SWI/SNF senses carbon starvation with a pH-sensitive low-complexity sequence

  1. J Ignacio Gutierrez
  2. Gregory P Brittingham
  3. Yonca Karadeniz
  4. Kathleen D Tran
  5. Arnob Dutta
  6. Alex S Holehouse
  7. Craig L Peterson
  8. Liam J Holt  Is a corresponding author
  1. Department of Molecular and Cell Biology, University of California, Berkeley, United States
  2. Institute for Systems Genetics, New York University Grossman School of Medicine, United States
  3. Program in Molecular Medicine, University of Massachusetts Medical School, United States
  4. Department of Cell and Molecular Biology, University of Rhode Island, United States
  5. Center for Science and Engineering of Living Systems (CSELS), Washington University in St. Louis, United States
  6. Department of Biochemistry and Molecular Biophysics, Washington University School of Medicine, United States

Abstract

It is increasingly appreciated that intracellular pH changes are important biological signals. This motivates the elucidation of molecular mechanisms of pH sensing. We determined that a nucleocytoplasmic pH oscillation was required for the transcriptional response to carbon starvation in Saccharomyces cerevisiae. The SWI/SNF chromatin remodeling complex is a key mediator of this transcriptional response. A glutamine-rich low-complexity domain (QLC) in the SNF5 subunit of this complex, and histidines within this sequence, was required for efficient transcriptional reprogramming. Furthermore, the SNF5 QLC mediated pH-dependent recruitment of SWI/SNF to an acidic transcription factor in a reconstituted nucleosome remodeling assay. Simulations showed that protonation of histidines within the SNF5 QLC leads to conformational expansion, providing a potential biophysical mechanism for regulation of these interactions. Together, our results indicate that pH changes are a second messenger for transcriptional reprogramming during carbon starvation and that the SNF5 QLC acts as a pH sensor.

Editor's evaluation

This study has considerable merit in providing evidence that the Q-rich low-complexity domain in Snf5, and the histidine residues located therein, functions as a sensor of the drop in intracellular pH that accompanies glucose starvation to mediate SWI/SNF recruitment and transcriptional activation of the battery of genes derepressed under these conditions in order to reprogram carbon utilization. The work is multifaceted in combining yeast genetics, single-cell assays of gene expression and intracellular pH, genome-wide analysis of gene expression changes by RNA-seq, and in vitro biophysical analysis of activator-dependent SWI/SNF recruitment and nucleosome remodeling in a purified system.

https://doi.org/10.7554/eLife.70344.sa0

Introduction

Biological processes are inherently sensitive to the solution environment in which they occur. A key regulated parameter is intracellular pH (pHi), which influences all biological processes by determining the protonation state of titratable chemical groups. These titratable groups are found across many biological molecules, from small-molecule osmolytes to the side chains of amino acids. While early work suggested that pHi was a tightly constrained cellular parameter, more recent technologies have revealed that pHi can vary substantially in both space and time (Llopis et al., 1998; Seksek and Bolard, 1996). Moreover, changes in pHi can regulate metabolism (Busa and Nuccitelli, 1984; Young et al., 2010), development (Needham and Needham, 1997), proliferation (Busa and Crowe, 1983), and cell fate (Okamoto, 1994), among other processes. Intriguingly, stress-associated intracellular acidification appears to be broadly conserved, suggesting that a drop in pHi is a primordial mechanism to coordinate the general cellular stress response (Drummond et al., 1986; Gores et al., 1989; Munder et al., 2016; O’Sullivan and Condon, 1997; Triandafillou et al., 2020; Yao and Haddad, 2004).

The budding yeast Saccharomyces cerevisiae is adapted to an acidic external environment (pHe), and optimal growth media is typically at pH 4.0–5.5. The plasma membrane (Pma1) and vacuolar (Vma1) ATPases maintain near neutral pHi of ~7.8 by pumping protons out of the cell and into the vacuole, respectively (Martínez-Muñoz and Kane, 2008). When cells are starved for carbon, these pumps are inactivated, leading to a rapid acidification of the intracellular space to pH ~6 (Kane, 1995; Orij et al., 2009). This decrease in intracellular pHi is crucial for viability upon carbon starvation and is thought to conserve energy, leading to storage of metabolic enzymes in filamentous assemblies (Petrovska et al., 2014), reduction of macromolecular diffusion (Joyner et al., 2016; Munder et al., 2016), decreased membrane biogenesis (Young et al., 2010), and possibly the noncovalent crosslinking of the cytoplasm into a solid-like material state (Joyner et al., 2016; Munder et al., 2016). These studies suggest that many physiological processes are inactivated when pHi drops. However, some processes must also be upregulated during carbon starvation to enable adaptation to this stress. These genes are referred to as ‘glucose-repressed genes’ as they are transcriptionally repressed in the presence of glucose (DeRisi et al., 1997; Zid and O’Shea, 2014). Recently, evidence was presented of a positive role for acidic pHi in stress-gene induction: transient acidification is required for induction of the transcriptional heat-shock response in some conditions (Triandafillou et al., 2020). However, the molecular mechanisms by which the transcriptional machinery senses and responds to pH changes remain mysterious.

The Sucrose Non-Fermenting genes (SNF) were among the first genes found to be required for induction of glucose-repressed genes (Neigeborn and Carlson, 1984). Several of these genes were later identified as members of the SWI/SNF complex (Abrams et al., 1986; Carlson, 1987), an 11-subunit chromatin remodeling complex that is highly conserved from yeast to mammals (Chiba et al., 1994; Peterson et al., 1994; Peterson and Herskowitz, 1992). The SWI/SNF complex affects the expression of ~10% of the genes in S. cerevisiae during vegetative growth (Sudarsanam et al., 2000). Upon carbon starvation, most genes are downregulated, but a set of glucose-repressed genes, required for utilization of alternative energy sources, are strongly induced (Zid and O’Shea, 2014). The SWI/SNF complex is required for the efficient expression of several hundred stress-response and glucose-repressed genes, implying a possible function in pH-associated gene expression (Biddick et al., 2008a; Sudarsanam et al., 2000). However, we still lack evidence for a direct role for SWI/SNF components in the coordination of pH-dependent transcriptional programs or a mechanism through which pH sensing may be achieved.

10/11 subunits of the SWI/SNF complex contain large intrinsically disordered regions (Figure 1—figure supplement 1), and in particular, 4/11 SWI/SNF subunits contain glutamine-rich low-complexity sequences (QLCs). QLCs are present in glutamine-rich transactivation domains (Kadonaga et al., 1988; Kadonaga et al., 1987) some of which, including those found within SWI/SNF, may bind to transcription factors (Prochasson et al., 2003) or recruit transcriptional machinery (Geng et al., 2001; Janody et al., 2001; Laurent et al., 1990). Intrinsically disordered regions lack a fixed three-dimensional structure and can be highly responsive to their solution environment (Holehouse and Sukenik, 2020; Moses et al., 2020). Moreover, the SWI/SNF QLCs contain multiple histidine residues. Given that the intrinsic pKa of the histidine sidechain is 6.9 (Whitten et al., 2005), we hypothesized that these glutamine-rich low-complexity regions might function as pH sensors in response to variations in pHi.

In this study, we elucidate SNF5 as a pH-sensing regulatory subunit of SWI/SNF. SNF5 is over 50% disordered and contains the largest QLC of the SWI/SNF complex. This region is 42% glutamine and contains seven histidine residues. We investigated the relationship between the SNF5 QLC and the cytosolic acidification that occurs during acute carbon starvation. By single-cell analysis, we found that intracellular pH (pHi) is highly dynamic and varies between subpopulations of cells within the same culture. After an initial decrease to pHi ~ 6.5, a subset of cells recovered their pHi to ~7. This transient acidification followed by recovery was required for expression of glucose-repressed genes. The SNF5 QLC and four embedded histidines were required for rapid gene induction. SWI/SNF complex histone remodeling activity was robust to pH changes, but recruitment of the complex to a model transcription factor was pH-sensitive, and this recruitment was mediated by the SNF5 QLC and histidines within. All-atom simulations indicated that histidine protonation causes a conformational expansion of the SNF5 QLC, perhaps enabling interaction with a different set of transcription factors and driving recruitment to the promoters of glucose-repressed genes. Thus, we propose changes in histidine charge within QLCs as a mechanism to sense pH changes and instruct transcriptional reprograming during carbon starvation.

Results

Induction of ADH2 upon glucose starvation requires the SNF5 glutamine-rich low-complexity sequence with native histidines

The SWI/SNF chromatin remodeling complex subunit SNF5 has a large low-complexity region at its N-terminus that is enriched for glutamine, the sequence of which is shown in Figure 1A. This sequence contains seven histidine residues, and we noticed a frequent co-occurrence of histidines within and adjacent to glutamine-rich low-complexity sequences (QLCs) of many proteins. Inspection of the sequence properties of proteins, especially through the lens of evolution, can provide hints as to functionally important features. Therefore, we analyzed the sequence properties of all glutamine-rich low-complexity sequences (QLCs) in the proteomes of several species.

Figure 1 with 9 supplements see all
Efficient induction of ADH2 upon glucose starvation requires the SNF5 glutamine-rich low-complexity sequence with native histidines.

(A) Sequence of the N-terminal low-complexity domain of SNF5. This domain was deleted in the ΔQsnf5 strain. The glutamine-rich domain is highlighted in orange. The 4/7 histidines that were mutated to alanine in the HtoASNF5 allele are highlighted in red. (B) The log2 of the frequency of each amino acid within QLCs divided by the global frequency of each amino acid in the proteome (S. cerevisiae). Values > 0 indicate enrichment in QLCs. (C) Left: schematic of the SWI/SNF complex engaged with a nucleosome. The SNF5 C-terminus is shown in gray, while the disordered N-terminal QLC is shown in orange. Right: schematic of the three main SNF5 alleles used in this study. (D) RT-qPCR results assessing levels of endogenous ADH2 mRNA in four strains grown in glucose (left) or after 4 hr of glucose starvation (right). Note: y-axes are different for each plot. (E) Representative histograms (10,000 cells) showing the fluorescent signal from a PADH2-mCherry reporter gene for four strains grown in glucose (left) or after 6 hr of glucose starvation (right). Statistical tests are Bonferroni-corrected t-tests, *p<0.05, **p<0.01, n.s., not significant.

We defined QLCs as protein subsequences with a minimum of 25% glutamine residues, a maximum interruption between any two glutamine residues of 17 residues, and a minimum overall length of 15 residues. These parameters were optimized empirically based on the features of glutamine-rich regions in the S. cerevisiae proteme (see Materials and methods and Figure 1—figure supplement 2). By these criteria, the S288c S. cerevisiae strain had 144 QLCs (Supplementary file 1). We found that proline and histidine were enriched (>50–100%-fold higher than average proteome abundance) in yeast QLCs (Figure 1B), with similar patterns found in Dictyostelium discoideum, and Drosophila melanogaster proteomes (Figure 1—figure supplement 3). Enrichment for histidine within QLCs was previously described across many Eukaryotes using a slightly different method (Ramazzotti et al., 2012). Interestingly, the codons for glutamine are a single base pair mutation away from proline and histidine. However, they are similarly adjacent to lysine, arginine, glutamate, and leucine, yet QLCs are depleted for lysine, arginine, and glutamate, suggesting that the structure of the genetic code is insufficient to explain the observed patterns of amino acids within QLCs. We also considered the possibility that histidines might be generally enriched in low-complexity sequences. In fact, this is not the case: histidines are 50% more abundant in yeast QLCs than in all other low-complexity sequences identified using Wootton–Federhen complexity (see Materials and methods). Thus, histidines are a salient feature of QLCs.

The N-terminus of SNF5 contains one of the largest QLCs in the yeast proteome and is in the top 3 QLCs in terms of number of histidines (Figure 1—figure supplement 2E and F). We compared the sequences of Snf5 N-terminal domains taken from 20 orthologous proteins from a range of Ascomycota (a fungal phylum) (Figure 1—figure supplement 4, Supplementary file 2). Despite the relatively poor sequence conservation across the N-terminal disordered regions in SNF5 (Figure 1—figure supplement 4A), every region consisted of at least 18% glutamine (max 43%) and all possessed multiple histidine residues (Figure 1—figure supplement 4B, Supplementary file 2; the phylogeny considered and the total number of QLCs for each species are shown in Figure 1—figure supplement 4C). A broader survey of the tree of life (Figure 1—figure supplement 5) indicates that the SNF5 QLC was likely gained in the lineage leading to the Ascomycota and is not present in most Metazoa (animals). In summary, enrichment for glutamine residues interspersed with histidine residues appears to be a conserved sequence feature, both in QLCs, in general, and in the N-terminus of SNF5, in particular, implying a possible functional role (Zarin et al., 2019).

To further investigate the functional importance of the glutamine-rich N-terminal domain in SNF5, we engineered three SNF5 mutant strains: a complete deletion of the SNF5 gene (snf5Δ); a deletion of the N-terminal QLC (ΔQsnf5); and an allele with four histidines within the QLC mutated to alanine (HtoAsnf5) (Figure 1A and C).

As previously reported (Laurent et al., 1990), snf5Δ strains grew slowly (Figure 1—figure supplement 6A). In contrast, growth rates of ΔQsnf5 and HtoAsnf5 were similar to WT during continuous growth in either fermentable (glucose) or poor (galactose or galactose/ethanol) carbon sources (Figure 1—figure supplement 6A–D) and showed minimal defects when grown in glucose, carbon-starved for 24 hr, and then reinoculated into glucose media. However, a strong growth defect was revealed for ΔQsnf5 and HtoAsnf5 strains when cells were carbon-starved for 24 hr and then switched to a poor carbon source (Figure 1—figure supplement 6E and F), suggesting that the SNF5 QLC is important for adaptation to new carbon sources. Deletion of the SNF5 gene has been shown to disrupt the architecture of the SWI/SNF complex, leading to loss of other subunits (Peterson et al., 1994; Yang et al., 2007). To test if deletion of the QLC leads to loss of Snf5p protein or failure to incorporate into SWI/SNF, we immunoprecipitated the SWI/SNF complex from strains with a tandem affinity purification (TAP) tag at the C-terminal of the core SNF2 subunit. We found that the entire SWI/SNF complex remained intact in both the ΔQsnf5 and HtoAsnf5 strains (Figure 1—figure supplement 7A). Silver stains of the untagged Snf5p and Western blotting of TAP-tagged SNF5 (Puig et al., 2001) strains showed that all SNF5 alleles were expressed at similar levels to wild-type both in glucose and upon carbon starvation (Figure 1—figure supplement 7B). Together, these results show that deletion of the SNF5 QLC is distinct from total loss of the SNF5 gene and that this N-terminal sequence is important for efficient recovery from carbon starvation.

We hypothesized that slow recovery of ΔQsnf5 and HtoAsnf5 strains after carbon starvation was due to a failure in transcriptional reprogramming. The alcohol dehydrogenase ADH2 gene is normally repressed in the presence of glucose and strongly induced upon carbon starvation. This regulation depends on SWI/SNF activity (Peterson and Herskowitz, 1992). Therefore, we used ADH2 as a model gene to test our hypothesis. We assayed SWI/SNF occupancy at the ADH2 promoter by chromatin immunoprecipitation (ChIP) of SWI/SNF complexes with a TAP-tag on the C-terminus of the SNF2 subunit from strains with various SNF5 alleles, followed by quantitative PCR (qPCR). These experiments showed that the wild-type complex is robustly recruited to the ADH2 promoter upon carbon starvation (Figure 1—figure supplement 8). However, this recruitment is reduced in ΔQsnf5 and HtoAsnf5 strains.

