A conserved ER protein prevents lipotoxicity by stimulating the key enzyme in glycerolipid synthesis Authors: Yan Li, Siwei Huang, Lam Shing Chan, Xuesong Li, Ning-Sum To, Xue Zhao, Kwangsek Jung, Xingyu Yang, Yu Chung Tse, Ningyi Xu, Yu Xia, Tom H Cheung, Jianan Qu, Ho Yi Mak Published: 2026 ABSTRACT AbstractThe synthesis of glycerolipid is essential to prevent the toxic effects of excess fatty acids. Glycerolipids are generated by the sequential attachment of acyl-CoA to glycerol by endoplasmic reticulum (ER)-localized acyltransferases, ultimately yielding non-toxic triacylglycerol that can be stored in lipid droplets. The key acyltransferase GPAT4 associates with the CHP1 protein, but the regulation of GPAT4 remains poorly understood. Here, we have used genetic, biochemical, and imaging techniques to identify TMEM120A as GPAT4-activating protein. We show that ER-localized TMEM120A and CHP1 synergistically activate GPAT4 and promote the incorporation of medium and long chain acyl-CoA into glycerolipid. C. elegans mutants of TMEM120A or CHP1 ortholog are susceptible to high-fat diet induced sterility, in part due to their deficiency in lipid droplet expansion. In mammalian cells, fatty acid supplementation and myristoylated CHP1 enhance the association of TMEM120A with GPAT4. Together, our results reveal an unexpected mechanism that alleviates lipotoxicity.Subject terms: Endoplasmic reticulum, Glycerides, Fatty acids Introduction Excess fatty acids are toxic to cells1. To prevent lipotoxicity, excess fatty acids are incorporated into neutral lipids, such as triacylglycerol (TAG), via the glycerol-3-phosphate (G-3-P) or the monoacylglycerol pathway2. The nontoxic TAG can then be stored in lipid droplets (LDs), which are conserved organelles for fat storage3–7. Central to the G-3-P pathway is the sequential addition of fatty acyl-coenzyme-A (acyl-CoA) molecules to the glycerol backbone by acyltransferases, with the G-3-P acyltransferase, GPAT4, as the rate-limiting enzyme8. However, how the G-3-P pathway is regulated in response to excess fatty acids is poorly understood. In the G-3-P pathway, purified acyltransferases show a clear preference to acyl-CoA with specific chain lengths9,10. Nevertheless, various lipidomic analyses revealed there to be a spectrum of acyl chains in TAG from animal cells11–13, suggesting a cryptic mechanism that enhances the ability of acyltransferases to accept less favorable substrates. Here, we report TMEM120A as a key enhancer of substrate plasticity. C. elegans has been used extensively as a model organism to link mechanisms of LD regulation to whole animal physiology14. From a genetic screen for mutants with altered LD size, we identified the C. elegans ortholog of TMEM120A (alternatively known as NET29 or TACAN), which is a member of a conserved family of transmembrane proteins. TMEM120A was first reported in a proteomic study of the nuclear envelope15. A subsequent functional analysis indicated that TMEM120A and its paralog TMEM120B are required for the differentiation of 3T3-L1 pre-adipocytes into mature adipocytes16. More recently, TMEM120A was reported to act at the cell surface as an ion channel that is sensitive to mechanical cues17. However, multiple independent studies did not support this conclusion18–21. Instead, structural analyses indicated that TMEM120A forms a symmetrical homodimer and each protomer binds specifically to a coenzyme-A (CoASH) molecule19–21. Although TMEM120A shares structural similarity to the fatty acid elongase, ELOVL722, results from the MacKinnon lab show that TMEM120A does not promote fatty acid elongation19. Therefore, the precise role of TMEM120A in lipid metabolism is still largely unknown. The lipid-modified, myristoylated form of calcineurin B homologous protein (CHP) 1 is the only known GPAT4 activator thus far23. Interestingly, CHP1 appears to play additional roles in regulating the maturation and trafficking of Na+/H+ exchangers (NHEs) to the cell surface24. Therefore, it remains unclear how CHP1 can engage different clients to modulate their functions in diverse biological pathways. In this paper, we report the identification of mutants that are defective in the sole C. elegans ortholog of TMEM120A and CHP1. We present new evidence, based on genetic, biochemical, imaging, and lipid analysis, that C. elegans TMEM-120/TMEM120A acts at the ER, in the same pathway as CHPF-1/CHP1, to promote the incorporation of medium and long chain acyl-CoA into glycerolipid. Furthermore, excess fatty acids promote the interaction between GPAT4, CHP1 and TMEM120A, which leads to the synergistic activation of GPAT4. Thus, our study reveals an unexpected convergence of regulatory mechanisms at the rate-limiting step of glycerolipid synthesis. Results TMEM-120 is required for LD expansion in C. elegans In response to demand for energy or membrane components, fatty acids that are stored in the form of TAG can be released from LDs and subsequently processed by β-oxidation pathways to yield acetyl-CoA25. We have previously shown that LDs undergo abnormal expansion in C. elegans daf-22/thiolase mutants, owing to a block in the peroxisomal β-oxidation pathway26. These animals accumulate more TAG through the concerted action of ACS-22/acyl-CoA synthetase and DGAT-2/diacylglycerol acyltransferase27, presumably to prevent the inappropriate release of fatty acids that cannot be further processed. Accordingly, a loss of ACS-22 or DGAT-2 function attenuates LD expansion in daf-22 mutant animals. In a genetic screen for additional daf-22 suppressors, we identified a complementation group consisting of two recessive alleles, hj49 and hj50. Molecular cloning revealed lesions in a gene annotated as M01G5.3, hereafter named tmem-120, which encodes a predicted transmembrane protein that is homologous to mammalian TMEM120A and TMEM120B16 (Fig. S1A). The hj50 nonsense allele (Glutamine 290 to amber) caused ~85% reduction of tmem-120 mRNA level (Fig. S1A, B), possibly due to nonsense mediated decay. Therefore, it is a strong loss-of-function allele. The hj49 missense allele caused the substitution of a conserved glycine to glutamate (G195E) (Fig. S1A). Based on the structure of TMEM120A, this conserved glycine residue is located in a flexible linker immediately upstream of a coenzyme A (CoASH) binding site19–21. Therefore, its replacement with glutamate may alter the conformation of the linker and, in turn, affect the orientation of the residues that constitute the CoASH binding site (including W193 of human TMEM120A and W197 of C. elegans TMEM-120). Since the phenotypes of mutant animals carrying hj49 and hj50 were indistinguishable (Fig. S1C–E), we concluded that the G195E substitution severely compromised TMEM-120 function. In subsequent experiments, we used animals carrying the hj50 nonsense allele for phenotypic analysis. To visualize the LDs of wild type and mutant animals, we employed a stimulated Raman scattering (SRS) microscopy system28. We focused on detecting C–H bond vibrations from neutral lipids, as these are highly concentrated in LDs. In agreement with previous results based on the use of vital dye or fluorescent protein markers27, LDs in the daf-22 mutants were larger than those in wild type animals (Fig. 1A). Furthermore, a loss of tmem-120 function reduced the size and blocked the expansion of LDs in both the wild type and daf-22 mutant animals (Fig. 1A, B). Similar results were observed when we used mRuby fused to DHS-3/17-beta-hydroxysteroid dehydrogenase 13 (HSD17B13), hereafter called DHS-3::mRuby, as a fluorescent LD marker29 (Fig. S1C). Therefore, TMEM-120 and DAF-22 could regulate LD expansion independently. Next, we quantified the SRS signals, which are proportional to the neutral lipid content30. We detected 23% more SRS signals in the daf-22 mutants than in the wild type animals (Fig. 1C, D). In contrast, tmem-120 and daf-22; tmem-120 mutant animals had 26% and 14% less SRS signals, respectively, than wild type animals (Fig. 1C, D). Such a reduction of neutral lipids was unlikely due to an alteration of the feeding rate (Fig. S1F). Furthermore, the total number of LDs was unchanged in tmem-120 mutant animals, suggesting that TMEM-120 has minimal involvement in the formation of LDs (Fig. S1G). Taken together, these results indicate that TMEM-120 specifically supports neutral lipid accumulation and LD expansion. TMEM-120/TMEM120A promotes LD expansion. A Visualization of LDs in wild type (WT), tmem-120(hj50), daf-22(ok693), and daf-22(ok693); tmem-120(hj50) young adult animals by stimulated Raman scattering (SRS). Representative images of the first two intestinal segments are shown: anterior left. Scale bar = 10 μm. B Quantification of LD diameter in L4 animals using the DHS-3::mRuby(hj200) marker. Each data point shows the average LD diameter in one animal. Mean ± SD of 10 animals are shown. C As in (A), showing images of entire L4 animals at the focal plane with the strongest SRS intensity. Scale bar = 100 μm. D Quantification of neutral lipid content of animals from (C). Mean ± SD of 20 animals per strain are shown except daf-22(-) (n = 19). The mean value for the WT group was normalized to 1 arbitrary unit (a.u.). E Visualization of LDs in L4 animals (DHS-3::mRuby (hj200) marker). Representative maximum intensity projections (7.5 μm z-stacks, second intestinal segment) of daf-22(-) (top left), daf-22(-); tmem-120(hj50) (bottom left), daf-22(-); tmem-120(hj50); hjEx29[tmem-120] (top right) and daf-22(-); tmem-120(hj50); hjEx30[mTmem120a] (bottom right) are shown. Arrowheads indicate enlarged LDs. Scale bar = 5 μm. F Visualization of LDs in oleic acid (OA)-treated WT and TMEM120A overexpressing (OE) COS7 cells with FAS dye. Representative confocal projections from three independent experiments are shown. Scale bar = 10 μm. G Box plot to show the LD with the largest diameter in WT (n = 183) and TMEM120A OE COS7 (n = 239) cells from (F). The boxes indicate the interquartile range, the whiskers show the minimum and maximum values, and the median value is marked by a horizontal black line. H Representative thin-layer chromatography analysis of lipids from WT and TMEM120A OE COS7 cells treated with OA ± DGAT1 inhibitor (DGAT1i, A922500) for 3.5 h. Triolein and OA were loaded as TAG and free fatty acid (FFA) standards, respectively. I Quantification of TAG signals from (H). Data are presented as mean ± SD, n = 4 per group. J Same as (I), but showing quantification of FFA signals. Statistical analyses: (B, D, I, J) two-way ANOVA followed by Tukey’s multiple comparisons test; G two-sided unpaired t-test. Next, we sought to determine if mammalian TMEM120A plays a conserved role in regulating fat storage. To