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

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.


Introduction
7 starvation ( Figure 1D). These results suggest a dual-role for SNF5 in ADH2 regulation, both 176 contributing to strong repression in glucose, and robust induction upon carbon starvation. The 177 ΔQsnf5 and HtoAsnf5 alleles separate these functions, maintaining WT-like repression while 178 showing a strong defect in induction. 179 The RT-qPCR assay reports on the average behavior of a population. To enable single-cell 180 analysis, we engineered a reporter strain with the mCherry (43)   The SNF5 QLC is required for ADH2 expression and recovery of neutral pH 199 Multiple stresses, including glucose-starvation, have been shown to cause a decrease in the pH 200 of the cytoplasm and nucleus (nucleocytoplasm) (8,9,13,44). Herein, we refer to 201 nucleocytoplasmic pH as intracellular pH, or pHi. To investigate the relationship between ADH2 202 expression and pHi, and how these factors depend upon SNF5, we engineered strains bearing 203 both the ratiometric fluorescent pH-reporter, pHluorin (45), and the PADH2-mCherry reporter. These 204 cell lines allow us to simultaneously monitor pHi and expression of ADH2. 205 Wild-type cells growing exponentially in 2% glucose had a pHi of ~ 7.8. Upon acute carbon 206 starvation, cells rapidly acidified to pHi ~ 6.5. Then, during the first hour, two populations arose: 207 an acidic population (pHi ~ 5.5), and a second population that recovered to pHi ~ 7 (Figure 2A). 208 Cells at pHi 7 proceeded to strongly induce expression of the PADH2-mCherry reporter, while cells 209 at pHi 5.5 did not. After 8 h of glucose-starvation > 70% of wild-type cells had induced ADH2 210 (Figure 2A, C). 211 We next analyzed cells harboring mutant alleles of the QLC of SNF5. Similarly to WT, both 212 ΔQsnf5 and HtoAsnf5 strains rapidly acidified upon carbon starvation. However, these strains were 213 defective in subsequent neutralization of pHi and in the expression of PADH2-mCherry. At the 4 h 214 time point, > 95 % of both ΔQsnf5, and HtoAsnf5 cells remained acidic with no detectable 215 expression, while > 60% of wild-type cells had neutralized and expressed mCherry (Figure 2A, 216 C). These results demonstrate that the SNF5 QLC is necessary for efficient recovery from 217 transient acidification. Eventually, after 24 h, the majority of mutant cells neutralized to pHi ~ 7 218 and induced expression of PADH2-mCherry (Figure 2 -figure supplement 1). Thus, the SNF5 219 QLC and histidines within are required for the rapid dynamics of both transient acidification and 220 transcriptional induction of PADH2-mCherry upon acute carbon starvation. 221 We hypothesized that mutant cells might fail to recover from acidification because transcripts 222 controlled by SWI/SNF are responsible for pHi recovery. In this model, SWI/SNF drives 223 expression of a set of genes that must be both transcribed and translated. To test this idea we 224 measured pHi in WT cells during carbon starvation in the presence of the cyclohexamine to 225 prevent translation of new transcripts. In these conditions, we found that cells experienced a drop 226 in pHi but were unable to recover neutral pH (Figure 2 -figure supplement 2). Thus, new gene 227 expression is required for recovery of pHi.

