Disturbed intramitochondrial phosphatidic acid transport impairs cellular stress signaling

Lipid transfer proteins of the Ups1/PRELID1 family facilitate the transport of phospholipids across the intermembrane space of mitochondria in a lipid-specific manner. Heterodimeric complexes of yeast Ups1/Mdm35 or human PRELID1/TRIAP1 shuttle phosphatidic acid (PA) mainly synthesized in the endoplasmic reticulum (ER) to the inner membrane, where it is converted to cardiolipin (CL), the signature phospholipid of mitochondria. Loss of Ups1/PRELID1 proteins impairs the accumulation of CL and broadly affects mitochondrial structure and function. Unexpectedly and unlike yeast cells lacking the CL synthase Crd1, Ups1-deficient yeast cells exhibit glycolytic growth defects, pointing to functions of Ups1-mediated PA transfer beyond CL synthesis. Here, we show that the disturbed intramitochondrial transport of PA in ups1Δ cells leads to altered unfolded protein response (UPR) and mTORC1 signaling, independent of disturbances in CL synthesis. The impaired flux of PA into mitochondria is associated with the increased synthesis of phosphatidylcholine and a reduced phosphatidylethanolamine/phosphatidylcholine ratio in the ER of ups1Δ cells which suppresses the UPR. Moreover, we observed inhibition of target of rapamycin complex 1 (TORC1) signaling in these cells. Activation of either UPR by ER protein stress or of TORC1 signaling by disruption of its negative regulator, the Seh1-associated complex inhibiting TORC1 complex, increased cytosolic protein synthesis, and restored glycolytic growth of ups1Δ cells. These results demonstrate that PA influx into mitochondria is required to preserve ER membrane homeostasis and that its disturbance is associated with impaired glycolytic growth and cellular stress signaling.


Edited by Dennis Voelker
Lipid transfer proteins of the Ups1/PRELID1 family facilitate the transport of phospholipids across the intermembrane space of mitochondria in a lipid-specific manner. Heterodimeric complexes of yeast Ups1/Mdm35 or human PRELID1/TRIAP1 shuttle phosphatidic acid (PA) mainly synthesized in the endoplasmic reticulum (ER) to the inner membrane, where it is converted to cardiolipin (CL), the signature phospholipid of mitochondria. Loss of Ups1/PRELID1 proteins impairs the accumulation of CL and broadly affects mitochondrial structure and function. Unexpectedly and unlike yeast cells lacking the CL synthase Crd1, Ups1-deficient yeast cells exhibit glycolytic growth defects, pointing to functions of Ups1mediated PA transfer beyond CL synthesis. Here, we show that the disturbed intramitochondrial transport of PA in ups1Δ cells leads to altered unfolded protein response (UPR) and mTORC1 signaling, independent of disturbances in CL synthesis. The impaired flux of PA into mitochondria is associated with the increased synthesis of phosphatidylcholine and a reduced phosphatidylethanolamine/phosphatidylcholine ratio in the ER of ups1Δ cells which suppresses the UPR. Moreover, we observed inhibition of target of rapamycin complex 1 (TORC1) signaling in these cells. Activation of either UPR by ER protein stress or of TORC1 signaling by disruption of its negative regulator, the Seh1-associated complex inhibiting TORC1 complex, increased cytosolic protein synthesis, and restored glycolytic growth of ups1Δ cells. These results demonstrate that PA influx into mitochondria is required to preserve ER membrane homeostasis and that its disturbance is associated with impaired glycolytic growth and cellular stress signaling.
The functional integrity of mitochondria depends on characteristic lipid compositions of their membranes, the mitochondrial outer and inner membrane (OM and IM) (1,2). This is exemplified by the hallmark lipid of mitochondria, cardiolipin (CL), which is pivotal for the structure and function of mitochondria. CL maintains respiration and cristae morphogenesis, ensures protein biogenesis and affects the fusion and fission of mitochondrial membranes (3)(4)(5)(6). Altered CL levels in the OM modulate mitophagy and apoptosis (7)(8)(9)(10)(11). Hence, reduced levels and aberrant acylation or peroxidation of CL compromise mitochondrial activities and are associated with aging and various pathophysiological conditions, including cardiomyopathies, skeletal myopathies, ataxias, or nonalcoholic fatty liver disease (12)(13)(14)(15).
CL is synthesized along an enzymatic cascade at the IM from phosphatidic acid (PA), which is mainly synthesized in the endoplasmic reticulum (ER) and imported into mitochondria (16). Although it has been reported that PA can be produced in mitochondrial membranes by phospholipases and kinases (17,18), the maintenance of mitochondrial lipid homeostasis requires extensive exchange of phospholipids between the ER and mitochondria (19,20). Similar to other membrane lipids, PA is transported from the ER to the OM at membrane contact sites, which facilitate the bi-directional transport of phospholipids between both organelles. Studies on the cellular distribution of newly synthesized phosphatidylserine (PS) revealed that phospholipid transfer to mitochondria can be driven by localized synthesis (21,22). After reaching the OM, phospholipids are transported across the intermembrane space by conserved, lipid-specific lipid transfer proteins (23)(24)(25)(26)(27)(28), which likely act in concert with MICOS membrane tethering complexes (24). Heterodimeric complexes of yeast Ups1 (human PRELID1) and Mdm35 (human TRIAP1) serve as PA-specific lipid transfer proteins (23,26). Crystal structures of various members of the conserved Ups1/ PRELID1 family of lipid transfer proteins show an internal lipid binding pocket, which is surrounded by a β-sheet and αhelices, reminiscent of other classes of lipid transfer proteins (29)(30)(31)(32). Loss of Ups1 or PRELID1 in yeast or human cells, respectively, inhibits intramitochondrial PA transport and limits CL synthesis at the IM, which is accompanied by mitochondrial deficiencies, such as impaired respiration and mitochondrial fragmentation, and increases the susceptibility of the cells toward apoptotic stimuli (23,26,(33)(34)(35).
High levels of PRELID1 and TRIAP1 were identified as an unfavorable prognostic marker in cancer (36). As cancer cells are often glycolytic, it is noteworthy that yeast cells lacking Ups1 also exhibit slow growth in glucose media (23,34,35). This phenotype is difficult to reconcile with reduced CL levels only, as cells lacking the CL synthase Crd1 in mitochondria This article contains supporting information. grow normally in glucose-containing media although they are virtually devoid of CL (37). It therefore appears that the inhibition of intramitochondrial PA transport upon loss of Ups1 does not only perturb the mitochondrial membrane homeostasis but also causes deficiencies presumably beyond mitochondrial functions. Here, we demonstrate that a disturbed intramitochondrial transport of PA affects TORC1 signaling and alters the phospholipid homeostasis of the ER membrane modulating the unfolded protein response (UPR).

