The PB components Pat1 and Dhh1 were each previously shown to be required for PB formation in yeast (Pilkington and Parker, 2008; Ramachandran et al., 2011; Mugler et al., 2016; Rao and Parker, 2017). To understand the relationship between Pat1 and Dhh1 in the regulation of PB assembly, we took advantage of the observation that overexpression of Pat1 leads to constitutive PB formation (Coller and Parker, 2005). This allowed us to characterize their respective regulatory contributions without external influences such as nutrient starvation signals. As expected, Pat1 overexpression from its endogenous locus using the galactose promoter in the presence of Dhh1 led to the formation of constitutive PBs, as visualized by co-localization of Dhh1-GFP and Dcp2-mCherry foci (Figure 1A). No foci were formed when Pat1 was not overexpressed or when we overexpressed the human influenza hemagglutinin (HA) tag alone (Figure 1—figure supplement 1A). Notably, when Pat1 was overexpressed in cells that lack Dhh1, a drastic reduction in the number of PBs was observed, as visualized by the bona fide PB marker Edc3-GFP and its co-localization with Dcp2-mCherry (Figure 1A,B). This demonstrates that formation of constitutive PBs upon Pat1 overexpression requires Dhh1.

Figure 1 with 1 supplement see all

Download asset Open asset

Pat1 is an 88 kDa protein that consists of an N-terminal un-structured domain (Pat1-N) followed by a proline-rich stretch plus middle domain and a C-terminal globular fold (Pat1-C). The N-terminal domain of Pat1 directly binds Dhh1’s RecA2 core via a conserved DETF motif (Sharif et al., 2013; Ozgur and Stoecklin, 2013). Furthermore, Pat1’s C-terminus harbors two serine residues that are phosphorylated by the cAMP-dependent protein kinase A (PKA) negatively regulating the interaction with Dhh1 and PB formation (Ramachandran et al., 2011).

In order to understand how these Pat1 domains influence LLPS and PB assembly, we wanted to characterize various Pat1 mutants both in vivo and in vitro. Unfortunately, we were unable to recombinantly express functional, full-length Pat1. Based on published crystal structures and in vitro interaction studies, we therefore developed a minimal Pat1 construct encompassing both N- and C-terminal domains that we refer to as Pat1-NC [amino acids: 5–79 (N) and 456–787 (C) connected with a (GGS)₄ linker] (Figure 1C). The N-terminal fragment of Pat1 (amino acids 5–79) was crystallized and shown to directly bind to Dhh1 (Sharif et al., 2013) whereas the C-terminus of Pat1 (amino acids 456–787) contains the PKA phosphorylation sites and mediates the interaction with the LSM-complex and RNA (Sharif et al., 2013; Fourati et al., 2014; Charenton et al., 2017; Ramachandran et al., 2011; Pilkington and Parker, 2008).

To investigate whether this PAT1-NC construct is functional in vivo and behaves similar to wild-type Pat1, we took advantage of the temperature-sensitive growth defect of a PAT1 deletion mutant (Fourati et al., 2014). Like full-length Pat1, expression of the PAT1-NC construct fully restored growth of pat1Δ cells at 37° C (Figure 1—figure supplement 1B and C). In addition, we also tested if the expression of PAT1-NC is able to rescue the synthetic lethality that was observed in pat1Δ edc3Δ scd6Δ cells (Fourati et al., 2014). The expression of PAT1-NC rescued the growth phenotype of pat1Δ edc3Δ scd6Δ cells at 37°C to a similar extent as achieved by wild-type PAT1 (Figure 1—figure supplement 1D).

Furthermore, cells expressing the PAT1-NC construct when starved of glucose induced PBs to a similar level as wild-type PAT1 (Figure 1D,E) indicating full functionality in vivo. The PBs in the PAT1-NC expressing cells rapidly dissolved upon re-addition of glucose demonstrating that they are reversible and dynamic rather than anomalous granules (Figure 1D). In addition, Pat1-NC-GFP itself localized to PBs upon stress in a reversible manner (Figure 1—figure supplement 1E). Finally, when we overexpressed Pat1-NC using a GAL promoter, robust PBs were induced in the absence of stress (Figure 1F,G). Taken together our data suggest that the PAT1-NC construct is fully functional in vivo.

We next examined whether Pat1-NC can interact with Dhh1 in vitro, and as expected, Pat1-NC displayed strong binding to full-length Dhh1 in a GST pull down assay (Figure 2A). However, this interaction was completely abolished for a Pat1-NC variant in which the N-terminal DETF motif was mutated to four alanines [Pat1^4A-Dhh1, see Supplementary file 2 table S2B for a list of all mutants used in this study] (Figure 2A).

Figure 2 with 1 supplement see all

Download asset Open asset

We next investigated the effect of the pat1^4A-Dhh1 mutant in the context of the full-length protein on PB formation in vivo. While cells expressing pat1^4A-Dhh1 showed a mild growth defect compared to the wild type PAT1, they grew significantly better than pat1Δ cells, indicating that this protein is at least partially functional (Figure 2—figure supplement 1A). To examine whether the interaction between Pat1 and Dhh1 is required for PB formation, we monitored PB assembly upon stress in the absence of glucose in wild-type (PAT1) or the pat1^4A-Dhh1 background. As expected, the number of PBs was drastically reduced in the complete absence of Pat1, as visualized by Dhh1-GFP (Figure 2—figure supplement 1B). Importantly, pat1^4A-Dhh1 expressing cells also demonstrated a drastic reduction in PB number compared to cells expressing PAT1 (Figure 2B,C) and the Pat1^4A-Dhh1-GFP protein itself showed a significant defect in localizing to PBs (Figure 2D,E).