Next, we assayed transcription of the ADH2 gene using reverse transcriptase quantitative polymerase chain reaction (RT-qPCR). We found that robust ADH2 expression after acute carbon starvation was dependent on the SNF5 QLC and the histidines within (Figure 1D). This defect was far stronger in the ΔQsnf5 and HtoAsnf5 strains than in snf5Δ strains; snf5Δ strains did not completely repress ADH2 expression in glucose and showed partial induction upon carbon starvation, while ΔQsnf5 strains tightly repressed ADH2 in glucose (similar to WT), but completely failed to induce expression upon starvation (Figure 1D). These results suggest a dual role for SNF5 in ADH2 regulation, both contributing to strong repression in glucose and robust induction upon carbon starvation. The ΔQsnf5 and HtoAsnf5 alleles separate these functions, maintaining WT-like repression while showing a strong defect in induction.

The RT-qPCR and ChIP assays report on the average behavior of a population. To enable single-cell analysis, we engineered a reporter strain with the mCherry (Shaner et al., 2004) fluorescent protein under the control of the ADH2 promoter integrated into the genome immediately upstream of the endogenous ADH2 locus (Figure 1E, Figure 1—figure supplement 9A). We found high cell-to-cell variation in the expression of this reporter in WT strains: after 6 hr of glucose starvation, PADH2-mCherry expression was bimodal; about half of the cells had high mCherry fluorescence and half were low. This bimodality was strongly dependent on preculture conditions and was most apparent upon acute withdrawal of carbon from early log-phase cells that had grown for >16 hr with optical density at 600 nm (O.D.) never exceeding 0.3 (see Materials and methods). If cells became partly saturated at any time during preculture, ADH2 induction was more rapid and uniform. Complete deletion of SNF5 eliminated this bimodal expression pattern; again, low levels of expression were apparent in glucose and induction during starvation was attenuated. As in the RT-qPCR analysis, the ΔQsnf5 strain completely failed to induce the PADH2-mCherry reporter at this time point and mutation of four central histidines to alanine was sufficient to mostly abrogate expression (Figure 1E). Mutation of a further two histidines had little additional effect (Figure 1—figure supplement 9B–D). Taken together, these results suggest that the dual function of SNF5 leads to switch-like control of ADH2 expression. In glucose, SNF5 helps repress ADH2. Upon carbon starvation, SNF5 is required for efficient induction of ADH2. The SNF5 QLC and histidine residues within seem to be crucial for switching between these states.

The SNF5 QLC is required for ADH2 expression and recovery of neutral pH

Multiple stresses, including glucose starvation, have been shown to cause a decrease in the pH of the cytoplasm and nucleus (nucleocytoplasm) (Dechant et al., 2014; Gores et al., 1989; Triandafillou et al., 2020; Yao and Haddad, 2004). Here, we refer to nucleocytoplasmic pH as intracellular pH (pHi). To investigate the relationship between ADH2 expression and pHi, and how these factors depend upon SNF5, we engineered strains bearing both the ratiometric fluorescent pH reporter, pHluorin (Miesenböck et al., 1998), and the PADH2-mCherry reporter. To calibrate the pHluorin sensor, we calculated the ratio of intensities of fluorescence emission after excitation with 405 and 488 nm light in cells that were ATP-depleted and permeabilized in media of known pH. We obtained a near linear relationship between ratios of fluorescence intensity and pH (Figure 2—figure supplement 1, Materials and methods). Therefore, these strains allowed us to simultaneously monitor pHi and expression of ADH2.

Wild-type cells growing exponentially in 2% glucose had a pHi of ~7.8. Upon acute carbon starvation, cells rapidly acidified to pHi ~ 6.5. Then, during the first hour, two populations arose: an acidic population (pHi ~ 5.5), and a second population that recovered to pHi ~ 7 (Figure 2A). Cells at pHi 7 proceeded to strongly induce expression of the PADH2-mCherry reporter, while cells at pHi 5.5 did not. We used fluorescence-activated cell sorting (FACS) to separate these two populations and found that cells that neither recovered neutral pH nor expressed the PADH2-mCherry reporter had lower fitness relative to the PADH2-mCherry-inducing population, as indicated by lower rates of proliferation on both rich and poor carbon sources, and lower tolerance of heat stress (Figure 2—figure supplement 2). After 8 hr of glucose starvation, >70% of wild-type cells had induced ADH2 (Figure 2A and C).

Figure 2 with 5 supplements see all
The SNF5 QLC is required for ADH2 expression and recovery of neutral pH.

(A) Representative flow cytometry for WT, ΔQsnf5, or HtoAsnf5 strains: the x-axis shows nucleocytoplasmic pH (pHi), while the y-axis shows fluorescence from the PADH2-mCherry reporter. Panels show cells grown in glucose (top) and then (second to bottom) after 0–8 hr of acute glucose starvation. Percentage of cells in each quadrant is indicated by gray numbers. (B) Schematic of quantification scheme: raw data from (A) was fit to a single or double Gaussian curve determined by a least-residuals method. (C) Quantification of pHi and PADH2-mCherry expression during acute starvation. The median of each Gaussian for pHi is plotted in (C, top), black and gray lines are from induced and uninduced populations, respectively. The height of bars in (C, bottom) indicates the fraction of maximal PADH2-mCherry reporter gene expression (WT cells, 8 hr glucose starvation) The darkness of the bars indicates the fraction of the population in the induced versus uninduced state. Mean and standard deviation of three biological replicates are shown.

We next analyzed cells harboring mutant alleles of the QLC of SNF5. Similarly to WT, both ΔQsnf5 and HtoAsnf5 strains rapidly acidified upon carbon starvation. However, these strains were defective in subsequent neutralization of pHi and in the expression of PADH2-mCherry. At the 4 hr time point, >95% of both ΔQsnf5 and HtoAsnf5 cells remained acidic with no detectable expression, while >60% of wild-type cells had neutralized and expressed mCherry (Figure 2A and C). Eventually, after 24 hr, the majority of mutant cells neutralized to pHi ~7 and induced expression of PADH2-mCherry (Figure 2—figure supplement 3). Again, complete deletion of SNF5 led to less severe phenotypes than the ΔQsnf5 and HtoAsnf5 alleles with only a modest delay in PADH2-mCherry expression (Figure 2—figure supplement 4), suggesting that SNF5 plays both activating and inhibitory roles in ADH2 expression. Thus, the SNF5 QLC and histidines within are required for the rapid dynamics of both transient acidification and transcriptional induction of PADH2-mCherry upon acute carbon starvation.

We hypothesized that mutant cells might fail to recover from acidification because transcripts controlled by SWI/SNF are responsible for pHi recovery. In this model, SWI/SNF drives expression of a set of genes that must be both transcribed and translated. To test this idea, we measured pHi in WT cells during carbon starvation in the presence of the cyclohexamine to prevent translation of new transcripts. In these conditions, we found that cells experienced a drop in pHi but were unable to recover neutral pH (Figure 2—figure supplement 5). Thus, new gene expression is required for recovery of pHi.

Transient acidification is required for ADH2 induction upon carbon starvation

The acidification of the yeast nucleocytoplasm has been shown to depend upon an acidic extracellular pH (pHe). We took advantage of this fact to manipulate the changes in pHi that occur upon carbon starvation. Cell viability was strongly dependent on pHe, decreasing drastically when cells were starved for glucose in media at pH ≥ 7.0 for 24 hr (Figure 3—figure supplement 1). Expression of PADH2-mCherry expression was also highly dependent on pHe, especially in SNF5 QLC mutants (Figure 3A, Figure 3—figure supplement 2). WT cells failed to induce PADH2-mCherry at pHe ≥ 7, but induced strongly at pHe ≤ 6.5. RT-qPCR showed similar behavior for the endogenous ADH2 transcript (Figure 3—figure supplement 3). ChIP experiments indicated that recruitment of SWI/SNF to the ADH2 promoter was also reduced when starvation was performed with media buffered to pHe 7.5 (Figure 1—figure supplement 7). Furthermore, we found that the nucleocytoplasm of all strains failed to acidify when the environment was held at pHe ≥ 7 (Figure 3—figure supplement 4). Therefore, we conclude that an acidic extracellular environment is required for a decrease in intracellular pH upon carbon starvation, and that this intracellular acidification is required for activation of ADH2 transcription.

Figure 3 with 4 supplements see all
Transient acidification is required for ADH2 induction upon carbon starvation.

(A) Expression of PADH2-mCherry reporter gene in WT, ΔQsnf5, or HtoAsnf5 strains 8 hr after acute carbon starvation in media titrated to various pH (pHe, see legend, right). Bar height indicates the fraction of maximal PADH2-mCherry reporter gene expression (WT cells, pHe 5.5). The darkness of the bars indicates the fraction of the population in the induced versus uninduced state (see legend, right). (B) Time courses of glucose starvation with media manipulations to perturb the intracellular pH response, either by changing media pH (pHe) or by adding sorbic acid. Top panels show nucleocytoplasmic pH (pHi), black and gray lines from induced and uninduced populations, respectively. Bottom panels quantify expression of the PADH2-mCherry reporter gene (as in A). All strains are WT except for the far-right panels, which are from a ΔQsnf5 strain.

Given that intracellular acidification is necessary for ADH2 promoter induction, we next wondered if it was sufficient. First, we used the membrane-permeable sorbic acid to allow intracellular acidification but prevent pHi recovery. These cells failed to induce PADH2-mCherry, indicating that nucleocytoplasmic acidification is not sufficient; subsequent neutralization is also required. Carbon starvation at pHe 7.4 prevented transient acidification and likewise prevented expression (Figure 3B, Figure 3—figure supplement 3). Cells that were first held at pHe 7.4, preventing initial acidification, and then switched to pHe 5, thereby causing late acidification, failed to express mCherry after 6 hr. Finally, starvation at pHe 5 for 2 hr followed by a switch to pHe 7.4, with a corresponding increase in pHi, led to robust PADH2-mCherry expression. Together, these results suggest that transient acidification immediately upon switching to carbon starvation followed by recovery to neutral pHi is the signal for the efficient induction of PADH2-mCherry.

Deletion of the SNF5 QLC leads to both failure to neutralize pHi and loss of ADH2 expression. We therefore wondered if forcing cells to neutralize pHi would rescue ADH2 expression in a ΔQsnf5 strain. This was not the case: the ΔQsnf5 strain still fails to express PADH2-mCherry, even if we recapitulate normal intracellular transient acidification (Figure 3B, right). Therefore, the SNF5 QLC is required for normal kinetics of transient acidification and for additional steps in ADH2 gene activation.

The SNF5 QLC and acidification of the nucleocytoplasm are required for efficient widespread transcriptional reprogramming upon carbon starvation

We wondered if transient acidification and the QLC of SNF5 were important for transcriptional reprogramming on a genome-wide scale. To test this, we performed Illumina RNA-sequencing analysis on triplicates of each strain (WT, ΔQsnf5, HtoAsnf5) either growing exponentially in glucose or after acute carbon starvation for 4 hr at pHe 5. In addition, to test the pH dependence of the transcriptional response, we analyzed WT strains carbon-starved at pHe 7, which prevents intracellular acidification (Figure 3B, Figure 3—figure supplement 4).

Principal component analysis showed tight clustering of all exponentially growing samples, indicating that mutation of the QLC of SNF5 does not strongly affect gene expression in rich media (Figure 4A). In contrast, there are greater differences between wild-type strains with mutant SNF5 alleles upon glucose starvation. The genes that accounted for most variation (the first two principal components) were involved in carbon transport, metabolism, and stress responses. We defined a set of 89 genes that were induced (greater than threefold) and 60 genes that were downregulated (greater than threefold) in WT strains upon starvation in media titrated to pHe 5. Many of these genes were poorly induced in ΔQsnf5 and HtoAsnf5 mutants, as well as in WT strains starved in media titrated to suboptimal pHe 7 (Figure 4B). Figure 4C and D show transcriptional differences between glucose-starved strains as volcano plots, emphasizing large-scale differences between WT and ΔQsnf5 strains, and similarities between ΔQsnf5 and HtoAsnf5.

The SNF5 QLC and acidification of the nucleocytoplasm are required for efficient widespread transcriptional reprogramming upon carbon starvation.

(A) Principal component (PC) analysis of three RNA-seq biological replicates for each condition tested. (B) Expression levels of genes that were greater than threefold induced or repressed upon carbon starvation in WT strains are plotted for each SNF5 allele. (C) Volcano plot showing the log2 ratio of expression levels in WT versus ΔQsnf5 strains (x-axis) and p-values for differential expression (y-axis). Genes with significantly different expression are indicated in red (log2 fold change > 1 and Wald test adjusted p-value<0.05). (D) Volcano plot as in (C) but comparing expression levels in HtoAsnf5 strains to ΔQsnf5 strains. (E) Hierarchically clustered heat map showing expression values of 149 genes with a significant change in expression upon starvation of WT cells (log2 fold change > 1 and Wald test adjusted p-value<0.05). Color code indicates gene expression relative to the mean expression of that gene across all strains and conditions, with red indicating high and blue indicating low values (see legend). Three biological replicates are shown for each experiment. Strain and condition identities are indicated at the bottom of each column. Four groups of genes with similar behavior are indicated to the left. Gene Ontology enrichment results for nine clusters of genes are shown to the right.

We next performed hierarchical clustering analysis (Euclidean distance) of the 149 genes that are strongly differentially expressed between strains or at suboptimal pHe 7 (Figure 4E). Based on this clustering and some manual curation, we assigned these genes to four groups. Group 1 genes (n = 42) were activated in starvation in an SNF5 QLC and pH-dependent manner. They are strongly induced in WT, but induction is attenuated both in mutants of the SNF5 QLC and when the transient acidification of pHi was prevented by starving cells in media titrated to pHe 7. Gene Ontology (GO) analysis revealed that these genes are enriched for processes that are adaptive in carbon starvation, for example, fatty acid metabolism and the TCA cycle. Group 2 (n = 64) genes were not strongly induced in WT, but were inappropriately induced during starvation in SNF5 QLC mutants and during starvation at pHe 7. GO analysis revealed that these genes are enriched for stress responses, perhaps because the failure to properly reprogram transcription leads to cellular stress. Group 3 genes (n = 51) were repressed upon carbon starvation in a pH-dependent but SNF5 QLC-independent manner. They were repressed in all strains, but repression failed at pHe 7. Finally, group 4 genes (n = 16) were repressed in WT cells in a pH-independent manner, but failed to repress in SNF5 QLC mutants.

We performed an analysis for the enrichment of transcription factors within the promoters of each of these gene sets using the YEASTRACT server (Teixeira et al., 2014). These enrichments are summarized in Supplementary file 3. Top hits for group 1 included the CAT8 and ADR1 transcription factors, which have previously been suggested to recruit the SWI/SNF complex to the ADH2 promoter (Biddick et al., 2008b).

