainly shows that Dcp2 can be recruited to the mRNA by alternative manners, which is not surprising. In my opinion the network of interactions is very redundant and deleting one interaction (e.g. through the removal of a Dcp2 motif) is backed up by the other interactions. Pat1 and Dhh1 interacts with Dcp1 directly for instance, maybe that is the default route and the routes via the Dcp2 C–terminal IDR might be less important. In that light, it has also been shown that the complete Dcp2 C–terminal region can be deleted without causing large in vivo effects.

Page 12: "combining the leucine–rich element deletions LD1–8, LD1–9, LD9–3, and LD9–2 with the Edc3–binding element deletions E3D or E3D1 caused additional substantial stabilization of the Edc3 substrate YRA1 pre–mRNA". This merely reflects that fact that Edc3 can interact with the HLMs and the Pat can interact with the HLM that the authors call E3 motif here.

Page 12: "...cells harboring E3D1 consistently had lower transcript levels than those harboring E3D for Edc3 and Dhh1 substrates ". I really don't see that. HA–dcp2–E3D and HA–dcp2–E3D1 are the same for instance (Figure 4—figure supplement 2, YRA1 and RPS28 mRNAs). Maybe I am looking at the wrong thing, it is sometimes hard to find the corresponding graph.

Page 12: "Similarly, cells harboring LD9–3 also had consistently lower transcript levels than those harboring LD9–2 for both Edc3 and Dhh1 substrates, suggesting that LD9–3 deletion maintains more function of Dcp2 than that of LD9–2 in decay of these mRNAs." Again, I don't see that, the bars are the same to me.

Page 13: " raising the possibility that Pat1/Lsm1 substrates can be decapped without the function of any decapping activators.". Or there are just additional interactions that recruit Dcp2 to the mRNA and that are independent of the Dcp2 C–terminal region. The Dcp2 C–terminal region is not required and only adds some additional functionality.

Page 14: "Deletion of EDC3 caused additional 7–8–fold stabilization of the Edc3 substrate YRA1 pre–mRNA in E3D or E3D1 cells". Really, maybe I am again looking at the wrong graphs, but in my interpretation the levels change from around 3 (E3D or E3D1) to around 10 (when DeltaEdc3 is there too). That is a 3.5 fold stabilization. Seem that the authors exaggerate the effect.

Page 14: " Deletion of EDC3 also caused additional stabilization of the Dhh1 substrates EDC1 and SDS23 mRNAs and the Pat1/Lsm1/Dhh1 substrates HSP12 and HXT6 mRNAs". Those effects appear very small to me. Is that statistically really relevant?

Page 17 "assembly of the Dcp1–Dcp2–Edc3–Upf1 decapping complex". Is this a new finding? It was known that Edc3 and Upf1 interact (10.1371/journal.pone.0026547). Why is this (direct) Edc3–Upf1 interaction not shown in 5D?

Page 17: "Interestingly, deletion of EDC3 enhanced Pat1:Dcp1 interaction". The most logical explanation is that Edc3 also binds to the HLMs. Deletion of Edc3 thus makes these HLMs available for the interaction with Pat1. The complete discussion in the paper seems too complex to explain this.

page 19:"suggesting that Xrn1 binding to Dcp2 may be dependent on a specific Dcp2 conformation.". The only logical explanation is that a sequence in the C–terminal region is inaccessible in a specific Dcp2 conformation. As many structures of Dcp2 are known this should be addressed in more detail. Currently the explanation is too handwaving.

Page 20:"We suspect that the observed Xrn1:Edc3 interaction may involve dimerization of exogenous Edc3 with endogenous Edc3 bound to Dcp2 in a decapping complex. ". Edc3 is always a stable dimer. Why would that only play a role here and not in the discussions above?

Page 22:"a new direct role of Edc3 in selective targeting of the decapping enzyme to Dhh1–regulated mRNAs. ". This appears trivial to me, as Dcp2:Edc3 and Edc3:Dhh1 interactions have been structurally described in detail (as the authors also write), so nothing new in my eyes.

Page 23: "Edc3–binding motif can promote assembly of at least one additional decapping complex, a Dcp1–Dcp2–Edc3–Edc3 complex….". That is trivial and can be extended to the known observation that Edc3 can bridge two Dcp2 proteins. In that manner Edc3 thus supports the formation very large complexes that contain many Dcp2 proteins that can all independently recruit factors.

Page 24:"both as a common core component of multiple decapping complexes and as a unique targeting component of specific decapping complexes provides a unified theory for explaining the apparently contradictory proposed functions for Edc3,". This, in my opinion, is generally accepted in the literature and not novel.

Page 28:" distinct multi–component decapping complexes". I am not sure if these really exist. The interactions are all weak and in a cellular environment constant rearrangements of the interactions will take place. Also, the decapping complexes are not isolated but form higher order complexes that are, when large enough. visible as P–bodies and that contain many Dcp2 proteins. The provided model is thus not relevant in my eyes as it oversimplifies things too much.

Reviewer #3:

Yeast Dcp2 is a large protein, containing a structured N terminus (residues 1–245) and a disordered C–terminus comprised of 725 amino acids. Prior studies by Jacobson and colleagues indicate that deletion of the C–terminus of Dcp2 results in dysregulation of 1/3 of all protein–coding transcripts in yeast (He et al., eLife 2018). They also previously showed that the C–terminus of Dcp2 harbors binding sites for cofactors such as Edc3, Pat1 and Upf1 which act on different classes of transcripts during normal mRNA decay and quality control pathways such as NMD. (He and Jacobson RNA, 2015).

In the present study, the authors perform an analysis of Dcp2 protein interactions with Edc3 , Pat1, and Upf1 and find the binding sites are separable with respect to the effects of steady–state levels of specific transcripts in yeast: mutation of a single Edc3 binding site in the C–terminus of Dcp2 increases steady–state levels of YRA1 and RPS28B mRNA compared to the Edc3 gene deletion to varying extents; mutation of both UPf1 binding sites in Dcp2 C–terminus partially stabilizes NMD transcripts compared to the Upf1 gene deletion; mutation of the Pat1 binding sites in the C–terminus of Dcp2 does not affect steady–state levels Pat1 sensitive transcripts. The results suggest decapping is rate–limiting for decay of Edc3 and Upf1 transcripts but not for those sensitive to Pat1.

The most significant finding in the manuscript is that Edc3 may play a broader role in mRNA decay than was previously appreciated.

EDC3 was previously shown to regulate two transcripts in budding yeast, RPS28B and YRA1 ( Badis et al., Mol Cell, 2004 ;Dong et al., Mol Cell, 2007). Surprisingly, Dhh1 substrates are upregulated when the Edc3 binding site on Dcp2 is mutated (E3D) but not when the EDC3 is deleted. This observation suggests one additional factor may bind the Edc3 binding site of Dcp2. Moreover, deletion of the EDC3 gene together with mutation in its binding site in Dcp2 causes additional stabilization of Edc3 sensitive YRA1 mRNA and, to a lesser degree, Dhh1 sensitive transcripts. These observations suggest an extra function for Edc3 outside of recruitment of Dcp2 to mRNA and provide additional evidence Edc3 is involved in decay of Dhh1 sensitive mRNAs.

Additionally, the authors uncover a failsafe mechanism for control of decapping and 5'–3' decay. Partial increases in steady–state levels of Edc3 and Dhh1 sensitive transcripts caused by mutation of Edc3 binding site (E3D) can be restored to levels observed in Edc3 and Dhh1 deletion strains when Pat1 binding sites are eliminated.

To understand the interplay between cofactors that bind the C–terminal domain of Dcp2, the authors used yeast two–hybrid analyses. The data suggest Edc3 is a core component of multiple decapping complexes that can promote interactions of the Dcp2 C–terminal domain with Upf1 and Xrn1 or antagonize its interactions with Pat1.

Some open–ended questions for future studies include: 1–to identify the functional consequences of Dcp2–Xrn1 interactions for RNA degradation. It would be interesting to determine if there is an overlap in residues of Xrn1 required for its binding to Dcp2 with those shown to bind the 80S ribosome (Tesina et al., NSMB, 20219). 2–determine the identity of cofactors that work together with Edc3 to promote degradation of Dhh1 sensitive transcripts. One wonders if Scd6 and Edc3 work together in this regard. 3–to understand the impact on how binding site mutations in the C–terminal domain Dcp2 impact subcellular localization. Prior studies showed the Edc3 binding element (E3) promotes localization of Dcp2 to P–bodies (Xing et al., eLife, 2020); deletion of SCD6 and EDC3 promotes nuclear localization of an inactive form of Dcp2 (Tishinov, J. Sci. Sci. 2021).

The manuscript would be suitable for publication in a venue such as eLife if the following concerns are addressed:

1. The authors suggest one additional factor besides Edc3 is required for the decay of Dhh1 sensitive transcripts. Might this factor be Scd6? (re Jacobson and Hinnebusch, PLoS Genet 2018).

2. Related, if both Edc3 and Scd6 bind E3, the deletion of this element might result in relocalization of Dcp2 in the nucleus as described for the Edc3 and Scd6 double deletion strain (Tishinov and Spang, J Cell Sci 2021). To test, does mutation of Dcp2 NLS (K450T, described by Tshinov et al) reduce steady–state levels of RPS28b mRNA in the Dcp2 E3D background compared to Dcp2 E3D in isolation?

3. The yeast two–hybrid data presented in Figures 5 and 6 are dense and confusing. EDC3 deletion in tester strains has differential effects on interactions of the c–terminal domain of Dcp2 with Pat1, Upf1 and Xrn1. Deletion of Edc3 promotes binding of the C–terminal domain to Pat1 but inhibits binding to Upf1 and Xrn1. Is the former result from a competition between Edc3 and Pat1 binding? Is the latter from the ability of Edc3 to remodel the C–terminal domain of Dcp2 for recruitment of Upf1 and Xrn1?

4. In the discussion, manuscript p24, the authors state "Our genetic experiments challenge this proposed function for Edc3 and suggest that the core Edc3 component of each decapping complex may inhibit the enzymatic activity or substrate binding of the decapping enzyme, consistent with the existence of an inhibitory element in the Dcp2 C–terminus." Is there any evidence to suggest Edc3 inhibits decapping, as stated in the above sentence? If there are published data showing lesions in Edc3 decrease transcript levels, it should be cited otherwise the authors should include evidence of this claim.

5. The report of Xrn1 binding to Dcp2 is preliminary, as there are no functional data to suggest binding of Xrn1 to Dcp2 is important for 5'–3' decay (Figure 6). What are the consequences of disrupting the Dcp2–Xrn1 interaction identified in this study on 5'–3' mRNA decay? As the authors note in discussion, Pat1 can also recruit Xrn1 to the decapping complex but in my opinion, it is difficult to conclude whether or not the newly identified Dcp2:Xrn1 interaction 'challenges the Pat1 coupling model' as stated on manuscript p28.

6. The model depicted in Figure 7 is one of several that are consistent with the data. Can they rule out the existence of multiple decapping complexes with Upf1, Pat1 and Edc3 linking Dcp2 to a core of an mRNP and substrate targeting is dictated by another RNP component (or components) that bind transfactors or RNA directly? It seems unlikely Dcp2 exist on its own in a cell as depicted in their model.

7. The model posits that Edc3 acts as a dimer on Edc3 and Dhh1 substrates but not on Pat1/Lsm1 or NMD substrates. I could not find data to support this. Do the authors have evidence that Edc3 acts as a monomer on the indicated transcripts? Biochemical structural and genetic data from Parker and Song labs indicate Edc3 is a dimer in solution and in yeast (Ling, MCB , 2008).

8. In addition, Figure 7 has inconsistencies with other figures. The authors show Pat1 and Edc3 co–occupy the C–terminal domain of Dcp2 in Figure 5D but in the model depicted in Figure 7 this co–occupancy is omitted? If the authors do not think Edc3 and Pat1 co–occpy Dcp2 C–terminus to promote decay of Edc3 sensitive transcripts, then how does one explain the failsafe mechanism?

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Dcp2 C–terminal Cis–Binding Elements Control Selective Targeting of the Decapping Enzyme by Forming Distinct Decapping Complexes" for further consideration by eLife. Your revised article has been evaluated by James Manley (Senior Editor), Ruben Gonzalez (Reviewing Editor), and only one of the previous two reviewers (Reviewer 3).

Please note that rather than asking Reviewer 2 to review your Response Letter and Revised Manuscript, we instead asked Reviewer 3 to comment on your assessment of and response to Reviewer 2's criticisms. As you will see from Reviewer 3's comments, outlined below for your attention, the manuscript has been improved but there are some remaining issues, including a few issues raised by Reviewer 2, that need to be addressed.

In addition to their formal comments below, Reviewer 3 also asked me to informally pass on an additional piece of constructive criticism. Specifically, Reviewer 3 found the tone of your response to be unnecessarily antagonistic, which they felt was not fully warranted. In addition, Reviewer 3 thought that, as written, the original manuscript was unfairly dismissive of much of the previous biochemical work and suspects that this dismissiveness might have impacted Reviewer 2's response to the original manuscript as well as the overall review process. Below are Reviewer 3's formal comments, which I would ask you to address in a final Response Letter and Revised Manuscript.

Reviewer #3:

The manuscript by Feng He et al. is an important contribution to the field–not only for mRNA decay but also for the area of intrinsically disordered proteins and proteins that have intrinsically disordered regions like Dcp2. It is quite challenging to reconstitute and resolve such proteins, and Feng He and colleagues have done a textbook job of mapping binding to short linear motifs (using Y2H) then assigning function to the motifs/binding partners in decay by a combination of genetic mutations and analyses of steady mRNA levels by Northern Blot. This type of analyses is 'gold standard' for moving from descriptive to truly mechanistic studies.

The authors have addressed most of my concerns, adding additional data to the manuscript, including an interaction analysis of Dcp2 with Scd6 and a discussion of the impact of E3 motif deletions on Dcp2 localization.

After considering the original submission, the revised manuscript and rebuttal I recommend publication in eLife. Some of the points in the revised manuscript require clarification, and some of the comments of Rev 2 should be addressed before the manuscript is accepted.

1. On p30 line 693 the authors speculate the Edc3 core component of each decapping complex may enhance the autoinhibitory activity of Dcp2. This implies Edc3 promotes ––instead of alleviates–– autoinhibition, contrary to what is implied in Feng et al. 2015 and Paquette et al., 2018. My question remains, what evidence is there that Edc3 is an inhibitor of decapping?

2. Related on page 25 lines 563–565, the authors suggest "Scd6 can bind full–length Dcp2 as a monomer, or dimer with the dimer having higher affinity and each monomer binding three different motifs, one from each of three pairs. " Is there literature indicating Scd6 can form a dimer? Why do the authors rule out that a single Scd6 can bind multiple motifs through its Lsm domain using allovalency, as elaborated by Klein, Pawson and Tyers (PMID: 14521832) and reviewed in PMID: 28597296. In my mind, and that of Rev 2, there are multiple models for how Pat1, Scd6 and Upf1 bind multiple sites in the Dcp2 C–terminal domain which require validation with biophysical methods beyond the scope of the authors manuscript. I suggest the authors present the simplest model consistent with their data, and mention that they cannot rule out other models. Nailing down the binding stoichiometry of cofactors for Dcp2 for a given class of mRNA would require biophysical methods beyond the scope of the present manuscript.

