ER sensing of lipid metabolism drives PRA family-dependent regulation of COPII vesicle transport

Abstract

Newly synthesized secretory proteins and lipids are transported from the endoplasmic reticulum (ER) to the Golgi prior to their ultimate destinations, which is tightly regulated during adaptation to environmental stress. However, regulatory pathways governing the formation of COPII vesicles budded from the ER remain insufficiently explored. Here, we present evidence indicating that COPII-mediated vesicle transport is transcriptionally controlled through the phosphatidic acid (PA)-dependent Opi1-Ino2/Ino4 regulatory circuit. Our analysis shows that YIP3, a target gene of Ino2/Ino4, exerts a negative regulatory impact on COPII-mediated vesicle transport. We demonstrate that Ino2/Ino4, but not Yip3 modulates Sar1 activation, the initial step in COPII vesicle formation, whereas Yip3 hinders Sec16 assembly on the ER membrane, thereby implying that Ino2/Ino4 governs COPII vesicle formation at multiple steps. Finally, we show that under ER stress conditions which are accompanied by elevated PA, vesicular transport is restricted in a PA and Yip3-dependent manner. Thus, this study provides the first evidence for an ER sensing system that transcriptionally fine-tunes vesicle formation in response to alterations in lipid composition of the ER membrane during ER stress.

How cells coordinate vesicular transport during ER stress remains unclear. This study shows that the PA-dependent Opi1–Ino2/Ino4 regulatory circuit transcriptionally controls COPII vesicle formation through Yip3, thereby restricting ER-to-Golgi transport during ER stress.

Introduction

Coat protein complex II (COPII) proteins form vesicles that mediate the transport of cargo proteins and lipids from the endoplasmic reticulum (ER) to the Golgi apparatus^1–4. COPII vesicle formation occurs at a specialized high-curvature domain known as the ER exit site (ERES). ERES is marked by Sec16, which is thought to serve as an essential scaffold for COPII assembly. COPII assembly is triggered by the activation of the small GTPase Sar1 by its guanine nucleotide exchange factor (GEF) Sec12, followed by recruitment of Sec23/Sec24. Subsequently, the Sar1-Sec23/Sec24 complex recruits the Sec13/Sec31 complex, which leads to the bending of membranes and thereby the budding of COPII vesicles from ERES. COPII-mediated trafficking is an essential process in all eukaryotes, and its dysfunction causes a variety of human diseases, including chylomicron retention disease, cranio-lenticulo-sutural dysplasia and several neurological disorders^2,5,6.

Despite its broad physiological significance, important aspects of how COPII-mediated vesicle transport is regulated in response to environmental cues or pathological conditions remain obscure. As the efflux of cargo proteins and lipids from the ER depends on demands within the ER, formation of COPII-coated vesicles must be tightly coupled to changes in protein and lipid abundance and composition in the ER membranes. Notably, some lipids contribute to the formation of COPII vesicles by acting directly on membrane deformation, COPII assembly or ERES organization. For example, cone-shaped lipids such as phosphatidic acid (PA) and diacylglycerol (DAG) or inverted cone-shaped lipids such as lysophospholipids are predicted to promote vesicle formation by inducing membrane curvature due to their physical properties^7–10. Previous studies have shown that, in mammalian cells, activation of phospholipase D, which catalyzes the formation of PA, is required to support COPII coat assembly and ER export¹¹. Phospholipase A1 p125, which produces lysophospholipids, has also been shown to be involved in the architecture of ERES^12,13. In mammalian cells, PtdIns4P also acts to support COPII assembly, possibly through the function of PtdIns4P-binding effectors implicated in deforming the membrane¹⁴. In addition, our previous studies in yeast have found that lysophosphatidylinositol (lyso-PI) increases the recruitment of COPII machinery to membranes and rescues the sec12-4 mutant by enhancing COPII vesicle formation¹⁵. It is also known that loss of phospholipid head groups affects assembly of Sec23 at ERES¹⁶. Thus, curvature-inducing lipids could regulate COPII vesicle formation.

While post-transcriptional modification of COPII coat proteins is utilized as a major mode of regulation of COPII vesicle formation for the immediate response to acute stimuli^1,4,17–19, adaptation of COPII vesicle formation in response to chronic stress due to cargo overload or abnormal organelle homeostasis would be transcriptionally mediated. Thus, an intracellular sensing system is predicted to exist that monitors the membrane state in the ER, transduces signals to the nucleus and controls COPII vesicle formation at the transcriptional level. There is evidence that COPII coat proteins are transcriptionally upregulated in response to ER stress upon physiological nutrient fluctuations or developmental stage in animals^20–23, suggesting that unfolded protein response (UPR) sensors control COPII vesicle formation via transcriptional regulation of COPII coat proteins, thereby promoting the transport of refolded proteins. Such transcriptional regulation that transmits from the ER would be conserved for COPII vesicle formation in eukaryotes, because in yeast, the SEC genes required for COPII vesicle formation are also upregulated by the UPR²⁴. However, the mechanisms of transcriptional regulation and adaptation for vesicular trafficking coupled with lipid metabolism on the ER membranes are not known.

Here, we show that COPII-mediated vesicle transport is controlled through the Opi1-Ino2/Ino4-dependent transcription regulation, which is modulated by PA levels in the ER. Moreover, our genetic screening combined with gene expression profiling reveals that YIP3, a target gene of the transcription factors Ino2/Ino4, is required for the regulation of COPII-mediated trafficking. We also show that Yip3 does not modulate the activation of Sar1, whereas Ino2/Ino4 does, thus suggesting that Ino2/Ino4 regulates COPII-mediated transport at multiple steps. Instead, we show that Yip3 hinders Sec16 assembly on the ER membrane. Remarkably, during ER stress conditions, which are accompanied by elevated PA, Sec16 assembly at the ER membrane is restricted in a PA-dependent manner involving Yip3. Based on these findings, we propose a model in which stress-induced increases in PA sequester Opi1 at the ER membrane, relieving repression of Ino2/Ino4 target genes and promoting Yip3 expression, which in turn inhibits vesicular transport. Thus, this negative regulatory mechanism may act as a rheostat that adjusts COPII vesicle–mediated transport to changes in ER membrane lipid composition during ER stress.

Results

Defects in COPII vesicle formation in the sec12-4 mutant are rescued by deletion of SLC1, a gene encoding lysophospholipid acyltransferase

The budding yeast SLC1 gene encodes a lysophospholipid acyltransferase to convert lysophospholipids to phospholipids (Fig. 1a)^25–29. Because we showed previously that a sec12-4 conditional mutation is rescued by lysophospholipid accumulation¹⁵, we analyzed whether the phenotypes of the sec12-4 temperature-sensitive mutant are suppressed by deletion of SLC1. The sec12-4 mutant strain exhibits growth defects at temperatures above 32 °C (Fig. 1b). Deletion of SLC1 rescued the growth defects of sec12-4 mutant cells at restrictive temperatures. Furthermore, testing the maturation of cargo, CPY (carboxypeptidase Y) through the secretory pathway by western blot analysis, we observed that the accumulation of immature forms in the sec12-4 mutant was suppressed by SLC1 deletion (Fig. 1c, lower panel). For quantitative rather than qualitative evaluation, we also analyzed the maturation process of cargo proteins using pulse-chase labeling experiments, revealing a partial rescue of the CPY maturation defect in the sec12-4 mutant cells lacking the SLC1 gene (Fig. 1d). The p2 form of CPY was often invisible. Since it is possible that maturation of CPY may reflect ER-phagy rather than Golgi delivery, we analyzed the maturation of another cargo, Gas1, a glycosylphosphatidylinositol (GPI)-anchored protein. The immature forms of Gas1 accumulated in the sec12-4 mutant were also suppressed by SLC1 deletion (Fig. 1c, upper panel). As sec12-4 mutant cells exhibit disorganized and huge aggregated ERES^15,30, we next examined whether the absence of Slc1 affects the aggregation of ERES in the sec12-4 mutant. To visualize the ERES, we used the strains expressing SEC13-VENUS, a COPII outer coat protein tagged with yellow fluorescent protein (Venus)³¹. As previously observed³², the number of aggregated Sec13-Venus puncta per sec12-4 mutant cell was increased compared to wild-type (Fig. 1e), and deletion of SLC1 rescued the phenotype, suggesting that Slc1 negatively regulates COPII vesicle budding from the ER.

