Establishment and Maintenance of Chromatin Architecture Are Promoted Independently of Transcription by the Histone Chaperone FACT and H3-K56 Acetylation in Saccharomyces cerevisiae

Using a combination of in vitro biochemistry, genetics, and genomics, McCullough and Pham et al. explore how the histone:DNA contact at the entry/exit site of nucleosomes affects the functions of the histone chaperone FACT in the yeast...

Abstract

FACT (FAcilitates Chromatin Transcription/Transactions) is a histone chaperone that can destabilize or assemble nucleosomes. Acetylation of histone H3-K56 weakens a histone–DNA contact that is central to FACT activity, suggesting that this modification could affect FACT functions. We tested this by asking how mutations of H3-K56 and FACT affect nucleosome reorganization activity in vitro, and chromatin integrity and transcript output in vivo. Mimics of unacetylated or permanently acetylated H3-K56 had different effects on FACT activity as expected, but the same mutations had surprisingly similar effects on global transcript levels. The results are consistent with emerging models that emphasize FACT’s importance in establishing global chromatin architecture prior to transcription, promoting transitions among different states as transcription profiles change, and restoring chromatin integrity after it is disturbed. Optimal FACT activity required the availability of both modified and unmodified states of H3-K56. Perturbing this balance was especially detrimental for maintaining repression of genes with high nucleosome occupancy over their promoters and for blocking antisense transcription at the +1 nucleosome. The results reveal a complex collaboration between H3-K56 modification status and multiple FACT functions, and support roles for nucleosome reorganization by FACT before, during, and after transcription.

FACT (FAcilitates Chromatin Transcription/Transactions) is a conserved histone chaperone that can both destabilize and assemble nucleosomes (Singer and Johnston 2004; Reinberg and Sims 2006; Winkler and Luger 2011; Formosa 2012). FACT is a heterodimer of Spt16 with either Pob3 (in yeast and fungi) or SSRP1 (in higher eukaryotes), with each subunit comprising multiple histone-binding modules connected by unstructured linkers (Winkler and Luger 2011; Kemble et al. 2015; Tsunaka et al. 2016). SSRP1 has an intrinsic DNA-binding domain not found in Pob3, but both versions of FACT require additional free-standing DNA-binding activity to support full activity (Valieva et al. 2016; McCullough et al. 2018). The High Mobility Group B (HMGB) family member Nhp6 provides this function in yeast (Stillman 2010; Valieva et al. 2016; McCullough et al. 2018). FACT is proposed to use these domains to bind multiple sites on nucleosomes, sequentially exposing and engaging additional buried sites to ultimately produce an altered structure, which we have called the “reorganized” nucleosome (Winkler and Luger 2011; Formosa 2012; Hondele and Ladurner 2013; Tsunaka et al. 2016; McCullough et al. 2018). The nucleosome assembly activity of FACT (Belotserkovskaya et al. 2003) is then proposed to involve reversal of these steps.

Nucleosomes stall the progression of RNA polymerase II (RNA Pol II) in vitro and FACT was initially named for its ability to partially reverse this block (Reinberg and Sims 2006; Hsieh et al. 2013). Based on this activity, FACT was inferred to facilitate the progression of transcription complexes in vivo by destabilizing chromatin barriers. Consistent with a role in transcription, FACT occupancy is generally proportional to the level of transcription (Mason and Struhl 2003; Pelechano et al. 2009; Mayer et al. 2010; Feng et al. 2016). However, this occupancy profile can also be explained by a more passive model in which FACT localization is driven by exposure of its binding sites during any process that disrupts nucleosomes, including transcription (Kemble et al. 2015; Tsunaka et al. 2016; Martin et al. 2018; Wang et al. 2018). In this model, FACT’s role is to rescue disrupted nucleosomes by reversing the reorganization pathway. This could explain the increased nucleosome turnover observed in highly transcribed regions upon acute depletion of FACT (Jamai et al. 2009), as nucleosomes destabilized during transcription could not be restored to the canonical state and would therefore need to be replaced. Consistent with this view, FACT can stabilize nucleosomes in vitro (Hsieh et al. 2013; Chang et al. 2014) and it is needed to restore nucleosome occupancy after transcription (Biswas et al. 2007; Fleming et al. 2008; Kaplan et al. 2008; Ransom et al. 2009; Hainer et al. 2011, 2012; Voth et al. 2014; Feng et al. 2016). However, FACT also has an established role in ejecting nucleosomes from promoters during transcriptional activation (Biswas et al. 2006; Ransom et al. 2009; Takahata et al. 2009a,b; Shakya et al. 2015). FACT can therefore either stabilize or destabilize nucleosomes, depending on the direction of the reorganization reaction being promoted. This raises questions about how the direction of the reaction is regulated to achieve opposite outcomes in different contexts.

The strength of the histone–DNA contact at the entry/exit point of the nucleosome is anticipated to have a key role in FACT activity, as breaking this contact is the first step toward reorganization, and forming it is among the last steps in nucleosome assembly. Lysine 56 of histone H3 (H3-K56) interacts with the phosphodiester backbone of the DNA at the entry/exit point of nucleosomes and acetylation of this residue weakens this contact (Andrews et al. 2010). H3-K56 is acetylated in nascent H3-H4 dimers, so newly formed nucleosomes carry this mark globally after S phase (Driscoll et al. 2007; Han et al. 2007; Tsubota et al. 2007; Kaplan et al. 2008; Kuang et al. 2014). Hst3 and Hst4 deacetylate H3-K56 during G2/M, but subsequent nucleosome deposition restores the acetylated form, so the presence of H3-K56ac in interphase reveals sites of nucleosome turnover (Rufiange et al. 2007; Kaplan et al. 2008; Kuang et al. 2014). Constitutive turnover appears to be an important mechanism for maintaining appropriate nucleosome occupancy in promoters (Parnell et al. 2015), but it is not known whether the H3-K56ac observed at these sites is solely a temporary consequence of turnover, if it has a functional role in promoting it, or if modification at this site has some other more persistent role in maintaining appropriate chromatin structure.

