1. Biochemistry and Chemical Biology
  2. Chromosomes and Gene Expression
Download icon

Specialization of the chromatin remodeler RSC to mobilize partially-unwrapped nucleosomes

  1. Alisha Schlichter
  2. Margaret M Kasten
  3. Timothy J Parnell
  4. Bradley R Cairns  Is a corresponding author
  1. Howard Hughes Medical Institute (HHMI), Department of Oncological Sciences, Huntsman Cancer Institute, University of Utah School of Medicine, United States
Research Article
  • Cited 0
  • Views 643
  • Annotations
Cite this article as: eLife 2020;9:e58130 doi: 10.7554/eLife.58130

Abstract

SWI/SNF-family chromatin remodeling complexes, such as S. cerevisiae RSC, slide and eject nucleosomes to regulate transcription. Within nucleosomes, stiff DNA sequences confer spontaneous partial unwrapping, prompting whether and how SWI/SNF-family remodelers are specialized to remodel partially-unwrapped nucleosomes. RSC1 and RSC2 are orthologs of mammalian PBRM1 (polybromo) which define two separate RSC sub-complexes. Remarkably, in vitro the Rsc1-containing complex remodels partially-unwrapped nucleosomes much better than does the Rsc2-containing complex. Moreover, a rsc1Δ mutation, but not rsc2Δ, is lethal with histone mutations that confer partial unwrapping. Rsc1/2 isoforms both cooperate with the DNA-binding proteins Rsc3/30 and the HMG protein, Hmo1, to remodel partially-unwrapped nucleosomes, but show differential reliance on these factors. Notably, genetic impairment of these factors strongly reduces the expression of genes with wide nucleosome-deficient regions (e.g., ribosomal protein genes), known to harbor partially-unwrapped nucleosomes. Taken together, Rsc1/2 isoforms are specialized through composition and interactions to manage and remodel partially-unwrapped nucleosomes.

Introduction

Nucleosomes regulate transcription in diverse ways and can either block or attract transcriptional regulators (Workman and Kingston, 1998; Iyer, 2012). At promoters, nucleosome positioning and/or occupancy plays a central role in regulating transcription factor binding, with transitions in nucleosome positioning typically accompanying activation. SWI/SNF-family ATP-dependent chromatin remodeling complexes (CRCs) have evolved to conduct nucleosome sliding and ejection, and enable transcription factor access to DNA. These CRCs are complex in both composition and mechanism; they utilize a catalytic ATPase to translocate DNA around nucleosomes to conduct nucleosome sliding and eviction, and contain an additional set of proteins to help target and regulate each complex (Clapier and Cairns, 2009; Lorch and Kornberg, 2017; Narlikar et al., 2013).

The SWI/SNF-family remodeler, RSC (Remodels the Structure of Chromatin), from the budding yeast S. cerevisiae, is both essential and abundant, and has long served as a prototype CRC. RSC complex (like others in the SWI/SNF family) is found in more than one compositional subtype, and contains either Rsc1 or its highly-related paralog, Rsc2 (Cairns et al., 1999). Rsc1 and Rsc2 are orthologs of the mammalian polybromo, as both contain multiple bromodomains, a bromodomain-adjacent homology (BAH) domain, and a DNA binding motif (AT Hook or HMG box). Additional RSC compositional variation has been suggested, involving the association of two additional paralogous RSC subunits, Rsc3 and Rsc30 (Campsteijn et al., 2007; Chambers et al., 2012), which are zinc cluster DNA-binding proteins with affinity for GC-rich sequences (Badis et al., 2008). RSC1 and RSC2 are redundant for viability (rsc1Δ rsc2Δ mutants are inviable), however loss of only one confers mild but dissimilar phenotypes, suggesting overlapping essential functions alongside limited unique functions (Baetz et al., 2004; Chambers et al., 2012; Cairns et al., 1999; Bungard et al., 2004). To gain further understanding regarding the roles of RSC1 and RSC2 complexes, we characterized Rsc1- and Rsc2-containing complexes through purification and in vitro biochemical assays, alongside in vivo genetic and genomic characterization. Both approaches converged to reveal an interplay of functional roles for Rsc1/2, Rsc3/30 and a partner high-mobility group (HMG) domain protein, Hmo1, in managing partially-unwrapped nucleosomes.

Partially-unwrapped nucleosomes are defined here as those in which the DNA has released (or displays a tendency to release) from the histone octamer at one or both of the symmetric locations where DNA enters/exits the nucleosome, while the central DNA gyre maintains association with the octamer. DNA sequences differ in their affinity for histone octamers and propensity to form nucleosomes (Anderson et al., 2002); stiff homopolymer AT tracts deter nucleosome formation (Segal and Widom, 2009) whereas those with short (5 bp) alternating AT and GC tracts more easily adopt nucleosomal curvature, providing a lower cost in energy for nucleosome formation. The commonly used ‘601’ nucleosome positioning sequence is synthetic, was selected for high affinity (Lowary and Widom, 1998), and displays a 5 bp AT/GC alternating pattern. In contrast, the 5S rRNA gene sequence is naturally occurring and of lesser affinity (comparable to genome averages) in comparison to 601 (Polach and Widom, 1995; Dong et al., 1990; Li and Widom, 2004; Zhou et al., 2019; Mauney et al., 2018). Here, there is some debate whether the entry/exit DNA ends of the 5S positioning sequence displays higher or lower rates of detachment from the octamer than the 601 sequence (North et al., 2012; Chen et al., 2014; Zhou et al., 2019). However, recent work using small-angle X-ray scattering (SAXS) with salt titration to compare the unwrapping dynamics of the 5S and 601 nucleosomes demonstrates that the 5S nucleosome unwraps more rapidly and at lower salt concentrations than the 601 nucleosome (Mauney et al., 2018; Chen et al., 2014).

Prior work has revealed the presence of partially-unwrapped ‘fragile’ nucleosomes at promoters with an especially wide nucleosome deficient region (NDR), and co-incidence of RSC, including at the majority of ribosomal protein genes (RPGs) (Kubik et al., 2015; Brahma and Henikoff, 2019; Knight et al., 2014). Indeed, in silico size fractionation of RSC-bound nucleosomes provides evidence that RSC occupies wrapped nucleosomes at the +1 and −1 promoter positions, as well as a partially-unwrapped nucleosome interposed between those two displaying lower/fractional occupancy – the basis for the appearance of a large NDR (Brahma and Henikoff, 2019). As partially-unwrapped nucleosomes are likely conformationally diverse (Bilokapic et al., 2018), RSC may have evolved to both recognize and remodel these nucleosomes. Prior work with the Rsc2-containing form of RSC revealed altered remodeling outcomes for partially-unwrapped (H3 R40A) nucleosomes (Somers and Owen-Hughes, 2009), further questioning whether an alternative form of RSC might better manage them. RSC mobilization of partially-unwrapped nucleosomes may allow sets of transcription factors regulated access to a section of DNA without keeping the region constitutively nucleosome-free – which may provide regulatory benefits and help maintain genome stability (see Discussion). Here, we provide multiple lines of evidence that a set of proteins in RSC (Rsc1, Rsc3/30), and interacting with RSC (Hmo1), cooperate to help remodel partially-unwrapped nucleosomes in vitro and in vivo.

Results

The Rsc3/30 heterodimer preferentially associates with the RSC1 complex via the CT2 domain

Rsc1 and Rsc2 are highly similar proteins (45% identical, 62% similar), with high homology present in the bromodomains, BAH domain, and the CT1 region – whereas the CT2 region, which is required for Rsc1 or Rsc2 assembly into RSC, is considerably more divergent (Cairns et al., 1999). We began by exploring whether these Rsc1- or Rsc2-containing subtypes differ in composition, beyond Rsc1/2 themselves. To determine, we purified RSC sub-complexes using a TAP tag (Puig et al., 2001) on either Rsc1 or Rsc2, which revealed Rsc3/30 at apparent stoichiometric levels in the Rsc1-containing complex, but substoichiometric levels in the Rsc2-containing complex (Figure 1A and BChambers et al., 2012). Furthermore, when purifications were conducted under increasing salt conditions, Rsc3/30 association with Rsc1-TAP was maintained, but was lost with Rsc2-TAP (Figure 1—figure supplement 1) confirming higher avidity of Rsc3/30 for the Rsc1 sub-type.

Figure 1 with 2 supplements see all
Rsc3/30 module has higher avidity for the RSC1 complex.

(A) Purified RSC1-TAP and RSC2-TAP complexes (2 µg) analyzed on 7.5% SDS-PAGE gel stained with Coomassie dye. (B) RSC1 and RSC2 complex compositions, with decreased opacity displaying the reduced association of the Rsc3/30 module in the RSC2 complex. (C) Domain structure and swaps of Rsc1 and Rsc2. (D) CT2 domain swaps complement for viability. TRP1-marked plasmids bearing RSC1 (p609), RSC2 (p604), RSC1 w/2CT2 (p3097), RSC2 w/1CT2 (p3098), or vector (pRS314) were transformed into rsc1rsc2∆ [RSC1.URA3] (YBC800), and spotted as 10x serial dilutions to SC-TRP or to SC-TRP+5FOA to force the loss of the RSC1.URA3 plasmid. ’ w/’ indicates ‘with’. One of four biological replicates shown. (E) The Rsc3/30 module associates more strongly with the CT2 region of Rsc1. Immunoprecipitations of Rsc1, Rsc2 w/1CT2, Rsc2, Rsc1 w/2CT2 from whole cell extracts. Blots were probed with anti-Sth1, then stripped and reprobed for anti-Rsc3 and anti-Rsc4. One of three technical replicates shown. Figure 1—figure supplement 1. The Rsc3/30 module associates with the RSC1 complex at high stringency. Figure 1—figure supplement 2. Additional swaps and truncations define the region of Rsc3/30 association.

To identify whether a particular region within Rsc1 mediates this preferential association with Rsc3/30, we performed domain swaps between Rsc1 and Rsc2 (Figure 1C) and checked for complementation and Rsc3 association. To assess this, a rsc1Δ rsc2Δ strain containing RSC1 on a URA3-marked plasmid was transformed with TRP1-marked plasmids containing Rsc1/2 domain swap derivatives, and complementation was demonstrated by their ability to lose the URA3-marked RSC1 plasmid with 5FOA (Figure 1D). Here, co-immunoprecipitation revealed strong Rsc3 association with Rsc1, and with Rsc2 derivatives only if they contained the CT2 region of Rsc1 (Figure 1E). Conversely, strong Rsc3 association with Rsc1 was lost when Rsc1 contained the CT2 region of Rsc2. Similar approaches involving internal Rsc1/2 deletions and additional swap positions provided further refinement, narrowing the region on Rsc1 responsible for strong Rsc3/30 association to residues 617–777 (Figure 1—figure supplement 2A–D). Several structures of RSC-nucleosome complexes have recently been published (Wagner et al., 2020; Ye et al., 2019; Patel et al., 2019). However, as Rsc1, Rsc2, Rsc3, and Rsc30 are within the flexible regions, these structures contain only partial models (or lack density/models). While the Rsc1-Rsc3/30 interaction has not been resolved structurally, crosslinks were observed between Rsc2 CT2 and Rsc3/30 heterodimer (Wagner et al., 2020), an interaction we find to be much stronger in Rsc1.

The RSC1 complex slides 5S nucleosomes better than the RSC2 complex

Having defined and isolated the four main RSC subtypes (RSC1 or RSC2, +/- Rsc3/30; Figure 2A), we then tested for differences in ATPase activity, and remodeling efficiency, by examining nucleosomes of typical wrapping/affinity, such as those formed with the sea urchin 5S DNA nucleosome positioning sequence (NPS) and recombinant yeast histones. First, all four RSC complexes displayed similar DNA-dependent ATPase activities in typical Vmax determinations with plasmid DNA (Figure 2B). The activity of one complex, RSC1-3/30, plateaued earlier than the other RSC complexes, but otherwise the two complexes behaved similarly. However, major differences were observed with 5S nucleosomes – RSC1 complex displayed much greater sliding activity than RSC2 (Figure 2C and Figure 2—figure supplement 1 for quantification). As shown previously for RSC, all sliding activities with RSC1 and RSC2 complexes are ATP dependent (Cairns et al., 1996 and Figure 2—figure supplement 2A). RSC1 and RSC2 complexes both bound 5S nucleosomes comparably in the absence of ATP (Figure 2—figure supplement 2B) indicating that their differences in activities occur after the initial engagement of the 5S nucleosome. Finally, the Rsc3/30 module did not affect their ATPase activities (Figure 2B) nor their binding to 5S nucleosomes (Figure 2—figure supplement 2B); however, in both the RSC1 and RSC2 complexes, Rsc3/30 inclusion moderately inhibited remodeling (Figure 2C). Taken together, RSC1 complex displays markedly higher sliding activity on 5S nucleosomes compared to RSC2 complex.

Figure 2 with 3 supplements see all
RSC1 and RSC2 complexes differ in remodeling activity on a sea urchin 5S mononucleosomal substrate.

(A) Alternative RSC1 and RSC2 complexes, with the Rsc3/30 module maintained or removed during purification. Purified RSC complexes (600 ng) were analyzed on a 6% polyacrylamide SDS-PAGE gel stained with silver. The RSC1 and RSC2 complexes are from the same gel, but were moved adjacent for the depiction. (B) ATPase time course of RSC1 and RSC2 with and without the Rsc3/30 module. Values are the mean +/- standard deviation from two separate RSC preps for each RSC complex assayed in triplicate. (C) Comparative sliding of 174 bp sea urchin 5S yeast mononucleosomes (20 nM) by RSC1 and RSC2 complexes (30 nM). The nucleosomal Start (green), Slid (blue), and free DNA (grey) bands were quantified and reported as a percent of the total signal.

We note that the RSC-remodeled 5S nucleosomal product migrates more slowly on native gels than the starting nucleosome in our sliding assay. This could result from an altered position of the octamer along the DNA, from the partial unwrapping of DNA from the octamer, or possibly from the creation of a hexasome during the remodeling reaction (which would create a more ‘open’ and unwrapped structure). To determine whether the slower-migrating RSC remodeled product was the result of H2A/H2B dimer loss from the nucleosome, we assembled 174 bp 5S nucleosomes using yeast octamers that were fluorescently labeled with Oregon Green (OG) on H2A (Q114C) and conducted sliding assays with RSC1-3/30 (Figure 2—figure supplement 3A). The ratio of H2A to DNA was calculated for both the ‘start’ and ‘slid’ bands at each time point, normalized to the starting nucleosomal band (Figure 2—figure supplement 3B). Here, the ‘slid’ band maintained a dimer to DNA ratio similar to the starting nucleosome, rather than that predicted for a hexasome or tetrasome, supporting the identity of the slower-migrating band being an intact nucleosome. We will, therefore, refer to this band on the native gel as the ‘slid’ nucleosome position.

rsc1Δ mutation is lethal in combination with histone mutations that confer partial unwrapping

In principle, a variety of factors might underlie better relative remodeling by RSC1 complexes on 5S nucleosomes. To provide insight into Rsc1/2 differences and to help guide further in vitro approaches, we conducted unbiased genetic screens to identify histone mutations that are selectively lethal with rsc1Δ or rsc2Δ mutations. Here, we utilized an alanine scanning approach in which each of the histone residues is separately mutated to alanine. To implement, we created rsc1Δ or rsc2Δ deletions (separately) within a ‘histone shuffle’ strain (rsc1Δ or rsc2Δ, h3-h4Δ [H3-H4.URA3]) and combined those with a library encoding all viable histone H3-H4 alanine substitutions on TRP1-marked plasmids (Nakanishi et al., 2008). We then assessed viability following forced loss of the wild-type H3-H4 plasmid on 5FOA-containing medium. Histone mutations that are lethal with both rsc1Δ and rsc2Δ, or lethal uniquely with rsc2Δ, were found distributed throughout the nucleosome. In striking contrast, those mutations that were lethal specifically with rsc1Δ mapped exclusively within the H3 αN helix region – the position where DNA enters/exits the nucleosome (Table 1, Figure 3A–B). Furthermore, the specific H3 αN helix mutations obtained in our screen overlap strongly with those reported to increase partial unwrapping using FRET formats (Ferreira et al., 2007), whereas alanine substitutions that did not show a phenotype (P43A, E50A, K56A) had little effect on FRET/unwrapping (Table 2). Given the results from the H3-H4 screen, we then performed a screen combining rsc1Δ or rsc2Δ with H2A-H2B mutations (Nakanishi et al., 2008). In keeping with the results above, we find rsc1Δ-specific synthetic lethality primarily with mutations in the H2A C-terminus which interact with the H3 αN helix, as well as with specific histone-DNA contacts (H2A R78 and H2A R30; Figure 3—figure supplement 1). Taken together, our genetic results, which encompass the entire nucleosome, strongly suggest that synthetic lethality in rsc1∆ strains is due to decreased ability of RSC2 complexes to remodel partially-unwrapped nucleosomes relative to RSC1 complexes.