To assure that the Pat1^4A-Dhh1 mutant is stably expressed, we checked its protein levels using Western blotting and observed that the mutant expresses to similar levels as wild-type Pat1 (Figure 2—figure supplement 1C). Moreover, in the Pat1^4A-Dhh1 mutant cells, the expression levels of Dhh1 were comparable to the Dhh1 levels in cells harboring a wild-type copy of Pat1 (Figure 2—figure supplement 1D). This demonstrates that the defect in PB formation in cells expressing the Pat1^4A-Dhh1 binding mutant is specifically due to impaired Pat1 function.

We were unable to overexpress the Pat1^4A-Dhh1 mutant in yeast. In order to investigate PB formation independently of carbon starvation stress, we therefore treated cells with the drug hippuristanol, which inhibits the eukaryotic translation initiation factor eIF4A (Bordeleau, M.E., et al., 2006). In consequence, hippuristanol prevents translation initiation and robustly induces PB formation within minutes in cells expressing wild-type PAT1 (Chan et al., 2018) [Figure 2F, Video 1]. Cells expressing pat1^4A-Dhh1 however, did not show PB formation until 2 hrs after hippuristanol addition (Figure 2F and Video 2).

Video 1

Download asset

Video 2

Download asset

It cannot be excluded that the Pat1^4A-Dhh1 mutant might interfere with the binding of Pat1 to some other factor with a role in PB formation in addition to Dhh1. We therefore also employed a complementary approach wherein we examined mutants on Dhh1 that impair interaction with Pat1. Two such Dhh1 mutants have been described: Dhh1 Mut3A (Dhh1^S292DN294D) and Dhh1 Mut3B (Dhh1^R295D) that partially or completely abolish binding to Pat1, respectively (Sharif et al., 2013). Cells expressing either wild-type Dhh1, Dhh1^S292DN294D, Dhh1^R295D or lacking Dhh1 altogether were starved of glucose and PB assembly was assessed. Whereas the Dhh1 variants expressed as well as wild-type Dhh1 (Figure 3—figure supplement 1), we observed a drastic reduction in the number of PBs in both Dhh1^S292DN294D, Dhh1 ^R295D cells (Figure 3A,B), and as published before also in dhh1Δ cells (Figure 3A,B, Mugler et al., 2016). The Dhh1^R295D had a more severe PB formation defect compared to the Dhh1^S292DN294D aligning with the in vitro binding abilities of these Dhh1 mutants to Pat1 (Sharif et al., 2013).

Figure 3 with 1 supplement see all

Download asset Open asset

To test PB formation independently of carbon starvation stress in these Pat1-binding mutants of Dhh1, we used the drug hippuristanol as before. Dhh1 wild-type cells when treated with hippuristanol formed PBs within minutes as judged by the co-localization of Dhh1-GFP and Dcp2-mCherry foci (Figure 3C,D and Video 3). However, both the Dhh1^S292DN294Dand Dhh1^R295D mutants did not form robust PBs until 3 hrs of hippuristanol treatment exhibiting a phenotype similar to dhh1Δ cells (Figure 3C,D and Videos 4, 5 and 6). Taken together, our data reveal that the Pat1-Dhh1 interaction is pivotal for robust PB assembly.

Video 3

Download asset

Video 4

Download asset

Video 5

Download asset

Video 6

Download asset

Pat1 is a target of the cAMP-dependent protein kinase A (PKA) and phosphorylation of two serines (amino acids 456/457) in the C-terminus of Pat1 under glucose-rich conditions was previously shown to negatively regulate PB formation (Ramachandran et al., 2011). In agreement with the published literature, cells expressing the pat1^EE mutant showed a defect in PB formation (Figure 4—figure supplement 1A,B and Ramachandran et al., 2011). To better understand how Pat1 phosphorylation controls PB assembly, we sought to characterize the influence of the phosphorylation status of Pat1 on PB formation in the absence of stress. Overexpression of PAT1^WT/SS (WT = wild type) and pat1^AA (non-phosphorylatable) led to constitutive PB formation as visualized by Dhh1-GFP and Dcp2-mCherry positive foci. However, upon overexpression of pat1^EE (phospho-mimetic), foci number was reduced (Figure 4A,B). The Pat1^EE-GFP mutant was expressed as well as wild-type Pat1-GFP and had no effect on Dhh1 expression (Figure 2—figure supplement 1C and D) emphasizing that the phospho-mimetic mutant specifically impairs Pat1 function.

Figure 4 with 1 supplement see all

Download asset Open asset

Pull-down experiments from cell extracts previously suggested that PKA-dependent phosphorylation of Pat1 diminishes its interaction with Dhh1 (Ramachandran et al., 2011). To test whether this effect is direct, we performed pull-down assays with the recombinant Pat1-NC protein and the Pat1-NC^EE variant. However, in this assay there was no significant difference in Dhh1 binding between Pat1-NC and Pat1-NC^EE (Figure 2A). Since the strong DETF binding site in Pat1-N might mask any weaker interaction in Pat1-C, we also tested for direct binding of Pat1-C to Dhh1 but were unable to detect any interaction (Figure 4C), indicating that the DETF-motif in Pat1-N provides the major interaction surface for Dhh1.

Pat1-C was previously shown to bind RNA (Pilkington and Parker, 2008; Chowdhury et al., 2014). We therefore tested whether RNA binding is regulated by PKA-dependent phoshphorylation by performing RNA oligo gel shift assays using recombinant proteins. Whereas wild-type Pat1-C robustly binds the RNA oligo and shifts it to higher molecular weight in a native PAGE gel, Pat1^EE displayed a significantly reduced RNA shift (quantified in Figure 4D). Taken together, our in vivo and in vitro results on the phospho-mimetic Pat1^EE mutant suggest that in addition to Pat1-Dhh1 binding, also the interaction between Pat1 and RNA is important for PB formation.