Here, we investigated how FACT activity and functions are affected by histone H3-K56 status. We probed the importance of the nucleosomal entry/exit point contact for FACT’s activities in vitro by weakening it with the acetylation mimic H3-K56Q, or by using FACT mutants with defects in promoting or resolving reorganization (Spt16-11 and Pob3-Q308K, respectively). We then compared the results with the effects of the same mutations in vivo to ask how altering this balance affects phenotypes, including global changes in nucleosome occupancy and positioning, and changes in steady-state transcript levels. Weakening the entry/exit point contact might be expected to make it easier for FACT to promote reorganization, but our previous results suggested that release of FACT from nucleosomes requires satisfying an “integrity checkpoint” (McCullough et al. 2013, 2015, 2018), which might be more difficult to accomplish with a weaker contact. Previous studies showed that acute withdrawal of FACT using unstable temperature-sensitive alleles had severe effects on transcription and chromatin structure (Feng et al. 2016), but loss of FACT is ultimately lethal. We used more subtle mutations with defined defects in specific properties to ask how each component contributes to chromatin integrity and how the resulting changes in chromatin affect transcript levels.

Our results support the expected importance of the strength of the entry/exit point contact for appropriate FACT function in vitro and in vivo, but reveal a complex relationship. Some changes observed were consistent with the expected role for FACT in maintaining chromatin in a repressive state in vivo, as the largest increases in transcript levels occurred in conditionally expressed genes with a “closed” promoter architecture (Parnell et al. 2015) and antisense transcript levels increased globally in mutants. However, other outcomes were more surprising, including good correlations for transcript profile changes when comparing mutations that were predicted to have different effects, poor correspondence between the effects of FACT mutations and the expected outcomes for defective transcription elongation, unexpected general depopulation of nucleosomes over the 3′ ends of genes irrespective of transcription frequency, and poor correspondence between changes in nucleosome occupancy and changes in transcript levels. The results therefore support an important role for FACT in establishing and maintaining chromatin architecture, but raise questions about how that architecture influences and is influenced by transcription, and the relative importance of distinct FACT functions before, during, and after transcription.

Materials and Methods

Yeast strains are listed in Supplemental Material, Table S1 and are isogenic with the W303 background (Thomas and Rothstein 1989). hht1-K56A/R/Q and hht2-K56A/R/Q mutations were integrated into the native loci without genetic markers (Storici et al. 2001). Translationally silent substitutions were introduced adjacent to the desired mutation to allow detection of these alleles using a combination of PCR amplification (with primers that distinguish HHT1 and HHT2; Table S4), SalI endonuclease digestion, melting curve analysis (Wittwer et al. 2003), and sequencing. Accurate integration was confirmed by DNA sequencing, then standard genetic methods were used to construct strains with the desired combinations of mutations (Rothstein 1991; Sherman 1991). For phenotype tests, strains were grown to saturation in rich medium, and 10-fold serial dilutions were plated on solid medium and incubated as described in each figure legend. MNase-sequencing (MNase-seq) and RNA-sequencing (RNA-seq) were performed with the strains listed in Table S1. Three independent cultures were processed in parallel for each genotype, and sequencing was performed using single-end reads on an Illumina platform. Library preparation and bioinformatic analysis followed the outline of previous work (Ramakrishnan et al. 2016; Sdano et al. 2017); a more detailed description is found in the Supplemental Methods. Electrophoretic mobility shift assays (EMSAs), complex disruption with unlabeled competitor DNA, and restriction endonuclease sensitivity quantification were performed as described (Xin et al. 2009). Western immunoblots were probed with an antibody against rabbit yeast histone H3 (Covance, special order using the peptide N-CKDIKLARRLRGER-C as the antigen) and a mouse antibody against yeast Pgk1 (Molecular Probes, Eugene, OR), along with anti-rabbit (800CW) and anti-mouse (680RD) secondary antibodies (Li-Cor), and scanned with an Odyssey CLx infrared scanner (Li-Cor).

Data availability

Strains and plasmids are available upon request. Supplemental material available at Figshare: https://doi.org/10.25386/genetics.7586207 as a single file containing Tables S1–S3 (strains used), Table S4 (primers used), Table S5 (comparison of correlation coefficients for scatter plots), Figures S1–S7, and Supplemental Methods describing genomic analysis. Gene expression and MNase-seq data are available at the Gene Expression Omnibus with the accession number: GSE118332.

Results

Effects of H3-K56Q on FACT activity in vitro

Reorganization of nucleosomes by FACT has been detected in vitro through the formation of slow-migrating FACT:nucleosome complexes, observed by native gel electrophoresis and by increased sensitivity to restriction endonucleases (Xin et al. 2009). The Spt16-Pob3(Q308K) mutant heterodimer is somewhat hyperactive for reorganization of nucleosomes (McCullough et al. 2013 and Figure 1A), but displays inefficient dissociation of the FACT:nucleosome complexes upon the addition of competitor DNA (McCullough et al. 2013 and Figure 1, B and C). This release defect has been attributed to failure of a nucleosome assembly quality control step (McCullough et al. 2013, 2018). We previously found that specific mutations in histone H4 can suppress phenotypes caused by pob3-Q308K in vivo and also promote efficient dissociation of FACT(Q308K):nucleosome complexes in vitro (McCullough et al. 2013), demonstrating that this assay accurately reports on a physiologically relevant function of FACT. In contrast, (Spt16-11)-Pob3 heterodimers are defective in producing reorganization (McCullough et al. 2011 and Figure 1A), but release from complexes normally (Figure 1C). These defects are suppressed both in vivo and in vitro by specific H2A-H2B mutations (McCullough et al. 2011), again validating these assays for measuring relevant functions of FACT in vivo. These mutant proteins are stable in vivo at 30° (VanDemark et al. 2008), so the phenotypes observed at this temperature reveal specific functional defects rather than generic effects of loss of an essential factor. We focus here on these alleles as they have opposite activity defects (failure to assemble or release from nucleosomes and failure to efficiently destabilize nucleosomes, respectively), and different profiles of suppression and enhancement of phenotypes in vivo (McCullough et al. 2011, 2013, 2015), so they provide probes for distinct functions of FACT.

Figure 1.