Fig. 1. Reduced levels of phosphatidic acid rescue sec12-4 mutant phenotypes.

a A scheme for the synthesis of phospholipids in yeast. Gro-3-P, glycerol 3-phosphate; Lyso-PA, lysophosphatidic acid; PA, phosphatidic acid; DAG, diacylglycerol; CDP-DAG, cytidine diphosphate diacylglycerol; PI, phosphatidylinositol; PS, phosphatidylserine; PE, phosphatidylethanolamine; PC, phosphatidylcholine; Lyso-PI, lysophosphatidylinositol; Lyso-PS, lysophosphatidylserine; Lyso-PE, lysophosphatidylethanolamine; Lyso-PC, lysophosphatidylcholine. b, f–h Five-fold serial dilutions of cells (b, f), cells transformed with pRS426 (empty) (g) or pRS426-PAH1 (pPAH1) (g, h) were spotted onto SD plates (f), SD plates without uracil (URA; g, h) or SD plates supplemented with uracil and 5-Fluoroorotic acid (5-FOA; h), and were incubated at the indicated temperatures for 5 days. 5-FOA was used to select yeast cells that lose the URA3 plasmid (pPAH1). Representative images; similar results were obtained in n = 3 independent biological replicates. c Cell extracts from the indicated cells grown at 32 °C for 60 min after preculture at 25 °C were analyzed by western blotting for CPY and Gas1. Left, representative western blot images. Indications for mature (m) forms of CPY and Gas1, ER (p1) and Golgi (p2) forms of CPY, and immature (im) Gas1 are shown. Right, quantification of the proportion of immature species (Gas1 and CPY p1/p2): Data represent the means ± sd from n = 3 independent biological replicates; circles indicate individual data points. Statistical significance was tested using the Tukey-Kramer multiple comparison test; exact P-values are reported on the figures. d Pulse-chase analysis of CPY was performed by labeling indicated cells with [³⁵S] methionine for 5 min and chasing at 32 °C for the indicated times. CPY was immunoprecipitated, followed by SDS-PAGE and analyzed using a phosphorimager. Representative autoradiographs from n = 2 independent biological replicates. The ER (p1), Golgi (p2) and mature (m) forms of CPY are indicated. The proportion of mature CPY from the representative images are shown. e Cells expressing Sec13-Venus were grown in SD medium at 32 °C for 60 min following 25 °C preculture and observed by fluorescence microscopy. Left, representative microscopy images. Right, analysis of the number of fluorescent dots larger than 0.3 µm² per cell. A total of N = 428, 360, 458, and 515 cells were analyzed for WT, sec12-4, slc1Δ, and sec12-4 slc1Δ, respectively (each biological replicate included >100 cells). Data represent the means ± sd from n = 3 independent biological replicates; circles indicate individual data points. Statistical significance was tested using the Tukey-Kramer multiple comparison test; exact P-values are reported on the figures. Scale bars, 5 μm. Source data are provided as a Source Data file.

This supports a proposed role of lysophospholipids in facilitating COPII vesicle formation¹⁵. Since Ale1/Slc4 and Cst26 are also lysophospholipid acyltransferases with different substrate specificity and localization^26,28,29,33, we next tested whether loss of function of their proteins could rescue the sec12-4 mutation. The ALE1/SLC4 disruption suppressed the temperature-sensitive growth defect and CPY maturation defect in the sec12-4 mutant, whereas the CST26 disruption did not (Supplementary Fig. 1a–c). These results suggest that Slc1 and Ale1/Slc4 play specific roles in COPII vesicle formation, and are consistent with the findings that Slc1 and Ale1/Slc4 localize to the ER, whereas Cst26 does not localize to the ER but to lipid droplets^26,29. Consistently, disruptions of other acyl transferases such as Tgl3, Tgl4 and Tgl5, which are known to localize to lipid droplets, did not suppress the growth defect of sec12-4 (Supplementary Fig. 1d).

Suppression of sec12-4 mutation by SLC1 deletion is related to reduced levels of PA in the ER membrane

We speculated that deletion of SLC1 affects the amounts of specific glycerophospholipids, which may participate in the suppression of sec12-4 phenotypes. Thus, we measured the glycerophospholipid (PC, PE, PI, PS) and lysophospholipid (lyso-PC, lyso-PE, lyso-PI, lyso-PS) composition of cells grown at 32 °C with electrospray ionization coupled to triple stage quadrupole (TSQ) Vantage mass spectrometry (Supplementary Data 2, Supplementary Fig. 2a, b). Neither sec12-4 or slc1Δ mutation changed glycerophospholipid profiles (Supplementary Figs. 2a, 3b). Interestingly, we detected significant decreases in lyso-PI and lyso-PS in slc1Δ or sec12-4 slc1Δ strain (Supplementary Fig. 2b), suggesting that the suppression of sec12-4 phenotypes by SLC1 deletion is not related to increased levels of lyso-PI, which facilitate COPII vesicle formation¹⁵.

The total amount of PA was not significantly affected by loss of Slc1 (Supplementary Fig. 3c). However, since Slc1 has the capacity to acylate lyso-PA to generate PA^25,27–29, we examined a possible correlation between PA metabolism and the suppression of sec12-4 mutation by SLC1 deletion. To test whether a decreased amount of PA could replicate the effect of SLC1 deletion on temperature-sensitive phenotypes in sec12-4 mutant, we generated a sec12-4 strain lacking DGK1, which encodes a diacylglycerol (DAG) kinase that phosphorylates DAG to generate PA^27,34. Compared with sec12-4, the sec12-4 dgk1Δ strain exhibited improved growth at restrictive temperatures (Fig. 1f). In addition, we overexpressed PAH1, which encodes an enzyme catalyzing the dephosphorylation of PA yielding DAG^27,35 in the sec12-4 mutant. Overexpression of PAH1 rescued sec12-4 (Fig. 1g). We also found that PAH1 deletion aggravates the growth phenotype of sec12-4 (Fig. 1h). These results suggest that reduced levels of PA may contribute to the suppression of sec12-4 mutation by SLC1 deletion.

To provide further support for the importance of PA levels, we next studied the localization of Opi1, a negative regulator of UAS_INO element/ICRE (inositol/choline-responsive element)-dependent transcription, because decreasing PA levels leads to the release of Opi1 from the ER and its entrance into the nucleus (Fig. 2a)^27,36–39. We took advantage of GFP-tagged Opi1 to address how much PA is present in the ER. Opi1-GFP was largely localized to the perinuclear ER region in the wild-type and sec12-4 mutant strains, whereas the majority of Opi1-GFP was present in the nucleus of slc1Δ and sec12-4 slc1Δ strains (Fig. 2b). This result suggests that the level of PA localized to the ER membrane is maintained relatively low in the absence of the SLC1 gene.

Fig. 2. Rescue of sec12-4 mutant by SLC1 deletion requires Opi1 transcriptional repressor.

a A scheme for the phosphatidic acid-dependent Ino2/Ino4-Opi1 regulatory pathway. PA, phosphatidic acid; UAS_INO, upstream activating sequence inositol-responsive element. b Cells expressing Opi1-GFP were grown at 32 °C in SD medium for 60 min (25 °C preculture) and observed by fluorescence microscopy. Top, representative microscopy images; scale bars, 5 μm. Bottom panels showing intensity plots along the yellow line in wild-type and slc1Δ cell images. Right, analysis of perinuclear ER/nucleoplasm intensity ratios from n = 3 independent biological replicates, with N = 100 cells counted per genotype per replicate. Each dot represents the signal ratio of a single cell and is colored according to the biological replicate. Colored dashed lines indicate the mean value of each replicate. Statistical significance was tested using the Tukey-Kramer multiple comparison test; exact P-values are reported on the figures. c Five-fold serial dilutions of the indicated cells were spotted onto SD plates and incubated at the indicated temperatures for 5 days. Representative images; similar results were obtained in n = 3 independent biological replicates. d Cell extracts from the indicated cells grown at 32 °C for 60 min (25 °C preculture) were analyzed by western blotting for CPY. Left, representative western blot images; p1, p2 and m represent the ER, Golgi, and mature/vacuole forms of CPY, respectively. Right, quantification of the proportion of immature CPY (p1/p2): Data represent the means ± sd from n = 3 independent biological replicates; circles indicate individual data points. Statistical significance was tested using the Tukey-Kramer multiple comparison test; exact P-values are reported on the figures. e Pulse-chase analysis of CPY was performed by labeling indicated cells with [³⁵S] methionine for 5 min and chasing at 32 °C for the indicated times. CPY was immunoprecipitated, followed by SDS-PAGE, and analyzed using a phosphorimager. Representative autoradiographs from n = 2 independent biological replicates. The ER (p1), Golgi (p2), and mature (m) forms of CPY are indicated. The proportion of mature CPY from the representative images are shown. Source data are provided as a Source Data file.

To confirm this, we isolated the ER-rich membrane fraction (P20) from cells (Supplementary Fig. 3d) and analyzed PA levels (Supplementary Fig. 3e). Consistent with total lipid analysis, ER-associated PA was elevated in sec12-4 compared to wild type. Moreover, consistent with changes in Opi1 localization, we found that the PA level in the ER-rich membrane fraction of the sec12-4 slc1 Δ mutant was significantly lower than that of the sec12-4 mutant. A similar result was observed in the sec12-4 dgk1Δ mutant.

The OPI1-INO2/INO4 pathway negatively regulates COPII-mediated vesicle transport

Once inside the nucleus, Opi1 binds to Ino2 and attenuates transcriptional activation by the Ino2-Ino4 complex^27,37,39,40. Therefore, the transcriptional repressor activity of Opi1 may contribute to the suppression of sec12-4 by SLC1 deletion. To test this hypothesis, we constructed double (sec12-4 opi1Δ) and triple mutant (sec12-4 slc1Δopi1Δ), which allowed us to study the effects of loss of OPI1 function on sec12-4 phenotypes such as temperature sensitivity and CPY maturation. OPI1 single null mutant can grow at temperatures above 32 °C (Fig. 2c) and exhibits normal cargo protein transport (Supplementary Fig. 4a, b), whereas the temperature sensitivity of sec12-4 was aggravated by OPI1 deletion, and the suppressive effect of SLC1 deletion on sec12-4 growth was also rescued by OPI1 deletion (Fig. 2c), suggesting that Opi1 is required for the suppression of sec12-4 by SLC1 deletion. Similar results were observed by analyzing the phenotypes in CPY maturation (Fig. 2d, e).