In conclusion, both pH changes and the SNF5 QLC are required for correct transcriptional reprogramming upon carbon starvation, but the dependencies are nuanced. Mutation of the SNF5 QLC or prevention of nucleocytoplasmic acidification appears to trigger a stress response (group 2 genes). Another set of genes requires pH change for their repression upon starvation, but this pH sensing is independent of SNF5 (group 3). A small set of genes requires the SNF5 QLC but not pH change for repression upon starvation (group 4). Finally, a set of genes, including many of the traditionally defined ‘glucose-repressed genes,’ require both the SNF5 QLC and a pH change for their induction upon carbon starvation (group 1). For these genes, point mutation of four histidines in the QLC is almost as perturbative as complete deletion of the QLC. We propose that the SNF5 QLC senses the transient acidification that occurs upon carbon starvation to elicit transcriptional activation of this gene set. It is striking that this set is enriched for genes involved in catabolism, TCA cycle, and metabolism, given that these processes are important for energetic adaptation to acute glucose starvation.

The SNF5 QLC mediates a pH-sensitive transcription factor interaction in vitro

We reasoned that pHi changes could affect the intrinsic nucleosome remodeling activity of SWI/SNF or alternatively might impact the interactions of SWI/SNF with transcription factors. Indeed, recent structural evidence (He et al., 2021) shows that the QLCs of not only SNF5 but also several other SWI/SNF subunits appear to be poised for interaction with transcription factors on DNA immediately downstream of the nucleosome (Figure 5—figure supplement 1). We used a fluorescence-based strategy in vitro to investigate these potential pH-sensing mechanisms. A center-positioned, recombinant mononucleosome was assembled on a 200 bp DNA fragment containing a ‘601’ nucleosome positioning sequence (Dechassa et al., 2008; Figure 5A). The nucleosomal substrate contained two binding sites for the Gal4 activator located upstream and 68 base pairs of linker DNA downstream of the nucleosome. The mononucleosome contained a Cy3 fluorophore covalently attached to the distal end of the template DNA, and Cy5 was attached to the H2A C-terminal domain. The Cy3 and Cy5 fluorophores can function as a Förster resonance energy transfer (FRET) pair only when the Cy3 donor and Cy5 acceptor are within an appropriate distance (see also Li and Widom, 2004). In the absence of SWI/SNF activity, the center-positioned nucleosome has a low FRET signal, but ATP-dependent mobilization of the nucleosome toward the distal DNA end leads to an increase in FRET (Brune et al., 1994; Luger et al., 1999; Sen et al., 2017; Smith and Peterson, 2005; Zhou and Narlikar, 2016; Figure 5). In the absence of competitor DNA, SWI/SNF does not require an interaction with a transcription factor to be recruited to the mononucleosome and thus intrinsic nucleosome remodeling activity can be assessed independently of recruitment. In this assay, SWI/SNF complexes containing either ΔQsnf5p or HtoAsnf5 retained full nucleosome remodeling activity (Figure 5B–D), as well as full DNA-stimulated ATPase activity (Figure 5—figure supplement 2). Furthermore, these activities were similar at pH 6.5, 7, or 7.6. Thus, we conclude that the SNF5 QLC does not sense pH by modifying its intrinsic ATPase and nucleosome remodeling activity, at least in this in vitro context.

Figure 5 with 2 supplements see all
The SNF5 QLC mediates a pH-sensitive transcription factor interaction in vitro.

(A) Schematic: a Cy3 donor fluorophore was attached to one end of the DNA, and the histone H2A C-termini were labeled with a Cy5 acceptor fluorophore. ATP-dependent mobilization of the nucleosome to the DNA increases Förster resonance energy transfer (FRET), leading to increased emission at 670 nm. (B) Representative kinetic traces for WT (B), ΔQsnf5p (C), and HtoAsnf5 (D) SWI/SNF complexes at pH 7.6 (blue), 7.0 (green), or 6.5 (orange). There is no competitor DNA, so these traces indicate intrinsic remodeling activity without requirement for recruitment by transcription factors. (E) Schematic: in the presence of excess competitor DNA, SWI/SNF-dependent remodeling requires recruitment by a transcription factor (Gal4-VP16). (D) Representative kinetic traces for WT (F), ΔQsnf5p (G), and HtoAsnf5 (H) SWI/SNF complexes at pH 7.6 (blue), 7.0 (green), or 6.5 (orange). Inset on the WT panel (F) shows the first 100 s of the assay after ATP addition. All traces are averages of 2–4 experiments and represent FRET normalized to values prior to addition of ATP.

Next, we assessed if the SNF5 QLC and pH changes could affect SWI/SNF interactions with transcription factors. SWI/SNF remodeling activity can be targeted to nucleosomes in vitro by Gal4 derivatives that contain acidic activation domains, an archetypal example of which is VP16 (Yudkovsky et al., 1999). Indeed, it was previously demonstrated that the QLC of Snf5p mediates interaction with the Gal4-VP16 transcription factor (Prochasson et al., 2003). To assess recruitment of SWI/SNF, we set up reactions with an excess of nonspecific competitor DNA. In these conditions, there is very little recruitment and remodeling without interaction with a transcription factor bound to the mononucleosome DNA (Figure 5E and F). In this context, we found that the QLC of SNF5 was required for rapid, efficient recruitment of SWI/SNF by the Gal4-VP16 activator, and that the pH of the buffer affected this recruitment (Figure 5F). Within the physiological pH range (6.5–7.6), recruitment and remodeling increased with pH. This behavior might correspond to the recruitment of SWI/SNF to genes that are active at high pHi during growth in glucose. We predict that interactions with transcription factors at glucose-repressed genes would show the opposite behavior, that is, recruitment would be increased at lower pHi. SWI/SNF complexes deleted for the SNF5 QLC (containing ΔQsnf5p) had constitutively lower recruitment and were completely insensitive to pH changes over this same range (Figure 5G). SWI/SNF complexes containing HtoAsnf5p were even more defective that the ΔQsnf5 allele with respect to recruitment to the VP16 transcription factor (Figure 5H); this recruitment was barely above background levels at all pH values. Therefore, we conclude that the SNF5 QLC can sense pH changes by modulating interactions between SWI/SNF and transcription factors. Furthermore, these results suggest that the histidines within the SNF5 QLC must be present and deprotonated to enable interaction with VP16.

Protonation of histidines leads to conformational expansion of the SNF5 QLC

How might pH change be sensed by SNF5? As described above (Figure 1B), glutamine-rich low-complexity sequences (QLCs) are enriched for histidines, and they are also depleted for charged amino acids (Figure 1B). Charged amino acids have repeatedly been shown to govern the conformational behavior of disordered regions (Mao et al., 2010; Müller-Späth et al., 2010; Sorensen and Kjaergaard, 2019). Given that histidine protonation alters the local charge density of a sequence, we hypothesized that the charge-depleted QLCs may be poised to undergo protonation-dependent changes in conformational behavior. To test this idea, we performed all-atom Monte Carlo simulations to assess the conformational ensemble of a 50 amino acid region of the SNF5 QLC (residues 71–120) that contained three histidines, two of which we had mutated to alanine in our experiments (Figure 6A). We performed simulations with histidines in both uncharged and protonated states to mimic possible charges of this polypeptide at the pH found in the nucleocytoplasm in glucose and carbon starvation, respectively. These simulations generated ensembles of almost 50,000 distinct conformations (representative images shown in Figure 6B). To quantify conformational changes, we examined the radius of gyration, a metric that describes the global dimensions of a disordered region (Figure 6C). Protonation of the wild-type sequence led to a striking increase in the radius of gyration, driven by intramolecular electrostatic repulsions (Figure 6D, left). In contrast, when 2/3 histidines were replaced with alanines, no such change was observed (Figure 6D, right). For context, we also calculated an apparent scaling exponent (νapp), a dimensionless parameter that can also be used to quantify chain dimensions. This analysis showed that protonation of the wild-type sequence led to a change in νapp from 0.48 to 0.55, comparable to the magnitude of changes observed in previous studies of mutations that fundamentally altered intermolecular interactions in other low-complexity disordered regions (Martin et al., 2020; Sorensen and Kjaergaard, 2019). These results suggest that small changes in sequence charge density can elicit a relatively large change in conformational behavior. An analogous (albeit less pronounced) effect was observed for the second QLC subregion that we mutated (residues 195–233) (Figure 6—figure supplement 1). Taken together, our results suggest that charge-depleted disordered regions (such as QLCs) are poised to undergo pH-dependent conformational rearrangement. This inference offers the beginnings of a mechanism for pH sensing by SWI/SNF: the conformational expansion of the QLC sequence upon nucleocytoplasmic acidification may tune the propensity for SWI/SNF to interact with transcription factors (Figure 6E).

Figure 6 with 1 supplement see all
Protonation of histidines leads to conformational expansion of the SNF5 QLC.

(A) Schematic of the SNF5 gene (center) with the N-terminal QLC in orange and the two simulated peptides in dark orange. Sequences of the simulated peptides and identities of histidines mutated in both the HtoAsnf5 yeast strain and in simulations are indicated. (B) Representative images of conformations sampled in Monte Carlo all-atom simulations. (C) Cartoon depicting quantification of radius of gyration (Rg). (D) Radius of gyration (Rg, y-axis) of simulations of amino acids 71–120 of the SNF5 QLC with histidines either neutral (pH 7.4) or protonated (pH 5.0). Left two datasets are for the native peptide, right two datasets are with 2/3 histidines (H106 and H109) replaced with alanine, mimicking the HtoAsnf5 allele. Points represent the mean Rg from all conformations sampled in each independent simulation (beginning from distinct random initial conformers). Bars represent the mean values of all simulations. (E) Model of SWI/SNF regulation during carbon starvation. (Top) In glucose (pHi ~ 7.8), the SNF5 QLC is unprotonated. SWI/SNF is engaged by transcription factors that prevent transcription of glucose repressed genes or that activate other genes (TFA). (Middle) Upon acute carbon starvation, pHi drops to ~6.5, leading to protonation of histidines in the SNF5 QLC. Conformational expansion of the QLC may aid the release of SWI/SNF from some transcription factors (TFA) and potentially drive recruitment to others (not shown). (Bottom) As the cell adapts to carbon starvation, pHi neutralizes to ~7.0. Histidines within the SNF5 QLC may be partially protonated? The pKa of histidine is highly context-dependent. The QLC may aid recruitment of SWI/SNF to the promoters of glucose-repressed genes, thus leading to their expression.

Discussion

Intracellular pH changes occur in many physiological contexts, including cell cycle progression (Gagliardi and Shain, 2013), the circadian rhythm of crassulacean acid metabolism plants (Hafke et al., 2001), oxidative stress (van Schalkwyk et al., 2013), heat shock (Triandafillou et al., 2020), osmotic stress (Karagiannis and Young, 2001), and changes in nutritional state (Jacquel et al., 2020; Orij et al., 2009). However, the physiological role of these pHi fluctuations and the molecular mechanisms to detect them remain poorly understood. Prior results have emphasized the inactivation of processes in response to cytosolic acidification (Joyner et al., 2016; Munder et al., 2016; Petrovska et al., 2014). However, it is unclear how necessary modifications to the cell can occur if cellular dynamics are uniformly decreased. Much less has been reported regarding a potential role of fluctuations in pHi as a signal to activate specific cellular programs. In this work, we found that transient acidification is required for activation of glucose-repressed genes. Therefore, our work establishes a positive regulatory role for nucleocytoplasmic pH changes during carbon starvation.

Previous studies of intracellular state during glucose starvation based on population averages reported a simple decrease in pHi (Orij et al., 2009). In this work, we used single-cell measurements of both pHi and gene expression, and found that two coexisting subpopulations arose upon acute glucose starvation, one with pHi ~ 5.5 and a second at ~6.5. The latter population recovered to neutral pHi and then induced glucose-repressed genes, while the former remained dormant in an acidified state. We have not yet determined the mechanism that drives the bifurcation in pH response. It is possible that this bistability provides a form of bet-hedging (Levy et al., 2012) where some cells attempt to respond to carbon starvation, while others enter a dormant state (Munder et al., 2016). However, we have yet to discover any condition where the population with lower pHi and delayed transcriptional activation has an advantage. An alternative explanation is that these cells are failing to correctly adapt to starvation, perhaps undergoing a metabolic crisis, as suggested in a recent study (Jacquel et al., 2020).

It is becoming clear that intracellular pH is an important mechanism of biological control. It was previously shown that the protonation state of phosphatidic acid (PA) determines binding to the transcription factor Opi1, coupling membrane biogenesis and intracellular pH (Young et al., 2010). We focused our studies on the N-terminal region of SNF5 because it is known to be important for the response to carbon starvation and contains a large low-complexity region enriched in both glutamine and histidine residues. Histidines are good candidates for pH sensors as they can change protonation state over the recorded range of physiological pH fluctuations, and their pKa can be tuned substantially depending on local sequence context. Consistent with this hypothesis, we found that the SNF5 QLC and the histidines embedded within were required for transcriptional reprogramming.

Our in vitro assays showed that the intrinsic ATPase and nucleosome remodeling activities of SWI/SNF are robust to pH changes from 6.5 to 7.6. However, recruitment of the SWI/SNF complex by a model transcription factor (GAL4-VP16) was pH-sensitive, and this pH dependence was dependent on both the SNF5 QLC and the four central histidines within this domain. In this case, the recruitment by GAL4-VP16 was inhibited at pH 6.5. We speculate that low pHi favors release of SWI/SNF from activators that it is bound to in glucose conditions, and then the subsequent partial recovery in pHi could allow it to bind to a different set of activators (e.g., ADR1 and CAT8), thus recruiting it to genes that are expressed during starvation. This model is consistent with the requirement for both acidification and subsequent neutralization for expression of ADH2 (Figure 3). In principle, the conformational dynamics of the SNF5 QLC could be distinct at all three stages (Figure 6E). There are almost certainly additional pH-sensing elements of the transcriptional machinery that also take part in this reprogramming; multiple candidates are present among the of transcription factors that were enriched in our RNA-seq experiments (Supplementary file 3).

Low-complexity sequences, including QLCs, tend to be intrinsically disordered and therefore highly solvent exposed. A recent large-scale study of intrinsically disordered sequences showed that their conformational behavior is inherently sensitive to changes in their solution environment (Holehouse and Sukenik, 2020; Moses et al., 2020). Similarly, our simulations revealed that histidine protonation may lead the SNF5 QLC to expand dramatically. This provides a potential mechanism for pH sensing: upon acidification, histidines become positively charged, leading QLCs to adopt a more expanded state, perhaps revealing short linear interaction motifs (SLIMs), reducing the entropic cost of binding to interaction partners, preventing polar-mediated protein-protein interactions, or facilitating electrostatic-mediated contacts. The enrichment of histidines in QLCs hints that this could be a general, widespread mechanism to regulate cell biology in response to pH changes.