Effects of H3-K56Q on FACT activities in vitro. Nucleosomes were constructed with recombinant yeast histones (H3 or H3-K56Q) and a 181-bp 5S rDNA fragment, as described previously (Xin et al. 2009). (A) The rate of DraI digestion at a site near the dyad was determined without or with 0.2 µM Spt16-Pob3, without or with 2 µM Nhp6. “FACT” indicates that both Spt16-Pob3 and Nhp6 were added, with mutated subunits substituted as indicated. Initial rates of digestion were determined and normalized to H3-containing nucleosomes treated with WT FACT in parallel. Error bars indicate the SD. For WT FACT, the average rate of digestion with H3-K56Q nucleosomes was 66% of the WT rate (P = 0.1%, paired Student’s t-test). (B) Nucleosomes were incubated alone or with Nhp6 and Spt16-Pob3 as in (A), then separated by native polyacrylamide gel electrophoresis without (top panel) or after adding unlabeled competitor DNA (bottom panel). Additional combinations that did not cause a release defect are not shown. (C) The fraction of retained complexes resistant to dissociation using the assay shown in (B) was quantitated, with the average and SD for multiple independent repeats shown. Complex retention with Spt16-Pob3(Q308K) and H3-K56Q was 90% of the level observed for H3 (P = 0.02%, paired Student’s t-test). FACT, FAcilitates Chromatin Transcription/Transactions; Nuc, nucleosome; WT, wild-type.

To assess how weakening the histone–DNA contact at the entry/exit point affects reorganization by FACT, we assembled nucleosomes using recombinant yeast histones with H3-K56Q in place of H3. H3-K56Q has been validated as a mimic of H3-K56ac in a number of genetic and biophysical assays, showing that it destabilizes the entry/exit point histone–DNA interaction (Chen et al. 2008; Watanabe et al. 2010; Shimko et al. 2011). H3-K56Q had minimal effects on the affinity of nucleosomes for FACT (Figure S1, A and B), but produced a lower rate of DraI digestion, revealing a moderate defect as a substrate for reorganization (Figure 1A). FACT(Q308K) produced the expected increased digestion rate and decreased efficiency of release from nucleosomes (McCullough et al. 2011, 2018), and the latter was weakly suppressed by H3-K56Q (Figure 1B). The suppression was statistically significant but minimal in magnitude (Figure 1C). We expected H3-K56Q to promote initiation of reorganization, but the results indicate that it instead made reorganization by FACT more difficult, and assembly or release easier.

Effects of H3-K56 status on FACT functions in vivo

To connect the activities of FACT observed in vitro with its roles in vivo, we combined mutations that affect the status of H3-K56 with a range of FACT mutations. The commonly used spt16-G132D allele encodes an unstable protein that is rapidly degraded when cultures are shifted to a nonpermissive temperature (VanDemark et al. 2008; Feng et al. 2016). While useful for testing the effects of acute depletion of FACT, this allele is less informative for dissecting distinct functions of FACT in viable cells. Therefore, we focused on the pob3-Q308K and spt16-11 alleles described above (Figure 1). We used three methods to alter the modification state of H3-K56: (1) loss of acetylation by deleting the machinery that writes this mark (the acetyltransferase encoded by RTT109; Schneider et al. 2006; Driscoll et al. 2007; Han et al. 2007), (2) increased acetylation by removing its erasers (the deacetylases encoded by HST3 and HST4; Celic et al. 2006; Maas et al. 2006; Kaplan et al. 2008), and (3) mutating H3-K56 to alanine (uncharged, no interaction with DNA), arginine (constitutive stabilization of the DNA contact), or glutamine (permanent weakening of the contact). Histone H3 is produced in Saccharomyces cerevisiae from two genes that encode identical proteins, HHT1 and HHT2 (Smith and Andresson 1983), so mutations are often tested by deleting the endogenous genes and supplying the mutated gene on a plasmid. This can result in copy number variation, and we have shown previously that histone genes on plasmids affect the phenotypes of FACT mutants and their interactions with histone mutations (McCullough et al. 2011, 2013). We also found that placing markers near the wild-type (WT) H3 genes caused slow growth when combined with FACT mutants (data not shown), so we instead introduced the K56A, K56R, and K56Q mutations into both HHT1 and HHT2 through markerless conversion (Storici et al. 2001), adding translationally silent substitutions to allow detection of the alleles in genetic crosses (see Materials and Methods).

The H3-K56 mutations had minimal or no effect on total or soluble H3 levels (Figure S2, A and B), and did not cause temperature sensitivity or sensitivity to hydroxyurea (HU) under the conditions tested [Figure 2A; phenotypes consistent with those reported in Masumoto et al. (2005) were observed under more stringent conditions; data not shown]. Combining pob3-Q308K with H3-K56 mutations caused synthetic defects ranging from severe with H3-K56A to minor with H3-K56R (Figure 2A). The effects of H3-K56A or H3-K56R could reflect known overlapping functions of FACT with the histone chaperone Rtt106, which, unlike FACT, binds preferentially to (H3-H4)₂ tetramers with the H3-K56ac modification (Su et al. 2012; Zunder et al. 2012; Kemble et al. 2013). The severe growth defect caused by H3-K56A made genetic interactions difficult to interpret, so it was not studied further. H3-K56Q enhanced the growth defect, temperature sensitivity, and HU sensitivity (HUs) caused by pob3-Q308K, while the effects of H3-K56R were more moderate and were limited to the HU sensitivity (Figure 2A), suggesting a defect during DNA replication.

Figure 2.

Genetic effects of manipulating H3-K56 status in FACT mutants. H3-K56Q, K56R, and K56A alleles were constructed by markerless integration into the normal genomic HHT1 and HHT2 sites, then combined with alleles of FACT by standard genetic methods (Table S2). Next, 10-fold dilutions were tested for growth under various conditions indicated (see Materials and Methods). D indicates rich medium (yeast extract, peptone, adenine, dextrose). Labels indicate the concentration of HU (mM), the temperature, and the length of incubation in days. (A) The temperature and HU sensitivity caused by pob3-Q308K were strongly enhanced by H3-K56Q or H3-K56A, while H3-K56R had no effect on the temperature-sensitive phenotype and caused only moderate enhancement of the HU-sensitive phenotype. spt16-11 displayed stronger synthetic defects with H3-K56R than with H3-K56Q at elevated temperatures, and a strong synthetic defect with H3-K56R on HU, but mild suppression of the HU-sensitive phenotype by H3-K56Q. (B) As in (A), except FACT mutations were combined with deletions of the H3-K56ac deacetylases HST3 and HST4. (C) As in (B), except the gene encoding the H3-K56 acetyltransferase RTT109 was deleted. Growth on –lys indicates the Spt⁻ phenotype resulting from incomplete repression of the cryptic promoter associated with insertion of a Ty1 transposon ∂ element in LYS2 (Simchen et al. 1984). The HU sensitivity caused by rtt109∆ (Schneider et al. 2006) complicates interpretation of the enhanced HU-sensitive phenotype. FACT, FAcilitates Chromatin Transcription/Transactions; HU, hydroxyurea; −lys, synthetic medium lacking lysine; YPAD, [insert definition].

