Multiple roles for the ESCRT machinery in maintaining plasma membrane homeostasis

The endosomal sorting complexes required for transport (ESCRT) execute evolutionary conserved membrane remodeling processes. Here we used budding yeast to explore how the ESCRT machinery contributes to plasma membrane (PM) homeostasis. In response to reduced membrane tension and inhibition of the target of rapamycin complex 2 (TORC2), ESCRT-III/Vps4 assemblies form at the PM and help to maintain membrane integrity. Conversely, the growth of ESCRT mutants strongly depends on TORC2-mediated homeostatic regulation of sphingolipid (SL) metabolism. This is caused by calcineurin phosphatase activity which causes Orm2 to accumulate at the endoplasmic reticulum (ER) in ESCRT mutants. Orm2 is a repressor of SL biosynthesis and its accumulation provokes increased membrane stress. This necessitates TORC2 signaling through its downstream kinase Ypk1 to control Orm2 protein levels and prevent a detrimental imbalance of SL metabolism. Our findings reveal new aspects of antagonistic calcineurin/TORC2 signaling for the regulation of SL biosynthesis and the maintenance of PM homeostasis, and suggest that the ESCRT machinery contributes directly and indirectly to these processes.


7
Results 162 163 ESCRT mutants depend on TORC2-Ypk1 signaling for cell growth and survival 164 The evolutionary conserved ESCRT complexes assemble into membrane remodeling 165 machineries that execute fundamental biological processes, including the biogenesis of 166 multivesicular bodies for lysosomal membrane protein degradation and the repair of diverse 167 cellular membranes (41). Given these key roles of the ESCRT machinery, it is surprising that 168 ESCRT mutant cells are viable. This implies that cells can at least in part compensate for the 169 loss of ESCRT function. Understanding how cells achieve this compensation would therefore 170 also provide a better understanding for the role of the ESCRT machinery in cellular homeostatic 171 processes. 172

173
To identify the processes that enable the survival of ESCRT-deficient cells, we used the 174 budding yeast, S. cerevisiae, as a model system and conducted a genome-wide synthetic genetic 175 interaction screen (9). Gene ontology (GO) term enrichment analysis revealed 'lipid 176 metabolism' among the top ranked hits that are required by ESCRT mutants (vps4∆) (Fig. 1A, 177 Table S1). This GO term included genes involved in the metabolism of various lipid classes, 178 but predominantly enzymes mediating non-essential steps of sterol biosynthesis, and genes 179 regulating SL homeostasis. We confirmed the synthetic growth defect of vps4∆ cells with four 180 of these genes (ypk1∆, sac1∆, csg2∆, erg2∆) in a different genetic background (the SEY6210 181 strain) (Fig. S1A). These results suggested that the loss of ESCRT-dependent membrane 182 remodeling rendered cells particularly sensitive to perturbations of membrane homeostasis. 183 184 We focused on the role of the protein kinase Ypk1 in ESCRT-deficient cells, because it 185 functions together with TORC2 to maintain membrane homeostasis (22,24,33,36,59,60). YPK1 186 deletion (but not deletion of YPK2) also caused synthetic growth defects with ESCRT-II 187 (vps25∆) and ESCRT-III mutants (vps2∆), but not with mutants of Chm7 (chm7∆) (Fig. S1B), 188 which recruits the ESCRT machinery to the ER and the inner nuclear membrane (51,52,61). 189 This result suggested that the function of Ypk1 was required to compensate for the loss of 190 ESCRT functions that are independent of Chm7. 