Table 1
Summary of Histone H3-H4 screen with rsc1∆ and rsc2∆.

Library of TRP1-marked plasmids containing H3-H4 residues mutated to alanine were transformed into h3-h4Δ [H3-H4.URA3] (YBC1939), rsc1Δ h3-h4Δ [H3-H4.URA3] (YBC2090) or rsc2Δ h3-h4Δ [H3-H4.URA3] (YBC3040), and spotted to SC-TRP, or SC-TRP + 5FOA to force the loss of the WT histone plasmid and test for synthetic lethality. Histone mutations that were lethal on their own in WT RSC are shaded grey, lethal in combination with rsc1Δ are in bold and highlighted yellow, lethal with rsc2Δ are italicized and highlighted in blue, and residues that were lethal with both rsc1Δ and rsc2Δ are highlighted in green.

H3123456789101112
AR2AT11AL20AT32AR40AR49AS57AL65AE73AL82AL92AV101A
BT3AG12AS22AG33AY41AE50AT58AP66AI74AR83AQ93AS102A
CK4AG13AK23AG34AK42AI51AE59AF67AQ76AF84AE94AL103A
DQ5AK14AR26AV35AP43AR52AL60AQ68AD77AQ85AS95AF104A
ET6AP16AK27AK36AG44AR53AL61AR69AF78AS86AV96AE105A
FR8AR17AS28AK37AT45AF54AI62AL70AK79AS87AE97AD106A
GK9AK18AP30AP38AV46AQ55AR63AV71AT80AI89AY99AT107A
HS10AQ19AS31AH39AL48AK56AK64AR72AD81AG90AL100AN108A
H3123456789101112
AL109AQ120AR129A
BI112AK121AL130A
CH113AK122AR131A
DK115AD123AG132A
ER116AI124AE133A
FV117AK125AR134A
GT118AL126AS135A
HI119AR128A
H4123456789101112
AS1AG9AH18AI26AR35AK44AE52AF61AS69AR78AV86AR95A
BG2AL10AR19AQ27AR36AR45AE53AL62AV70AK79AV87AT96A
CR3AG11AK20AG28AL37AI46AV54AE63AT71AT80AY88AL97A
DG4AK12AI21AI29AR39AS47AR55AS64AY72AV81AL90AY98A
EK5AG13AL22AT30AR40AG48AV57AV65AT73AT82AK91AG99A
FG6AG14AR23AK31AG41AL49AL58AI66AE74AS83AR92AF100A
GG7AK16AD24AP32AG42AI50AK59AR67AH75AL84AQ93AG101A
HK8AR17AN25AI34AV43AY51AS60AD68AK77AD85AG94AG102A
Lethal w/rsc1Lethal w/rsc2∆Lethal w/rsc1∆ and rsc2Lethal w/WT
Figure 3 with 5 supplements see all
Mutations in the H3 αN helix are lethal in combination with rsc1∆, but not rsc2∆, and they reduce RSC remodeling of the 5S nucleosome.

(A) Histone H3 αN helix mutations that are lethal with rsc1∆. TRP1-marked plasmids containing WT H4, and H3 mutations within the αN helix were transformed into h3-h4∆ [H3-H4.URA3] (YBC1939), rsc1h3-h4∆ [H3-H4.URA3] (YBC2090) or rsc2h3-h4∆ [H3-H4.URA3] (YBC3040), and spotted to SC-TRP or SC-TRP+5FOA to force the loss of the WT histone plasmid. Mutations that were lethal on their own without mutated RSC are shaded in grey. Mutations that were lethal in rsc1∆ but not rsc2∆ are shaded in purple and underlined. Transformants were grown at 30°C for 2 days. Shown is one of two biological replicates. (B) Location of the synthetic lethal rsc1∆ H3 αN helix mutations are depicted in purple on the nucleosome, PDB code 1ID3. (C) The RSC1 CT2 region complements the synthetic lethal rsc1∆ H3 αN helix mutations. rsc1rsc2h3-h4∆ [RSC1.URA3] with [H3.WT, R40A, or G44A-H4.WT. LYS2] (YBC3466, YBC3444, YBC3433) transformed with TRP1-marked plasmids bearing RSC1 (p609), RSC2 (p604), RSC1 w/2CT2 (p3097), RSC2 w/1CT2 (p3098), or vector (pRS314) and spotted as 10x serial dilution to SC-TRP-LYS, or SC-TRP-LYS+5FOA. Shown is one of four biological replicates. (D) High-copy RSC3 or RSC30 will partially suppress the Tsˉ phenotype of rsc1H3V46A. Strain rsc1h3-h4∆ [H3.V46A-H4.WT.TRP] (YBC3586) transformed with URA3-marked high copy (2μ) plasmids containing RSC1 (p705), RSC3 (p1310), RSC30 (p916), HMO1 (p3390), or vector (pRS426), and spotted as 10x serial dilutions at 30°C or 37°C. Shown is one of two biological replicates. (E) Comparative sliding and ejection of 174 bp sea urchin 5S NPS H3 R40A yeast mononucleosomes (20 nM) by RSC1 and RSC2 complexes (30 nM). The nucleosomal Start (green), Slid (blue), and free DNA (grey) bands were quantified and reported as a percent of the total signal.

Table 2
*DNA end to end FRET measurements on mononucleosomes containing H3 αN helix mutations from Ferreira et al., 2007 with the phenotype when combined with RSC, rsc1∆, or rsc2∆.
Mutation* FRET %Phenotype
WT100 ± 6No phenotype
R40A71 ± 7Lethal w/rsc1
Y41A72 ± 7Lethal w/WT
K42A54 ± 3Lethal w/rsc1
P43A91 ± 3No phenotype
G44A68 ± 9Lethal w/rsc1
T45A52 ± 5Lethal w/WT
V46A89 ± 5Lethal w/rsc1
L48A86 ± 4Lethal w/WT
R49A67 ± 3Lethal w/rsc1
E50A98 ± 6No phenotype
I51A81 ± 13Lethal w/WT
R52A78 ± 3Lethal w/rsc1
R53A80 ± 4No phenotype
F54A96 ± 3Lethal w/WT
Q55A69 ± 11Lethal w/WT
K56A95 ± 4No phenotype
S57A102 ± 5No phenotype
K56Q82 ± 2Lethal w/rsc1

Increased Rsc3/30 association with RSC suppresses phenotypes associated with histone αN helix mutant combinations

We then explored whether the genetic differences observed with histone αN helix mutations in combination with rsc1Δ or rsc2Δ involve differential interaction of Rsc1/2 with Rsc3/30. First, we tested whether the CT2 domain of Rsc1, which interacts better with Rsc3/30 than its counterpart in Rsc2, could confer growth when placed within a Rsc2 derivative. Here, growth was clearly restored to rsc1Δ H3 G44A and rsc1Δ H3 R40A combinations with a plasmid encoding Rsc1 or Rsc2 bearing the Rsc1 CT2, but not with Rsc2 or Rsc1 bearing the Rsc2 CT2 (Figure 3C). Additional domain swap experiments localize this complementation to the Rsc3/30 association region in Rsc1 (aa 617–777) (Figure 3—figure supplement 2A). As a complementary approach, we tested for RSC3 or RSC30 high-copy plasmid suppression. While high-copy RSC3 or RSC30 could not rescue rsc1Δ αN helix histone combined lethality (data not shown), suppression of the strong temperature sensitivity (Ts-) phenotype of rsc1Δ H3 V46A was observed (Figure 3D). Taken together, improving the association/functionality of Rsc3/30 suppresses phenotypes associated with αN helix histone mutations – further linking Rsc1/2 and Rsc3/30 to partial nucleosome unwrapping.

Rsc1/2 differences are largely independent of bromodomain-histone interactions

We then tested the alternative hypothesis that the bromodomains of Rsc1/2 might interact differently with the main acetylated histone residue withinthe αN helix, H3 K56. However, the H3 K56A mutation was not lethal in combination with either rsc1Δ or rsc2Δ (Figure 3A). We then further tested H3 K56Q, which mimics the acetylated form. The K56Q mutation confers synthetic lethality when combined with rsc1Δ, but not with rsc2Δ (Figure 3—figure supplement 2B). Furthermore, domain swaps involving the highly homologous bromodomains and BAH domains between Rsc1 and Rsc2 did not confer phenotypic differences, nor did these swaps alter the synthetic lethality of rsc1Δ with histone αN helix mutations (Figure 3—figure supplement 2A). Notably, H3 K56A has little effect on DNA unwrapping, whereas K56Q promotes unwrapping (Ferreira et al., 2007; Masumoto et al., 2005) – and is lethal with rsc1Δ, further supporting an unwrapping function as being responsible for the phenotype. Thus, the lethality with rsc1Δ does not appear linked to bromodomains or histone acetylation, in agreement with findings that H3K56 acetylation does not enhance RSC binding (Neumann et al., 2009).

RSC2 complexes are deficient in remodeling partially-unwrapped nucleosomes

Our genetic results prompted the examination of sliding by the RSC1 and RSC2 complexes on 5S nucleosomes bearing a mutation (e.g., H3 R40A) predicted to confer partial unwrapping. Although sliding of nucleosomes bearing H3 R40A by either form of RSC is reduced relative to wild-type (WT) 5S nucleosomes, RSC1 complexes were markedly more active than RSC2 complexes (Figure 3E and Figure 3—figure supplement 3 for quantification), reinforcing the difference between RSC1 and RSC2. Additionally, both RSC1 and RSC2 complexes bind H3 αN helix mutant 5S nucleosomes similarly (Figure 3—figure supplement 4), demonstrating that initial nucleosome binding is not inhibited by this octamer mutation, suggesting downstream remodeling activity as the affected step. To confirm and better define the extent of partial unwrapping observed on our 174 bp 5S nucleosome with yeast octamers, we conducted ExoIII-S1 nuclease mapping (which removes DNA that is either outside of, or not well wrapped in a nucleosome; Flaus, 2011), and combined this with a high-throughput paired-end sequencing approach to define the endpoints and proportion of the nuclease-protected species. We found WT 174 bp 5S yeast nucleosomes display a fully-wrapped side (position 158), and a side of partial unwrapping, in agreement with asymmetric 5S nucleosome unwrapping previously shown (Winogradoff and Aksimentiev, 2019; Chen et al., 2014). Here, WT nucleosomes displayed a much higher proportion of largely-wrapped species (≥135 bp) than did H3 R40A nucleosomes (Figure 3—figure supplement 5). Since RSC2 is more deficient than RSC1 in repositioning H3 R40A nucleosomes, the mapping supports the hypothesis that RSC1 complexes manage partially-unwrapped nucleosomes better than RSC2 complexes. We note that H3 R40A 5S nucleosomes are likely to have a distinct conformation that is not distinguished in our nuclease protection assay, since remodeling by both RSC1 and (more so) by RSC2 complexes is inhibited by the H3 αN helix mutation (compare Figure 3E with Figure 2C). These mutant octamers may enforce a greater degree of openness or distance between the 5S DNA ends, as demonstrated previously (Ferreira et al., 2007), and thereby inhibit RSC activity (perhaps the transition of binding to DNA translocation, see Discussion) – a nucleosome perturbation better managed by RSC1 complexes.

RSC cooperates with Hmo1 to remodel partially-unwrapped nucleosomes

Hmo1 is an HMGB family protein that stabilizes fragile/partially-unwrapped nucleosomes, particularly at rRNA and ribosomal protein gene promoters (Hall et al., 2006; Panday and Grove, 2017). Unlike other HMGB proteins which have an acidic CTD that promotes bending and nucleosome destabilization, Hmo1 is unique in containing a basic lysine rich C-terminal extension, and has been shown to stabilize chromatin and perform the functions of a linker histone (Panday and Grove, 2016). Hmo1 is proposed to bind near the nucleosome dyad and use its basic extension to bind linker DNA and prevent bending (Panday and Grove, 2017), which may also improve wrapping and thus cooperate with RSC1 or RSC2 to promote remodeling. We first tested for rsc/hmo1 genetic interactions by combining rsc1Δ or rsc2Δ with hmo1Δ. Notably, we observe synthetic phenotypes in rsc1Δ hmo1Δ mutants, but not rsc2Δ hmo1Δ mutants (Figure 4A), suggesting a greater reliance of Rsc2 on Hmo1 for functional cooperativity. Furthermore, as we saw with the H3 αN helix mutations, this synthetic sickness was partially complemented by the presence of the Rsc1 CT2, and partially suppressed by high-copy RSC3 or RSC30, providing further support that these proteins work together in a modular manner (Figure 4B).

Figure 4 with 6 supplements see all
Hmo1 cooperates with RSC to remodel fragile or partially-unwrapped nucleosomes.

(A) An hmo1 null mutation is synthetically sick with rsc1Δ, but not rsc2Δ. WT (YBC604), rsc1Δ (YBC774), rsc2Δ (YBC82), hmo1Δ (YBC3509), rsc1∆ (YBC774), rsc2Δ (YBC82), rsc1Δ hmo1Δ (YBC3514), rsc2hmo1∆ (YBC3515) spotted as 10x serial dilutions to YPD 30°C, YPD 38°C, and SC+20 µg/ml mycophenolic acid (MPA). One of two or more biological replicates shown. (B) The rsc1Δ hmo1Δ synthetic sickness is suppressed by high copy RSC3, RSC30, or RSC1 CT2. Strain rsc1hmo1∆ (YBC3514) transformed with TRP1-marked RSC1 (p609), RSC2 (p604), RSC1 w/2CT2 (p3097), RSC2 w/1CT2 (p3098), 2µ.RSC3 (p929), 2µ.RSC30 (p911), or vector (pRS314) spotted as 10x serial dilutions to SC-TRP 30°C or SC-TRP 35°C. One of four biological replicates shown. (C) Co-IP of Rsc1 and Rsc2 with Hmo1. Sonicated chromatin extracts from RSC1.9XMYC HMO1.V5 (YBC3558) and RSC2.9XMYC HMO1.V5 (YBC3559) were immunoprecipitated using anti-Myc or anti-V5. Western blots were probed with anti-Myc or anti-V5 antibodies. One of three biological replicates shown. (D) Comparative sliding by RSC1 and RSC2 complexes (10 nM) of 174 bp sea urchin 5S yeast mononucleosomes (20 nM) pre-incubated with increasing concentrations of Hmo1 protein. Reactions were conducted at 30°C for 20 min. The Start (green) and Slid (blue) bands were quantified and reported as percent of the total signal. The free DNA band was negligible and not quantified. (E) Comparative sliding and ejection of Widom 601 yeast mononucleosomes (20 nM) by RSC1 and RSC2 complexes (10 nM). The nucleosomal Start (green), Slid (blue), and free DNA (grey) bands were quantified and reported as a percent of the total signal.