Because ALE1/SLC4 deletion rescued sec12-4 mutant phenotypes (Supplementary Fig. 1a, b), we wanted to determine whether loss of Opi1 function could cancel out the effects of ALE1/SLC4 deletion. As expected, the rescue in growth seen in the sec12-4 ale1Δ strain was diminished by additional deletion of OPI1 (Supplementary Fig. 1e).

We have previously shown that the sec12-4 mutation is rescued by loss of Scs2 and Scs22, the coreceptor for Opi1^27,37,39,41, and that the suppressive effect is phenocopied by mutations in Scs2 that are unable to associate with the Opi1 FFAT motif³⁰. Therefore, we tested the effect of OPI1 deletion on the rescue of growth defects in sec12-4 scs2Δ or sec12-4 scs2Δscs22Δ. OPI1 deletion also canceled the suppressive effect of scs2Δ/scs22Δ mutation (Supplementary Fig. 1e). This suggests that the negative regulation of Scs proteins in COPII-mediated vesicle transport is dependent on Opi1.

We next asked if the phenotypes of sec12-4 could be rescued when INO2 and INO4 genes were deleted. The temperature-sensitive growth defect of sec12-4 was rescued by INO2 or INO4 deletion (Fig. 3a). Western blot analysis revealed a partial restoration of CPY and Gas1 maturation (Fig. 3b). Compared to when SLC1 was deleted (Fig. 1c), the suppression by INO4 deletion was less effective. This may reflect a dual role for SLC1 deletion, one direct on membrane biophysical properties and one indirect via the Ino2/Ino4 pathway. The aggregation of ERES in the sec12-4 mutant was also rescued by INO4 deletion (Fig. 3c). Furthermore, with the same approaches, we examined the effect of INO4 deletion on temperature-sensitive phenotypes of sec23-1 and sec16-2 mutants. Both temperature-sensitive growth defects of sec23-1 and sec16-2 mutants were suppressed in the absence of INO4 (Fig. 3d). Cargo maturation defects in sec23-1 were also restored (Fig. 3e). Taken together, these results suggest that the OPI1-INO2/INO4 pathway negatively regulates COPII-mediated trafficking. Consistent with this, deletion of SLC1 or DGK1, which catalyze reactions that generate PA, rescued the temperature-sensitive growth defect of the sec16-2 mutant. (Supplementary Fig. 1f).

Fig. 3. Deletion of INO2 and INO4 rescues sec mutant phenotypes.

a, d Five-fold serial dilutions of the indicated cells were spotted onto SD plates and were incubated at the indicated temperatures for 5 days. Representative images; similar results were obtained in n = 3 independent biological replicates. b, e Cell extracts from the indicated cells grown at 32 °C for 60 min after preculture at 25 °C were analyzed by western blotting for CPY and Gas1. Representative western blot images, similar results were obtained in n = 3 independent biological replicates. Indications for mature (m) forms of CPY and Gas1, ER (p1) and Golgi (p2) forms of CPY, and immature (im) Gas1 are shown. Right, quantification of the proportion of immature species (Gas1 and CPY p1/p2): Data represent the means ± sd  from n = 3 independent biological replicates; circles indicate individual data points. Statistical significance was tested using the Tukey-Kramer multiple comparison test; exact P-values are reported on the figures. c Cells expressing Sec13-Venus were grown in SD medium at 32 °C for 60 min following a 25 °C preculture and observed by fluorescence microscopy. Left, representative microscopy images. Right, analysis of the number of fluorescent dots larger than 0.3 µm² per cell. A total of N = 438, 464, 443, and 369 cells were analyzed for WT, sec12-4, ino4Δ, and sec12-4 ino4Δ, respectively (each biological replicate included >100 cells). Data represent the means ± sd from n = 3 independent biological replicates; circles indicate individual data points. Statistical significance was tested using the Tukey-Kramer multiple comparison test; exact P-values are reported on the figures. Scale bars, 5 μm. Source data are provided as a Source Data file.

Screening for candidate genes involved in regulating COPII-mediated vesicle transport

By comparing the transcriptomic profiles of sec12-4 ino2Δ ino4Δ and sec12-4 mutants cultured at 25 °C—a temperature at which vesicular transport defects are present without inducing strong lethality—we aimed to identify gene expression changes specifically associated with suppression of the sec12-4 phenotype (Supplementary Data 3). Expression of 152 known genes was significantly downregulated by more than twofold upon deletion of INO2/INO4 and the functional categorization showed that different functional categories of genes were regulated by the OPI1-INO2/INO4 pathway (Fig. 4a). Consistent with previous reports^42,43, the expression of OPI3 (relative amounts are 0.52 ± 0.01, p < 0.001, mean ± sd n = 3), CHO1 (0.56 ± 0.13, p < 0.05) and CHO2 (0.67 ± 0.09, p < 0.05) that contain the UAS_INO element in their promoters was significantly down-regulated compared to the expression level of control strain, although the decreases in expression were not more than two-fold. In addition, we found that there are no significant changes in expression of genes involved in phospholipid metabolism such as SLC1, ALE1, DGK1, PAH1, PLB1, PLB2 and PLB3, except NTE1, and genes functioning upstream of INO2/INO4 such as SCS2, SCS22 and OPI1 (Supplementary Fig. 5a). These results suggest that suppression of sec12-4 mutation by INO2/INO4 deletion is not due to changes in their gene expression. We also did not observe significant 1.5-fold or greater increases in expression of SEC genes involved in COPII vesicle formation (Supplementary Fig. 5b), implying that other genes are involved in the OPI1-INO2/INO4 pathway-dependent regulation of COPII vesicle formation.

Fig. 4. Deletion of YIP3, a target gene of Ino2/Ino4, suppresses sec12-4 mutant phenotypes.

a Functional groups among 152 genes whose expression decreased more than twofold in sec12-4 ino2Δ ino4Δ cells compared to sec12-4 cells cultured at 25 °C. 7 genes that cause a growth defect when overexpressed in wild-type cells (Supplementary Fig. 6, 7) are shown in red. b Tetrad analysis reveals that deletion of TRF5 or YIP3 suppresses the slow-growth defect of the sec12-4 mutant. Diploids heterozygous for both sec12-4 and disruption of seven candidate genes for negative regulators of COPII vesicle formation shown in red (a) were sporulated, and tetrads were dissected and incubated on YPD plates for 5 days. Blue circles, squares, and red circles represent sec12-4, XXXΔ, and sec12-4 XXXΔ haploid cells, respectively. Colony size was quantified using ImageJ by drawing circles around the perimeter of each colony in the acquired images and measuring the diameter. Bottom graphs show the mean difference in colony diameter (%) ± sd  for the indicated genotypes from n = 3 independent biological replicates; circles indicate individual data points. For each biological replicate, one sec12-4 and one sec12-4 trf5Δ or sec12-4 yip3Δ colony derived from the same tetrad were analyzed. Statistical significance was assessed by a two-tailed unpaired t-test, and exact P-values are reported on the figures. c Five-fold serial dilutions of cells were spotted onto SD plates and incubated at the indicated temperatures for 5 days. Representative images; similar results were obtained in n = 3 independent biological replicates. d, f Cell extracts from the indicated cells (d) or from cells carrying either an empty vector or a GPD promoter-driven 2 µ YIP3 plasmid (YIP3 OE, f) were analyzed by western blotting for CPY and Gas1. Cells were grown in appropriately supplemented SD medium at 32 °C for 60 min after preculture at 25 °C. Left, representative western blot images. Indications for mature (m) forms of CPY and Gas1, ER (p1) and Golgi (p2) forms of CPY, and immature (im) Gas1 are shown. Right, quantification of the proportion of immature species (Gas1 and CPY p1/p2): Data represent the means ± sd from n = 3 independent biological replicates; circles indicate individual data points. Statistical significance was tested using the Tukey-Kramer multiple comparison test (d) and two-tailed unpaired t-test (f); exact P-values are reported on the figures. e Pulse-chase analysis of CPY was performed by labeling indicated cells with [³⁵S] methionine for 5 min and chasing at 32 °C for the indicated times. CPY was immunoprecipitated, followed by SDS-PAGE, and analyzed using a phosphorimager. Representative autoradiographs from n = 2 independent biological replicates. The ER (p1) and mature (m) forms of CPY are indicated. The proportion of mature CPY from the representative images are shown. g Five-fold serial dilutions of cells carrying either an empty vector or a GPD promoter–driven 2 µ YIP3 plasmid (YIP3 OE) were spotted onto SD plates and incubated at the indicated temperatures for 5 days. Representative images; similar results were obtained in n = 3 independent biological replicates. h Cells expressing endogenously tagged Sec13-GFP and carrying either an empty vector or a GPD promoter-driven 2 µ YIP3 plasmid (YIP3 OE) were grown in SD medium at 25 °C and imaged by fluorescence microscopy. Top, representative microscopy images. Bottom, analysis of the number of fluorescent dots larger than 0.1 µm² per cell. A total of N = 341 and 394 cells were analyzed for WT + EV and WT + YIP3 OE, respectively (each biological replicate included >100 cells). Data represent the means ± sd from n = 3 independent biological replicates; circles indicate individual data points. Statistical significance was tested by a two-tailed unpaired t-test; exact P-values are reported on the figures. Scale bars, 5 μm. Source data are provided as a Source Data file.