Combining spt16-11 with H3-K56Q/R enhanced the temperature-sensitive (Ts⁻) phenotype, but in this case H3-K56R was more detrimental than H3-K56Q (Figure 2A). Importantly, H3-K56Q moderately suppressed the HU sensitivity caused by spt16-11. Distinct genetic interactions with these FACT alleles are expected based on previous studies (McCullough et al. 2011, 2013) and their different activity defects in vitro (Figure 1). We did not expect weakening of the entry/exit point contact (H3-K56Q) to make reorganization more difficult, but given that outcome (Figure 1), it is sensible that the H3-K56Q mutation was detrimental in strains with altered reorganization capacity. However, it is puzzling that the FACT allele with stronger reorganization capability in vitro (pob3-Q308K) was more negatively impacted than one with an explicit defect in this activity (spt16-11; pob3-Q308K displayed enhanced Ts⁻ and HUs phenotypes, and spt16-11 displayed enhanced Ts⁻ but reduced HUs phenotypes). It also remains unclear why strengthening the contact (H3-K56R) was beneficial in a pob3-Q308K strain but detrimental in an spt16-11 strain, although this result might point to the importance of establishing this contact prior to the release of FACT. These results therefore support the importance of reorganization as a core physiological function of FACT and demonstrate a role for H3-K56 status in this activity, but leave open questions regarding what processes the distinct functions of FACT promote and how the failure of each activity contributes to gross phenotypes measured in these plating assays.

To compare the H3-K56Q acetylation mimic with the effects of H3-K56ac itself, we deleted HST3/HST4 to produce increased levels of H3-K56ac (Celic et al. 2006; Maas et al. 2006; Kaplan et al. 2008; Thurtle-Schmidt et al. 2016). This strongly enhanced the Ts⁻ phenotypes caused by both FACT mutations and the HUs phenotype caused by pob3-Q308K, just as H3-K56Q did (Figure 2B). However, the suppression of the HUs caused by spt16-11 was not observed, presumably because of the known additional functions of these deacetylases aside from modifying H3-K56 (Madsen et al. 2015). Eliminating H3-K56ac by deleting RTT109 should mimic H3-K56R, and both manipulations had minimal effects when combined with pob3-Q308K but severe effects with spt16-11 (Figure 2, A and C). These results show that H3-K56ac affects the physiological functions of FACT, that being able to switch between stable and unstable configurations at this site is particularly important in FACT mutants, and that reorganization may have a prominent role during DNA replication when high levels of nucleosome disassembly and assembly occur globally.

FACT and H3 mutations caused similar defects in transcript accumulation

To ask how chronic FACT and H3-K56 defects affect gene expression, we used RNA-seq to examine the changes in steady-state transcript profiles. We then compared the expression changes measured in triplicate for each mutant strain as normalized log₂ ratios relative to the WT strain (log₂FC). Each set of strains with the same genotype gave consistent changes as determined by principal component analysis (Figure S3A). Figure 3A shows the number of genes whose normalized transcript levels increased or decreased at least twofold at a false discovery rate of 1%, as determined by DESeq2 (Love et al. 2014). The largest effects on both increased and decreased transcript levels were observed in the strain with both H3-K56Q and pob3-Q308K. This increased effect upon combining these mutations suggests that FACT activity and the strength of the entry/exit point contact collaborate to accomplish normal regulation of transcript levels.

Figure 3.

Global effects of mutations on transcript levels. Three biological replicates of WT and mutant strains were grown at 30° (strains used are listed in Table S1), and RNA-seq was performed as described in the Materials and Methods. rlog₂ values were calculated for each mRNA (using only the reads representing the sense strand for each annotated gene) and the ratio to WT was determined (log₂FC). (A) The number of genes with more than twofold changes in transcript level at an FDR < 1% is shown for each strain. (B) The log₂FC values for the pob3-Q308K strain are shown plotted against log₂(RNA Pol II occupancy) values from a published chromatin immunoprecipitation-sequencing data set (Pelechano et al. 2010). The Spearman correlation coefficient and the slope of a linear regression analysis are indicated. (C) As in (B), except the log₂FC values are plotted against the rlog₂ values for total sense-strand mRNA for the WT strain in this study, or against the log₂ of the annotated length of each transcript in base pairs (D). Similar analyses of the other mutants described here and testing of other metrics for transcription frequency are shown in the supplemental material (Figure S3, B–F). FACT, FAcilitates Chromatin Transcription/Transactions; FDR, false discovery rate; log₂FC, normalized log₂ ratios relative to the WT strain; rlog₂, regularized log₂; RNA Pol II, RNA polymerase II; RNA-seq, RNA-sequencing; WT, wild-type.

Initial models suggested that FACT promotes the progression of polymerases through chromatin by weakening the barrier function of each nucleosome encountered (Reinberg and Sims 2006). However, attempts to detect the predicted decrease in the rate of RNA Pol II elongation in FACT mutants have had mixed results, with some reporting indirect effects (Fleming et al. 2008) but others finding no changes (Mason and Struhl 2005; Biswas et al. 2006; Kuryan et al. 2012). Recent work indicates that FACT occupancy is driven by exposure of its binding sites as nucleosomes are disrupted by transcription, suggesting that FACT responds to transcription rather than facilitating it (Martin et al. 2018). If FACT or H3-K56ac promote polymerase progression, mutations could affect initiation (the ability to convert an initiation complex to an elongation complex by overcoming the initial barrier posed by the +1 nucleosome) or elongation (the ability to progress efficiently through each successive nucleosome encountered throughout transcription units). The average gene in yeast is transcribed about once every 8 min (Pelechano et al. 2010), so inefficient progression through the +1 nucleosome might have little impact on steady-state transcript levels for most genes, but could impair the optimal production of frequently transcribed genes that require rapid cycles of initiation. Similarly, inefficient elongation might be expected to affect longer genes disproportionately due to the accumulated effects of delays at many consecutive barriers. Yeast cells have mechanisms for buffering the total RNA pool size (Sun et al. 2013), so we examined the steady-state RNA levels for evidence of skewing within the pool that would indicate disproportionate effects related to transcription frequency or gene length.