To test RSC-Hmo1 associations in vivo, we performed co-immunoprecipitations between Rsc1 and Rsc2 with Hmo1. Hmo1 was endogenously tagged at its C-terminus with a V5 epitope in Myc-tagged Rsc1 and Rsc2 strains. Here, crosslinked chromatin extracts were prepared from log phase cells and sonicated, or treated with micrococcal nuclease, resulting primarily in mononucleosomes. Immunoprecipitation with anti-Myc or anti-V5 antibody followed by immunoblot analysis revealed that both Rsc1 and Rsc2 co-immunoprecipitate with Hmo1, which represents either a direct interaction between Hmo1 and RSC or colocalization on the same nucleosome(s) (Figure 4C, Figure 4—figure supplement 1). Taken together, Hmo1 shows physical interaction on chromatin with Rsc1 and Rsc2, but strong functional/genetic interaction primarily with Rsc2 (RSC2 complexes are reliant on Hmo1, in a rsc1Δ strain).

Hmo1 strongly stimulates the sliding activity of Rsc2, and moderately stimulates Rsc1

Hmo1 stimulates the sliding activity of RSC and SWI/SNF complex in vitro on 601 nucleosomes (Hepp et al., 2014), but it is not known whether stimulation applies equally to RSC1 and RSC2 complexes, or has a differential effect on their action upon partially-unwrapped nucleosomes. One explanation for why rsc1Δ hmo1Δ mutants grow more poorly than rsc2Δ hmo1Δ mutants is that RSC1 is able to remodel those locations in the absence of Hmo1, but RSC2 is not. As Hmo1 enhances RSC remodeling activity (Hepp et al., 2014), we preincubated the 5S nucleosomal template with increasing amounts of purified Hmo1 and tested for stimulation of RSC1/2 sliding activity. Interestingly, Hmo1 moderately stimulates RSC1 activity, whereas Hmo1 greatly stimulates RSC2 activity (Figure 4D and Figure 4—figure supplement 2 for quantification), a property that is even more evident in conditions of limiting remodeler or at early time points (Figure 4—figure supplement 3). Moreover, both RSC1 and RSC2 slide the 5S nucleosome at lower remodeler concentrations in the presence of Hmo1 (30 nM RSC without Hmo1 compare with 10 nM RSC with Hmo1), further supporting a role for Hmo1 in stimulating RSC remodeling activity.

Well-wrapped nucleosomes are remodeled comparably by both RSC1 and RSC2 complexes

Beyond Hmo1 addition, we explored the ability of RSC1 and RSC2 complexes to remodel nucleosomes bearing a very strong positioning sequence: the optimized 601 sequence. Here, RSC1 and RSC2 complexes both displayed robust sliding activity with 601 nucleosomes, with RSC1 activity slightly higher (Figure 4E and Figure 4—figure supplement 4 for quantification). We note a correlation here and in prior work (Clapier et al., 2016) between ejection and the use of strong positioning sequences – their initial stability from favored ‘phasing’ may convert to high instability and disfavored ‘phasing’ after 5 bp of DNA translocation – which may underlie the progressive ejection observed with 601 nucleosomes.

We then explored whether an extremely strong positioning sequence (601) might remain wrapped in the presence of an αN helix mutation and rescue RSC2 remodeling. Indeed, we found robust remodeling of 601 H3 R40A nucleosomes by RSC2 (Figure 4—figure supplement 5A and Figure 4—figure supplement 6 for quantification), and (by nuclease susceptibility) found that these nucleosomes remain largely wrapped (Figure 4—figure supplement 5B). Our data support the model that while the H3 R40A mutation can result in an open and more loosely wrapped nucleosome conformation, this can be overcome by a strong DNA positioning sequence. In addition, the data indicate that the αN helix H3 R40A mutation itself does not impair remodeling by RSC2 unless that mutation confers partial unwrapping due to the underlying DNA sequence. Lastly, our mapping data confirm that 5S nucleosomes produced with yeast octamers are less well-wrapped than 601 nucleosomes, and have DNA ends that are more susceptible to nuclease digestion.

Rsc1 and Rsc2 both occupy wide NDRs, with Rsc1 specifically occupying tDNAs

RSC has been examined by chromatin co-immunoprecipitation (ChIP) in several formats (Damelin et al., 2002; Yen et al., 2012; Vinayachandran et al., 2018; Brahma and Henikoff, 2019), though only one study has examined Rsc1/2 differences. This early ChIP approach with microarrays examined Rsc1 and Rsc2 differences, and revealed similar promoter targets for RSC1 and RSC2 complexes (Ng et al., 2002). However, possible issues of sensitivity with earlier data combined with the ability to perform more advanced approaches and analyses prompted ChIP-seq experiments to reveal possible differences. Recent examination of Sth1/RSC occupancy by the ‘cut and run’ MNase approach (which would have superimposed and not distinguished between Rsc1 and Rsc2 occupancy) revealed general enrichment for RSC complexes at many +1 and −1 nucleosomes, as well as high enrichment of RSC within the NDR of genes with very wide NDRs (e.g., ribosomal protein genes). Notably, these wide NDR regions/promoters have been shown to contain RSC bound to partially-unwrapped nucleosomes, rather than the complete lack of nucleosomes (Brahma and Henikoff, 2019). Therefore, we will hereafter refer to them as ‘partially-unwrapped regions’ when appropriate, and use the term ‘NDR’ to generally refer to the nucleosome-deficient regions between the −1 and +one nucleosome.

To further examine the genome-wide location of RSC1 and RSC2 complexes and their relation to nucleosome wrapping, we utilized MYC-tagged derivatives of Rsc1, Rsc2 and Rsc3 (tagged at their endogenous loci) and performed MNase-based ChIP-seq from logarithmically growing cells. We observe RSC at locations similar to prior work (Brahma and Henikoff, 2019), and largely comparable profiles between Rsc1 and Rsc2 at most Pol II genes (as described previously Ng et al., 2002). However, we observe differential occupancy in two regions: promoters with wide NDRs and tDNAs.

First, we observed high occupancy of RSC (Rsc1, Rsc2 and Rsc3; Figure 5A) at promoters with wide NDRs, with Rsc1 appearing more enriched than Rsc2. As promoters with partially-unwrapped nucleosomes often contain the HMGB protein, Hmo1, we also compared the occupancy of Hmo1 using the ChIP-seq data (Knight et al., 2014) reprocessed using our parameters. Notably, Hmo1 occupancy positively correlates with regions displaying the widest NDRs and bearing the highest Rsc1 and Rsc2 occupancy, with an apparent higher correlation with Rsc1 (Figure 5A–D). The loci with the highest Hmo1 occupancy displayed a mean region size of 336 bp, and a remarkable two-thirds of those genes were ribosomal protein genes (RPGs). Ribosomal protein gene promoters contain GC-rich sequences within their ‘NDR’ region, and in keeping, we observe Rsc3 enrichment, consistent with Rsc3/30 involvement in targeting or retention.

Figure 5 with 3 supplements see all
Rsc1 and Rsc2 occupy promoters with wide NDRS, with preferential occupancy of RSC1 at tDNAs.

(A) Heat maps showing enrichment of nucleosomes, Rsc1, Rsc2, Rsc3, and Hmo1 at promoters, sorted by NDR length. (B) Violin plot of Rsc1, Rsc2, and Hmo1 occupancy at promoters at three categories of NDR length. (C) Plot of Rsc1, Rsc2, and Hmo1 occupancy compared to NDR length. (D) Plot of Rsc1 and Rsc2 enrichment by Hmo1 occupancy. (E) Heat maps showing enrichment of nucleosomes, Rsc1, Rsc2, Rsc3, and Hmo1 at all tDNAs (tRNA encoding genes). (F) Violin plot of Rsc1, Rsc2, and Hmo1 mean log2 fold enrichment at tDNAs. (G) Number of genes affected by rsc and hmo1 deletions. For each mutation the number of genes up or downregulated two fold or more compared to WT, rsc1Δ (↑129↓45), rsc2Δ (↑160↓129), hmo1Δ (↑44↓132), rsc1∆ hmo1∆ (↑838↓1028), rsc2∆ hmo1∆ (↑1336↓1131). (H) Gene Expression changes. Violin plots of RNA expression for each mutation at all pol II genes (6145 genes) and at ribosomal protein genes (132 genes) as compared to WT expression. ChIP-seq and RNA-seq data shown represents averages of two and three biological replicates, respectively.

The most striking difference between Rsc1 and Rsc2 occupancy was observed at tDNAs. tDNAs encode tRNAs, and are approximately the size of a single nucleosome - though they are among the most nucleosome-depleted loci in the yeast genome, due at least in part to the action of RSC (Parnell et al., 2008) and their high fractional occupancy by RNA polymerase III transcription factors (Kumar and Bhargava, 2013). Here, Rsc1 complexes are markedly more enriched than Rsc2 complexes, and Hmo1 is notably absent (Figure 5E,F). Notably, many tDNAs are flanked by highly AT-rich sequences (Giuliodori et al., 2003), which could (in principle) confer partial unwrapping during the uncommon times (e.g., after replication) when tDNAs are transiently unoccupied by Pol III; here, RSC1 complexes might conduct nucleosome sliding to reveal tDNA sequences to the Pol III machinery. Indeed, the high fractional occupancy of Pol III complexes likely underlies the lack of Hmo1 at tDNAs, as Pol III factors would likely compete with Hmo1 for DNA binding and can remove nucleosomes during transcription. Taken together, Rsc1 and Rsc2 both highly occupy Pol II promoters known to contain partially-unwrapped nucleosomes and Hmo1, whereas Rsc1 preferentially occupies tDNAs.

RSC and HMO1 cooperate to regulate gene expression

To address the transcriptional effects of each of these proteins, we performed RNAseq on logarithmically growing WT, rsc1Δ, rsc2Δ, hmo1Δ, rsc1Δ hmo1Δ, and rsc2Δ hmo1Δ cells cultured in SC media. We observe modest differences in transcriptional profiles between WT, rsc1Δ, and rsc2Δ (Figure 5G–H), which we interpret to be reflective of the known redundancy between Rsc1 and Rsc2 for most functions.

However, combining rsc1∆ or rsc2∆ with hmo1∆ resulted in a strong transcriptional shift, resulting in the upregulation (>2 fold) and downregulation (>2 fold) of ~1000 genes (Figure 5G), with the most affected class of genes including RPGs, which are downregulated (Figure 5H). Additionally the transcriptional shift observed in the double mutants largely mirrors the shift of both up-regulation and down-regulation observed in response to environmental stress response (ESR) genes (Gasch et al., 2000; Brion et al., 2016Figure 5—figure supplement 1). These results suggest that Rsc1 and Rsc2 both cooperate with Hmo1 to promote the transcription of ribosomal protein genes, and that the burden for chromatin remodeling at these and other growth regulated loci requires the action of both Rsc1 and Rsc2, if Hmo1 is absent.

As Rsc3/30 also interacts with and affects Rsc1/2 function, we further analyzed rsc1rsc30∆, and rsc2rsc30∆ mutants (Figure 5—figure supplement 2). Remarkably, only rsc2rsc30∆ mutants strongly impact RP genes – suggesting that the RSC1 complex, which interacts more strongly with Rsc3/30, is more reliant on Rsc3/30 than is the RSC2 complex, and that in a rsc2rsc30∆ both complexes are impaired (Figure 5—figure supplement 2C).

Given our exploration of Rsc1/2 paralog and Rsc3/30 paralog function, we also explored their evolution. A whole genome duplication (WGD) occurred within the Saccharomyces lineage approximately 150 million years ago (Wolfe and Shields, 1997) resulting in RSC1/2 paralogs. We find the single RSC1/2 ortholog present in species that did not undergo the WGD (e.g., Zygosaccharomyces, Ashbya, and Lachancea), more closely related to RSC2 than RSC1, suggesting RSC2 as the more ancient ortholog (Figure 5—figure supplement 3). Notably, species more distant to S. cerevisiae lack both RSC3 and RSC30. Following the appearance of RSC3, the WGD event then created RSC1/2 and RSC3/RSC30 orthologs, along with duplications of the ribosomal protein genes, a large fraction of which have been maintained in Saccharomyces cerevisiae. Finally, HMO1 predates the appearance of RSC3 and the WGD and is found in Schizosaccharomyces pombe (Albert et al., 2013). Thus, specialization of Rsc1 to preferentially bind Rsc3/30 and play a role in the regulation of the ribosomal protein genes was enabled by the WGD, and may have arisen to help rapidly and properly regulate RPGs in response to growth conditions.

Taken together, our results are consistent with RSC1 complexes bearing a higher intrinsic ability to mobilize partially-unwrapped nucleosomes compared to RSC2, augmented by partner proteins that preferentially assist either RSC1 (Rsc3/30) or RSC2 (Hmo1) complexes to more efficiently remodel at locations with partially-wrapped nucleosomes (Figure 6).

Model for RSC1 and RSC2 action on wrapped versus partially-unwrapped nucleosomes.

Here, flexible/positioning DNA sequences and the protein Hmo1 promote wrapping (Anderson et al., 2002; Iyer, 2012; Panday and Grove, 2017). In contrast, partial unwrapping can be facilitated by ‘stiff’ (AT-rich) DNA sequences, acetylation (e.g., H3 K56ac), the binding of general regulatory transcription factors (not shown) to entry/exit DNA, or by mutation of residues within/near the H3 αN helix (which normally binds entry/exit DNA) (Segal and Widom, 2009; Neumann et al., 2009; Knight et al., 2014; Ferreira et al., 2007). Whereas fully-wrapped nucleosomes are remodeled well by RSC1 or RSC2 complexes, partially-unwrapped nucleosomes are better managed and remodeled by RSC1 complexes.

Discussion

Genomes are diverse in DNA sequence composition, and the biophysical properties of DNA – in particular, DNA curvature and stiffness – creating a spectrum of affinities for nucleosomes, the main packaging unit of chromatin. This spectrum can undergo selection to create regions where nucleosome formation, positioning, and/or turnover is favored or disfavored, properties which can be utilized at promoters and enhancers to help regulate transcription factor binding, transcription, and ultimately fitness. These biophysical properties work in concert (and sometimes in opposition) with the action of chromatin remodelers, which utilize ATP to move nucleosomes to either favored or disfavored positions, and to eject nucleosomes to provide regulated access of transcription factors to DNA. The commonness in yeast of stiff/disfavored DNA sequences at proximal promoters, juxtaposed to bendable/favorable sequences at the −1 and +1 positions (especially at constitutive or highly transcribed genes), raises the possibility that chromatin remodelers may have undergone specialization to manage the remodeling of both fully wrapped and partially-unwrapped nucleosomes.

We begin by briefly discussing prior studies that address the extent of partial unwrapping of 5S or 601 DNA sequences. It has been established that the 601 sequence displays higher overall affinity for the histone octamer than does the 5S sequence (Thåström et al., 1999). However, a separate report found that the 601 DNA sequence (with Xenopus histones) displayed more unwrapping from the nucleosome edge (entry/exit) than does the Xenopus 5S sequence (North et al., 2012). It is worth noting, this work inserted a LexA binding sequence into both the 5S and 601 sequences (by replacing the existing sequence) which involves the region that unwraps from the octamer. Notably, a set of subsequent studies (Chen et al., 2014; Mauney et al., 2018) used small angle X-ray scattering (SAXS) to compare the native/unaltered sea urchin 5S to 601 nucleosomes, which showed that the 5S sequence unwraps more rapidly and at lower salt concentrations than does 601.