The loss of Ups1 impairs basal UPR
To examine how the loss of Ups1 affects cell growth under glycolytic conditions, we used quantitative mass spectrometry (qMS) to determine the phospholipid profile of cellular membranes in wild-type and ups1Δ cells which were grown in glucose-containing media. In agreement with previous findings (23, 34), we observed reduced CL levels in ups1Δ cells, reflecting impaired PA transport to CL-synthesizing enzymes at the IM (Fig. 1, A and B and Fig. S1A). We also noted a significant increase of phosphatidylcholine (PC) and a decrease of phosphatidylethanolamine (PE), resulting in a low PE/PC ratio in cellular membranes (Fig. 1, A-C and Fig. S1A). Unexpectedly, deletion of UPS1 also affected the phospholipid composition of ER membranes. We observed an increased total phospholipid content of an ER-enriched membrane fraction, which was accompanied by a significant increase of PC and phosphatidylinositol (PI) (Fig. 1, D and E and Fig. S1B), and a low PE/PC ratio (Fig. 1F).
Previous studies have reported that an increased PE/PC ratio in ER membranes activates the UPR independent of protein stress (38)(39)(40)(41). We therefore monitored UPR induction in ups1Δ cells harboring GFP under the control of the UPR element (UPRE) in their genome (42). Whereas GFP accumulated in wild-type cells under basal conditions, GFP expression was decreased in cells lacking Ups1 which were grown in glucose- (Fig. 1, G and H) or galactose-containing media (Fig. S1, C and D). Moreover, we observed reduced mRNA levels of spliced HAC1, encoding a transcription factor involved in UPR (Fig. 1I) (43)(44)(45). We therefore conclude that basal gene expression under the control of UPRE is suppressed in ups1Δ cells. Notably and in contrast to ups1Δ cells, cells lacking other enzymes involved in CL synthesis (such as Tam41, Pgs1, or Crd1) did not show significant reduction of the GFP expression under UPRE (Fig. S1, E-H), pointing that CL defect per se does not lead to the suppression of UPR.
A central sensor of UPR is the ER kinase Ire1, which senses unfolded proteins in the ER lumen as well as lipid bilayer stress in the membrane (43,(46)(47)(48). To assess the contribution of lipid bilayer stress specifically, we exploited the previously described ΔIII mutant of Ire1, which is able to sense lipid bilayer stress but not unfolded ER proteins (49,50). Cells expressing Ire1(ΔIII) recapitulated the suppression of the UPR by the loss of Ups1 (Fig. 1, J and K). These results suggest that the altered phospholipid composition of ER membranes limits UPR activation in ups1Δ cells.