191 192 The growth of vps4∆ ypk1∆ double mutants was restored by re-expression of VPS4 and converge to mitigate membrane stress. Therefore, we first defined the molecular nature of 276 membrane stress in ESCRT mutants that is controlled by TORC2-Ypk1 signaling. Ypk1 is 277 activated by TORC2 through phosphorylation primarily of Ser 644 and Thr 662 (Fig. 1C) 278 (23,69). Importantly, ESCRT mutants not only had slightly higher steady state Ypk1 protein 279 levels, but also higher levels of active Ypk1 phosphorylated on Thr 662 (pT662) (Fig. 3A, Fig.  280 S3A, B). We found no evidence for ESCRT-dependent lysosomal turnover of Ypk1 (70) in 281 growing cells (Fig. S3C, D). The Ypk1 activation level in ESCRT mutants was substantial, but 282 still lower compared to the hyperactivation of TORC2 signaling upon treatment with the SPT 283 inhibitor myriocin (Fig. 3A, lane 1). Of note myriocin treatment for 150 min also reproducibly 284 resulted in the up-regulation of Ypk1 protein levels (Fig. 3A, lane 1). The elevated Ypk1 protein 285 levels that were detected in ESCRT mutants and in response to myriocin treatment might thus 286 reflect Ypk1 stabilization by TORC2 phosphorylation as described earlier (69). We also 287 observed a slight increase of YPK1 mRNA in vps4∆ cells (Fig. S3E). 288 289 Ypk1 has a range of different cellular targets (31,32,59). Among those, the paralogous ER 290 proteins Orm1 and Orm2 are the essential targets of 31). In ESCRT mutant 291 cells Orm2 protein levels and its Ypk1-dependent phosphorylation were increased (Fig. 3B). 292 Deletion of YPK1 reduced but did not completely abrogate steady state phosphorylation of 293 Orm2 (presumably due to the presence of Ypk2), and caused upregulation of Orm2 protein 294 levels (Fig. 3B). In vps4∆ ypk1∆ double mutants Orm2 protein levels were higher than in either 295 single mutant. 296 11 297 TORC2-Ypk1 dependent phosphorylation of Orm2 and its paralogue Orm1 is required to de-298 repress the synthesis of LCBs by the SPT complex (15,22) (Fig. 3C). TORC2 and Ypk1 also 299 stimulate a subsequent step in the SL biosynthetic pathway by phosphorylating Lag1 and Lac1, 300 two components of the ER-resident ceramide synthase (CS) (71,72) (Fig. 3C). This concerted 301 regulation avoids the toxic accumulation of biosynthetic intermediates and helps to maintain 302 SL homeostasis (32). 303 304 We assessed the steady state levels of LCBs and ceramides using liquid chromatography -mass 305 spectrometry (LC-MS) analysis. Despite the accumulation of Orm2 in vps4∆ cells, total LCBs 306 (C18-PHS and C18-DHS) and ceramides were not lower than in WT cells (Fig. 3D). Also in 307 vivo labeling of the SL pool with [ 3 H]-L-serine in WT cells and vps4∆ cells showed only subtle 308 differences in the steady state levels of mannosylated SL species, but the overall formation of 309 [ 3 H]-labelled complex SL was comparable (Fig. 3E). It seemed that steady state levels of SL 310 synthesis in ESCRT mutants were maintained by Ypk1-dependent phosphorylation of Orm2. 311 Consistently, loss of Ypk1 in vps4∆ mutants skewed SL biosynthesis. In vps4∆ ypk1∆ double 312 mutants, the steady state LCB levels were increased (> 3-fold), while ceramide levels remained 313 similar to vps4∆ single mutants (Fig. 3D). The loss of Ypk1 signaling in ESCRT mutants 314 apparently caused an imbalance of SPT and CS activity, implying that the activity of CS might 315 have become a bottleneck. Consistently, we observed that vps4∆ ypk1∆ double mutants 316 produced less complex SL (predominantly inositolphosphoryl-ceramide (IPC)) than both WT 317 and vps4∆ cells (Fig. 3E). Expression of a phospho-mimetic variant of CS can promote the 318 formation of ceramides and complex SL independently of Ypk1 activity (32). Indeed, ectopic 319 expression of the LAG1-S23,24E allele, which mimics constitutive Ypk1 phosphorylation (32), 320 partially restored the growth of vps4∆ ypk1∆ cells (Fig. 3F) To understand which