Glutamine-rich low-complexity sequences have been predominantly studied in the context of disease. Nine neurodegenerative illnesses, including Huntington’s disease, are thought to be caused by neurotoxic aggregation seeded by proteins that contain polyglutamines created by expansion of CAG trinucleotide repeats (Fan et al., 2014). However, polyglutamines and glutamine-rich sequences are relatively abundant in Eukaryotic cells: more than 100 human proteins contain QLCs, and the Dictyostelium and Drosophilid phyla have QLCs in ~10% and ~5% of their proteins, respectively (Schaefer et al., 2012). Furthermore, there is clear evidence of purifying selection to maintain polyQs in the Drosophilids (Huntley and Clark, 2007). This prevalence and conservation suggest an important biological function for these sequences. Recent work in Ashbya gossypii has revealed a role for QLC-containing proteins in the organization of the cytoplasm through phase separation into liquid droplets to enable subcellular localization of signaling molecules (Zhang et al., 2015). More generally, polyglutamine has been shown to drive self-association into a variety of higher-order assemblies, from fibrils to nanoscopic spheres to liquid droplets (Crick et al., 2013; Peskett et al., 2018; Posey et al., 2018). Taken together, these results imply that QLCs may offer a general mechanism to drive protein-protein interactions. In this study, we have identified a role for QLCs in the SWI/SNF complex as pH sensors. Our current model (Figure 6E) is that the SNF5 QLC partakes in heterotypic protein interactions that are modulated by protonation of histidines when the cell interior acidifies. However, we do not rule out the possibility for homotypic interactions and higher-order assembly of multiple SWI/SNF complexes.

All cells must modify gene expression to respond to environmental changes. This phenotypic plasticity is essential to all life, from single-celled organisms fighting to thrive in an ever-changing environment, to the complex genomic reprogramming that must occur during development and tissue homeostasis in plants and animals. Despite the differences between these organisms, the mechanisms that regulate gene expression are highly conserved. Changes in intracellular pH are increasingly emerging as a signal through which life perceives and reacts to its environment. This work provides a new role for glutamine-rich low-complexity sequences as molecular sensors for these pH changes.

Materials and methods

Key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
Gene (Saccharomyces cerevisiae)SNF5https://www.yeastgenome.org/SGD:S000000493
Gene (S. cerevisiae)SNF2https://www.yeastgenome.org/SGD:S000005816
Gene(pHluorin)pHluorindoi:10.1099/mic.0.022038-0
Strain, strain background (S. cerevisiae S288c)BY4741doi:https://doi.org/10.1002/(SICI)1097-0061(19980130)14:2<115::AID-YEA204>3.0.CO;2-2All strains used in this study are derived form BY4741
OtherLH3647ADH2::PADH2-mCherry-URA3 snf2::SNF2-TAP-His3MX6Yeast strain used to purify SWI/SNF complex
OtherLH3649ΔQsnf5-HIS3 ADH2::PADH2-mCherry-URA3 snf2::SNF2-TAP-kanMX6Yeast strain used to purify SWI/SNF complex containing ∆Qsnf5
OtherLH3652HtoAsnf5-HIS3 ADH2::PADH2-mCherry-URA3 snf2::SNF2-TAP-kanMX6Yeast strain used to purify SWI/SNF complex containing HtoAsnf5
Recombinant DNA reagentPlasmid (pRS316)GenBank: U03442Used to complement SNF5 gene in snf5∆ strains prior to removal using 5FOA
Recombinant DNA reagentPlasmid(pRS306)GenBank: U03438SNF5 and snf5 mutant alleles were all cloned into pRS306 and pRS303
Recombinant DNA reagentPlasmid(pRS303)GenBank: U03435SNF5 and snf5 mutant alleles were all cloned into pRS306 and pRS303
AntibodyRabbit polyclonal IgGSigmaCat# 12-370
AntibodyFluorescently labeled goat anti-rabbit polyclonalLI-COR BiosciencesCat# 926-68071Western blot (1:15,000 dilution)
AntibodyRabbit polyclonalanti-glucokinaseUS BiologicalCat# H2035-01Western blot (1:3000 dilution)
AntibodyFluorescently labeled goat anti-rabbit polyclonalLI-COR BiosciencesCat# 926-32211Western blot (1:15,000 dilution)

Yeast strains used in this study

Request a detailed protocol

All strains were derived from LH2145.

StrainGenotype
LH2145WT, Mat a from sporulation of BY4743: ura3∆0 his3∆0 leu22∆0 met15∆0
LH2090ΔQsnf5::kanMX6
LH2971SNF5-TAP-His3MX6
LH2973ΔQsnf5-TAP-His3MX6
LH2974HtoAsnf5-HIS3
LH2975HtoAsnf5-TAP-kanMX6
LH2991ADH2::PADH2-mCherry-URA3
LH2992ΔQsnf5-kanMX6 ADH2::PADH2-mCherry-URA3
LH2993HtoAsnf5-HIS3 ADH2::PADH2-mCherry-URA3
LH3486met15∆0 SNF5::kanMX6 (CEN/ARS-SNF5::URA3)
LH3513snf5Δ::kanMX6 ADH2::PADH2-mCherry-URA3 (CEN/ARS-SNF5::URA3)
LH3632snf5Δ::kanMX6 ADH2::PADH2-mCherry-URA3 TRP1::pHluorin-natMX6 (CEN/ARS-SNF5::URA3)
LH3647ADH2::PADH2-mCherry-URA3 snf2::SNF2-TAP-His3MX6
LH3649ΔQsnf5-HIS3 ADH2::PADH2-mCherry-URA3 snf2::SNF2-TAP-kanMX6
LH3652HtoAsnf5-HIS3 ADH2::PADH2-mCherry-URA3 snf2::SNF2-TAP-kanMX6
LH3705SNF5 ADH2::PADH2-mCherry-URA3 leu2::pHluorin-LEU2
LH3707ΔQsnf5::kanMX6 ADH2::PADH2-mCherry-URA3 leu2::pHluorin-LEU2
LH3713HtoAsnf5-HIS3 ADH2::PADH2-mCherry-URA3 leu2::pHluorin-LEU2

Plasmids used in this study

Request a detailed protocol
PlasmidIdentity
pLH226pFA6a- ΔQsnf5-GFP(S65T)-KANMX6
pLH416pFA6a-SNF5-GFP-KANMX6
pLH887pRS316-SNF5 (CEN/ARS plasmid)
pLH931pFA6a-4HtoAsnf5-KANMX6
pLH963pFA6a-SNF5-TAP-KANMX6
pLH964pFA6a-SNF5-TAP-HIS3MX6
pLH998pRS306-PADH2-mCherry
pLH1085pFA6a-6HtoAsnf5-HIS3MX6
pLH1093pFA6a-3’snf2-TAP-KANMX6
pLH1097pRS305-PTDH3-pHluorin
pLH1206pFA6a-3’snf2-TAP-NATMX

Cloning and yeast transformations

Request a detailed protocol

Yeast strains used in this study were all in the S288c strain background (derived from BY4743). The sequences of all genes in this study were obtained from the S. cerevisiae genome database (http://www.yeastgenome.org/).

We cloned the various SNF5 alleles into plasmids from the Longtine/Pringle collection (Longtine et al., 1998). We assembled plasmids by PCR or gene synthesis (IDT gene blocks) followed by Gibson cloning (Gibson et al., 2009). Then, plasmids were linearized and used to overwrite the endogenous locus by sigma homologous recombination using homology to both ends of the target gene.

The ΔQsnf5 gene lacks the N-terminal 282 amino acids that comprise a glutamine-rich low-complexity domain. Methionine 283 serves as the ATG for the ΔQ-SNF5 gene. In the HtoAsnf5 allele, histidines 106, 109, 213, and 214 were replaced by alanine using mutagenic primers to amplify three fragments of the QLC region, which were combined by Gibson assembly into an SNF5 parent plasmid linearized with BamH1 and Sac1.

We noticed that the slow growth null strain phenotype of the snf5Δ was partially lost over time, presumably due to suppressor mutations. Therefore, to avoid these spontaneous suppressors, we first introduced a CEN/ARS plasmid carrying the SNF5 gene under its own promoter and the URA3 auxotrophic selection marker. Then, a kanMX6 resistance cassette, amplified with primers with homology at the 5′ and 3′ of the SNF5 gene, was used to delete the entire chromosomal SNF5 ORF by homologous recombination. We subsequently cured strains of the CEN/ARS plasmid carrying WT SNF5 by negative selection against its URA3 locus by streaking for single colonies on 5-FOA plates immediately before each experiment to analyze the snf5Δ phenotype.

The PADH2-mCherry reporter was cloned into integrating pRS collection plasmids (Chee and Haase, 2012). URA3 (pRS306) or LEU2 (pRS305) were used as auxotrophic selection markers. The 835 base pairs upstream of the ADH2 gene were used as the promoter (PADH2). PADH2 and the mCherry ORF were amplified by PCR and assembled into linearized pRS plasmids (Sac1/Asc1) by Gibson assembly. These plasmids were cut in the middle of the ADH2 promoter using the Sph1 restriction endonuclease and integrated into the endogenous ADH2 locus by homologous recombination.

The pHluorin gene was also cloned into integrating pRS collection plasmids. URA3 (pRS306) and LEU2 (pRS305) were used for selection. The plasmid with the pHluorin gene was obtained as described in Orij et al., 2009. We amplified the pHluorin gene and the strong TDH3 promoter and used Gibson assembly to clone these fragments into pRS plasmids linearized with Sac1 and Asc1. Another strategy was to clone the pHluorin gene and a natMX6 cassette into the integrating pRS304 plasmid (that contains TRP1), which was then linearized within the TRP1 cassette using HindIII and integrated into the TRP1 locus.

A C-terminal TAP tag was used to visualize Snf5 and Snf2 proteins in Western blots. pRS plasmids were used, but the cloning strategy was slightly different. A 3′ fragment of the SNF5 and SNF2 genes was PCR amplified without the Stop codon. This segment does not contain a promoter or an ATG codon for translation initiation. The TAP tag was then amplified by PCR and cloned together with the 3′ of SNF5 and SNF2 ORFs by Gibson assembly into pRS plasmids with linearized Sac1 and Asc1. Plasmids were linearized in the 3′ of the SNF5 or SNF2 ORFs with StuI and XbaI, respectively, to linearize the plasmid allowing integration it into the 3′ of each gene locus by homologous recombination. Therefore, transformation results in a functional promoter at the endogenous locus fused to the TAP tag.

The SNF5-GFP strain was obtained from the yeast GFP collection (Huh et al., 2003), a gift of the Drubin/Barnes laboratory at UC Berkeley. The SNF2-GFP fused strain was made by the same strategy used for the TAP tagged strain above.

Supplementary files 6 and 7 list strains and plasmids generated in this study.

Culture media

Request a detailed protocol

Most experiments, unless indicated, were performed in synthetic complete (SC) media (13.4 g/L yeast nitrogen base and ammonium sulfate; 2 g/L amino acid mix and 2% glucose). Carbon starvation media was SC media without glucose, supplemented with sorbitol, a nonfermentable carbon source to avoid osmotic shock during glucose starvation (6.7 g/L YNB + ammonium sulfate; 2 g/L amino acid mix and 100 mM sorbitol). The pH of starvation media (pHe) was adjusted using NaOH.

Growth assays

Request a detailed protocol

Growth rates were determined in an Infinite M200 plate reader (Tecan) in 96-well microtiter plates using 200 μL total volume, cultured at 30°C and agitated at 800 rpm. Cells were pre-cultured overnight to log-phase (or subjected to other indicated pre-culture conditions) and then seeded at an A600 of 0.1 (based on a path length of ~0.3 cm) in SC media with various carbon sources. All measurements were performed in triplicate.

Glucose starvation

Request a detailed protocol

Cultures were incubated in a rotating incubator at 30°C and grown overnight (14–16 hr) to an OD between 0.2 and 0.3. Note that it is extremely important to prevent culture OD from exceeding 0.3, and results are different if cells are allowed to saturate and then diluted back. Thus, it is imperative to grow cultures from colonies on plates for >16 hr without ever exceeding OD 0.3 to obtain reproducible results. Typically, we would inoculate 3 mL cultures and make a series of 4–5 1/5 dilutions of this starting culture to be sure to catch an appropriate culture the following day. 3 mL of OD 0.2–0.3 culture were centrifuged at 6000 rpm for 3 min and resuspended in 3 mL starvation media (SC sorbitol at various pHe). This spin and resuspension was repeated two more times to ensure complete removal of glucose. Finally, cells were resuspended in 3 mL of starvation media. For flow cytometry, 200 μL samples were transferred to a well of a 96-well plate at each time point. During the course of time-lapse experiments, culture aliquots were set aside at 4°C. An LSR II flow cytometer with an HTS automated sampler was used for all measurements. 10,000 cells were analyzed at each time point.

Nucleocytoplasmic pH measurements

Request a detailed protocol

Nucleocytoplasmic pH (pHi) was measured by flow cytometry or microscopy. The ratiometric, pH-sensitive GFP variant, pHluorin, was used to measure pH based on the ratio of fluorescence from two excitation wavelengths. The settings used for LSR II flow cytometer were AmCyan (excitation 457, emission 491) and FITC (excitation 494, emission 520). AmCyan emission increases with pH, while FITC emission decreases. A calibration curve was made for each strain in each experiment. To generate a calibration curve, glycolysis and respiration were poisoned using 2-deoxyglucose and azide. This treatment leads to a complete loss of cellular ATP, and the nucleocytoplasmic pH equilibrates to the extracellular pH. We used the calibration buffers published by Patricia Kane’s group (Diakov et al., 2013): 50 mM MES (2-(N-morpholino) ethanesulfonic acid), 50 mM HEPES (4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid), 50 mM KCl, 50 mM NaCl, 0.2 M ammonium acetate, 10 mM sodium azide, 10 mM 2-deoxyglucose. Buffers were titrated to the desired pH with HCl or NaOH. Sodium azide and 2-deoxyglucose were always added fresh.

RT-qPCR

Request a detailed protocol

For qPCR and RNA-seq, RNA was extracted with the ‘High pure RNA isolation kit’ (Roche) following the manufacturer’s instructions. Three biological replicates were performed. cDNAs and qPCR were made with iSCRIPT and iTAQ universal SYBR green supermix by Bio-Rad, following the manufacturer’s instructions. Samples processed were exponentially growing culture (+Glu) or acute glucose starvation for 4 hr in media titrated to pH 5.5 or 7.5. Primers for qPCR were taken from Biddick et al., 2008a; for ADH2 and FBP1 genes: forward (GTC TAT CTC CAT TGT CGG CTC), reverse (GCC CTT CTC CAT CTT TTC GTA), and forward (CTT TCT CGG CTA GGT ATG TTG G), reverse (ACC TCA GTT TTC CGT TGG G). ACT1 was used as an internal control; primers were: forward (TGG ATT CCG GTG ATG GTG TT), reverse (TCA AAA TGG CGT GAG GTA GAG A).

RNA-sequencing

Request a detailed protocol

We performed RNA-sequencing analysis to determine the extent of the requirement for the SNF5 QLC in the activation of glucose-repressed genes. Three biological replicates were performed. Total RNA was extracted from WT, ΔQ-snf5, and HtoAsnf5 strains during exponential growth (+Glu) and after 4 hr of acute glucose starvation. In addition, WT strains were acutely starved in media titrated to pH 7. Next, poly-A selection was performed using Dynabeads and libraries were performed following the manufacturer’s indications. Sequencing of the 32 samples was performed on an Illumina HiSeq on two lanes. RNA-seq data were aligned to the University of California, Santa Cruz (UCSC), sacCer2 genome using Kallisto (0.43.0, http://www.nature.com/nbt/journal/v34/n5/full/nbt.3519.html) and downstream visualization and analysis was in R (3.2.2). Differential gene expression analysis, heat maps, and volcano plots were created using DESeq2. A Wald test was used to determine differentially expressed genes. Euclidean distance was used to calculate clustering for heat maps, with some manual curation to remove small clusters with no significant GO hits, and to consolidate clusters that had similar behavior. RNA-seq R-code can be found at https://github.com/gbritt/SWI_SNF_pH_Sensor_RNASeq., (copy archived at swh:1:rev:802f3d233210c02c66b745e414a6f7aa1385e379). RNA-seq datasets are deposited at GEO accession number GSE174687 (available here).