Increased PC levels cause suppression of basal UPR
To examine how a decreased PE/PC ratio in ER membranes affects UPR, we induced de novo synthesis of PC via the Kennedy pathway by supplementing the growth medium with choline (51) and monitored UPR activity. Choline supplementation increased PC levels in ER-enriched membrane fractions of wild-type cells and led to a lower PE/PC ratio in these membranes (Fig. 2, A and  B). Similar to cells lacking Ups1, the decreased PE/PC ratio in ER membranes was accompanied by an impaired basal UPR (Fig. 2, C and D), which was not altered in cells expressing mutant Ire1(ΔIII) (Fig. 2, E and F). We conclude from these experiments that the basal UPR is impaired if the PE/PC ratio in ER membranes is decreased, whereas an increased PE/PC ratio in ER membranes induces the UPR (41).
To unambiguously demonstrate that Ups1 affects UPR activation by modulating the phospholipid profile in ER membranes, we manipulated the PE/PC ratio in ups1Δ cells genetically. Yeast cells lacking Cho2 or Opi3, which are required for synthesis of PC from PE (51), accumulate PE relative to PC (Fig. 2G). Accordingly, deletion of CHO2 or OPI3 induces the UPR (41). We therefore deleted CHO2 or OPI3 in ups1Δ cells, which strongly increased the PE/PC ratio in an ER-enriched membrane fraction (Fig. 2, G and H). This was accompanied by UPR activation in the absence of Ups1 (Fig. 2, I and J). Thus, ups1Δ cells maintain the ability to induce the UPR upon lipid bilayer stress, but the reduced PE/PC ratio in the ER membrane of ups1Δ cells impairs the UPR under basal conditions.

UPR activation restores glycolytic growth of ups1Δ cells
Having established the link between Ups1 and UPR activation, we examined in further experiments whether the reduced UPR activity can explain the impaired glycolytic growth of ups1Δ cells. We therefore treated ups1Δ cells with tunicamycin, an inhibitor of protein glycosylation, which causes ER protein stress and induces the UPR. Treatment with tunicamycin activated the UPR both in wild-type and in ups1Δ cells (Fig. 3A) and allowed growth of ups1Δ cells on fermentable medium (Fig. 3B). Consistently, activation of the UPR upon overexpression of Ire1 ( Fig. 3C) (47,52) or deletion of CHO2 or OPI3 (Fig. 2, I and J) suppressed growth deficiencies of ups1Δ cells on glucose-containing medium (Fig. 3, D and E). The restoration of cell growth by tunicamycin and overexpression of Ire1 depended on Ire1-Hac1 signaling (Fig. 3, F and G), demonstrating that UPR activation is sufficient to promote glycolytic growth of ups1Δ cells. Notably, previous studies correlated the growth of ups1Δ cells on glucosecontaining media with mitochondrial CL levels (53). However, overexpression of Ire1 did not alter the accumulation of CL (Fig. 3G). We therefore conclude that UPR activation is sufficient to allow growth of ups1Δ cells on fermentable carbon sources independent of CL levels in mitochondrial membranes.

Reduced cellular protein synthesis in ups1Δ cells
We noted that the loss of UPR activity in ire1Δ and hac1Δ cells did not limit growth in glucose-containing medium Data represented as log2 fold change at x-axis with −log 10 p value of student's t test at y-axis. Dashed line represents p = 0.05. C, PE/PC ratio of WT and ups1Δ cells. D, phospholipid composition in ER-enriched microsome fraction from WT and ups1Δ cells grown to log phase in SCD medium. Data represent mean ± SD (n = 3). *p < 0.05, **p < 0.01. E, changes in the absolute abundance of phospholipids in ER-enriched microsome fraction of ups1Δ. Data represented as log2 fold change at x-axis with −log 10 p value of Student's t test at y-axis. Dashed line represents p = 0.05. F, PE/PC ratio in ER-enriched microsome fraction from WT and ups1Δ cells. G, WT and ups1Δ cells expressing 4xUPRE-GFP grown to log phase in SCD medium were subjected to Western blotting. Pgk1 and ponceau staining were monitored as a loading control. H, GFP in (G) was quantified. GFP signals were normalized to Pgk1 and expressed relative to WT cells (set as one). Data represent mean ± SD (n = 3). I, WT and ups1Δ cells were grown to log phase in SCD medium. Spliced HAC1 mRNA expression was analyzed by real time PCR and normalized to ACT1 mRNA expression. Data represent mean ± SD (n = 5). J, WT and ups1Δ cells expressing Ire1(ΔIII)-3HA-GFP and 4xUPRE-GFPgrown to log phase in SCD medium were subjected to Western blotting. K, GFP in (J) was quantified as described in (H). Data represent mean ± SD (n = 3). ER, endoplasmic reticulum; PC, phosphatidylcholine; PE, phosphatidylethanolamine; SCD, synthetic complete glucose; UPRE, UPR element. Figure 2. Basal UPR is suppressed in high PC levels. A, phospholipid composition in ER-enriched microsome fraction from wild-type (WT) cells grown to log phase in SCD medium with or without 1 mM choline. Data represent mean ± SD (n = 3). B, PE/PC ratio in ER-enriched microsome fraction from WT with or without choline. C, WT cells expressing 4xUPRE-GFP grown to log phase in SCD medium with or without 1 mM choline were subjected to Western blotting. Pgk1 and ponceau staining were monitored as a loading control. D, GFP in (C) was quantified. GFP signals were normalized to Pgk1 and expressed relative to WT cells (set as one). Data represent mean ± SD (n = 3). E, WT cells expressing Ire1(ΔIII)-3HA-GFP and 4xUPRE-GFP grown to log phase in SCD medium with or without 1 mM choline were subjected to Western blotting. F, GFP in (E) was quantified as described in (D). Data represent mean ± SD (n = 3). G, phospholipid composition in ER-enriched microsome fraction from WT, ups1Δ, cho2Δ and ups1Δcho2Δ (ΔΔ) cells grown to log phase in SCD medium. Data represent mean ± SD (n = 3). *p < 0.05, **p < 0.01. H, PE/PC ratio of ER-enriched microsome fraction from WT, ups1Δ, cho2Δ and ups1Δcho2Δ (ΔΔ) cells. I, WT, cho2Δ, opi3Δ, ups1Δ, ups1Δcho2Δ, and ups1Δopi3Δ cells expressing 4xUPRE-GFP grown to log phase in SCD medium were subjected to Western blotting. J, GFP in (I) was quantified as described in (D). Data represent mean ± SD (n = 3). ER, endoplasmic reticulum; PC, phosphatidylcholine; PE, phosphatidylethanolamine; SCD, synthetic complete glucose; UPR, unfolded protein response; UPRE, UPR element.
Loss of Ups1 limits UPR activation & impairs TORC1 signaling ( Fig. 3F). This contrasts ups1Δ cells and therefore points to additional deficiencies in ups1Δ cells that limit cell growth. We therefore determined transcriptome and proteome profiles of ups1Δ cells under glycolytic growth conditions (Tables S1 and  S2). Consistent with the observed impaired UPR activity, canonical UPR target genes were expressed at lower levels in ups1Δ cells when compared with wild-type cells (Fig. S2A). Both transcriptome and proteome profiles pointed to reduced protein synthesis in ups1Δ cells (Fig. 4, A and B). The expression of genes involved in processes driving protein synthesis, such as rRNA processing or ribosome biogenesis, was most prominently downregulated in ups1Δ cells ( Fig. 4A and Table S3). Consistently, proteins regulating rRNA processing accumulated at reduced levels in ups1Δ cells ( Fig. 4B and Table S4). Together, these data demonstrate that protein synthesis is inhibited in Ups1-deficient cells. The expression of genes related to protein synthesis is decreased under stress conditions (54). Consistently, tunicamycin treatment limited the expression of genes involved in cellular protein synthesis (Fig. 4C). However, it partially promoted cellular protein synthesis in ups1Δ cells ( Fig. 4D and Fig. S2C), consistent with the improved cell growth under these conditions (Fig. 3B).