Our lab previously showed that Not1 promotes PB disassembly by activating the ATPase activity of Dhh1, and cells expressing a Not1 mutant that cannot bind to Dhh1 (not1^9X-Dhh1) form constitutive PBs in non-stressed cells. Yet these PBs are less intense and fewer in number than those formed upon glucose starvation (Mugler et al., 2016), indicating that additional mechanisms prevent PB formation in glucose-replete conditions.

Since in glucose-rich media PB assembly is also inhibited by the PKA-dependent phosphorylation of Pat1 (Ramachandran et al., 2011), we wanted to test whether expression of the non-phosphorylatable pat1^AA variant enhances PB formation in the not1^9X-Dhh1 strain background. While expression of pat1^AA alone was not sufficient to induce constitutive PBs as observed before (Ramachandran et al., 2011) and Figure 4—figure supplement 1A,B), expression of pat1^AA in the presence of not1^9X-Dhh1 significantly increased the number of constitutive PBs compared to cells expressing not1^9X-Dhh1 alone (Figure 5A and B). Yet, these PBs are still not as bright and numerous as in stressed wild-type cells. Interestingly, PB intensity strongly increased when these strains were further treated with hippuristanol, a drug that blocks initiation and liberates mRNA molecules from polysomes. The percentage of large PBs was highest in the pat1^AA + not19^9X-Dhh1 strain, followed by the not1^9X-Dhh1 and then pat1^AA (Figure 5A,B).

Figure 5

Download asset Open asset

Overall, this suggests that there are at least three determinants that act cooperatively to enhance the formation of PBs: (a) direct interactions between Pat1-Dhh1 and Pat1-RNA; (b) lack of Not1 binding to Dhh1 and in consequence low Dhh1 ATPase activity (Mugler et al., 2016); and (c) the availability of PB client mRNAs that are not engaged in translation. Furthermore, our results suggest that Pat1 and Not1 act antagonistically on PB dynamics. Whereas Pat1 binding to Dhh1 enhances the assembly of PBs, the interaction of Not1 with Dhh1 promotes PB disassembly, likely through activation of the ATPase activity of Dhh1.

At least two mechanisms of how Pat1 promotes PB formation via Dhh1 could be envisioned: first, Pat1 could slow the ATPase cycle of Dhh1, for example by preventing Not1 binding and/or inhibiting ATP hydrolysis, second, Pat1 could directly promote oligomerization and condensation of Dhh1 on RNA.

To test the first hypothesis, we used an in vitro ATPase assay using purified protein components in which Not1 robustly stimulates the ATPase activity of Dhh1 (Mugler et al., 2016). However, we did not observe a significant inhibition of the Not1-stimulated ATPase activity upon addition of Pat1-NC (Figure 6—figure supplement 1A), suggesting that Pat1 does not directly interfere with the mechanism of ATPase activation.

We had previously shown that PB dynamics are regulated by the ATPase activity of Dhh1 (Carroll et al., 2011; Mugler et al., 2016). To test Pat1’s effect on the Dhh1 ATPase cycle in vivo, we therefore also examined the turnover of PBs that form upon Pat1 overexpression. The drug cycloheximide (CHX) traps mRNAs on polysomes and thereby stops the supply of new RNA clients to PBs. Since active Dhh1 constantly releases mRNA molecules from PBs, a lack of mRNA influx eventually leads to their disassembly and in consequence the catalytic dead mutant of Dhh1 that is locked in the ATP- state (Dhh1^DQAD) inhibits PB disassembly and was shown to have negligible turnover rates (Mugler et al., 2016; Kroschwald et al., 2015). We observed that PBs formed upon Pat1 overexpression were more dynamic than PBs formed in the presence of Dhh1^DQAD and their disassembly kinetics was comparable to PBs formed upon stress in a wild-type DHH1 background (Figure 6—figure supplement 1B,C and Videos 7, 8 and 9). Thus, overall, our in vitro and in vivo findings are not consistent with the conclusion that Pat1 blocks the ATPase activation of Dhh1.

Video 7

Download asset

Video 8

Download asset

Video 9

Download asset

The second hypothesis is that Pat1 promotes higher-order mRNP assembly by directly or indirectly promoting Dhh1 oligomerization, thereby acting as a scaffold and providing additional protein–protein or protein-RNA interactions (Coller and Parker, 2005; Rao and Parker, 2017). We have previously shown that recombinant Dhh1 can undergo LLPS in the presence of RNA and ATP, and that these Dhh1 droplets can recapitulate aspects of in vivo PB dynamics. (Mugler et al., 2016). We therefore utilized this in vitro system to test Pat1’s impact on the phase separation behavior of Dhh1.

Dhh1 droplets were assembled from purified components as described previously (Mugler et al., 2016). Interestingly, while no LLPS of Pat1-NC alone was detected (Figure 6—figure supplement 2A), addition of increasing concentrations of wild-type Pat1-NC strongly enhanced the LLPS of Dhh1 in the presence of ATP and RNA as judged by an increase in the area * intensity of Dhh1 droplets (Figure 6A). Furthermore, Pat1-NC itself enriched in Dhh1 droplets, suggesting that Pat1 and Dhh1 co-oligomerize with RNA to form a composite phase-separated compartment (Figure 6A).