Western blots

Request a detailed protocol

Strains containing SNF5 and SNF2 fused to the TAP tag were used. Given the low concentration of these proteins, they were extracted with trichloroacetic acid (TCA): 3 mL culture was pelleted by centrifugation for 2 min at 6000 rpm and then frozen in liquid nitrogen. Pellets were thawed on ice and resuspended in 200 µL of 20% TCA, ~0.4 g of glass beads were added to each tube. Samples were lysed by bead beating four times for 2 min with 2 min of resting in ice in each cycle. Supernatants were extracted using a total of 1 mL of 5% TCA and precipitated for 20 min at 14,000 rpm at 4°C. Finally, pellets were resuspended in 212 µL of Laemmli sample buffer and pH adjusted with ~26 µL of Tris buffer pH 8. Samples were run on 7–12% gradient polyacrylamide gels with Thermo Fisher PageRuler prestained protein ladder 10–18 kDa. Proteins were transferred to a nitrocellulose membrane, which was then blocked with 5% nonfat milk and incubated with a rabbit IgG primary antibody (which binds to the protein A moiety of the TAP tag) for 1 hr and then with fluorescently labeled goat anti-rabbit secondary antibody IRDye 680RD goat-anti-rabbit (LI-COR Biosciences, Cat# 926-68071, 1:15,000 dilution). Anti-glucokinase was used as a loading control (rabbit-anti-Hxk1, US Biological, Cat# H2035-01, RRID:AB_2629457, Salem, MA, 1:3,000 dilution) followed by IRDye 800CW goat-anti-rabbit (LI-COR Biosciences, Cat# 926-32211, 1:15,000 dilution). Membranes were visualized using a LI-COR Odyssey CLx scanner with Image Studio 3.1 software. Fluorescence emission was quantified at 700 and 800 nM.

Co-immunoprecipitation of SWI/SNF complex

Request a detailed protocol

To evaluate the assembly state of the SWI/SNF complex, we immunoprecipitated Snf2p. To enable this experiment, we constructed strains in which the SNF2 gene was tagged at the C-terminus with a TAP tag (Puig et al., 2001). For each purification, 6 L of cells were grown in YPD to an OD of 1.2. Cells were broken open using glass beads in buffer A (40 mM HEPES [K+], pH 7.5, 10% glycerol, 350 mM KCl, 0.1% Tween-20, supplemented with 20 µg/mL leupeptin, 20 µg/mL pepstatin, 1 µg/mL benzamidine hydrochloride, and 100 µM PMSF) using a BioSpec bead beater followed by treatment with 75 units of benzonase for 20 min (to digest nucleic acids). Heparin was added to a final concentration of 10 µg/mL. The extract was clarified by first spinning at 15,000 rpm in a SS34 Sorvall rotor for 30 min at 4°C, followed by centrifugation at 45,000 rpm for 1.5 hr at 4°C in a Beckman ultracentrifuge. The soluble extract was incubated with IgG Sepharose beads for 4 hr at 4°C using gentle rotation. IgG Sepharose bound proteins were washed five times in buffer A and once in buffer B (10 mM Tris-HCl, pH 8.0, 10% glycerol, 150 mM NaCl, 0.5 mM EDTA, 0.1% NP40, 1 mM DTT, supplemented with 20 µg/mL leupeptin, 20 µg/mL pepstatin, 1 µg/mL benzamidine hydrochloride, and 100 µM PMSF). Bound protein complexes were incubated in buffer B with TEV protease overnight at 4°C using gentle rotation. The eluted protein was collected, CaCl2 was added to a final concentration of 2 mM and bound to Calmodulin Sepharose beads for 4 hr at 4°C using gentle rotation. Following binding, the protein-bound Calmodulin Sepharose beads were washed five times in buffer C (10 mM Tris-HCl, pH 8.0, 10% glycerol, 150 mM KCl, 2 mM CaCl2, 0.1% NP40, 1 mM DTT, supplemented with 20 µg/mL leupeptin, 20 µg/mL pepstatin, 1 µg/mL benzamidine hydrochloride, and 100 µM PMSF). The bound proteins were eluted in buffer D 10 mM Tris-HCl, pH 8.0, 10% glycerol, 150 mM KCl, 2 mM EGTA, 0.1% NP40, 0.5 mM DTT, supplemented with 20 µg/mL leupeptin, 20 µg/mL pepstatin, 1 µg/mL benzamidine hydrochloride, and 100 µM PMSF. The protein complexes were resolved by SDS-PAGE and visualized by silver staining.

Chromatin immunoprecipitation of SWI/SNF

Request a detailed protocol

For ChIP of the SWI/SNF complex, we constructed strains in which the SNF2 gene was tagged at the C-terminus with a TAP tag, as above. 1.25 × 108 cells were collected for each mutant and condition and fixed on 1% formaldehyde for 20 min to crosslink proteins to chromatin, and then the reaction was stopped with 136 mM glycine. Cells were pelleted and frozen in liquid nitrogen. Cells were then resuspended in 400 µL lysis buffer (0.1% deoxycholic acid, 1 mM EDTA, 50 mM HEPES pH 7.5, 140 mM NaCl, 1% Triton X-100, and 5 mM phenanthroline), mixed with 400 µL glass beads, and then lysed by vortexing for 15 min. The same lysis buffer was used to rinse the glass beads once more to recover remaining lysate. Lysates were then sonicated for 10 s, six times in ice to sheer chromatin, and then incubated with 40 µL of IgG-conjugated magnetic beads per sample (1 × 108 beads) and incubated for 24 hr at 4°C on a nutator (Dynabeads m-270 epoxy [Thermo Fisher 14301] conjugated to IgG from rabbit serum [Sigma-Aldrich I5006]; for conjugation protocol, see here).

After binding, samples were washed once with 600 µL buffer 2 (0.1 % deoxycholic acid, 1 mM EDTA, 50 mM HEPES pH 7.5, 500 mM NaCl, 1% Triton X-100, and 5 mM phenanthroline) and then washed once with 600 µL buffer 3 (0.5% deoxycholic acid, 1 mM EDTA, 250 mM LiCl, 0.5% NP-50, 10 mM Tris pH 7.9, and 5 mM phenanthroline), and finally washed once with 600 µL buffer TE.

The crosslinking between DNA and proteins was reversed by heating in elution buffer (50 mM Tris-HCl pH 7.5, 10 mM EDTA, and 1% SDS) for 2 hr at 42°C and then for 8 hr at 65°C. Eluted DNA was purified using QIAGEN kit (28104) according to the manufacturer’s instructions.

qPCR was performed using a Roche LightCycler 480 SYBR green master mix (04707516001) following the manufacturer’s instructions.

Two sets of primers were used to amplify for ADH2 (Parua et al., 2014):

  • 1F: ACC ATC CAC TTC ACG AGA CTG A, 1R:AAA AGT CGC TAC TGG CAC TC

  • 2F: GAG TGC CAG TAG CGA CTT TTT, 2R: ACT TGC CGT TGG ATT CGT AG

Data fitting

Request a detailed protocol

Fluorescence intensity from the PADH2-mCherry reporter and ratiometric fluorescence measurements from pHluorin were fit with a single or double Gaussian curve for statistical analysis using MATLAB (MathWorks). The choice of a single or double Gaussian fit was determined by assessing which fit gave the least residuals. For simplicity, the height (mode) of each Gaussian peak was used to determine the fraction of cells in each population rather than the area because peaks overlapped in many conditions.

Sequence analysis of QLCs

Identification of QLCs

Request a detailed protocol

QLCs were defined as subregions of the proteome that have an average fraction of glutamine residues of 25% or higher (minimum fraction), the maximum interruption between any two glutamine residues is less than 17 residues, and the whole QLC is at least 15 residues in length (minimum length) (Figure 1—figure supplement 2A). All XLCs (low-complexity subssequences for all amino acids, including glutamine) are provided online for further exploration and analysis (see GitHub). Secondly, systematic variation of the maximum interruption size to ask how the number of QLCs and number of residues found revealed that 17 residues was the value that maximized the number of QLCs and the number of residues found within QLCs, offering an optimally permissive value under the 0.25 or greater fraction of glutamine threshold.

Computation of per-residue conservation

Request a detailed protocol

Per-residue conservation was calculated by taking orthologous fungal proteins from the yeast genome order browser, aligning those using Clustal Omega, and calculating the Jensen–Shannon divergence as implemented by Caprah and Singh using the BLOSUM62 matrix (Byrne and Wolfe, 2005; Capra and Singh, 2007; Henikoff and Henikoff, 1992; Lin, 1991; Sievers et al., 2011).

Proteome-wide analysis

Request a detailed protocol

S. cerevisiae, Dictyostelium, Drosophila, and human proteins were obtained from UniProt. Sequence analysis was performed with SHEPHARD (https://shephard.readthedocs.io/). Predicted disorder scores, IDR identification, and predicted pLDDT scores were performed by metapredict (Emenecker et al., 2021; ). QLCs and full proteomes are provided at here.

Proteome-wide per-residue enrichment or depletion in QLCs

Request a detailed protocol

To compute the enrichment or depletion of specific amino acid residues in QLCs, we determined the fraction of non-glutamine residues in QLCs compared to the fraction of non-glutamine residues across the entire proteome. Specifically, for each proteome (S. cerevisiae, D. discoideum, D. melanogaster, and Homo sapiens) we first computed the proteome-wide background by taking the complete set of all protein sequences, removing all glutamine residues from all proteins, and then computing the fraction of the proteome made up of the remaining 19 amino acids. For each proteome, we then identified the full set of QLCs and repeated the analysis. The log2 of the ratio of the fraction of each amino acid in a QLC vs. across the proteome was used to compute enrichment or depletion for different amino acids within QLCs.

Proteome-wide per-residue enrichment or depletion of QLCs with respect to all XLCs

Request a detailed protocol

To compute enrichment of different amino acids in QLCs compared to other low-complexity domains (XLCs), we repeated the analysis above using XLCs defined by enrichment for non-glutamine residues, and then re-computed non-glutamine enrichment as was done for the whole proteome. The complete set of all XLC subsequences for all four proteomes is provided.

Nucleosome remodeling assays

SWI/SNF purification

Request a detailed protocol

SWI/SNF complexes were purified from yeast strains with a TAP protocol as previously described (Smith and Peterson, 2005). Cells were grown in YPAD media and harvested at OD600 = 3, and flash frozen and stored at –80°C. Yeast cells were lysed using a cryomill (PM100 Retsch). Ground cell powder was resuspended in E buffer (20 mM HEPES, 350 mM NaCl, 0.1% Tween-20, 10% glycerol, pH 7.5), with fresh 1 mM DTT and protease inhibitors (0.1 mg/mL phenylmethylsulfonyl fluoride, 2 µg/mL leupeptin, 2 µg/mL pepstatin, 1 mM benzamidine) and incubated on ice for 30 min. The crude lysate was clarified first by centrifugation 3K rpm for 15 min, and then 40K rpm for 60 min at 4°C. The clear lysate was transferred to a 250 mL Falcon tube and incubated with 400 µL IgG resin slurry (washed previously with E buffer without protease inhibitors) for 2 hr at 4°C. The resin was washed extensively with E buffer and protease inhibitors, and the protein-bound resin was incubated with 300 units TEV protease overnight at 4°C. The eluent was collected, incubated with 400 µL calmodulin affinity resin, washed previously with E buffer with fresh protease inhibitors, DTT and 2 mM CaCl2, for 2 hr at 4°C. Resin washed with the same buffer and SWI/SNF was eluted with E buffer with protease inhibitors, DTT, and 10 mM EGTA. The eluent was dialyzed in E buffer with PMSF, DTT, and 50 µM ZnCl2 at least three times. The dialyzed protein was concentrated with a Vivaspin column, aliquoted, flash frozen, and kept at –80°C. SWI/SNF concentration was quantified by electrophoresis on 10% SDS-PAGE gel alongside a BSA standard titration, followed by SYPRO Ruby (Thermo Fisher Scientific) staining overnight and using ImageQuant 1D gel analysis.

Mononucleosome reconstitutions

Request a detailed protocol

Recombinant octamers were reconstructed from isolated histones as described previously (Luger et al., 1999). In summary, recombinant human H2A (K125C), H2B, and H3 histones and Xenopus laevis H4 were isolated from Escherichia coli (Rosetta 2 [DE3] with and without pLysS). In order to label human H2A, a cysteine mutation was introduced at residue K125 via site-directed mutagenesis, which was labeled with Cy5 fluorophore attached to maleimide group (Zhou and Narlikar, 2016). DNA fragments were generated from 601 nucleosome positioning sequence and 2x Gal4 recognition sites with primers purchased from IDT. For FRET experiments, PCR amplification of labeled DNA fragments was as follows: 500 nM Cy3 labeled (5′-Cy3/TCCCCAGTCACGACGTTGTAAAAC-3′) and unlabeled primers (5′-ACCATGATTACGCCAAGCTTCGG-3′), 200 µM dNTPs, 0.1 ng/µL p159-2xGal4 plasmid kindly donated by Blaine Bartholomew, 0.02 U/µL NEB Phusion DNA polymerase, 1× Phusion High Fidelity Buffer. For ATPase assays, two unlabeled primers used (PrimerW: 5′-GTACCCGGGGATCCTCTAGAGTG-3′, PrimerS: 5′-GATCCTAATGACCAAGGAAAGCA-3′) under same PCR conditions with NEB Taq DNA Polymerase with 1× NEB ThermoPol Buffer. 400 nM fluorescently labeled and unlabeled mononucleosomes were reconstituted via salt gradient at 4°C with a peristaltic pump as described previously (Luger et al., 1999), with 600 mL high salt buffer (10 mM Tris-HCl, pH 7.4, 1 mM EDTA, 2 M KCl, 1 mM DTT) exchanged with 3 L of low salt buffer (10 mM Tris-HCl, pH 7.4, 1 mM EDTA, 50 mM KCl, 1 mM DTT) over 20 hr. The quality of the nucleosomes was checked by visualizing proteins on a 5% native-PAGE gel and scanning fluorescence ratios of labeled nucleosomes on an ISS PC1 spectrofluorometer.