Here, we provide several lines of evidence that the RSC1 complex can slide partially-unwrapped nucleosomes better than its paralog, the RSC2 complex. Furthermore, as RSC is assisted by additional proteins (e.g., Hmo1) that augment this function in vivo, RSC can be thought of as two functional entities: ‘Rsc2 with Hmo1’ and ‘Rsc1 with Rsc3-30’, with strong phenotypes only observed when both entities are impaired (Figure 6). First, we find that RSC1 acts more efficiently on 5S nucleosomes, and retains activity on nucleosomes bearing H3 R40A, a mutation that favors unwrapping, whereas RSC2 is relatively impaired. Second, an unbiased genetic screen revealed synthetic lethality with rsc1Δ αN helix combinations, but not rsc2Δ αN helix combinations, strongly suggesting that the RSC2 complex has more difficulty remodeling partially-unwrapped nucleosomes. Third, Hmo1 stabilizes nucleosomes (Panday and Grove, 2016; Panday and Grove, 2017), and we observe synthetic phenotypes with rsc1Δ hmo1Δ mutants, but not with rsc2Δ hmo1Δ mutants (Figure 4A), suggesting that Rsc2 relies much more on Hmo1 for functional cooperativity than does Rsc1. This result is paralleled in our remodeling experiments, which show a more pronounced rescue of 5S nucleosome sliding by RSC2 complexes than RSC1 complexes with recombinant Hmo1 (Figure 4D). We note that as nucleosome binding affinity is very similar between RSC1 and RSC2 complexes, the defect/challenge in remodeling with the RSC2 complex may involve a step downstream of binding, perhaps involving the ability of the complex to commit to the initiation of DNA translocation, which may be sensitive to the conformation of the nucleosome. Alternatively, nucleosome conformation/wrapping may impact the efficiency of DNA translocation – termed ‘coupling’ – which involves the probability of each ATP hydrolysis resulting in 1 bp of productive DNA translocation, a property known to be regulated in RSC complex by the actin-related proteins, Arp7 and Arp9 (Clapier et al., 2016; Szerlong et al., 2008). Finally, we emphasize that partially-unwrapped nucleosomes are not preferred by RSC1 complex over wrapped nucleosomes; RSC1 simply manages partially-unwrapped nucleosomes better than does RSC2 complex.

Initially, it may seem counter-intuitive for partial unwrapping of a nucleosome to impede rather than enhance RSC remodeling. However, we note that partial unwrapping is not an intermediate to full unwrapping by CRCs – as full unwrapping followed by full octamer re-wrapping is not the mechanism of nucleosome sliding. Therefore, a partially-unwrapped nucleosome is not necessarily a remodeling intermediate, and therefore not necessarily stimulatory. Instead, we suggest that sliding involves a 1 bp/ATP DNA translocation mechanism imposed on a fully wrapped, or partially-wrapped nucleosome. Here, we suggest that partial unwrapping results in a nucleosome conformation that can indeed be encountered/bound by RSC1/2 complexes, but resists transition to the DNA translocation phase of remodeling – with RSC2 complexes less able to conduct this transition than RSC1 complexes (Figure 6). A second related model would be that RSC1 complex might better induce or stabilize a wrapped nucleosome conformation that is more conducive to initiating or continuing remodeling.

Above, we show that the product of remodeling (using 5S nucleosomes formed with yeast octamers) migrates more slowly than does the initial substrate, and (by using Oregon Green-labeled H2A) that the product is not a hexasome or tetrasome involving H2A-H2B dimer loss. Nucleosome migration is determined by both the octamer position along the DNA (center vs end-positioned), and the overall shape of the octamer-DNA complex, including the extent to which the product/slid nucleosome is wrapped – as partial unwrapping and an ‘open’ conformation would be predicted to result in a slower migrating species through a native gel, just as a more ‘compact’ conformation results in a faster migrating species (Chakravarthy et al., 2012).

The mammalian polybromo protein (PBRM1) helps define the PBAF sub-complex of mammalian SWI/SNF complex, and is similar in domain composition to the combination of Rsc1/2 and the Rsc4 protein (which contains multiple bromodomains). PBRM1 also contains an HMG domain, which is notably absent in the Rsc1, 2, and 4 combination of domains. Here, we speculate that the Hmo1 protein and the HMG domain of PBRM1 may have functional similarities in managing DNA wrapping, which can be tested in future work.

RSC contains two proteins with affinity for GC-rich sequences: the paralogs Rsc3 and Rsc30. Here, we reveal a higher avidity of the Rsc3/30 module for the RSC1 complex, localize the region of interaction of Rsc3/30 with Rsc1/2 to the beginning of the CT2 domain, and provide genetic evidence that the preferred interaction of Rsc3/30 with Rsc1 has functional consequences. A current curiosity is the observation – true for both Rsc1 and Rsc2 complexes – that the presence of Rsc3/30 moderately inhibits remodeling, while the genetics supports a positive role for Rsc3/30 in assisting Rsc1/2 function. One obvious role for Rsc3/30 is in targeting RSC to GC-rich promoter sequences such as ribosomal protein genes, which likely underlies the essential nature of Rsc3/30 function. Here, future work may explore whether Rsc3/30 serve a regulatory role in the remodeling reaction, with their modest attenuation function relieved in the proper regulatory contexts of DNA sequence (e.g., GC richness) and protein composition (e.g., Hmo1, others) to help confer environmental sensing and properly regulate RPGs.

In favorable growth conditions, approximately 50% of the transcriptional effort of RNA Pol II is directed at RPGs (Warner, 1999). Our work supports and greatly extends the prior work of others (Knight et al., 2014; Hepp et al., 2017; Wade et al., 2004) that S. cerevisiae has evolved many ways to ‘poise’ chromatin at RPG promoters in a relatively ‘open and ready’ format. These promoters have evolved to be less favorable to nucleosomes by bearing regions of ‘stiff’ AT-rich DNA (Segal and Widom, 2009) (note: Rsc1/2 each bear an AT hook), are punctuated by GC-rich DNA (known Rsc3/30 binding sites), and use histone acetylation (e.g., H3 K56ac) (Neumann et al., 2009), Hmo1 (Hall et al., 2006), DNA-binding general regulatory factors (e.g., Rap1, Abf1) and chromatin remodeling complexes such as RSC and SWI/SNF complex – that together help keep these regions nucleosome deficient and partially-unwrapped (Kubik et al., 2018; Hepp et al., 2017; Brahma and Henikoff, 2019; Reja et al., 2015). While RSC1 and RSC2 complexes appear largely redundant at most genes, and can both be assisted at RPG promoters by Hmo1, our work supports the notion that the RSC1 complex, through its ability to better manage partially-unwrapped nucleosomes and preferential association with Rsc3/30, has become specialized to help perform this role (Figure 6). This combination of specialization and partial redundancy provides this system the needed robustness and the ability to conduct rapid and sophisticated activation/regulation of this important RPG class – attributes which may contribute to fitness in diverse environments.

Materials and methods

Key resources table
Reagent type
(species) or resource
DesignationSource or
reference
IdentifiersAdditional
information
Gene Saccharomyces cerevisiaeYeast GenomeUCSCSacCer3
Strain, strain background (Escherichia coli)BL21-CodonPlus (DE3)RILAgilentCat# 230245
AntibodyAnti-Sth1
(Rabbit polyclonal)
Cairns et al., 1996(1:1000)
AntibodyAnti-Rsc3
(Rabbit polyclonal)
Angus-Hill et al., 2001(1:1000)
AntibodyAnti-Rsc30
(Rabbit polyclonal)
Angus-Hill et al., 2001(1:1000)
AntibodyAnti-Rsc2
(Rabbit polyclonal)
Kasten et al., 2004(1:1000)
AntibodyAnti-Rsc4
(Rabbit polyclonal)
Kasten et al., 2004(1:1000)
AntibodyAnti-HA
(mouse monoclonal)
Cairns et al., 1999(1:1000)
AntibodyAnti-Myc
(mouse monoclonal)
AbcamCat# ab56
RRID:AB_304876
(1:1000)
AntibodyAnti-V5
(mouse monoclonal)
Thermo ScientificCat#
R960-25
RRID:AB_2556564
(1:1000)
Peptide, recombinant protein8XHIS.HMO1This paper, Figure 4purified from E. coli BL21-CodonPlus(DE3)-RIL cells
Peptide, recombinant proteinHistone H2A,
Oregon green, yeast octamers
Xin et al., 2009
Peptide, recombinant proteinHistone H3 R40AThis paper, Figure 3purified from E. coli BL21-CodonPlus(DE3)-RIL cells
Peptide, recombinant proteinAcTEV proteaseThermo ScientificCat# 12575015
Peptide, recombinant proteinMicrococcal NucleaseUSBCat# 70196Y
Peptide, recombinant proteinExonuclease IIINew England BiolabsCat# M0206
Peptide, recombinant proteinS1 nucleaseThermo ScientificCat#
18001–016
Peptide, recombinant proteinKlenow fragmentNew England BiolabsCat# M0212L
Commercial assay or kitNEBNext ChIP-Seq MasterMix SetNew England BiolabsCat# E6240L
Commercial assay or kitRNEasyQiagenCat# 74106
Commercial assay or kitRiboPure RNA purification kit, YeastThermo ScientificCat# AM1926
Commercial assay or kitTruSeq Stranded Total RNA Library Prep Kit with Ribo Zero GoldIlluminaCat# RS-122–2301
Commercial assay or kitBioRad protein assay reagentBioRadCat#
500–0006
Commercial assay or kitMinelute PCR purification kitQiagenCat#28006
Chemical compound, drug5-Fluoroorotic Acid (5FOA)Toronto Research ChemicalsCat# F59500
Chemical compound, drugMycophenolic Acid (MPA)CalbiochemCat# 475913
Software, algorithmPrismGraphPad Software IncRRID:SCR_002798Version 8.0.2
Software, algorithmDESeq2BioconductorRRID:SCR_015687Version 1.21
Software, algorithmSTARDobin et al., 2013RRID:SCR_015899Version 2.5.4
Software, algorithmRR Development Core Team, 2018RRID:SCR_001905
Software, algorithmpHeatmapKolde, 2019RRID:SCR_016418
Software, algorithmNovoalignNovocraftRRID:SCR_014818Version 3.8.2
Software, algorithmUMI ScriptsHuntsman Cancer InstituteSee Materials and methods for github link
Software, algorithmBioToolBox packagesTJ Parnell, Huntsman Cancer InstituteSee Materials and methods for github link
Software, algorithmImageQuant TLGE HealthcareRRID:SCR_018374
OtherIgG Sepharose Fast FlowGE HealthcareCat#
GE 17-0969-01
OtherCalmodulin Affinity ResinAgilentCat#
214303
OtherNickel-NTA Agarose BeadsQiagenCat#
30230
OtherSlide-A-Lyzer mini dialysis unitsThermo ScientificCat#
69560
OtherDynabeads Pan Mouse IgGThermo ScientificCat#11041

Strains, Media, Yeast growth, and Assay Replication

Request a detailed protocol

Resources used in this study are provided in the Key Resources Table. Rich media (YPD), synthetic complete (SC), and sporulation media were prepared using standard methods. Standard procedures were used for transformations, sporulation, tetrad analysis and spotting. A null mutation of Hmo1 was obtained from Invitrogen and crossed into rsc1 and rsc2 deletion strains. Rsc1 was C-terminally TAP-tagged as described by Rigaut et al., 1999, and Hmo1 C-terminally tagged with V5 as described in Funakoshi and Hochstrasser, 2009. Full genotypes of yeast strains are available in Supplementary file 1. A technical replicate is defined as a test of the same sample multiple times, whereas a biological replicate is defined as the same test run on multiple biological samples of independent origin. The number and type of replicates for each experiment is indicated in the figure legend.

Plasmid construction

Request a detailed protocol

DNA encoding Rsc1 w/Rsc2 CT2 was created by PCR amplification of RSC2 CT2 from position K617 through the end of the gene, using primers containing 40 bp homology to RSC1 or vector sequence. PCR product was co-transformed with linearized p609 (314.RSC1) into YBC800 (rsc1rsc2∆ [316.RSC1]) and homologous recombination inserted the DNA encoding CT2 into RSC1 beginning at position K617. Rsc2 w/1CT2 at position K656 was created by amplifying RSC1 CT2 on p609 (314.RSC1) and co-transforming into linearized p604 (314.RSC2) into YBC803 (rsc1rsc2∆ [316.RSC2]) and repaired by homologous recombination. Plasmid was isolated from TRP positive colonies, sequenced, and retransformed to check for complementation. Additional swaps were constructed similarly. Internal deletions in Rsc1 and Rsc2 were created by digesting with restriction enzymes with blunt or compatible ends, removing the small fragment and ligating ends back together. Plasmid markers were swapped as needed as described in Cross, 1997. A full list of plasmids used in this study is provided in Supplementary file 2.

RSC1 and RSC2 TAP purification

Request a detailed protocol

RSC complexes were purified from S. cerevisiae strains BCY211 (RSC2-TAP) and BCY516 (RSC1-TAP) harboring an integrated version of the C-terminally Tandem Affinity Purification (TAP)-tagged on the construct of the indicated gene. Purifications were performed essentially as described (Rigaut et al., 1999Szerlong et al., 2008) with the following modifications. After harvesting, cells were washed with 0.5x PBS + 10% glycerol, pelleted, and frozen in liquid nitrogen as pea-size pieces. Frozen cells were pulverized in a SPEX SamplePrep 6850 Freezer/Mill at a rate of 10, for 10 cycles of 3 min ‘on’ and 2 min ‘off’, with the resulting powder stored at −80°C until purification.

RSC1+3/30 complex was purified from BCY516 essentially as described (Szerlong et al., 2008) with the following modifications since Rsc1 is found at lower levels than Rsc2 in yeast cells. Pulverized cells (160 g) were solubilized in Lysis Buffer (50 mM HEPES [pH 7.5], 400 mM KOAc, 10 mM EDTA, 20% glycerol, 0.5 mM BME, protease inhibitors). The KOAc concentration of the lysate was brought to 450 mM and nucleic acids were precipitated with 0.1–0.2% polyethyleneimine. The cleared lysate was incubated with 5 ml IgG Sepharose 6 Fast Flow (GE Healthcare) for 3 hr, the resin sequentially washed with Lysis Buffer, IgG Wash Buffer (20 mM Tris-Cl, [pH 8], 400 mM KOAc, 0.1% Igepal, 1 mM EDTA, 0.5 mM BME, protease inhibitors), and IgG Elution Buffer (20 mM Tris-Cl [pH 8], 200 mM KOAc, 10% glycerol, 0.1% Igepal, 0.5 mM BME). The RSC1-TAP complex was cleaved from the IgG resin with 500 U AcTEV enzyme (Invitrogen) gently rotating the slurry at 15°C for 1 hr followed by 4°C rotation overnight. The eluate was collected, brought to 3 mM CaCl2, and bound to 1.5 ml Calmodulin Affinity Resin (Agilent Technologies) overnight. The resin was washed with 250 Calmodulin Wash Buffer (20 mM Tris-Cl [pH 7.5], 250 mM NaCl, 1 mM Mg Acetate, 1 mM imidazole, 2 mM CaCl2, 10% glycerol, 0.1% Igepal, 0.5 mM BME, protease inhibitors). The RSC1 complex was eluted with 250 Calmodulin Elution Buffer (20 mM Tris-Cl [pH 7.5], 250 mM NaCl, 2 mM EGTA, 10% glycerol, 0.1% Igepal, 0.5 mM BME, protease inhibitors) in successive 500–750 µl fractions. The RSC1-containing fractions were determined, pooled, and residual nucleic acid was removed by a final DEAE Sepharose clean-up step. The complex was concentrated on a Vivaspin 30K PES concentrator, concentration was determined by Bradford assay (Bio-Rad) and confirmed by silver stained polyacrylamide SDS-PAGE analysis.