Figure 6 with 2 supplements see all

Download asset Open asset

In order to test if the enhancement in phase separation of Dhh1 via Pat1 is RNA-dependent, we perfomed this experiment also in the absence of RNA. Interestingly, we did not observe any Dhh1 droplets upon addition of increasing concentrations of Pat1-NC when RNA was omitted from the reaction (Figure 6—figure supplement 2B) suggesting that under these in vitro conditions RNA plays a critical role in the Pat1-promoted enhancement of Dhh1 oligomerization and higher-order phase separation. Consistent with this, the Pat1-N- terminus alone which cannot bind to RNA also does not enhance the phase separation of Dhh1 even in the presence of RNA and ATP (Figure 6—figure supplement 2C).

Overall, our in vitro data suggests that a combination of both the Pat1-N (binding to Dhh1) and Pat1-C (binding to RNA) is critically required to enhance the phase separation of Dhh1 and promote higher-order oligomeric mRNP structures akin to in vivo PBs.

To test the specificity of LLPS stimulation, we tested the Pat1^4A-Dhh1 (Dhh1 binding) and Pat1^EE (phospho-mimetic) mutants both of which drastically reduce PB formation in vivo (see Figures 2B and 4A,B and Figure 4—figure supplement 1A). Consistent with the in vivo results, Dhh1-RNA droplet formation was either not at all or only mildly stimulated upon addition of these Pat1 variants (Figure 6A).

We previously demonstrated that the MIF4G domain of Not1 prevents formation of Dhh1-RNA droplets, presumably by activating the ATPase cycle of Dhh1 (Mugler et al., 2016). We therefore next analyzed whether the Pat1-Dhh1 droplets are still responsive to Not1. Similar to what we observed in vivo (Figure 5), Not1^MIF4G also diminishes formation Dhh1 droplets in the presence of Pat1 in vitro (Figure 6B).

Taken together our results suggest that Pat1 enhances the multivalency of protein–protein and protein-RNA interactions in a Dhh1-Pat1 mRNP and in consequence promotes the formation of higher-order, liquid-like mRNP droplets akin to in vivo PBs.

To examine how well our in vitro droplets resemble in vivo PBs we investigated the stoichiometric ratio of Dhh1:Pat1 in these granules. In order to determine the stoichiometry in vivo, we measured the number of PBs and the fluorescence intensity of GFP-tagged Dhh1 and Pat1 in these foci. Briefly, we glucose starved both Pat1-GFP and Dhh1-GFP expressing cells to induce PBs (Figure 7—figure supplement 1A) and used the single particle tracking software Diatrack to count the number and intensity of PBs in an unbiased and automated manner in each strain (Vallotton et al., 2017). Surprisingly, despite the fact that the cellular concentration of Pat1 is ten-times lower than Dhh1 (Figure 7—figure supplement 1B), the ratio of the two PB components was approximately 2:1 ± 0.18 (Dhh1:Pat1) after 0.5–4 hr of starvation (Figure 7A). Extended starvation resulted in enhanced Dhh1 recruitment to PBs until reaching a Dhh1:Pat1 ratio of 2.5:1 ± 0.23, suggesting that PB composition matures over time (Figure 7A).

Figure 7 with 2 supplements see all

Download asset Open asset

In order to determine the stoichiometry of Pat1 and Dhh1 droplets in vitro, we imaged Dhh1-mCherry and Pat1-GFP droplets with a confocal microscope (Figure 7—figure supplement 1C). The protein concentration of both Dhh1 and Pat1 in the droplet was calculated from a standard curve determined from values measured for different concentrations of the respective soluble fluorophores (Figure 7B). Remarkably, we found a stoichiometric ratio of 2.7:1 ± 0.13 of Dhh1 to Pat1 in the in vitro phase-separated droplets, closely resembling the Dhh1:Pat1 ratio of mature PBs in vivo (Figure 7B).

Thus, while other mRNA decay factors also contribute to PB formation in vivo, our data demonstrate that with a minimal number of constituents, namely Dhh1, RNA, ATP and Pat1 higher-order dynamic liquid droplets can be formed in vitro. The in vitro droplets recapitulate various properties of PBs formed in vivo such as (a) the stoichiometry of PB components, and (b) the dynamics of Pat1-stimulated assembly and Not1-stimulated disassembly of in vivo PBs identifying Pat1 as an important player in PB formation promoting LLPS via Dhh1.

The ability of proteins and RNA to undergo LLPS has emerged as an essential fundamental biological process allowing cells to organize and concentrate their cellular components without the use of membranes. Yet, how these membraneless organelles assemble, are kept dynamic and are regulated remains enigmatic. One such cellular phase-separated compartment is the PB and our results presented here together with our prior work (Mugler et al., 2016) identify and characterize both positive and negative regulators of PB assembly and provides novel insight into the regulation of membraneless organelle formation.

Overall, our work uncovers a function of Pat1 as an enhancer of PB formation that acts via Dhh1 and counteracts the inhibitory function of Not1. Based on our results, we propose that there are at least three inputs that cooperatively regulate PB dynamics in vivo (Figure 7C). First, the availability of mRNA clients acting as seeding substrates to initiate PB formation. mRNA availability is inversely proportional to the translation status, which is regulated by different stress responses and the metabolic state of the cell (Figure 2C). Second, the activity of cAMP-dependent PKA negatively regulates PB assembly in nutrient-rich conditions, at least in part through phosphorylation of Pat1. However, since expression of non-phosphorylatable Pat1^AA at endogenous levels (in contrast to overexpression) is not sufficient to induce PBs, Pat1 availability appears to be limiting (Figures 5 and 4A, and Figure 4—figure supplement 1A and Ramachandran et al., 2011). Third, activation of the Dhh1 ATPase cycle by the CCR4-Not1 deadenylation complex stimulates mRNA release and negatively regulates PB assembly. It is of note that Not1 itself might also be subject to post-translational regulation, in response to metabolic state and/or cell cycle stages of the cell (Mugler et al., 2016; Braun et al., 2014).