3. I appreciate the authors looking into whether or not deletion of the E3 motif affects decapping by altering the localization of Dcp2 but am not convinced that the data in their manuscript convincingly address this question. For example, they show steady–state levels of RPS28B mRNA increase by E3D but are not affected by the double mutation containing E3D and U1D1 (which "eliminates 12 of 18 residues of the NLS, including the four critical lysine residues changed in the K450T mutation" described by Tishnov and Spang). On ms p26 line 594 the authors tentatively conclude loss of E3 may not affect the nuclear localization of Dcp2 or the NLS identified in Dcp2 may not function as an NLS. Is it possible that there is an increase in cytoplasmic Dcp2 containing U1D1 and E3D mutations compared to Dcp2–E3D but no effect on decapping (due to lack of binding to Edc3 sensitive transcripts)? Because Tishinov and Spang directly monitor nuclear localization by fluorescence, whereas the authors of the present manuscript did not, I would suggest toning down the conclusion that the "NLS in Dcp2 identified (by Tishnov and Spang) is not an NLS." Likewise, if they think K450T abrogates nuclear localization of Dcp2 by abrogating binding of Pat1, then they ought to show that with a binding experiment, such as Y2H. Either the section should be re–written with more explicit mention of the caveats associated with inferring nuclear localization from steady–state mRNA levels, compared to direct methods to quantify localization by fluorescence microscopy as reported by Tishinov and Spang or the discussion of nuclear localization (p 25 lines 567 to p27 line 613) should be removed from this otherwise dense manuscript.

4. The authors have responded reasonably to Rev 2 's comments they deem 'appropriate'. In particular, I agree with the authors that there are an extraordinary number of northern blots (157) and 90% of them have error bars. The concern of this reviewer about rigor is unfair.

5–The authors' responses to Rev 2's comments that they deem are 'wrong' or confused are well–founded with the exception of two points that warrant further discussion or modification. (i)––the authors show in full–length Dcp2, leucine–rich motifs L1 to L5 control selective binding to Pat1 to Dcp2, most likely with a contribution from each motif, in contrast to the proposed mode of Pat1 binding to Dcp2 based on structural data (Charenton et al., 2017)." I agree with Reviewer 2 that this effect could be due to avidity, or if only one binding site on Pat1 engages a single motif (L1,L2, L3, L4 or L5) at a time , through allovalency (see point 2 above). The authors may discuss how multiple sites L1–L5 could collaborate to bind a single pocket on Pat1, as described with the crystal structure, consistent with allovalency. (ii)––There is a disagreement about whether full–length Edc3 always forms a dimer as asserted by Rev 2, or if in cells it may be regulated and form a monomer, as asserted by the authors. The authors' argument that hEdc3 is not a dimer is specious: sedimentation velocity–AUC by Ling et al., 2008 reveals a molecular mass of 94500 which is 15% smaller than the dimer (predicted to be 112,154 Da) but by no means is this magnitude of discrepancy between predicted and measured molar mass in SV–AUC unreasonable given the assumptions that go into fitting (such as partial specific volume, which will depend on protein conformation ––e.g. if the Lsm and YjefN domains are beads on a string or pack together to form a globular unit.) While Ling et al. noted that hEdc3 was somewhat aggregated in equilibrium AUC, there was no evidence of aggregation in the SV–AUC data, so I do not think Ling misinterpreted their AUC data. Note well, equilibrium AUC makes greater demands on sample stability than SV–AUC, because the former takes several days whereas the latter can be performed overnight: the aggregation observed in equilibrium AUC is simply a manifestation of a protein that is unstable over timescales of days. To be careful, Ling et all addressed whether human Edc3 is a monomer or dimer using size–exclusion chromatography, noting "full–length hEdc3 is a dimer at all concentrations tested. Again, these experiments are relatively quick (hours) compared to equilibrium AUC (days). Last, yeast two–hybrid experiments by Ling et al. on the budding yeast Edc3 are also consistent with a dimer. There are no data in the author's manuscript to suggest Edc3 is a monomer as part of the decapping mRNPs depicted in Figure 8, so like Reviewer 2, I suggest the authors take more care with their interpretation of dimeric and monomeric states of Edc3 (and Scd6 for that matter).

https://doi.org/10.7554/eLife.74410.sa1

Author response

[Editors’ note: The authors appealed the original decision. What follows is the authors’ response to the first round of review.]

Reviewer #2:

The authors should discuss any (growth) phenotypes of the strains they prepared. In case any of the deletions or truncations are functionally important I would expect that those strains have serious defects.

In contrast to the reviewer’s expectations, all our Dcp2 element mutants grow well, with doubling times similar to those of wild–type cells. We added the growth phenotypes of Dcp2 element mutants at lines 1017–1019 on pages 46–47.

Pull–down experiments between the different components would be able to complement the Y2H experiments. This is important as I wonder to what degree do the endogenous proteins in the Y2H experiments interfere with the results? E.g. is the WT Dcp2 present in the cells that address the behavior of specific Dcp2 version.

If the reviewer is recommending experiments to further confirm the existence of different decapping complexes, we totally agree. In fact, such experiments will be our major research undertaking for the next few years, but we believe that such experiments are far beyond the scope of this data–heavy manuscript. If the reviewer is suggesting that pull–down experiments be employed to confirm the binary interactions between different dcp2 mutants and Upf1, Edc3, and Pat1 shown in Figures 1B, 2A, and 4A, we disagree because these experiments were all complemented by our genetic analyses of mRNA decay phenotypes, and our genetic data are consistent with our Y2H results.

With regard to the reviewer’s concerns about Y2H experiments, two points should be noted: (i) We have substantial expertise with Y2H analysis and have used this technique successfully to identify factors in mRNA decay and map their domain–domain interactions (He et al., 1996, 1997; He et al., 2013; He and Jacobson, 1995, 2015; He et al., 2014). Our results from Y2H analyses of mRNA decay factors were consistent with the results generated by co–IP from the Izaurralde group for both fly and mammalian systems, but showed significant discrepancies with the results of pull–down experiments from the Parker group. We previously discussed these discrepancies including the Dcp2–Dhh1(Decker et al., 2007) and Dcp1 or Dcp2–Pat1(Nissan et al., 2010) interactions (He and Jacobson, 2015). We believe that Y2H and co–IP are less prone to artifacts than pull–down experiments, at least with decay factor interactions. Thus, the suggested pull–down experiments could generate additional data, but would probably create unnecessary confusion, and (ii) Our Y2H experiments in Figures 1B, 2A, and 4A assessed direct interactions between different dcp2 alleles and Upf1, Edc3, and Pat1, and the results from these experiments were very unlikely to be affected by the presence of wild–type Dcp2 in the tester strains, as Dcp2 does not form dimers and dcp2 mutant alleles were overexpressed relative to endogenous Dcp2. Only when we were assessing indirect interactions that are bridged by Dcp2 (as in the experiments of Figures 5 and 6) would the presence of Dcp2 or its specific elements have significant effects on the final outcomes. For our Y2H experiments shown in Figures 5 and 6, all the tester strains contain just the endogenous DCP2 and no extra copy of DCP2 is involved at all. For all these reasons we did not change the text in response to this comment.

Finally, I would like to mention that it is often hard to link the text to the figures. The figures are very large and finding the exact gels/ bars quickly is not always easy.

We agree with the reviewer on this point. To ensure an extremely thorough study, we analyzed 50 different dcp2 alleles, resulting in several panels that were large and busy. To address this problem, we have now added subtitles for the panels of Figures 5, 6, and 7.

Page 3: "Dcp2 is a 970–amino acid protein" That depends strongly on the organism, this statement should be refined.

We have replaced “yeast” with “Saccharomyces cerevisiae” at line 23 on page 3.

Page 3: "Edc3 was also originally thought …". Edc3 has also been reported to be important for the formation of processing bodies. This should be mentioned.

Our manuscript focuses on mRNA decay and processing bodies do not play a role in mRNA decay (Decker et al., 2007; Eulalio et al., 2007). Hence, no revisions were made to address this point.

Page 5: "Deletion of either the first (U1D1) or the second (U1D2) Upf1–binding motif had no discernible effect on Upf1's binding to Dcp2, but loss of both Upf1–binding motifs (U1D1–U1D2) eliminated Upf1 binding to Dcp2 (Figure 1B).". It appears that E3D–ID–U1D2 has an effect on the Dcp2–UPF interaction. Is that a real effect or not?

We believe that E3D–ID–U1D2 does have an effect on Dcp2–Upf1 interaction. However, this subtle effect was not caused by loss of U12, but most likely by loss of E3. This explanation is based on the results presented in Figures 5A and 5B, showing that Edc3 enhances Upf1 binding to the U11 element. Hence, no revisions were made to address this point.

Page 6: "However, loss of both Upf1–binding motifs caused approximately two– to three–fold increases in the levels of these transcripts". Based on the bar–graphs the effect appears less than two–fold in some cases and rarely more than two–fold.

The manuscript includes all the relevant quantitative data (see Figure 1—figure supplement 1, where the fold increases in dcp2–U1D1–U1D2 cells are: 2.6±0.4 for CYH2 pre–mRNA, 2.0±0.0 for can1–100 mRNA, 1.6±0.4 for ade2–1 mRNA, and 2.1±0.3 for trp11 mRNA). These fold changes indicate that our description was accurate, and no changes were made to the text.

Page 7: The discussion of the Edc3 effects is very handwaving. For instance: "RPS28B mRNA is not likely degraded by an alternative pathway." Why not, there can be an alternative pathway that is completely independent of Edc3. Or: "…YRA1 pre–mRNA and Edc3 may play an additional role in the decay of this transcript." What roles of Edc3 would that be if it is not substrate recruitment, and why would that other role not be important for other mRNAs? Edc3 is known to enhance Dcp2 decapping efficiency in general.

We adjusted the wording at line 117 on page 7 to make our statement more precise.

Page 8:"The partial stabilization of Dhh1 substrates caused by loss of E3 can be explained similarly as described above for YRA1 pre–mRNA.". I am not sure what the authors mean. They should be more explicit.

We adjusted the wording at lines 135–137 on page 8 to make our description more explicit.

Page 8: "Edc3 binds to the conserved 17–amino acid segment." (E3–1). It is known that the Edc3 Lsm domain interacts with leucine rich motifs. Both E3–1 and E3–2 potentially contain such motifs (Figure S1). How do the authors explain then that only E3–1 is important? They could do some simple modeling to address this on a structural level.

We did consider how the 17–amino acid segment (E3–1) might engage the Edc3 Lsm domain and determined that our E3–1 element matches exactly the Edc3–bingding helix of K. lactis Dcp2 in the Dcp1–Dcp2–Edc3–m7–GDP structure solved by the Graille group (Charenton et al., 2016). We also identified the residues of E3–1 that are important for Edc3 binding. A similar motif is found in S. pombe Dcp2, but is interrupted by an insertion of eight amino acids. Because our manuscript focuses on Dcp2 element functions, we did not discuss structural aspects of E3–1, E3–2, and all other HLMs. That said, our new data in Figure 7 further confirmed that Edc3 binds only to E3–1 and that Scd6 can bind to both E3–1 and E3–2.

Page 14: " Deletion of EDC3 also caused additional stabilization of the Dhh1 substrates EDC1 and SDS23 mRNAs and the Pat1/Lsm1/Dhh1 substrates HSP12 and HXT6 mRNAs". Those effects appear very small to me. Is that statistically really relevant?

Those effects are indeed very small, but they are reproducible and statistically relevant (Figure 4E, Figure 4—figure supplement 5), leading us to believe they are correct and biologically relevant. These small effects can be easily explained by our multiple–step kinetic model. No changes were made to the text in response to this comment.

Reviewer #3:

Yeast Dcp2 is a large protein, containing a structured N terminus (residues 1–245) and a disordered C–terminus comprised of 725 amino acids. Prior studies by Jacobson and colleagues indicate that deletion of the C–terminus of Dcp2 results in dysregulation of 1/3 of all protein–coding transcripts in yeast (He et al., eLife 2018). They also previously showed that the C–terminus of Dcp2 harbors binding sites for cofactors such as Edc3, Pat1 and Upf1 which act on different classes of transcripts during normal mRNA decay and quality control pathways such as NMD. (He and Jacobson RNA, 2015).

In the present study, the authors perform an analysis of Dcp2 protein interactions with Edc3, Pat1, and Upf1 and find the binding sites are separable with respect to the effects of steady–state levels of specific transcripts in yeast: mutation of a single Edc3 binding site in the C–terminus of Dcp2 increases steady–state levels of YRA1 and RPS28B mRNA compared to the Edc3 gene deletion to varying extents; mutation of both UPf1 binding sites in Dcp2 C–terminus partially stabilizes NMD transcripts compared to the Upf1 gene deletion; mutation of the Pat1 binding sites in the C–terminus of Dcp2 does not affect steady–state levels Pat1 sensitive transcripts. The results suggest decapping is rate–limiting for decay of Edc3 and Upf1 transcripts but not for those sensitive to Pat1.

The most significant finding in the manuscript is that Edc3 may play a broader role in mRNA decay than was previously appreciated.

EDC3 was previously shown to regulate two transcripts in budding yeast, RPS28B and YRA1 (Badis et al., Mol Cell, 2004 ;Dong et al., Mol Cell, 2007). Surprisingly, Dhh1 substrates are upregulated when the Edc3 binding site on Dcp2 is mutated (E3D) but not when the EDC3 is deleted. This observation suggests one additional factor may bind the Edc3 binding site of Dcp2. Moreover, deletion of the EDC3 gene together with mutation in its binding site in Dcp2 causes additional stabilization of Edc3 sensitive YRA1 mRNA and, to a lesser degree, Dhh1 sensitive transcripts. These observations suggest an extra function for Edc3 outside of recruitment of Dcp2 to mRNA and provide additional evidence Edc3 is involved in decay of Dhh1 sensitive mRNAs.

Additionally, the authors uncover a failsafe mechanism for control of decapping and 5'–3' decay. Partial increases in steady–state levels of Edc3 and Dhh1 sensitive transcripts caused by mutation of Edc3 binding site (E3D) can be restored to levels observed in Edc3 and Dhh1 deletion strains when Pat1 binding sites are eliminated.

To understand the interplay between cofactors that bind the C–terminal domain of Dcp2, the authors used yeast two–hybrid analyses. The data suggest Edc3 is a core component of multiple decapping complexes that can promote interactions of the Dcp2 C–terminal domain with Upf1 and Xrn1 or antagonize its interactions with Pat1.

Some open–ended questions for future studies include: (1) to identify the functional consequences of Dcp2–Xrn1 interactions for RNA degradation. It would be interesting to determine if there is an overlap in residues of Xrn1 required for its binding to Dcp2 with those shown to bind the 80S ribosome (Tesina et al., NSMB, 20219). (2) determine the identity of cofactors that work together with Edc3 to promote degradation of Dhh1 sensitive transcripts. One wonders if Scd6 and Edc3 work together in this regard. 3–to understand the impact on how binding site mutations in the C–terminal domain Dcp2 impact subcellular localization. Prior studies showed the Edc3 binding element (E3) promotes localization of Dcp2 to P–bodies (Xing et al., eLife, 2020); deletion of SCD6 and EDC3 promotes nuclear localization of an inactive form of Dcp2 (Tishinov, J. Sci. Sci. 2021).