this end, we expressed mouse TMEM120A in daf-22; tmem-120 mutant animals. We found that TMEM120A could rescue the LD phenotype in a similar manner to C. elegans TMEM-120 (Fig. 1E). As a result, large LDs reappeared in intestinal cells of daf-22; tmem-120 mutant animals. Finally, the size of LDs in daf-22; tmem-120 and daf-22; dgat-2; tmem-120 mutant animals are comparable (Fig. S1H, I), which is consistent with the notion that TMEM-120 acts in the same pathway as DGAT-2. Therefore, we concluded that TMEM-120 and TMEM120A share a deeply conserved function in regulating cellular fat storage. TMEM120A promotes LD expansion in mammalian cells Next, we tested if human TMEM120A is sufficient to promote LD expansion in mammalian COS7 cells. To this end, we generated COS7 cells that stably overexpressed TMEM120A using the Sleeping Beauty transposon system31. When oleic acid (OA) was added to the culture medium, we detected significantly larger LDs in cells that overexpressed TMEM120A when compared with those in the parental cells (Fig. 1F, G). Therefore, TMEM120A is sufficient to promote LD expansion, plausibly by enhancing the incorporation of exogenous fatty acids into TAG. To further explore the impact of TMEM120A overexpression, we compared the effect of DGAT1 inhibition after OA supplementation in wild type and TMEM120A overexpressing (OE) cells. Consistent with our imaging results, TMEM120A OE cells produced more TAG upon short-term OA supplementation (Fig. 1H, I). Simultaneous DGAT1 inhibition and OA supplementation sharply reduced the TAG levels in both cell lines when compared to the OA controls (Fig. 1H, I). In addition, DGAT1 inhibition triggered free fatty acids (FFA) accumulation, which was more severe in TMEM120A OE than wild type cells (Fig. 1H, J), consistent with enhanced upstream flux that was poorly buffered once DGAT1 was inhibited. This excess FFA was associated with impaired mitochondrial function, as assessed by Seahorse respirometry. TMEM120A OE cells showed elevated basal and maximal respiration with OA alone (Fig. S2A, B), but these rates were reduced upon OA and DGAT1 inhibitor (DGAT1i) co‑treatment (Fig. S2C, D). Prolonged OA+DGAT1i treatment induced near-complete cell death in TMEM120A OE cells, whereas wild type cells were only moderately (~40–50%) affected (Fig. S2E, G). This vulnerability was also observed during PA+DGAT1i co-treatment (Fig. S2F, H). Furthermore, we demonstrated that the OA + DGAT1i induced cell death was dose-dependent and could be reproduced with a second DGAT1i (Fig. S2I, J). Together, these data indicate that TMEM120A contributes to DGAT1‑dependent TAG synthesis by promoting the sequestration of excess fatty acids in LDs. When this storage route is compromised, the resulting FFA accumulation and mitochondrial dysfunction likely contribute to lipotoxic cell death in TMEM120A OE cells. TMEM-120 promotes the incorporation of fatty acids into neutral lipids Cellular neutral lipid homeostasis is dependent on the balance between TAG synthesis and mobilization at LDs32. TAG synthesis, in turn, relies on the availability of dietary or de novo synthesized fatty acids. We sought to determine if the decrease in neutral lipid content in tmem-120 mutant animals was caused by a decrease in its accumulation or an increase in its mobilization by lipolysis. Based on quantitative real-time PCR analysis, the expression levels of genes known to regulate lipogenesis and lipolysis were not significantly changed in the tmem-120 mutants, except fat-7/fatty acyl desaturase (Fig. S2K). Next, we used SRS to detect exogenously supplied deuterium-labeled fatty acids33,34. We tuned our system to detect C–D bond vibrations, to avoid interference from endogenous fatty acids. To measure neutral lipid accumulation, we fed young adult stage wild type and tmem-120 mutants with deuterium-labeled monounsaturated oleic acid (OA-d34) and imaged live animals by SRS at regular intervals over a period of 30 h (Fig. 2A). We detected a progressive increase in SRS signals from LDs in both strains (Fig. 2C). However, the rate of increase in the mutants was significantly slower than that in the wild type animals (Fig. 2E). Similar observations were made when animals were fed deuterium-labeled saturated palmitic acid (PA-d31) (Fig. 2G). We concluded that a loss of TMEM-120 function impaired the incorporation of exogenous fatty acids into neutral lipids. TMEM-120 promotes the incorporation of fatty acids into neutral lipids. Schematics to show the experimental design used to monitor deuterated oleic acid-d34 (OA-d34) accumulation (A) and turnover (B). Visualization of OA-d34 incorporation (C) and turnover (D) by SRS in WT and tmem-120(hj50) animals. These are representative fluorescence images showing a single focal plane through the first and second intestinal segments, which exhibited the strongest SRS signal. Scale bar = 10 μm. Quantification of OAd34 accumulation (E) and turnover (F) in the WT and tmem-120(hj50) animals shown in (C, D), respectively. The data indicate the mean ± SD of independent biological replicates, with sample sizes ranging from n = 5 to 9 (E) and n = 7 to 10 (F) worms per group at each time point. Exact n values for each time point are provided in the Source data file. Quantification of palmitic acid-d31 (PA-d31) accumulation (G) and turnover (H) in WT and tmem-120(hj50) worms. The data indicate the mean ± SD of independent biological replicates, with sample sizes ranging from n = 5 to 9 (G) and n = 4 to 9 (H) worms per group at each time point. Exact n values for each time point are provided in the Source data file. In E–H, the straight dashed lines and equations were generated based on linear regression analysis of each group. Statistical analyses: E–H two-sided unpaired t-test (for each time point). To measure lipolysis, we fed newly hatched larval stage L1 animals with OA-d34 until they reached the young adult stage. We then removed labeled fatty acids from their diet and imaged them by SRS at regular intervals (Fig. 2B). The dissipation of SRS signals over time reflected the rate of lipolysis, as the labeled fatty acids stored as neutral lipids in LDs were metabolized. We found that the rate of decrease of SRS signals was not significantly different between the wild type and tmem-120 mutant animals (Fig. 2D, F). Similar observations were made when animals were fed with PA-d31 (Fig. 2H). We concluded that the turnover of labeled fatty acid-derived neutral lipids, as measured by SRS, was not altered in the tmem-120 mutants. Taken together, these results suggest that the reduction of neutral lipid content in tmem-120 mutants may be attributed to a retardation of fatty acid incorporation into neutral lipids. TMEM-120 mutant animals are resistant to diet-induced TAG synthesis We challenged wild type and tmem-120 mutant animals with a high-fat diet (HFD), instigated by the addition of 2% glucose to the nematode growth medium. It has previously been shown that glucose supplementation increases the cellular fatty acid level in C. elegans, either directly or indirectly through dietary E. coli that synthesize more fat in response to glucose35,36. Accordingly, a significant increase in LD size was observed in wild type animals under HFD (Fig. 3A, B). In contrast, tmem-120 mutant animals were resistant to such treatment. This is consistent with our model that TMEM-120 is required for efficient incorporation of fatty acids into neutral lipids. tmem-120 mutant animals are resistant to high fat diet induced TAG synthesis. A Visualization of LDs (with DHS-3::mRuby (hj200)) in WT and tmem-120(hj50) L4 animals fed E. coli OP50 in the absence (control) or presence of 2% glucose (high fat bacterial diet; HFD). Max-intensity projections (7.5 μm depth, second intestinal segment) are shown. The regions outlined by the white squares were magnified 4× and are shown as insets. Scale bar = 5 μm. B Quantification of LD diameter in WT and tmem-120(hj50) L4 animals shown in (A). Each data point shows the average LD diameter in one animal. Mean ± SD of 10 animals are shown. C Schematic of glycerolipid synthesis via the glycerol-3-phosphate (G-3-P) and monoacylglycerol (MAG) pathways. GPAT glycerol-3-phosphate acyltransferase, AGPAT acylglycerolphosphate acyltransferase, MGAT monoacylglycerol acyltransferase, DGAT diacylglycerol acyltransferase. Quantification of TAG (D), DAG (E), PC (F) and PE (G) in WT and tmem-120(hj50) L4 animals fed a control diet or HFD, by LC-MS. Mean ± SD of three independent biological samples (~3000 animals each); WT (control) group was normalized to 1. H Visualization of LDs in L4 WT and tmem-120(hj266) animals treated with L4440 (vector control) or fasn-1 RNAi. Intestine-specific RNAi was achieved by using the sid-1(qt78); hjSi493[vha-6p::sid-1] genetic background. I As in (B), but with 5 animals of each experimental group shown in (H). J Stacked bar chart to show the percentage growth rate of WT and tmem-120(hj50) animals fed a control diet or HFD. Animals were synchronized at the L1 stage and quantified after 48 h. The total number of animals quantified: WT (control) = 259, tmem-120(hj50) (control) = 245, WT (HFD) = 229, and tmem-120(hj50) (HFD) = 209. K Quantification of brood size of WT and tmem-120(hj50) animals fed a control diet or HFD. Total number of animals counted: WT (control) = 11, tmem-120(hj50) (control) = 12, WT (HFD) = 12, and tmem-120(hj50) (HFD) = 12. The data show the mean ± SD. Statistical analyses: two-way ANOVA followed by Tukey’s multiple comparisons test. To define more precisely the block imposed by TMEM-120 deficiency on neutral lipid synthesis, we compared the lipid profiles of wild type and tmem-120 mutant animals, which were raised on a control diet or HFD, by liquid chromatography-mass spectrometry (LC-MS). Principal component analysis (PCA) of all detectable lipid species (including but not limited to TAG, PC, PE, DAG, PI, and PS) in the glycerolipid synthesis pathways (Fig. 3C) suggested a strong difference in the overall lipid profiles of wild type and tmem-120 mutant animals, in the absence or presence of HFD (Fig. S4A). As expected, we found that tmem-120 mutants had almost 30% less TAG than wild type animals (Fig. 3D). If TMEM-120 acted exclusively with DGATs, then it is plausible that excess fatty acids might be diverted to phospholipid synthesis, according to the canonical glycerolipid synthesis pathway (Fig. 3C). Interestingly, such a compensatory increase in PC and PE levels was not observed in tmem-120 mutants under HFD. Instead, the DAG, PC, and PE levels were all lower in the tmem-120 mutants than in the wild type animals (Fig. 3E–G), suggesting that TMEM-120 acts at an early step of