In this model, a tug-of-war between Pat1 and Not1 regulates PB assembly and disassembly: whereas a tight Pat1-Dhh1 interaction in the presence of RNA clients, promotes PB formation, Not1, induces PB turnover via stimulating the ATPase activity of the DEAD box ATPase Dhh1 (Figure 7C). Remarkably, our in vivo findings are corroborated in vitro wherein, Pat1 enhances the LLPS of Dhh1 and RNA, a step critical for the assembly of large mRNP granules, and by contrast, Not1 reverses the LLPS of Pat1-Dhh1-RNA droplets (Figures 6 and 7C). Furthermore, the observed stoichiometry of Dhh1 and Pat1 in PBs is recapitulated in our in vitro phase separation assay (Figure 7A,B).

Inducing PB formation by Pat1 overexpression or hippuristanol treatment allowed us to dissect distinct mechanisms that govern PB assembly and characterize the crucial role of the Pat1-Dhh1 interaction in this process. The benefit of these modes of PB formation are that they bypass the characteristic stresses associated with PB formation, such as nutrient starvation, that might have more widespread and confounding effects on translation, mRNA degradation, or on the regulation of other PB components.

Our experiments reveal that Pat1 functions in PB formation through Dhh1 since PB induction by Pat1 overexpression is strictly dependent on the presence of Dhh1 (Figure 1A). Furthermore, a mutant in the DETF motif in the N-terminus of Pat1 mediating direct binding to Dhh1 (Pat1^4A-Dhh1 variant) and mutants of Dhh1 that abolish Pat1 binding fail to promote PB assembly (Figure 2). These results were further confirmed using the drug hippuristanol that inhibits translation initiation wherein PBs were induced in wild-type cells but not in cells expressing the Pat1^4A-Dhh1 mutant, and the Dhh1 mutants that abolish Pat1 binding (Dhh1^R295D and Dhh1 ^S292DN294D) [Figures 2F and 3C and Chan et al., 2018]. This suggests that high mRNA load alone is inadequate for PB formation and that a direct physical interaction between Pat1 and Dhh1 is obligatory as well. Since the DETF motif in Pat1 is conserved across evolution and was suggested to regulate PB assembly in human cells as well (Sharif et al., 2013; Ozgur and Stoecklin, 2013), the Pat1-Dhh1 interaction likely plays a critical role in PB formation across species.

Our results seem to be at odds with a previous publication, which suggests that Pat1 acts independently of Dhh1 to promote PB formation (Coller and Parker, 2005). However, in this study Pat1 was overexpressed from a plasmid, in the presence of the endogenous copy. When recapitulating these conditions, we observed that PB components mislocalize to the nucleus [observed by DAPI and a nuclear rim marker, see Figure 7—figure supplement 2A,B] and it is possible that in the absence of additional organelle markers, this nuclear accumulation was misinterpreted as PB formation.

How does Pat1 enhance PB formation? One hypothesis is that Pat1 directly counteracts the Not1-stimulated ATPase activation of Dhh1, but we do not favor such a model since i) Pat1 does not inhibit Dhh1's ATPase activity at physiological protein ratios in vitro (Figure 6—figure supplement 1A), ii) Not1^MIF4G dissolves Pat1-Dhh1 condensates, while droplets formed from catalytic-dead Dhh1^DQAD are Not1-resistant (Figure 6 and Mugler et al., 2016), and iii) in vivo, PBs formed upon Pat1 overexpression are much more dynamic than those formed from Dhh1^DQAD (Figure 7—figure supplement 1B).

Instead, our data are consistent with a model in which Pat1 promotes PB assembly by directly enhancing the LLPS of Dhh1. DDX6, the human homolog of Dhh1 can oligomerize (Ernoult-Lange et al., 2012), and given the fact that both DDX6 and Dhh1 are found in molar excess over mRNA, we propose that Pat1 functions as a scaffold, providing additional multivalent interactions with Dhh1 and RNA, thereby aiding in the fusion of Dhh1 LLPS droplets and promoting assembly of microscopically-detectable mRNP condensates (Figure 6). Interestingly, both functional binding to Dhh1 in Pat1-N and RNA-binding in Pat1-C are required for this mechanism as individually mutations in each of the domains alter PB assembly in vivo and fail to enhance Dhh1’s phase separation in vitro. Rather, Pat1-N alone diminishes the extent of LLPS (Figure 6—figure supplement 2C). It is interesting to note that Pat1-N was previously shown to interfere with the ability of Dhh1 to interact with RNA (Sharif et al., 2013) and we thus speculate that Pat1-N might outcompete RNA from Dhh1, and in the absence of the C-terminal Pat1 RNA-binding site destabilize droplets. This might explain why RNA-binding by Pat1’s C-terminus that is under the regulation of PKA (Figure 4A,B, Figure 4—figure supplement 1A) is critical for LLPS and PB assembly. In vivo, we find that the Pat1^4A-Dhh1 (Dhh1 binding mutant) and Pat1^EE (phospho-mimetic mutant) not only have a drastically reduced number of PBs, but also much lower intensity (reduced by 7 and 8 fold respectively, Figure 7—figure supplement 2C) compared to Pat1 wild-type PBs.

Our recent findings also show that the un-structured poly-Q rich C-terminal tail of Dhh1 is required for PB formation in vivo and Dhh1 LLPS in vitro (Hondele et al., submitted). Thus, Pat1 might potentially enhance Dhh1 multimerization via the low complexity tails of Dhh1. Structural studies will now be needed to clarify how Pat1 facilitates oligomerization and PB formation of the Dhh1-Pat1- RNA complex.