The reviewer’s brief history of our research and his/her highlights of our manuscript’s most significant findings are concise and accurate. Further, the open–ended questions that ought to be addressed for future research are thought provoking, mechanistically relevant, and biologically important. We certainly intend to pursue them.

The manuscript would be suitable for publication in a venue such as eLife if the following concerns are addressed:

1. The authors suggest one additional factor besides Edc3 is required for the decay of Dhh1 sensitive transcripts. Might this factor be Scd6? (re Jacobson and Hinnebusch, PLoS Genet 2018).

This reviewer is correct about Scd6. We have strong evidence supporting the involvement of Scd6 in the decay of Dhh1–sensitive transcripts: (i) Scd6 interacts with multiple decapping factors including forming direct interactions with Dcp2 and Pat1, and forming Dcp2 bridged interactions with Dcp1 and Edc3; (ii) Scd6 binds multiple elements on Dcp2 including the Edc3–binding element and the leucine–rich motifs; (iii) Scd6’s binding to Dcp2 requires three distinct pairs of binding motifs E3–1 and E3–2, L3 and L5, and L8 and L9, and the two motifs in each of these three binding motif pairs all have redundant Scd6 binding activities, suggesting Scd6 can bind to full–length Dcp2 either as a monomer or as a dimer with the dimer having a higher affinity, and each monomer engaging three different binding motifs, one from each of the three pairs; (iv) Edc3 promotes the joining of Scd6 into the decapping complex to form a Dcp1–Dcp2–Edc3–Scd6 complex; and (v) in the absence of Edc3, Pat1 promotes the joining of Scd6 to the decapping complex to form a Dcp1–Dcp2–Scd6–Pat1 complex.

These new results indicate that Scd6 can collaborate with either Edc3 or Pat1 in promoting mRNA decapping and that Scd6 likely exists in both Edc3–containing and non–Edc3–containing decapping complexes. Our current hypothesis is that the E3–1 and E3–2 motifs on Dcp2 promote the assembly of three distinct decapping complexes containing different dimeric forms of Edc3 and Scd6, i.e., Edc3:Edc3, Edc3:Scd6, and Scd6:Scd6. Decapping complexes containing Edc3–Edc3–Dhh1 or Edc3–Scd6–Dhh1 target Dhh1–sensitive transcripts. Decapping complexes containing Edc3–Scd6–Pat1 or Scd6–Scd6–Pat1 target the Pat1/Lsm1–sensitive transcripts. Partitioning Edc3, Scd6, and Pat1 into different decapping complexes explains the genetic redundancy observed for the Edc3–binding site and the inhibitory activity of Edc3 observed for Dcp1–Pat1 interaction (Figure 5A).

We added the Scd6 data as Figure 7, described the new Scd6 results under two separate subtitles in the Results section (lines 461 to 541, pages 21 to 25), and incorporated the Scd6 results into our model in Figure 8.

2. Related, if both Edc3 and Scd6 bind E3, the deletion of this element might result in relocalization of Dcp2 in the nucleus as described for the Edc3 and Scd6 double deletion strain (Tishinov and Spang, J Cell Sci 2021). To test, does mutation of Dcp2 NLS (K450T, described by Tshinov et al) reduce steady–state levels of RPS28b mRNA in the Dcp2 E3D background compared to Dcp2 E3D in isolation?

This reviewer is suggesting a very insightful experiment whose purpose is to address whether loss of the E3 element might cause Dcp2 localization to the nucleus. To address this question, this reviewer is asking us to construct a dcp2 E3D–K450 double mutant and analyze its decapping activity by comparing it to a dcp2 E3D mutant using the steady–state levels of RPS28B mRNA as an indirect readout, as decay of this mRNA is mostly decapping limited and may be sensitive to subtle changes in Dcp2 subcellular localization.

Although we have not done the precise experiment suggested by Reviewer #3, the answer to this question is likely to already be in our data. The NLS identified by Tishinov and Spang maps to Dcp2 residues 450–467. This NLS is two residues downstream of the Pat1–binding motif L1 (residues 443–447) and has a twelve–residue overlap with the first Upf1–binding U11 motif (residues 456–475). The Dcp2 K450T mutation contains the NLS–inactivating changes in four key lysine residues (K460T, K461T, K463Tand K465T) that were eliminated by our Upf1–binding element deletion U1D1. To answer the reviewer’s question, it is necessary to compare the RPS28B mRNA levels in dcp2–E3D and dcp2–E3D–U1D1 cells. As shown in Figure 1C and Figure 1—figure supplement 2, the RPS28B mRNA levels in dcp2–E3D and dcp2–E3D–U1D1 cells are comparable. Compared to wild–type cells, fold increases for the mRNA are 1.9 in edc3∆ cells, 1.7 in dcp2–E3D cells, and 1.9 in dcp2–E3D–U1D1 cells. Thus, in the E3D background, loss of the four key lysine residues of the NLS had no detectable effect on the level of RPS28B mRNA. From this negative result, we can draw two tentative conclusions: loss of E3 may not affect the nuclear localization of Dcp2, or the NLS identified in Dcp2 may not function as an NLS.

Interestingly, we noticed that, in the E3D background, loss of the Pat1–binding motifs L1 to L8 (LD1–8) or L1 to L9 (LD1–9) consistently yielded lower RPS28B mRNA levels. As shown in Figure 4B and Figure 4–Supplement 2, the E3D–LD1–8 and E3D–LD1–9 alleles yielded lower RPS28B mRNAs levels than the E3D allele. Compared to wild–type cells, the RPS28B mRNA had about 1.60–fold increases in edc3∆, 1.40 in dcp2–E3D, and 1.2 in dcp2–E3D–LD1–8, and 1.1 in dcp2–E3D–LD1–9 cells. The effects accompanying loss of the Pat1–binding motifs on RPS28B mRNA levels were also evident in the experiments shown in Figure 4C and Figure 4–Supplement 2. Although this experiment did not include the dcp2–E3 allele as a control, edc3∆ cells serve as a proxy. E3D–U1D2–LD1–9, E3D–U1D1–LD1–9, or E3D–U1D1–U1D2–LD1–9 alleles all yielded lower RPS28B mRNA levels than the edc3∆ allele (respective fold increases of 1.1, 1.3, 1.2 vs 1.7). These results indicate that, in the E3D background, loss of the Pat1–binding leucine–rich motifs appear to result in increased cytoplasmic decapping activity, suggesting that the Pat1–binding leucine–rich motifs, but not the NLS, may actually control the nuclear import of Dcp2. This implies that Pat1 may promote Dcp2 nuclear localization under certain conditions, an idea consistent with a report that Pat1 is a shuttling protein (Teixeira and Parker, 2007). To reconcile our results with Tishinov and Spang’s observations, we propose that, in the absence of Edc3 and Scd6, the K450T mutation may abolish Pat1 binding to Dcp2 and thus block Pat1–mediated Dcp2 nuclear import.

We added the above results of the combined deletions of E3D and U1D1 or LD1–8 and L1–9 into the Results section under a subtitle “The Pat1–binding Leucine–rich Motifs May also Control the Nuclear Import of Dcp2” (lines 543 to 589, pages 25 to 27), and discussed the implication of these results in the Discussion section (lines 808 to 811 on page 36).

3. The yeast two–hybrid data presented in Figures 5 and 6 are dense and confusing. EDC3 deletion in tester strains has differential effects on interactions of the c–terminal domain of Dcp2 with Pat1, Upf1 and Xrn1. Deletion of Edc3 promotes binding of the C–terminal domain to Pat1 but inhibits binding to Upf1 and Xrn1. Is the former result from a competition between Edc3 and Pat1 binding? Is the latter from the ability of Edc3 to remodel the C–terminal domain of Dcp2 for recruitment of Upf1 and Xrn1?

We agree with the reviewer that our two–hybrid data in Figures 5 and 6 are dense and consequently confusing and we are open to any suggestion that might make our data presentation more effective. To make the specific question being tested in each experiment clear, we now added a title for each panel in Figures 5 and 6 and added similar titles to the components of our new Figure 7.

The reviewer’s interpretation that there are distinct effects of EDC3 deletion on different molecular interactions is correct. Based on the data presented in our manuscript, we postulate that Edc3 is partitioned into two different pools. One pool exists in the Dcp1–Dcp2–Edc3 core decapping complex and another pool functions as a targeting component for the Edc3 and Dhh1 substrates. Edc3 in the targeting pool competes with Pat1 for binding to the Dcp1–Dcp2–Edc3 core complex. As discussed in our responses to comment #1, the E3–1 and E3–2 motifs likely promote the assembly of three distinct decapping complexes containing different dimeric forms of Edc3 and Scd6. Among these three dimeric Edc3 and Scd6–containing complexes, only the Edc3:Scd6 and Scd6:Scd6, but not Edc3:Edc3–containing decapping complexes, can recruit Pat1. The common requirement of the E3–1 and E3–2 motifs for the binding of these different Edc3 and Scd6 dimers and the distinct binding specificities of Pat1 for these dimer–containing decapping complexes explains our observed competition between Edc3 and Pat1. Our genetic data are consistent with the interpretation that Edc3 joining to the Dcp1–Dcp2 complex either remodels the C–terminal domain of Dcp2, or provides additional weak but specific binding surfaces for factors such Upf1, Scd6, and Xrn1 to promotes more efficient assembly of distinct decapping complexes.

We incorporated the new Scd6 results into our model in Figure 8 that now can easily explain the observed competition between Edc3 and Pat1. We also discussed the role of the core Edc3 component in the assembly of different decapping complexes (see lines 676–682 on pages 30–31).

4. In the discussion, manuscript p24, the authors state "Our genetic experiments challenge this proposed function for Edc3 and suggest that the core Edc3 component of each decapping complex may inhibit the enzymatic activity or substrate binding of the decapping enzyme, consistent with the existence of an inhibitory element in the Dcp2 C–terminus." Is there any evidence to suggest Edc3 inhibits decapping, as stated in the above sentence? If there are published data showing lesions in Edc3 decrease transcript levels, it should be cited otherwise the authors should include evidence of this claim.

This is good advice and “challenge” is probably not the appropriate word to use because it implies some certainty. We have changed the problematic sentence to “Since the Dcp2 C–terminal domain also contains an autoinhibitory element (He and Jacobson, 2015; Paquette et al., 2018), we speculate that the core Edc3 component of each decapping complex may function to enhance the autoinhibitory activity of Dcp2 before the decapping enzyme is finally targeted to specific mRNPs.” This revision is located at lines 668–671 on page 30.

We did not have direct evidence to suggest Edc3 inhibits decapping. Given the dual roles of Edc3 (i.e., functioning as a core component in different decapping complexes and as a targeting component of Edc3 and Dhh1substrate–specifc decapping complexes) that we proposed in our manuscript, it is difficult to assess the potential inhibitory function of Edc3 on mRNA decapping. However, in our published work (He and Jacobson, 2015), we did have indirect evidence indicating that Edc3, and possibly also Scd6, have the postulated inhibitory activities, as loss of the Dcp2 E3 element consistently leads to a more active decapping enzyme.

5. The report of Xrn1 binding to Dcp2 is preliminary, as there are no functional data to suggest binding of Xrn1 to Dcp2 is important for 5'–3' decay (Figure 6). What are the consequences of disrupting the Dcp2–Xrn1 interaction identified in this study on 5'–3' mRNA decay? As the authors note in discussion, Pat1 can also recruit Xrn1 to the decapping complex but in my opinion, it is difficult to conclude whether or not the newly identified Dcp2:Xrn1 interaction 'challenges the Pat1 coupling model' as stated on manuscript p28.

This, too, is good advice in that we still have much to learn about Xrn1 binding to Dcp2 and the role of that interaction in decapping regulation. Accordingly, we have rewritten the last two sentences of the paragraph, changing “challenge” to “suggest” and “speculate” as in the following: “Since Pat1 targets only a subset of yeast transcripts (He et al., 2018), and loss of the Pat1–binding motifs had no effect on all tested decapping substrates (Figure 2B) yet loss of the Edc3– and Upf1–binding motifs each yielded specific effects (Figure 1C), we suggest that Pat1–mediated coupling may be limited to a small number of mRNAs and may not make a significant contribution to the overall decay of these mRNAs. We speculate that the observed Pat1:Xrn1 interaction may not recruit Xrn1 to decapped transcripts, but instead may serve to dissociate Xrn1 from decapping complexes after Pat1–mediated decapping.” These revisions are located on lines 760–766 on page 34.

6. The model depicted in Figure 7 is one of several that are consistent with the data. Can they rule out the existence of multiple decapping complexes with Upf1, Pat1 and Edc3 linking Dcp2 to a core of an mRNP and substrate targeting is dictated by another RNP component (or components) that bind transfactors or RNA directly? It seems unlikely Dcp2 exist on its own in a cell as depicted in their model.

We think it is very unlikely that multiple decapping complexes with Upf1, Pat1, and Edc3 link Dcp2 to a core of an mRNP. The following points argue against the “Dcp2–linked multiple decapping complexes” model and strongly favor our “target–specific distinct decapping complexes” model:

First, the available evidence in the literature indicates that mRNA decapping mostly occurs on polyribosomes, a conclusion supported by clear experimental results: (i) decapped mRNAs are associated with translating ribosomes (Hu et al., 2010; Hu et al., 2009; Pelechano et al., 2015); (ii) the decapping regulators Upf1–3 (Atkin et al., 1997), Dhh1 (Sweet et al., 2012), Pat1 (Wyers et al., 2000), Lsm1 (Bonnerot et al., 2000), and Scd6 (Weidner et al., 2014) are all associated with polyribosomes; and (iii) the decay enzymes Dcp2 (our unpublished data) and Xrn1 are also mostly polyribosome–associated (Tesina et al., 2019).

Second, Upf1 and Pat1 both use structured domains (Upf1–CH and Pat1–C) to bind Dcp2. The major binding elements that we identified for these two factors, U11 and L1 (Figure 5B), are just five residues apart and this would make the simultaneous binding of both factors to the same Dcp2 molecule very unlikely. We also tested for possible interactions between Upf1 and Pat1 or Scd6 in the two–hybrid system and found that they do not interact. In addition, the observed competition between Edc3 and Pat1 for Dcp2 binding (Figures 5A and 5B) argues against the “Dcp2–linked multiple decapping complexes” model, but strongly favors our “target–specific distinct decapping complexes” model.

Third, most decapping factors accumulate to some extent in P–bodies (Parker and Sheth, 2007). Localization of these decapping factors in P–bodies could potentially have some regulatory functions in mRNA decapping, but P–bodies are not the sites where mRNA decapping occurs (Decker et al., 2007; Eulalio et al., 2007). P–bodies are also not visible under normal growth conditions, but are only formed when cells are subjected to specific stress conditions or when the cellular decapping or 5’ to 3’ exonucleolytic decay activities are severely compromised (Parker and Sheth, 2007). The Parker and Rosen groups recently analyzed the composition and dynamics of yeast P–bodies (Xing et al., 2020) and all their results were generated from dcp1∆ cells that completely lack decapping activity. Their observed associations of Dcp2, Edc3, Pat1, and Upf1 with P–bodies in dcp1∆ cells are likely the consequence of the loss of decapping activity and are mostly irrelevant to the present study that was conducted under normal cellular and environmental conditions.