of these steps are affected in ESCRT mutants, we compared them to 344 EGAD mutants. Orm2 protein levels and its phosphorylation were similar in ESCRT (vps4∆) 345 and EGAD (tul1∆) mutants (Fig. 4A). However, Orm2 accumulated in different subcellular 346 compartments. In tul1∆ mutants GFP-Orm2 accumulated at the ER and on post-ER 347 compartments, including Golgi, endosomes and the vacuolar limiting membrane, as reported 348 earlier (9). In contrast, in the majority of vps4∆ cells, GFP-Orm2 was detected almost 349 exclusively at the ER (Fig. 4B). Only in some cells, we observed a partial colocalization of 350 GFP-Orm2 with the ESCRT substrate mCherry-CPS in class E compartments (Fig. S4C). Thus,351 it seemed that in ESCRT mutants Orm2 accumulated mainly at the ER. 352 353 Consistent with earlier reports (73), we detected Tul1 and Ubx3, two essential components of 354 the Dsc complex, at the class E compartment in ESCRT mutants (Fig. S4D). Hence, the Dsc 355 complex could degrade Orm2 once it reached the class E compartment. Indeed, GFP-Orm2 356 strongly accumulated in class E compartments when the EGAD pathway was additionally 357 compromised in vps4∆ tul1∆ cells (Fig. 4B), in agreement with our earlier findings (9). The 358 mutant Orm2-K25,33R, which is no longer ubiquitinated by the Dsc complex and hence no 359 longer degraded, was also found predominantly in class E compartments in ESCRT mutants 360 ( Fig. S4E). Native immunoprecipitation (IP) of Flag-Orm2 demonstrated that a fraction of 361 Orm2 still physically interacted with the Dsc complex in vps4∆ mutants (Fig. 4C), and 362 denaturing IP of Flag-Orm2 showed that it was ubiquitinated (Fig. 4D), although to a lower 363 extent compared to WT cells. 364 13 365 Consistently, Orm2 was still degraded in vps4∆ and vps25∆ mutants in cycloheximide chases, 366 although the half-life was substantially longer (Fig. 4E). In WT cells Orm2 was degraded with 367 a half-life of approximately 90 minutes, and in vps4∆ mutants the half-life of Orm2 was 368 approximately 180 minutes (Fig. 4E, F). In contrast, Orm2 degradation was fully blocked in 369 mutants of the Dsc complex (tul1∆) (9) (Fig. 4E, F) Our results so far present a conundrum: In ESCRT mutants Orm2 is phosphorylated by TORC2-378 Ypk1 (Fig. 3B, Fig. 4A), but still accumulates at the ER. Mutation of Ser 46,47,48 to alanine 379 in Orm2 (Orm2-3A) prevented phosphorylation by Ypk1 (22), reduced Orm2 ER export and 380 subsequent EGAD (9), and therefore Orm2-3A accumulated in WT cells and in vps4∆ mutants 381 ( Fig. 5A). We also reported earlier that Ser 46,47,48 to aspartic acid mutations (Orm2-3D) 382 mimic constitutive phosphorylation and trigger constitutive ER export, leading to degradation 383 via EGAD. Therefore, Orm2-3D protein levels are low in WT cells. Importantly also in vps4∆ 384 mutants, Orm2-3D protein levels were strongly reduced (Fig. 5A). This result implied that in 385 vps4∆ mutants the phospho-mimetic Orm2-3D mutant could be exported more efficiently from 386 the ER, and then was degraded by EGAD. Remarkably, expression of the Orm2-3D mutant, 387 but not Orm2-3A, improved also the growth of vps4∆ ypk1∆ cells (Fig. 5B Calcineurin activity also rendered ESCRT mutants dependent on TORC2-Ypk1 signaling, since 415 deletion of CNB1 partially restored the growth of vps4∆ ypk1∆ mutants (Fig. 5E). Likewise, 416 pharmacological inhibition of calcineurin activity (2µg/ml FK-506) or addition of the calcium-417 chelator EGTA to the growth medium improved growth of vps4∆ ypk1∆ cells (Fig. 5F). 418 Calcineurin inhibition with FK-506 markedly decreased Orm2 protein levels in the vps4∆ 419 ypk1∆ double mutants (Fig. 5G). The inhibitory function of calcineurin was likely independent 420 of the transcription factor Crz1 which is a major downstream effector of calcineurin (78,79). 