RSC1-3/30 complex was purified as described above except the IgG Wash Buffer had 750 mM NaCl instead of KOAc, and the IgG resin was rotated overnight during the IgG Wash step to remove the Rsc3/30 module. The bound IgG resin was equilibrated into IgG Elution Buffer before cleaving with AcTEV.

RSC2+3/30 complex was purified as above with the following modifications. Approximately 80 g of pulverized BCY211 cells were solubilized in Lysis Buffer. Cleared lysates were incubated with IgG Sepharose, washed and eluted with AcTEV enzyme as above. The eluate was collected, brought to 3 mM CaCl2, and bound to 1.5 ml Calmodulin Affinity Resin for 3 hr. The resin was washed with 200 KOAc Calmodulin Wash (same as Calmodulin Wash but with 200 mM KOAc replacing the NaCl) followed by 150 NaCl Calmodulin Wash (same as Calmodulin Wash but with 150 mM NaCl). The RSC2 complex was eluted with 150 Calmodulin Elution Buffer (same as above but with 150 mM NaCl) in successive 500–750 μl fractions. The RSC2-containing fractions were determined and pooled. The salt was increased to 250 mM NaCl and residual nucleic acid was removed on DEAE Sepharose. The complex was immediately diluted to 125 mM NaCl while maintaining the other buffer component concentrations. The complex was concentrated and the concentration was determined as above.

RSC2-3/30 complex was purified essentially as the RSC2+3/30 complex with the following modifications. After binding the Calmodulin Affinity Resin, the resin was washed with 200 KOAc Calmodulin Wash Buffer, followed by four 20 min washes with 500 mM NaCl Calmodulin Wash Buffer, and a final rinse with 150 NaCl Calmodulin Wash Buffer to equilibrate the resin. The RSC2-3/30 complex was eluted from the Calmodulin Affinity Resin, residual nucleic acid removed on DEAE Sepharose, concentrated and quantified as above for the RSC2+3/30 complex.

Hmo1 purification

Request a detailed protocol

Recombinant Hmo1 was purified from bacteria as an N-terminal fusion to 8xHIS (plasmid from T. Formosa). Briefly, E. coli BL21-CodonPlus(DE3)RIL transformed with pET.8xHIS.Hmo1 was induced with IPTG at 30°C for 3 hr to express the fusion protein. Cells were harvested, resuspended in HIS Lysis Buffer (20 mM Tris [pH 8], 500 mM NaCl, 10 mM imidazole, 10% glycerol, 0.1% Igepal, 1.5 mM BME, protease inhibitors), and lysed by sonication. The lysate was cleared and bound to pre-equilibrated Ni-NTA agarose beads (Qiagen) for 1.5 hr. The bound resin was washed sequentially with HIS tag Lysis Buffer, Ni Wash Buffer A (same as HIS Lysis Buffer but with 30 mM imidazole), and Ni Wash Buffer B (same as HIS Lysis Buffer but with 50 mM imidazole). The protein was eluted with Ni Elution Buffer (20 mM Tris [pH 8], 500 mM NaCl, 250 mM imidazole, 10% glycerol, 0.005% Igepal, 1.5 mM BME, protease inhibitors) and sized on a Superdex 200 gel filtration column (GE). Hmo1 ran as a soluble aggregate (approximately 440 kD) larger than predicted from its molecular weight (27.5 kD). The Hmo1 protein was dialyzed into 150 mM Sizing Buffer (20 mM Tris [pH 8], 150 mM NaCl, 10% glycerol, 1.5 mM BME, protease inhibitors) and concentrated before storing at −80°C in aliquots.

RSC1 and RSC2 Rsc3/30 Stringency Testing

Request a detailed protocol

Stringency testing on Rsc3/30 association with RSC1 and RSC2 was conducted on IgG Sepharose. Cells were grown and pulverized as for RSC purification. Sixty grams of RSC1-TAP cells and 20 g of RSC2-TAP cells were solubilized in Lysis Buffer and the lysate was cleared as above. One milliliter of IgG resin was incubated with the cleared lysate for 2 hr. The beads were washed with Lysis Buffer, and then with IgG Wash Buffer, before being split into 10 separate Eppendorf tubes for testing. The beads were gently pelleted, rinsed three times with the specific IgG Wash Buffer containing 150, 250, or 500 mM NaCl, as indicated. The specific IgG Wash Buffer was added for 1 hr or 16 hr (overnight), as indicated. All beads were washed three times with IgG Elution Buffer, resuspended in IgG Elution Buffer, and RSC was released from the beads with AcTEV enzyme as above and 1/10th of the eluate was analyzed by silver stain on a 6% SDS polyacrylamide gel.

Co-Immunoprecipitations

Request a detailed protocol

Whole cell extracts were prepared and co-immunoprecipitations for RSC were performed as described, (Cairns et al., 1999) with the following modification. Anti-Myc or anti-HA antibodies (0.6 µg) were bound to a 25 µl slurry of pan-mouse magnetic Dynabeads (Thermo Fisher) and incubated with 1000 µg of extract. Blots were probed with anti-Sth1 (Saha et al., 2002), anti-Rsc3, anti-Rsc30 (Angus-Hill et al., 2001), anti-Rsc2 and anti-Rsc4 (Kasten et al., 2004).

Extracts for Rsc1 and Rsc2 chromatin co-IPs with Hmo1 were prepared from cells grown to OD600 = 0.8, Cultures were crosslinked in 1% formaldehyde final for 30 min at RT, quenched with 0.2M glycine, and lysed by bead-beating (mini-beadbeater, Biospec) in LB140 buffer. Samples were sonicated in a Bioruptor Pico (Diagenode) to release chromatin or MNase treated with 100U of micrococcal nuclease for 15 min at 37°C. Immunoprecipitations were performed with 1000 µg of extract and a 25 µl slurry of pan-mouse Dynabeads bound to anti-Myc (Abcam) or anti-V5 (Thermo-Fisher) antibodies.

Nucleosome assembly

Request a detailed protocol

Yeast octamers were produced using purified S. cerevisiae histones expressed in E. coli BL21-CodonPlus(DE3)RIL. Either wild-type histone H3 or the H3 R40A mutant was assembled into yeast octamers by salt-dialysis, essentially as described previously (Dyer et al., 2004). Mononucleosomes were assembled from a linear salt gradient dialysis (2 M to 50 mM KCl) using Slide-A-Lyser Mini Dialysis units with a 7000 molecular weight cutoff (Thermo Scientific) essentially as described (Clapier et al., 2016) using one of several DNA positioning sequences. The 205 bp Widom 601 positioning sequence (Lowary and Widom, 1998) were produced from pUC12 × 601 digested with AvaI and purified using preparative electrophoresis (PrepCell, Bio-Rad). The 174 bp sea urchin 5S positioning sequence was prepared by PCR amplification using a plasmid containing a copy of the sea urchin 5S sequence as a template. The PCR product was precipitated before purifying on a PrepCell. The exact 601 and 5S positioning sequences used are given in Supplementary file 2. Assembled mononucleosomes were separated from free DNA by gradient sedimentation on a 10–30% sucrose gradient as previously described (Wittmeyer et al., 2004). Yeast octamers bearing H2A fluorescently labeled with Oregon Green on Q114C were a gift from L. McCullough and T. Formosa (Xin et al., 2009). These labeled yeast octamers were assembled into nucleosomes with the 174 bp 5S NPS and purified as described above.

ATPase assays

Request a detailed protocol

Measurement of ATP hydrolysis was as described previously (Saha et al., 2002; Wittmeyer et al., 2004) using a color (malachite green) absorbance assay that quantitatively measures released free phosphate. Time courses were performed on two separate purifications of each type of RSC complex. Reactions were performed in triplicate. Activities were measured at Vmax using double-stranded Bluescript plasmid as the substrate.

Nucleosome sliding assay

Request a detailed protocol

Nucleosome sliding assays were performed as described (Clapier et al., 2016) with the following modifications. Reactions were conducted in 10 mM Tris [pH 7.4], 50 mM KCl, 3 mM MgCl2, 0.1 mg/ml BSA, 1 mM ATP with 20 nM mononucleosomes at 30°C and 400 rpm shaking in a Thermomixer (Eppendorf). The amount of remodeler varied. Sliding reactions on 205 bp 601 nucleosomes used 10 nM RSC. Sliding reactions on 174 bp 5S nucleosomes used 30 nM RSC unless otherwise indicated. Aliquots were removed at each time point and reactions were stopped by adding 10 mM EDTA + 200 ng competitor DNA (Bluescript plasmid). Samples were loaded with 10% glycerol on a 4.5% (37.5:1) native polyacrylamide gel and run in 0.4x TBE for 45 min at 110 V constant. Gels were stained with ethidium bromide solution and scanned on a Typhoon Trio (Amersham, GE).

Nucleosome gel shift assays

Request a detailed protocol

Nucleosome gel shift assays were conducted similarly to the nucleosome sliding assay described above with the following modifications. Reactions were conducted in the absence of ATP and at 30°C for 20 min with 20 nM mononucleosomes. The RSC remodelers were added at 30 and 60 nM. The reactions were loaded on to 3.8% (37.5:1) native polyacrylamide gels and run in 0.4x TBE for 55 min at 110 V constant without addition of any stop solution or competitor. Gels were stained with ethidium bromide solution and scanned on a Typhoon Trio.

Histone mutant screen

Request a detailed protocol

Null mutations in rsc1 and rsc2 were crossed into the H3/H4 shuffle strain WZY42 (Zhang et al., 1998). Strains YBC1939 (WZY42), YBC2090 and YBC3040 and YBC3221 and YBC3547 were transformed in 96 well plate format with a TRP-marked histone H3-H4 or H2A-H2B mutant plasmid library respectively, obtained from and screened as described previously (Nakanishi et al., 2008). Transformants were spotted to SC-TRP plates, replica plated again to SC-TRP plates after 2 days, followed by replica plating to SC-TRP and SC-TRP + 5FOA. Synthetic lethality with mutations in the H3 αN helix were confirmed by a second round of individual transformations and shuffles.

Recombinant nucleosome mapping

Request a detailed protocol

The position and wrapping of the recombinant nucleosomes were determined by sequencing the protected DNA fragment after treating assembled nucleosomes with Exonuclease III and S1 nuclease. Approximately 400–800 fmol of nucleosomes purified from sucrose gradients was digested with a 5–25U titration of ExoIII enzyme (New England Biolabs) for 1, 2, or 3 min at 37°C in ExoIII Buffer (10 mM Tris [pH 8], 50 mM NaCl, 3 mM MgCl2). Reactions were moved to ice, S1 Buffer and NaCl were added (30 mM NaOAc [pH 4.6], 1 mM ZnOAc, 5% glycerol, and 300 mM NaCl final concentration), and treated with 50U S1 for 30 min at room temperature. Tris [pH 8.8] and EDTA were added to final concentrations of 88 mM and 14 mM, respectively, and heated to 70°C for 10 min. SDS was added to 1%, vortexed, and iced. The protected DNA fragments were cleaned up on a Qiagen MinElute PCR purification column and eluted in 30 μl EB. The level of digestion was determined for each sample on a 4.5% (37.5:1) native polyacrylamide 0.4x TBE gel as described above.

Libraries were made from the above protected fragments using the NEBNext ChIP-Seq MasterMix Set (New England Biolabs) with the following modifications. Samples did not go through the initial End-Repair. A custom adaptor with an 8 bp unique molecular identifier (UMI) was ligated onto the dA-tailed samples. The UMI Adaptor was created from an oligo based on the standard NEBNext adaptor sequence incorporating eight random nucleotides at the 5’ end of the oligo (see oligo sequence in File Supplement 2). To create the UMI Adaptor, 25 μM oligo (synthesized by IDT) was first heated to 95°C and slow cooled to room temperature in Duplex Buffer (100 mM KOAc, 30 mM HEPES [pH 7.5]). The ends of the annealed UMI Adaptor were filled in using Klenow (New England Biolabs). The reaction was stopped with EDTA and heat. The adaptor was cleaned up on a Micro Bio-Spin six chromatography column (Bio-Rad) equilibrated with water. A dT-tail was added to the adaptor with Klenow exo- (New England Biolabs). The UMI Adaptor was cleaned up on a Micro Bio-Spin six column and eluted in water. The UMI Adaptor was diluted to 1.14 μM final concentration for use in library preparations. High-throughput sequencing was performed by Illumina’s protocol for 50 bp paired-end runs on an Illumina HiSeq 2500 or a MiSeq.

The embedded UMI code was first extracted from the Fastq files using the script embedded_UMI_extractor from the package UMIScripts (https://github.com/HuntsmanCancerInstitute/UMIScripts; Parnell, 2019a). Output Fastq sequences were then aligned to an index comprised of the recombinant sea urchin 5S or 601 sequence using Novocraft Novoalign (version 3.8.2), giving the adapter sequences for trimming. After alignment, PCR-duplicate reads were identified and marked based on the UMI information with the UMISripts application bam_umi_dedup. The 5’ start positions for each alignment, discarding duplicates, were recorded as a bigWig file with 1 bp resolution using the application bam2wig from the BioToolBox package (https://github.com/tjparnell/biotoolbox; Parnell, 2020a). Separate bigWig files were generated for each length of alignments (92–169 bp). To normalize for sequencing read depth, alignment counts were scaled to an equivalent of 100K reads (calculated from the total sum of alignments without regard to alignment length). Count matrices for each length at each position on the reference sequence and for each sample were then collected using the BioToolBox application get_datasets from the bigWig files. Count matrices were analyzed with Microsoft Excel employing a 5% cutoff of the normalized fragment count.

ChIP seq and analysis

Request a detailed protocol

Yeast cultures from yHN1 (YBC1544-RSC1.9XMYC) and yHN2 (YBC1545; RSC2.9XMYC) (Ng et al., 2002) were grown in SC-TRP at 30°C with two biological replicates. Cultures were harvested at OD600 = 0.8, and crosslinked with 1% formaldehyde final for 30 min at room temperature. Cultures were quenched with 0.2M glycine for 5 min. Chromatin extracts were prepared by bead-beating, and chromatin was liberated with micrococcal nuclease. Immunoprecipitation was performed using anti-cMyc 9E11 (Abcam). DNA was isolated for input and IP samples and assembled into a library using Illumina protocols and sequenced as single end reads on an Illumina sequencer.

Fastq sequences were aligned to the yeast genome (UCSC version SacCer3) with NovoCraft Novoalign (version 3.8.2), allowing for one random alignment for multi-mapping reads. To maintain processing consistency between single-end and paired-end alignments, paired-end was aligned as single-end by ignoring the second read. Alignments were processed using the MultiRepMacsChIPSeq pipeline (version 10.1, https://github.com/HuntsmanCancerInstitute/MultiRepMacsChIPSeq; Parnell, 2020b). Since MNase-digested chromatin yields high levels of coordinate-duplicate alignments (observed mean of 65%, range 57% to 74%), duplicate alignments were randomly subsampled to a uniform rate of 40% to remove sample bias while retaining relative signal intensity. Alignments over ribosomal DNA, telomeric sequences, mitochondrial chromosome, and other high copy sequences were excluded. Fragment coverage tracks were generated by extending alignments in the 3’ direction by 160 bp in all fragments. We note that there may be some between-sample biases in the distribution of MNase-digested fragment lengths, particularly with RSC-enriched fragments, but rationalized that uniformity of processing should help minimize these biases. Replicates were depth-normalized (Reads Per Million) prior to combining as an average after confirming reasonable similarity to each through standard correlation metrics. Log2 fold enrichment of ChIP signal over non-enriched nucleosome signal (Input) was generated with Macs2 (Zhang et al., 2008) without background lambda correction. Hmo1 ChIPs (GSM1509041) (Knight et al., 2014) were re-processed in a similar manner.