As non-membrane bound organelles, it is of interest to understand how cells assemble large RNP granules and regulate their dynamics. PBs are prominent membraneless cellular compartments, which have been shown to be not only involved in numerous aspects of mRNA turnover, but also crucial for survival under various cellular stresses. With an in vitro phase separation tool in hand that faithfully reconstitutes certain aspects of PB formation, we are now in a position to dissect how PBs form and characterize their biophysical behavior. This will aid us not only in elucidating key components of PB formation, but also help us elucidate the function of PBs in regulating mRNA turnover.

At this stage, our in vitro systems are limited to Dhh1, Pat1, Not1 and RNA and for the sake of simplicity; we present a model only depicting the role of Pat1 and Dhh1 in PB formation (Figure 7C). However, it is important to note that several additional PB components have been characterized. For example, this includes Ecd3, members of the Lsm1-7 complex, or Dcp2 that play crucial roles either alone or together with Pat1 and Dhh1 in PB assembly. Both Edc3 and Lsm4 contain low complexity-domains that were shown to be critical for PB formation in vivo, and it was recently demonstrated a threshold concentration of low-complexity domain containing proteins is required for robust PB assembly (Protter et al., 2018). In addition, it was previously shown that both Pat1 (C-terminus) and the Lsm1-7 complex individually have low RNA binding ability but a reconstituted Lsm1-7-Pat1-C complex displays an enhanced ability to bind RNA (Chowdhury et al., 2014). This might suggest that in vivo the Pat1-C- terminus functions together with the Lsm1-7 complex to bind RNA and enhance PB formation. Furthermore, it cannot be excluded that direct binding of Dcp2 to the C-terminus of Pat1 also regulates PB assembly and RNA turnover (Fourati et al., 2014). It will thus be critical to examine additional PB components in future experiments and to evaluate how they function together with Pat1 and Dhh1 in order to understand the complex interplay of various inputs that regulate PB assembly and dynamics. Nonetheless, based on our data and the aforementioned literature evidence, we propose that Pat1 provides a central PB scaffold and participates in an intricate network of protein–protein and protein-RNA interactions that promote PB assembly and influence RNA turnover rates in vivo.

The physical basis of LLPS has attracted a great deal of attention recently, at least in part because of the critical role that proper mRNP assembly plays in pathological neurodegenerative diseases and stress responses (Ramaswami et al., 2013; Alberti and Hyman, 2016; Patel et al., 2015; Aguzzi and Altmeyer, 2016; Hyman et al., 2014). An important challenge is to understand how these granules are kept dynamic and functional under some conditions whereas the formation of solidified and aberrant aggregates are triggered under others. Robust in vitro reconstitution systems that can recapitulate mRNP granule assembly, dynamics and stoichiometry will be invaluable in addressing these research questions and increase our understanding of the molecular mechanism of how cells age and deal with stress.

S. cerevisiae strains used in this study are derivatives of W303 and are described in Supplementary file 1. ORF deletion strains and C-terminal epitope tagging of ORFs was done by PCR-based homologous recombination, as previously described (Longtine et al., 1998). Plasmids for this study are described in Supplementary file 2—table S2A. Mutations in Pat1 were generated by introducing the mutation in the primer used to amplify the respective Pat1 regions and stitched together with the selection marker from the plasmid (Supplementary file 2—table S2B). Mutations in Dhh1 and Not1 were generated as in Mugler et al. (2016). Primer sequences for strain construction are listed in table S2C.

The samples were grown overnight in synthetic media containing 2% raffinose, diluted to OD₆₀₀ = 0.05 or 0.1 the following day, and grown to mid-log phase (OD₆₀₀ = 0.3–0.4). The culture was split into two and to one-half galactose was added to 2% final concentration and the corresponding protein induced for 2–3 hr. The cells in both raffinose and galactose were imaged using a wide-field fluorescence microscope.

PBs were induced via glucose starvation stress. Samples were grown overnight in synthetic media containing 2% glucose, diluted to OD₆₀₀ = 0.05 or 0.1 the following day, and grown to mid-log phase (OD₆₀₀ = 0.3–0.8). Cells were harvested by centrifugation and washed in ¼ volume of fresh synthetic media ± 2% glucose, then harvested again and re-suspended in 1 vol of fresh synthetic media ± 2% glucose and grown 30 min at 30°C. Cells were then transferred onto Concanavalin A-treated MatTek dishes (MatTek Corp., Ashland, MA) and visualized at room temperature. For cycloheximide experiments, final concentration of 50 ug/mL (Sigma-Aldrich, CH) were used to examine PB disassembly kinetics. For PB induction and to determine the kinetics of PB assembly we used hippuristanol (a generous gift of Junichi Tanaka, University of the Ryukyus) at a final concentration of 10 µM.

Cells were transferred onto Concanavalin A-treated MatTek dishes (MatTek Corp., Ashland, MA) and visualized at room temperature Microscopy was performed using an inverted epi-fluorescence microscope (Nikon Ti) equipped with a Spectra X LED light source and a Hamamatsu Flash 4.0 sCMOS camera using a 100x Plan-Apo objective NA 1.4 and the NIS Elements software. Representative images were processed using ImageJ software. Brightness and contrast were adjusted to the same values for images belonging to the same experiment and were chosen to cover the whole range of signal intensities. Image processing for PB analysis was performed using Diatrack 3.05 particle tracking software (Vallotton and Olivier, 2013) as described below.