glycerolipid synthesis. Our results indicate that TMEM-120 deficiency causes a broad impairment in the incorporation of excess fatty acids into glycerolipid. To explore the role of TMEM-120 in promoting the incorporation of de novo synthesized fatty acids, we performed intestine-specific RNAi against fasn-1/fatty acid synthase, a key enzyme for lipogenesis37, using previously established method38. Attenuating fatty acid synthesis significantly reduced the LD size in wild type animals (Fig. 3H, I). In contrast, there was no further reduction in LD size when fasn-1 was knocked down in tmem-120 mutants, suggesting that FASN-1 and TMEM-120 act in the same pathway. We concluded that TMEM-120 contributes to the incorporation of de novo synthesized fatty acids into glycerolipid. Enhanced susceptibility of tmem-120 mutants to HFD The incorporation of excess dietary fatty acids into glycerolipid is a crucial protective mechanism against lipotoxicity. Hence, the increase in dietary fatty acids resulting from an HFD was tolerated by wild type animals, with only a minor impact on their growth rate and brood size (Fig. 3J, K). In contrast, when tmem-120 mutants were fed an HFD, we observed a severe delay in their growth and sterility. The correlation between membrane fluidity, growth rate and fertility in C. elegans was previously established39. Therefore, we entertained the hypothesis that the altered acyl chain composition of glycerolipid in tmem-120 mutant animals might affect their membrane fluidity and, in this way, contribute to their susceptibility to dietary fatty acids. We determined the fluorescence recovery after photobleaching (FRAP) of a membrane-associated prenylated green fluorescent protein (GFP-CAAX) as an indicator of membrane fluidity (Fig. S3A, B)39. We detected an increase in recovery half-time and a decrease in diffusion coefficient, indicating a reduction in membrane fluidity, which was specific to the tmem-120 mutants raised under HFD (Fig. S3C, D). Our results correlate with the enhanced susceptibility of tmem-120 mutant animals to excess fatty acids and further reinforce the notion that TMEM-120 plays a protective role against lipotoxicity. TMEM-120 promotes the incorporation of specific acyl chains into TAG We next compared the acyl chain sum composition of TAG and its intermediates in the glycerolipid synthesis pathway in wild type and tmem-120 mutant animals using LC-MS. With a normal diet, TAG and PC with relatively low acyl chain sums appeared to be preferentially depleted in tmem-120 mutant animals (Fig. 4A, C). However, there was no clear trend that suggested a concomitant depletion of specific acyl chains in DAG and PE (Fig. 4B, D). Interestingly, we detected a corresponding increase in sphingomyelin (SM) with low acyl chain sums in tmem-120 mutants (Fig. S4B, C) even though the overall SM level was not significantly changed in these animals (Fig. S4D). Therefore, specific dietary fatty acids that could not be incorporated into glycerolipid were partially diverted to form SM. When animals were subjected to an HFD, the depletion of specific classes of PC and TAG was further exaggerated (Fig. 4A, C). Altered glycerolipid profiles of tmem-120 mutant animals. Relative abundance of each group of triacylglycerol (TAG) (A), DAG (B), PC (C), and PE (D) species with the same total chain length in WT and tmem-120(hj50) worms fed a control diet or HFD. Three independent biological samples are shown for each group. For each dietary condition, the mean value for the WT (in each dietary group) was assigned as 1, and the log2 fold change was plotted. In B, C39 could not be detected, and so this is shown in grey. Bubble plots showing the TAG chain length determined from absolute lipidomic analysis of WT and tmem-120(hj50) mutants fed a control diet (E) or HFD (F). The y-axis represents the total acyl chain length of each TAG species, and the x-axis shows the relative abundance (%) of each TAG in the tmem-120(hj50) mutants when compared with the WT (hj50/WT). The color of each bubble indicates the adjusted p-value (Benjamini–Hochberg method) for statistical significance derived from unpaired two-sided t-tests, and the size of each bubble corresponds to the absolute abundance of that TAG species in each WT worm (ng). G, H Same as (E, F), but the y-axis indicates the number of carbon–carbon double bonds in TAG species for worms fed a control diet (G) or HFD (H). We performed a second round of LC-MS analysis to determine how TMEM-120 deficiency affected the absolute quantity of individual lipid species. The different treatment groups produced distinct lipid profiles based on PCA, with the pooled quality control samples occupying the central position of the plot, thus demonstrating the robustness of the analysis (Fig. S4E). TAG, with an acyl chain sum of 52 carbons (C52 TAG), was the most abundant species in the wild type group, such that 0.92 ± 0.05 and 2.02 ± 0.66 ng were detected per animal when they were fed the control and high fat diet, respectively. In comparison, in tmem-120 mutants, only 0.41 ± 0.01 and 0.73 ± 0.08 ng C52 TAG were detected in the control and HFD groups, respectively. Consistent with our initial analysis, TAG species with lower acyl chain sums were preferentially depleted in the tmem-120 mutants, independent of their intrinsic abundance (Fig. 4E, F). A similar trend was observed for PC and to a lesser extent PE (Fig. S4G, H). In contrast, when acyl chains with different numbers of C=C bonds were considered, there was a relatively uniform reduction of all detectable TAG species (Fig. 4G, H). Taken together, our observations suggest a possible role for TMEM-120 in promoting the incorporation of medium or long acyl chains, regardless of their degree of saturation, during TAG synthesis. TMEM-120 is an ER resident protein We conducted a TOPCONS analysis40, and showed that the C. elegans TMEM-120 and its human orthologs share the same predicted membrane topology, with their N- and C-termini facing the cytoplasm (Fig. 5A). According to their cryo-EM structures, membrane embedded TMEM120A and its paralog TMEM120B have 6 transmembrane helices and form homodimers18–21. Intriguingly, a consensus has yet to emerge regarding the subcellular localization of TMEM120A. Mammalian TMEM120A was previously reported to reside at the nuclear envelope16 or the ER41. In contrast, the assignment of TMEM120A as a mechanosensory channel placed it at the plasma membrane when overexpressed17. We sought to determine the localization of C. elegans TMEM-120 when it was expressed at the endogenous level in live animals. To this end, we used CRISPR mediated genome editing42 to insert the coding sequence of GFP into the endogenous tmem-120 locus (Fig. 5B). We generated two independent knockin alleles, hj239 and hj270 (Fig. 5B). In one strain, GFP was fused to the N-terminus of TMEM-120 via a flexible linker, whereas in the other, GFP was expressed as part of the predicted C-terminal cytoplasmic tail, leaving the extreme C-terminus unaltered. This design was necessitated by the presence of a KxHxx motif at the C-terminus of TMEM-120, which functions similar to the KxKxx motif for ER retention43,44. We found that both TMEM-120 fusion proteins were functional because their expression did not significantly alter the lipid content of daf-22 mutant animals, as observed when the function of TMEM-120 was lost (Fig. S5A–C). We used established markers of intestinal ER and LDs to ascertain the localization of TMEM-120 (Fig. 5C, D) and found that it co-localized extensively with the ER marker. TMEM-120 is an ER resident protein. A The topology of TMEM-120 based on its homology with TMEM120A. B Schematic representation of TMEM-120 GFP fusion proteins encoded by tmem-120(hj239) and tmem-120(hj270) knockin alleles. Visualization of TMEM-120 GFP fusion protein (green) and either an intestinal-specific luminal ER marker SEL-1(18–79)::mCherry::HDEL (hjSi158) (magenta) (C) or an LD marker DHS-3::mRuby (hj200) (magenta) (D) in L4 animals. Representative single optical sections through the first and second intestinal segments are shown, and in each image, the intestine is enclosed by white dashed lines. In the merged images, the regions outlined by the white squares were magnified 5×, and these are shown as separate panels. Scale bar = 10 μm. E Schematic diagram showing the GFP nanobody::ZIF-1 mediated degradation of cytoplasmic GFP fusion proteins. In contrast, GFP targeted to the ER lumen is protected from degradation. F Visualization of an intestinal-specific luminal ER marker SEL-1(18–79)::GFP::HDEL (hjSi528) and TMEM-120 GFP fusion proteins in L4 animals in the absence (control) or presence of GFP nanobody::ZIF-1 (hjSi524). Each image is a projected stack of confocal optical sections spanning a depth of 5 μm through the second intestinal segment, and the intestine is enclosed by white dashed lines. GFP signals in the hypodermis (i.e., regions outside the dashed lines) were unaffected by GFP nanobody::ZIF-1. Scale bar = 10 μm. Next, we determined the localization of TMEM-120(G195E) by inserting the GFP coding sequence into the 5′ end of tmem-120(hj49) to yield a new allele (hj297) (Fig. S5A). We found that GFP::TMEM-120(G195E) was expressed at a comparable level as the wild type protein and localized correctly to the ER (Fig. S5D–F). Therefore, we conclude that the G195E mutation most likely attenuates TMEM-120 function by interfering with CoASH binding, as predicted by the TMEM120A structure. Finally, we experimentally verified the location of the N- and C-terminal tails of TMEM-120. To this end, we took advantage of the anti-GFP nanobody (vhhGFP4) directed protein degradation system45. When a GFP nanobody::ZIF-1 fusion protein is expressed in the intestinal cytoplasm, GFP fusion proteins with the GFP moiety exposed in the cytoplasm are subject to degradation and the fluorescence signals are lost (Fig. 5E). However, if the GFP moiety is in the ER lumen, then the fusion protein is protected, and fluorescence is maintained. Indeed, we detected fluorescence signals in animals that expressed a luminal ER GFP marker, SEL-1(18–79) (hjSi528), even when GFP nanobody::ZIF-1 was co-expressed in the intestine (Fig. 5F). In contrast, both versions of the GFP fusion protein with TMEM-120 were subject to degradation, resulting in a lack of fluorescence when GFP nanobody::ZIF-1 was co-expressed (Fig. 5F). As a control, hypodermal GFP fusions with TMEM-120 remained detectable (Fig. 5F), indicating that the anti-GFP nanobody directed, post-translational degradation, was tissue-restricted as designed. Taken together, our results firmly suggest that TMEM-120 is an ER-resident protein, with its N- and