Transient acidification is required for ADH2 induction upon carbon starvation 230
The acidification of the yeast nucleocytoplasm has been shown to depend upon an acidic 231 extracellular pH (pHe). We took advantage of this fact to manipulate the changes in pHi that occur 232 upon carbon starvation. Cell viability was strongly dependent on pHe, decreasing drastically when 233 cells were starved for glucose in media at pH ≥ 7.0 for 24 h (Figure 3 -figure supplement 1). 234 Expression of PADH2-mCherry expression was also highly dependent on pHe, especially in SNF5 235 QLC mutants ( Figure 3A, Figure 3 -figure supplement 2). WT cells failed to induce PADH2-236 mCherry at pHe ≥ 7, but induced strongly at pHe ≤ 6.5. RT-qPCR showed similar behavior for the 237 endogenous ADH2 transcript (Figure 3 -figure supplement 2). Furthermore, we found that the 238 nucleocytoplasm of all strains failed to acidify when the environment was held at pHe ≥ 7 ( Figure  239 3 -figure supplement 3). Therefore, we conclude that an acidic extracellular environment is 240 required for a drop in intracellular acidity upon carbon starvation, and that this intracellular 241 acidification is required for activation of ADH2 transcription. 242 Given that intracellular acidification is necessary for ADH2 promoter induction, we next 243 wondered if it was sufficient. First, we used the membrane permeable sorbic acid to allow 244 intracellular acidification but prevent pHi recovery. These cells failed to induce PADH2-mCherry, 245 indicating that nucleocytoplasmic acidification is not sufficient; subsequent neutralization is also 246 required. Carbon starvation at pHe 7.4 prevented transient acidification and likewise prevented 247 expression ( Figure 3B, Figure 3 -figure supplement 3). Cells that were first held at pHe 7.4, 248 preventing initial acidification, and then switched to pHe 5, thereby causing late acidification, failed 249 to express mCherry after 6 h. Finally, starvation at pHe 5 for 2 h followed by a switch to pHe 7.4, 250 with a corresponding increase in pHi led to robust PADH2-mCherry expression. Together, these 251 results suggest that transient acidification immediately upon switching to carbon starvation 252 followed by recovery to neutral pHi is the signal for the efficient induction of PADH2-mCherry. 253 Deletion of the SNF5 QLC leads to both failure to neutralize pHi and loss ADH2 254 expression. We therefore wondered if forcing cells to neutralize pHi would rescue ADH2 255 expression in a ΔQsnf5 strain. This was not the case: the ΔQsnf5 strain still fails to express PADH2-256 mCherry, even if we recapitulate normal intracellular transient acidification (Figure 3B, left). We next performed hierarchical clustering analysis (Euclidean distance) of the 149 genes that 280 are strongly differentially expressed between strains, or at suboptimal pHe 7 ( Figure 4E). Based 281 on this clustering and some manual curation, we assigned these genes to four groups. Group 1 282 genes (n = 42) were activated in starvation in a SNF5 QLC and pH-dependent manner. They are 283 strongly induced in WT but induction is attenuated both in mutants of the SNF5 QLC and when 284 the transient acidification of pHi was prevented by starving cells in media titrated to pHe 7. GO 285 analysis revealed that these genes are enriched for processes that are adaptive in carbon 286 starvation, for example fatty acid metabolism and the TCA cycle. Group 2 (n = 64) genes were 287 not strongly induced in WT, but were inappropriately induced during starvation in SNF5 QLC 288 mutants and during starvation at pHe 7. GO analysis revealed that these genes are enriched for 289 stress responses, perhaps because the failure to properly reprogram transcription leads to cellular 290 stress. Group 3 genes (n = 51) were repressed upon carbon-starvation in a pH-dependent but 291 SNF5 QLC-independent manner. They were repressed in all strains, but repression failed at pHe 292 7. Finally, Group 4 genes (n = 16) were repressed in WT cells in a pH-independent manner, but 293 failed to repress in SNF5 QLC mutants. 294

14
We performed an analysis for the enrichment of transcription factors within the promoters of 295 each of these gene sets using the YEASTRACT server (46). These enrichments are summarized 296 in Supplemental Table 2. Top hits for Group 1 included the CAT8 and ADR1 transcription factors, 297 which have previously been suggested to recruit the SWI/SNF complex to the ADH2 promoter 298 (47). 299 In conclusion, both pH changes and the SNF5 QLC are required for correct transcriptional 300 reprogramming upon carbon starvation, but the dependencies are nuanced. Mutation of the SNF5 301 QLC or prevention of nucleocytoplasmic acidification appears to trigger a stress response (Group 302 2 genes). Another set of genes requires pH change for their repression upon starvation, but this 303 pH sensing is independent of SNF5 (Group 3). A small set of genes requires the SNF5 QLC but 304 not pH change for repression upon starvation (Group 4). Finally, a set of genes, including many 305 of the traditionally defined "glucose-repressed genes", require both the SNF5 QLC and a pH 306 change for their induction upon carbon starvation (Group 1). For these genes, point mutation of 4 307 histidines in the QLC is almost as perturbative as complete deletion of the QLC. We propose that 308 the SNF5 QLC senses the transient acidification that occurs upon carbon starvation to elicit 309 transcriptional activation of this gene-set. It is striking that this set is enriched for genes involved 310 in catabolism, TCA cycle and metabolism, given that these processes are important for energetic 311 adaptation to acute glucose-starvation. 312 Gene ontology enrichment results for 9 clusters of genes are shown to the right.