TORC1 inhibition limits glycolytic growth of ups1Δ cells
The TORC1 regulates ribosome biogenesis in response to the metabolic status of cells (55). We noted an increased expression of genes coding core autophagy proteins in ups1Δ cells when compared with wild-type cells ( Fig. S2B and Table S1), suggesting inhibition of TORC1 and induction of autophagy in ups1Δ cells. To directly assess TORC1 activity in ups1Δ cells, we monitored the level and phosphorylation of one of its substrates, Sch9 (56), an AGC family protein kinase, which is homologous to mammalian S6 and Akt kinases and regulates ribosome biogenesis (57,58). We found that total Sch9 protein levels were not changed (Fig. 5, A and B), but Sch9 phosphorylation was significantly suppressed in ups1Δ cells compared with wild-type cells (Fig. 5, C and D). These results indicate low TORC1 activity in ups1Δ cells, which agrees with the reduced protein synthesis in these cells.
We hypothesized that, similar to UPR activity, an altered membrane lipid composition may inhibit TORC1 activity in ups1Δ cells. However, neither loss of the CL synthase Crd1 to decrease CL levels (Fig. S3A) nor choline supplementation of wild-type cells to mimic the PE/PC ratio of ups1Δ cells affected TORC1 activity and phosphorylation of Sch9 (Fig. S3, B-E). Similarly, decreased PE levels upon loss of the mitochondrial PS transfer protein Ups2 did not alter TORC1 activity (Fig. S3, B and C). TORC1 activity is dependent on PA (59, 60), but disturbed intramitochondrial PA transport in ups1Δ cells did not result in the accumulation of PA in ER-enriched membrane fractions (Fig. 1, D and E). Moreover, PA accumulation upon inhibition of the PA phosphatase in nem1Δ or spo7Δ cells did not affect Sch9 phosphorylation (Fig. S3, F and G). Thus, other deficiencies appear to inhibit TORC1 activity in ups1Δ cells.
To examine whether TORC1 inhibition limits the glycolytic growth of ups1Δ cells, we used a genetic approach. The Seh1-associated complex inhibiting TORC1 consisting of Npr2, Npr3, and Iml1 functions as negative regulator of TORC1 in yeast (61,62). Constitutive activation of TORC1 upon deletion of one of the Seh1-associated complex inhibiting TORC1 subunits restored cell growth of ups1Δ cells (Fig. 5, E and F). Moreover, expression of a phosphomimetic mutant variant of Sch9, Sch9 2D3E , greatly restored the growth of ups1Δ (Fig. 5G). We therefore conclude that inhibition of TORC1-Sch9 signaling impairs the growth of ups1Δ cells on a fermentable carbon source such as glucose. Notably, TORC1 activation upon expression of Sch9 2D3E in ups1Δ cells did not restore UPR activity (Fig. 5, H and I), indicating that both pathways promote glycolytic growth of ups1Δ cells independently.