We also think that it is very unlikely that substrate targeting is dictated by another RNP component. In our 2015 RNA paper (He and Jacobson, 2015), we analyzed the interactions between Dcp2 and the decapping activators Edc3, Pat1, Lsm1, Dhh1, Upf1, Upf2, and Upf3. We found that Edc3, Pat1 and Upf1 bind directly to Dcp2 and that, in contrast, Lsm1, Dhh1, Upf2, and Upf3 do not bind to Dcp2 and thus are likely components of specific mRNPs. The targeting components in our model (Edc3, Upf1, and Pat1; Figure 8) are all RNA–binding proteins, and each of these factors also interacts directly with at least one specific mRNP component. Edc3 binds directly to specific RNA elements in RPS28B mRNA and YRA1 pre–mRNA (Badis et al., 2004; Dong et al., 2010; He et al., 2014) and thus is itself a component of specific mRNPs; Edc3 self–associates and binds to Dhh1 (He and Jacobson, 2015); Pat1 binds to the Lsm1–7 complex (Bouveret et al., 2000; Sharif and Conti, 2013; Wu et al., 2014), and Dhh1 (Sharif et al., 2013); and Upf1 binds to Upf2 (He et al., 1997). Thus, the targeting components Edc3, Pat1, and Upf1 directly connect the decapping enzyme to specific mRNPs.

As noted above, Dcp2 contains an autoinhibitory element (residues 344–379). We recently investigated Dcp2’s 3D structure predicted by AlphaFold and found that the inhibitory element binds at the catalytic center of Dcp2 and likely can block its substrate binding. In addition, loss of the entire Dcp2 C–terminal domain yields a disregulated but still active decapping enzyme in vivo (He et al., 2018; He and Jacobson, 2015). Collectively, these observations and results lead us to believe that Dcp2 may be able to exist on its own and that assembly of the final decapping complexes may involve multiple sequential steps. Hence, we have not changed the text to address the reviewer’s question.

7. The model posits that Edc3 acts as a dimer on Edc3 and Dhh1 substrates but not on Pat1/Lsm1 or NMD substrates. I could not find data to support this. Do the authors have evidence that Edc3 acts as a monomer on the indicated transcripts? Biochemical structural and genetic data from Parker and Song labs indicate Edc3 is a dimer in solution and in yeast (Ling, MCB, 2008).

It is correct that biochemical, structural, and genetic data from the Parker and Song labs indicate that Edc3 is a dimer in both solution and in crystal structures (Ling et al., 2008). However, this conclusion was drawn from a human Edc3 C–terminal fragment (Residues 250–507), i.e., the YjeN domain, but not the full–length protein. This Edc3 fragment lacks the Lsm domain (residues 1–82) and the disordered FDF domain (residues 83–249). Ling et al. noticed that full–length Edc3 protein formed aggregates in solution and had a sedimentation coefficient smaller than that expected for a dimer. Our published genetic experiments also indicate that yeast Edc3 forms a dimer in vivo (He and Jacobson, 2015; He et al., 2014). However, in these experiments, Edc3 was overexpressed from a high copy number plasmid.

We also have strong genetic data indicating that Edc3 can function as a monomer and that Edc3 dimerization may subject to regulation. In our 2014 MCB paper (He et al., 2014), we analyzed the Edc3 requirement for decay of YRA1 pre–mRNA and RPS28B mRNA. We found that the Edc3 Lsm domain alone is sufficient to promote efficient decay of YRA1 pre–mRNA. In contrast, decay of RPS28B mRNA requires both the Lsm and YjeN domains of Edc3. Autoregulated RPS28B mRNA decay also requires the Rps28b protein. Rps28b binds to the RB motif (residues 201–231) located in the Edc3 FDF domain (residues 70–276) (He et al., 2014; Kolesnikova et al., 2013) and loss of the RB motif stabilizes the RPS28B mRNA (Kolesnikova et al., 2013). Interestingly, loss of the entire FDF domain including the RB motif promotes constitutive destabilization of RPS28B mRNA (He et al., 2014). These data indicate that the Edc3 FDF domain encodes additional regulatory activities. As RPS28B mRNA decay requires the binding of an Edc3 dimer to its decay–inducing element in the 3’–UTR (He et al., 2014), these genetic data are consistent with the idea that Rps28b regulates the dimerization of Edc3. In our unpublished data, we mapped the Dhh1–binding region of Edc3 to residues 71–196 of the FDF domain. This fragment of Edc3 binds weakly to Dhh1. However, the binding of the fragment to Dhh1 is greatly enhanced by the YjeN domain of Edc3, indicating that Dhh1likely binds to an Edc3 dimer in vivo.

We do not have experimental evidence indicating that Edc3 acts as a monomer on NMD and Pat1/Lsm1 substrates. We proposed that monomeric Edc3 is a component of the NMD and Pat1/Lsm1 substrate–specific decapping complexes because this Edc3 subunit functions as a shared core component of all decapping complexes. The targeting of the NMD and Pat1/Lsm1–specific decapping complexes to their respective substrates is controlled by the binding of Upf1 and Pat1 to their specific Dcp2 cis elements. We proposed that an Edc3 dimer is present in the Edc3 and Dhh1 substrate–specific decapping complexes because our results suggest that one monomer functions as a core component and another monomer functions as a targeting component. In this scheme targeting of the Edc3 and Dhh1–specific decapping complexes to their respective substrates is controlled by homodimerization of the targeting and core Edc3 components. Our new Scd6 data in Figure 7 indicate that Scd6 can collaborate with either Edc3 or Pat1 to form distinct decapping complexes, suggesting that the Edc3:Scd6 dimer–containing complex may also target the Dhh1 substrates, and that the Edc3:Scd6 or Scd6:Scd6 dimer–containing decapping complexes may target the Pat1/Lsm1 substrates (Figure 8).

We proposed that monomeric Edc3 is present in the NMD and Pat1/Lsm1 substrate–specific decapping complexes also because: (i) Dcp2 contains only one Edc3–binding motif mapped to a 17–amino acid fragment (E3–1) (Figure 1D) and this fragment can only engage one Edc3 monomer for binding in the K. lactis Dcp1–Dcp2–Edc3–m7–GDP structure (Charenton et al., 2016) and (ii) Edc3 and Upf1 or Pat1 co–occupy Dcp2 C–terminus (Figure 5A) and the joining of Upf1 and Pat1 to the decapping complexes is sensitive to Edc3 copy number, as endogenous Edc3 inhibits both Edc3–Upf1 and Edc3–Pat1 interactions in the two–hybrid assay (Figure 5C).

To clarify this issue, we included some of the information discussed above in the paper (see lines 784–795 on page 35).

8. In addition, Figure 7 has inconsistencies with other figures. The authors show Pat1 and Edc3 co–occupy the C–terminal domain of Dcp2 in Figure 5D but in the model depicted in Figure 7 this co–occupancy is omitted? If the authors do not think Edc3 and Pat1 co–occpy Dcp2 C–terminus to promote decay of Edc3 sensitive transcripts, then how does one explain the failsafe mechanism?

The original Figures 7 and 5D were correct. We omitted Pat1 in the Edc3 substrate–specific decapping complex because Edc3:Edc3, Edc3:Scd6,and Scd6:Scd6 appear to compete for binding to the E3–1 and E3–2 motifs for decapping complex formation as we described in our responses to comments 1 and 3 above. Scd6 binds directly to Pat1 (Figure 7C) and thus likely provides at least one additional binding site besides the L1–L5 motifs for Pat1 in the decapping complexes. Our hypothesis is that Pat1 can joining the Edc3:Scd6 or Scd6:Scd6 dimer–containing decapping complexes, but not the Edc3:Edc3 dimer–containing decapping complexes that Edc3–sensitive transcripts.

To explain the failsafe mechanism for Edc3–sensitive substrates, we proposed that these transcripts are normally decapped rapidly through a deadenylation–independent mechanism by the Dcp1–Dcp2–Edc3–Edc3 decapping complex (Badis et al., 2004; Dong et al., 2007). When this pathway is blocked by either cis or trans mutations, the transcripts proceed through the slower deadenylation–dependent pathway and are most likely decapped by the Dcp1–Dcp2–Edc3:Scd6–Pat1 or Dcp1–Dcp2–Scd6–Scd6–Pat1 complexes. The combined element deletions of E3D or E3D1 and LD1–8 or LD1–9 block the assembly of all three possible decapping complexes and thus cause significant stabilization of Edc3–sensitive substrates.

The new Scd6 results in Figure 7 and the adjusted model in Figure 8 support our explanations.

– Methods: "With a few exceptions, each northern blotting experiment was repeated independently at least two times. Bar graphs in relevant figures were generated by generated by GraphPad Prism 9, mostly using the average {plus minus} SEM data.". This description is not satisfactory at all. What are "few exceptions" (in a large number of the bar–graphs error–bars are lacking, which is not acceptable; a single blot does not seem to be an exception) and why are there exceptions at all (all experiments should be in triplicates)? And what was repeated, where those technical replicates or biological replicates? In summary, I would think that one needs to have at least 3 real replicates (independent growth of the cells, independent gel, blot and analysis). Two technical replicates (where the same blot was run twice with the same sample) is not very useful. Also, all gels should be available in full size in some supplement. The differences are small in many cases and this requires data and analysis that is much more rigorous as what is currently done.

The reviewer jumped to the last paragraph of the methods section and exaggerated the meaning of our words “a few exceptions.” There are an extraordinary number of northern blots in our paper (157 blots in total), and only 16 out of 157 (about 10%) of the blots did not have error bars. The 16 graphs that do not have error bars were all presented in our initial phenotypic analyses of dcp2 element mutants (Figures 1C and 2B). Eight blots (RPS28B, EDC1, SDS23, HXT6, HSP12, LSM3, BUR6, and DIF1 mRNAs) in Figure 1C did not have replicates (Figure 1—figure supplement 2). However, the phenotypic analyses for each of these substrates in the relevant dcp2 element mutants were independently repeated in our subsequent experiments shown in Figures 1D, 3B, 4B, and 4E. Likewise, eight blots (HSP12, can1–100, ade2–1, RPS28B, EDC1, and SDS23 mRNAs, and CYH2 and YRA1 pre–mRNAs) in Figure 2B did not have replicates (Figure 2—figure supplement 1C). However, the phenotypic analyses for six out eight of these substrates in relevant dcp2 element mutants were independently repeated in our subsequent experiments (Figure 4B). In short, only two mRNAs in Figure 2B, HSP12 and ade2–1, did not have repeat experiments. These two mRNAs were added to Figure 2B as “bonuses” and could certainly be eliminated from the figure without affecting any of our conclusions. Further, in the experiments of Figure 2B, deletion of the leucine–rich elements has no effect. From a genetics perspective each of the individual mRNA substrates comprises an independent experiment which thus has thirteen independent repeats. Our data for this experiment are thus extremely strong even without error bars.

In sum, our analyses did follow rigorous standards, with three or more independent biological replicates. This is now noted specifically in the methods section of the paper (lines 1136 to 1151 on page 52). Clearly, we believed that experimental repeats were critically important as most changes in mRNA levels that we observed in the dcp2 element mutants were small, but biologically significant. In fact, it was because of these small changes that we were able to infer the rate–limiting steps for different decapping substrates and the likely existence of upstream functions for different decapping activators implemented before decapping enzyme recruitment. Finally, we did note that source data associated with our northern blotting figures were deposited in the Dryad repository (https://datadryad.org/stash), and within that site can be found at doi:10.5061/dryad.pc866t1px.

– The argumentation in the manuscript assumes that specific substrates are degraded by specific mechanism. I am not sure if that is really the case. A substrate is likely degraded by a combination of multiple mechanisms. For some substrates the decay rate might be slightly enhanced by one factor, for other substrates another factor might slightly shift the use of the different pathways. Often the authors just assume that substrates are only degraded in a specific manner, which is not true.

We disagree. We did not make any assumptions in the manuscript. We made the dcp2 element mutants, analyzed their respective decay phenotypes, interpreted the data, and then proposed a new model for mRNA decapping that was more consistent with our results than with previous models. Our data clearly indicate that the Edc3 binding element promotes Edc3 binding to Dcp2 and enhances the decapping of both Edc3 and Dhh1 substrates. Similarly, the two Upf1–binding motifs promote Upf1 binding to Dcp2 and enhance the decapping of NMD substrates. Our inference of different decay pathways for different decapping substrates was exclusively based on specific results, not assumptions.

– The authors also assume that the motifs in Dcp2 have a single function, either Edc3 binding, Upf1 binding or Pat1 binding. Based on previous studies, it is however clear that Edc3 can also bind to HLMs (that the authors refer to as Pat binding motifs only). Furthermore, the region around the motif that the authors call E3 has also been shown to have an inhibitory effect on the Dcp2 enzyme (e.g. 10.1093/nar/gky233). This effect has not been discussed in detail. Deletion of such elements can of course change mRNA levels, independent of any binding to activators. Deletion of a motif can thus have multiple effects (lower recruitment of a factor and increase/ decrease of activity through direct Dcp2 inhibition).

We disagree. We never assumed that “the motifs in Dcp2 have a single function” and in fact, we were eager and excited to identify new functions for different Dcp2 elements because these new functions may give us the hints about potential genetic redundancy or unusual regulatory activities. Our analyses of the Edc3–binding element provide proof that we did not have any prior assumptions. We observed that cis deletion of the Edc3–binding element, but not trans deletion of the EDC3 gene causes selective stabilization of Dhh1 substrates. Based on this observation, we inferred that the Edc3–binding motif likely has a second function and promotes the binding of another non–Edc3 factor (indeed Reviewer #3 thought that this non–Edc3 factor might be Scd6, and our new data in Figure 7 indicate that Scd6 indeed binds to the E3–1 motif). Our results indicate that the two Upf1–binding motifs promote Upf1 binding to Dcp2 and enhance the specific decapping of NMD substrates and that leucine–rich motifs L1–L5 mostly promote Pat1 binding and can enhance the specific decapping of Dhh1 substrates under some conditions. These are experimental results, not assumptions. In fact, we are still actively searching for additional factors that may bind to each of the Dcp2 elements.

With respect to Edc3–binding to HLMs, this reviewer was undoubtedly referring to early published work from the Sprangers and Izaurralde groups for S. pombe Dcp2 (Fromm et al., 2012). Using in vitro pull–down assays and NMR titration techniques, the authors of that paper identified four HLMs that bind to the Lsm domains of both Edc3 and Scd6. In their experiments, the Edc3 Lsm domain appears to have stronger binding to these HLMs than that of Scd6. While these in vitro observations likely still hold, in vivo evidence for either Edc3 or Scd6 binding and the role of such binding in mRNA decapping for each of these HLMs is still lacking. We refer the nine HLMs from S. cerevisiae Dcp2 as Pat1 binding motifs because they bind to Pat1 both in vivo (He and Jacobson, 2015a) and in vitro (Charenton et al., 2017), and each of these HLMs does not bind Edc3 (He and Jacobson, 2015a). Based on these facts, we refer to these HLMs as Pat1–binding motifs, and not Edc3–binding motifs. It would make no sense to name these S. cerevisiae Dcp2 HLMs as Edc3–binding motifs because similar S. pombe Dcp2 HLMs bind to Edc3 since the available evidence indicates that S. pombe decapping is more similar to that of mammalian systems than to that of S. cerevisiae. For example, S. pombe also contains an Edc4 homolog that is present in mammalian systems, but absent in S. cerevisiae (Wang et al., 2013). Simply put, decapping regulation in S. pombe and S. cerevisiae has sufficient significant differences that render expectations of similarities between the two systems to be unwarranted. Our new data in Figure 7 confirmed that Edc3 binds solely to E3–1, but Scd6 can bind to multiple motifs in Dcp2 including E3–1, E3–2, L3, L5, L8, and L9.