NDRs were determined by first determining all nucleosome positions in the genome using the software package biotoolbox-nucleosome (https://github.com/tjparnell/biotoolbox-nucleosome; Parnell, 2019b) with the Input MNase sample alignments. To reduce mapping noise and increase efficacy of calling positions, ‘skinny’ nucleosome coverage was used rather than midpoint data. Fragment coverage was generated by shifting the 5’ alignment coordinate in the 3’ direction by 37 bp and extending 74 bp, essentially recording the predicted central portion of the nucleosome. After mapping nucleosomes, NDRs were extracted by calculating all inter-nucleosomal intervals and selecting for those with lengths between 75 and 600 bp and occurring over or adjacent to a defined protein-coding gene TSS. NDRs were sorted by decreasing length and aligned by the edge closest to the TSS; NDRs between divergent promoter pairs are represented once.

Data were collected with applications from the BioToolBox package (https://github.com/tjparnell/biotoolbox; Parnell, 2020a). Mean occupancy of RSC and Hmo1 were collected over the NDRs using the application get_datasets, while spatial data surrounding NDRs and tDNAs were collected with get_relative_data. Heat maps were generated in R (R Development Core Team, 2018) with pHeatmap (Kolde, 2019). Dot plots and violin plots were generated with GraphPad Prism.

RNA seq and analysis

Request a detailed protocol

Yeast cultures for RNASeq were grown in SD media supplemented for auxotrophic amino acids at 30°C in three biological replicates. RNA from logarithmically growing cells was purified using Ambion Ribo-pure yeast kit. Samples were additionally DNAse-treated and cleaned up with Qiagen RNeasy kit. Illumina Ribo Zero yeast kit was used for library preparation, and sequencing performed on an Illumina sequencer.

Fastq reads were processed following an RNASeq pipeline (https://github.com/HuntsmanCancerInstitute/hciR; Stubben, 2020). Briefly, reads were aligned with STAR (version 2.5.4, Dobin et al., 2013, counts obtained with Subread featureCounts (version 1.6.3) based on Ensembl annotation (release 90), and differential gene expression performed with DESeq2 (version 1.21). Dot plots and violin plots were generated with GraphPad Prism (version 8.0.2).

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
    Pretty Heatmaps
    1. R Kolde
    (2019)
    Pretty Heatmaps.
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52
  53. 53
  54. 54
  55. 55
  56. 56
  57. 57
  58. 58
  59. 59
    R: A Language and Environment for Statistical Computing
    1. R Development Core Team
    (2018)
    R Foundation for Statistical Computing, Vienna, Austria.
  60. 60
  61. 61
  62. 62
  63. 63
  64. 64
  65. 65
  66. 66
  67. 67
  68. 68
  69. 69
  70. 70
  71. 71
  72. 72
  73. 73
  74. 74
  75. 75
  76. 76
  77. 77
  78. 78
  79. 79
  80. 80
  81. 81

Decision letter

  1. Stephen Buratowski
    Reviewing Editor; Harvard Medical School, United States
  2. Kevin Struhl
    Senior Editor; Harvard Medical School, United States
  3. Blaine Bartholomew
    Reviewer; University of Texas M.D. Anderson Cancer Center, United States

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

Acceptance summary:

It is important to understand the differences between closely related chromatin remodeling complexes, and your studies of the two yeast RSCs provides interesting insights into this question. The combination of in vivo and in vitro studies used here make for a strong story.

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 your work entitled "Specialization of the chromatin remodeler RSC to mobilize partially-unwrapped nucleosomes" for consideration by eLife. Your article has been reviewed by three peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by a Reviewing Editor (me) and a Senior Editor (Kevin Struhl). The following individual involved in review of your submission has agreed to reveal their identity: Blaine Bartholomew (Reviewer #1).

Based on the individual reviews (below) and then further consultations, we regret to inform you that your paper was not accepted for publication in eLife. Everybody appreciated the work that went into isolating and analyzing the different RSC variants. There were some concerns about interpretation and quantitation of the in vitro remodeling experiments, although those concerns might be addressed with a little additional work. However, the main criticism was voiced by reviewer 2, who felt that there was a disconnect between the in vitro and in vivo experiments, and therefore that the conclusion that Rsc1 was specialized for partially unwrapped nucleosomes was not strongly justified. After consultation between the reviewers and editors, we felt that these issues would be difficult to address within the window that eLife sets for revisions. We hope that these reviews will help you with a future version of the paper.

Reviewer #1:

The authors have carefully characterized the functional differences between the Rsc1 and Rsc2 forms of the RSC ATP-dependent chromatin remodeler and their interplay with Rsc3/Rsc30, two DNA binding subunits of the RSC complex. They found the key difference to be in the two repeats present at the C-terminus of Rsc1 and Rsc2 with the repeat closest to the end helping to regulate the association of Rsc3/30 to the complex. The remodeling differences were only observed between the RSC1 and RSC2 complexes with 5S DNA nucleosomes and not with 601, which suggested the difference might reside in the ability to handle nucleosomes that could be partially unwrapped from the beginning. The combination of biochemistry and genetic experiments are powerful and complementary in this study. However, some of the biochemical experiments were not as well designed as needed to effectively compare the enzymatic differences between the two enzymes or to ascertain as to the experimental reproducibility. The quantitation of nucleosome remodeling assays like that shown in Figure 2C with their corresponding replicates is not displayed and would be clearer if plotted in a line-graph with standard deviation shown for each time point rather than displayed for only of the replicates as a histogram. As it is currently, it is not possible to know what the deviation was like between the replicates. It also seems that they if properly quantitated the authors would also be able to determine the rates of nucleosome movement. The ATPase assays also don't seem to be as effective for comparing difference as they only measure Vmax and measure the extent of ATPase hydrolysis at only one time point with a high concentration of ATP. In their approach it would be easy to miss if there were indeed differences in their ability to hydrolyze ATP. The binding assays are also deficient in that there are only two titration points, there is no quantitation of this data and thus it is also not possible to ascertain as to the standard deviation between the technical replicates similar to the remodeling assays. I was also concerned since remodeling with the 5S DNA nucleosome seems to be generally less efficient than expected and raises questions as to why this might be. Could the linker DNA be too short in the 5S nucleosomes for the Rsc3/30 to bind and be the reason why the addition of Rsc3/30 is inhibitory when based on the genetic data is expected to stimulate the activity of the complex?

They use a powerful approach with a series of histone point mutations to identify histone region(s) that might contribute to the enzymatic differences between the RSC1 and RSC complexes and observed that mutations in the H3 α N helix are synthetically lethal when only RSC2 was present. These are very complementary observations and provide an orthogonal approach to indicate it is the high propensity of unwrapping DNA from the end of the nucleosome that is crucial for these differences. I did however find it odd that in the context of the 5S nucleosomes that H3 R40A mutation appears to weaken these interactions (Figure 3—figure supplement 3); whereas in the context of the 601 nucleosome the same mutation appears to stabilize these interactions over wild type (Figure 4—figure supplement 2 panel B). Does this mean that the destabilizing effect observed with H3 R40A is nucleosome specific and may vary depending on the DNA sequence and would this complicate the authors interpretation? The experiments in Figure 3E indicate that both RSC1 and RSC2 are inhibited by DNA unwrapping from the nucleosome surface, but to different extents. Is it a problem that DNA unwrapping is inhibitory for both forms of RSC and wouldn't it change some of the later interpretations?

The experimental setup in which Hmo1 is added is not adequate to effectively compare how much Hmo1 stimulates the RSC1 versus the RSC2 complex because only 1 time point is used for both enzymes in the remodeling assays. In essence remodeling with RSC1 is much further along because of the time window used, while RSC2 is not and as such you are not able to determine the difference in the stimulation of remodeling when Hmo1 is added. Also, I wasn't clear how an HMG-like protein would help stabilize the wrapping of DNA onto the nucleosome?

The MNase ChIP-seq was a good addition as it helped re-examined the potential differences in the genome-wide Rsc1 and Rsc2 localization and the difference is evident at the tRNA genes and ribosomal protein genes. The RNA-seq also aids in this regard and combined with the other experiments provides a compelling story.

Reviewer #2:

In this manuscript, Cairns and colleagues investigate the potential for in vitro and in vivo distinctions between two forms of the RSC remodeling enzyme that contain either the Rsc1 or Rsc2 subunit. The first half of the manuscript describes extensive biochemical studies with mononucleosomes, comparing the activities of RSC1 and RSC2 complexes. The authors report that the RSC1 complex remodels a 5S nucleosome substrate more efficiently than the RSC2 complex, and that the RSC2 complex has greater problems remodeling a 5S nucleosome harboring a H3-R40A substitution derivative. H4-R40 is located in the H3 N-helix, and the authors find nice synthetic lethality phenotypes between many substitution derivatives in the H3-N helix and a rsc1 deletion strain. Together the data suggest that RSC2 may have a general problem (compared to RSC1) remodeling 5S nucleosomes with alterations in the H3-N helix. Previous work from the Owen-Hughes group has reported that these H3-N helix substitutions lead to increased breathing (spontaneous unwrapping) of entry/exit DNA and increased nucleosome thermal mobility. The authors suggest that RSC1 may be specialized to remodel partially unwrapped nucleosomes. Consistent with this view, they how that RSC1 and RSC2 are proficient at remodeling a more stably wrapped, 601 nucleosome. These in vitro studies lead the authors to predict that RSC1 may play a larger function at genes where RSC is known to engage partially unwrapped nucleosomes, specifically the ribosomal protein (RP) genes. To this end, the carry out extensive ChIP-seq analyzes of Rsc1 and Rsc2, as well as RNA-seq analyzes in rsc1 and rsc2 deletion strains. Contrary to the model, these data reinforce the view that Rsc1 and Rsc2 are functionally redundant at the majority of target genes, including RP genes. The only large different is that Rsc1 appears to play a dominant role at rDNAs. Notably, rDNAs are not known to contain alternative nucleosome structures that would link these functional data with the biochemistry. In general, this work illustrates differences between RSC1 and RSC2 complexes for remodeling a 5S mononucleosome substrate (but not a 601), distinct genetic interactions with histone substitutions, but no clear functional correlation of these results with gene targets or unwrapped nucleosomes in vivo.

1) One major concern with the biochemistry data is the remodeling results with the 5S substrates. This substrate appears to be a center-positioned nucleosome (10N16), but after remodeling, the "slid" product migrates much slower on the native gels. This is counter to every other published sliding assay for RSC or SWI/SNF-like enzymes. These enzymes always move nucleosomes towards or off DNA ends. These always lead to faster migrating species on native gels. For examples, see work from Owen-Hughes with the MMTV A nucleosome (Ferreira et al., 2007) or the 601 nucleosome shown here. Given the mobility of this remodeled species, it is more likely that this product represents a histone dimer loss event, leading to a hexosome or tetrasome. Alternatively, perhaps it is some type of product where the nucleosome is slid off the end and free DNA has re-captured an exposed histone surface (Kassabov et al., 2003). It is unclear why the authors see such a product specifically with a 5S nucleosome, as faster migrating species have been observed with this element previously (Jaskeliof et al., 2000), and early DNase I footprinting assays using the same template shown here gives rise to a "naked DNA" pattern after RSC remodeling, consistent with the nucleosome moved off the DNA ends (Cairns et al., 1996). The identity of this novel species is important for interpreting the data for how histone substitutions or Rsc3/Rsc30 impacts this novel reaction. It also leads to worries that this particular substrate is not representative of typical remodeling events.

2) The authors interpret the impact of H3-N helix substitutions as due to their impact on nucleosome unwrapping. They also argue that 5S is distinguished from 601, based on unwrapping dynamics. It is important to realize that the unwrapping phenotypes were determined by the Owen-Hughes lab using an MMTV A nucleosome, and in some cases also a 601. Note that a H3-K56Q substitution has a large impact on unwrapping a 601 (Neumann et al., 2009). Furthermore, on an MMTV A nucleosome (and/or a 601 nuc), these substitutions stimulate the remodeling activity of the RSC2 complex. This is certainly true for H3-R40A and for H3-K56Q. Indeed, the Owen-Hughes group has argued that RSC2 activity is stimulated by histone substitution derivatives that enhance unwrapping and intrinsic thermal mobility. This is completely opposite of the conclusions and models described here.

3) Figure 4C. This is not an optimal experiment to measure interactions between Hmo1 and RSC in vivo. Sonication of chromatin may yield an average size of ~250-500bp, but the distributions are normally large. This experiment should at least be performed with Mnase trimmed cores. Most likely this will not work, as previous work has shown that Hmo1 needs linker DNA to bind to mononucleosomes (Hepp et al., 2014).

4) Figure 3—figure supplement 3. These mapping data for the 5S mononucleosomes seem odd. Historically, the 5S rRNA gene is not considered a perfect nucleosome positioning sequence. Previous mapping data has indicated that the major translational position only reflects 50% of the population, with the remaining nucleosomes offset by 10 bp intervals. It is surprising then that the mapping work yields a unique right boundary at position 158. Could the "unwrapped" left boundaries really be alternate translational frames, and the right boundaries that might be positioned at the full-length ends (position 174) were not included? It also seems odd that only the left end is "unwrapped" by the H3-R40A substitution.

5) The general interest of this work relies on what is stated in the title that RSC1 is specialized to remodel partially unwrapped nucleosomes and such nucleosome remodeling specialization is important for a particular gene target, such as RP genes. Unfortunately, the in vivo data do not support this model. All of the results show that Rsc1 and Rsc2 share a nearly identical binding profile (with the exception of tDNA), and single deletion strains have nearly no impact on RNA levels, especially RP genes. Furthermore, it is very odd that the only mutant that impacts RP gene expression is the rsc2 rsc30 double mutant. In contrast, the rsc1 rsc30 has little impact. In a simple interpretation, this suggests that Rsc2 is sufficient to support RP gene expression, even without Rsc30. This is not consistent with the main conclusions of the biochemistry or the title.

Reviewer #3:

In this paper, Schlichter et al., compare the Rsc1 and Rsc2 versions of RSC remodeling complex. They find that Rsc2 complexes have reduced affinity for Rsc3/30, and map this difference to the CT2 domain. While the four different complexes (Rsc1 or Rsc2, +/- Rsc3/30) have similar affinity for nucleosomes and ATPase activity, the Rsc1 complexes have stronger nucleosome remodeling activity. They screened histone mutant libraries and found specific synthetic lethality with rsc1 δ in H3 alphaN domain or H3 K56Q (but not A) mutants. These differences correlate with partial nucleosome unwrapping (citing an earlier FRET study). Interestingly, the H3 alphaN synthetic lethality is suppressed if Rsc1 CT2 domain swapped into Rsc2. Using in vitro assays and ChIP experiments the authors build a case that Rsc1 functions preferentially at promoters with wide NDRs (particularly ribosomal protein genes and tDNAs), where partially unwrapped nucleosomes may be more prevalent. Overall, the combination of in vitro biochemistry and correlational in vivo work seems persuasive. The paper is well written and the data relatively clear. Understanding the functional differences between the closely related RSC complexes (or other remodeler families) is of significant interest to the field.

Some specific questions and comments:

1) Does swapping the CT2 domain of Rsc1 into Rsc2 also suppress the synthetic lethality with K56Q?

2) Subsection “RSC2 complexes are deficient in remodeling partially-unwrapped nucleosomes”: The authors conclude that Rsc1 complex is more active for remodeling unwrapped nucleosomes because it has more activity on the R40A nucleosomes than Rsc2. But that is also true for WT 5S nucleosomes. Given that R40A is unlikely to completely disrupt wrapping, isn't it possible that remodeling only occurs when the nucleosomes are wrapped, and R40A simply reduces the time spent in the wrapped conformation? This idea seems supported by the fact that R40A has less effect in the 601 DNA sequence. Following this line of reasoning, could Rsc1 do a better job in inducing unwrapped nucleosomes to adopt the wrapped "remodelable" configuration? The discussion seems to be moving in this direction, but it this is the authors' model it could be made clearer and perhaps foreshadowed more strongly in the results.