Discussion
Ups1-dependent transport of PA to the IM promotes CL synthesis (23). Here, we have unraveled an unexpected link between intramitochondrial PA transport and cellular signaling (Fig. 6). Disturbed PA transport in mitochondria lacking Ups1 limits the activation of UPR and inhibits TORC1. TORC1 signaling controls ribosome biogenesis and protein synthesis, which is suppressed in ups1Δ cells and limits glycolytic cell growth. Accordingly, activation of either UPR or TORC1 is sufficient to suppress growth deficiencies in the absence of Ups1. Thus, the flux of PA into mitochondria does not only ensure mitochondrial CL synthesis and membrane biogenesis but is coupled to the phospholipid homeostasis of other cellular membranes affecting cell signaling.
Several lines of evidences reveal that the disturbed intramitochondrial PA transport in ups1Δ cells impairs independently CL synthesis and UPR and TORC1 signaling. The activation of the UPR allowed glycolytic growth of ups1Δ cells but did not restore the accumulation of CL in mitochondrial membranes (Fig. 3G) or the phosphorylation of Sch9 (Fig. 5, C and D). On the other hand, mimicking the PE/PC ratio observed in ups1Δ cells by choline supplementation of wild-type cells suppressed UPR but did not alter the phosphorylation of Sch9 (Fig. S3, D and E). Conversely, expression of phosphomimetic Sch9 2D3E mutant did not restore basal UPR in ups1Δ cells (Fig. 5,  H and I). These findings are consistent with a previous study demonstrating that activation of TORC1 does not affect the UPR (63). Moreover, CL deficiencies in cells lacking the CL synthase Crd1 did not suppress TORC1 signaling (Fig. S3, B and C) or basal UPR (Fig. S1, E and F). Therefore, we conclude that the impaired intramitochondrial PA transport rather than the CL deficiency leads to the alteration of stress signaling cascades outside of ups1Δ mitochondria.
Inhibition of PA transport across the intermembrane space in ups1Δ cells is associated with an increased PC and PI and a decreased PE/PC ratio but not with a significant accumulation of PA in the ER membrane. This indicates efficient bi-directional transport of PA between the OM and the ER membrane and suggests that PA is rapidly metabolized in the ER, if it is not consumed for CL synthesis in ups1Δ mitochondria. Because PA serves as a common precursor for the synthesis of PC and CL, PA is likely converted to PC resulting Loss of Ups1 limits UPR activation & impairs TORC1 signaling in lowered PE/PC ratios in the ER membrane and the inhibition of UPR. Rapid conversion to PC can also explain why impaired PA utilization in mitochondria does not suppress Opi1, the transcriptional repressor of phospholipid synthesis, which is regulated by PA (51). Transcript profiling revealed no increased expression of Opi1 target genes in ups1Δ cells, consistent with the unaltered PA levels and increased PI in ER membranes in ups1Δ cells. The uncontrolled accumulation of PC and PI in the ER membrane may also retard functionality of the ER in addition to the limited UPR activation, contributing to the impaired glycolytic growth in ups1Δ cells.
The low PE/PC ratio in ER membranes suppresses the basal UPR in ups1Δ cells, complementing previous findings which revealed UPR activation upon an increase of the PE/PC ratio (41). The ratio of PE and PC determines membrane packing and curvature stress in the ER membrane, which appears to be sensed by the Ire1 kinase. Indeed, an amphipathic helix adjacent to the transmembrane domain of Ire1 was found to detect biophysical changes at the ER membrane (64). It therefore can be envisioned that the low PE/PC ratio may improve membrane packing and restrain the formation of the X-shaped transmembrane dimer of Ire1.  Figure 3. Glycolytic growth of ups1Δ cells is restored upon UPR activation. A, wild-type (WT) and ups1Δ cells expressing 4xUPRE-GFP grown to log phase in SCD medium were treated with 1 μg/ml tunicamycin, collected at the indicated time points, and subjected to Western blotting. Pgk1 and ponceau staining were monitored as a loading control. B, serial dilutions of WT and ups1Δ cells were spotted on SCD medium with or without tunicamycin (concentration as 0.25 or 0.5 μg/ml) and incubated at 30 C for 2 days (n = at least 3). C, 4xUPRE-GFP expressing WT and ups1Δ cells transformed with an empty vector or a plasmid encoding Ire1 (pIRE1) were grown to log phase in SCD medium and subjected to Western blotting. D, serial dilutions of WT and ups1Δ cells transformed with an empty vector or pIRE1 were spotted on SCD medium and incubated at 30 C for 2 days (n = at least 3). E, serial dilutions of WT, cho2Δ, opi3Δ, ups1Δ, ups1Δcho2Δ, and ups1Δopi3Δ cells were spotted on SCD medium and incubated at 30 C for 2 days (n = at least 3). F, serial dilutions of WT, ire1Δ, hac1Δ, ups1Δ, ups1Δire1Δ, ups1Δhac1Δ cells were spotted on SCD medium with or without 0.25 μg/ml tunicamycin and incubated at 30 C for 2 days (n = at least 3). G, serial dilutions of WT, ire1Δ, hac1Δ, ups1Δ, ups1Δhac1Δ cells transformed with an empty vector or pIRE1 were spotted on SCD medium and incubated at 30 C for 2 days (n = 3). H, phospholipid composition in whole cell of WT and ups1Δ cells transformed with an empty vector or pIRE1. Data represent mean ± SD (n = 3). SCD, synthetic complete glucose; UPR, unfolded protein response; UPRE, UPR element.
Protein synthesis is suppressed under stress conditions (54). Our transcriptome data correspondingly show that expression levels of genes involved in ribosome biogenesis are decreased in wild-type cells treated with tunicamycin. On the other hand, tunicamycin induces a number of those genes in ups1Δ cells in accordance with the recovery of glycolytic growth of the cells. One remaining question is how tunicamycin treatment induces genes involved in protein synthesis, although it does not restore TORC1 signaling in ups1Δ cells. In contrast to the previous indication that the suppression of protein synthesis with tunicamycin proceeds independent of the UPR (65), the growth recovery of ups1Δ cells by tunicamycin depends on Ire1 and Hac1. Therefore, our results imply the existence of an UPR-dependent pathway promoting protein synthesis that is induced by proteotoxic stress in ups1Δ cells.
The loss of Ups1 also impairs TORC1 signaling independent of UPR inhibition. TORC1 inhibition in ups1Δ cells is not caused by alterations in the membrane lipid composition of the ER membrane, as neither a decreased PE/PC ratio, nor the accumulation of PA by other means (e.g., inhibition of PA phosphatase complex in nem1Δ or spo7Δ cells; Fig. S3, F and G) resulted in the inhibition of TORC1 activity. Notably, A C D B Figure 4. ups1Δ cells show suppressed ribosome synthesis. A, symplot representing the changed genes within an enriched biological process (BP) GO terms in ups1Δ cells against wild-type (WT) cells. Top ten significant terms from yeast GO-slim are plotted. Percentage of downregulated (blue) and upregulated (red) genes in each term is shown. The size of the central bars was directly proportional to the number of genes in the query belonging to the respective term, and the enrichment value is represented by the color of the bar as previously reported (71). B, 1D annotation enrichment analysis of proteome in ups1Δ cells against WT cells is shown. Annotation is biological process (BP) GO terms (76). Score represents difference of the distribution of the ratios for the proteins corresponding to GO terms from the ratios of the distribution for all proteins. Top ten significant terms are indicated. C, symplot representing the changed genes within a BP GO terms in WT cells treated with 1 μg/ml tunicamycin for 3 h against non-treated WT cells. Top ten significant terms from Yeast GO-slim are plotted. D, symplot representing the changed genes within an enriched BP GO terms in ups1Δ cells treated with 1 μg/ml tunicamycin for 3 h against nontreated ups1Δ cells. Top ten significant terms from yeast GO-slim are plotted. GO, Gene Ontology.