With respect to the Edc3–binding motif and its nearby inhibitory activity, this reviewer was referring to published work from the Gross group, again for S. pombe Dcp2 (Paquette et al., 2018). Based mostly on in vitro kinetic analysis of different Dcp1–Dcp2 complexes, Paquette et al. identified two Dcp2 segments that manifest inhibitory activities and named these segments as IM1 and IM2. In S. pombe Dcp2, IM1 is located immediately downstream of the second HLM, one of the Edc3–binding motifs identified in vitro by the Sprangers and Izaurralde groups as described above. IM1 is similar to the inhibitory element (IE) that we identified in S. cerevisiae Dcp2 (He and Jacobson, 2015a). We did not discuss the inhibitory activities of S. pombe Dcp2 in our paper because we felt it was not relevant to our results. In S. cerevisiae Dcp2, the Edc3–binding element of Dcp2 is 81 amino acids away from the inhibitory element. In our experiments, two deletions generated around the Edc3 binding region (E3Dand E3D1) did not touch any of the residues in the inhibitory element, but both caused selective stabilization of the Edc3 and Dhh1 substrates. We did not observe additional effects for these two deletions.

– Page 4: "Edc1 and Edc2 were isolated as high–copy suppressors …. but they do not appear to be required for mRNA decapping in vivo. ". Deletion of Edc1 and Edc2 impairs mRNA decay (e.g. 10.1093/genetics/157.1.27). Based on that they are required for in vivo mRNA decapping. Please clarify.

The reviewer was referring to published work from the Parker lab (Dunckley et al., 2001). In that paper, it was shown that deletion of EDC1 or EDC2 had no effect on decapping of the reporter MFA2 mRNA in otherwise wild–type cells, but can have an effect on mRNA decapping in strains compromised for decapping activity. “Edc1p and Edc2p are not rate limiting for mRNA decay” and “The edc1∆ and edc2∆ slow mRNA decapping in strains compromised for decapping activity” are two paragraph titles from the Results section of that paper. Reviewer #2’s statement that “Deletion of Edc1 and Edc2 impairs mRNA decay” is incorrect, i.e., it does not reflect the findings of the original paper.

– Figure 5: Why is there an interaction between UPF1 and Dcp1 in the DelteEdc3 background in Figure 5A (row3, column2), but this interaction is not there in Figure 5B (row1, column4).

The reviewer failed to understand our data. We clearly stated that “deletion of EDC3 diminished and loss of the entire C–terminal domain of Dcp2 eliminated Upf1:Dcp1 interaction”. Upf1:Dcp1 interactions were documented in both Figure 5A (row 3, column 2) and Figure 5B (row 1, column 4). The color in Figure 5B (row 1, column 4) is simply less intense because Edc3 has been lost.

– Page 5: "Wild–type (WT) Dcp2 exhibited a strong two–hybrid interaction with Upf1". This has been shown before: doi.org/10.1371/journal.pone.0026547. And that manuscript should thus be referenced.

The reviewer was referring to a paper published by the Parker group (Swisher and Parker, 2011). This paper showed that Upf1 binds to the catalytic domain of Dcp2 (aa 102 to 300) and that this Dcp2:Upf1 interaction is bridged by Edc3. Our results in the present manuscript show that Upf1 binds directly to Dcp2 through two distinct sequence elements located in the C–terminal domain of Dcp2. We have previously discussed these discrepancies (He and Jacobson, 2015a) and we do not see any value to cite this reference. It should be noted Dcp2–Upf1 interaction was originally identified in 1995 in our initial yeast two–hybrid screen using Upf1 as a bait. At the time we named the gene as NMD1 (He and Jacobson, 1995). Our more recent work showed that Dcp2–Upf1 interaction is mediated by two independent Upf1–binding motifs in Dcp2 (He and Jacobson, 2015a). We did not feel we needed a citation here.

– Page 6 "These increases were much smaller in magnitude than those caused by deletion of UPF1, which usually led to >10–fold increases for these transcripts.". This finding strongly suggests that additional UPF1 binding motifs must be present in Dcp2, or that Dcp2 does not play an important role in NMD. The authors should be more specific here, and not only mention that "decapping is maybe not rate–limiting" (what would then be rate limiting?) or by mentioning that "other pathways could be important" (which ones?). In the end, it raises the question if the Dcp2:Upf1 interaction is biologically important.

Here the reviewer is doing his/her best to ignore very important and provocative results in favor of simplistic notions of what should be going on in decapping of NMD substrates. We searched for 30 years by multiple approaches, only found two Upf1–binding motifs in Dcp2, and no additional Upf1–binding motifs. Why, then, does deletion of these binding sites not lead to mRNA stabilization equivalent to the loss of Upf1 itself? The answer most likely lies in a multistep NMD pathway that includes more than one function for Upf1. Our current models for NMD (He and Jacobson, 2015b) suggest that completion of the NMD pathway requires at least three steps, namely recognition and dissociation of a prematurely terminating ribosome, mRNA decapping, and 5’ to 3’ exonucleolytic decay (in that order). If in this pathway ribosome dissociation is the rate–limiting step, and if Upf1 functions at both ribosome dissociation and recruitment of the decapping enzyme, then loss of Upf1 will make the ribosome dissociation step even slower, markedly decrease overall decay rates of NMD substrates, and cause strong accumulation of these mRNAs. However, if loss of the two Upf1–binding motifs in Dcp2 only slows down the decapping step but not ribosome dissociation, then the overall decay rate will only be slightly reduced, resulting in weak accumulation of NMD substrates. In thinking about this explanation, it’s important to remember that many biological pathways have both rate–limiting and non–rate–limiting steps. Clearly, the decapping step could become rate–limiting if the activity of the decapping enzyme is greatly reduced, e.g., under certain cellular or environmental conditions. In short, our observation that loss of the two Upf1–binding motifs causes 2–3–fold selective stabilization of NMD substrates, provides important and novel insights into the functioning of the NMD pathway, and we do not question the biological importance of Dcp2–Upf1 interaction.

– Page 7: The discussion of the Edc3 effects is very handwaving. For instance: "RPS28B mRNA is not likely degraded by an alternative pathway." Why not, there can be an alternative pathway that is completely independent of Edc3. Or: "…YRA1 pre–mRNA and Edc3 may play an additional role in the decay of this transcript." What roles of Edc3 would that be if it is not substrate recruitment, and why would that other role not be important for other mRNAs? Edc3 is known to enhance Dcp2 decapping efficiency in general.

Here the reviewer gets several facts wrong and reveals that he/she is not up to date on the relevant literature. Our discussion of the Edc3 effects is not handwaving, but is based on solid genetic principles of unique functions and functional redundancies that operate in all biological systems. These principles led us to formulate our hypothesis, design follow–up experiments, and uncover the backup decay pathway for both Edc3 and Dhh1 substrates. The reviewer’s question (“why would that other role not be important for other mRNAs?”) clearly indicates that he/she is not familiar with important questions in the field.

With regard to the in vivo function of Edc3, published work from the Jacquier group as well as ours clearly established that Edc3 functions as a transcript–specific decapping activator, as deletion of EDC3 only caused stabilization of two transcripts in the entire transcriptome, RPS28B mRNA and YRA1 pre–mRNA (Badis et al., 2004; Dong et al., 2007). The two transcripts targeted by Edc3, RPS28B mRNA and YRA1 pre–mRNA, are very unusual and both contain specific Edc3–binding elements. RPS28B mRNA contains an Edc3–binding element in its long 3’–UTR (Badis et al., 2004; He et al., 2014) and YRA1–pre–mRNA contains two Edc3–binding elements in its large intron (Dong et al., 2010; Dong et al., 2007). Although both transcripts are targeted by Edc3, our experiments indicate that the decay mechanisms for these two transcripts are largely distinct (He et al., 2014). Further, based on genetic analyses of the Edc3–binding element presented in our current manuscript, Edc3 may only have one role in RPS28 mRNA decay, namely to recruit the decapping enzyme. In contrast, Edc3 likely has two roles in YRA1 pre–mRNA decay, recruitment of the decapping enzyme and repressing translation or modifying the mRNP.

It has been suggested by the Parker group that Edc3 functions as a general mRNA decapping activator (Kshirsagar and Parker, 2004), but this proposed general function of Edc3 does not fit with the absence of any effect of EDC3 deletion on any mRNAs other than RPS28B mRNA and YRA1 pre–mRNA. Indeed, the Parker group did not see any effect of EDC3 deletion on their reporter MFA2 mRNA. Hence, the big question to emerge from results demonstrating Edc3’s transcript–specific effects has been “is Edc3 really a general decapping activator?” In short, to the best our knowledge, there is no evidence in the literature indicating that S. cerevisiae Edc3 enhances Dcp2 decapping efficiency in general. That said, we should note that the experiments in our manuscript did uncover additional functions for Edc3, but not that of a general decapping activator (see below).

– Page 7: "…E3 also controls targeting of the decapping enzyme to Dhh1–regulated mRNAs". Is there any evidence that the E3 motif can directly interact with the Dhh1 helicase? Does E3 motif e.g. contain an FDF–like motif? Furthermore, it is known that Edc3 and Dhh1 directly interact. as the authors note and is it thus not most likely that Dhh1 is recruited to Dcp2 with the E3 motif and Edc3? This could be tested by changing the FDF motif in Edc3.

Here, too, the reviewer appears to lack familiarity with the relevant literature or to have misunderstood the data presented in our manuscript. First, in our earlier published work, we demonstrated that Dhh1 does not bind Dcp2 in vivo (He and Jacobson, 2015a). In the same paper, we also showed that additional factors such Upf2, Upf3, and Lsm1 also do not bind Dcp2. Second, the data we presented in Figures 5A and B clearly show that Dhh1 does not bind Dcp2 directly and that Dhh1 joins the decapping complex through Edc3. The latter is substantiated by our demonstration that loss of Edc3 eliminates Dhh1:Dcp1 interaction.

– Page 9: "These results indicate that in the context of full–length Dcp2, leucine–rich motifs L1 to L5 control the selective binding of Pat1 to Dcp2, most likely with a contribution from each motif, in contrast to the proposed mode for Pat1 binding to Dcp2 based on structural data (Charenton et al., 2017).". I disagree. First: based on the data the interaction between Dcp2 and Pat1 appears to require around 4 HLMs. Those can be either 1–4 (as in LD9–5) or 5–9 (as in LD1–4). The L1 to L5 are not more important than the others. Second, the Pat1:Dcp2 structure clearly shows that one HLM motif interacts with the C–terminal domain from Pat1, forming a 1:1 complex. Avidity effects can than result a more efficient recruitment of Pat1 to Dcp2 when multiple HLMs (any random ones) are present in Dcp2. The off–rates are just slower when more motifs are present. The data is thus fully consistent with the structure.

We disagree with this reviewer on both points. First, our data in Figure 2A clearly show that full–length Dcp2 lacking the first five motifs (L1–5), but containing the last four L6–9 motifs (LD1–5 allele) does not bind Pat1. This indicates that the last four L6–9 motifs have no Pat1 binding activity. This is a straightforward observation and it is consistent with our C–terminal element deletion analysis showing that loss of last four L6–9 had no effect on Pat1–binding (LD9–6 allele), indicating that the last four L6–9 motifs are not required for Pat1 binding activity. We made our conclusion based on these two facts and can’t rationalize an alternative conclusion based on avidity theory. Second, it is true that the Pat1–Dcp2 structural work from the Graille group (Charenton et al., 2017) showed that “one HLM motif interacts with the C–terminal domain from Pat1, forming a 1:1 complex.” However, this reviewer should know how the Graille group generated the Pat1–Dcp2 crystals. Based on our understanding of their paper, the Graille group used a C–terminal fragment of Pat1 (residues 435 to 796) and incubated it individually with short Dcp2 peptides of HLM2, 3, and 10 motifs. They never tested whether the Pat1 C–terminal fragment may bind to two or more different HLM motifs. Whether the full–length Pat1 may have additional HLM–binding regions is also a big question for us. Third, we also have yeast two–hybrid Pat1–Xrn1 interaction data indicating that the Pat1 C–terminal domain is inefficient for Xrn1 binding, suggesting the Pat1–Xrn1 HLM interaction observed in crystals by the Graille group may not occur in vivo or on full–length proteins.

– Figure 2—figure supplement 1: Why does deletion of Pat or Lsm1 stabilize the YRA1 and RPS28B mRNAs? These effects are on the same level as the effects that the authors discuss in the other figures and based on that I assume that they are real. However, the bar–graphs again have no error–bars.

This claim by the reviewer is totally false. Deletion of Pat1 or Lsm1 does not cause stabilization YRA1pre–mRNA and RPS28B mRNA. Rather, we see a slight destabilization of these two transcripts that is consistent with our model for competition between different decapping complexes (in Figure 2—figure supplement 1: when compared to the corresponding levels in wild–type cells, YRA1 pre–mRNA showed a level of 0.6 in both pat1∆ and lsm1∆ cells, and RPS28B mRNA also showed a level of 0.6 in both pat1∆ and lsm1∆ cells).

– Page 9: "…one possible explanation for this surprising observation is that decapping is not rate–limiting". Do these mRNAs contain strong secondary structure elements in the 5' end that would slow down Xrn1? Xrn1 is normally very processive and rapid and thus not rate limiting. What could be rate limiting if not decapping?

The reviewer has missed our point, namely that the rate limiting step in decay of these mRNAs is upstream of decapping, not downstream of decapping. And, no, the mRNAs do not contain strong secondary structures.

– Page 9: "…Pat1 performs an unidentified major function upstream…". Please be more specific in what that could be. This is unsatisfactory.

Our manuscript is largely structured on principles of genetics to define genes or their functions. A term like “an unidentified major function upstream or downstream” is standard in genetics.

– Page 11: "This indicates that these decapping substrates can all be decapped by an alternative route when the normal Dcp2 cis–element–mediated active recruitment of the decapping enzyme is blocked.". This plainly shows that Dcp2 can be recruited to the mRNA by alternative manners, which is not surprising. In my opinion the network of interactions is very redundant and deleting one interaction (e.g. through the removal of a Dcp2 motif) is backed up by the other interactions. Pat1 and Dhh1 interacts with Dcp1 directly for instance, maybe that is the default route and the routes via the Dcp2 C–terminal IDR might be less important. In that light, it has also been shown that the complete Dcp2 C–terminal region can be deleted without causing large in vivo effects.