FRET-based nucleosome remodeling

Request a detailed protocol

The fluorescence resonance energy transfer between Cy3-labeled DNA and Cy5-labeled octamer was used to measure the remodeling and recruitment activity of SWI/SNF using an ISS PC1 spectrofluorometer. The remodeling activity was measured by the increase in FRET signal in that occurred as a consequence of nucleosome sliding the DNA template. The reaction was performed under three different pH conditions: pH 6.5 (25 mM MES, 0.2 mM EDTA, 5 mM MgCl2, 70 mM KCl, 1 mM DTT), pH 7 (25 mM Tris, 0.2 mM EDTA, 5 mM MgCl2, 70 mM KCl, 1 mM DTT), and pH 7.6 (25 mM HEPES, 0.2 mM EDTA, 5 mM MgCl2, 70 mM KCl, 1 mM DTT). Remodeling reactions contained 2 nM or 4 nM (WT or mutant) SWI/SNF, 5 nM nucleosome, and 100 µM ATP or AMP-PNP. A 100 s pre-scan of the reaction was taken before the reaction started and the time-dependent fluorescence measurements started after addition of ATP or AMP-PNP for 1000s at room temperature. Similarly, recruitment assays were performed in three different buffer conditions: pH 6.5, pH 7, and pH 7.6. The recruitment assays contained 2 nM or 4 nM (WT or mutant) SWI/SNF, 5 nM nucleosome, 4 nM competitor DNA, 100 µM Gal4–VP16 (Protein One, P1019-02) and 100 µM ATP or AMP-PNP, together with respective controls (Sen et al., 2017). 100 s of pre-scans and 1000s of time-dependent enzyme kinetics were measured. At least 2–4 kinetic traces were collected per reaction. Data were normalized to their respective pre-scans to account for variation between reactions. The time-dependent FRET signals were excited at 530 nm and measured at 670 nm. The data analysis was performed using the OriginLab software package.

ATPase activity measurements

Request a detailed protocol

7-Diethylamino-3-[N-(2-maleimidoethyl)-carbamoyl]-coumarin-conjugated phosphate binding protein A197C (MDCC-PBP) (Brune et al., 1994) was used to detect inorganic phosphate (Pi) release from ATPase activity in real time. Before the reaction, ATP was cleared of free Pi by performing a mopping reaction. In order to mop the ATP, 10 mM ATP was incubated with 1 U/mL PNPase (Sigma, N2415-100UN) and 200 µM 7-methylguanosine (Sigma, M0627-100MG) in mopping buffer (25 mM HEPES, 75 mM NaCl, 5 mM MgCl2, 1 mM DTT) for 2 hr at room temperature. ATPase assay reaction conditions were 2 nM SWI/SNF, 5 nM nucleosome, and 100 µM ATP in respective pH buffers; pH 6.5 (25 mM MES, 0.2 mM EDTA, 5 mM MgCl2, 70 mM KCl, 1 mM DTT), pH 7 (25 mM Tris, 0.2 mM EDTA, 5 mM MgCl2, 70 mM KCl, 1 mM DTT), or pH 7.6 (25 mM HEPES, 0.2 mM EDTA, 5 mM MgCl2, 70 mM KCl, 1 mM DTT). The measurements were performed on a Tecan Infinite 1000, with excitation at 405 nm and emission at 460 nm. Pre-scan measurements were taken to detect the basal level of signal per reaction. The time-dependent measurements were taken after starting the reaction by ATP addition. At least 3–4 kinetic traces were analyzed using the steady-state equation using GraphPad Prism 8 software.

All-atom simulations

Request a detailed protocol

All-atom simulations were run with the ABSINTH implicit solvent model and CAMPARI Monte Carlo simulation (V2.0; http://campari.sourceforge.net/; Vitalis and Pappu, 2009). The combination of ABSINTH and CAMPARI has been used to examine the conformational behavior of disordered proteins with good agreement to experiment (Cubuk et al., 2020; Fuertes et al., 2017; Martin et al., 2020).

All simulations were started from randomly generated nonoverlapping random-coil conformations, with each independent simulations using a unique starting structure. Monte Carlo simulations perturb and evolve the system via a series of moves that alter backbone and sidechain dihedral angles, as well as rigid-body coordinates of both protein sequences and explicit ions. Simulation analysis was performed using CAMPARITraj (http://www.ctraj.com/) and MDTraj (McGibbon et al., 2015).

ABSINTH simulations were performed with the ion parameters derived by Mao et al. and using the abs_opls_3.4.prm parameters (Mao et al., 2010). All simulations were run at 15 mM NaCl and 325 K, a simulation temperature previously shown to be a good proxy for bona fide ambient temperature (Das et al., 2016; Martin et al., 2020). A summary of the simulation input details is provided in Supplementary file 5. For SNF571-120 simulations, 20 independent simulations were run for each combination of pH (as defined by histidine protonation state) and mutational state. For SNF5195-223, the high glutamine content made conformational sampling challenging, as has been observed in previous glutamine-rich systems, reflecting the tendency for polyglutamine to undergo intramolecular chain collapse (Crick et al., 2006; Newcombe et al., 2018; Warner et al., 2017). To address this challenge, we ran hundreds of short simulations (with a longer equilibration period than in SNF71-120) that are guaranteed to be uncorrelated due to their complete independence (Vitalis and Caflisch, 2010). Simulation code and details can be found at https://github.com/holehouse-lab/supportingdata/tree/master/2021/Gutierrez_QLC_2021.

Bioinformatic analysis

Request a detailed protocol

All protein sequence analyses were performed with localCIDER, with FASTA files read by protfasta (https://github.com/holehouse-lab/protfasta; Holehouse et al., 2017; Holehouse, 2021). Sequence alignments were performed using Clustal Omega (Sievers et al., 2011). Sequence conservation was computed using default properties in with the score_conservation program as defined by Capra and Singh, 2007. Proteomes were downloaded from UniProt Consortium, 2015.

Low-complexity sequences were identified using Wootton-Federhen complexity (Ginell and Holehouse, 2020; Wootton and Federhen, 1993). Sequence complexity is calculated over a sliding window size of 15 residues, and a threshold of 0.6 was used for binary classification of a residue as ‘low’ or ‘high’ complexity. After an initial sweep, gaps of up to three ‘high-complexity residues’ between regions of low-complexity residues were converted to low-complexity. Finally, contiguous stretches of 30 residues or longer were taken as the complete set of low-complexity regions in the proteome. The full set of those SEG-defined LCDs for human, Drosophila, Dictyostelium, and Cerevisiae proteomes is provided as FASTA files available here.

Data availability

Simulation code and details can be found at: https://github.com/holehouse-lab/supportingdata/tree/master/2021/Gutierrez_QLC_2021, (copy archived at swh:1:rev:bafdd4e42c496ecdf1da134c9ca5ac709b273ae5; path=/2021/Gutierrez_QLC_2021/) RNA-seq R-code can be found at: https://github.com/gbritt/SWI_SNF_pH_Sensor_RNASeq, (copy archived at swh:1:rev:802f3d233210c02c66b745e414a6f7aa1385e379) datasets are deposited at GEO accession number GSE174687 https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE174687.

The following data sets were generated
    1. Brittingham GP
    2. Holt LJ
    3. Gutierrez JI
    (2021) NCBI Gene Expression Omnibus
    ID GSE174687. SWI/SNF senses carbon starvation with a pH-sensitive low complexity sequence.

References

    1. Okamoto YKK
    (1994)
    Cytoplasmic Ca+2 and H+ concentrations determine cell fate in Dictyostelium discoideum
    Access 28:2423–2427.
    1. Seksek O
    2. Bolard J
    (1996)
    Nuclear pH gradient in mammalian cells revealed by laser microspectrofluorimetry
    Journal of Cell Science 109 (Pt 1):257–262.

Decision letter

  1. Alan G Hinnebusch
    Reviewing Editor; Eunice Kennedy Shriver National Institute of Child Health and Human Development, United States
  2. Kevin Struhl
    Senior Editor; Harvard Medical School, United States

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

[Editors' note: this paper was reviewed by Review Commons.]

Decision letter after peer review:

Thank you for submitting your article "SWI/SNF senses carbon starvation with a pH-sensitive low complexity sequence" for consideration by eLife. We have evaluated the three reviews from Reviews Common and your rebuttal, and also carefully read the manuscript ourselves. The Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

1. In addition to making all of the proposed revisions and additions of new analysis described in the authors' rebuttal letter to the three Review Commons reviews, the following additional experiments are required:

2. As requested by more than one of the referees (#1 and #2), it is necessary to extend the in vitro analysis of Figure 5 to include the HtoA variant of Snf5. As the authors noted, the N-terminal domain of Snf5 was shown previously to interact with the VP16 activation domain. It is critical to show that the histidine residues in this domain are key to this interaction using their elegant in vitro assay for SWI/SNF recruitment. This request does not involve simply "investigating more subtle alleles", as the effect of the HtoA variant is central to their model; and it does not involve developing a new reconstitution assay-at odds with the main arguments in the authors' rebuttal for declining to conduct this key requested experiment.

3. It is necessary to demonstrate by ChIP assay that recruitment of SWI/SNF to the key target gene ADH2 in response to carbon starvation is impaired by the Snf5 HtoA mutation and by an external pH of 7 in WT cells to provide in vivo evidence that protonation of the His residues in the Snf5 QLC stimulates SWI/SNF recruitment to a target promoter activated during glucose starvation. This is particularly important because recruitment of SWI/SNF by Gal4-VP16 was lower at pH6.5 versus pH7.0 in the in vitro recruitment assay of Figure 5, and thus does not mimic the effect of pH on expression of ADH2 and other genes induced by glucose starvation in yeast cells, as was noted by Ref. #1. This requested ChIP experiment falls far short of the ChIP-seq and ATAC-sec or MNase ChIP-seq experiments requested by Ref. #3 in terms of time and expense, while providing essential support for the proposed mechanistic model.

https://doi.org/10.7554/eLife.70344.sa1

Author response

Essential revisions:

1. In addition to making all of the proposed revisions and additions of new analysis described in the authors' rebuttal letter to the three Review Commons reviews, the following additional experiments are required:

2. As requested by more than one of the referees (#1 and #2), it is necessary to extend the in vitro analysis of Figure 5 to include the HtoA variant of Snf5. As the authors noted, the N-terminal domain of Snf5 was shown previously to interact with the VP16 activation domain. It is critical to show that the histidine residues in this domain are key to this interaction using their elegant in vitro assay for SWI/SNF recruitment. This request does not involve simply "investigating more subtle alleles", as the effect of the HtoA variant is central to their model; and it does not involve developing a new reconstitution assay-at odds with the main arguments in the authors' rebuttal for declining to conduct this key requested experiment.

We have completed these additional in vitro experiments. We found that the SWI/SNF complex containing the HtoA Snf5p variant was fully functional in terms of basal chromatin remodeling activity (new figure 5D) but is completely defective in terms of recruitment to the VP16 transcription activation domain (new figure 5H.) We interpret this result to mean that the Snf5 QLC requires unprotonated histidines to be present to allow interaction with VP16, and that alanines are not able to substitute for this interaction. This experiment confirms the importance of the four central histidines in the Snf5 QLC for interaction with transcription factors.

3. It is necessary to demonstrate by ChIP assay that recruitment of SWI/SNF to the key target gene ADH2 in response to carbon starvation is impaired by the Snf5 HtoA mutation and by an external pH of 7 in WT cells to provide in vivo evidence that protonation of the His residues in the Snf5 QLC stimulates SWI/SNF recruitment to a target promoter activated during glucose starvation. This is particularly important because recruitment of SWI/SNF by Gal4-VP16 was lower at pH6.5 versus pH7.0 in the in vitro recruitment assay of Figure 5, and thus does not mimic the effect of pH on expression of ADH2 and other genes induced by glucose starvation in yeast cells, as was noted by Ref. #1. This requested ChIP experiment falls far short of the ChIP-seq and ATAC-sec or MNase ChIP-seq experiments requested by Ref. #3 in terms of time and expense, while providing essential support for the proposed mechanistic model.

We have completed these additional ChIP experiments (new Figure 1—figure supplement 8). We find that recruitment of SWI/SNF to the ADH2 promoter is impaired in ΔQsnf5 and HtoAsnf5 strains and when extracellular pH is buffered to 7.5. These results are consistent with our hypothesis that protonation of the SNF5 QLC is important for recruitment to this glucose-repressed promoter.

[Editors' note: we include below the reviews that the authors received from Review Commons, along with the authors’ responses.]

Major experiments and analyses added to the new manuscript in response to Review Commons Reviews:

Regarding bioinformatic analyses of amino-acid enrichment in glutamine-rich low-complexity sequences (QLCs) and of the evolution of low-complexity sequences in SWI/SNF.

– The operational definition of QLCs was somewhat arbitrary. We have updated our analysis using an optimized definitions of QLC to better understand amino-acid enrichments throughout evolution. The parameter optimization is show in new Figure 1—figure supplement 2.

– We previously only explored the Snf5 QLC in the Ascomycota. In the revised manuscript, we expanded our analysis to explore Eukaryotic diversity more broadly. These results are presented in the new Figure 1—figure supplement 5. These results indicate that the SNF5 QLC was gained in the fungal lineage that includes the Basidiomycetes and Ascomycota.

Regarding the differences between SWI/SNF and human BAF complex.

– We only really talked about human BAF in one paragraph of the discussion. We have completely deleted this paragraph in the current version. We have also added an annotated diagram comparing the human BAF and yeast SWI/SNF complexes indicating the positions of the QLCs and other major low complexity sequences – new Figure 5 —figure supplement 1.

Responses to individual reviews:

Reviewer #1 (Evidence, reproducibility and clarity (Required)):

Gutiérrez et al. report an intriguing set of experimental results on the transcriptional role of SWI/SNF subunit Snf5 in pH sensing in the context of carbon starvation. More generally, the authors propose that histidine residues embedded in polyglutamine stretches sense pH and mechanistically transduce this as protein conformational changes that,in their system, affect TF interactions with polyglutamine stretches present in coTF subunits.

Overall the data are solid and the paper is very well written. Still, the discussion is a little lengthy, including emphasis on extrapolation of the yeast results to human cells, which reads more like a grant proposal than a discussion.

We are happy to reduce the discussion. We have deleted the paragraph that speculates regarding relevance to human BAF complex in cancer.

Although somewhat speculative, the conclusions of the authors are supported by high-quality data. I therefore support publication of a revised manuscript.

We thank the reviewer.

Minor:

– Mediator – ADR1 interactions are independent of SWI/SNF according to PMID: 18250152 which is consistent with the author's observations. Could this aspect of ADH2 transcriptional regulation be discussed?

Our interpretation of this study is that ADH2 induction becomes independent of SWI/SNF only if ADR1 is artificially fused to a mediator subunit. The results of the study are consistent with our model.

– What about RNA stability to explain the results? PMID: 26667037

This study shows that elements in the ADH2 promoter influence mRNA stability, particularly leading to destabilization of the transcript in the presence of glucose. We don’t explore the inactivation of ADH2 in glucose here, but it could be interesting to think about in the future.

– In relation to the speculations of the authors in the discussion; do yeast recover their pH if put back into 2% glucose? Could cell sorting be applied to study the short and long-term fate of the low and hi pH subpopulations.

We have sorted cells to see if there are any differences in fate for the high and low pH subpopulations. We have found that the high pH cells (cells that recover pH) are always better, in terms of fitness and heart stress tolerance, than the low pH cells. We have added discussion of these data (new Figure 2 supplement 2).

– The RNAseq data should be deposited in a public database (GEO or SRA).

We have deposited the data in the GEO database with accession number GSE174687.

Figures:

Figure 2C, 3B black vs grey lines? It is not immediately obvious from the legends of Figures 2 and 3 that the grey and black lines concern the low and high pHi subpopulation.

We have clarified this point in the legend.