Loss of Ups1 limits UPR activation & impairs TORC1 signaling
previous studies have reported that TORC1-dependent Sch9 phosphorylation is inhibited by mitochondrial depolarization (66), coupling mitochondrial OXPHOS activity with cellular nutrient signaling. Because mitochondrial membrane potential is low in ups1Δ cells grown in glucose media (35), it is conceivable that mitochondrial depolarization upon inhibition of PA transport interferes with TORC1 signaling in ups1Δ cells. Interestingly, the strong induction of Pdr3, Pdr5, and Pdr15 in ups1Δ cells suggests induction of the mitochondrial compromised protein import response (67), which is in line with mitochondrial depolarization and import defects in ups1Δ cells (35). Therefore, it is tempting to speculate that signaling cascades in response to mitochondrial protein import stress might contribute to the TORC1 inhibition in ups1Δ cells  Figure 5. TORC1-Sch9 signaling is impaired in ups1Δ cells. A, wild-type (WT) and ups1Δ cells expressing Sch9-6HA grown to log phase in SCD medium were treated with DMSO (D), 1 μg/ml tunicamycin (Tu), or 0.2 μg/ml rapamycin (Rap) for 3 h and subjected to Western blotting. Pgk1 and ponceau staining were monitored as a loading control. B, Sch9-6HA of nontreated cells in (A) was quantified. Total Sch9-6HA signals were normalized to Pgk1 and expressed relative to wild-type cells (set as one). Data are means ± SD (n = 3). C, WT and ups1Δ cells expressing Sch9-6HA grown to log phase in SCD medium were treated with DMSO (D), 1 μg/ml tunicmycin (Tu), or 0.2 μg/ml rapamycin (Rap) for 3 h. For the analysis of Sch9 phosphorylation, lysates were treated with NTCB and subjected to Western blotting. D, Phosphorylated Sch9-6HA ratio in (C) was quantified. Phosphorylated Sch9-6HA signals were divided with total Sch9-6HA signals. Data represent mean ± SD (n = 3). *p < 0.05. E, Serial dilutions of WT, npr2Δ, ups1Δ, and ups1Δnpr2Δ cells were spotted on SCD medium and incubated at 30 C for 2 days (n = 3). F, Serial dilutions of WT, npr3Δ, ups1Δ, and ups1Δnpr3Δ cells were spotted on SCD medium and incubated at 30 C for 2 days (n = 3). G, Serial dilutions of WT and ups1Δ cells transformed with an empty vector or a plasmid encoding Sch9 WT (pSCH9), Sch9 5A (T723A, S726A, T737A, S758A, S765A), or Sch9 2D3E (T723D, S726D, T737E, S758E, S765E) were spotted on SCD medium and incubated at 30 C for 2 days (n = at least 3). H, 4xUPRE-GFP expressing WT and ups1Δ cells transformed with an empty vector or a plasmid encoding Sch9 WT , Sch9 5A , or Sch9 2D3E were grown to log phase in SCD medium and subjected to Western blotting. I, GFP in (H) was quantified. GFP signals were normalized to Pgk1 and expressed relative to WT cells (set as one). Data represent mean ± SD (n = 3). N.S., not significant; NTCB, 2-nitro-5-thiocyanobenzoic acid; SCD, synthetic complete glucose; UPRE, UPR element. (68). While further studies are needed to elucidate the link between mitochondrial depolarization and TORC1 signaling, our findings highlight that disturbances in the intramitochondrial transport of phospholipids and alterations in the phospholipid composition of mitochondrial membranes can broadly affect cellular signaling.