In this comment the reviewer made one assumption and several other points that he/she treated as facts. Both the assumption and the “facts” are false. The reviewer stated that “Pat1 and Dhh1 interacts with Dcp1 directly for instance”, but, to the best our knowledge, there are no data in the literature indicating that S. cerevisiae Pat1 and Dhh1 interact with Dcp1 directly. The data in this manuscript (Figure 5A) clearly established that Pat1 and Dhh1 both have no direct interaction with Dcp1. Further, the reviewer stated that “the complete Dcp2 C–terminal region can be deleted without causing large in vivo effects.” This statement, too, is false. As evidence, here’s title of the first paragraph from our 2018 eLife paper: “Elimination of the large Dcp2 C–terminal domain causes significant changes in genome–wide mRNA expression” (He et al., 2018). To be precise, that study demonstrated that 1530 transcripts were significantly affected by deletion of the large Dcp2 C–terminal domain. As for the assumption (“This plainly shows that Dcp2 can be recruited to the mRNA by alternative manners”), our published work indicated that decapping of some mRNAs does not require enhanced recruitment of Dcp2 at all (He and Jacobson, 2015a). For these mRNAs, decay is unaffected even by loss of the entire Dcp2 C–terminal domain and Pat1, Lsm1, Dhh1, and Edc3. These data, as well as our observations of the decay behaviors for Pat1/Lsm1 substrates, can be easily explained by a kinetic model without evoking network interactions.

– Page 12: "combining the leucine–rich element deletions LD1–8, LD1–9, LD9–3, and LD9–2 with the Edc3–binding element deletions E3D or E3D1 caused additional substantial stabilization of the Edc3 substrate YRA1 pre–mRNA". This merely reflects that fact that Edc3 can interact with the HLMs and the Pat can interact with the HLM that the authors call E3 motif here.

Once again, the reviewer failed to recognize facts and made false assumptions. Our earlier Dcp2 element mapping experiments clearly demonstrated that Edc3 and Pat1 bind to specific motifs in Dcp2 (He and Jacobson, 2015a). Edc3 binds to the E3–1 element and Pat1 binds to the L1–9 elements. Our Dcp2 element deletion analyses in Figures 1A, 2A, and 4A further established the binding specificity of Edc3 and Pat1. As for the assumptions, the results with three dcp2 alleles proved them wrong, e.g., (i) “Edc3 can interact with the HLMs.” Please examine the E3D and E3D1 alleles in Figure 1B. These two alleles have all nine L1–9 motifs intact, yet show no Edc3–binding at all. Thus, Edc3 does not bind HLMs; (ii) “Pat can interact with the HLM that the authors call E3 motif here.” Please examine the U1D1–UD2–LD1–9 allele in Figure 4B. This dcp2 allele has an intact E3 element, but lacks all other elements, yet it has no Pat1–binding at all. Thus, Pat1 does not bind the HLM that we named E3.

In contrast to the reviewer’s erroneous assumptions and interpretations, our genetic data clearly suggest that decay of Edc3 and Dhh1 substrates occurs in ordered, multiple–step processes and that Dhh1 and Pat1 function at different steps of the mRNA decay pathway.

– Page 12: "..cells harboring E3D1 consistently had lower transcript levels than those harboring E3D for Edc3 and Dhh1 substrates ". I really don't see that. HA–dcp2–E3D and HA–dcp2–E3D1 are the same for instance (Figure 4—figure supplement 2, YRA1 and RPS28 mRNAs). Maybe I am looking at the wrong thing, it is sometimes hard to find the corresponding graph.

The reviewer was looking at the wrong dcp2 alleles. Here we are analyzing the genetic relationships between the Edc3–binding element and the Pat1–binding elements L1–9 in Figure 4B and Figure 4—figure supplement 2. The relevant dcp2 alleles are those containing the combined deletions of E3D or E3D1 and LD1–8, LD1–9, LD9–2, or LD9–3, but not the dcp2 alleles containing single element deletions of E3D or E3D1. Our described differences are clear in both the northern blots shown in Figure 4B and the bar graphs shown in Figure 4—figure supplement 2. The mRNA decay phenotypes of the single element deletions of E3D and E3D1 were already established in Figures 1C,1D, and 3B.

– Page 12: "Similarly, cells harboring LD9–3 also had consistently lower transcript levels than those harboring LD9–2 for both Edc3 and Dhh1 substrates, suggesting that LD9–3 deletion maintains more function of Dcp2 than that of LD9–2 in decay of these mRNAs." Again, I don't see that, the bars are the same to me.

Again, the reviewer was looking at the wrong dcp2 alleles. The described differences are evident in both the northern blots shown in Figure 4B and the bar graphs shown in Figure 4—figure supplement 2.

– Page 13: " raising the possibility that Pat1/Lsm1 substrates can be decapped without the function of any decapping activators.". Or there are just additional interactions that recruit Dcp2 to the mRNA and that are independent of the Dcp2 C–terminal region. The Dcp2 C–terminal region is not required and only adds some additional functionality.

Although the reviewer suggests alternative interpretations for our observation, we think it is most easily explained by a multiple–step kinetic model.

– Page 14: "Deletion of EDC3 caused additional 7–8–fold stabilization of the Edc3 substrate YRA1 pre–mRNA in E3D or E3D1 cells". Really, maybe I am again looking at the wrong graphs, but in my interpretation the levels change from around 3 (E3D or E3D1) to around 10 (when DeltaEdc3 is there too). That is a 3.5 fold stabilization. Seem that the authors exaggerate the effect.

We did not exaggerate the effect. The reviewer ignored our explanation for the quantitative difference. We always compare the deletion mutant cells to wild–type cells and, here, the cis–element deletion causes 3–fold increases relative to wild–type whereas the double deletions cause 10–fold increases relative to wild–type. Hence, the additional net effect of deleting EDC3 is 10–3=7–fold.

– Page 17 "assembly of the Dcp1–Dcp2–Edc3–Upf1 decapping complex". Is this a new finding? It was known that Edc3 and Upf1 interact (10.1371/journal.pone.0026547). Why is this (direct) Edc3–Upf1 interaction not shown in 5D?

The reviewer counters our result by citing an older paper from the Parker group in which it was suggested that Edc3 and Upf1 interact well (Swisher and Parker, 2011). However, we have previously refuted the Parker lab Edc3–Upf1 result (He and Jacobson, 2015a) and, in the same paper, established that Edc3, Pat1, and Upf1 bind to distinct elements of Dcp2. The experiments we presented in Figures 5 and 6 were specifically designed to dissect the basis of several molecular interactions that are bridged by Dcp2, i.e., interactions that are not direct. Bridged interactions between Edc3 and Upf1 were analyzed in Figures 5A and 5C and from those data we clearly stated in the Results section that “These results indicate that Dcp2 bridges an interaction between Upf1 and Edc3.” We are very troubled that this reviewer did not appear to understand our concepts, our experimental design, our results, and our conclusions.

– Page 17: "Interestingly, deletion of EDC3 enhanced Pat1:Dcp1 interaction". The most logical explanation is that Edc3 also binds to the HLMs. Deletion of Edc3 thus makes these HLMs available for the interaction with Pat1. The complete discussion in the paper seems too complex to explain this.

This reviewer is proposing a new explanation for our data and deems it to be the most logical explanation. However, our experimental evidence proved that his/her explanation is wrong. As shown in Figure 1B, we have nine dcp2 alleles lacking the E3 element but containing all nine intact HLMs (E3D, E3D–ID, E3D–U1D1, E3D–U1D2, E3D–ID–U1D1, E3D–ID–U1D2, E3D–U1D1–U1D2, E3D–ID–U1D1–U1D2, and E3D1). None of these dcp2 alleles binds to Edc3, a result indicating that Edc3 does not bind to the HLMs. As shown in Figure 5B, in cells harboring the dcp2–N300 allele with an intact E3 element (and two HLMs), Pat1 does not promote the assembly of any decapping complexes in the absence Edc3. Thus, Pat1 does not bind to the E3 element in the absence of Edc3.

– Page 19:"suggesting that Xrn1 binding to Dcp2 may be dependent on a specific Dcp2 conformation.". The only logical explanation is that a sequence in the C–terminal region is inaccessible in a specific Dcp2 conformation. As many structures of Dcp2 are known this should be addressed in more detail. Currently the explanation is too handwaving.

The reviewer’s proposed explanation is exactly the same as ours, yet when he/she presents it the explanation is the “only logical explanation” and ours is “too handwaving.” Further, the reviewer’s statement “many structures of Dcp2 are known” is inaccurate. To the best our knowledge, the only known Dcp2 structures are from S. cerevisiae, S. pombe, and K. lactis and these structures principally address just the catalytic domains and lack the regulatory domains. For S. cerevisiae, the catalytic domain contains 245 amino acids, and the regulatory domain contains 745 amino acids. The entire regulatory domain of Dcp2 appears disordered and no real structures are available. This includes the most recent results from AlphaFold.

– Page 20:"We suspect that the observed Xrn1:Edc3 interaction may involve dimerization of exogenous Edc3 with endogenous Edc3 bound to Dcp2 in a decapping complex. ". Edc3 is always a stable dimer. Why would that only play a role here and not in the discussions above?

The reviewer’s statement “Edc3 is always a stable dimer” has no definitive support in the literature. Edc3 contains three domains (Lsm, FDF, and YjeN), of which the Lsm and YjeN domains form defined structures and the FDF region is mostly disordered. Published work from the Parker and Song groups (Ling et al., 2008) demonstrated that a truncated human Edc3 YjeN domain (residues 250–507) could form dimers in both solution and in crystals. And full–length human Edc3 formed aggregates in solution and had a sedimentation coefficient smaller than that expected for a dimer (Ling et al., 2008). However, whether full–length Edc3 is a monomer, dimer, or both in vivo is unknown. In our functional analysis of Edc3, we showed that monomeric Edc3 can promote YRA1 pre–mRNA decay in vivo (He et al., 2014). We consider it likely that Edc3 dimer formation is regulated in vivo and have preliminary evidence indicating that Dhh1 and the Rps28b protein may carry out that regulation.

– Page 22:"a new direct role of Edc3 in selective targeting of the decapping enzyme to Dhh1–regulated mRNAs. ". This appears trivial to me, as Dcp2:Edc3 and Edc3:Dhh1 interactions have been structurally described in detail (as the authors also write), so nothing new in my eyes.

This reviewer is completely wrong on this point. Our finding that Edc3 controls the selective targeting of the decapping enzyme to Dhh1 substrates is one of the most significant discoveries in our manuscript. This finding broadens the list of in vivo Edc3–regulated substrates from just two transcripts to more than one thousand (to be precise, 1098 transcripts) (He et al., 2018) and establishes how Dhh1 substrates are decapped in vivo. In our opinion, and that of Reviewer #3, the identification of new substrates and delineation of a new decapping mechanism are very significant mechanistic advances. Unfortunately, Reviewer #2 used fragmented and misleading information to diminish the importance of our finding as “nothing new in my eyes.” Let’s get the facts straight. First, in several model systems, biochemical or structural data show that Dcp2:Edc3 (Charenton et al., 2016; Fromm et al., 2012; Mugridge et al., 2018) and Edc3:Dhh1 (Sharif et al., 2013) interactions occur in vitro. However, in most model systems, the cellular substrates for each of these factors are unknown. Without knowing the in vivo substrates for these factors, the biological significance of these interactions always remains an important unanswered question (but apparently not for the reviewer). Second, in essentially all the in vitro structural studies, either isolated domains or small peptides were used. As noted above, this also leads to the question of whether the observed binding or activities may still hold for the full–length proteins, not even to mention whether the observed Dcp2:Edc3 and Edc3:Dhh1 in vitro interactions be all be applied in vivo.

– Page 23: "Edc3–binding motif can promote assembly of at least one additional decapping complex, a Dcp1–Dcp2–Edc3–Edc3 complex….". That is trivial and can be extended to the known observation that Edc3 can bridge two Dcp2 proteins. In that manner Edc3 thus supports the formation very large complexes that contain many Dcp2 proteins that can all independently recruit factors.

Defining the composition of different decapping complexes is critically important for an understanding both the functions and the mechanisms of actions of mRNA decay factors. The reviewer used a baseless claim (“the known observation that Edc3 can bridge two Dcp2 proteins”) and a biased opinion (“In that manner Edc3 thus supports the formation very large complexes”) to dismiss our finding. We’re pretty sure that the reviewer would be unable to provide a reference showing that yeast “Edc3 can bridge two Dcp2 proteins.”

– Page 24:"both as a common core component of multiple decapping complexes and as a unique targeting component of specific decapping complexes provides a unified theory for explaining the apparently contradictory proposed functions for Edc3,". This, in my opinion, is generally accepted in the literature and not novel.

The reviewer made another baseless claim here. Our proposition of potential dual functions of Edc3 in mRNA decay is indeed novel and we would challenge the reviewer to provide us with a reference in which similar propositions have been put forward before our manuscript.

– Page 28:" distinct multi–component decapping complexes". I am not sure if these really exist. The interactions are all weak and in a cellular environment constant rearrangements of the interactions will take place. Also, the decapping complexes are not isolated but form higher order complexes that are, when large enough. visible as P–bodies and that contain many Dcp2 proteins. The provided model is thus not relevant in my eyes as it oversimplifies things too much.

This reviewer’s uncertainty as to whether distinct multi–component decapping complexes may really exist suggests that he/she did not examine our data carefully or did not understand our data. Using two baseless claims combined with his/her biased assumptions, the reviewer blindly rejected our fact–based model as “not relevant in my eyes” or “oversimplifies things too much.” There’s no evidence in the literature showing that “the interactions are all weak” or that “the decapping complexes are not isolated but form higher order complexes.” Contrary to the reviewer’s baseless claims or assumptions, many of the pertinent interactions are strong with nM affinities (Dutta et al., 2011; Webster et al., 2019), and several distinct decapping complexes or subcomplexes have already been purified from different biological systems (Bouveret et al., 2000; Fenger–Gron et al., 2005). We consider it appalling that a reviewer for eLife would fail to use objective criteria and use baseless claims to dismiss our model. At a minimum we would have expected his/her critique to justify his/her dismissal of our work by finding fault in our experimental designs, finding specific data to be unconvincing for reasons other than a few missing error bars, documenting why specific interpretations are wrong or which conclusions are not solid.

The authors address how specific mRNA decapping components arrange into larger complexes and how this influences mRNA decapping/ degradation. To that end, the authors have used a tour–de–force experimental approach and have performed a very large number of Y2H experiments and northern blots.

– I feel that the data is not always very strong (lack of error–bars).

We disagree. Our data are strong and follow rigorous standards. We have independent biological replicates and the mRNA decay phenotypes for each of the key dcp2 mutants were all confirmed with replicates in several independent experiments. There are an extraordinary number of northern blots in our paper (157 blots in total), and only 16 out of 157 (about 10%) of the blots did not have error bars. The 16 graphs that do not have error bars were all presented in our initial phenotypic analyses of dcp2 element mutants (Figures 1C and 2B). Eight blots (RPS28B, EDC1, SDS23, HXT6, HSP12, LSM3, BUR6, and DIF1 mRNAs) in Figure 1C did not have replicates (Figure 1—figure supplement 2). However, the phenotypic analyses for each of these substrates in the relevant dcp2 element mutants were independently repeated in our subsequent experiments shown in Figures 1D, 3B, 4B, and 4E. Likewise, eight blots (HSP12, can1–100, ade2–1, RPS28B, EDC1, and SDS23 mRNAs, and CYH2 and YRA1 pre–mRNAs) in Figure 2B did not have replicates (Figure 2—figure supplement 1C). However, the phenotypic analyses for six out eight of these substrates in relevant dcp2 element mutants were independently repeated in our subsequent experiments (Figure 4B). In short, only two mRNAs in Figure 2B, HSP12 and ade2–1, did not have repeat experiments. These two mRNAs were added to Figure 2B as “bonuses” and could certainly be eliminated from the figure without affecting any of our conclusions. Further, in the experiments of Figure 2B, deletion of the leucine–rich elements has no effect. From a genetics perspective each of the individual mRNA substrates comprises an independent experiment which thus has thirteen independent repeats. Our data for this experiment are thus extremely strong even without error bars.