C-termini both oriented towards the cytoplasm. TMEM120A associates with GPAT4 and CHP1 in the ER To elucidate the mechanism by which TMEM120 family members promote glycerolipid synthesis in the ER, we sought to identify their protein partners in mammalian cells by TurboID-mediated labeling41. TurboID is an engineered enzyme that biotinylates proteins in its proximity41. Transgenic HEK293T cells were generated using CRISPR-mediated insertion into the AAVS1 safe-harbor site46. After clonal selection, cell lines that stably expressed TurboID-TMEM120A or TurboID-TMEM120B fusion proteins, or the negative control TurboID-NES (nuclear export signal), were established (Fig. 6A). All fusion proteins were expressed at comparable levels (Fig. 6B). We further showed that the TurboID-TMEM120A and TurboID-TMEM120B fusion proteins were functional because they promoted neutral lipid accumulation upon OA loading, as indicated by an increase in LipidTOX red signals (Fig. 6C). In contrast, cells expressing the cytoplasmic TurboID-NES control showed no response. As proof of principle, we detected a similar spectrum of biotinylated proteins from cells that expressed either TurboID-TMEM120A or TurboID-TMEM120B, but these were largely distinct from those extracted from wild type parental or TurboID-NES expressing cells (Fig. 6E). Our results indicate that the sequence and structural similarities of TMEM120A and TMEM120B enable them to bind similar protein partners. Identification of TMEM120A and TMEM12B interacting proteins. A Schematic representations of TurboID-V5-TMEM120A, TurboID-V5-TMEM120B, and TurboID-V5-NES (nuclear export signal; negative control) fusion proteins. B Detection of TurboID fusion proteins from HEK293T AAVS1-site stable insertion cell lines by Western blotting using anti-V5 antibodies. The results shown are representative of three independent experiments. C Comparison of LipidTOX red signal intensity between WT and TurboID fusion proteins expressing HEK293T cells by FACS. Cells were treated with oleic acid (OA) for 20 h prior to LipidTOX red staining. D The workflow of TurboID-mediated proximity labeling and subsequent capture and identification of biotinylated proteins by mass spectrometry (MS). E Detection of biotinylated proteins by Western blotting with streptavidin-HRP. The results shown are representative of three independent experiments. Cells were pre-treated with OA, followed by incubation with either biotin or DMSO. F Identification of proteins that interacted with both TMEM120A and TMEM120B. The log2 of the average MS intensities of proteins biotinylated and enriched in TMEM120A and TMEM120B samples over the NES sample are plotted. Each data point represents the mean MS intensity of three biological replicates. CHP1 calcineurin- like EF-hand protein 1, GPAT4 glycerol-3-phosphate acyltransferase 4. G Representative images of HEK293T cells expressing GFP1-10-TMEM120A with CHP1-GFP11, GPAT4-GFP11, or sarcoplasmic/endoplasmic reticulum Ca2+ ATPase 2 (ATP2A2)-GFP11. The results shown are representative of three independent experiments. The reconstitution of a GFP signal indicates the proximity of TMEM120A with CHP1 and GPAT4 at the ER. The cell nuclei were stained with Hoechst (blue). Scale bar = 10 μm. Because TMEM120A and TMEM120B are ER-resident proteins, we enriched the ER membranes by differential centrifugation prior to streptavidin-mediated capture of biotinylated proteins; these were then identified by mass spectrometry (Fig. 6D). Proteins that were more abundant in either the TurboID-TMEM120A or TurboID-TMEM120B samples when compared with TurboID-NES were first selected. We then focused on a subset of proteins that ranked high in both lists (Fig. 6F). Strikingly, a group of acyltransferases that are known to promote glycerolipid synthesis were among the most highly enriched proteins in our analysis, with GPAT4 as the top hit. We also found CHP1. Because CHP1 and GPAT4 were previously shown to act together23, we focused our subsequent analysis on how TMEM120A engaged with these proteins to promote glycerolipid synthesis. First, we verified the interaction between TMEM120A and GPAT4 or CHP1 in HEK293T cells using a split-GFP assay47. The reconstitution of GFP was facilitated by two interacting proteins, which were fused to either the GFP1-10 or GFP11 fragment. GFP fluorescence in the ER was detected in cells that co-expressed GFP1-10-TMEM120A and CHP1-GFP11 or GPAT4-GFP11 (Fig. 6G). However, no fluorescence was detected when GFP1-10-TMEM120A and the negative control ER-resident ATP2A2-GFP11 were co-expressed (Fig. 6G). Taken together, our results demonstrate that TMEM120A specifically associates with CHP1 and GPAT4 at the ER in vivo. TMEM120A forms a stable complex with CHP1 and GPAT4 We further validated the interaction of TMEM120A with CHP1 and GPAT4 by co-immunoprecipitation experiments. To this end, we expressed HA-epitope tagged GPAT4, CHP1, or GFP in HEK293T cells that stably expressed V5-epitope tagged TMEM120A. We found that GPAT4 efficiently pulled down TMEM120A when they were co-expressed (Fig. 7A, lane 6). This interaction was enhanced ~2-fold when cells were treated with OA (Fig. 7B, compare lanes 5 and 7). In contrast, similar to the GFP negative control, CHP1 was unable to pull down TMEM120A in the absence or presence of OA (Fig. 7A, lanes 7 and 8; Fig. 7B, lanes 6 and 8). However, the expression of CHP1 boosted the ability of GPAT4 to pull down TMEM120A by more than 2-fold (Fig. 7C, lanes 5 and 6). Interestingly, we observed a slight, but reproducible increase in the amount of TMEM120A recovered when CHP1 was used to pull down GPAT4 (Fig. 7C, lanes 7 and 8). Our results suggest that a fraction of TMEM120A forms a stable complex with CHP1 and GPAT4, possibly with GPAT4 as the anchoring protein. Physical association of TMEM120A, GPAT4, and CHP1. A Coimmunoprecipitation of TurboID-V5-TMEM120A with GPAT4-2xHA (but not CHP1-2xHA or GFP-2xHA) in HEK293T cells treated with oleic acid (OA). The lysate from cells that did not express any HA-epitope tagged proteins served as the negative control. Input lanes (1–4) were loaded with 11% of the lysate used for immunoprecipitation. Samples were separated by SDS-PAGE and transferred to a single membrane, which was sectioned and probed with the indicated antibodies. B–D As in (A), but with lysates from cells that were treated with ethanol (EtOH; solvent control) or OA (B); with lysates from OA-treated cells that co-expressed TurboID-V5-TMEM120A with the indicated HA- and/or FLAG-tagged proteins (C); or with lysates from OA-treated cells that co-expressed TurboID-V5-TMEM120A with GPAT4-2xHA and WT or mutant CHP1-FLAG (D). Enrichment ratios were calculated as the signal intensity in eluate lanes divided by that in corresponding input lanes. Data represent the mean of three independent experiments for (A–D). E Relative TMEM120A mRNA levels during OP9 cell differentiation were measured by real-time PCR at the indicated time points. Data are presented as mean ± SD from three independent biological samples. For each biological sample, measurements were made in triplicates, and the mean value was used for statistical analysis. F Schematic diagram illustrating the CRISPR/Cas9-mediated integration of the puromycin R-T2A-3xFLAG-GFP cassette into the 5′ end of the endogenous Tmem120a gene in OP9 cells. G As in (A), but with lysates for immunoprecipitation from undifferentiated (Undiff.) or differentiated (Diff.) TMEM120A knock-in (KI) OP9 cells. WT OP9 cells were included as negative controls. Data are representative of three independent experiments. Statistical analysis: E two-sided one-way ANOVA followed by Dunnett’s multiple comparisons test. It is known that CHP1 needs to be myristoylated for it to activate GPAT423. We therefore asked if such lipid modification of CHP1 is also required for its complex formation with TMEM120A and GPAT4. We found that the myristoylation-defective CHP1(G2A S6A) mutant could not bind GPAT4 and failed to promote TMEM120A-GPAT4 interaction (Fig. 7D, compare lanes 6 and 7). Another loss-of-function mutant, CHP1(G44R), discovered through a C. elegans genetic screen (see next “Section”), partially retained its ability to bind GPAT4, but the recruitment of TMEM120A to the complex was impaired (Fig. 7D, compare lanes 6 and 8). The expression levels of CHP1(G2A S6A) and CHP1(G44R) were comparable to that of the wild type protein (Fig. 7D, lanes 2–4). Thus, we conclude that a key function of CHP1 is to stabilize the GPAT4-CHP1-TMEM120A complex. Finally, to demonstrate the interaction between endogenous TMEM120A and GPAT4, we used OP9 mouse stromal cells that can be differentiated into adipocytes48 with a concomitant increase in TMEM120A expression (Fig. 7E). Using CRISPR, we generated a knock-in (KI) cell line that expressed a 3xFLAG-GFP fusion protein of TMEM120A from the endogenous locus (Fig. 7F). Upon differentiation, a robust up-regulation of the fusion protein, together with endogenous GPAT4, was observed (Fig. 7G, lanes 2 and 4). Importantly, GPAT4 co-immunoprecipitated with TMEM120A from lysates prepared with the differentiated cells (Fig. 7G, lane 8). This indicates that TMEM120A interacts with GPAT4 under physiological conditions. Synergistic activation of GPAT4 by CHP1 and TMEM120A We devised an in vitro acyltransferase assay to assess the functional impact of GPAT4-CHP1-TMEM120A complex formation. Lysates from cells that expressed GPAT4 with or without TMEM120A and/or CHP1 were prepared, and then further processed by differential centrifugation to yield microsomal membrane fractions (Fig. 8A). NBD- palmitoyl-CoA was used as the substrate to measure GPAT4 activity, based on the detection of fluorescent NBD-labeled lysophosphatidic acid (LPA), which is the product of the GPAT4 reaction (Fig. 8A). Under our experimental conditions, GPAT4 alone did not yield a significant amount of NBD-LPA (Fig. 8B, C, lane 3). We reproduced the observation that CHP1 enhanced GPAT4 activity23, and so did TMEM120A (Fig. 8B, C, lanes 4 and 5). Remarkably, co-expression of GPAT4, CHP1, and TMEM120A increased the abundance of NBD-LPA by >40-fold (Fig. 8B, C, lane 6). Since HA-tagged GPAT4 was present at comparable levels in each reaction (Fig. 8D), the synergistic activation of GPAT4 was unlikely due to a boost of its steady state level by CHP1 and TMEM120A. Next, we tested GPAT4(H248A) and GPAT4(R292A) mutants, which are predicted to be catalytically inactive49. These mutants responded poorly to the co-expression of CHP1 and TMEM120A and yielded little NBD-LPA (Fig. 8E, F, compare lanes 3–5), despite being expressed at higher levels (Fig. 8G, lanes 3 and 4). Furthermore, the GPAT4(H248A) and