Thus, we screened for the 152 downregulated genes to identify the genes that cause growth defects when overexpressed (Supplementary Fig. 6), because the genes of interest negatively regulate the formation of COPII vesicles. For this purpose, we used a gene expression method, named genetic tug-of-war (gTOW), that causes the overexpression of the target gene due to an increase in the copy number, which is inversely correlated with the leucine concentration in the growth medium⁴⁴. We found that overexpression of 7 genes, including BAP2, GGC1, FET3, TRF5, MSG5, YIP3, and YOR338W, led to growth defects in wild-type cells of the BY background strain (Fig. 4a, shown in red and Supplementary Figs. 6, 7). To validate and refine the candidate genes obtained by our primary screen, we next investigated whether disruption of those genes could rescue the growth defect of sec12-4 mutant, because sec12-4 cells grow slowly in comparison to wild-type cells at permissive temperature (25 °C). We generated heterozygous diploid double mutant BY background strains, sporulated and subjected them to tetrad dissection. After spores were allowed to form colonies for several days, we determined the colony size and genotypes. The colonies of sec12-4 trf5Δ and sec12-4 yip3Δ double mutants were larger than that of the single sec12-4 mutant (Fig. 4b), indicating that deletion of TRF5 or YIP3 rescues the slow growth of the sec12-4 mutant. Furthermore, we observed that deletion of those genes partially suppressed the temperature-sensitive growth defects of the sec12-4 mutant (Supplementary Fig. 8). Thus, we identified two potential genes that are responsible for the regulation of COPII-mediated vesicle transport via the OPI1-INO2/INO4 pathway.

YIP3 upregulated by Ino2/Ino4 negatively regulates COPII-mediated vesicle transport

Trf5 is a noncanonical poly(A) polymerase and assembles into the heteromeric Trf5-Air1-Mtr4 (TRAMP) complex that promotes the degradation of RNA substrates by interacting with the nuclear exosome^45–47. Because TRF5 deletion increases RNA abundance for RNAs showing Trf5 binding⁴⁸, the most straightforward explanation, which fits with our findings, could lie in the role of the TRAMP complex in RNA levels. Hence, increased expression levels of proteins involved in the formation of COPII vesicles might be a reason for the suppression of the sec12-4 mutation by deletion of TRF5.

On the other hand, YIP3 encodes a protein known to localize to COPII vesicles⁴⁹ and has been proposed to be involved in vesicular transport between the ER and Golgi apparatus because it physically interacts with yeast Rab proteins^50,51, but its function is unknown. No changes in the localizations of yeast Rab proteins and in the maturation of CPY were observed in a yip3Δ mutant⁵¹, whereas overproduction of Yip3 caused a severe growth defect (Supplementary Figs. 6, 7) and aberrant ER membrane proliferation⁵¹, suggesting the role of Yip3 as a negative regulator of ER homeostasis. As ER membranes have been shown to proliferate in mutants defective in COPII vesicle formation⁵², we therefore examined whether Yip3 negatively regulates COPII-mediated vesicle transport. First, we created a sec12-4 yip3Δ double mutant with a different genetic background than BY and confirmed that the suppressive effect of YIP3 deletion on the growth defect of sec12-4 is not background specific (Fig. 4c). Next, we tested whether immature cargos accumulated in the sec12-4 mutant could be restored by YIP3 deletion. Western blot analysis revealed the restoration of CPY and Gas1 maturation by YIP3 deletion (Fig. 4d). Furthermore, pulse-chase labeling experiments showed that the CPY maturation defect in the sec12-4 mutant is partially restored in the absence of Yip3 (Fig. 4e). These results suggest that Yip3 negatively regulates COPII-mediated vesicle transport. This is supported by results showing the delayed maturation of cargo proteins caused by YIP3 overexpression (Fig. 4f). In addition, YIP3 overexpression greatly exacerbated the temperature sensitivity of sec16-2 and sec12-4 mutants (Fig. 4g) and increased the number of aggregated ERES labeled with Sec13-GFP (Fig. 4h), which are consistent with the negative effect of Yip3. Moreover, YIP3 overexpression reversed the rescue of growth in sec12-4 dgk1Δ mutant (Supplementary Fig. 1g). Thus, when the results are taken together, we conclude that the OPI1-INO2/INO4 pathway in response to changes in PA metabolism regulates anterograde vesicular transport from the ER via YIP3 transcription.

The PA-Opi1-Ino2/Ino4 system plays a critical role in controlling COPII vesicle formation by tuning multiple steps

What roles can Yip3 play in COPII-mediated trafficking? Because phenotypes of sec12-4 mutant, which is defective in GDP-to-GTP exchange on Sar1, have been reportedly rescued by Sar1 overexpression^53–55, it is possible that the suppression by YIP3 deletion is due to an increased level of Sar1 protein, although no effect of INO2/INO4 disruption on mRNA levels of SAR1 gene was observed (Supplementary Fig. 5b). Therefore, we first studied the protein levels of Sar1. Western blot analysis showed that the steady-state levels of Sar1 were unaffected by YIP3 deletion and overexpression, as well as by INO4 deletion (Supplementary Fig. 9a–c).

We next tried to determine if Yip3 functions to inhibit the GDP-to-GTP exchange on Sar1. Sar1 is activated by Sec12, which catalyzes the exchange of GDP for GTP and thereby promotes Sar1 association with the ER membrane^53–55. We therefore assessed ER membrane association of Sar1–AcGFP, a C-terminally AcGFP-tagged wild-type Sar1 that cycles between GDP- and GTP-bound states in vivo⁵⁶. In wild-type cells, ER membrane/cytosol signal ratios (an index of ER membrane association) of Sar1-AcGFP were between 1.3 and 1.5 (Fig. 5a). On the other hand, as expected, the ratios of Sar1D32G-AcGFP, a GDP- locked form of Sar1, were significantly reduced to about 1.1. A reduction in the ER membrane association of Sar1-AcGFP was also observed in the sec12-4 mutant strain. These results are consistent with the model that activated Sar1-GTP associates with the ER membrane. We analyzed the ER membrane association of Sar1-AcGFP in the sec12-4 yip3Δ mutant and showed that YIP3 deletion fails to rescue the ER membrane association defect in the sec12-4 mutant (Fig. 5b), suggesting that Yip3 is not involved in regulating Sar1 activity. Interestingly, we found that the ER membrane association defect in the sec12-4 mutant was partially rescued by the loss of Ino2 or Ino4 (Fig. 5c). This result suggests that there is an unknown factor in the Ino2/Ino4 target genes that functions to negatively regulate the ER membrane association of Sar1. Taken together, these results support a critical role of the PA-Opi1-Ino2/Ino4 system in regulating COPII vesicle formation by tuning multiple steps.

Fig. 5. Deletion of INO2 or INO4 but not YIP3 rescues Sar1 membrane association defect in sec12-4 mutant.

a–c Cells expressing Sar1-AcGFP or Sar1D32G-AcGFP and Kar2-SS-mRFP-HDEL were grown at 25 °C in SD medium, shifted to 32 °C for 60 min and observed by fluorescence microscopy. The left panel of a shows representative microscopy images. Bottom panels are intensity plots along the yellow lines in representative images. Scale bars, 5 μm. Superplots show the AcGFP ER membrane/cytosol intensity ratios from n = 3 independent biological replicates, with N = 16 cells counted per genotype per replicate in (a) and N = 50 cells counted per genotype per replicate in (b, c). Each dot represents the signal ratio of a single cell and is colored according to the biological replicate. Colored dashed lines indicate the mean value of each replicate. Statistical significance was tested using the Tukey-Kramer multiple comparison test; exact P-values are reported on the figures. Source data are provided as a Source Data file.

Yip3 is involved in the regulation of Sec16 assembly on the ER membrane

Yip3 was not involved in the regulation of Sar1 activity. Therefore, we considered that Yip3 may function in the later stages of the Sar1 activity. Sar1 on the ER membrane recruits the inner coat Sec23/Sec24 complex from the cytosol, which then recruits the outer coat Sec13/Sec31 complex to deform the membrane and form COPII vesicles^1–4. As Sec16 interacts with all of the COPII coat subunits, it is thought to act as a scaffold for COPII assembly to form ERES⁵⁷. Sec16 is also proposed to act as a Sec23 GAP inhibitor to contribute to stable coat assembly, and a recent study has shown that Sec16 localizes to the ERES in a manner that is dependent on Sed4⁵⁸. To characterize whether Yip3 is associated with Sec16 function, we assessed the effect of Yip3 deletion on the assembly of Sec16 to the ERES⁵⁸. For this purpose, we generated yeast strains that express Sec16-AcGFP from a CEN plasmid under the control of the endogenous promoter or Sec16-GFP from its endogenous genomic locus and conducted fluorescence microscopy. Interestingly, we found that in wild-type cells, Sec16-AcGFP displayed low signal in the cytosol, whereas it was detected in the ERES but also showed a diffuse cytoplasmic signal in sec12-4 mutant cells (Fig. 6a). We reasoned that if Yip3 hinders Sec12-dependent assembly of Sec16 into the ERES, deletion of YIP3 would rescue the Sec16 localization defect in the sec12-4 mutant. Indeed, we observed that the Sec16 localization defect was fully restored by YIP3 deletion as evidenced by increased ERES/cytosol signal ratio of Sec16-AcGFP, suggesting that Yip3 has the ability to inhibit the assembly of Sec16 into the ERES.