The normalized log₂FC values for the pob3-Q308K strains relative to WT were plotted against the frequency of transcription determined by RNA Pol II occupancy (Figure 3B; Pelechano et al. 2010), the total abundance of transcripts for each gene from this study (Figure 3C), and the length of each transcription unit (Figure 3D). The Spearman correlation coefficient and the slope of the correlation determined by linear regression are given, as this reveals both the quality of the correlation as well as the magnitude of the dose response. The pob3-Q308K FACT mutant gave low-to-moderate negative correlations with these measures of transcription frequency (r ∼ −0.3) with low dose responses (m = −0.084 indicates a twofold decrease in transcript level is associated with ∼4000-fold difference in the normal transcription frequency). Similar results were obtained with other single mutants (Figure S3, B–D). Other methods of determining transcription frequency, including metabolic labeling and Native Elongating Transcript sequencing (NET-Seq), gave lower correlations than those shown here (Figure S3, E and F). The strongest effects were again observed with the H3-K56Q pob3-Q308K combined mutant, with moderate negative correlations (r = −0.44 and −0.36) and higher dose responses (m = −0.11 and −0.2) for mRNA levels and Pol II occupancy, respectively.

To compare the effects of chronic FACT defects with acute withdrawal, we analyzed a published RNA-seq data set generated with the spt16-G132D allele after a 45-min incubation at the restrictive temperature of 37°, which leads to acute loss of FACT (Feng et al. 2016). The base RNA levels from these data sets gave good correlations with one another (r = +0.81, m = 0.85 for both WT and spt16-G132D strains grown under permissive conditions; Figure S3G panels a and b). The changes in transcript levels upon acute FACT withdrawal produced a strong negative correlation and dose response with RNA Pol II occupancy (Figure S3G, d; r = −0.48, m = −0.84) and transcript abundance (Figure S3G panel e; r = −0.79, m = −0.63). Neither our analysis nor the published data showed strong correlation between changes in transcript level and gene size (Figure 3D and Figure S3G panel f), although the acute withdrawal data showed a weak positive correlation that appears to suggest increased proficiency of elongation in the absence of FACT. These results show that defects in different phases of transcription can be detected through skewing of the distribution of the steady-state levels of transcripts, and indicate that while acute loss of FACT appears to affect the efficiency of initiation of transcription, more subtle defects in FACT activity or H3-K56 mutations (that nonetheless cause robust phenotypes in vivo) do not have this effect unless combined with one another. Neither acute loss of FACT nor more subtle mutations had differential effects on the accumulation of long transcripts that were expected with defective elongation, suggesting that either elongation was not severely impacted or that this metric is not sensitive enough to detect a change in this feature of transcription.

To compare the effects of the different mutations on individual genes, we plotted the log₂FC values in pairwise combinations (Figure 4). These comparisons produced strong positive correlations and dose responses, indicating that the different mutations affected individual genes similarly. The spt16-G132D strain grown here under permissive conditions gave the weakest effects, and correlations were also modest when we compared the log₂FC(37°/25°) values for the published spt16-G132D data set with our mutant strains (Feng et al. 2016; Figure S4A). The combined H3-K56Q pob3-Q308K mutant that displayed additive phenotypes above (Figure 2A and Figure 3A) also showed roughly additive effects on individual genes (Figure 4; H3-K56Q vs. pob3-Q308K had a slope of m = 0.52, pob3-Q308K compared to itself has a slope of 1, and H3-K56Q pob3-Q308K vs. pob3-Q308K had a slope of m = 1.3, with complete additivity predicting a value of m = 1.52). The combined mutant also produced the best correlation with acute withdrawal of FACT (Figure S4A, F; r = +0.52, m = 0.25), suggesting that a FACT defect and an H3-K56Q mutation affect distinct but overlapping features of transcriptional control, and that combining them begins to approach acute loss of FACT activity.

Figure 4.

A range of mutations had similar effects on transcript levels from the same genes. The log₂FC values for the changes in transcript levels relative to WT were calculated as in Figure 3, then plotted against each other in pairs. (A–E) Each mutant (y-axis) compared to the pob3-Q308K strain (x-axis). (F) The H3-K56R strain compared to H3-K56Q. The Spearman correlation coefficients and regression line slopes are indicated. log₂FC, normalized log₂ ratios relative to the WT strain; WT, wild-type.

We expected weakening and strengthening of the entry/exit point contact to have opposite effects, but H3-K56Q and H3-K56R produced the highest correlation observed among pairs of mutants (Figure 4F; r = +0.78, m = 1.0). This surprising result reveals a complex role for this site in controlling transcript levels, potentially indicating that the ability to both install and remove modifications is more important than the state of modification. Similarly, we expected spt16-11 and pob3-Q308K alleles to have different effects because, as noted above, they affect FACT activity differently and often display distinct genetic interactions. However, these strains also produced similar changes in transcripts from the same genes (Figure 4B; r = +0.66, m = 0.49). The effects of these mutations on transcript accumulation are therefore more complex than their effects on nucleosome reorganization.

A screen of a large number of viable yeast deletion strains revealed a common pattern of changes in transcript accumulation that was attributed to skewing of the distribution of cells among cell cycle phases in many slow-growing strains (O’Duibhir et al. 2014). To see if this effect was causing the similar outcomes among our strains, we removed this “slow-growth signature” from our data and repeated our analyses. Comparing the effects of each mutant with and without this correction revealed only small changes (Figure S4B; the largest effects and strongest evidence for this profile were observed in the H3-K56R strain), and this did not alter the overall similarity of mutant–mutant comparisons (Figure S4C), or the correlations with transcription frequency or gene size (data not shown). The similar effects of these mutations on transcript accumulation from individual genes therefore do not appear to be the result of triggering the known pattern of transcript changes associated with cell cycle redistribution.

Effects of mutations on nucleosome occupancy and positioning

FACT and H3-K56ac mutations are predicted to affect transcription by altering the chromatin landscape. To test this idea, we performed MNase-seq in triplicate with a subset of the strains described above to measure changes in nucleosome occupancy and positioning. Principal component analysis indicated good clustering of the outcomes by genotype (Figure S5A).