GPAT4(R292A) mutants retained their ability to bind CHP1 and TMEM120A (Fig. 8H, lanes 6–8). Therefore, NBD-LPA was specifically produced by functional GPAT4 in our in vitro assay. Taken together, our results indicate that TMEM120A and CHP1 act synergistically to enhance GPAT4 activity. TMEM120A and CHP1 synergistically enhance GPAT4 activity. A Schematic to show the workflow of the in vitro GPAT4 acyltransferase assay. B Representative TLC analysis of the NBD-labeled GPAT4 acyltransferase products. Microsomal fractions from HEK293T cells transfected with combinations of V5-TMEM120A, GPAT4-2xHA, and CHP1-FLAG were assessed, with 20 μg protein used per assay and the lipids were visualized via NBD fluorescence. Ori origin, LPA lysophosphatidic acid, PA phosphatidic acid. C Quantification of NBD-labeled product intensity from (B). Data are shown as mean ± SD from three independent experiments. The mean value of the GPAT4-only group was assigned as 1, and the fold change relative to this group was plotted. D A representative immunoblot showing the expression levels of V5-TMEM120A, GPAT4-2xHA, and CHP1-FLAG in each of the microsomal fractions. Three independent experiments were performed. E TLC analysis of the enzymatic activity of WT or catalytically inactive GPAT4 mutants (H248A, R292A) when co-expressed with CHP1-FLAG and V5-TMEM120A in HEK293T cells. F Quantification of NBD-labeled products from (E). Data are shown as mean ± SD from three independent experiments. The mean value of the GPAT4(WT) + CHP1-FLAG + V5-TMEM120A group was set to 1, and fold changes relative to this group were calculated and plotted. G As in (D), a representative immunoblot showing the expression levels of epitope-tagged proteins in each of the microsomal fractions. Three independent experiments were performed. H Co-immunoprecipitation of TurboID-V5-TMEM120A with wild type GPAT4-2xHA, catalytically inactive mutants GPAT4(H248A)-2xHA or GPAT4(R292A)-2xHA, and CHP1-FLAG. Following treatment with oleic acid (OA), cell lysates were subjected to immunoprecipitation and analyzed using the same procedure as described in Fig. 7. Data are representative of three independent experiments. Statistical analysis: one-way ANOVA followed by Dunnett’s multiple comparisons test. CHPF-1/CHP1 is required for LD expansion in C. elegans From the genetic screens that led to the discovery of tmem-120, we recovered 4 loss-of-function alleles of chpf-1, which encodes the C. elegans ortholog of CHP1 (Fig. S6A). The glycine residue at position 2 is the conserved site for myristoylation. The recovery of the G2R mutant, therefore, underscored the importance of myristoylation to CHPF-1 function. According to the CHP1 structure and the AlphaFold3 prediction of the CHPF-1 structure50, glutamate and phenylalanine at positions 13 and 35, respectively, reside in α-helices, whereas glycine at position 44 resides in a flexible region (Fig. S6B). Therefore, it is conceivable that missense mutations in these conserved residues may impair the structural integrity of CHPF-1. We used chpf-1(hj47) as the reference allele, since it was predicted to disrupt the myristoylation and hence the function of CHPF-1. We showed that the loss of CHPF-1 function reduced the size and blocked the expansion of LDs in wild type and daf-22 mutant animals (Fig. 9A, B). However, importantly, there was no difference in LD size in daf-22; chpf-1 and daf-22; chpf-1; tmem120 mutant animals, suggesting that chpf-1 and tmem-120 acted in the same genetic pathway. Consistent with a recent report51, the localization of our DHS-3/17-beta-hydroxysteroid dehydrogenase 13 (HSD17B13)-based LD marker was unaffected in chpf-1 mutant animals (Fig. 9A). The chpf-1 mutants were also remarkably similar to tmem-120 mutants in that they were rendered almost completely sterile under HFD (Fig. 9C). They were also resistant to HFD-induced LD expansion (Fig. 9D, E). In summary, our genetic C. elegans CHPF-1 data converge with the biochemical CHP1 data from mammalian cells and provide a strong argument that TMEM-120/TMEM120A and CHPF-1/CHP1 act together to promote glycerolipid synthesis and LD expansion. CHPF-1 acts together with TMEM-120 in C. elegans. A Visualization of LD size in WT, chpf-1(hj47), daf-22(-), daf-22(-); chpf-1(hj47) and daf-22(-); chpf-1(hj47); tmem-120(hj50) L4 animals. DHS-3::mRuby (hj200) was used as the LD marker. The regions outlined by the small white squares were magnified 4× and are shown as insets. Scale bar = 5 μm. Subsequent panels use the same marker and scale unless stated otherwise. B Quantification of the diameter of LDs in the animals from (A). The data show the mean ± SD LD diameter of n = 10 animals per group, and each data point shows the average LD diameter in one animal. The total number of LDs quantified: WT = 1181, chpf-1(hj47) = 1380, daf-22(-) = 871, daf-22(-); chpf-1(hj47) = 1982, and daf-22(-); chpf-1(hj47); tmem-120(hj50) = 2321. C Quantification of brood size of WT and chpf-1(hj47) animals raised on a control diet or HFD. Each data point represents the mean ± SD brood size of each of 10 animals. D Visualization of the LDs in L4 stage WT or chpf-1(hj47) animals fed with a control diet or HFD. Scale bar = 5 μm. E Quantification of LD diameter in animals as in (B). Each data point represents the average LD diameter of each of 10 animals. Total number of LDs quantified: WT (control) = 1546, chpf-1(hj47) (control) = 1612, WT (HFD) = 1461, and chpf-1(hj47) (HFD) = 1953. F Model illustrating the promotion of TAG synthesis by the GPAT4-CHP1-TMEM120A complex at the ER. Created in BioRender. Huang, S. (2026) https://BioRender.com/f57f610. Statistical analyses: B, C, E two-way ANOVA followed by Tukey’s multiple comparisons test. Discussion In this paper, we combined genetic, imaging and biochemical approaches to demonstrate an evolutionarily conserved function of TMEM-120/TMEM120A, when in a complex with GPAT4 and CHP1, to promote glycerolipid synthesis (Fig. 9F). Lipid profiling in C. elegans indicated that one role of TMEM-120 might be to expand the spectrum of fatty acids that can be incorporated into TAG and phospholipids. We propose that such plasticity is important to prevent lipotoxicity upon shifts in the profile of dietary fatty acids, for example, during dieting for weight or metabolic disease management. TMEM120A (alternatively known as NET29 or TACAN) was first reported as a nuclear envelope protein15. Subsequent work suggested a role for TMEM120A in adipogenesis16,52 but, perplexingly, it was also reported to serve as a cell surface mechanosensitive ion channel17. Our examination of TMEM-120 at the endogenous level in live animals infers the ER as its primary site of action (Fig. 5), and we showed that TMEM120A, GPAT4 and CHP1 interact at the ER (Fig. 6G). Thus, our results are consistent with the notion that the ensemble of TAG synthesis proteins can all be found in the ER, which conceivably enables the transfer of biosynthetic intermediates within the membrane. Based on our lipid analysis (Figs. 4 and S4), we propose that TMEM-120/TMEM120A assists in the presentation of medium to long chain acyl-CoA to GPAT4, regardless of their degree of unsaturation. GPAT4, and its paralog GPAT3, catalyze the first step of glycerolipid synthesis from glycerol-3-phosphate (G-3-P) in distinct tissues in mammals9. Besides acting in the ER, where most GPAT activity is detected, GPAT4 has also been shown to re-localize to LDs53. However, it was unclear how cellular GPAT activity was modulated until myristoylated CHP1 was found to activate GPAT423, potentially achieved in part by the mutual stabilization of CHP1 and GPAT4 at a post-transcriptional level23. Additional CHP1 partners have been studied, including Na+/H+ exchangers (NHEs), which modulate cellular pH by coupling the influx of extracellular Na+ and efflux of intracellular H+ at the cell surface54. Structural and biochemical studies suggest that CHP1 associates with NHEs in a conformation-specific manner, and it regulates their maturation and trafficking to the cell surface24,55. Moreover, according to the Human Protein Atlas56, CHP1 is more highly expressed than its clients in general. Taken together, it is tempting to speculate that myristoylated CHP1 gains access to cellular membranes where it engages its various partners in specific conformations. Accordingly, we propose that CHP1 bound GPAT4 adopts a conformation that favors its interaction with TMEM120A (Fig. 7C). For example, CHP1 may favor the ER-residency over LD-residency by GPAT4, thereby enhancing its availability for interaction with the ER-resident TMEM120A (Fig. 9F). The simultaneous engagement of CHP1 and TMEM120A with GPAT4 greatly enhances the latter’s acyltransferase activity (Fig. 8B, C). Our results place TMEM120A at one of the early steps of glycerolipid synthesis, thus explaining the reduction of TAG, DAG, PC and PE in tmem-120 mutant animals (Fig. 3D–G), especially when they are under HFD. The tmem-120 and chpf-1 loss-of-function mutant animals share remarkably similar phenotypes, including their deficiency in promoting LD expansion, and HFD sensitivity (Fig. 9). Importantly, the severity of these phenotypes did not change in tmem-120; chpf-1 compound mutant animals, suggesting that TMEM-120 and CHPF-1, and presumably their mammalian orthologs, indeed act together to promote glycerolipid synthesis. Nevertheless, CHP1 and its paralogs interact with a broader range of proteins than TMEM120A57. Therefore, further analysis of chpf-1 mutant animals may reveal additional physiological functions of CHPF-1/CHP1. In conclusion, our functional studies of the GPAT4-CHP1-TMEM120A complex reveal a hitherto unknown mechanism that overcomes the intrinsic substrate preference of acyltransferases, thus enabling the glycerolipid synthesis pathways to accept a range of fatty acids with comparable efficiency. This can be achieved by an increase in the local concentration of medium and long chain fatty acyl-CoA and their subsequent presentation to GPAT4 by TMEM120A. Pharmacological modulation of the TMEM120 family of proteins should therefore have a direct impact on fatty acid absorption and utilization. Methods Strains and transgenes The wild type C. elegans strain used was Bristol N2. The following alleles were obtained from the Caenorhabditis Genetics Center (CGC), which is funded by NIH Office of Research Infrastructure Programs (P40 OD010440): LG I, EG6701 (ttTi4348); LG II, daf-22(ok693); LG IV, EG6703 (cxTi10816); and LG V, sid-1(qt78). For simplicity, daf-22(ok693) is referred to as daf-22(-) in the figures. The dagt-2(hj44), tmem-120(hj49), tmem-120(hj50), and chpf-1(hj47) alleles were obtained by ethyl methane sulfonate (EMS) mediated mutagenesis and confirmed by sequencing. The following transgenes or CRISPR-generated alleles were used: hjSi158[vha-6p::sel-1(a.a.1-79)::mCherry::HDEL::let-858 3′UTR] I hjSi493[vha-6p::sid-1 exons 1-3::sid-1 cDNA exons 4-7::dhs-28 3′UTR] IV hjSi524[vha-6p::vhhgfp4::zif-1::let-858 3′UTR] I hjSi528[vha-6p::sel-1(a.a.1-79)::GFP::HDEL::tbb-2 3′UTR] IV hjSi557[vha-6p::gfp::3xFLAG::tmem-120(codon modified) cDNA::dhs-28 3′-UTR] IV dhs-3(hj200 [dhs-3::mRuby]) I tmem-120(hj239 [tmem-120(a.a.1-368)::gfp::3xFLAG::tmem-120(a.a.369 to stop)]) III tmem-120(hj266 [tmem-120::gfp-3xFLAG::C-term with loxP in intron 2 and intron 4]) III tmem-120(hj270 [GFP::3xFLAG::tmem-120]) III tmem-120(hj297 [GFP::3xFLAG::TMEM-120(G195E)]) III In addition, the following extrachromosomal array strains were used: hjEx29[vha-6p::tmem-120 cDNA::sl2::gfp::let-858 3′UTR] hjEx30[vha-6p::mouse tmem120a cDNA::sl2::gfp::let-858 3′UTR] All the strains were outcrossed with the wild type (WT) N2 at least twice before further characterization, and all the experimental animals were maintained at 20 °C. Cell culture COS7 cells (ATCC-CRL-1651) and HEK293T cells (a gift from Tuan Anh Nguyen, HKUST) were maintained in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and 1% antibiotic-antimycotic (all from Life Technologies). OP9 cells (CRL-2749) were maintained in MEM-α supplemented with 20% FBS and 1% antibiotic-antimycotic (all from Life Technologies). For adipogenic differentiation48, OP9 cells were seeded at approximately 70% confluence in six-well plates. Before reaching full confluence, the growth medium was replaced with insulin-oleate differentiation medium (MEM-α, 0.2% FBS, 175 nM insulin, and 900 μM oleate bound to albumin (5.5:1, molar ratio), containing 1% antibiotic-antimycotic) for 3 days. All the cell lines were maintained in a humidified incubator at 37 °C with 5% CO2. Genetic screens The tmem-120 and chpf-1 alleles were isolated in a genetic screen for suppressors of the expanded LD phenotype in daf-22(ok693) mutants. Complementation tests indicated that hj49 and hj50 belong to the same complementation group. Using an SNP-based mapping strategy with the Hawaiian C. elegans isolate CB485658, we mapped hj50 to LGIII. Mutant animals were then subjected to whole genome sequencing (Illumina). Molecular lesions observed in hj49 and hj50 in tmem-120 were confirmed by Sanger sequencing. The daf-22(ok693); tmem-120(hj50) mutant was rescued by a vha-6p::tmem-120::sl2::gfp transgene (Fig. 1E). We employed the same molecular cloning strategy for chpf-1 as we did for tmem-120. Fluorescence imaging of C. elegans and mammalian cells Fluorescence imaging of live animals and mammalian cells was performed as previously described38. In brief, intact animals were mounted on 8% agarose pads with Polybead microspheres (Polysciences) prepared with 1× phosphate-buffered saline (PBS) and 0.025% sodium dodecyl sulphate (SDS). Fluorescence images of animals at the stages indicated were acquired with an Axio Observer Z1 (Carl Zeiss) spinning disk confocal microscope equipped with a piezo Z stage using either a 63×, NA 1.4 or a 100×, NA 1.46 objective lens. Images were acquired using a Neo sCMOS camera (Andor), which was controlled by iQ3 software (Andor). The following excitation and emission wavelengths, respectively, were used: GFP, 488 nm and 500–550 nm; mRuby/mCherry, 561 nm and 580.5–653.5 nm; autofluorescence (from lysosome-related organelles), 488 nm and 580.5–653.5 nm; BFP, 405 and 437 nm; and FAS LD dye, 405 and 617 nm. Optical sections were acquired at 0.5-µm intervals, and z stacks of 8–10 µm were exported from iQ3 to Imaris 8 (Bitplane) for processing. At least three imaging sessions were conducted on independent biological samples on different days. To quantify the LD number and size, all LDs present in the second intestinal segment of animals were considered. The size of each LD was measured by manual fitting with the spot function in Imaris. LDs with a diameter < 0.5 μm (i.e., below the resolution limit) were excluded, after which the datasets were further processed in Excel, and graphs were plotted in GraphPad Prism. Live imaging by stimulated Raman scattering To measure endogenous neutral lipids, stimulated Raman scattering (SRS) was conducted as described34. In brief, live animals were mounted on an 8% agarose pad and immersed in 1× PBS containing 0.2 mM levamisole. Signals from LDs located in a central focal plane of the first and second intestinal segments of young adult animals were collected using a 40×/1.15 NA water immersion objective lens (UAPO40XW3/340, Olympus). To measure the neutral lipid content of whole animals, signals were collected at the focal plane that exhibited the maximal SRS intensity, as previously described34, with a 20×/0.8 NA air objective (Plan-Apochromat, Zeiss). The vibrational frequencies of the C–H bonds were detected at 2863.5 cm−1 and quantified as previously described34. Fatty acid absorption and lipolysis in live animals was also measured by SRS, with the Olympus 40×/1.15 NA water immersion objective lens and method described above34. Saturated bacterial cultures of E. coli OP50 were mixed with 4 mM deuterium labeled PA-d31 or OA-d34 (Sigma) and then seeded onto nematode growth medium (NGM) plates. To measure fatty acid uptake, young adults that had been raised in the absence of deuterium labeled fatty acids, were transferred to plates containing PA-d31 or OA-d34. In addition, to measure lipolysis, L1 larvae were raised on plates containing PA-d31 or OA-d34 and then transferred to E. coli OP50 seeded plates without deuterium labeled fatty acids when they were young adults. In both cases, the animals were imaged by SRS at 5–6-h intervals for 28–30 h, and the C–D bond vibration was detected at 2116.8 cm−1. Images were imported into ImageJ59 and then processed for quantification in MATLAB_R2015a. Real-time quantitative PCR for C. elegans samples Each experimental sample comprised ~400 animals (either WT or tmem-120(hj50)) synchronized at the L1 larval stage. Animals were harvested at the L4 stage (i.e., 43 h after L1 for WT, and 46 h after L1 for tmem-120(hj50)), and the total RNA was extracted with a Total RNA mini kit (Geneaid), after which 500 ng total RNA was reverse transcribed with a PrimeScript™ 1st strand cDNA synthesis kit (TaKaRa, 6110 A). The real-time PCR was carried out using a Roche LightCycler system with SYBR Green Master Mix (Roche, 04913850001). For each strain, technical triplicates were performed for each biological sample and the raw Ct values were analyzed using the delta-delta Ct (ΔΔCt) method. Primer sequences are listed in Supplementary Table 1. Real-time quantitative PCR for OP9 cells OP9 cells were harvested at different times during differentiation, after which RT-qPCR was performed following the protocol described above for C. elegans. Primer sequences are listed in Supplementary Table 2. Western blotting C. elegans samples: Animals were synchronized at the L1 larval stage (~3000 animals per sample) and harvested at the L4 stage (i.e., 43 h after L1 for WT and tmem120(hj270), and 46 h after L1 for tmem-120(hj297)). After harvesting, 2× SDS sample buffer (160 mM Tris-HCl, 8% glycerol, 0.4% SDS, 0.01% bromophenol blue, 400 mM urea,10% β-mercaptoethanol) was added to the worm pellets. Mammalian cell samples: For experiments with mammalian cells, lysis was performed in RIPA buffer (50 mM Tris, 150 mM NaCl, 0.1% (wt/vol) SDS, 0.5% (wt/vol) sodium deoxycholate, 1% (vol/vol) Triton X-100, pH 7.5) or MAPK buffer (10 mM Tris-HCl, pH 8.0, 50 mM NaCl, 1 mM EDTA, 1% IGEPAL CA-630 (Sigma, I3021)) and centrifuged at 13,000 × g for 10 min at 4 °C. The protein concentration of the supernatants was measured with a bicinchoninic acid (BCA) protein assay kit (Thermo Scientific, 23227). Each sample, containing the same amount of protein, was mixed with 1× reducing SDS sample buffer (10 mM dithiothreitol (DTT), 62.5 mM Tris base, 10% glycerol, and 0.01% bromophenol blue). Common processing steps: The worm and cell lysate samples in SDS sample buffer were boiled at 70 °C for 10 min, after which they were sonicated in an ice-cold water bath (at intervals of 2 min on and 30 s off) for a total of 7 min, before being subjected to SDS-PAGE. The separated proteins were then transferred to a nitrocellulose membrane and blocked with 5% fat-free milk (Bio Rad) at room temperature (~22 °C) for 1 h before being incubated with primary and HRP-conjugated secondary antibodies. The proteins of interest were visualized by treating the membrane with the ClarityTM ECL substrates kit (Bio Rad), after which images were acquired and quantified with Image Lab (Bio Rad). Signals from the total amount of protein in each lane were used for the normalization of sample loading. Pharyngeal pumping rate The pharyngeal pumping rate of L4 stage animals was measured using bright-field videos acquired with an Olympus SZX16 stereo microscope. One day before imaging, the animals were transferred to a newly seeded NGM plate and left beside the microscope for acclimatization at room temperature. For each experimental group, up to 10 animals were prepared, and a 2-min video focusing on the pharynx of each worm was captured. Each video was trimmed to a 60-s clip and played back at 0.3× speed. The number of pharyngeal contractions was quantified visually. Growth rate measurements To evaluate the growth rate of animals, NGM plates seeded with E. coli OP50, with or without 2% glucose, were prepared 2 days prior to the addition of larvae. Approximately 50 synchronized, newly hatched L1 larvae were aliquoted onto each plate, and after 48 h, the number of animals at the different developmental stages was quantified. The data from 4 plates per strain per diet were combined, and the percentage of animals at each stage was calculated and plotted. Two independent experiments were performed. Glucose supplementation for high fat diet A high fat diet was provided to animals via glucose supplementation of NGM plates35. A stock solution was prepared with autoclaved MilliQ water containing 40% D-(+)- glucose (Sigma, G7021) and sterilized by filtration through a 0.22 µm filter. This glucose stock solution was added to freshly autoclaved NGM agar solution that had cooled to ~ 50 °C at 2% final concentration, before being poured into 6-cm plates. Control NGM plates were prepared by substituting the glucose stock solution with the same volume of MilliQ water alone. The plates were then dried in a laminar flow hood for 3 h, after which they were seeded with a fresh overnight culture of E. coli OP50 and left at room temperature for ~ 48 h before use. Five P0 animals at the L4 larval stage were picked onto each