313
The SNF5 QLC mediates a pH-sensitive transcription factor interaction in vitro 314 We reasoned that pHi changes could affect the intrinsic nucleosome remodeling activity of 315 SWI/SNF, or alternatively might impact the interactions of SWI/SNF with transcription factors. We 316 used a fluorescence-based strategy in vitro to investigate these potential pH-sensing 317 mechanisms. A center-positioned, recombinant mononucleosome was assembled on a 200 bp 318 DNA fragment containing a "601" nucleosome positioning sequence (48) (Figure 1A). The 319 nucleosomal substrate contained two binding sites for the Gal4 activator located upstream, and 320 68 base pairs of linker DNA downstream of the nucleosome. The mononucleosome contained a 321 Cy3 fluorophore covalently attached to the distal end of the template DNA, and Cy5 was attached 322 to the H2A C-terminal domain. The Cy3 and Cy5 fluorophores can function as a Förster 323 Resonance Energy Transfer (FRET) pair only when the Cy3 donor and Cy5 acceptor are within 324 an appropriate distance (see also Li and Widom, 2004). In the absence of SWI/SNF activity, the 325 center-positioned nucleosome has a low FRET signal, but ATP-dependent mobilization of the 326 nucleosome towards the distal DNA end leads to an increase in FRET (49-53) (Figure 5). In the 327 absence of competitor DNA, SWI/SNF does not require an interaction with a transcription factor 328 to be recruited to the mononucleosome and thus intrinsic nucleosome remodeling activity can be 329 assessed independently of recruitment. In this assay, SWI/SNF complex containing ΔQsnf5p 330 retained full nucleosome remodeling activity (Figure 5A), as well as full DNA-stimulated ATPase 331 activity ( Figure 5 -figure supplement 1). Furthermore, these activities were similar at pH 6.5, 332 7, or 7.5. Thus, we conclude that the SNF5 QLC does not sense pH by modifying its intrinsic 333 ATPase and nucleosome remodeling activity, at least in this in vitro context. 334 Next, we assessed if the SNF5 QLC and pH changes could affect SWI/SNF interactions 335 with transcription factors. SWI/SNF remodeling activity can be targeted to nucleosomes in vitro 336 by Gal4 derivatives that contain acidic activation domains, an archetypal example of which is 337 VP16 (Yudkovsky et al., 1999). Indeed, it was previously demonstrated that the QLC of Snf5p 338 mediates interaction with the Gal4-VP16 transcription factor (32). To assess recruitment of 339 SWI/SNF we set up reactions with an excess of nonspecific competitor DNA. In these conditions, 340 there is very little recruitment and remodeling without interaction with a transcription factor bound 341 to the mononucleosome DNA (Figure 5C, D). In this context, we found that the QLC of SNF5 was 342 required for rapid, efficient recruitment of SWI/SNF by the Gal4-VP16 activator, and that the pH 343 of the buffer affected this recruitment ( Figure 5D). Within the physiological pH-range (pH 6.5 to 344 7.5), recruitment and remodeling increased with pH. SWI/SNF complexes deleted for the SNF5 345 QLC (containing ΔQsnf5p) had constitutively lower recruitment and were completely insensitive 346 to pH changes over this same range (Figure 5D, right). Therefore, we conclude that the SNF5 347 DQsnf5p (right) SWI/SNF complexes at pH 7.6 (blue), 7.0 (green), or 6.5 (orange). Inset on the left panel shows the first 100 seconds of the assay after ATP addition. All traces represent FRET normalized to values prior to addition of ATP.