To determine nucleosomal spacing around the transcription start site (TSS), we plotted the cumulative predicted midpoints of the nucleosomal fragments (calculated as 74 bp from the 5′ end of each sequenced fragment, assuming MNase digested all linkers efficiently). Nucleosomes downstream of the TSS shifted further downstream in the mutants, with most of the change attributable to the FACT mutation pob3-Q308K (Figure 5A). To quantitate this effect, we estimated the change in position at each gene for the −1 through the +4 nucleosome for each mutant relative to WT (Figure S5B). This analysis showed that the shifts were linearly cumulative, with slopes indicating that each successive nucleosome added 0.8, 2.5, or 2.1 bp to the mean spacing in H3-K56Q, pob3-Q308K, and the combined mutant, respectively.

Figure 5.

Effects of mutations on nucleosome positioning and occupancy. MNase-sequencing was performed in triplicate using strains with the pob3-Q308K mutation, hht1-K56Q hht2-K56Q (H3-K56Q), or all three mutations (see Table S1 for full genotypes). (A) Nucleosome midpoints were determined and aligned by the TSS for each of the 5337 genes. Vertical dashed lines indicate the accumulated shift at the fourth nucleosome downstream of the TSS in the H3-K56Q pob3-Q308K strain relative to WT. Additional analysis of this shift is presented in Figure S5, B–D in the supplemental material. (B) Nucleosome occupancy at each nucleotide was determined and averaged in 40 bins for each gene to produce the average normalized gene profile. Occupancy is also shown for the 250 bp US and DS of each gene without normalization. (C and D) As in (B), except genes were aligned by the TSS or by the last recognizable nucleosome in the transcription unit (the terminal nucleosome). Also see Figure S5, E–H in the supplemental material. DS, downstream; TSS, transcription start site; TTS, transcription termination site; US, upstream; WT, wild-type.

A potential explanation for this shift could be delayed redeposition of nucleosomes after passage of RNA polymerase. In support of this idea, the magnitude of the shift downstream correlated moderately with transcription frequency (Figure S5C). While the overall Spearman correlation coefficients for scatter plots of nucleosome shift vs. rlog₂(mRNA abundance) were 0.42, 0.41, and 0.30 (H3-K56Q, pob3-Q308K, and combined, respectively; Figure S5D), unpaired Student’s t-tests showed that the lowest quintile based on transcript level differed from all other quintiles for all strains with P-values of at least 10⁻⁹, indicating a significant effect of transcription frequency. While the normal pattern of nucleosome positioning has mainly been attributed to inherent properties of DNA sequences and the effects of ATP-dependent remodeling (Hughes et al. 2012; Parnell et al. 2015; Lai and Pugh 2017), these results indicate that the histone chaperone FACT and the status of H3-K56 also affect this property of chromatin, with at least a small compounding effect with the frequency of transcription.

FACT mutants might be predicted to cause either increased nucleosome occupancy (due to failed nucleosome destabilization) or decreased occupancy (due to failed tethering of nucleosomal components or lack of assembly activity). Therefore, we examined the distribution of nucleosome density (measured as the cumulative nucleosomal fragment depth where alignments were extended to 148 bp from each 5′ end) across a length-normalized gene body (Figure 5B). Occupancy increased at the 5′ ends of genes in pob3-Q308K mutants but decreased over 3′ ends in each single mutant, with additivity of the 3′ end effect in the pob3-Q308K H3-K56Q combined mutant. This pattern is not an artifact of gene averaging or normalization, as it is readily observable in browser tracks at most genes (see Figures S6B and S7, A and B). A defect in reassembly would cause loss of nucleosomes, but it is not clear why this effect would be more prominent over the 3′ ends of genes, with little correlation of the level of depopulation with transcription frequency (Figure S5E).

Our WT strain produced the expected decrease in nucleosome occupancy across the bodies of highly transcribed genes, but the effect was less pronounced in pob3-Q308K mutants (Figure S5F, compare the red traces; see Figure S5G for quantitation). If decreased occupancy in highly transcribed genes is caused by disruption of nucleosomes by encounters with RNA Pol II, then a failure to observe this effect could indicate that mutating FACT made nucleosome reassembly in the wake of transcription more efficient or that it decreased the frequency of transcription. However, (1) we observed an apparent delay in nucleosome redeposition in this mutant above, (2) acute withdrawal of FACT is associated with inefficient recovery of chromatin after transcription (Jamai et al. 2009), (3) the pob3-Q308K allele has a defect in nucleosome assembly/release in vitro (Figure 1, B and C), and (4) we did not observe a strong decrease in transcripts from this set of genes (Figure 3C and Figure S3C). The unexpected relative increase in nucleosome occupancy over frequently transcribed genes in a FACT mutant is therefore puzzling, but may point to a role for FACT in establishing and maintaining chromatin in processes that are independent of ongoing transcription (see the Discussion).

To remove potential distortion introduced by normalizing gene lengths, we aligned genes by their TSSs (Figure 5C) or by the nucleosome nearest to the transcription termination site (TTS) (the terminal nucleosome; Figure 5D). These plots confirmed that the 5′ increase/3′ decrease noted above is independent of normalization, that the effects extended to several nucleosomes near the TSS/TTS, and that the increased nucleosome occupancy observed in frequently transcribed genes is not an artifact of normalization (Figure S5H). The region immediately upstream of a yeast gene typically contains its promoter, and due to the dense packing of the yeast genome the region immediately downstream usually contains either the neighboring gene’s promoter or the TTSs for both genes. We parsed the data by the orientation of the downstream neighbor and found that this did affect the patterns of nucleosome occupancy at the 3′ ends of genes (Figure S5I), but that WT cells and mutants were similar, as were frequently and infrequently transcribed genes, so the orientation of the neighboring gene did not explain the 3′ end depopulation. H3-K56ac and FACT therefore collaborate to establish or maintain nucleosome occupancy over the 3′ ends of genes, but this does not appear to be a consequence of ongoing transcription or a mechanism for regulating it (see the Discussion).

The relationship between altered nucleosome occupancy and changes in transcript level

To examine the consequences of altered chromatin structure, we compared changes in nucleosome occupancy with changes in transcript output from individual genes. We calculated log₂FC values relative to WT for nucleosome occupancy in three regions of each gene: the nucleosome depleted region (NDR, defined here as −150 to −30 bp upstream of the TSS, typically containing the gene’s promoter), the +1 nucleosome region (−20 to +140 bp relative to the TSS), and the averaged gene body (TSS to TTS). We then plotted these values against the change in transcript level for each gene (Figure 6A).

Figure 6.