Yeast strains
Saccharomyces cerevisiae strains used in this study are based on S288c or W303. The genotype of the yeast strains is described in the strain list table (Table S5). Genes were deleted by PCR-targeted homologous recombination with the respective forward and reverse primers. Plasmids encoding disruption maker, HphNT1, NatNT2 or KanMX6, were used as template. C-terminal 6xHA tag of Sch9 at its endogenous loci was constructed using PCR-targeted homologous recombination with plasmid encoding 6HA-NatNT2 (69). For construction of cells expressing both Ire1(ΔIII)-3HA-GFP and 4xUPRE-GFP, 4xUPRE-GFP was integrated by PCR-targeted homologous recombination at the URA3 locus in cells expressing Ire1(ΔIII)-3HA-GFP kindly provided by R. Ernst (PZMS, Homburg). 4xUPRE-GFP-URA3 was amplified using as a template genomic DNA of a yeast strain, which contained this gene and which was kindly provided by M. Graef (MPI aging, Cologne) as a template.
For cell growth assay, cells were grown in SCD plate for 1 day at 30 C and suspended in water at a concentration of 0.1 OD 600 units. Four microliter of 5-fold serial dilutions of cells were spotted on SCD plates and incubated for 2 days at 30 C.

Preparation of ER-enriched microsomes
Hundred OD 600 units of cells grown to log phase in SCD medium were collected by centrifugation (3000g, 5  Whole cell homogenates were subjected to centrifugation (1200g, 5 min, 4 C). Membrane and microsomes soluble fraction were separated by centrifugation (17,500g, 5 min, 4 C). Supernatant was transferred to new tubes and subjected to centrifugation (40,000g, 40 min, 4 C). The pellet was resuspended in 200 μl of pure water, and the protein concentration was determined by BCA assay. Aliquots of 50 μg protein were kept at −80 C and were subjected to lipid analysis by qMS.
QuantStudio 5 (Thermo Fisher Scientific) was used for measurement. Data were analyzed according to the ΔΔCT method normalized to housekeeping gene ACT1.

RNA sequence for transcriptomics
Libraries were prepared using the Illumina TruSeq mRNA stranded sample preparation Kit. Library preparation started with 1 μg total RNA. After poly-A selection (using poly-T oligo-attached magnetic beads), mRNA was purified and fragmented using divalent cations under elevated temperature. The RNA fragments underwent reverse transcription using random primers. This is followed by second strand cDNA synthesis with DNA Polymerase I and RNase H. After end repair and A-tailing, indexing adapters were ligated. The products were then purified and amplified (14 PCR cycles) to create the final cDNA libraries. After library validation and quantification (Agilent tape station), equimolar amounts of library were pooled. The pool was quantified by using the Peqlab KAPA Library Quantification Kit and the Applied Biosystems 7900HT Sequence Detection System. The pool was sequenced with a PE100 run on a NovaSeq6000 sequencer.

Bioinformatics and data analysis for transcriptomics
Raw reads were quantified using the alignment-free quantification tool kallisto version 0.45.0. Reference genome is sacCer3. Gene counts were imported to R version 3.5.1 and normalized to library size using DESeq2 version 1.22.2. General differential gene expression was determined in pairwise comparisons using DESeq2. Functional enrichment of differentially expressed genes was performed using the DAVID API as previously reported (71). The Yeast Gene Ontology-slim list and information of gene function were obtained from the Saccharomyces Genome Database.