Protonation of histidines leads to conformational expansion of the SNF5 QLC 353
How might pH change be sensed by SNF5? As described above (Figure 1B), Q-rich low-354 complexity sequences (QLCs) are enriched for histidines, and they are also depleted for charged 355 amino acids (Figure 1B). Charged amino acids have repeatedly been shown to govern the 356 conformational behavior of disordered regions (54-56). Given that histidine protonation alters the 357 local charge density of a sequence, we hypothesized that the charge-depleted QLCs may be 358 poised to undergo protonation-dependent changes in conformational behavior. To test this idea, 359 we performed all-atom Monte-Carlo simulations to assess the conformational ensemble of a 50 360 amino acid region of the SNF5 QLC (residues 71-120) that contained 3 histidines, 2 of which we 361 had mutated to alanine in our experiments ( Figure 6A). We performed simulations with histidines 362 in both uncharged and protonated states to mimic possible charges of this polypeptide at the pH 363 found in the nucleocytoplasm in glucose and carbon starvation respectively. These simulations 364 generated ensembles of almost 50,000 distinct conformations (representative images shown in 365 Figure 6B). To quantify conformational changes, we examined the radius of gyration, a metric 366 that describes the global dimensions of a disordered region ( Figure 6C). Protonation of the 367 wildtype sequence led to a striking increase in the radius of gyration, driven by intramolecular 368 electrostatic repulsions (Figure 6D, left). In contrast, when 2/3 histidines were replaced with 369 alanines, no such change was observed (Figure 6D, right). For context, we also calculated an 370 apparent scaling exponent (ν app ), a dimensionless parameter that can also be used to quantify 371 chain dimensions. This analysis showed that protonation of the wildtype sequence led to a change 372 in ν app from 0.48 to 0.55, comparable to the magnitude of changes observed in previous studies 373 of mutations that fundamentally altered intermolecular interactions in other low-complexity 374 disordered regions (56, 57). These results suggest that small changes in sequence charge density 375 can elicit a relatively large change in conformational behavior. An analogous (albeit less 376 pronounced) effect was observed for the second QLC subregion that we mutated (residues 195-377 233) (Figure 6 -figure supplement 1). Taken together, our results suggest that charge-depleted 378 disordered regions (such as QLCs) are poised to undergo pH-dependent conformational re-379 arrangement. This inference offers the beginnings of a mechanism for pH-sensing by SWI/SNF: 380 the conformational expansion of the QLC sequence upon nucleocytoplasmic acidification may 381 tune the propensity for SWI/SNF to interact with transcription factors (Figure 6E). Previous studies of intracellular state during glucose starvation based on population 396 averages reported a simple decrease in pHi (15). In this work, we used single-cell measurements 397 of both pHi and gene expression, and found that two co-existing subpopulations arose upon acute 398 glucose-starvation, one with pHi ~ 5.5 and a second at ~ 6.5. The latter population recovered to 399 neutral pHi and then induced glucose-repressed genes, while the former remained dormant in an 400 acidified state. We have not yet determined the mechanism that drives the bifurcation in pH 401 response. It is possible that this bistability provides a form of bet-hedging (63) where some cells 402 attempt to respond to carbon starvation, while others enter a dormant state (19). However, we 403 have yet to discover any condition where the population with lower pHi and delayed transcriptional 404 activation has an advantage. An alternative explanation is that these cells are failing to correctly 405 adapt to starvation, perhaps undergoing a metabolic crisis, as suggested in a recent study (62). 406 It is becoming clear that intracellular pH is an important mechanism of biological control. 