Fig. 6. Yip3 inhibits Sec16 assembly into the ERES.

a Cells expressing Sec16-AcGFP were grown at 25 °C in SD medium and were either analyzed directly or shifted to 32 °C for 60 min and observed by fluorescence microscopy. Left, representative microscopy images. Scale bars, 5 μm. Right, Sec16-AcGFP fluorescence intensities in ERES (puncta, A) and in cytosol (B) were measured⁵⁸, background-subtracted, and quantified as the ERES/cytosol intensity ratio (A/B). Bottom superplots show the Sec16-AcGFP in ERES/cytosol intensity ratios from n = 3 independent biological replicates, with N = 50 cells counted per genotype per replicate. Each dot represents the signal ratio of a single cell and is colored according to the biological replicate. Colored dashed lines indicate the mean value of each replicate. Statistical significance was tested using the Tukey-Kramer multiple comparison test; exact P-values are reported on the figures. b, d, e Subcellular fractionation followed by Western blotting reveals that Sec16-GFP membrane association is modulated by Yip3 and Slc1. Total lysates, ER-enriched membrane fractions (P20), and cytosolic fractions (S20) were prepared from the indicated cells expressing endogenously tagged Sec16-GFP after preculture at 25 °C and incubation at 32 °C for 0 min (d), 60 min (b), or 90 min (e) in SD medium. All samples were treated with NTCB prior to Western blot analysis using an anti-GFP antibody. Cells used in (d) and (e – right panel) carried either an empty vector or a plasmid expressing YIP3 under the control of the GPD promoter (YIP3 OE). Data represent the means ± sd from n = 5 (b) and n = 3 (d, e) independent biological replicates; circles indicate individual data points. Statistical significance was tested using the Tukey-Kramer multiple comparison test (b, e – left panel) and two-tailed unpaired t-test (d, e – right panel); exact P-values are reported on the figures. c Fluorescence microscopy of endogenously Sec16-GFP tagged wild-type cells carrying either an empty vector or a plasmid expressing YIP3 under the control of the GPD promoter (YIP3 OE) cultured at 25 °C in SD medium. The ratio of intensities of Sec16-GFP in ERES and cytosol was determined as described in (a). Top, representative microscopy images. Scale bars, 5 μm. Bottom superplot shows the Sec16-GFP in ERES/cytosol intensity ratios from n = 3 independent biological replicates, with N = 100 cells counted per genotype per replicate. Each dot represents the signal ratio of a single cell and is colored according to the biological replicate. Colored dashed lines indicate the mean value of each replicate. Statistical significance was tested by a two-tailed unpaired t-test; exact P-values are reported on the figures. Source data are provided as a Source Data file.

To confirm our in vivo results biochemically, we fractionated lysates from Sec16-GFP expressing cells into ER-rich membrane and supernatant fraction, and analyzed the distribution by Western blot (Fig. 6b). Given the very large size of Sec16-GFP (~270 kDa), analysis of native Sec16-GFP using SDS-PAGE migration shift is difficult. Therefore, we employed a reagent NTCB (2-nitro-5-thiocyanatobenzoic acid), which selectively cleaves cysteine residues under alkaline conditions⁵⁹. Without NTCB treatment, Sec16-GFP could not be detected in total yeast extracts from wild-type cells, whereas NTCB treatment yielded a highly reproducible fragment (Supplementary Fig. 9d–f) because there are two cysteine residues in the N-terminal upstream region of GFP. The fact that Sec16-GFP was not detected without NTCB suggests that Sec16-GFP is not degraded within the cell, and we did not observe degradation products in any of the strains we analyzed (Supplementary Fig. 9d–f). This biochemical analysis using NTCB revealed that in the wild-type cells, approximately twice the amount of Sec16-GFP was present in the ER-rich membrane fraction, whereas in the sec12-4 mutant cells, nearly equal amounts were in both the supernatant and the membrane fraction (Fig. 6b). In addition, fractionation analysis showed increased ER membrane localization of Sec16–GFP in sec12-4 yip3Δ cells compared with sec12-4 cells, confirming our in vivo results and strongly suggesting that Yip3 represses the assembly of Sec16-GFP onto membranes.

To verify this, we next assessed the impact of YIP3 overexpression on Sec16 assembly in the wild-type cells. We observed that overexpressing YIP3 hinders Sec16 assembly into the ERES as detected by microscopy (Fig. 6c) and by fractionation analysis (Fig. 6d). Since we have already shown that the sec12-4 slc1Δ strain exhibits reduced PA level in the ER (Supplementary Fig. 3e), relocalization of Opi1 into the nuclear lumen (Fig. 2b) and a recovery in growth and vesicular transport compared to sec12-4 (Fig. 1b–d), we biochemically assessed the effect of SLC1 deletion on Sec16 assembly into the ERES. Our results showed that deletion of SLC1 in sec12-4 significantly recovered the aberrant high cytosolic Sec16-GFP phenotype of the sec12-4 strain (Fig. 6e, left panel). In addition, YIP3 overexpression reversed the recovery of Sec16 assembly in sec12-4 slc1Δ mutant (Fig. 6e, right panel). Collectively, our findings suggest that Yip3 serves as a negative regulator of COPII vesicle formation that inhibits Sec16 assembly into the ERES depending on PA level.

The PA-Opi1-Ino2/Ino4 system and Yip3 function are involved in the delay of protein transport during ER stress

A functional interaction exists between two mechanisms: Vesicular transport of proteins from the ER and the UPR^{22–24,60,61}. In yeast, ER stress inducers such as the reducing agent dithiothreitol (DTT) and the N-glycosylation inhibitor tunicamycin, which activate the UPR, delay the transport of cargo proteins^62–64. Therefore, to test whether the transport delay caused by ER stress inducers is modulated by PA-dependent Yip3, we first examined the effect of DTT on Yip3 protein levels, because we confirmed that SLC1 deletion, which reduces PA levels in the ER (Supplementary Fig. 3e), decreases Yip3 protein levels (Fig. 7a). The results showed that DTT increases Yip3 levels during the treatment period and at the same time inhibits cargo protein maturation (Fig. 7b). Additionally, an increase in PA levels by DTT treatment was also observed (Fig. 7c). We also found that the inhibition of cargo protein maturation by DTT treatment is Slc1-dependent, and the suppressive effect of SLC1 deletion is canceled by loss of Opi1 function (Fig. 7d). Furthermore, the inhibition of cargo protein maturation was partially rescued by YIP3 deletion (Fig. 7e), suggesting the involvement of Yip3 in DTT-induced protein transport delay. Consistent with this, DTT treatment reduced the amount of Sec16 bound to the ER membrane (Fig. 7f, g). Elevated PA (Fig. 7h) and Yip3 levels (Fig. 7i) and decreased membrane binding of Sec16 (Fig. 7j, k) were also observed under tunicamycin treatment. These results suggest that the inhibitory effect of Yip3 on Sec16 membrane localization due to increased PA levels is at least partially involved in the delay in protein transport caused by ER stress.

Fig. 7. The PA-Opi1-Ino2/Ino4 system and Yip3 contribute to the delay of protein transport during ER stress.

a, i Cell extracts from cells expressing endogenously HA-tagged Yip3 grown at 25 °C in SD medium (a, i; -TM) or SD medium containing 2 µg/ml of tunicamycin for 4 h (i; +TM) were analyzed by Western blotting using anti-HA and anti-Wbp1 (a) or anti-PGK1 (i) antibodies. Representative western blot images are shown. Data represent the mean Yip3-HA/Wbp1 (a) and Yip3-HA/PGK1 (i) ratios ± sd from n = 3 independent biological replicates; circles indicate individual data points. Statistical significance was tested by a two-tailed unpaired t-test; exact P-values are reported on the figures. b Cell extracts from wild-type cells expressing endogenously HA-tagged Yip3, grown at 25 °C in SD medium, and treated with 5 mM dithiothreitol (DTT) for the indicated times were analyzed by Western blotting for Gas1, CPY, Yip3-HA, and Wbp1. Left, representative western blot images. Mature (m) forms of CPY and Gas1 are indicated; p1 and p2 represent the ER and Golgi forms of CPY, respectively, and immature (im) indicates the ER form of Gas1. Right, quantification of the proportion of immature species (Gas1 and CPY p1/p2) and Yip3-HA/Wbp1 ratio. Data represent the means ± sd from n = 3 independent biological replicates; circles indicate individual data points. Statistical significance was tested using the Tukey-Kramer multiple comparison test; exact P-values are reported on the figures. c, h Wild-type cells were treated with or without 5 mM DTT (c) or 2 µg/ml tunicamycin (TM, h) for 4 h at 25 °C in SD medium. Phospholipids were extracted and analyzed by double one-dimensional TLC. Data represent the means ± sd from n = 3 independent biological replicates; circles indicate individual data points. Statistical significance was assessed by a two-tailed unpaired t-test; exact P-values are reported on the figures. PI phosphatidylinositol; PS phosphatidylserine; PA phosphatidic acid. d, e Cell extracts from the indicated cells, grown in SD medium at 25 °C with or without 5 mM DTT for 4 h (d) or for the indicated durations (e), were analyzed by Western blotting for Gas1 and CPY. Left, representative western blot images. Indications for mature (m) forms of CPY and Gas1, ER (p1) and Golgi (p2) forms of CPY, and immature (im) Gas1 are shown. Right, quantification of the proportion of immature species (Gas1 and CPY p1/p2). Data represent the means ± sd from n = 3 independent biological replicates; circles indicate individual data points. Statistical significance was tested using the Tukey-Kramer multiple comparison test; exact P-values are reported on the figures. f, j Fluorescence microscopy of cells endogenously expressing GFP-tagged Sec16 cultured at 25 °C in SD medium with or without 5 mM DTT (f) or 2 µg/ml of tunicamycin (j) for 4 h. The ratio of intensities of Sec16-GFP in ERES and cytosol was determined as described in Fig. 6a. Left, representative microscopy images. Scale bars, 5 μm. Right, superplot shows the Sec16-GFP in ERES/cytosol intensity ratios from n = 3 independent biological replicates, with N = 30 cells counted per genotype per replicate. Each dot represents the signal ratio of a single cell and is colored according to the biological replicate. Colored dashed lines indicate the mean value of each replicate. Statistical significance was tested by a two-tailed unpaired t-test; exact P-values are reported on the figures. g, k Cells expressing endogenously tagged Sec16-GFP were cultured at 25 °C in SD medium, treated with or without 5 mM DTT (g) or 2 µg/ml of tunicamycin (k) for 4 h and analyzed by subcellular fractionation followed by Western blotting. Total lysates, ER-enriched membrane fractions (P20), and cytosolic fractions (S20) were prepared and treated with NTCB prior to Western blot analysis using an anti-GFP antibody. Top, representative western blot images. Bottom, data represent the means ± sd from n = 3 independent biological replicates; circles indicate individual data points. Statistical significance was tested by a two-tailed unpaired t-test; exact P-values are reported on the figures. Source data are provided as a Source Data file.