Protein lysis and digest for proteomics
One OD 600 units of cells were lysed in 40 μl of 2% SDC in 100 mM Tris-HCl [pH 8.0], and lysate was cleared by centrifugation (12,000 rpm, 10 min, 25 C). Supernatant was subjected to protein concentration determination. In total, 30 μg of protein were used for protein digestion. Proteins were reduced and alkylated by TCEP (10 mM) and CAA (20 mM) for 60 min at 45 C. 1 μg of LysC endopeptidase was added and incubated at 37 C or 2 h, followed by addition of 1 μg of Trypsin for digestion at 37 C for 16 h. Digestion was stopped by addition of TFA to a final concentration of 0.5%. Lysates were cleared (SDC precipitates) by centrifugation, and the supernatant was subjected for desalting using the StageTip (material: SDB-RPS, Affinisep) technique (72).
To generate the peptide spectral library, 2 μl of each sample (all conditions pooled 1:1) was pooled and subjected to high pH reversed phase chromatography. The instrumentation consisted out of a ZirconiumTM Ultra HPLC and a PAL RTC autosampler. The buffer systems consisted out of two buffers. (A) 10 mM ammonium hydroxide and (B) 80% acetonitrile and 10 mM ammonium hydroxide. Peptides were separated according to their hydrophobicity using an in-house packed column (length = 40 cm, inner diameter = 200 μm, 2.7-μm beads, PoroShell, Agilent Technologies) column. The instrument communicated and were controlled using the software Chronos (Axel Semrau GmbH). The gradient length was 60 min, and in total, 12 fractions were collected (1/60 s) and subsequently concentrated using a SpeedVac to complete dryness. Peptides were dissolved in 10 μl 2% formic acid, 2.5% acetonitrile of which 3 μl were injected per liquid chromatography and tandem mass spectrometry run.

Liquid chromatography and tandem mass spectrometry
For the MS/MS spectra library generation, the QExactive HF-x operated in a Top22 data-dependent mode. MS1 spectra were acquired in a mass range of 350 to 1750 m/z, using an AGC target of 3 × 106 and a resolution at 200 m/z of 60,000. MS/MS spectra were acquired at 15,000 resolution using an AGC target of 5 × 105 and a maximal injection time of 22 ms.
For data-independent acquisition measurements, MS1 spectra were acquired using a resolution of 60,000 and an AGC target of 1 × 106. For MS/MS independent spectra acquisition, 48 windows were acquired at an isolation m/z range of 15 Th and the isolation windows overlapped by 1 Th. The isolation center range covered a mass range of 385 to 1043 m/z. Fragmentation spectra were acquired at a resolution of 15,000 at 200 m/z using a maximal injection time of 22 ms and stepped normalized collision energies of 24, 27, 30. The default charge state was set to 4.

Bioinformatics and data analysis for proteomics
The MS/MS data-dependent spectra library was analyzed with MaxQuant 1.6.3.4 and the implemented Andromeda search engine (73) using default settings. The acquired MS2 spectra were correlated against the Uniprot reference Yeast proteome (downloaded 01.2019, 6007 entries). The maximum number of missed cleavages was set to 2. Carbamidomethylation at cysteine residues was defined as a fixed modification, whereas methionine oxidation and protein N-terminal acetylation were defined as variable modification. The false discovery rate (FDR) on the peptide spectrum match and protein level was estimated to 0.01 using the implemented decoybased algorithm. The mass tolerance on the peptide precursor level to 20 ppm (first search) and 4.5 ppm (main search). On the fragment level (MS2), the mass tolerance was set to 20 ppm. The output txt folder was then used to build the spectral library in Spectronaut (v. 13.0.190604). Data-independent acquisition runs were analyzed using the following settings in Spectronaut (v. 14.10.201222) utilizing the DDA library. Acquired runs were aligned using Precision iRT algorithm in Spectronaut by applying nonlinear regression. The algorithm is based on correlation of extracted ion currents to a consensus elution profile (74). Mass tolerances were set as default (40 ppm for MS1 and MS2 level). Q-value cutoff on precursor and protein level was set to 0.01 using the implemented decoy method "mutated". A maximum of three precursors were picked per peptide. Quantification was done on the MS2 level using the are under curve. Protein identifications with single peptide identifications in the library were removed as well as proteins that were identified by a single library peptide in DDA runs.
The data were exported in pivot-table format and further processed in Perseus (75). For pairwise comparison a twosided t test was utilized. The FDR was calculated by a permutation-based approach using 500 permutations and a fudge factor s0 of 0.1 A protein was considered to be significantly differently expressed at a FDR of 5%. No cutoff for the fold change was considered. 1D enrichments (76) were performed in Perseus software using the Gene Ontology annotations of the Uniprot database.

Statistical analysis
Quantitative data are presented as arithmetic means ± standard deviation. The statistical significance in related figures was assessed using two-tailed Student's t test. A p values and the number of experiments (n) are described in corresponding figure legend.

Data availability
All data are contained within the manuscript. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository (77) with the data set identifier: PXD022321. Transcriptomic data have been deposited to the GEO omnibus with the data set identifier GSE160861.