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

Thank you for submitting your article "Specialization of the Chromatin Remodeler RSC to Mobilize Partially-Unwrapped Nucleosomes" for consideration by eLife. Your article has been reviewed by three peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Kevin Struhl as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Blaine Bartholomew (Reviewer #3).

The reviewers appreciated the new experiments and additional statistical analysis done in response to their comments. I'm happy to say they all agree that the revised paper will be appropriate for eLife with just a few additional changes to the text. This letter is to help you prepare a revised submission.

The reviewers state that this paper has several key results that are solid and significant:

1) The RSC1 enzyme has evolved to use the Rsc3/30 module, and this module facilitates expression of RP genes and helps to counteract histone substitution alleles. Surprisingly, this module has an inhibitory impact in vitro on both enzymes.

2) Hmo1 functions together with RSC1 and RSC2 (though more so with RSC2) to facilitate nucleosome sliding in vitro and gene expression.

3) Rsc1 and Rsc2 show distinct genetic interactions with histone residues.

4) The functional interactions between Rsc1, Rsc2, Hmo1, and Rsc3/30 are enlightening.

5) RSC1 remodels a 5S nucleosome much better than RSC2, though they show more equivalent activity on a 601 nucleosome.

However, the one concern that remains is whether the paper over-simplifies things by making the strong, generalized conclusion (stated in the title and at various points in the paper) that the differential activity of RSC1 and RSC2 is linked to partial unwrapping of the nucleosome. Although the genetic evidence is consistent with your model, reviewer 2 (supported by reviewer 3) cites published data saying the difference in 5S versus 601 unwrapping is not as clear as suggested in your text:

"Introduction. The authors cite 4 papers in support of the statement that "entry/exit DNA displays higher level of detachment from the octamer" (for 5S in comparison to 601). However, none of these papers compare 5S to 601. Polach and Widom, 1995 used 5S, and Li and Widom, 2004 used 601. In the cited review (Zhou et al., 2019), they discuss in detail the comprehensive study by the Poirier group (North et al., 2012; not cited) where they do directly compare 5S and 601 nucleosomes in FRET-based unwrapping assays. Contrary to expectations, a 5S nucleosome unwraps less than 601 at the edge. This is shown to be due to the first 7bp of the 5S sequence at the entry/exit site. Thus, even though 601 has a higher overall affinity, the unwrapping dynamics are higher than 5S. These data are also consistent with published work showing that the 5S sequence has a higher affinity for H2A/H2B dimers (cited in Zhou et al., 2019). Notably, the higher affinity of 601 for octamers is due to AT repeats surrounding the dyad (and contacting the H3/H4 tetramer), not really the edges. As discussed by the authors, differences in how they interpret their data and other work (e.g., older Owen-Hughes work on α N-helix substitutions) may be due to the differences in histone octamer source – yeast vs Xenopus. Perhaps yeast dimers have a higher affinity for 601, rather than 5S. However, without extensive discussion, the authors' current conclusions seem to contradict several previously published papers. In general, the remodeling assays don't support a simple model where RSC1 remodels unwrapped nucleosomes better than RSC2. Note that I am not questioning the data at all – they are solid and very well done. I worry that the conclusions may just not be so simple."

In your revised manuscript, please address this point. I took a look at the paper cited by the reviewer (North et al., 10215) and it does seem to claim that 601 unwraps its ends more that 5S. If the reviewer's summary of the literature is accurate, please modify your statements (and model Figure 6) concerning 5S versus 601 accordingly, and factor that into your discussion and interpretations. If you feel those results can be reconciled, or are not applicable, to your model, please explain why in your Discussion section.

https://doi.org/10.7554/eLife.58130.sa1

Author response

[Editors’ note: the authors resubmitted a revised version of the paper for consideration. What follows is the authors’ response to the first round of review.]

We carefully considered the reviewer comments and your editorial comments – and over the last several months we successfully addressed the vast majority, including those that were indicated as the most important. These efforts have substantially improved our study.

Reviewer #1:

The authors have carefully characterized the functional differences between the Rsc1 and Rsc2 forms of the RSC ATP-dependent chromatin remodeler and their interplay with Rsc3/Rsc30, two DNA binding subunits of the RSC complex. They found the key difference to be in the two repeats present at the C-terminus of Rsc1 and Rsc2 with the repeat closest to the end helping to regulate the association of Rsc3/30 to the complex. The remodeling differences were only observed between the RSC1 and RSC2 complexes with 5S DNA nucleosomes and not with 601, which suggested the difference might reside in the ability to handle nucleosomes that could be partially unwrapped from the beginning. The combination of biochemistry and genetic experiments are powerful and complementary in this study. However, some of the biochemical experiments were not as well designed as needed to effectively compare the enzymatic differences between the two enzymes or to ascertain as to the experimental reproducibility. The quantitation of nucleosome remodeling assays like that shown in Figure 2C with their corresponding replicates is not displayed and would be clearer if plotted in a line-graph with standard deviation shown for each time point rather than displayed for only of the replicates as a histogram. As it is currently, it is not possible to know what the deviation was like between the replicates. It also seems that they if properly quantitated the authors would also be able to determine the rates of nucleosome movement.

We agree that additional replicates and display modes would improve the manuscript. We have quantified replicates of the sliding assays as the reviewer recommends and these are now included as additional supplemental figures for each assay. Figure 2—figure supplement 1, Figure 3—figure supplement 3, Figure 4—figure supplement 2, Figure 4—figure supplement 3B, Figure 4—figure supplement 4, and Figure 4—figure supplement 6.

The ATPase assays also don't seem to be as effective for comparing difference as they only measure Vmax and measure the extent of ATPase hydrolysis at only one time point with a high concentration of ATP. In their approach it would be easy to miss if there were indeed differences in their ability to hydrolyze ATP.

We have replaced the ATPase bar graph with an ATPase time course in Figure 2B.

The binding assays are also deficient in that there are only two titration points, there is no quantitation of this data and thus it is also not possible to ascertain as to the standard deviation between the technical replicates similar to the remodeling assays.

We observe strong differences in RSC remodeling between wt 5S and H3R40A 5S yNucs using nuc:RSC ratios of 1:1.5. Our nucleosome binding assays were conducted at nuc:RSC ratios of 1:1.5, as well as 1:3. Since the nuc binding assay utilizes the conditions used in the remodeling assay, this indicates that RSC1 and RSC2 bind to wt and R40A 5S yeast nucleosomes similarly under these conditions. Additional titration points would require increasing the amount of RSC complex in the assay relative to the amount of nucleosome. Currently this would be challenging because the remodeler concentration required for these higher titration points would be difficult to obtain while remaining within assay parameters for a crisp RSC:Nuc gelshift band. In addition, the higher remodeler concentration would be needed for all four RSC complexes, which would require four new preps which are not feasible to produce at this time. However, we have included the technical replicates for the current experiments as support in the Supplementary Figure.

I was also concerned since remodeling with the 5S DNA nucleosome seems to be generally less efficient than expected and raises questions as to why this might be. Could the linker DNA be too short in the 5S nucleosomes for the Rsc3/30 to bind and be the reason why the addition of Rsc3/30 is inhibitory when based on the genetic data is expected to stimulate the activity of the complex?

We note that we do conduct our work under catalytic conditions, with excess substrate. Regarding length, we have lengthened 5S NPS to 198bp with similar, though more complicated, remodeling results because this NPS resulted in 2 different nucleosome ‘slid’ positions. However, as there was no striking difference with the Rsc3/30 in vitro effect (same result as with shorter DNA), we have not presented experiments using this longer NPS in the paper. Furthermore, it is clear RSC sliding is much more efficient on the 5S nucleosome in the presence of Hmo1, which we discuss in depth in the paper.

They use a powerful approach with a series of histone point mutations to identify histone region(s) that might contribute to the enzymatic differences between the RSC1 and RSC complexes and observed that mutations in the H3 α N helix are synthetically lethal when only RSC2 was present. These are very complementary observations and provide an orthogonal approach to indicate it is the high propensity of unwrapping DNA from the end of the nucleosome that is crucial for these differences. I did however find it odd that in the context of the 5S nucleosomes that H3 R40A mutation appears to weaken these interactions (Figure 3—figure supplement 3); whereas in the context of the 601 nucleosome the same mutation appears to stabilize these interactions over wild type (Figure 4 —figure supplement 2 panel B). Does this mean that the destabilizing effect observed with H3 R40A is nucleosome specific and may vary depending on the DNA sequence and would this complicate the authors interpretation?

Indeed, the H3 R40A mutation causes further unwrapping of the 5S nucleosome, whereas it has little effect on a 601 nucleosome. We agree that the destabilizing effect is likely nucleosome specific depending on DNA sequence, and that the effect is greater with a DNA sequence that has a lesser affinity for the octamer. We would expect the greatest effect on further destabilization to be at regions with sequences that are inherently partially unwrapped (similar to 5S) such as those found at RPGs and other wide NDRs (“fragile/MNase sensitive nucleosomes”).

We note that reviewer 3’s comments (see below) for grouping the wrapped and unwrapped products together and listing percentages for the individual species was indeed helpful, will assist in the interpretation here. In keeping, we have made these changes to Figure 3—figure supplement 5 and Figure 4—figure supplement 5.

The experiments in Figure 3E indicate that both RSC1 and RSC2 are inhibited by DNA unwrapping from the nucleosome surface, but to different extents. Is it a problem that DNA unwrapping is inhibitory for both forms of RSC and wouldn't it change some of the later interpretations?

The reviewer is correct, and we needed to make this point more clearly in our revision, as two reviewers had questions here. To clarify, our data and model are indeed that wrapped nucleosomes are better substrates for both RSC1 and RSC2 complexes, but RSC1 can remodel the unwrapped version better than RSC2 – especially when unaided by other proteins. We have now stated this clearly in two locations in the revision – both in the Results section and in the Discussion section. Furthermore, our data supports RSC complex as two redundant functional entities that manage partially-wrapped nucleosomes: ‘Rsc2 with Hmo1’ and ‘Rsc1 with Rsc3-30’ – and one must significantly impair both entities for a strong phenotype (see below).

The experimental setup in which Hmo1 is added is not adequate to effectively compare how much Hmo1 stimulates the RSC1 versus the RSC2 complex because only 1 time point is used for both enzymes in the remodeling assays. In essence remodeling with RSC1 is much further along because of the time window used, while RSC2 is not and as such you are not able to determine the difference in the stimulation of remodeling when Hmo1 is added.

We agree, and for this reason different time point and remodeler concentrations are shown in supplemental Figure 4—figure supplement 3, so that the “fast” remodelers are tested when they aren’t “further along.” These results reinforce our conclusions.

Also, I wasn't clear how an HMG-like protein would help stabilize the wrapping of DNA onto the nucleosome?

Hmo1 has been shown to stabilize chromatin and perform the functions of a linker histone. Unlike other HMGB proteins which have an acidic CTD that promotes bending, Hmo1 is relatively unique in this class as it contains a basic lysine rich C-terminal extension. Hmo1 is predicted to bind near the nucleosome dyad and use its basic extension to bind linker DNA and prevent bending and improve wrapping. Additionally, nucleosomes in an hmo1 null display increased MNase sensitivity. (Panday and Grove, 2016; Panday and Grove, 2017). We have incorporated additional text in the manuscript to convey this to the reader.

The MNase ChIP-seq was a good addition as it helped re-examined the potential differences in the genome-wide Rsc1 and Rsc2 localization and the difference is evident at the tRNA genes and ribosomal protein genes. The RNA-seq also aids in this regard and combined with the other experiments provides a compelling story.

We thank the reviewer for appreciating these experiments.

Reviewer #2:

[…] 1) One major concern with the biochemistry data is the remodeling results with the 5S substrates. This substrate appears to be a center-positioned nucleosome (10N16), but after remodeling, the "slid" product migrates much slower on the native gels. This is counter to every other published sliding assay for RSC or SWI/SNF-like enzymes. These enzymes always move nucleosomes towards or off DNA ends. These always lead to faster migrating species on native gels. For examples, see work from Owen-Hughes with the MMTV A nucleosome (Ferreira et al., 2007) or the 601 nucleosome shown here. Given the mobility of this remodeled species, it is more likely that this product represents a histone dimer loss event, leading to a hexosome or tetrasome. Alternatively, perhaps it is some type of product where the nucleosome is slid off the end and free DNA has re-captured an exposed histone surface (Kassabov et al., 2003). It is unclear why the authors see such a product specifically with a 5S nucleosome, as faster migrating species have been observed with this element previously (Jaskeliof et al., 2000), and early DNase I footprinting assays using the same template shown here gives rise to a "naked DNA" pattern after RSC remodeling, consistent with the nucleosome moved off the DNA ends (Cairns et al., 1996). The identity of this novel species is important for interpreting the data for how histone substitutions or Rsc3/Rsc30 impacts this novel reaction. It also leads to worries that this particular substrate is not representative of typical remodeling events.

This is a highly relevant point, and we thus thought it important to quantitatively assess whether the slower migrating species in the sliding assay could be due to dimer loss during RSC remodeling. We addressed this issue directly by examining 5S sliding on fluorescently-labeled yeast nucleosomes. Here, we assembled 174bp 5S yeast nucleosomes labeled with Oregon Green at Q114C of H2A, and used them in RSC sliding reactions. RSC1 complex remodeled these labeled nucleosomes resulting in a slower migrating band, as we have seen with remodeled unlabeled 5S nucleosomes. If H2A/H2B dimer was lost from the slid product during remodeling, then we would expect a reduction in the dimer to DNA ratio when normalized to the ratio observed in the starting nucleosomes. However, the dimer to DNA ratio does not decrease in the slid product, and instead is very similar to the starting nucleosomes. This data has been added to the paper as Figure 2—figure supplement 3.

The reviewer states that all nucleosomes positioned near or off of DNA ends will always lead to faster, not slower, migrating species on native gels. A complex’s migration through a native gel is affected by its overall shape as well as charge. It is known that different DNA sequences have different degrees of inherent stiffness and flexibility. It seems reasonable that some DNA sequences protruding from a nucleosome could alter its migration on a native gel. This would be a property of the specific NPS and nucleosomes. For example, Chakravarthy et al., (2012) (Bowman lab) reported remodeled nucleosomes that ran aberrantly fast on native gels which showed a smaller effective size for the particle and greater compaction.

After examining, we did not consider the mononucleosome sliding results of the 5S template used in the Jaskelioff, 2000 paper as comparable to our 5S mononucleosome sliding. The Jaskelioff template is 416 bp with 2 copies of the 5S NPS (compatible with di-nucleosome formation and examination) while ours is a 174 bp fragment with a single 5S NPS, for mononucleosome formation and examination. Thus, it seems entirely possible that these two templates would migrate differently on native gels before or after remodeling.

Although we believe our Oregon Green labeling has resolved the H2A/B dimer issue for our experiments, we will mention that some hexasomes migrate slower than nucleosomes (Chen et al., 2017), while others run faster (Levendosky et al., 2016, Brehove et al., 2019, Mazurkiewicz et al., 2006). Indeed, the Mazurkiewicz, 2006 paper shows Sea Urchin 5S hexasomes and nucleosomes on 146 bp and 207 bp NPS fragments and their experiments show hexasomes running faster than the corresponding nucleosome.

Regarding the early 5S DNase I footprinting after RSC remodeling (Cairns, 1996), where the product resembles naked DNA. I agree with the reviewer that this observation is consistent with dimer loss – reinforcing the need for conducting the Oregon Green experiment. As the dimers are shown to remain, I believe that RSC has moved these nucleosomes out of their initial/preferred rotational phase to a superposition of all rotational phases, and thus the DNaseI patterns will resemble naked DNA – with the fraction of end position nucleosomes lacking (as the reviewer notes) octamer protection.