407 It was previously shown that the protonation state of phosphatidic acid (PA) determines binding 408 to the transcription factor Opi1, coupling membrane biogenesis and intracellular pH (4). We 409 focused our studies on the N-terminal region of SNF5 because it is known to be important for the 410 response to carbon starvation and contains a large low-complexity region enriched in both 411 glutamine and histidine residues. Histidines are good candidates for pH sensors as they can 412 change protonation state over the recorded range of physiological pH fluctuations, and their pKa 413 can be tuned substantially depending on local sequence context. Consistent with this hypothesis, 414 we found that the SNF5 QLC and the histidines embedded within were required for transcriptional 415 reprogramming. 416 23 Global analysis revealed that genes that require pHi oscillation and the SNF5 QLC for their 417 induction during carbon starvation are involved in metabolic processes including the TCA cycle, 418 fatty acid metabolism and the glyoxylate cycle. The upregulation of these metabolic pathways 419 may provide alternative energy sources. It will be interesting to see if human SWI/SNF undergoes 420 similar pH-dependent regulation. Cancer biology hints that this may be the case. It has been 421 observed that about 20% of human cancers have mutations in the SWI/SNF complex (64). Human 422 SNF5 (SMARCB1) was the first subunit of the SWI/SNF to be linked to cancer, where it is mutated 423 in most cases of pediatric malignant rhabdoid tumor (65, 66). It is known that mutations of the 424 SWI/SNF that lead to cancer generally result in misregulation of fatty acid synthesis, which is 425 required for cancer proliferation (67, 68). The pH-sensing QLC found in yeast SNF5 is absent in 426 the human orthologue, SMARCB1, but QLCs and regions of extreme histidine enrichment are 427 present in the Arid1a, Arid1b and Arid2 subunits of human SWI/SNF, and loss of Arid1a is a 428 leading cause of ovarian and uterine cancers (69). An acidic pH is a prominent feature of the 429 tumor microenvironment (70, 71) and intracellular pH tends to be elevated in tumor cells. These 430 observations motivate the future study of pH-sensing by SWI/SNF in humans. 431 Our in vitro assays showed that the intrinsic ATPase and nucleosome remodeling activities 432 of SWI/SNF are robust to pH changes from 6.5 to 7.6. However, recruitment of remodeling activity 433 by a model transcription factor (GAL4-VP16) was pH-sensitive, and this pH dependence was 434 dependent on the SNF5 QLC. In this case, the recruitment by GAL4-VP16 was inhibited at pH 435 6.5. We speculate that low pHi favors release of SWI/SNF from activators that it is bound to in 436 glucose conditions, and then the subsequent partial recovery in pHi could allow it to bind to a 437 different set of activators, thus recruiting it to genes that are expressed during starvation. This 438 model is consistent with the requirement for both acidification and subsequent neutralization for 439 expression of ADH2 (Figure 3). In principle, the conformational dynamics of the SNF5 QLC could 440 be distinct at all three stages ( Figure 6E). There are almost certainly additional pH-sensing 441 elements of the transcriptional machinery that also take part in this reprogramming. 442 Low complexity sequences, including QLCs, tend to be intrinsically disordered and 443 therefore highly solvent exposed. A recent large-scale study of intrinsically disordered sequences 444 showed that their conformational behavior is inherently sensitive to changes in their solution 445 environment (36,37). Similarly, our simulations revealed that histidine protonation may lead the 446 SNF5 QLC to expand dramatically. This provides a potential mechanism for pH-sensing: upon 447 acidification, histidines become positively charged leading QLCs to adopt a more expanded state, 448 perhaps revealing short linear interaction motifs (SLIMs), reducing the entropic cost of binding to 449 interaction partners, preventing polar-mediated protein-protein interactions, or facilitating 450 24 electrostatic mediated contacts. The enrichment of histidines in QLCs hints that this could be a 451 general, widespread mechanism to regulate cell biology in response to pH changes. 