In sum, our analyses did follow rigorous standards, with three or more independent biological replicates. This is now noted specifically in the methods section of the paper (lines 1136 to 1151 on page 52). Clearly, we believed that experimental repeats were critically important as most changes in mRNA levels that we observed in the dcp2 element mutants were small, but biologically significant. In fact, it was because of these small changes that we were able to infer the rate–limiting steps for different decapping substrates and the likely existence of upstream functions for different decapping activators implemented before decapping enzyme recruitment.

– I fear that the hypothesis that is made (one element interacts with one binding partner; one mRNA is degraded in one manner) is not always correct and additional interactions that can take have not been considered.

We disagree. The binding elements that we identified in Dcp2 were not based on a hypothesis, but were based on solid, published experimental evidence (He and Jacobson, 2015a). Edc3 binds to one element, Upf1 binds to two elements, and Pat1 binds multiple elements, and these are all facts, not theoretical suggestions. Our finding that Scd6 also binds to the Edc3–binding motif (E3–1) (Figure 7E) comprises further evidence that the notion is not hypothetical. We did not and would never hypothesize that Edc3 binds to multiple elements if there were no data to support the claim. Our observations that decapping of NMD substrates is enhanced by the Upf1 binding elements and that decapping of both Edc3 and Dhh1 substrates is enhanced by the Edc3 binding element are also facts, not hypotheses. Further, we did not hypothesize that “one mRNA is degraded in one manner.” Rather, we actively searched for the alternative decay pathways for individual mRNAs in different dcp2 element mutants and dedicated two whole Figures (Figures 3 and 4) to this effort. We also found an alternative decay pathway for both Edc3 and Dhh1 substrates when their Edc3–mediated decay is blocked (Figure 4B).

– The final model that the authors propose does not contain many new aspects and neglects known aspects.

We disagree. Three functions have been proposed for decapping activators, including repressing translation, activating the decapping enzyme, and sensing codon optimality. Based on the experimental data presented in our manuscript, we proposed that some decapping activators control the targeting specificity of the decapping enzyme and do so by forming distinct decapping complexes. These are new functions and a new mechanism for decapping activators. Contrary to what was implied by the reviewer, we did not neglect known aspects of higher order complexes as found in P–bodies. We chose not to discuss them because these aspects were irrelevant to our paper. First, P bodies occur under stress conditions and when cellular decapping or 5’ to 3’ exonucleolytic activities are severely compromised while our experiments were all carried out under normal cellular and growth conditions. Second, P–bodies may sequester some biomolecules or inhibit some enzymatic activities, but available evidence indicates that P–bodies accumulate only a small proportion of decay factors from the cytoplasm (Leung et al., 2006; Xing et al., 2020). Third, it is now well established that P bodies are not required for mRNA decay (Decker et al., 2007; Eulalio et al., 2007). Fourth, we favor a model of co–translational decapping for most mRNAs because: (i) decapped normal and nonsense–containing mRNAs are associated with polyribosomes (Hu et al., 2010; Hu et al., 2009; Pelechano et al., 2015); (ii) most decay factors, including Xrn1 (Tesina et al., 2019), Dhh1 (Sweet et al., 2012), and Pat1/Lsm1 (Bonnerot et al., 2000; Wyers et al., 2000) are associated with polyribosomes; and (iii) our recent unpublished data also indicate that the vast majority of Dcp2 is associated with polyribosomes.

–Unfortunately, I feel that the data is not always very strong. I fear that the hypothesis that is made (one element interacts with one binding partner; one mRNA is degraded in one manner) is not correct and additional interactions that can take have not been considered. In that light I have a large number of remarks that show I often disagree with the drawn conclusions or that I don't see much novelty. In that light I think that the paper is not suitable for publication in eLife, but would be a better fit for a more specialized journal (e.g. RNA).

We disagree. This reviewer has ignored the results in our paper that appear to contradict his/her preconceptions about decapping activators and the general mechanism of mRNA decapping.

References

Atkin, A.L., Schenkman, L.R., Eastham, M., Dahlseid, J.N., Lelivelt, M.J., and Culbertson, M.R. (1997). Relationship between yeast polyribosomes and Upf proteins required for nonsense mRNA decay. J Biol Chem 272, 22163–22172.

Badis, G., Saveanu, C., Fromont–Racine, M., and Jacquier, A. (2004). Targeted mRNA degradation by deadenylation–independent decapping. Mol Cell 15, 5–15.

Bonnerot, C., Boeck, R., and Lapeyre, B. (2000). The two proteins Pat1p (Mrt1p) and Spb8p interact in vivo, are required for mRNA decay, and are functionally linked to Pab1p. Mol Cell Biol 20, 5939–5946.

Bouveret, E., Rigaut, G., Shevchenko, A., Wilm, M., and Seraphin, B. (2000). A Sm–like protein complex that participates in mRNA degradation. EMBO J. 19, 1661–1671.

Charenton, C., Taverniti, V., Gaudon–Plesse, C., Back, R., Seraphin, B., and Graille, M. (2016). Structure of the active form of Dcp1–Dcp2 decapping enzyme bound to m7GDP and its Edc3 activator. Nat Struct Mol Biol 23, 982–986.

Decker, C.J., Teixeira, D., and Parker, R. (2007). Edc3p and a glutamine/asparagine–rich domain of Lsm4p function in processing body assembly in Saccharomyces cerevisiae. J Cell Biol 179, 437–449.

Dong, S., Jacobson, A., and He, F. (2010). Degradation of YRA1 Pre–mRNA in the cytoplasm requires translational repression, multiple modular intronic elements, Edc3p, and Mex67p. PLoS Biol 8, e1000360.

Dong, S., Li, C., Zenklusen, D., Singer, R.H., Jacobson, A., and He, F. (2007). YRA1 autoregulation requires nuclear export and cytoplasmic Edc3p–mediated degradation of its pre–mRNA. Mol Cell 25, 559–573.

Eulalio, A., Behm–Ansmant, I., Schweizer, D., and Izaurralde, E. (2007). P–body formation is a consequence, not the cause, of RNA–mediated gene silencing. Mol Cell Biol 27, 3970–3981.

He, F., Brown, A.H., and Jacobson, A. (1996). Interaction between Nmd2p and Upf1p is required for activity but not for dominant–negative inhibition of the nonsense–mediated mRNA decay pathway in yeast. Rna 2, 153–170.

He, F., Brown, A.H., and Jacobson, A. (1997). Upf1p, Nmd2p, and Upf3p are interacting components of the yeast nonsense–mediated mRNA decay pathway. Mol Cell Biol 17, 1580–1594.

He, F., Celik, A., Wu, C., and Jacobson, A. (2018). General decapping activators target different subsets of inefficiently translated mRNAs. ELife 7.

He, F., Ganesan, R., and Jacobson, A. (2013). Intra– and intermolecular regulatory interactions in Upf1, the RNA helicase central to nonsense–mediated mRNA decay in yeast. Mol Cell Biol 33, 4672–4684.

He, F., and Jacobson, A. (1995). Identification of a novel component of the nonsense–mediated mRNA decay pathway by use of an interacting protein screen. Genes Dev 9, 437–454.

He, F., and Jacobson, A. (2015). Control of mRNA decapping by positive and negative regulatory elements in the Dcp2 C–terminal domain. RNA 21, 1633–1647.

He, F., Li, C., Roy, B., and Jacobson, A. (2014). Yeast Edc3 targets RPS28B mRNA for decapping by binding to a 3' untranslated region decay–inducing regulatory element. Mol Cell Biol 34, 1438–1451.

Hu, W., Petzold, C., Coller, J., and Baker, K.E. (2010). Nonsense–mediated mRNA decapping occurs on polyribosomes in Saccharomyces cerevisiae. Nat Struct Mol Biol 17, 244–247.

Hu, W., Sweet, T.J., Chamnongpol, S., Baker, K.E., and Coller, J. (2009). Co–translational mRNA decay in Saccharomyces cerevisiae. Nature 461, 225–229.

Kolesnikova, O., Back, R., Graille, M., and Seraphin, B. (2013). Identification of the Rps28 binding motif from yeast Edc3 involved in the autoregulatory feedback loop controlling RPS28B mRNA decay. Nucleic Acids Res.

Ling, S.H., Decker, C.J., Walsh, M.A., She, M., Parker, R., and Song, H. (2008). Crystal structure of human Edc3 and its functional implications. Molecular and cellular biology 28, 5965–5976.

Nissan, T., Rajyaguru, P., She, M., Song, H., and Parker, R. (2010). Decapping activators in Saccharomyces cerevisiae act by multiple mechanisms. Mol Cell 39, 773–783.

Paquette, D.R., Tibble, R.W., Daifuku, T.S., and Gross, J.D. (2018). Control of mRNA decapping by autoinhibition. Nucleic Acids Res 46, 6318–6329.

Parker, R., and Sheth, U. (2007). P bodies and the control of mRNA translation and degradation. Mol Cell 25, 635–646.

Pelechano, V., Wei, W., and Steinmetz, L.M. (2015). Widespread Co–translational RNA Decay Reveals Ribosome Dynamics. Cell 161, 1400–1412.

Sharif, H., and Conti, E. (2013). Architecture of the Lsm1–7–Pat1 complex: a conserved assembly in eukaryotic mRNA turnover. Cell Rep 5, 283–291.

Sharif, H., Ozgur, S., Sharma, K., Basquin, C., Urlaub, H., and Conti, E. (2013). Structural analysis of the yeast Dhh1–Pat1 complex reveals how Dhh1 engages Pat1, Edc3 and RNA in mutually exclusive interactions. Nucleic Acids Res.

Sweet, T., Kovalak, C., and Coller, J. (2012). The DEAD–box protein Dhh1 promotes decapping by slowing ribosome movement. PLoS Biol 10, e1001342.

Teixeira, D., and Parker, R. (2007). Analysis of P–body assembly in Saccharomyces cerevisiae. Molecular Biology of the Cell 18, 2274–2287.

Tesina, P., Heckel, E., Cheng, J., Fromont–Racine, M., Buschauer, R., Kater, L., Beatrix, B., Berninghausen, O., Jacquier, A., Becker, T., et al. (2019). Structure of the 80S ribosome–Xrn1 nuclease complex. Nat Struct Mol Biol 26, 275–280.

Weidner, J., Wang, C., Prescianotto–Baschong, C., Estrada, A.F., and Spang, A. (2014). The polysome–associated proteins Scp160 and Bfr1 prevent P body formation under normal growth conditions. Journal of Cell Science 127, 1992–2004.

Wu, D., Muhlrad, D., Bowler, M.W., Jiang, S., Liu, Z., Parker, R., and Song, H. (2014). Lsm2 and Lsm3 bridge the interaction of the Lsm1–7 complex with Pat1 for decapping activation. Cell Res 24, 233–246.

Wyers, F., Minet, M., Dufour, M.E., Vo, L.T., and Lacroute, F. (2000). Deletion of the PAT1 gene affects translation initiation and suppresses a PAB1 gene deletion in yeast. Mol Cell Biol 20, 3538–3549.

Xing, W., Muhlrad, D., Parker, R., and Rosen, M.K. (2020). A quantitative inventory of yeast P body proteins reveals principles of composition and specificity. ELife 9, e56525.

[Editors’ note: what follows is the authors’ response to the second round of review.]

Please note that rather than asking Reviewer 2 to review your Response Letter and Revised Manuscript, we instead asked Reviewer 3 to comment on your assessment of and response to Reviewer 2's criticisms. As you will see from Reviewer 3's comments, outlined below for your attention, the manuscript has been improved but there are some remaining issues, including a few issues raised by Reviewer 2, that need to be addressed.

In addition to their formal comments below, Reviewer 3 also asked me to informally pass on an additional piece of constructive criticism. Specifically, Reviewer 3 found the tone of your response to be unnecessarily antagonistic, which they felt was not fully warranted. In addition, Reviewer 3 thought that, as written, the original manuscript was unfairly dismissive of much of the previous biochemical work and suspects that this dismissiveness might have impacted Reviewer 2's response to the original manuscript as well as the overall review process. Below are Reviewer 3's formal comments, which I would ask you to address in a final Response Letter and Revised Manuscript.

Reviewer #3 (Recommendations for the authors):

The manuscript by Feng He et al. is an important contribution to the field–not only for mRNA decay but also for the area of intrinsically disordered proteins and proteins that have intrinsically disordered regions like Dcp2. It is quite challenging to reconstitute and resolve such proteins, and Feng He and colleagues have done a textbook job of mapping binding to short linear motifs (using Y2H) then assigning function to the motifs/binding partners in decay by a combination of genetic mutations and analyses of steady mRNA levels by Northern Blot. This type of analyses is 'gold standard' for moving from descriptive to truly mechanistic studies.

The authors have addressed most of my concerns, adding additional data to the manuscript, including an interaction analysis of Dcp2 with Scd6 and a discussion of the impact of E3 motif deletions on Dcp2 localization.

After considering the original submission, the revised manuscript and rebuttal I recommend publication in eLife. Some of the points in the revised manuscript require clarification, and some of the comments of Rev 2 should be addressed before the manuscript is accepted.

1. On p30 line 693 the authors speculate the Edc3 core component of each decapping complex may enhance the autoinhibitory activity of Dcp2. This implies Edc3 promotes ––instead of alleviates–– autoinhibition, contrary to what is implied in Feng et al. 2015 and Paquette et al., 2018. My question remains, what evidence is there that Edc3 is an inhibitor of decapping?

We do not have direct evidence indicating that Edc3 is an inhibitor of decapping. However, we do have evidence indicating that the Edc3–binding motif has some inhibitory activity. In our model, we partition Edc3 into two different populations. One fraction functions as a core component of multiple decapping complexes and the other functions as a targeting component of specific decapping complexes. We hypothesized that the core Edc3 component in decapping complexes functions to enforce Dcp2 autoinhibition and that this autoinhibition is released at targeted mRNPs through Edc3 dimerization. Our in vivo experiments from our 2015 RNA paper do not contradict this model, but more experiments are needed to provide definitive proof. Paquette et al.’s in vitro experiments from their 2018 NAR paper showed that Edc3 alleviates Dcp2 autoinhibition. However, they used a four–fold molar excess of Edc3 relative to the decapping enzyme. The enhancement effects observed in these in vitro experiments could still result from Edc3 dimerization. Since we do not have actual experimental evidence for Edc3 as an inhibitor of decapping, we eliminated the problematic sentences from line 683 on page 30 to line 687 on page 31 of the original main text. In addition, to be more cautious, we also eliminated the sentence from line 779 to 781 at the end of Page 34 and the beginning of Page 35, in which we had speculated on the role of Pat1:Xrn1 interaction in mRNA decay.