In order to quantify PB formation in live cells, we used an automated image analysis in a manner similar to Mugler et al. (2016). First, PBs were counted using Diatrack 3.05 particle tracking software using local intensity maxima detection, followed by particle selection by intensity thresholding and particle selection by contrast thresholding with a value of 5% (Vallotton et al., 2017). To speed up the analysis, we renamed all our images in a form that can be recognized as a time-lapse sequence by Diatrack, and placed them all in a single directory, such that they all will be analyzed using exactly the same image analysis parameters. Renaming and copying was done by a custom script (Source Code 1), which also performed cell segmentation using a method adapted from (Hadjidemetriou et al., 2008). Briefly, the method first detects all edges using a Laplacian edge detection step, and then traces normals to those edges in a systematic manner. These normals tend to meet at the cell centre where the high density of normals is detected, serving as seeds to reconstruct genuine cells. Our script thus counts cells and reports their number for each image - information which is output to an excel table. The results from Diatrack PB counting are imported from a text file into that table, and the number of PB is divided by the number of cells for each image.

Dhh1 and Not1 were purified as described previously (Mugler et al., 2016). V5-Pat1-C, GST-Pat1-N and GST-Pat1-NC constructs as described in Supplementary file 2—table S2A were cloned into pETM-CN vectors. These expression vectors were transformed into chemically competent E. coli BL21 DE3 under the selection of ampicillin and chloramphenicol. Pre-cultures were grown in LB at 37°C over-night, and diluted 1:100 into rich medium the next morning. Cells were grown at 37°C to an OD600 of 0.6 and induced with 200 mM IPTG (final concentration). Cells were then grown overnight at 18°C, harvested and resuspended in 30 mL lysis buffer (500 mM (Pat1-C, Pat1-N) or 300 mM NaCl (Pat1-NC), 25 mM Tris-HCl pH 7.5, 10 mM imidazole, protease inhibitors, 10% glycerol) per cell pellet from 2 L of culture. After cell lysis by EmulsiFlex (Avestin Inc, Ottawa, CA), the 6xHis tagged proteins were affinity extracted with Ni²⁺ sepharose in small columns, dialyzed into storage buffer (MH200G (Pat1-C, Pat1-N): 200 mM NaCl, 25 mM Tris pH 7.5 (RT), 10% glycerol, 2 mM DTT), MH300G (Pat1-NC): same with 300 mM NaCl) with simultaneous protease cleavage of the His-tag and GST-tag (unless required for pull-down assays) and further purified by size exclusion with a Superdex 200 column on an AEKTA purifier (both GE Life Sciences, Marlborough, MA) in storage buffer. Protein expression levels, His eluates and gel filtration fractions were analyzed by SDS-PAGE. Clean Superdex elution fractions were pooled, concentrated using Millipore Amicon Centrifugation units and snap frozen as ~20 µl aliquots in siliconized tubes in liquid nitrogen.

ATPase assays were performed according as described (Montpetit et al., 2011) with the following modifications: final concentration 2 µM Dhh1-mCherry was mixed with 0.5 µM Not1 and Pat1-mCherry (1 and 3 µM) as indicated and protein volumes equalized with storage buffer (MH200G). 2 µL 10x ATPase buffer (300 mM HEPES-KOH pH 7.5, 1 M NaCl, 20 mM MgCl₂), 4 µL 10 mg/ml polyU (unless indicated otherwise), RNase inhibitors, 13.3 µL 60% glycerol, 2.7 µL 10 mg/mL BSA, were added to a final volume of 36 µL. Reactions were set up in triplicate in a 96-well NUNC plate. The assay was initiated by the addition of 40 µL of a master mix containing 1x ATPase buffer, 2.5 mM ATP (from a 100 mM stock in 0.5 M HEPES-KOH pH 7.5), 1 mM DTT, 6 mM phosphoenolpyruvate, 1.2 mM NADH (from a 12 mM stock in 25 mM HEPES-KOH pH 7.5) and 125–250 units/mL PK/LDH. NADH absorption was monitored with a CLARIOstar plate reader (BMG Labtech, Ortenberg, Germany) at 340 nm in 30 s intervals for 400 cycles.

Reactions were pipetted in 384-well microscopy plates (Brooks 384 well ClearBottom Matriplate, low glass). Proteins were diluted to 50 µM (Dhh1) or 100 µM (Not1, Pat1) stocks with storage buffer and mixed as droplets at the side of the well; volumes were equalized to 5 µl with storage buffer. Next, a master mix of 12.5 µl 150 mM KCl buffer (150 mM KCl, 30 mM HEPES-KOH pH 7.4, 2 mM MgCl₂), 2 µl ATP reconstitution mix (40 mM ATP, 40 mM MgCl₂, 200 mM creatine phosphate, 70 U/mL creatine kinase), 2.5 µl 2 mg/ml polyU (in H₂O) as an RNA analog, 1 µl 1M HEPES-KOH pH 6.4 buffer and 2 ml 10 mg/ml BSA were added and mixed by pipetting, the droplets still at the side of the well, since this later on produced more equal droplet distribution. First imaging was performed after 20 min incubation at room temperature; just prior to imaging, droplets were spun down at 100 g for 1 min. For subsequent analysis, plates were stored in the fridge. Pictures displayed were recorded after 1 hr incubation.

In order to estimate the relative amount of Dhh1 to that of Pat1 in PBs in vivo, we labelled both proteins with eYGFP in two separate strains, and imaged them in exactly the same conditions after inducing carbon (glucose) starvation stress for indicated times (Figure 6A). Using exactly the same endogenous probe for both proteins allowed us to deduce ratios of abundances directly from ratios of intensities (no spectral corrections are necessary). We then measured the intensity of the 10 brightest PBs as a function of time and in both cases took the median intensity value as representative of the more visible PBs (at least 10 PBs were present in every image for both strains and at each time point). Finally, we divided the median value obtained for Dhh1 by that for Pat1 and plotted that ratio as a function of time post-stress.