2) The authors interpret the impact of H3-N helix substitutions as due to their impact on nucleosome unwrapping. They also argue that 5S is distinguished from 601, based on unwrapping dynamics. It is important to realize that the unwrapping phenotypes were determined by the Owen-Hughes lab using an MMTV A nucleosome, and in some cases also a 601. Note that a H3-K56Q substitution has a large impact on unwrapping a 601 (Neumann et al., 2009). Furthermore, on an MMTV A nucleosome (and/or a 601 nuc), these substitutions stimulate the remodeling activity of the RSC2 complex. This is certainly true for H3-R40A and for H3-K56Q. Indeed, the Owen-Hughes group has argued that RSC2 activity is stimulated by histone substitution derivatives that enhance unwrapping and intrinsic thermal mobility. This is completely opposite of the conclusions and models described here.

There are indeed apparent differences with prior results, but we believe they can be reconciled. First, the type of NPS sequence and DNA length are important factors for nucleosome wrapping and H3-α N helix mutations may behave differently in the context of the Sea Urchin 5S nucleosomes examined here than in the context of other NPS previously studied. More to the point, we consider it reasonable and likely that histone origin (e.g., yeast, Drosophila, Xenopus, or chicken) influences nucleosome wrapping, and underlies differences in observations. Our experiments utilize yeast remodeler with yeast octamers (nucleosomes), whereas the papers cited (and the majority in the field) use a variety of octamer sources for the in vitro work, but rarely have people used yeast remodelers with yeast nucleosomes. Our lab has observed differences in remodeler activities based on octamer source, and therefore we elected to use only yeast octamers to test yeast RSC derivatives, as we believe this provides a higher degree of confidence in the results.

We mapped the positioning of the wild-type and mutant recombinant nucleosomes used in this study, which supports the hypothesis that canonical Sea Urchin 5S sequence with yeast nucleosomes are less well wrapped than original Widom 601 yeast nucleosomes, and that RSC1 and RSC2 behave differently on these two templates. The H3 R40A mutation was not sufficient to unwrap the yeast Widom 601 nucleosome nor did it affect RSC sliding on yeast 601 nucs in our experiments.

Importantly, while the Owen-Hughes group does indeed show an increase in the initial rate of repositioning (1.2-2x -measured as the sum of all remodeled products) by RSC for the same H3 mutations where we see lethality with rsc1∆ (e.g., H3R40A; subsection “Well-wrapped nucleosomes are remodeled comparably by both RSC1 and RSC2 complexes”). Importantly, they also show that when these mutant nucleosomes are remodeled by RSC2 and run on a native polyacrylamide gel, they show significantly altered products compared to WT – so the distribution of products is highly affected. So, while there may be aspects of remodeling that are enhanced, proper remodeling to completion appears impaired (see subsection “Well-wrapped nucleosomes are remodeled comparably by both RSC1 and RSC2 complexes”). This improper remodeling by RSC2 of R40A is therefore supported by our in vivo genetic studies.

3) Figure 4C. This is not an optimal experiment to measure interactions between Hmo1 and RSC in vivo. Sonication of chromatin may yield an average size of ~250-500bp, but the distributions are normally large. This experiment should at least be performed with Mnase trimmed cores. Most likely this will not work, as previous work has shown that Hmo1 needs linker DNA to bind to mononucleosomes (Hepp et al., 2014).

This is a good point, and our DNA fragment size following sonication using the Bioruptor Pico is predominantly 150-300 bp. In addition, we have now repeated the experiment from crosslinked MNase-treated chromatin extracts as suggested by the reviewer and still see co-immunoprecipitation. This is included as Figure 4—figure supplement 1.

4) Figure 3—figure supplement 3. These mapping data for the 5S mononucleosomes seem odd. Historically, the 5S rRNA gene is not considered a perfect nucleosome positioning sequence. Previous mapping data has indicated that the major translational position only reflects 50% of the population, with the remaining nucleosomes offset by 10 bp intervals. It is surprising then that the mapping work yields a unique right boundary at position 158. Could the "unwrapped" left boundaries really be alternate translational frames, and the right boundaries that might be positioned at the full-length ends (position 174) were not included? It also seems odd that only the left end is "unwrapped" by the H3-R40A substitution.

We used paired-end sequencing after ExoIII and S1 digestion to map the recombinant yeast nucleosomes and agree the 5S is not a perfect NPS. Paired-end sequencing gives fragment length, start and end points for each individual fragment, and number of reads indicating strength of fragment representation, which we provide. We have improved the display of our mapping data as suggested by reviewer 3. Our wild type 5S nucleosome mapping is in agreement with the asymmetric unwrapping previously shown for the 5S NPS by Winogradoff and Aksimentiev, 2019).

5) The general interest of this work relies on what is stated in the title that RSC1 is specialized to remodel partially unwrapped nucleosomes and such nucleosome remodeling specialization is important for a particular gene target, such as RP genes. Unfortunately, the in vivo data do not support this model. All of the results show that Rsc1 and Rsc2 share a nearly identical binding profile (with the exception of tDNA), and single deletion strains have nearly no impact on RNA levels, especially RP genes. Furthermore, it is very odd that the only mutant that impacts RP gene expression is the rsc2 rsc30 double mutant. In contrast, the rsc1 rsc30 has little impact. In a simple interpretation, this suggests that Rsc2 is sufficient to support RP gene expression, even without Rsc30. This is not consistent with the main conclusions of the biochemistry or the title.

I believe these comments are a result of our inadequate description of our interpretations and model. First, we agree that (beyond tDNAs) the binding profiles of Rsc1 and Rsc2 are nearly identical. That fact was stated in the manuscript and consistent with our model, but we did not do an adequate job stating the model. To rectify, we now have sections in the expanded Discussion section on the issue of how RSC1 complexes better manage partially-unwrapped nucleosomes after (independent of) the initial nucleosome binding step. Therefore, if the initial binding is not affected, then the genomics should show them both occupying the same Pol II genes.

Second, we clarify in key locations how the in vitro and in vivo data are largely consistent, and that although Rsc1/2 are at the same locations, they can have partially overlapping/redundant, and partially unique functions – summarized in Figure 5—figure supplement 2 and in the Model figure (Figure 6). Put succinctly, our data supports RSC as two partially redundant functional entities: ‘Rsc2 with Hmo1’ and ‘Rsc1 with Rsc3-30’ – and one must significantly impair both entities for a strong phenotype in vivo (see Figure 5—figure supplement 2C). We note that in vitro tests on particular nucleosome substrates cannot encompass the complexity of genome sequences that RSC manages. However, there is general agreement in the data to support differences in the ability of RSC1 versus RSC2 to manage partially-unwrapped nucleosomes – as the best current model to explain our genetic, biochemical and genomic data.

Regarding RP gene expression in particular, we find that yeast prior to the whole genome duplication do not have Rsc1, and as neither rsc1 nor rsc30 nulls are inviable, RSC2 (with Hmo1) must be sufficient to support RP gene expression. The rsc1, rsc2, and rsc1 rsc30 mutations have mild effects because of this built in redundancy that Rsc2/Hmo1 can perform the role when RSC1 is absent. In the rsc2 rsc30 mutation, RSC2 complex is absent, and the RSC1 complex is now impaired due to loss of RSC30, which causes a significant impact on RP expression. As mentioned above, we have included a table depicting the redundancy relationship in Figure 5—figure supplement 2C.

We speculate that Rsc1, as a later evolutionary adaptation, seems to be specialized to help remodel these regions. RSC1 is only 20% of total RSC complexes, yet has a slightly higher occupancy at RPGs, wide NDRs and tDNAs than RSC2. While the differences in occupancy and expression may not appear that substantial as a result of the redundancy, having nucleosomes poised in a specialized state that is “better” handled by one form of RSC provides for rapid and regulated action in response to environmental signals, and likely provides a fitness advantage to yeast in the wild.

Reviewer #3:

[…] Some specific questions and comments:

1) Does swapping the CT2 domain of Rsc1 into Rsc2 also suppress the synthetic lethality with K56Q?

Yes it does. This is shown in Figure 3—figure supplement 2B.

2) Subsection “RSC2 complexes are deficient in remodeling partially-unwrapped nucleosomes”: The authors conclude that Rsc1 complex is more active for remodeling unwrapped nucleosomes because it has more activity on the R40A nucleosomes than Rsc2. But that is also true for WT 5S nucleosomes. Given that R40A is unlikely to completely disrupt wrapping, isn't it possible that remodeling only occurs when the nucleosomes are wrapped, and R40A simply reduces the time spent in the wrapped conformation? This idea seems supported by the fact that R40A has less effect in the 601 DNA sequence. Following this line of reasoning, could Rsc1 do a better job in inducing unwrapped nucleosomes to adopt the wrapped "remodelable" configuration? The discussion seems to be moving in this direction, but it this is the authors' model it could be made clearer and perhaps foreshadowed more strongly in the results.

Yes, the idea that Rsc1 could do a better job inducing the wrapped conformation, likely via its onboard Rsc3/30 module which could bind the DNA and help “wrap” or stabilize the nucleosome (much like Hmo1) fits nicely within our model. The alternative is that RSC1 complex could be more tolerant of a partially-wrapped nucleosome to initiate DNA translocation. We include both of these possibilities in the new Discussion.

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

[…] In your revised manuscript, please address this point. I took a look at the paper cited by the reviewer (North et al., 10215) and it does seem to claim that 601 unwraps its ends more that 5S. If the reviewer's summary of the literature is accurate, please modify your statements (and model Figure 6) concerning 5S versus 601 accordingly, and factor that into your discussion and interpretations. If you feel those results can be reconciled, or are not applicable, to your model, please explain why in your Discussion section.

Again, we would like to thank the reviewers and editors for their efforts in reviewing and helpful comments – which improved the manuscript – and for their appreciation of the quality and significance of the work. We were delighted to have the paper provisionally accepted at eLife. Below we address the one remaining request from the editors and reviewer.

One reviewer expresses concern that our paper conveys a more simplified conclusion (nucleosome partial unwrapping) than merited for differences between Rsc1 versus Rsc2, given that a prior paper has found that the 601 DNA sequence displays more unwrapping from the nucleosome edge (entry/exit) than does the 5S sequence (North et al., 2012). The reviewer is right to point out this paper, and we agree that additional discussion would be helpful to reconcile their observations with ours. To fully address, we will discuss a major caveat with the North et al., paper, discuss subsequent published work showing higher unwrapping with 5S compared to 601, and emphasize that with our nucleosomes (using yeast octamers with either sea urchin 5S or the Widom 601 sequence), the 5S displays more unwrapping than does the 601 nucleosome.

First, we utilized the sea urchin 5S sequence, while the North et al., paper used Xenopus, 5S – so the DNA sequences are slightly different. Second, and much more importantly, North et al. make a significant alteration to both the 5S and 601 sequences by inserting the LexA binding sequence within the region that unwraps from the octamer. In addition, the authors state that without LexA bound, 5S has a lower FRET efficiency than 601, possibly due to increased unwrapping by the first few base pairs of the 5S. Third, there are additional studies, subsequent to North et al. (e.g., Chen et al., 2014 and Mauney et al., 2018) that used small angle X-ray scattering (SAXS) to compare native/unaltered sea urchin 5S to the Widom 601 sequence, which showed that the 5S unwraps more rapidly and at lower salt concentrations than does 601. Fourth, our studies utilize yeast octamers to form nucleosomes, which we consider best suited to the study of yeast remodelers, rather than the more common usage of either Drosophila or Xenopus histones. Here, yeast octamers might behave somewhat differently in their wrapping properties.

Importantly, rather than simply relying on prior studies, and given that our studies use yeast octamers with the 5S and 601 sequences, we directly examined the extent of unwrapping with the 5S and 601 nucleosomes used in our studies. To this end, we performed ExoIII-S1 nuclease mapping combined with high throughput sequencing to determine the extent of unwrapping in our starting nucleosomal templates, and found that 5S nucleosomes are unwrapped to a greater extent than 601 nucleosomes, and that the H3 R40A mutation enhances the unwrapping of the 5S, but not the 601 template. (Figure 3—figure supplement 5, and Figure 4—figure supplement 5B). Our mapping studies are thus in agreement with the findings from Chen et al., 2014 and Mauney et al., 2018. Taken together with our genetics results, and the positive impact (in vitro and in vivo) of Hmo1, which is predicted to improve wrapping – we conclude that partial nucleosome unwrapping provides the most parsimonious explanation for the differences observed in remodeling capacity between Rsc1 and Rsc2 complexes.

To address this issue in the manuscript, we have added content and these additional references to the Introduction, made a clearer statement about our mapping of the unwrapping of the 5S and 601 starting templates in the Results section, and added a paragraph to the Discussion section.

https://doi.org/10.7554/eLife.58130.sa2

Article and author information

Author details

  1. Alisha Schlichter

    Howard Hughes Medical Institute (HHMI), Department of Oncological Sciences, Huntsman Cancer Institute, University of Utah School of Medicine, Salt Lake City, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Writing - original draft, Writing - review and editing
    Contributed equally with
    Margaret M Kasten
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-1928-9396
  2. Margaret M Kasten

    Howard Hughes Medical Institute (HHMI), Department of Oncological Sciences, Huntsman Cancer Institute, University of Utah School of Medicine, Salt Lake City, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Writing - original draft, Writing - review and editing
    Contributed equally with
    Alisha Schlichter
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-5753-7122
  3. Timothy J Parnell

    Howard Hughes Medical Institute (HHMI), Department of Oncological Sciences, Huntsman Cancer Institute, University of Utah School of Medicine, Salt Lake City, United States
    Contribution
    Data curation, Formal analysis, Methodology, Writing - original draft
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-3632-3691
  4. Bradley R Cairns

    Howard Hughes Medical Institute (HHMI), Department of Oncological Sciences, Huntsman Cancer Institute, University of Utah School of Medicine, Salt Lake City, United States
    Contribution
    Conceptualization, Supervision, Funding acquisition, Writing - original draft, Writing - review and editing
    For correspondence
    brad.cairns@hci.utah.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-9864-8811

Funding

Howard Hughes Medical Institute (Brad Cairns Investigator)

  • Alisha Schlichter
  • Margaret M Kasten
  • Bradley R Cairns

National Cancer Institute (CA042014)

  • Timothy J Parnell

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

Acknowledgements

We thank Brian Dalley and the High-Throughput Sequencing Shared Resource for sequencing data. We thank Ali Shilatifard, Tim Formosa, Jef Boeke, Kevin Struhl, Anne Grove, Roger Kornberg and Sharon Dent for strains, plasmids, and labeled yeast octamer. We thank Patrick Murphy and Edward Grow for helpful discussions on data analysis and experimental design and Naveen Verma, Tim Formosa, and Mahesh Chandrasekharan for feedback on the manuscript. We thank HHMI for support of AS, MK and BC, and thank NCI CA042014 to Huntsman Cancer Institute for support of core facilities.

Senior Editor

  1. Kevin Struhl, Harvard Medical School, United States

Reviewing Editor

  1. Stephen Buratowski, Harvard Medical School, United States

Reviewer

  1. Blaine Bartholomew, University of Texas M.D. Anderson Cancer Center, United States

Publication history

  1. Received: April 22, 2020
  2. Accepted: June 3, 2020
  3. Accepted Manuscript published: June 4, 2020 (version 1)
  4. Version of Record published: June 22, 2020 (version 2)

Copyright

© 2020, Schlichter 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

  • 643
    Page views
  • 161
    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)

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

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

Further reading

    1. Biochemistry and Chemical Biology
    2. Plant Biology
    Pengxiang Fan et al.
    Research Article
    1. Biochemistry and Chemical Biology
    2. Cell Biology
    Senem Aykul et al.
    Research Article Updated