Discussion

How the COPII vesicle-dependent transport system adapts to changes in physiological cues and stress is not well understood. Post-transcriptional modifications have emerged as major modes of regulation of the COPII pathway^1,4,17–19. Nevertheless, extensive regulation occurring at the transcriptional level must also be important for environmental adaptation. In this study, we provide evidence that PA sensing in the ER transcriptionally regulates COPII-mediated vesicle transport. Our data showed that decreased PA levels in the ER membranes suppresses the secretory defect of the sec12-4 mutant. This effect is specific to the sec12-4 mutant, as lowering PA through slc1Δ or dgk1Δ single deletion does not alter ERES morphology or trafficking behavior. The suppression is caused by inhibiting the Ino2/Ino4 transcriptional activator complex through its interaction with Opi1, which is sequestered in the ER by PA and when released from the ER, is translocated into the nucleus to repress Ino2/Ino4 target genes. Furthermore, we found that YIP3, an Ino2/Ino4 target gene, negatively regulates COPII-mediated vesicle transport. This readily explains the negative effects of Yip3 overexpression on yeast cell growth and ER morphology⁵¹. Consistent with the role of Yip3 as a negative regulator, we showed that Yip3 inhibits Sec16 assembly into the ERES. Thus, our findings suggest that Yip3 negatively regulates COPII vesicle formation via transcriptional expression by the PA-Opi1-Ino2/Ino4 pathway (Fig. 8), and propose that this pathway may serve as a rheostat linking the ER membrane status to COPII vesicle formation.

Fig. 8. A model of ER sensing of phosphatidic acid (PA) metabolism in regulating COPII vesicle formation.

Yip3 and Xxx (unknown protein), encoded by YIP3 and XXX, target genes of the Ino2/Ino4, act as negative regulators of COPII vesicle formation. When PA levels in the ER membranes decrease, Opi1 is translocated to the nucleus, binds to Ino2 and attenuates transcriptional activation by the Ino2/Ino4 complex. This leads to the repression of Yip3 and Xxx expression and hence the release of negative regulation of COPII vesicle formation. The fact that deletion of INO2 or INO4 but not YIP3 rescues the Sar1 membrane association defect in sec12-4 mutant reveals that Yip3 and Xxx proteins regulate different steps in COPII vesicle formation. Yip3 hinders Sec16 assembly into the ERES.

Loss of SLC1 did not affect the total amount of phospholipid species. Given the roles of Slc1 in the fatty acid remodeling of phospholipids^25–29, we asked how the lack of SLC1 influences fatty acyl chain composition of phospholipids. As previously reported²⁶, relative amounts of phospholipid species such as 26:0, 26:1, 26:2, 28:0, 28:1 and 28:2 composed of short fatty acids decreased (Supplementary Data 4, Supplementary Fig. 10). This suggests that Slc1 preferentially utilizes short fatty acids as substrates to remodel phospholipids. Remarkably, slc1Δ strain showed decreased levels of lyso-PI and lyso-PS. Decreases in lyso-PI 16:0 and lyso-PI 18:1 species, the majority of lyso-PI, have also been observed in slc1Δ cells²⁶. Because Slc1 has preferential substrate specificity for PI and PS, like PA²⁵, the decreases in lyso-PI and lyso-PS in slc1Δ cells may be related to the substrate specificity of Slc1. Although the local abundance of lysophospholipids in the ER membrane is unknown, the finding that SLC1 deletion does not increase the total amounts of lysophospholipids such as lyso-PI, -PS, -PE and -PC suggests that suppression of the sec12-4 mutant is not due to their increased lysophospholipids, which facilitate COPII vesicle formation as proposed previously¹⁵. Like lysophospholipids, PA has a unique physical property that generates membrane curvature^7,9,10, thereby facilitating the budding of vesicles. To determine the local abundance of PA in the ER, we analyzed the binding of Opi1-GFP to the ER membrane by fluorescence microscopy and found that the amount of Opi1-GFP bound to the membrane decreased in the absence of Slc1. This was consistent with data obtained from biochemical analysis of PA levels in the ER-rich membrane fraction and suggests that deletion of SLC1 reduces local levels of PA in the ER, giving rise to the possibility that a local increase in lyso-PA in the ER may be responsible for the suppression of the sec12-4 mutation. However, the suppression by SLC1 deletion was canceled by loss of Opi1, implying that the negative effect of Slc1 on COPII-mediated vesicle transport is dependent on Opi1 rather than a local increase in lyso-PA level.

Decreased PA dissociates Opi1 from the ER and thereby represses expression of Ino2/Ino4 target genes through binding of Opi1 to the Ino2^27,37,39. Consistently, we showed that deletion of INO2/INO4 also rescued the sec12-4 mutation, suggesting that COPII-mediated vesicle transport is negatively regulated by genes that are transcriptionally upregulated by Ino2/Ino4. The PA-Opi1-Ino2/Ino4 system maintains ER homeostasis during ER stress by promoting ER expansion^65,66. ER stress also upregulates both genes involved in protein folding and degradation and genes required for COPII vesicle formation²⁴. This suggests that UPR acts to reduce the level of misfolded protein not only by promoting folding to the natural state but also by increasing the rate of protein degradation and facilitating the exit of proteins from the ER via the secretory pathway once refolding is complete. The latter is consistent with the evidence that mutations affecting COPII vesicle formation have negative genetic interactions with mutations conferring defects in UPR⁶⁷ and that the constitutive activation of the UPR suppressed sec12-4 mutant phenotypes⁶⁰. In mammals, it has also been shown that the IRE1 and XBP1, the key regulators of the UPR, are functionally connected to COPII-mediated trafficking, and transcriptional regulation of COPII components by XBP1 contributes to the promotion of COPII-dependent trafficking²³. However, our DNA microarray analysis in the sec12-4 mutant background showed that disruption of INO2/INO4 hardly changed the expression of UPR target genes encoding a chaperone (SCJ1; 1.09 ± 0.07 fold, p = 0.16) and required for ERAD (DER1; 1.15 ± 0.06 fold, p = 0.05). Moreover, we observed that SEC genes required for COPII vesicle formation are not significantly upregulated in the absence of Ino2/Ino4. These results suggest that the suppression of the sec12-4 mutation by Ino2/Ino4 deletion is independent of the activity of the UPR signaling pathway and is not due to increased expression of COPII components.

By genetic screening combined with gene expression profiling, we identified YIP3 as a negative regulator of COPII-dependent trafficking. YIP3 was downregulated more than two-fold in the absence of Ino2/Ino4. This is in agreement with a report showing that YIP3 is an Opi1-regulated gene⁴³. Previous studies showed that the YIP3‑encoded protein Yip3 localizes to COPII vesicle fractions, the ER, and the Golgi, and proposed a role in trafficking through the early secretory pathway^{49–51,68,69}. Our findings confirm and refine this view. We observed that deletion of YIP3 suppresses the growth and trafficking defects of the sec12‑4 mutant, indicating that Yip3 acts as a negative regulator of COPII‑mediated vesicle transport and that transcriptional repression of YIP3 is the main mechanism by which loss of Ino2/Ino4 alleviates sec12‑4 phenotypes. Moreover, our colocalization analysis of Yip3-GFP with markers of the secretory pathway, quantified using Pearson and Mander’s coefficients (Supplementary Fig. 11a, b), showed that colocalization of Yip3 with ERES marker Sec16 and Sec13, as well as trans-Golgi Sec7, is low, while co-localization with ERGIC/cis-Golgi Sed5 is high. These results indicate that Yip3 may cycle between ERES and ERGIC. Importantly, unlike Yip3 deletion, GFP tagging of Yip3 did not alter the sec12‑4 phenotype (Supplementary Fig. 1h), confirming that the observed localization reflects the behavior of functional Yip3.