For the analysis of the in vitro phase transition of Dhh1 in the presence of Pat1, 2 µM mCherry-tagged Dhh1 and 4 µM GFP-tagged Pat1-NC were added to 0.1 mg/ml polyU and 5 mM ATP in 150 mM KCl, 30 mM HEPES-KOH, pH 7.4 and 2 mM MgCl₂. After 1 hr incubation, the phase-separated droplets were imaged in both the mCherry and the GFP channel by confocal fluorescence microscopy (Leica TCS SP8) with a 63x NA 1.4 oil objective (Leica). The protein concentration was evaluated from the fluorescence intensity via a standard curve obtained by measuring a series of samples at different known concentrations for each fluorophore. The size distributions were obtained by analyzing the images obtained by fluorescence microscopy with an in-house code written in Matlab.

25 µl of 100 µM recombinant purified GST Pat1-NC was bound to 30 µl GST-slurry (Sigma), and one aliquot of GST-beads was included without added Pat1-NC as negative control. Reactions were incubated for 30mins rotating at 4°C and washed 4-times with wash buffer (300 mM NaCl, 10 mM Tris pH 7.5, 0.05 % NP40). 4 µl 250 µM Dhh1 was added in 500 µl wash buffer and incubated rotating at 4°C for 1 hr. Reactions were washed 5-times, proteins eluted by boiling with 4x SDS loading buffer and separated on a 12% acrylamide gel and stained with Coomassie. For V5-pulldown assays, the bait (15 µl 100 µM Pat1-C construct) was bound to 15 ml V5-slurry (Sigma), the bait 3 µl 200 µM Dhh1-mCherry, and incubation/wash buffers as above with salt concentration as indicated in the blot. Final washes were only performed three times.

Proteins were diluted to 20 µM. 0, 2 or 4 µl of protein solution was filled to 4 µl with storage buffer (200 mM NaCl, 25 mM Tris pH 7.5, 10% glycerol, 2 mM DTT). Proteins were mixed with a master mix containing per reaction 2 µl 10x ATPase buffer (300 mM HEPES-KOH pH 7.5, 1 M NaCl, 20 mM MgCl₂), 0.3 µl Cy5-labeled in vitro transcribed RNA oligo (29nt, 1.1 µg/µl), 2 µl 2.5M NaCl, 13.7 µl H₂O and 3 µl 60% glycerol. Reactions were incubated on ice for 20 min and separated on a precast Bis-Tris Native gradient Gel (Invitrogen) in 1x running buffer (50 mM Bis-Tris 50 mM Tricine pH 6.8) at 100V 4°C for about 2.5 hr.

Overnight yeast cultures were grown in permissive conditions in SCD media. Growth curves were acquired using CLARIOstar automated plate reader (BMG Labtech, Ortenberg, Germany) at 30°C in 24-well plastic dishes (Thermo Fisher) from overnight pre-cultures diluted 1:125 in the specified synthetic liquid media. For spotting assays on plates the indicated strains were grown overnight at 25°C to saturation. Next day equal number of ODs for each strain was spotted (with a 1:10 series dilution) on YPD plates and incubated at 25°C, 30°C and 37°C for 2–3 overnights.

For Western blot analysis, roughly 5 OD₆₀₀ units of log growing cells were harvested and treated with 0.1M NaOH for 15 mins at room temperature. The samples were vortexed in between and finally centrifuged to remove the NaOH. SDS sample buffer was added and homogenates were boiled. Proteins were resolved by 4–12% Bolt Bis-Tris SDS PAGE (Thermo Fischer, Waltham, MA), then transferred to nitrocellulose membrane (GE Life Sciences, Marlborough, MA). Membranes were blocked in PBS with 5% non-fat milk, followed by incubation with primary antibody overnight. Membranes were washed four times with PBS with 0.1% Tween-20 (PBS-T) and incubated with secondary antibody for 45 min. Membranes were imaged and protein bands quantified using an infrared imaging system (Odyssey; LI-COR Biosciences, Lincoln, NE). The following primary antibodies were used for detection of tagged proteins at the indicated dilutions: rabbit-anti-Dhh1 (1:5000) as described in (Fischer and Weis, 2002), (Weis Lab ETH Zurich Cat# Weis_001, RRID:AB_2629458), mouse-anti-GFP (1:1000) (Roche Cat# 11814460001, RRID:AB_390913), and rabbit-anti-Hxk1 (1:3000) (US Biological Cat# H2035-01, RRID:AB_2629457, Salem, MA). IRdye 680RD goat-anti-rabbit (LI-COR Biosciences Cat# 926–68071, RRID:AB_10956166) and IRdye 800 donkey-anti-mouse (LI-COR Biosciences Cat# 926–32212, RRID:AB_621847) were used as secondary antibodies.

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

We would like to thank Justine Kusch and the ETHZ ScopeM facility for technical assistance with confocal microscopy, Elisa Dultz, Carmen Weber, Stephanie Heinrich, Mostafa Zedan, Sarah Khawaja Evgeny Onishchenko, Annemarie Kralt and Jan Nagler for their critical reading of this manuscript and members of the Weis lab for helpful discussions and comments. We would like to thank Bram Hofland for assisting in cloning of yeast constructs. We would also like to thank the Séraphin lab for sharing their yeast strains. MH was supported by a Human Frontier Science Program (HFSP) postdoctoral fellowship (LT000914/2015) and an ETH postdoctoral fellowship (FEL-37-14-2). This work was supported by NIH/NIGMS (R01GM058065 and R01GM101257 to KW) and the Swiss National Science Foundation (SNF 31003A_159731 and 31003A_179275 to KW).

© 2019, Sachdev 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.