Changes in transcript level compared with changes in nucleosome occupancy in different regions of genes. (A) The log₂FC in nucleosome occupancy relative to the WT strain was determined for each mutant, then the average values were calculated for the NDR of each gene (150–30 bp upstream of the TSS, left panels), the +1 nucleosome region (20 bp upstream to 140 bp downstream of the TSS, middle panels), or the full gene body (right panels). These were plotted against the log₂FC in transcript level for each gene, with the Spearman correlation coefficients and regression line slopes shown. (B) Average nucleosome occupancies near the TSS were calculated as in Figure 5C and binned according to the log₂FC value for the change in transcript level in each mutant. Occupancy profiles are shown for genes with the largest decreases (1–10th percentile), the middle of the distribution (45–55th percentile), and the highest increases in transcript level (91–100th percentile), with the profile for all genes in the WT strain included for reference. log₂FC, normalized log₂ ratios relative to the WT strain; NDR, nucleosome depleted region; TSS, transcription start site; WT, wild-type.

Consistent with the expectation that nucleosomes in the promoter region would impair transcription, changes in occupancy of this region produced negative correlations with changes in transcript levels (increased nucleosome occupancy was associated with decreased transcript levels; Figure 6A, left panels). However, the correlations were low-to-moderate (r = −0.17 to −0.34). The +1 nucleosome marks the transition between initiation and elongation, so we expected that failure to establish or maintain this feature would affect transcript production significantly. It was therefore surprising that changes in occupancy over the TSS or the total gene body correlated poorly with changes in the level of transcripts (Figure 6A, middle and right panels).

To determine whether the genes experiencing the greatest changes in expression displayed any similarities with one another, we analyzed the nucleosome occupancy profiles for gene classes separated into deciles by transcript level changes. The decile with the greatest increase in transcripts in each case was a set of genes with weak depletion of nucleosomes upstream of the gene and lower occupancy at the +1 nucleosome site (Figure 6B and data not shown). The unusually high nucleosome occupancy upstream of these genes has been described as a closed promoter architecture and is associated with conditional transcription, higher occupancy by both coactivating and corepressing ATP-dependent remodelers, and high rates of nucleosome turnover (Parnell et al. 2015). Importantly, while this set of genes displayed increased transcript levels in the mutant strains, nucleosome occupancy in the NDR was unchanged and occupancy downstream of the TSS was only slightly reduced (Figure 6B, compare the solid and dotted red lines). A browser track of an example of a repressed gene in this class, CHA1, is shown in Figure S6A, illustrating the large increase in transcript level without a decrease in nucleosome occupancy.

In contrast, the decile of genes with the largest decreases in transcripts had stronger than average NDR depletion but, like the derepressed genes, also had lower than normal occupancy at the +1 nucleosome position (Figure 6B, blue lines). While the pob3-Q308K mutant had increased nucleosome occupancy in all regions for this set of genes, the other mutants did not (compare the solid and dotted blue lines). Changes in transcript levels in the mutants therefore aligned with promoter class and with chromatin architecture at the TSS, but the gross features of the architecture did not change dramatically and changes did not correlate well with changes in transcript accumulation. The interplay among FACT, H3-K56ac status, gross nucleosome occupancy, and transcript output therefore appears to be more complicated than anticipated by a simple barrier model, suggesting that the genomic tools used here may not be adequate to assess the precise features or dynamics important for proper control of transcription initiation.

H3-K56Q and H3-K56R have opposite effects on antisense transcription near the TSS

Acute FACT withdrawal causes elevated antisense transcripts (Feng et al. 2016). We also observed this effect in our FACT mutants as well as in H3-K56Q/R mutants (Figure 7). Antisense transcripts are likely to be unstable, so our measurements may underestimate their actual level, but we noted a distinct change in the spatial distribution of signal between the gene body and the region upstream of the TSS, with a maximum near the average position of the +1 nucleosome. This suggests that the +1 nucleosome might participate in a mechanism that blocks antisense transcription from progressing out of the gene body. Again, combining pob3-Q308K with H3-K56Q caused a more severe phenotype. Notably, while H3-K56Q and H3-K56R caused similar levels of antisense transcripts upstream of the TSS (to the left in Figure 7), they caused distinct effects near the +1 nucleosome and in genes. H3-K56R caused production of antisense signals to levels similar to those observed in the WT strain, while H3-K56Q aligned more closely with the FACT mutants. This suggests that the +1 nucleosome has an antisense barrier function, that stabilizing the histone–DNA contact (permanently with H3-K56R or conditionally by deacetylating H3-K56ac in normal strains) strengthens this obstacle, and that FACT activity is needed to maintain this feature. This is a novel function for the +1 nucleosome in yeast, but is consistent with its proposed activity in regulating sense transcription and with observations in human cells (Mayer et al. 2015). Browser tracks showing distinct profiles of antisense transcript accumulation are shown in Figure S7, A and B. In this context, FACT functions to enhance the barrier function of a specific nucleosome, the opposite of facilitating transcription through chromatin.

Figure 7.

Effects of mutations on antisense transcript signals near the TSS. Reads were parsed into sense and antisense for ∼5700 genes, and the log₂ value of the normalized number of RPM was calculated to determine coverage near the TSS. The approximate location of the average +1 nucleosome is indicated at 70 bp downstream of the TSS. RPM, reads per million fragments; TSS, transcription start site; WT, wild-type.

Discussion

The contact between histones and DNA at the nucleosomal entry/exit site has a key role during the reorganization and assembly cycle promoted by FACT, so altering the strength of this interaction by manipulating H3-K56 status was expected to affect FACT’s activities in vitro and its functions in vivo. We found that a permanent mimic of H3-K56ac that weakens this contact made reorganization less favorable in vitro, tipping equilibrium away from the open, nuclease-accessible form and toward nucleosome assembly (Figure 1). Altering H3-K56 status in vivo affected the phenotypes of FACT mutants, but the effects were complex and allele-specific (Figure 2). The results support the physiological relevance of the interaction between H3-K56Q status and FACT function, but show that this relationship is complicated. The range of effects caused by manipulating entry/exit point stability presumably reflects whether FACT’s capacity to destabilize or assemble nucleosomes is dominant for the specific process being studied.