421 Unlike the deletion of CNB1, deletion of CRZ1 did not rescue growth of vps4∆ ypk1∆ cells 422 (Fig. S5A). This is consistent with a role of calcineurin in controlling either directly or indirectly 423 Orm2 phosphorylation. Thus, calcineurin activity in ESCRT mutants reduced the ER export 424 and degradation of Orm2, which contributed to accumulation of Orm2 in the ER and rendered 425 the growth of ESCRT mutants dependent on TORC2-Ypk1 signaling. 426 427 Clearly, the accumulation of Orm2 constituted a major membrane stress factor in ESCRT 428 mutants: In vps4∆ orm2∆ double mutants Ypk1 activation by TORC2 (pT662) was no longer 429 elevated (Fig. 5G). In addition, vps4∆ orm2∆ cells grew better in presence of PalmC than vps4∆ 430 single mutants, suggesting a decreased susceptibility to PM stress (Fig. S5B). Furthermore, 431 deletion of ORM2 in vps4∆ ypk1∆ double mutants restored membrane integrity and decreased 432 15 the fraction of PI-positive cells to orm2∆ single mutant levels ( Fig S5C). Finally, the deletion 433 of ORM2 fully rescued the growth defect of vps4∆ ypk1∆ cells (Fig. 5H). A partial rescue was 434 also observed upon deletion of ORM1. 435

436
Our results suggest that calcineurin activity in ESCRT mutant cells hampered the homeostatic 437 regulation of Orm2 ER export. In turn, Orm2 accumulated at the ER and was no longer 438 efficiently degraded by EGAD. The accumulation of Orm2 at the ER renders the growth of 439 ESCRT mutants dependent on TORC2-Ypk1 signaling to prevent imbalance in SL homeostasis 440 and loss of plasma membrane integrity (Fig. S5D) How ESCRT-III was recruited to these structures is currently unclear, but several scenarios 462 appear possible. ESCRT-III/Vps4 recruitment could be driven by ESCRT-0, -I or -II complexes 463 interacting with the ubiquitinated membrane proteins in these stalling endocytic buds. Indeed, 464 ESCRT-II mutants also genetically interact with YPK1 (Fig. S1B). Alternatively, Bro1 -like 465 Alix in human cells -might help to recruit ESCRT-III directly to the PM. In budding yeast, 466 ESCRT-III subunits (Snf7) are also recruited to the PM during adaptation to alkaline pH and 467 are critically required for the regulator of Ime2 (RIM) pathway (80). Alternatively, changes in 468 PM tension might recruit the ESCRT machinery as described (81,82 The results from our genetic screen indicated that ESCRT-deficient cells are particularly 480 sensitive to perturbations in the homeostasis of two classes of lipids: the fungal cholesterol 481 analog ergosterol and sphingolipids. In concert, these lipids are known to promote membrane 482 rigidity and also to stabilize many membrane proteins (83-86). In yeast, they are important for 483 the formation of eisosomes (87) Lag1 and Lac1 (32). Failure to phosphorylate CS also causes accumulation of LCBs and 502 diminished formation of complex SL. Thus, we speculate that in vps4∆ ypk1∆ cells the 503 coordination of SPT and CS activity is lost. Probably SPT activity can still be to some degree 504 de-repressed by TORC2 via Ypk2, whereas CS activity cannot be efficiently stimulated when 505 Ypk1 is deleted (32). This may additionally be aggravated by the accumulation of Orm2 in the 506 ER, because high Orm2 levels can diverge Ypk1 (and hence probably also Ypk2) activity away 507 from other important targets (32). Collectively, this leads to a built up of LCBs in vps4∆ ypk1∆ 508 cells and the ensuing defects in membrane integrity. Consistently, mimicking phosphorylation 509 of Lag1 slightly improved the growth of vps4∆ ypk1∆ mutants (Fig. 3F). 510

18