plate to generate offspring. For analysis, 10 F1 animals at the L4 stage per group were imaged, and three independent experiments were performed. Brood size quantification For quantification of brood size, five WT, chpf-1(hj47) or tmem-120(hj50) P0 animals at the L4 larval stage were picked onto NGM plates in the absence or presence of 2% glucose, prepared following the procedure described above. Individual F1 animals at the L4 larval stage were transferred onto new NGM plates in the absence or presence of 2% glucose, with 12 worms per group. F2 animals that had hatched were counted and removed from the plates at or before the L4 larval stage for the next few days, until no F2 generation animals remained. Fluorescence recovery after photobleaching Fluorescence recovery after photobleaching (FRAP) was optimized, based on a previously published protocol60. FRAP was conducted on the Zeiss Axio Observer Z1 spinning disk confocal microscope described above, using the 63×/NA 1.4 objective lens, Neo sCMOS camera and iQ3 software, described above. A FRAPPA calibration was conducted every time prior to use. WT and tmem-120(hj50) animals, both expressing the intestinal cell surface marker (hjSi601, designed according to ref. 39), were raised for at least three generations on NGM plates seeded with a fresh overnight culture of E. coli OP50. Three WT or tmem-120(hj50) animals at the L4 larval stage were then picked onto NGM plates with or without 2% glucose. Approximately 15 offspring at the L4 larval stage were mounted on 8% agarose plates in 2 μl 20 mM levamisole. Standardized (15-pixel diameter) regions on the intestinal cell surface were chosen for photobleaching, with no more than two regions used for each animal, and at least 10 animals were imaged in each group. Ten images were captured before photobleaching, after which a region of interest (ROI) was bleached twice, each time using a laser power of 20% and a duration of 60 μs. The recovery of fluorescence was then recorded by imaging at maximal speed for 200 frames. The whole process took approximately 1 min 10 s. The images were analyzed with the “FRAP analysis” function in iQ3. Images in which the worms were moving, or the ROI was out of focus, were discarded. For FRAP analysis, apart from the ROI that was bleached, the fluorescence signals in another circular region (50-pixel diameter) were used as a reference for pre-bleach correction. Four FRAP periods (i.e., pre-bleach, bleach, recovery and steady state) were defined manually. A corrected curve and recovery fitting curve were generated, and the half time and diffusion coefficient were calculated according to Axelrod et al.61. The data were exported and further processed in Excel (Microsoft), and graphs were plotted using GraphPad Prism. Representative images were processed using Imaris 8 (Bitplane; Oxford Instruments) and exported as videos from ImageJ. Experiments were repeated three times. Lipid analysis Two lipidomic approaches were employed, namely relative and absolute lipidomics. For both experimental strategies, synchronized L1 larvae were used: ~3000 larvae per sample for relative lipidomics and ~6000 larvae per sample for absolute lipidomics. Animals were cultured on 10-cm NGM plates ± 2% glucose such that 20× concentrated overnight cultures of E. coli OP50 were used for seeding the plates. Animals were harvested at the L4 stage as follows: With a normal diet, 43 h after L1 for WT animals and 46 h after L1 for tmem-120(hj50) animals. With a diet supplemented with 2% glucose, 50 h after L1 for WT animals and 59 h after L1 for tmem-120(hj50) animals. All the animals were washed with detergent-free PBS at least three times and then transferred into organic solvent resistant Eppendorf tubes, after which they were frozen in liquid nitrogen and stored at −80 °C. For relative lipidomic analysis, lipid extraction was conducted using methyl-tert-butyl ether (MTBE) as described62,63, with some modifications. In brief, the samples were thawed on ice, and 250 μl methanol (precooled to −20 °C; RDH, for HPLC) and 875 µl MTBE (VWR, for HPLC) were added. The animals were lysed in an ice-cold ultrasonic bath with cycles of 2 min sonication followed by 30 s rest, repeated for a total of 7 min. Phase separation was induced by the addition of 210 µl water, and the samples were sonicated for a further 15 min, keeping the MTBE:methanol:H2O ratio to 10:3:2.5 (vol/vol/vol). They were then centrifuged at 16,100 × g for 15 min at 4 °C, after which the upper organic phase was transferred into a glass vial. MTBE (325 µl) was added to the lower phase, and the sample was centrifuged once again at 16,100 × g for 15 min at 4 °C to re-extract the lipids. The upper organic phase was combined with those from previously collected samples and then dried under a stream of nitrogen at room temperature. For the untargeted analysis of lipid profiles, the dried extracts were dissolved in acetonitrile (RCI Labscan, for HPLC), isopropanol (RNH) and water at a 65:30:5 (vol/vol/vol) ratio and stored at −80 °C. Lipid analysis was performed as previously described29. A 100-µl aliquot of extract was analyzed using the Bruker Elute UPLC system with three independent biological samples and two technical injections per sample. The mass spectrometry data were analyzed using MetaboScape (ver5.0; Bruker), annotated with the following spectral libraries: MSDIAL-Tandem Mass Spectral Atlas-VS68-pos and MSDIAL-Tandem Mass 736 Spectral Atlas-VS68-neg. The intensity was normalized using the probabilistic quotient normalization method64, and datasets were exported to Excel for further processing. A principal component analysis (PCA) was conducted using ClustVis65. Before quantitative analysis, lipid species with variations > 10-fold between the technical replicates were manually excluded. Then, the mean value of the technical replicates of each lipid species was calculated for each biological sample. Only lipids detected in all three biological samples were retained for further analysis. For the targeted analysis of sphingomyelin, the dried extracts were dissolved and injected for analysis as previously described66 with three independent biological samples and one technical injection per sample. The intensity was normalized to the protein content using the BCA assay. The data were exported to Excel for further processing. For the absolute lipidomic analysis, samples were thawed on ice, mixed with 200 μL water and 20 μL internal lipid standard mixture (SPLASH™ LIPIDOMIX™, Avanti, 330707), followed by vortexing. Subsequently, 800 μL MTBE was added, and the mixture was vortexed for 30 s. Then, 240 μL pre-cooled methanol was added, and the mixture was vortexed again. The mixture was subjected to ultrasonic extraction in an ice-water bath for 20 min and then allowed to stand at room temperature for 30 min. Centrifugation was performed at 14,000 × g for 15 min at 10 °C, after which the upper organic phase was collected and dried under a stream of nitrogen. For mass spectrometry analysis, this residue was reconstituted with 200 μL 90% isopropanol/acetonitrile solution, vortexed thoroughly, and 90 μL of the reconstituted solution was centrifuged at 14,000 × g for 15 min at 10 °C. The supernatant was then collected for analysis by LC-MS. Pooled quality control (QC) samples were prepared by combining equal amounts (i.e., 10 μL) of each extraction mixture. The samples were separated using a UHPLC Nexera LC-30A ultra-high-performance liquid chromatography system (SHIMADZU) equipped with a C18 column (Waters, ACQUITY UPLC CSH C18, 1.7 µm, 2.1 mm × 100 mm column). The column temperature was maintained at 45 °C, with a flow rate of 300 μL/min. The mobile phase consisted of (A) acetonitrile-water (6:4, v/v) containing 0.1% formic acid and 0.1 mM ammonium formate, and (B) acetonitrile-isopropanol (1:9, v/v) containing 0.1% formic acid and 0.1 mM ammonium formate. The gradient elution program was conducted as follows: 0–2 min, (B) maintained at 30%; 2–25 min, (B) linearly increased from 30 to 100%; 25–35 min, (B) returned to and held at 30%. Throughout the analysis, samples were stored in an autosampler at 10 °C. To minimize the impact of instrument signal fluctuations, the samples were analyzed consecutively in a random order. Mass spectrometric detection was performed in both electrospray ionization (ESI) positive and negative ion modes. After UHPLC separation, the samples were analyzed with a Q Exactive series mass spectrometer (Thermo Scientific). The ESI source parameters were as follows: heater temperature, 300 °C; sheath gas flow rate, 45 arbitrary units (AU); auxiliary gas flow rate, 15 AU; sweep gas flow rate, 1 AU; spray voltage 3.0 kV; capillary temperature, 350 °C; S-Lens RF level, 50%; and MS1 scan range, 200–1800 m/z. The mass-to-charge ratios of the lipid molecules and fragments were acquired by the following method: after each full scan (MS1), ten MS2 fragmentation spectra (HCD) were recorded. For MS1 and MS2, the resolution was set to 70,000 at m/z 200 and 17,500 at m/z 200, respectively. The LipidSearch software was used for peak identification, peak extraction, and lipid identification (including MS/MS-based identification) of the lipid molecules and internal standard lipids. The main parameters were as follows: precursor tolerance, 5 ppm; product tolerance, 5 ppm; and product ion threshold, 5%. Generation of COS7 cells overexpressing TMEM120A The Sleeping Beauty transposon system was used to generate COS7 cells that overexpressed human TMEM120A. Cells were co-transfected with the Sleeping Beauty transposon plasmid pSBi-Hyg-BFP (hTMEM120A cDNA), and the Sleeping Beauty transposase plasmid pCMV (CAT) T7-SB100, with Lipofectamine2000 (Life Technologies). Three days after transfection, cells were maintained in selection medium (DΜΕΜ growth medium containing 400 μg/mL hygromycin) for at least 7 days. Cells that survived drug selection were sorted using the Aria III system (Becton Dickinson) and sub-divided into “low”, “medium”, and “high” populations based on BFP fluorescence intensity (a surrogate of TMEM120A expression). The “high” cell population was used in subsequent experiments. Generation of HEK293T cells expressing TurboID fusion proteins The transgenes expressing TurboID fusion proteins were integrated into the AAVS1 safe harbor locus using CRISPR/Cas9 homology-directed repair (HDR)41. Repair template plasmids expressing TurboID fusion proteins for HDR were constructed by PCR amplification and Gibson assembly based on the pSH-EFIRES-B-Seipin-miniIAA7-mEGFP plasmid (a gift from Elina Ikonen, University of Helsinki; Addgene plasmid # 129719)67 and C1(1-29)-TurboID-V5 (a gift from Alice Ting, Stanford