2. Related on page 25 lines 563–565, the authors suggest "Scd6 can bind full–length Dcp2 as a monomer, or dimer with the dimer having higher affinity and each monomer binding three different motifs, one from each of three pairs. " Is there literature indicating Scd6 can form a dimer? Why do the authors rule out that a single Scd6 can bind multiple motifs through its Lsm domain using allovalency, as elaborated by Klein, Pawson and Tyers (PMID: 14521832) and reviewed in PMID: 28597296. In my mind, and that of Rev 2, there are multiple models for how Pat1, Scd6 and Upf1 bind multiple sites in the Dcp2 C–terminal domain which require validation with biophysical methods beyond the scope of the authors manuscript. I suggest the authors present the simplest model consistent with their data, and mention that they cannot rule out other models. Nailing down the binding stoichiometry of cofactors for Dcp2 for a given class of mRNA would require biophysical methods beyond the scope of the present manuscript.

There is no literature indicating that Scd6 can form a dimer. However, typical Lsm domain proteins do form higher order ring assemblies. We hypothesized that the Scd6 Lsm domain may form a binding–induced dimer in decapping complexes, as its two binding sites in Dcp2 (E3–1 and E3–2) are so close to each other. The binding patterns of Scd6 and Pat1 to full–length Dcp2 in the binary two–hybrid assays are complex and we struggled to generate a consistent model to explain the binding pattern for each factor. For Scd6, we derived a model in which three different regions of Scd6 engage three distinct Dcp2 binding motifs in a cooperative manner. This model explains our data reasonably well. For Pat1, we did not have a good model that can explain our observations. The binary two–hybrid data in Figure 2A suggest that two or more regions of Pat1 may engage two or more Dcp2 binding motifs, and thus can be explained by the avidity model suggested by Reviewer #2. However, Pat1 binding to decapping complexes shown in Figures 5B and 5C suggests that a single region of Pat1 engages a single Dcp2 binding motif. We hadn’t considered an allovalency model for both Scd6 and Pat1 binding to Dcp2 and are certainly open to any model that can better explain our data. We think that allovalency explains the Pat1 binding pattern of Dcp2 reasonably well and appreciate the suggestion. The allovalency model may also provide an alternative explanation for Scd6 binding to Dcp2. We have modified the text description for Pat1 and Scd6 binding to Dcp2, and now discuss the potential modes of action for Pat1 and Scd6 engagement with Dcp2. The pertinent changes to the manuscript are on page 9, lines 170 to 175, and page 25, lines 548 to 555.

3. I appreciate the authors looking into whether or not deletion of the E3 motif affects decapping by altering the localization of Dcp2 but am not convinced that the data in their manuscript convincingly address this question. For example, they show steady–state levels of RPS28B mRNA increase by E3D but are not affected by the double mutation containing E3D and U1D1 (which "eliminates 12 of 18 residues of the NLS, including the four critical lysine residues changed in the K450T mutation" described by Tishnov and Spang). On manuscript p26 line 594 the authors tentatively conclude loss of E3 may not affect the nuclear localization of Dcp2 or the NLS identified in Dcp2 may not function as an NLS. Is it possible that there is an increase in cytoplasmic Dcp2 containing U1D1 and E3D mutations compared to Dcp2–E3D but no effect on decapping (due to lack of binding to Edc3 sensitive transcripts)? Because Tishinov and Spang directly monitor nuclear localization by fluorescence, whereas the authors of the present manuscript did not, I would suggest toning down the conclusion that the "NLS in Dcp2 identified (by Tishnov and Spang) is not an NLS." Likewise, if they think K450T abrogates nuclear localization of Dcp2 by abrogating binding of Pat1, then they ought to show that with a binding experiment, such as Y2H. Either the section should be re–written with more explicit mention of the caveats associated with inferring nuclear localization from steady–state mRNA levels, compared to direct methods to quantify localization by fluorescence microscopy as reported by Tishinov and Spang or the discussion of nuclear localization (p 25 lines 567 to p27 line 613) should be removed from this otherwise dense manuscript.

Our manuscript focuses on the targeting mechanisms of the yeast decapping enzyme. While Dcp2 nuclear localization is interesting and potentially important for decapping regulation, we think that a discussion of Dcp2 nuclear localization in this manuscript is a distraction from the overall message in the paper. In addition, the data we presented on Dcp2 nuclear localization raised new issues that need further experimentation. Hence, we chose to take the advice of this reviewer and have removed the entire discussion of Dcp2 nuclear localization (originally pages 25–27, lines 557 to 603). We also eliminated the last sentence in the Discussion that addressed Dcp2 nuclear localization (page 36, lines 823 to 826).

As a scientific aside, our proposal that Pat1 mediates Dcp2 nuclear location may have some merit. We believe that cis deletions of the Edc3 and Scd6 binding motifs in Dcp2 and trans deletions of Edc3 and Scd6 have different effects on the composition of Pat1–containing decapping complexes and thus may have different effects on Dcp2 nuclear localization. Scd6 interacts directly with Pat1 (Figure 7). In Dcp2 cis E3 deletion mutant cells, Scd6 may still associate with Pat1 in the decapping complex and this association could mask a surface of Pat1, potentially its nuclear localization signals (hypothetical). In contrast, in trans EDC3 and SCD6 double deletion mutants, Pat1 associates with Dcp2 in the decapping complex and the Scd6 binding surface of Pat1 is likely exposed. The exposed surface of Pat1 could be the target of nuclear localization. In addition, the major NLS identified by Tishinov and Spang mapped to the first Upf1–binding site U11 and Upf1 binding to this site could also block nuclear import receptor binding to Dcp2. In yeast cells, Upf1 outnumbers Dcp2 (molecules/per cell: 6905 for Upf1 vs 5761 for Dcp2) and Upf1’s inhibitory activity on the NLS–mediated nuclear import of Dcp2 could be very significant.

4. The authors have responded reasonably to Rev 2 's comments they deem 'appropriate'. In particular, I agree with the authors that there are an extraordinary number of northern blots (157) and 90% of them have error bars. The major concern of this reviewer about rigor is unfair.

We thank this reviewer for taking on the added responsibility of assessing the lengthy arguments in our responses to the comments from reviewer #2.

5. The authors' responses to Rev 2's comments that they deem are 'wrong' or confused are well–founded with the exception of two points that warrant further discussion or modification. (i) The authors show in full–length Dcp2 , leucine–rich motifs L1 to L5 control selective binding to Pat1 to Dcp2, most likely with a contribution from each motif, in contrast to the proposed mode of Pat1 binding to Dcp2 based on structural data (Charenton et al., 2017)." I agree with Reviewer 2 that this effect could be due to avidity, or if only one binding site on Pat1 engages a single motif (L1,L2, L3, L4 or L5) at a time , through allovalency (see point 2 above). The authors may discuss how multiple sites L1–L5 could collaborate to bind a single pocket on Pat1, as described with the crystal structure, consistent with allovalency. (ii) There is a disagreement about whether full–length Edc3 always forms a dimer as asserted by Rev 2, or if in cells it may be regulated and form a monomer, as asserted by the authors. The authors' argument that hEdc3 is not a dimer is specious: sedimentation velocity–AUC by Ling et al., 2008 reveals a molecular mass of 94500 which is 15% smaller than the dimer (predicted to be 112,154 Da) but by no means is this magnitude of discrepancy between predicted and measured molar mass in SV–AUC unreasonable given the assumptions that go into fitting (such as partial specific volume, which will depend on protein conformation –e.g. if the Lsm and YjefN domains are beads on a string or pack together to form a globular unit.) While Ling et al. noted that hEdc3 was somewhat aggregated in equilibrium AUC, there was no evidence of aggregation in the SV–AUC data, so I do not think Ling misinterpreted their AUC data. Note well, equilibrium AUC makes greater demands on sample stability than SV–AUC, because the former takes several days whereas the latter can be performed overnight: the aggregation observed in equilibrium AUC is simply a manifestation of a protein that is unstable over timescales of days. To be careful, Ling et all addressed whether human Edc3 is a monomer or dimer using size–exclusion chromatography, noting "full–length hEdc3 is a dimer at all concentrations tested. Again, these experiments are relatively quick (hours) compared to equilibrium AUC (days). Last, yeast two–hybrid experiments by Ling et al. on the budding yeast Edc3 are also consistent with a dimer. There are no data in the author's manuscript to suggest Edc3 is a monomer as part of the decapping mRNPs depicted in Figure 8, so like Reviewer 2, I suggest the authors take more care with their interpretation of dimeric and monomeric states of Edc3 (and Scd6 for that matter).

Again, we thank reviewer #3 for assessing the lengthy arguments in our responses to reviewer #2’s comments.

As for point (i): As described above in our response to item 2, we originally thought that the avidity model suggested by reviewer #2 could partially explain the binary two–hybrid data (Figure 2A), but appeared to contradict Pat1’s binding pattern to the decapping complex shown in Figures 5B and 5C. We now think that the allovalency model is a better fit for the Pat1 binding pattern of Dcp2. As noted in item 2 (above), we now discuss this binding mode of Pat1 in the revised text of our manuscript.

As for point (ii): We did realize that our arguments based on human Edc3 in vitro analytical data from the Ling et al., 2008 MCB paper were weak and we certainly were not trying to suggest that Ling et al. misinterpreted their data. We used their analytical data to argue that Edc3 has a large, disordered region between the Lsm and YjeF–N domains and that this disordered region likely has regulatory functions, e.g., regulating Edc3 dimerization in Edc3–mediated RPS28B mRNA decay. Consistent with our model that Edc3 may exist as both monomeric and dimeric forms, the Ling et al. graph of molar mass distribution of full–length human Edc3 SV (Figure 1E of Ling et al., 2008) indicates two forms of human Edc3. One minor peak has a molar mass about 55,000 Da, and one major peak has a molar mass slightly under 100,000 Da. The minor peak appears to be an Edc3 monomer and the major peak appears to be an Edc3 dimer. Our response to Reviewer #2 sought to call attention to the possibility that human Edc3 has some unusual biochemical properties that may regulate its ability to dimerize. Reviewer #3 has provided valuable technical insights into the Ling et al. paper, but our data remain consistent with the existence of dimeric and monomeric states of both Edc3 and Scd6. We summarized our reasoning for different Edc3 states in the Discussion section (page 35, lines 799 to 810), and thus did not make any changes to the text of our manuscript that addressed this point further.

Finally, we have eliminated all words that were potentially dismissive of biochemical studies. We apologize for the apparent disrespect and assure you that was not our intent.

https://doi.org/10.7554/eLife.74410.sa2

Article and author information

Author details

  1. Feng He

    Department of Microbiology and Physiological Systems, University of Massachusetts Chan Medical School, Worcester, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Validation, Writing – original draft, Writing – review and editing
    For correspondence
    feng.he@umassmed.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-7523-2602
  2. Chan Wu

    Department of Microbiology and Physiological Systems, University of Massachusetts Chan Medical School, Worcester, United States
    Contribution
    Data curation, Formal analysis, Visualization, Writing – review and editing
    Competing interests
    No competing interests declared
  3. Allan Jacobson

    Department of Microbiology and Physiological Systems, University of Massachusetts Chan Medical School, Worcester, United States
    Contribution
    Conceptualization, Formal analysis, Funding acquisition, Project administration, Resources, Supervision, Writing – original draft, Writing – review and editing
    For correspondence
    allan.jacobson@umassmed.edu
    Competing interests
    is co-founder, director, and Scientific Advisory Board chair for PTC Therapeutics Inc
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-5661-3821

Funding

National Institute of General Medical Sciences (MIRA 1R35GM122468)

  • Allan Jacobson

The funders had no role in study design, data collection, and interpretation, or the decision to submit the work for publication.

Acknowledgements

This work was supported by a grant to AJ (1R35GM122468) from the U.S. National Institutes of Health. We thank Robin Ganesan and Kotchaphorn Mangkalaphiban for comments on the manuscript.

Senior Editor

  1. James L Manley, Columbia University, United States

Reviewing Editor

  1. Ruben L Gonzalez, Columbia University, United States

Publication history

  1. Preprint posted: October 1, 2021 (view preprint)
  2. Received: October 4, 2021
  3. Accepted: May 22, 2022
  4. Accepted Manuscript published: May 23, 2022 (version 1)
  5. Version of Record published: June 6, 2022 (version 2)

Copyright

© 2022, He et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 490
    Page views
  • 221
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Feng He
  2. Chan Wu
  3. Allan Jacobson
(2022)
Dcp2 C-terminal cis-binding elements control selective targeting of the decapping enzyme by forming distinct decapping complexes
eLife 11:e74410.
https://doi.org/10.7554/eLife.74410
  1. Further reading

Further reading

    1. Chromosomes and Gene Expression
    2. Structural Biology and Molecular Biophysics
    Lena Maria Muckenfuss, Anabel Carmen Migenda Herranz ... Martin Jinek
    Research Article Updated

    3′ end formation of most eukaryotic mRNAs is dependent on the assembly of a ~1.5 MDa multiprotein complex, that catalyzes the coupled reaction of pre-mRNA cleavage and polyadenylation. In mammals, the cleavage and polyadenylation specificity factor (CPSF) constitutes the core of the 3′ end processing machinery onto which the remaining factors, including cleavage stimulation factor (CstF) and poly(A) polymerase (PAP), assemble. These interactions are mediated by Fip1, a CPSF subunit characterized by high degree of intrinsic disorder. Here, we report two crystal structures revealing the interactions of human Fip1 (hFip1) with CPSF30 and CstF77. We demonstrate that CPSF contains two copies of hFip1, each binding to the zinc finger (ZF) domains 4 and 5 of CPSF30. Using polyadenylation assays we show that the two hFip1 copies are functionally redundant in recruiting one copy of PAP, thereby increasing the processivity of RNA polyadenylation. We further show that the interaction between hFip1 and CstF77 is mediated via a short motif in the N-terminal ‘acidic’ region of hFip1. In turn, CstF77 competitively inhibits CPSF-dependent PAP recruitment and 3′ polyadenylation. Taken together, these results provide a structural basis for the multivalent scaffolding and regulatory functions of hFip1 in 3′ end processing.

    1. Cell Biology
    2. Chromosomes and Gene Expression
    Brandt Warecki, Simon William Abraham Titen ... William Sullivan
    Research Article

    Wolbachia, a vertically transmitted endosymbiont infecting many insects, spreads rapidly through uninfected populations by a mechanism known as cytoplasmic incompatibility (CI). In CI, a paternally delivered modification of the sperm leads to chromatin defects and lethality during and after the first mitosis of embryonic development in multiple species. However, whether CI-induced defects in later stage embryos are a consequence of the first division errors or caused by independent defects remains unresolved. To address this question, we focused on ~1/3 of embryos from CI crosses in Drosophila simulans that develop apparently normally through the first and subsequent pre-blastoderm divisions before exhibiting mitotic errors during the mid-blastula transition and gastrulation. We performed single embryo PCR and whole genome sequencing to find a large percentage of these developed CI-derived embryos bypass the first division defect. Using fluorescence in situ hybridization, we find increased chromosome segregation errors in gastrulating CI-derived embryos that had avoided the first division defect. Thus, Wolbachia action in the sperm induces developmentally deferred defects that are not a consequence of the first division errors. Like the immediate defect, the delayed defect is rescued through crosses to infected females. These studies inform current models on the molecular and cellular basis of CI.