452 Glutamine-rich low-complexity sequences have been predominantly studied in the context 453 of disease. Nine neurodegenerative illnesses, including Huntington's disease, are thought to be 454 caused by neurotoxic aggregation seeded by proteins that contain polyglutamines created by 455 expansion of CAG trinucleotide repeats (72). However, polyglutamines and glutamine-rich 456 sequences are relatively abundant in Eukaryotic cells: More than 100 human proteins contain 457 QLCs, and the Dictyostelium and Drosophilid phyla have QLCs in ~ 10% and ~ 5% of their 458 proteins respectively (73). Furthermore, there is clear evidence of purifying selection to maintain 459 polyQs in the Drosophilids (74). This prevalence and conservation suggest an important biological 460 function for these sequences. Recent work in Ashbya gosypii has revealed a role for QLC-461 containing proteins in the organization of the cytoplasm through phase separation into liquid 462 droplets to enable subcellular localization of signaling molecules (75). More generally, 463 polyglutamine has been shown to drive self-association into a variety of higher-order assemblies, 464 from fibrils to nanoscopic spheres to liquid droplets (76-78). Taken together, these results imply 465 that QLCs may offer a general mechanism to drive protein-protein interactions. In this study, we 466 have identified a role for QLCs in the SWI/SNF complex as pH-sensors. Our current model 467 ( Figure 6E) is that the SNF5 QLC partakes in heterotypic protein interactions that are modulated 468 by protonation of histidines when the cell interior acidifies. However, we don't rule out the 469 possibility for homotypic interactions and higher-order assembly of multiple SWI/SNF complexes. 470 All cells must modify gene expression to respond to environmental changes. This 471 phenotypic plasticity is essential to all life, from single celled organisms fighting to thrive in an 472 ever-changing environment, to the complex genomic reprogramming that must occur during 473 development and tissue homeostasis in plants and metazoa. Despite the differences between 474 these organisms, the mechanisms that regulate gene expression are highly conserved. Changes 475 in intracellular pH are increasingly emerging as a signal through which life perceives and reacts 476 to its environment. This work provides a new role for glutamine-rich low-complexity sequences as 477 molecular sensors for these pH changes. 478 479 25

Cloning and yeast transformations 482
Yeast strains used in this study were all in the S288c strain-background (derived from BY4743). 483 The sequences of all genes in this study were obtained from the Saccharomyces cerevisiae 484 genome database (http://www.yeastgenome.org/). 485 We cloned the various SNF5 alleles into plasmids from the Longtine/Pringle collection 486 (79). We assembled plasmids by PCR or gene synthesis (IDT gene-blocks) followed by Gibson 487 cloning (80). Then, plasmids were linearized and used to overwrite the endogenous locus by 488 sigma homologous recombination using homology to both ends of the target gene. 489 The ΔQsnf5 gene lacks the N-terminal 282 amino acids that comprise a glutamine rich 490 low complexity domain. Methionine 283 serves as the ATG for the ΔQ-SNF5 gene. In the HtoAsnf5 491 allele, histidines 106, 109, 213 and 214 were replaced by alanine using mutagenic primers to 492 amplify three fragments of the QLC region which were combined by Gibson assembly into a SNF5 493 parent plasmid linearized with BamH1 and Sac1. 