What, then, is the function of Yip3? Yip3 is the only yeast protein belonging to the PRA family, which is named for its role as a ‘Prenylated Rab acceptor’, and is highly conserved from yeast to humans⁶⁹. Human PRA1 was proposed to function as a GDI-displacement factor that dissociates Rab GTPases from Rab-GDP-dissociation inhibitor GDI⁷⁰. Like PRA1, Yip3 interacts with a member of the yeast Rab family Ypt1^50,51, which is involved in ER-to-Golgi trafficking. However, distribution of Ypt1 between the cytosol and membranes was not affected in Yip3-deleting or -overproducing cells⁵¹, suggesting that the negative regulation of Yip3 in ER-to-Golgi trafficking found in this study may not be linked to a mechanism that requires Ypt1. This is consistent with evidence that Ypt1 is not required for COPII vesicle formation^71,72. If, like Yip3, the other PRA family negatively regulates COPII vesicle-mediated transport from the ER, its expression would result in ER accumulation of immature cargo proteins. Indeed, immature glycosylation patterns of the glutamate transporter have been observed in mammalian cells expressing PRA family member, PRA2/GTRAP3-18⁷³. Similar immature glycosylation of cargo proteins was observed in Arabidopsis expressing the PRA isoform, AtPRA1.B6⁷⁴. While YIP3 deletion failed to rescue the ER membrane association defect of Sar1 in the sec12-4 mutant, it rescued the Sec16 localization defect in sec12-4. These results suggest that Yip3 acts to inhibit Sar1-dependent Sec16 assembly at the ERES (Fig. 8). This was confirmed by the results obtained upon Yip3 overexpression. Furthermore, this proposal is consistent with a previous report that Yip3 overexpression causes ER membrane expansion and growth arrest⁵¹. Recently, it was demonstrated that Sed4, a homolog of Sec12, is essential for the assembly of Sec16 to ERES through its interaction with Sec16⁵⁸. Thus, Yip3 might impede the assembly of Sec16 to ERES by inhibiting the binding between Sed4 and Sec16. Otherwise, since Yip3 may physically interact with Sec23, Sec24 and Sec13⁷⁵, it is possible that Yip3 may disturb Sec16 localization to ERES through its interaction with the COPII components. However, we could not observe by immunoprecipitation that Yip3 interacts with Sec16 (Supplementary Fig. 11c). Therefore, this result supports the notion that Yip3 negatively regulates Sec16 assembly indirectly rather than directly. Further studies will be required to determine whether Yip3 is involved in the dynamics or turnover of Sec16 and ERES. Methods for analyzing dynamics like fluorescence loss photobleaching (FLIP) might provide further insight into the role of Yip3 in the regulation of COPII vesicle formation.

Another striking feature from this study is that the defect in ER membrane association of Sar1 in the sec12-4 mutant was rescued by the loss of Ino2 or Ino4. This suggests that the target genes of Ino2/Ino4 contain a factor that regulates Sar1 activity. Because loss of Ino2/Ino4 did not increase the mRNA levels of SEC12 and because YIP3 deletion failed to rescue the defect in the ER membrane association of Sar1 in sec12-4 mutant, an unknown factor distinct from them may be involved in the activation of Sar1 (Fig. 8). Alternatively, it is possible that structural changes in the ER membrane by loss of Ino2 or Ino4 affect Sar1 activity, since Ino2/Ino4 complex is required for ER membrane expansion⁶⁵. Thus, we propose that the PA-Opi1-Ino2/Ino4 system fine-tunes COPII vesicle formation by coordinating multiple steps in response to changes in the lipid composition of the ER membrane.

Finally, we found that ER stress inducers increase both PA and Yip3 levels and reduce the membrane association of Sec16, which leads to the delayed transport of cargo proteins. Membrane expansion induced by the action of Ino2/Ino4 was shown as a mechanism to overcome ER stress⁶⁵. It has been proposed that this membrane expansion is driven by lipid biosynthesis, but another plausible explanation is that suppression of bulk lipid flow from the ER due to inhibition of COPII vesicle formation is the cause. Yip3 deletion suppressed DTT-induced transport delay of cargo proteins, although the suppression is partial, suggesting that ER stress-induced transport delay is caused by two mechanisms: Yip3-dependent and Yip3-independent. Interestingly, although Yip3 and Sec16 seem not to interact directly, treatment of cells with DTT increased colocalization of Yip3 with Sec16-mRFP marked ERES (Supplementary Fig. 11a, b). However, further study is needed to understand the mechanism by which ER stress causes Yip3-dependent and independent transport delays. Also, the mechanism by which ER stress increases PA levels remains to be elucidated. Nevertheless, we propose that the Ino2/Ino4 activity needed for ER membrane expansion and subsequent rise in Yip3 level, which suppresses cargo transport from the ER, is an integral cellular program to overcome ER stress. Actively reducing vesicle budding during ER stress possibly protects ER homeostasis by lowering export load, preserves limiting membrane lipid precursors for membrane remodeling, and coordinates secretory output with cellular stress and metabolic state to prevent further aggravation of the imbalance. It would be interesting to know whether mammalian homologs of the Yip3/PRA family alleviate ER stress.

Methods

Yeast strains and plasmids

All strains of S. cerevisiae used in this study are listed in Supplementary Data 1. Double mutants were constructed by crossing haploid yeast strains containing single-gene mutations in the same backgrounds, sporulation and subsequent dissection of the spores. The genotypes of spores were verified by colony PCR. To construct a plasmid (pPAH1) overexpressing PAH1, the DNA fragment containing its own promoter and open reading frame of PAH1 was amplified by PCR and cloned into pRS426 (2μ, URA3). The plasmid for expression of Opi1-GFP was constructed as follows. The BamHI-BglII fragment containing GFP(S56T) coding sequence from pFA6a-GFP(S65T)-kanMX6 was inserted into the BamHI site of pRS416 (CEN, URA3) to obtain pRS416-GFP. Subsequently, the EcoRI-BamHI fragment containing the own promoter and open reading frame (without stop codon) of OPI1 was amplified by PCR and cloned into pRS416-GFP to obtain pOPI1-GFP. To construct a plasmid expressing Kar2-SS-mRFP-HDEL, the DNA fragment containing Kar2 signal-peptide sequence (1–135)-mRFP-HDEL was amplified by PCR and cloned into pRS415 (CEN, LEU2) with GPD1 promoter. The plasmid for Yip3 overexpression was constructed by cloning the BamHI-HindIII fragment containing the open reading frame of YIP3 into pRS426 (2 μ, URA3 or HIS3) with the GPD1 promoter. Plasmids expressing Sec7-mRFP⁷⁶, mRFP-Sed5⁷⁶, and Sec13-mCherry⁷⁷ were described previously.

Culture conditions

Strains were grown either in rich YP medium (1% yeast extract, 2% peptone) supplemented with 0.2% adenine and containing 2% glucose (YPD) as carbon source or in synthetic dextrose (SD) minimal medium (0.15% yeast nitrogen base, 0.5% ammonium sulfate, 2% glucose) supplemented with the appropriate amino acids and bases as nutritional requirements or with 0.1% 5-fluoroorotic acid (5-FOA). To test the temperature sensitivity of strains, fivefold serial dilutions of cells were made in sterile water, spotted onto SD plates and incubated for 5 days⁷⁸. For the quantification of colony size formed after tetrad dissection, ImageJ was used.

Western blot analysis of cargo protein maturation, cargo protein level and immunoprecipitation assay

Accumulation of immature CPY or Gas1 was analyzed by SDS-PAGE followed by immunoblotting⁷⁹. Blots were probed with rabbit polyclonal antibodies against CPY and Gas1, and detected by chemiluminescence using a peroxidase-conjugated affinity-purified goat anti-rabbit IgG antibody. Pulse-chase analysis for CPY maturation was carried out exactly as described³⁰. Cells were cultured in SD without methionine at 25 °C to 0.5–2×10⁷/ml, harvested, and resuspended in SD medium without methionine. 3×10⁷ cells were used for each time point and preincubated at 32 °C for 15 min and labeled with 100 μCi of EasyTAG^TM express protein labeling mix, [³⁵S] (1175 Ci/mmol in aqueous solution; PerkinElmer Life Sciences) for 5 min. The chase was initiated by the addition of cold methionine and cysteine in 0.3 M (NH₄)₂SO₄ to 1% final concentration, respectively. The reactions were terminated by adding both NaN₃ and NaF to a 10 mM final concentration. The radiolabeled cells were suspended in TEPI (100 mM Tris-HCl, pH 7.5, 10 mM EDTA, protease inhibitors) and lysed with glass beads. The lysates were boiled in the presence of 1% SDS for 5 min and centrifuged to remove insoluble material. The supernatant was diluted four times with TNET (100 mM Tris-HCl, pH 8, 100 mM NaCl, 5 mM EDTA, 1% Triton X-100) and used for immunoprecipitation of Gas1p with anti-Gas1p rabbit antiserum and protein A-Sepharose (Amersham Biosciences). The samples were separated by SDS-PAGE and analyzed and quantified using a phosphorimager. The percentage of mature Gas1p was determined by calculating the ratio of the 125-kDa mature form to the total signal (125- and 105-kDa ER form) and multiplying by 100.