The major culprit for the accumulation of Orm2 at the ER in ESCRT mutants appears to be its 512 premature dephosphorylation in a calcineurin-dependent manner, which decreases ER export 513 of Orm2 and thus slows down Orm2 degradation through the EGAD pathway. Calcineurin is 514 activated by elevated intracellular calcium levels, and calcineurin activity is probably increased 515 in ESCRT mutants (76). ESCRT mutants have been found to be sensitive to calcium, which 516 can be suppressed by calcineurin inhibition or by stimulating intracellular calcium storage 517 (75,96). vps4∆ mutants upregulate the Golgi calcium pump Pmr1 (97), possibly to decrease 518 intracellular calcium. Pmr1 is also essential for the survival of ESCRT mutants (9). Similarly, 519 Csg2 is required for growth under high calcium conditions (98) and for growth of vps4∆ 520 mutants ( Fig. S1A), indicating tight links between the ESCRT machinery, calcium signaling 521 and the regulation of SL biosynthesis. 522

523
The growth defect of vps4∆ ypk1∆ double mutants was alleviated by calcineurin inhibition and 524 extracellular calcium chelation, suggesting that increased calcium uptake drives detrimental 525 calcineurin activity in these cells. Increased calcium influx in ESCRT mutants may be caused 526 by deregulated trafficking or turnover of the plasma membrane calcium channel (Cch1/Mid1), 527 which is likely a substrate of the ESCRT machinery. Mid1 could also respond to increased 528 membrane stress in ESCRT mutants, since Mid1 is a stretch-activated channel and allows 529 calcium influx and calcineurin signaling to respond to mechanical membrane stress (99). 530 Alternatively, ER-PM contact sites may be corrupted in ESCRT mutants. The disruption of 531 these contact sites leads to increased calcium influx and calcineurin activity, which increased 532 LCBs and decreased ceramide levels (60). 533 534 At the moment, it is unclear whether calcineurin directly dephosphorylates Orm2. The sequence 535 of Orm2 does not contain an obvious calcineurin binding site (100,101). The Cdc55-containing 536 protein phosphatase 2A (PP2A) controls Orm2 phosphorylation levels during heat shock (102). 537 Inactivation of Cdc55-PP2A or another phosphatase upon calcineurin-inhibition cannot be 538 ruled out as an indirect cause leading to dephosphorylation of Orm2, but to our knowledge has 539 not been reported. However, since calcineurin antagonizes TORC2 and Ypk1 activity also 540 towards other targets (29,32,74,103), it might also directly contribute to the phosphorylation 541 status of Orm2. In addition to lowering Orm2 protein levels, inhibition of calcineurin is also 542 expected to promote phosphorylation of the CS subunits Lag1/Lac1 and thereby formation of 543 ceramides, which mitigates the buildup of harmful LCBs (32). Thus, calcineurin inactivation 544 probably has several beneficial effects in ESCRT mutants, which collectively render their 545 growth largely independent of Ypk1 signaling. 546 547 In summary, we propose that ESCRT function maintains PM homeostasis on several levels: (i) 548 through the MVB pathway by degrading membrane proteins and preventing proteotoxic stress 549 (104), (ii) by preserving or repairing the PM in response to membrane stress, and (iii) by 550 maintaining calcium homeostasis, which is important for controlled output of TORC2-Ypk1 551 signaling and regulation of Orm2. The latter promotes biosynthetic SL flux and thereby 552 preserves PM homeostasis. In ESCRT mutants, regulation of Orm2 is disturbed because of 553 increased calcineurin-dependent dephosphorylation, and hence its ER export and turnover 554 through EGAD is impaired. The buildup of Orm2 protein levels in the ER constitutes a stress 555 that must be counteracted through increased TORC2-Ypk1 activity (Fig. S5D). As long as this 556 is possible, ESCRT mutants have little fitness