494 We noticed that the slow growth null strain phenotype of the snf5Δ was partially lost over 495 time, presumably due to suppressor mutations. Therefore, to avoid these spontaneous 496 suppressors, we first introduced a CEN/ARS plasmid carrying the SNF5 gene under its own 497 promoter and the URA3 auxotrophic selection marker. Then a kanMX6 resistance cassette, 498 amplified with primers with homology at the 5' and 3' of the SNF5 gene was used to delete the 499 entire chromosomal SNF5 ORF by homologous recombination. We subsequently cured strains 500 of the CEN/ARS plasmid carrying WT SNF5 by negative selection against its URA3 locus by 501 streaking for single colonies on 5-FOA plates immediately before each experiment to analyze the 502 snf5Δ phenotype. 503 The PADH2-mCherry reporter was cloned into integrating pRS collection plasmids (81). 504 URA3 (pRS306) or LEU2 (pRS305) were used as auxotrophic selection markers. The 835 base 505 pairs upstream of the ADH2 gene was used as the promoter (PADH2). PADH2, and the mCherry ORF 506 were amplified by PCR and assembled into linearized pRS plasmids (Sac1/Asc1) by Gibson 507 assembly. These plasmids were cut in the middle of the ADH2 promoter using the Sph1 restriction 508 endonuclease and integrated into the endogenous ADH2 locus by homologous recombination. 509 The pHluorin gene was also cloned into integrating pRS collection plasmids. URA3 510 (pRS306) and LEU2 (pRS305) were used for selection. The plasmid with the pHluorin gene was 511 obtained described in (15). We amplified the pHluorin gene and the strong TDH3 promoter and 512 used Gibson assembly to clone these fragments into pRS plasmids linearized with Sac1 and 513 26 Asc1. Another strategy was to clone the pHluorin gene and a natMX6 cassette into the integrating 514 pRS304 plasmid (that contains TRP1), which was then linearized within the TRP1 cassette using 515 HindIII and integrated into the TRP1 locus. 516 A C-terminal TAP tag was used to visualize Snf5 and Snf2 proteins in Western blots. pRS 517 plasmids were used but the cloning strategy was slightly different. A 3' fragment of the SNF5 and 518 SNF2 genes were PCR amplified without the Stop codon. This segment does not contain a 519 promoter or an ATG codon for translation initiation. The TAP tag was then amplified by PCR and 520 cloned together with the 3' of SNF5 and SNF2 ORFs by Gibson assembly into pRS plasmids with 521 linearized Sac1 and Asc1. Plasmids were linearized in the 3' of the SNF5 or SNF2 ORFs with 522 StuI and XbaI respectively to linearize the plasmid allowing integration it into the 3' of each gene 523 locus by homologous recombination. Therefore, transformation results in a functional promoter at 524 the endogenous locus fused to the TAP tag. 525 The SNF5-GFP strain was obtained from the yeast GFP collection (82), a gift of the 526 Drubin/Barnes laboratory at UC Berkeley. The SNF2-GFP fused strain was made by the same 527 strategy used for the TAP tagged strain above. 528 Supplemental Tables 6 and 7 list strains and plasmids generated in this study. 529 530

Culture media 531
Most experiments, unless indicated, were performed in synthetic complete (SC) media (13.4 g/L 532 yeast nitrogen base and ammonium sulfate; 2 g/L amino acid mix and 2% glucose). Carbon 533 starvation media was SC media without glucose, supplemented with sorbitol, a non-fermentable 534 carbon source to avoid osmotic shock during glucose-starvation (6.7 g/L YNB + ammonium 535 sulfate; 2g/L Amino acid mix and 100 mM sorbitol). The pH of starvation media (pHe) was adjusted 536 using NaOH. 537 538

Glucose-starvation 539
Cultures were incubated in a rotating incubator at 30 Ο C and grown overnight (14 -16 h) to an OD 540 between 0.2 and 0.3. Note: it is extremely important to prevent culture OD from exceeding 0.3, 541 and results are different if cells are allowed to saturate and then diluted back. Thus, it is imperative 542 to grow cultures from colonies on plates for > 16 h without ever exceeding OD 0.3 to obtain 543 reproducible results. Typically, we would inoculate 3 ml cultures and make a series of 4 -5 1/5 544 dilutions of this starting culture to be sure to catch an appropriate culture the following day.