To analyze protein levels, samples prepared from cell lysates or in the membrane fractionation assay and immunoprecipitation assay were analyzed by SDS-PAGE and immunoblotting using antibodies against Sar1, Pgk1, Emp47, Wbp1, HA and GFP. Cell lysis was done in 100 μl of urea buffer (50 mM Tris [pH 7.5], 5 mM EDTA, 6 M urea, 1% SDS, 1 mM PMSF, and 0.53 × PPi) with glass beads in a bead beater with subsequent heating for 10 min to 65 °C. For NTCB cleavage 30 μl of 0.5 M CHES (pH 10.5) and 20 μl of NTCB (7.5 mM in H2O) were added, and samples were incubated overnight at RT before 1 volume of 2 × sample buffer (+20 mM TCEP and 0.53 × PPi) was added as previously described⁵⁹. Immunoprecipitation using HA trap beads (Chromo Tek) was performed as previously described⁸⁰. Wild-type cells expressing endogenously tagged Yip3-HA together with either endogenously tagged Rtn1-GFP or Sec16-GFP were grown at 25 °C in SD medium and treated with or without 5 mM DTT for 4 h. Cells corresponding to 100 optical density units at 600 nm (OD₆₀₀) were washed twice with TNE buffer [50 mM tris-HCl (pH 7.5), 150 mM NaCl, 5 mM EDTA, 1 mM phenylmethylsulfonylfluoride, and protease inhibitor cocktail; Roche Diagnostics] and disrupted with glass beads. Cell debris and glass beads were removed by centrifugation, and the resulting supernatant was centrifuged at 17,000 g for 15 min at 4 °C. The pellet was resuspended in TNE buffer, and digitonin was added to a final concentration of 1%. After incubation for 1 h at 4 °C with rotation, insoluble material was removed by centrifugation at 13,000 g for 60 min at 4 °C. For immunoprecipitation of Yip3-HA, samples were preincubated with empty agarose beads (ChromoTek) for 1 h at 4 °C, followed by incubation with HA-TrapA (ChromoTek) for 3 h at 4 °C. Immunoprecipitated beads were washed five times with TNE buffer containing 0.2% digitonin, eluted in SDS sample buffer, separated by SDS-PAGE, and analyzed by immunoblotting. For the detection of Sec16-GFP, samples were treated with NTCB prior to SDS-PAGE. The following antibodies were used in this study: Mouse monoclonal anti-GFP antibody (Roche; 11814460001), rat monoclonal anti-HA antibody (Roche; 11867423001), Molecular probe (A-6457) and mouse monoclonal anti-PGK (Molecular Probes; A-6457). Monoclonal antibodies against CPY, Gas1, Wbp1, Emp47 and Sar1 were derived from rabbits and were originally produced.

Fluorescence microscopy

For ERES visualization, cells expressing Sec13-Venus or Sec13-GFP were grown at 25 °C in SD medium, shifted to 32 °C for 60 min and observed by conventional fluorescence microscopy. Cells were chosen randomly, and ERESs were counted manually. For visualization of Opi1-GFP localization, cells were transformed with pOPI1-GFP. The cells were grown at 25 °C in SD medium and imaged by fluorescence microscopy. Perinuclear ER/nucleoplasm intensity ratio was analyzed by ImageJ (version 1.53 k). At least 100 cells were evaluated for each sample per experiment. For ER membrane association of Sar1, cells expressing Sar1-AcGFP with Sar1-AcGFP (pRS316) plasmid⁵⁶ and an ER marker with Kar2-SS-mRFP-HDEL plasmid were grown at 25 °C in SD medium, shifted to 32 °C for 60 min and imaged by fluorescence microscopy. For Sec16 localization, Sec16-GFP was expressed from the endogenous locus on the chromosome, and Sec16-AcGFP was expressed from its own promoter on a pRS314-derived plasmid pTYY42-Sec16-AcGFP (CEN, TRP1)⁵⁶. ER membrane/cytosol intensity ratio was analyzed using ImageJ. Pearson’s correlation coefficient was calculated from the normalized GFP and mCherry/mRFP intensity profiles obtained from ImageJ using the CORREL function in Microsoft Excel. The scale bars in all images represent 5 µm.

Subcellular fractionation analysis

Cells were grown overnight in SD medium, and 12 OD₆₀₀ units of cells were collected per sample and suspended in 300 µl of TNE buffer. A 50-µl aliquot of the total cell lysate was collected as the total fraction. The remaining lysate was centrifuged at 20,000 × g for 30 min at 4 °C to separate the supernatant (S20) and pellet fractions (P20). The S20 fraction was collected, and the P20 fraction was resuspended and adjusted to a final volume of 250 µl with TNE buffer. A 50-µl aliquot was then taken from each fraction for further analysis. To generate a detectable Sec16-GFP cleavage fragment, each 50-µl sample was treated with 18 µl of 0.5 M CHES (pH 10.5) and 12 µl of 7.5 mM NTCB as described previously⁵⁹. Samples were incubated overnight at room temperature under alkaline conditions. After incubation, 30 µl of 4 × SDS sample buffer was added, and samples were heated at 65 °C for 10 min. Proteins were separated by SDS-PAGE and analyzed by western blotting using anti-GFP antibodies.

DNA microarray analysis

Microarray analysis was performed as described previously⁸¹, using the Gene Chip Yeast Genome 2.0 Array (Affymetrix). Total RNA was extracted from the yeast cells grown at 25 °C in YPD medium, and mRNA was purified using the Oligotex TM-dT30 <Super> mRNA Purification Kit (Takara Bio Inc., Japan). Biotinylated cRNA was prepared from 500 ng of mRNA according to the standard Affymetrix protocol, and 5 μg of cRNA was hybridized for 16 h at 45 °C on the GeneChip Yeast Genome 2.0 Array. GeneChips were washed and stained using the Hybridization, Wash, and Stain Kit (Affymetrix). Data were analyzed with Operating Software (GCOS; version 1.4), using the Affymetrix default analysis settings and global scaling as the normalization method. The trimmed mean target intensity of each array was arbitrarily set to 500. Microarray data can be retrieved from Gene Expression Omnibus (GEO) under the accession code GSE168638. DNA microarray data were obtained for n = 3 independent biological replicates from sec12-4 and sec12-4 ino2Δ ino4Δ cells.

MS analysis of glycerophospholipids

Mass spectrometer (MS) analysis of glycerophospholipids was carried out exactly as described¹⁵. Cells were grown at 32 °C in SD medium, and the lipids were extracted with the extraction solvent (ethanol, water, diethyl ether, pyridine, and 4.2 N ammonium hydroxide [15:15:5:1:0.018, v/v/v/v/v]) with an internal standard mix. Glycerophospholipids were analyzed by using a Triple Stage Quadrupole (TSQ) Vantage Mass Spectrometer (Thermo Scientific) equipped with a robotic nanoflow ion source Nanomate HD (Advion Biosciences, Ithaca, NY). The abundance of glycerophospholipids was calculated by their signal intensities relative to the internal standards and standard curves constructed using the internal standard lipids. The data presented are not corrected for +2 isotopes, as the data are presented as relative to wild type, and the vast majority of the isotope correction would be canceled out.

TLC analysis of phosphatidic acid

Phospholipids from cells or ER-rich membrane fractions were extracted and separated by double one-dimensional TLC on 20 cm × 20 cm TLC plates as described previously⁸². Lipid samples were applied to an origin line positioned 10.5 cm from the bottom edge of the plate. For the first dimension, plates were developed with chloroform/methanol/ammonium hydroxide (65:35:7.5, v/v/v) until the solvent front reached 1 cm from the top edge. Plates were then air-dried overnight. Standard phosphatidylcholine (PC), which was applied to both edges of the plate, was visualized by selectively spraying the edge regions with 0.05% primulin in acetone/water (8:2, v/v), followed by detection under ultraviolet light. The upper part of the plate containing the PC standard was then removed. The remaining plate was inverted and developed in the second dimension with chloroform/acetone/methanol/acetic acid/water (50:20:10:15:5, v/v/v/v/v) until the solvent front reached 1 cm from the top edge of the inverted plate. After the second development, plates were air-dried overnight and sprayed with 0.05% primulin in acetone/water (8:2, v/v), and lipids were visualized under ultraviolet light.

Statistical analysis

All experiments were performed with n = 3 independent biological replicates except Fig. 6b (n = 5), Supplementary Fig. 3e (n = 6) and pulse-chase analyses (Figs. 1d, 2d and Supplementary Fig. 4b; n = 2). Statistical analyses were performed using a two-tailed unpaired Student’s t-test to compare the means of two independent groups, based on at least three biological replicates. For comparisons involving three or more groups, one-way ANOVA followed by the Tukey–Kramer multiple comparison test was used to compare the means of at least three biological replicates per group. Data are presented as mean ± sd from at least n = 3 independent biological replicates. Graphs in this study were generated using either Excel or R software (version 4.5.1; R Foundation for Statistical Computing). Analysis and data presentation of superplots were performed as described previously⁸³.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

K.H., M.N., P.S., H.N., Me.K., R.I., M.I., K.E., T. K., A.I. and Y.Y. constructed strains and plasmids, performed biochemical studies, fluorescent microscopy experiments and data analysis. J.M.-L., A.A-R., S.S.-B. and A.P.-L. performed pulse-chase experiments with help from M.M., Me.K., A.I., Y.Y. and Mu.K. performed microarray experiments with help from H.I., I.R. and H.R. performed MS analysis for lipids. K.F. designed the experiments and wrote the manuscript with input from all other authors.

Peer review

Peer review information

Nature Communications thanks Alexandre Toulmay and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

This work was funded by the Grants-in-Aid for Scientific Research from Japan Society for the Promotion of Science, Japan [JP19H02922, JP21K19088 to K.F.], by the MICIU/AEI/ 10.13039/501100011033 and “ERDF A way of making Europe” grant PID2023-151267NB-I00 to M.M., and by the SNSF (grants 51NF-40-185898 and 310030_184949) and the Leducq Foundation to H.R.

Data availability

The DNA microarray data for sec12-4 and sec12-4 ino2Δ ino4Δ generated in this study have been deposited under the GEO Accession Number GSE168638 in NCBI and are provided as a supplemental dataset (Supplementary Data 3). Lipidomics data related to this study are provided as supplementary datasets (Supplementary Data 2, 4). All other raw data are included in the source data files submitted with the manuscript and are fully accessible. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-75057-x.