defects. However, if phosphorylation of Orm2 is 557 additionally impaired by any means (deletion of YPK1; inhibition of TORC2), or if plasma 558 membrane homeostasis is acutely challenged, ESCRT mutants are no longer able to 559 compensate. Now their PM integrity becomes compromised, and viability decreases 560 dramatically. Importantly, the mutual dependence of the ESCRT machinery and TORC2-Ypk1 Genetic modifications were performed by PCR and/or homologous recombination using 574 standard techniques. Plasmid-expressed genes including their native promoters and terminators 575 were amplified from yeast genomic DNA and cloned into centromeric vectors (pRS series) 576 (105). Tagged version of Orm1 and Orm2 (3xHA, 3xFLAG, or GFP) and their respective 577 mutants were expressed from plasmids in the respective deletion mutants replacing the 578 endogenous protein (with exception of Fig. 5D, where FLAG-Orm2 was co-expressed). All 579 constructs were analyzed by DNA-sequencing and transformed into yeast cells using standard 580 techniques. Genotypes of yeast strains and plasmids used in this study as well as primer for 581 PCR-based genetic modifications and cloning are listed in Table S2. 582 583 All S. cerevisiae strains in this study were SEY6210 derivatives. For liquid cultures, cells were 584 incubated in YNB synthetic medium supplemented with amino acids (according to respective 585 auxotrophies) and 2% glucose at 26°C in a shaker and were grown to midlog phase (OD600=0.5 586 -0.8). For growth on agar plates, yeast cells were diluted to OD600nm = 0.05 and spotted in 587 serial dilutions on YPD or YNB (auxotrophic selection medium) plates at the indicated 588 conditions. Rapamycin (Sigma) was added at the indicated concentrations from a 1 mM stock 589 in DMSO and myriocin (Sigma) was added to 1.5 µM from a 5 mM stock in methanol. 590 Untreated controls were supplied with the appropriate amount of the respective solvent. For the 591 assessment of growth in presence of PalmC, cells were grown into log phase, diluted to an 592 OD600nm of 0.1 into fresh YNB medium containing PalmC (Sigma) at the indicated 593 concentration (from a 5 mM stock in DMSO) and grown at 26°C, 180 rpm. For the assessment 594 of growth in presence of FK-506 (Sigma; stock 20 mg/ml in DMSO) or EGTA (Sigma; stock 595 500 mM in water) cells were grown into log phase, diluted to an OD600nm of 0.05 into fresh 596 YPD medium containing 2 µg/ml FK-506 or 0.1 mM EGTA. After 24 hours the OD600nm was 597 measured. 598 599

Analysis of genetic interaction data 600
Gene Ontology (GO) enrichment analysis (106) was performed using the 119 genes listed in 601 Appendix Table S1 of (9) as synthetically lethal with vps4∆ in two biological replicates or lethal 602 in one and sick in the other replicate. They were analyzed with the 'generic GO-Slim: process' 603 terms using the GO Slim mapper of the Saccharomyces genome database. GO term fusion and 604 enrichment analysis was performed as described (9). Only significantly scored GO terms (p < 605 0.05) are presented in Fig. 1A. The full analysis is presented in Table S1. 606 607

Preparation of whole cell protein extracts, Western blot analysis and immunodetection 608
To prepare whole cell lysates, proteins were extracted by alkaline extraction (107). When 609 protein phosphorylation was also analyzed, extraction was done by a modified protocol with 610 phosphatase inhibition as described (9). Protein extracts were denatured in Lämmli sample 611 buffer, separated by SDS-PAGE (Biorad Mini Protean) and transferred to PVDF membranes 612 by semi-dry electro-blotting. Phos-tag SDS PAGE as well as immunodetection of ubiquitinated 613 Orm2 were done as described (9). Antibodies used in this study are listed in Table S2. 614