Figure 5: 5B shows that the DeltaQsnf5 SWI/SNF complex is pH sensitive, even though it lacks the proposed pH sensor domain? 5D I would have expected the low pH to favor Gal4-VP16 recruitment, if this reconstitution of TF-mediated recruitment mimics the situation at ADH2. The result is the opposite. This apparently paradoxical result is only addressed in the discussion, but is somewhat misleading in the results parts. Importantly, why did the authors not assay the 4HtoASNF5 version of SWI/SNF in vitro?

In figure 5B (now figure 5C), there is no significant effect of pH on the basal rate of nucleosome remodeling for either WT or dQ-snf5.

In figure 5D (now figure 5F), we don’t think this is a paradox. We believe that the change in structure of the SNF5 QLC favors interactions with some transcription factors and disfavors others. We have addressed this point in the Results section to improve readability.

We performed the assay with the 4HtoA allele (figure 5D and H). We found that the intrinsic activity of this complex was equal to WT and (similar to WT) unaffected by pH (figure 5D). Interestingly, the 4HtoA was even more defective that the dQ allele with respect to recruitment to the VP16 transcription factor (figure 5H). This recruitment was barely above background levels at all pH values.

“development and tissue homeostasis in plants and metazoa.” > plants and animals.

We made this change to improve readability.

…“to its environment. This work provides a new role for glutamine-rich low-complexity sequences as molecular sensors for these pH changes.” > histidine-bearing Q-rich low complexity sequences.…. Would be more accurate phrasing of the proposed mechanistic model.

We made this change to improve accuracy.

Reviewer #3 (Significance (Required)):

The authors propose a novel mechanism of dbTF-coTF interaction modulation (via intracellular pH modulation of His proteonation) and provide evidence to support it.

The major shortcoming of this study concerns the reconstitution of pH-dependent TF-mediated recruitment of the 4HtoASnf5-SWI/SNF complex in vitro. In fact, reconstitution of SNF5-ADR1 interaction might be the best experiment, since Gal4-VP16 behaves opposite to the prediction of the model. Alternatively, Gal4-AH could be used?

We have added the 4HtoA allele to figure 5 (see response above).

Gal4-AH is predicted to behave just like VP16 – they are both considered to be classical acidic activators. We have thought about Adr1, but no one has defined the activation domain, thus making this experiment currently infeasible. We have added discussion of the model that pH allows SWI/SNF to reassort itself to other activators, like Adr1.

Referee Cross-commenting

The editor letter could include the suggestion by one of the reviewers to include rolox – vitamin E in their remodeling reaction. Journal of Fluorescence volume 17, pages785-795 (2007) Single-Pair FRET Microscopy Reveals Mononucleosome Dynamics.

This is an anti-oxidant used in single molecule FRET studies to eliminate photo-blinking events. This is not necessary for ensemble FRET, as the timescale is longer (blinking doesn’t matter) and we look at the entire population. This approach is not commonly used in ensemble FRET.

I would like to support Major Revision as a decision.

In my opinion, reviewer 3 very correctly cast the results in a physiological perspective when saying "Their model is.… After a transient acidification, the conformational expansion is reversed leading to interaction with different transcription factors and ultimately to an altered transcriptional response." The genetic involvement of the histidines in Snf5 appears to be solid even though more details can be provided by the authors. To my knowledge, the molecular mechanism has not yet been proposed, and therefore, has value as a hypothesis, even if it is restricted to fungi.

We all agree that the reconstitution experiments with the 4 histidine versus wt complex is missing and showing pH titration results would be great (reviewer 2).

Aren't those good reasons to ask whether the authors are in a position to provide more experimental data leading to a major revision?

We have added the pH titration curve (new Figure 2 supplement 1).

We have further explored the bioinformatics across the broader Eukaryotic tree.

This analysis suggests that the SNF5 QLC regulatory module was gained in Ascomycota.

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

Summary:

This paper addresses how yeast cells adapt to carbon starvation, a process that has long been known to depend on gene regulation by the SWI/SNF chromatin remodeling complex. The authors find that a Qrich low complexity sequence (QLC) in the SNF5 subunit of SWI/SNF, and His residues within this sequence, are required for the response to glucose starvation. Sequence analysis indicates that QLCs with an enrichment of His are common in many eukaryotes. Using a fluorescent ADH2 reporter to track the transcription response to glucose starvation, and a fluorescent pH reporter. Single cell measurements of ADH2 induction and intercellular pH during carbon starvation at different external pH conditions led the authors to conclude that the signal for induction of ADH2 is the transient intracellular acidification that occurs in response to glucose starvation. SNF5 QLC is required both for normal kinetics of transient acidification and for additional steps during ADH2 induction. RNA-seq of WT and SNF5 QLC mutants at different conditions identified 4 groups of genes with different regulatory patterns. Bioinformatic analyses showed that genes involved in acute glucose-starvation response are pH and SNF5 QLC-dependent. in vitro nucleosome remodeling assays were used to demonstrate that neither the SNF5 QLC nor pH (over the relevant range) regulate ATPase or nucleosome remodelling activity. However, using a remodeling assay that depends on transcription factor recruitment, wild type SWI/SNF, but not SWI/SNF with the SNF5 QLC deleted, shows decreased recruitment at lower pH. All atom Monte Carlo simulations predict that segments of the QLC undergo protonation dependent conformational expansion, which is suggested to form the mechanistic basis of pH-sensitive transcription factor interactions. Sequence analysis demonstrates that QLCs with His residues are conserved in SWI5 across many fungi, and that QLCs are and more particular histidines in these low complexity regions, were proposed to play an evolutionarily conserved role in many eukaryotes based on in silico sequence analyses.

The conclusions of the authors are mostly well supported by their experimental findings. Starting with the in silico sequence analyses that indicate a broad evolutionary relevance of QLCs and their histidines, the authors move on to dissect QLC function at multiple scales. Growth assays in combination with the pADH2 fluorescence reporter (verified by RT-qPCR) and with pHluorin examine cells at the population and single-cell level. The in vitro assays and Monte Carlo simulations aim at deciphering the underlying molecular mechanism of the observed QLC perturbations.. Both data and methods are carefully presented, with detailed descriptions of the methodological approach (e.g. snf5∆ slow growth phenotype loss and how this was solved). The authors further provide analysis code and resources online, which supports reproducibility. For most experiments, quantification is provided, and statistical tests used appropriately.

We thank the reviewer.

Major comments:

1) The one experiment that needs additional explanation/analysis is the in vitro nucleosome remodeling assay. First, do the graphs show single traces or are they averages?

These are averages of 2-4 independent kinetic traces. This is stated in the methods, but we will have also added this information to the legend. There is some day to day variability but the patterns that we highlight are consistent. We will have clarified the revised version.

Similarly, in the supplemental data, the ATPase data were fit, but the fits (and accompanying errors) are not provided-only a representative trace.

This is a representative example. No consistent differences were found. We will clarify in the legend.

Second, the analysis would be more complete if the His mutant of the QLC was included.

We have added these data to figure 5.

Third, he authors could describe their conclusions from these data more clearly. In Figure 5D, the decrease in remodeling with decreasing pH is clear for the wild type traces. It also seems clear that the δ QLC is not pH responsive. However, what is the interpretation of the higher baseline (no TF) signal, including the pattern at the end of the trace, and lower overall remodeling (at all pH). Is it the author’s model that both the QLC and another region in SWI/SNF are required for interaction with VP16, and that the QLC interaction is pH sensitive (thus without the QLC, there is still recruitment but it is not pH sensitive)?

We believe that the small differences in baselines of negative controls is just technical variation.

Recognizing that this assay is not the focus of the paper, and that the authors are careful not to over interpret it, it does provide an important level of mechanistic insight.

We thank the reviewer. Indeed, reconstitution is inherently limited due to its reductionist nature. However, we can distinguish between two key models: modulation of intrinsic chromatin remodeling activity of SWI/SNF versus modulation of recruitment to chromatin. This assay nicely supports the latter mechanism. Furthermore, our new data suggest that the presence of histidines in a low-protonation state is required for interaction with VP16.

2) Since calibration curves are quite central for the pH measurements, exemplary curves should be provided in the supplementary data.

We have added these curves (new Figure 2 supplement 1).

Minor comments:

1. While the pADH2 reporter assay is well-controlled with ADH2 mRNA RT-qPCR, it would be interesting to know if the induction effects are also observable on the protein level. Clearly, the authors show that translation of new transcripts is required for pH recovery using cyclohexamide. However, even if no additional experiments (like ADH2 Western blot) are performed, at least some information (possibly from previous publications) about how the findings on the transcriptional level/on the level of ADH2 mRNA induction correspond to the protein level.

We don’t have an antibody for ADH2p, however the main assay we use in the paper requires both transcription and translation of the reporter gene (pADH2-mCherry). The manuscript is very clearly focused on the role of SWI/SNF in transcriptional activation, therefore we won’t investigate possible effects of pH or SWI/SNF on protein stability at this time. This is an interesting topic for future studies though!

2. The statistical test used for RT-qPCR data in Figure 1D should be provided (for example in Figure 4C, the statistical test and the figure are nicely described).

These are Bonferroni-corrected t-tests. We have added this information to the legend.

3. In Figure 4B, it is difficult to see the underlying box plots.

We have made the data points a more transparent to make the box plots more apparent.

4. For the sake of completeness, the authors might want to add how the results differ if culture OD exceeds OD 0.3 before the assays (lines 541+542).

ADH2 induction happens much more quickly if the culture exceeds OD 0.3, and the effects of the SNF5 alleles are less apparent. We believe that this is because the regulatory mechanism under study is important for the very earliest stages of transcriptional remodeling. We have clarified these points in the manuscript.

5. Some typos in the methods section should be revised (e.g. line 710, 755, 758, 759, 766).

We have corrected these typos in the manuscript.

6. It is confusing to refer to a nucleosome as a "model promoter" (as done in the abstract).

We have changed this language to:

“Furthermore, the SNF5 QLC mediated pH-dependent recruitment of SWI/SNF to an acidic transcription factor in a reconstituted nucleosome remodeling assay.”

Reviewer #2 (Significance (Required)):

Significance

This study represents a conceptual advance. The authors present an exciting model of pH sensing, in which few histidines govern pH-sensitive transcription factor interactions upon carbon starvation in S. cerevisae. By dissecting how a QLC can act as a pH sensor to translate carbon starvation into transcription changes needed for adaptation, the authors provide new insight into disordered protein function (potentially identifying a class of pH sensitive LCDs), and a novel mechanism linking environment (i.e. pH) to transcription control. The authors provide interesting speculation as to how similar mechanisms could be broadly conserved, and highlight examples of pH changes that occur in cells (including human cells). This work is therefore of broad general interest to the field of disordered proteins, transcription, and (potentially) cellular metabolism and its links to disease.

We thank the reviewer.

Expertise lacking: molecular dynamics simulations

Referee Cross-commenting

I also agree that the authors should revise this interesting manuscript to address the points raised by the reviewers.

We thank the reviewer.

Reviewer #3 (Evidence, reproducibility and clarity (Required)):

The intracellular milieu, including pH, is important for proper functioning of cellular processes. To be able to react to changes, cells need sensing and subsequently correction mechanisms. In the manuscript of Ignacio Gutiérrez et al., the authors showed that in S. cerevisiae histidines in a glutamine-rich low complexity sequence (QLC) in the N-terminus of SNF5, a subunit of the SWI/SNF chromatin remodeling complexes, are important for transcription response during carbon starvation. Their model is that a transient acidification leads to protonation of the histidines in the QLC of SNF5 resulting in a

conformation change. After a transient acidification, the conformational expansion is reversed leading to interaction with different transcription factors and ultimately to an altered transcriptional response. It is an interesting idea that chromatin remodelers, which are highly important transcription regulators, serve as a pH sensor to adapt gene expression.

Thank you, this indeed is the intended scope of the study.

A bit of the enthusiasm is gone, as the QLC of SNF5 is not conserved in humans, where the authors speculate in the discussion that QLCs of other SWI/SNF subunits could be important here. However, human QLCs are not enriched for histidines according to their analyses.

We have gone back and re-evaluated our bioinformatic analysis with a revised definition of QLCs based on a more rational argument (new Figure 1—figure supplement 2). In this new analysis, QLCs are enriched for histidine in all organisms that we investigated, including humans. This enrichment is especially apparent when comparing QLCs to other low complexity sequences (Figure 1 —figure supplement 3).

As QLC including histidines are frequent across many proteins in several other species, the authors claim that these regions could commonly function as pH sensors, but evidence for this is also missing.

We started to investigate the hypothesis that SNF5 is a pH-sensor, “enrichment for glutamine residues interspersed with histidine residues appears to be conserved sequence feature, both in QLCs in general, and in the N-terminus of SNF5 in particular, implying a possible functional role (40).” It is beyond the scope of our current work to demonstrate pH sensing in multiple organisms, or in a large number of different protein complexes. However, we plan to follow up with these kinds of studies in the future.

While the authors address an important topic and show SNF5 QLC-dependent effects, their data also leave many open questions and the model needs further support of additional data.

Major comments:

The authors define QLCs as “stretches of low-complexity sequences containing at least 10 glutamines” (p.5). They show an enrichment of alanine, proline and histidine in QLCs in yeast, however, this is not very evident in the SNF5 case directly in the defined QLCs. The N-terminus contains 2 QLCs (26 aa, 64aa), but the relevant histidines are mainly adjacent to the QLCs.

This is a good point. We carefully reconsidered our definition of QLCs in the revised manuscript since our previous working definition was a good starting point, but somewhat arbitrary. Our new definition of a QLC is a sub-region of the proteome in with an average fraction of glutamine residues of 25% or higher (minimum fraction), with a maximum interruption between any two glutamine residues of less than 17 residues, and the where the whole QLC is at least 15 residues in length (minimum length) (new Figure 1—figure supplement 2A). Systematic variation of the maximum interruption size revealed that 17 residues was the value that maximized the number of QLCs and the number of residues found within QLCs, offering an optimally permissive value under the 0.25 or greater fraction of glutamine threshold. Using this new definition of QLCs, we found that the N-terminus of SNF5 is mostly QLC, including all of the relevant histidines in this study (new figure 1A).

Due to additional glutamines in the vicinity, the authors refer here as SNF5 QLC of a 282 amino acid region. This region is further poorly conserved, but a range of ascomycetes showed also high glutamine and several histidine residues in the N-terminus of SNF5. However, it is not clear how evolutionary conserved this feature is across other species.

We have zoomed out to encompass proteomes that more broadly cover the Eukaryotic tree of life. This allowed us to determine that this sequence feature was gained in basidiomycete and ascomycete fungi, indicating conservation for ~1 billion years, but was not present in. the ancestor of the metazoa (animals).

Moreover, the QLC is not conserved in humans and QLCs in humans are not enriched for histidines. It is unclear why the authors only focused on ascomycetes to look at evolutionary conservation. They should extend their studies to provide an idea in which species their proposed model might play a role and what the evolutionary advantage might be to lose the SNF5 QLC. As histidines are also not enriched in human QLCs, the authors should comment on the relevance of their proposed mechanisms for humans (e.g. number of QLCs, QLCs including histidines).

See above. Metazoa don’t appear to have lost the sequence, rather it is gained in the fungi. We also used recent structures to compare BAF and SWI/SNF. Both complexes contain large clusters of low complexity sequences poised near the DNA exiting the nucleosome (see new Figure 5—figure supplement 1). Interestingly, BAF does have one QLC in that position.