A broad range of mutations had surprisingly similar effects on transcripts produced by individual genes (Figure 4). For example, H3-K56R and H3-K56Q have opposite effects on the strength of the H3-K56:DNA contact, but they affected transcript levels from the same genes in similar ways. Likewise, pob3-Q308K and spt16-11, two alleles of FACT with distinct defects in vitro and different profiles of genetic interactions in vivo (McCullough et al. 2011, 2013), affected transcript output from the same genes similarly. The effects of acute withdrawal of FACT on transcript levels were different from the more targeted defects, although combining FACT and H3-K56 mutations produced changes more like those caused by full loss of FACT. This suggests that the individual mutations impair distinct activities and that combining multiple mutations begins to approximate full loss-of-function. Notably, changes in transcript levels upon acute loss of FACT correlated well with the frequency of transcription, with more significant impact at frequently transcribed genes. This suggests that FACT has an important role in transcription initiation, with loss of FACT having greater consequences for genes that have shorter cycling times between initiation events. Simpler mutations that impaired specific activities of FACT did not have this effect unless combined, again suggesting that FACT and H3-K56 status have several roles, and that the initiation of transcription is not impacted until multiple functions are affected.

Initial models suggested that FACT’s ability to destabilize nucleosomes is required to support transcription elongation. However, FACT mutations do not appear to affect the rate of elongation in vivo (Mason and Struhl 2005; Biswas et al. 2006; Kuryan et al. 2012), and recent results show that FACT localizes to transcription units primarily as a response to exposure of its binding sites by ongoing transcription (Martin et al. 2018). In this emerging view, FACT has a more prominent role in recovering from chromatin disruptions caused by transcription than in supporting the progression of transcription. We examined the changes in transcript levels in our strains for evidence of defects in elongation, which we reasoned might be detected as inefficient completion of longer transcripts that are more dependent on the efficient progression of polymerase. Neither acute nor chronic defects displayed this effect. Other mechanisms might compensate for reduced progression of polymerase, but our results fail to provide support for a role for FACT or H3-K56ac as key transcription elongation factors. Our observation that a FACT mutation caused increased spacing of nucleosomes in proportion to the frequency of transcription could indicate that repeated disruption of nucleosomes, coupled with inefficient reassembly, leads to repositioning of the nucleosomes further downstream. Our results therefore provide further support for a role for FACT and its collaboration with H3-K56 status in restoring chromatin after transcription.

Our MNase-seq data revealed two changes in global nucleosome occupancy that are difficult to explain. First, FACT and H3-K56Q mutations each caused a general loss of nucleosomes over the 3′ ends of genes, with additive effects in the combined mutant. Histone modifications are differentially distributed across transcription units, and histone and FACT mutations have been observed to cause increased FACT occupancy over the 3′ ends of some genes (Duina et al. 2007). Therefore, 3′ end depopulation could be a consequence of an interaction between FACT and a histone modification, or with the process of transcription termination. However, the magnitude of the effect did not correlate well with either the normal level of transcription or changes in transcript levels, so the depopulation does not seem to be directly driven by transcription. Second, highly transcribed genes normally have lower nucleosome occupancy across their lengths, which is thought to be a result of displacement of the nucleosomes by repeated cycles of disruption during transcription. Surprisingly, this diminished nucleosome occupancy in frequently transcribed genes was less prominent in a FACT mutant, and this could not be explained by decreased expression of these genes.

To explain these puzzling results, we speculate that chromatin architecture might be the result of a dynamic process that includes contributions from FACT and H3-K56 status, in which nucleosome occupancy and positioning are established, and maintained locally irrespective of ongoing transcription. The principle that nucleosome occupancy and positioning are features of genes that can be altered prior to the initiation of transcription is well established for the NDR upstream of genes. We propose extending this idea to include the level of nucleosome occupancy across individual gene bodies, and suggest that FACT and H3-K56 have roles in establishing or maintaining this property. In this view, the low nucleosome occupancy in frequently transcribed genes is not necessarily a consequence of transcription, but instead could be a regulated feature that supports transcription, and our mutants were unable to maintain this feature. Further, we suggest that a specific mechanism could act to maintain normal levels of nucleosome occupancy over the 3′ ends of genes, whether or not the genes are being transcribed, and that FACT and H3-K56 status have a role in this process.

The largest increases in transcript levels in FACT mutants were associated with a class of genes with a closed architecture lacking a prominent NDR (Parnell et al. 2015). This suggests that FACT was important for keeping these genes in an “off” state by maintaining repressive chromatin. This is consistent with the known role of FACT in activating expression of the HO gene in yeast and the Oct4 gene in mouse cells (Biswas et al. 2006; Takahata et al. 2009a,b; Shakya et al. 2015), where FACT promotes the eviction of specific nucleosomes in promoters prior to activation. However, in contrast to our expectation, the profile of nucleosome occupancy over the promoters of the upregulated class of genes did not change dramatically. As is the case for HO and Oct4 expression, the effects of FACT on nucleosome occupancy might be transient, or they might involve subtle changes in positioning that are not captured in our population averages over hundreds of genes. In any case, we were unable to detect expected relationships between nucleosome occupancy and transcriptional regulation, as changes in transcript levels did not correlate well with changes in nucleosome occupancy.

Repressing cryptic transcription in both sense and antisense orientations depends on efficient restoration of chromatin after transcription (Kaplan et al. 2003; Hainer et al. 2011, 2012; Feng et al. 2016). We did not observe significant activation of a cryptic transcription reporter with the mutants described here (data not shown), but we did detect elevated antisense transcript levels with both FACT and H3-K56 mutants (Figure 7). Notably, while they had similar effects on mRNA levels overall, H3-K56Q and H3-K56R had different effects on antisense levels near the TSS. The results suggest that deacetylating H3-K56 to strengthen the histone–DNA contact enhances the ability of the +1 nucleosome to block antisense transcription from proceeding into the gene’s promoter, and that normal FACT activity is needed to support this barrier function. If this is the case, FACT makes a nucleosome a more potent barrier to polymerase elongation. Overall, our results support translating the acronym FACT more broadly as FAcilitates Chromatin Transactions.

FACT is essential for viability in many organisms, which has been attributed to central roles in both transcription and replication (Formosa 2012). However, emerging evidence indicates that differentiated cells of higher organisms remain viable and transcriptionally active without FACT (Gurova et al. 2018), and that its dominant role might involve maintaining the existing chromatin landscape and promoting transitions to new ones during cell fate changes (Shen et al. 2018). Our results support the importance of cooperation between FACT and H3-K56 status in establishing and maintaining chromatin architecture, but leave open questions regarding which physiological processes are impacted by this collaboration.