RT-qPCR 568
For qPCR and RNA seq, RNA was extracted with the "High pure RNA isolation kit" (Roche) 569 following the manufacturer's instructions. Three biological replicates were performed. cDNAs and 570 qPCR were made with iSCRIPT and iTAQ universal SYBR green supermix by Bio-Rad, following 571 the manufacturer's instructions. Samples processed were: exponentially growing culture (+Glu), 572 or acute glucose-starvation for 4 h in media titrated to pH 5.5 or 7.5. Primers for qPCR were taken We performed RNA sequencing analysis to determine the extent of the requirement for the 581 SNF5 QLC in the activation of glucose-repressed genes. Three biological replicates were 582 performed. Total RNA was extracted from WT, ΔQ-snf5 and HtoAsnf5 strains during exponential 583 growth (+Glu) and after 4 hours of acute glucose starvation. In addition, WT strains were acutely 584 starved in media titrated to pH 7. Next, poly-A selection was performed using Dynabeads and 585 libraries were performed following manufactures indications. Sequencing of the 32 samples was 586 performed on an Illumina Hi-seq on two lanes. RNA-seq data were aligned to the University of 587

Western blots 595
Strains containing SNF5 and SNF2 fused to the TAP tag were used. Given the low concentration 596 of these proteins, they were extracted with Trichloroacetic acid (TCA): 3 mL culture was pelleted 597 by centrifugation for 2 min at 6000 RPM and then frozen in liquid nitrogen. Pellets were thawed 598 on ice and re-suspended in 200 uL of 20% TCA, ~ 0.4 g of glass beads were added to each tube. 599 Samples were lysed by bead beating 4 times for 2 min with 2 min of resting in ice in each cycle. 600 Supernatants were extracted using a total of 1 mL of 5% TCA and precipitated for 20 min at 14000 601 RPM at 4 C. Finally, pellets were re-suspended in 212 uL of Laemmli sample buffer and pH 602 adjusted with ~26 uL of Tris buffer pH 8. Samples were run on 7 -12% gradient polyacrylamide 603 gels with Thermo-Fisher PageRuler prestained protein ladder 10 to 18 KDa. Proteins were 604 transferred to a nitrocellulose membrane, which was then blocked with 5% nonfat milk and 605 incubated with a rabbit IgG primary antibody (which binds to the protein A moiety of the TAP tag) 606 for 1 hour and then with fluorescently labelled goat anti-rabbit secondary antibody IRdye 680RD 607 goat-anti-rabbit (LI-COR Biosciences Cat# 926-68071, 1:15,000 dilution). Anti-glucokinase was 608 used as a loading control (rabbit-anti-Hxk1, US Biological Cat# H2035-01, RRID:AB_2629457, 609 Salem, MA, 1:3,000 dilution) followed by IRDye 800CW goat-anti-rabbit (LI-COR Biosciences 610 Cat# 926-32211, 1:15,000 dliution). Membranes were visualized using a LI-COR Odyssey CLx 611 scanner with Image Studio 3.1 software. Fluorescence emission was quantified at 700 and 800 612 nM. 613 614 29

Data fitting 639
Fluorescence intensity from the PADH2-mCherry reporter and ratiometric fluorescence 640 measurements from pHluorin were fit with a single or double Gaussian curve for statistical 641 analysis using MATLAB (MathWorks). The choice of a single or double Gaussian fit was 642 determined by assessing which fit gave the least residuals. For simplicity, the height (mode) of 643 each Gaussian peak was used to determine the fraction of cells in each population rather than 644 the area, because peaks overlapped in many conditions. 645 646 Sequence analysis of QLCs 647