While the growth of S. cerevisiae under poor carbon sources seems not to depend on SNF5 QLC or the histidines in this region, the authors observe a strongly reduced growth rate if the switch to poor carbon sources is preceded by a 24h glucose starvation. How is the growth rate if they are provided with glucose again after the starvation or pH is kept low for longer? Glucose starvation will also lead to a drop of intracellular ATP concentration and consequently impair ATP-dependent chromatin remodeling. After starvation, the cells may depend more on remodelers to restore the chromatin. The authors should do additional experiments to be able to discriminate the source of reduced cell growth after starvation.

We have added the experiment switching from glucose to starvation and back to glucose Figure 1—figure supplement 6.

With respect to: "do additional experiments to be able to discriminate the source of reduced cell growth after starvation." We believe that it’s likely that failure to induce a key set of starvation response genes to metabolically adapt to carbon starvation is part of the reason for reduced growth (figure 4).

While the C-terminus of SNF5 is interacting with the acidic patch of the nucleosomes (e.g. Han et al., 2020), the N-terminus of SNF5 has been shown to be important for interaction with acidic transcription activators (Neely et al., 2002). Having the snf5Δ mutant, the authors should include also this clone as control in Figure 1—figure supplement 5 and quantify the results, in order to clearly show their statement that deletion of SNF5 QLC is distinct from total loss of the SNF5 gene.

It's very clear from growth phenotypes (e.g. Figure 1—figure supplement 6A) that deletion of SNF5 QLC is distinct from total loss of the SNF5 gene. It has been previously published that complete deletion of SNF5 leads to disruption of the SWI/SNF complex. We believe that repeating these published results in Figure 1—figure supplement 5 would not add much, and because the snf5∆ complex falls apart, the experiment would not be a valid comparison.

Moreover, snf5Δ mutant showed a different phenotypes compared to the N-terminal SNF5 mutants, but authors failed mostly to follow up this mutant in the experiments following figure 1. They should include as much data as possible also for the SNF5 knockout strain, in order to elucidate where the differences arise.

There is a major problem with the snf5Δ mutant strain. It is extremely sick and we believe that it rapidly acquires suppressor mutations (or perhaps undergoes poorly defined epigenetic conversions). For this reason, we have not included a great deal of snf5Δ strain data to prevent confusion. We have added a supplemental figure (Figure 2 —figure supplement 4) showing the snf5Δ results equivalent to figure 2.

The authors followed consequently to investigate gene expression changes genome-wide. This is important for their conclusions; however, the design of the experiment is not entirely clear. What was the rational for choosing these conditions?

These mutants and conditions were designed to test the hypothesis that (A) the N-terminus of SNF5 and Histidines within are important for transcriptional reprogramming and (B) a pH change is required.

Moreover, some transcriptional changes are likely hidden in total RNA-seq and would have benefitted from nascent RNA-seq.

Nascent RNA-seq might reveal some changes in stable mRNAs that we miss in our experiments, but capturing a few more genes doesn’t further our model and it is not our goal to discover every gene regulated by SWI/SNF. Therefore, given the high cost and effort required, we will not perform these experiments.

On top, they showed that ADH2 is only up-regulated in a subset of cells. In order to elucidate this bimodal response, single-cell gene expression analyses would have been more appropriate. The latter two points might have also contributed to the rather limited effects. The results further show that most of SWI/SNF target genes are not altered under these conditions and therefore "normal" SWI/SNF function is mostly maintained.

This is beyond the scope of our study. At the point of response to these reviews, scRNA-seq had only been successfully attempted twice in S. cerevisiae (PMID: 31985403, PMID: 32420869). Each of these experiments was an eLife paper. We don’t believe that these experiments would further our model. Therefore, given the high cost and effort required, we will not perform these experiments.

In order to investigate the mechanistic effects of the loss or mutation of the SNF5 QLC, the authors should extend their in vitro assays and investigate if it is leading to altered protein-protein interaction (as proposed), altered chromatin binding (e.g. by ChIP-seq) or affects the SWI/SNF remodeling activity/ chromatin accessibility genome-wide (e.g. by ATAC-seq or MNase-seq) in yeast.

Again, our manuscript is not about the genomics of SWI/SNF, and this topic has been extensively researched before, including by coauthors of this study. We already know that recruitment of SWI/SNF to chromatin is the fundamental molecular mechanism for transcriptional control. None of the above experiments would significantly further the model proposed in the title and abstract. Therefore, given the high cost and effort required, we will not perform these experiments. Nevertheless, we have undertaken more focused ChIP experiments to test the recruitment of SWI/SNF to the ADH2 promoter (Figure 1—figure supplement 8)

Also in their model, they assume that the histidines may partially be still protonated after pH recovery leading to altered interaction partners. However, experimental evidence is missing for that and what the impact on this low complexity region and the interactions with other proteins are.

It’s unclear what the experiment would be here. It's impossible to know the in vivo protonation state of SNF5, but we believe that it’s reasonable to propose that this protonation state changes over the intracellular pH ranges measured in our experiments, given our knowledge of chemistry and the pKa of the histidine side chain.

The authors need to add statistical information, e.g. number of replicates, type of replicates, error bar information, statistical tests used.

We have added this information.

Minor comments:

The authors should describe in the method section how they assessed growth rate.

We have added these methods.

The authors have several formatting or reference inconsistencies (e.g. Figure 1 sup 2C (p.6));

We have addressed this formatting.

Figure 2A, C instead of Figure 2A, B (p.9);

This is actually correct – 2B is a guide to help interpret the quantification in 2C.

Figure 3—figure supplement 3 and 4 (p.11);

This change is included in the current version of the manuscript.

Figure 3B, right (p.11);

This change is included in the current version of the manuscript.

Figure 5 (within figures Ph 7.6, in text Ph 7.5),

This change is included in the current version of the manuscript.

Yudkovsky et al. 1999 (p.17), Supplementary Tables) and should improve their labeling for certain graphs (e.g. Figure 1D, Figure 2A (addition of % values per quadrant), Figure 2C.

We have added these values to the revised manuscript.

Figure 2C, 3B: pHi – legend for black and grey lines are missing. Are these two replicates? If so, the authors should also show the third replicate mentioned in the legend (2C).

We have clarified the legend.

Data for the experiment with sorbic acid are not shown (p.11).

We are not sure which data the reviewer is referring to here.

Please indicate which clustering was used for Figure 4E and which “manual curation” was performed.

We used Euclidean clustering. Manual curation was when to removed small clusters that didn't have significant GO hits, and to consolidate clusters that had similar behavior. We have described these methods in more detail.

Figure 5B: There seem to be slight differences between WT and the mutant. An overlay of them would be nice to better compare the effects.

These slight differences are technical variation. Overlay of WT and dQ at pH 7.6 is shown in Author response image 1.

Author response image 1

Figure 5—figure supplement 1: A, How was significance determined?

In fact, it is difficult to speak to significance here. We will rephrase to say “only minor changes in ATPase activity were observed”. This is consistent with no impact of pH on nucleosome sliding activity in the absence of activator recruitment.

Reviewer #1 (Significance (Required)):

Cells need to sense and adjust changes in the intracellular environment. The authors propose that QLCs in important gene expression regulators – here a subunit of the SWI/SNF chromatin remodeler – can sense alterations in pH and consequently adjust gene expression. While the concept is interesting, further experiments are required to prove their model. Also due to a lack of conservation of this region, this particular QLC is not relevant for human. It would be great, if the authors could address the impact of QLCs in general for pH sensing and response and investigate, if similar mechanisms hold true for humans.

We agree that extension of the project would be very interesting. However, the title of the paper clearly indicates that the scope is for SWI/SNF. We are analyzing further pH sensing mechanisms in current/future work. We think that this demonstration that a crucial chromatin remodeling complex is a pH sensor is very interesting.

Referee Cross-commenting

I agree with reviewer 1 and 2 to provide the authors the opportunity to revise their manuscript and address the points raised.

https://doi.org/10.7554/eLife.70344.sa2

Article and author information

Author details

  1. J Ignacio Gutierrez

    Department of Molecular and Cell Biology, University of California, Berkeley, Berkeley, United States
    Present address
    Weill Cornell Medicine, New York, United States
    Contribution
    Conceptualization, Formal analysis, Investigation, Methodology, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-9017-8384
  2. Gregory P Brittingham

    Institute for Systems Genetics, New York University Grossman School of Medicine, New York, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Software, Validation, Visualization, Writing – review and editing
    Competing interests
    No competing interests declared
  3. Yonca Karadeniz

    Program in Molecular Medicine, University of Massachusetts Medical School, Worcester, United States
    Contribution
    Formal analysis, Investigation, Methodology
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-8299-551X
  4. Kathleen D Tran

    Department of Cell and Molecular Biology, University of Rhode Island, Kingston, United States
    Contribution
    Formal analysis, Investigation, Methodology
    Competing interests
    No competing interests declared
  5. Arnob Dutta

    Department of Cell and Molecular Biology, University of Rhode Island, Kingston, United States
    Contribution
    Formal analysis, Investigation, Methodology, Writing – review and editing
    Competing interests
    No competing interests declared
  6. Alex S Holehouse

    1. Center for Science and Engineering of Living Systems (CSELS), Washington University in St. Louis, St Louis, United States
    2. Department of Biochemistry and Molecular Biophysics, Washington University School of Medicine, St. Louis, United States
    Contribution
    Conceptualization, Formal analysis, Investigation, Methodology, Resources, Software, Visualization, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-4155-5729
  7. Craig L Peterson

    Program in Molecular Medicine, University of Massachusetts Medical School, Worcester, United States
    Contribution
    Conceptualization, Formal analysis, Funding acquisition, Investigation, Supervision, Writing – review and editing
    Competing interests
    No competing interests declared
  8. Liam J Holt

    Institute for Systems Genetics, New York University Grossman School of Medicine, New York, United States
    Contribution
    Conceptualization, Funding acquisition, Investigation, Methodology, Project administration, Visualization, Writing – original draft, Writing – review and editing
    For correspondence
    Liam.Holt@nyulangone.org
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-4002-0861

Funding

Becas Chile

  • J Ignacio Gutierrez

National Science Foundation (Graduate Research Fellows Program)

  • Gregory P Brittingham

Pershing Square Sohn Cancer Research Award

  • Liam J Holt

National Cancer Institute (R37 CA240765)

  • Liam J Holt

National Institute of General Medical Sciences (R01 GM132447)

  • Liam J Holt

American Cancer Society Cornelia T. Bailey Foundation Research Scholar Grant (RSG-19-073-01-TBE)

  • Liam J Holt

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

Acknowledgements

We thank Conor Howard for help with initial bioinformatics and conception of this project and Morgan Delarue for help with MATLAB analysis. We thank David Truong, Sudarshan Pinglay, and JoAnna Klein for help in writing the manuscript; Ivan Tarride for help with figure design; and Karsten Weis, Jeremy Thorner, and Douglas Koshland for advice, strains, plasmids, and reagents. We thank Cindy Hernandez for help with growth curves. We gratefully acknowledge funding from the William Bowes Fellows program, the Vilcek Foundation, the HHMI HCIA summer institute, NIH R01 GM132447 and R37 CA240765, the American Cancer Society Cornelia T Bailey Foundation Research Scholar Grant, RSG-19-073-01-TBE, and the Pershing Square Sohn Cancer Research Award (LJH); Becas Chile (JIG); and the National Science Foundation Graduate Research Fellows Program (GB).

Senior Editor

  1. Kevin Struhl, Harvard Medical School, United States

Reviewing Editor

  1. Alan G Hinnebusch, Eunice Kennedy Shriver National Institute of Child Health and Human Development, United States

Publication history

  1. Preprint posted: March 3, 2021 (view preprint)
  2. Received: May 19, 2021
  3. Accepted: February 6, 2022
  4. Accepted Manuscript published: February 7, 2022 (version 1)
  5. Version of Record published: March 2, 2022 (version 2)

Copyright

© 2022, Gutierrez 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

  • 1,519
    Page views
  • 267
    Downloads
  • 0
    Citations

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

Download links

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

Downloads (link to download the article as PDF)

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

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

  1. J Ignacio Gutierrez
  2. Gregory P Brittingham
  3. Yonca Karadeniz
  4. Kathleen D Tran
  5. Arnob Dutta
  6. Alex S Holehouse
  7. Craig L Peterson
  8. Liam J Holt
(2022)
SWI/SNF senses carbon starvation with a pH-sensitive low-complexity sequence
eLife 11:e70344.
https://doi.org/10.7554/eLife.70344

Further reading

    1. Biochemistry and Chemical Biology
    Sudipta Mondal et al.
    Research Article

    Chain-length specific subsets of diacylglycerol (DAG) lipids are proposed to regulate differential physiological responses ranging from signal transduction to modulation of the membrane properties. However, the mechanism or molecular players regulating the subsets of DAG species remains unknown. Here, we uncover the role of a conserved eukaryotic protein family, DISCO-interacting protein 2 (DIP2) as a homeostatic regulator of a chemically distinct subset of DAGs using yeast, fly and mouse models. Genetic and chemical screens along with lipidomics analysis in yeast reveal that DIP2 prevents the toxic accumulation of specific DAGs in the logarithmic growth phase, which otherwise leads to endoplasmic reticulum stress. We also show that the fatty acyl-AMP ligase-like domains of DIP2 are essential for the redirection of the flux of DAG subspecies to storage lipid, triacylglycerols. DIP2 is associated with vacuoles through mitochondria-vacuole contact sites and such modulation of selective DAG abundance by DIP2 is found to be crucial for optimal vacuole membrane fusion and consequently osmoadaptation in yeast. Thus, the study illuminates an unprecedented DAG metabolism route and provides new insights on how cell fine-tunes DAG subspecies for cellular homeostasis and environmental adaptation.

    1. Biochemistry and Chemical Biology
    2. Microbiology and Infectious Disease
    Atanas Radkov et al.
    Research Article

    Members of the bacterial T6SS amidase effector (Tae) superfamily of toxins are delivered between competing bacteria to degrade cell wall peptidoglycan. Although Taes share a common substrate, they exhibit distinct antimicrobial potency across different competitor species. To investigate the molecular basis governing these differences, we quantitatively defined the functional determinants of Tae1 from Pseudomonas aeruginosa PAO1 using a combination of nuclear magnetic resonance (NMR) and a high-throughput in vivo genetic approach called deep mutational scanning (DMS). As expected, combined analyses confirmed the role of critical residues near the Tae1 catalytic center. Unexpectedly, DMS revealed substantial contributions to enzymatic activity from a much larger, ring-like functional hot spot extending around the entire circumference of the enzyme. Comparative DMS across distinct growth conditions highlighted how functional contribution of different surfaces is highly context-dependent, varying alongside composition of targeted cell walls. These observations suggest that Tae1 engages with the intact cell wall network through a more distributed three-dimensional interaction interface than previously appreciated, providing an explanation for observed differences in antimicrobial potency across divergent Gram-negative competitors. Further binding studies of several Tae1 variants with their cognate immunity protein demonstrate that requirements to maintain protection from Tae activity may be a significant constraint on the mutational landscape of tae1 toxicity in the wild. In total, our work reveals that Tae diversification has likely been shaped by multiple independent pressures to maintain interactions with binding partners that vary across bacterial species and conditions.