University; Addgene plasmid # 107173)68. The following plasmids were used: pSH-HygroR_P2A_BFP_RPBSA_EF1_TurboID-V5-TMEM120A pSH-HygroR_P2A_BFP_RPBSA_EF1_TurboID-V5-TMEM120B pSH-HygroR_P2A_BFP_RPBSA_EF1_TurboID-V5-NES HEK293T cells were co-transfected with the repair template plasmid and pCas9-sgAAVS1-1 plasmid (a gift from Elina Ikonen; Addgene plasmid # 129726), at a 6:4 ratio in a 12-well plate. Two days post-transfection, the cells were transferred to a 10-cm plate to allow the formation of single-cell colonies in medium containing 100 μg/mL hygromycin. Single-cell colonies were picked and cultured in a 96-well plate. Individual clones were then evaluated by FACS for their response to OA (NU-CHEK PREP, U-46-A), as described below. Generation of OP9 cells with endogenous knock-in of TMEM120A To generate OP9 cells expressing endogenous TMEM120A tagged with 3xFLAG-GFP, the coding sequence of a puromycin R-T2A-3xFLAG-GFP fusion cassette was integrated into the 5′ region of the endogenous Tmem120a locus via CRISPR/Cas9-mediated HDR. A repair template plasmid containing the fusion cassette and a PX458 plasmid encoding the sgRNA targeting murine Tmem120a (5′-gctctccgcggccagtggcg-3′) were co-transfected into OP9 cells using Lipofectamine 2000. Three days post-transfection, the cells were subjected to selection with MEM-α growth medium containing 3 μg/mL puromycin and maintained under selection for successfully edited cells. For adipogenic differentiation, OP9 cells were seeded at approximately 70% confluence in six-well plates. Before reaching full confluence, the growth medium was replaced with the insulin-oleate differentiation medium mentioned above. Cells were maintained in this differentiation medium for 3 days prior to co-immunoprecipitation (CoIP). Fatty acid supplementation Fatty acid supplementation was performed according to a previously published method38,69. In brief, the cell culture medium was pre-heated to 60 °C for 5 min prior to the addition of fatty acids (400 µM OA or palmitic acid (NU-CHEK PREP, N-16-A)). The medium was then equilibrated to 37 °C before use. To induce LD expansion, cells were incubated with 400 μM fatty acids for 20 h. To visualize the LDs, the cells were stained with 10 μM FAS dye38 for 15 min prior to imaging. DGAT1 inhibitor and fatty acid treatment The cell culture medium was pre-heated to 60 °C for 5 min prior to the addition of fatty acids (400 μM OA or 400 μM palmitic acid) or ethanol (as a control), with DMSO containing one of the following DGAT1 inhibitors: A922500 or PF04620110 (all R&D system) or alone (control). The medium was equilibrated to 37 °C before use. The cells were treated with OA for 20 h or with palmitic acid for 60 h, after which they were stained with crystal violet. Quantification of crystal violet was performed according to a published protocol70. Lipid extraction and thin-layer chromatography (TLC) COS7 cells at 80–90% confluency in 6‑well plates were treated with 400 μM OA together with either 50 nM DGAT1 inhibitor (A922500) or DMSO for 3.5 h, washed with DPBS, and harvested by scraping. Lipids were extracted using MTBE:methanol:water (600:180:150 μL), vortexed, and centrifuged (13,200 × g, 15 min, 4 °C). The upper organic phase was collected, dried under nitrogen, and resuspended in 20 μL isopropanol. The lower aqueous phase was processed for protein normalization: proteins were pelleted, resuspended in 400 μL lysis buffer (0.3 N NaOH, 0.1% SDS), clarified by rotation, and quantified with the BCA assay. Lipids equivalent to 100 μg total protein were resolved by TLC using hexane:diethyl ether:acetic acid (75:25:2, v/v/v) as the solvent system. Standards (3 μg triolein (Sigma, T7140) and 5 μg OA) defined the TAG and FFA positions, respectively. Plates were sprayed with 10% CuSO₄ in 8% H₃PO₄ and visualized by ChemiDoc colorimetric imaging after heating (130 °C, 10–15 min). Seahorse measurement of COS7 cells The Seahorse experiment was conducted using the Seahorse XFp FluxPak (Agilent, 103022-100), which contains XFp sensor cartridges, XFp Cell Culture Miniplates, and XF Calibrant Solution. COS7 cells were seeded into an 8-well XFp cell culture miniplate at a density of 24,000 cells per well one day prior to the assay. The outer wells (top and bottom rows) were filled with an equal volume of culture medium to serve as background controls. On the day of the experiment, cells were treated with 50 nM DGAT1 inhibitor and 400 μM OA for 3.5 h. Following treatment, the culture medium was replaced with XF DMEM (pH 7.4, Agilent, 103680-100) supplemented with 10 mM glucose, 1 mM pyruvate, and 2 mM glutamine. The XFp sensor cartridge was calibrated with XF calibrant, according to the manufacturer’s instructions and then loaded with oligomycin (Aladdin, O111756), FCCP (Aladdin, C423663), and rotenone (Aladdin, R426175)/antimycin A (Sigma, A8674) diluted in XF DMEM. During the assay, cells were sequentially treated with oligomycin (final concentration 1.5 μM), FCCP (final concentration 2 μM), and rotenone/antimycin A (final concentration 0.5 μM). Oxygen consumption rate (OCR) was measured using a Seahorse XFp Analyzer (Agilent) following the manufacturer’s instructions. Basal and maximal respiration rates were subsequently calculated from the raw OCR traces using the Agilent online analysis platform. Lipid quantification by FACS The accumulation of cellular lipid was quantified according to a published method71. In brief, a 6-well plate was coated with 50 μg/ml poly-D-lysine (Sigma, P9155). HEK293T cells expressing TurboID fusion proteins were then seeded at 60%-70% confluency and treated with OA (400 μM) for 20 h. Subsequently, they were washed with Dulbecco’s phosphate-buffered saline (DPBS) and stained with HCS LipidTOX™ Red (Life Technologies, H34476) for 15 min at 37 °C. After staining, the cells were detached using trypsin (Life Technologies, 15090046) and harvested for subsequent FACS analysis using an AriaIII or AriaIIIu (BD) with excitation and emission wavelengths of 561 and 610 nm, respectively. Biotinylation and cell fractionation ER membranes were enriched using biotinylation and cell fractionation, as previously described41,72. In brief, HEK293T cells were seeded in 15-cm dishes and treated with OA (400 μM) for 20 h, after which biotinylation was initiated by incubation with 50 μM biotin for 10 min at 37 °C. The cells were then washed with ice-cold DPBS before harvesting. The cell pellets were resuspended in 2 ml IB-1 buffer (225 mM mannitol, 75 mM sucrose, 30 mM Tris-HCl, pH 7.4, 0.1 mM EGTA, and protease inhibitors (Roche, 04693159001)) and lysed by sonication using a Q125 sonicator with probe (QSONICA) at 100% amplitude for 30 s. The lysates were centrifuged at 600 × g for 5 min to eliminate intact cells, debris, and the nuclear fraction. The supernatants were then centrifuged at 7000 × g for 20 min to separate the mitochondrial fraction, after which the supernatants were centrifuged at 20,000 × g for 30 min. The resulting pellet was resuspended in IB-2 buffer (225 mM mannitol, 75 mM sucrose, and 30 mM Tris-HCl, pH 7.4) and then subjected to an additional round of centrifugation at 20,000 × g for 30 min to eliminate any remaining contaminants. The final pellet, representing the isolated ER fraction, was resuspended in 1 ml RIPA buffer supplemented with protease inhibitors. The protein concentration was determined using the BCA assay, and an equal amount of protein from each biological sample was used for further processing and analysis. Sample preparation for mass spectrometry The protein samples were subjected to affinity binding using streptavidin beads, as previously described41. The purified ER fractions were incubated with streptavidin beads overnight at 4 °C. The beads were collected with a magnetic rack, and a series of wash steps were performed using RIPA lysis buffer (for 1 min), 1 M KCl (for 1 min), 0.1 M Na2CO3 (for 10 s), 10 mM Tris-HCl (pH 8.0) containing 2 M urea (for 10 s) and RIPA lysis buffer (for 1 min). The proteins bound to the streptavidin beads were washed, first with 50 mM Tris-HCl (pH 7.5) and then with 50 mM Tris HCl (pH 7.5) containing 2 M urea. To elute the enriched proteins, the beads were incubated with 50 mM Tris-HCl (pH 7.5) containing 1 M urea, 1 mM DTT and 0.5 µg trypsin (ThermoFisher, 90058) at 37 °C for 1 h, after which they were washed with 50 mM Tris-HCl (pH 7.5) containing 1 M urea alone. The initial eluates and washes were combined, and disulfide bonds were reduced by incubation with DTT (to a final concentration of 4 mM) at 37 °C for 30 min. The protein samples were then alkylated with 10 mM iodoacetamide and incubated at 37 °C for 45 min. To generate small peptides for MS analysis, the samples were treated with 0.5 µg trypsin at 37 °C overnight. To desalt the peptides, a C18 desalting column (ThermoFisher, 89870) was employed, following the manufacturer’s protocol, and the desalted samples were dried in a vacuum evaporator (Thermo, SAVANT ISS110). Liquid chromatography-tandem mass spectrometry (LC-MS/MS) The dried peptides were reconstituted in MilliQ water containing 0.1% formic acid before being analyzed using a timsTOF Pro mass spectrometer coupled with a nanoElute system (Bruker) using the default dia-PASEF-short built-in method with 1400 V spray voltage. Peptides were separated using an analytical column (25 cm × 0.075 µm, 1.7 µm, IonOpticks Aurora Ultimate CSI) at 50 °C with a flow rate of 300 nL/min using a 32-min gradient. Details of the gradient are as follows: from 2% to 5% B in 0.5 min, from 5% to 30% B in 26.5 min, from 30% to 95% B in 0.5 min, in 95% B for 0.5 min, from 95% to 2% B in 0.1 min, and in 2% B for 3.9 min (A = 0.1% formic acid in H2O, B = 0.1% formic acid in acetonitrile). Mass spectrometry data analysis A total of nine independent biological samples were analyzed, three for each cell line. All mass spectra were analyzed with the Spectronaut software (Biognosys AG, ver 18). MS/MS spectra were searched against the Homo sapiens Uniprot protein sequence database (downloaded in March 2023). The default setting of the Direct-DIA mode was used, except for a change in the minimal peptide length to 6. Enzyme specificity was set to Trypsin with a maximum of 2 missed cleavages. Carbamidomethyl (C) was set as a fixed modification, while Oxidation (M) and Acetyl (Protein N-term) were set as variable modifications. Statistical significance was determined using a two-tailed Student’s t test, and the resulting p-values were adjusted for multiple hypothesis testing using the Benjamini–Hochberg method. The enrichment ratio was calculated as the fold change of TurboID-TMEM120A/B intensity over the TurboID-NES control, expressed as log2 (Fold Change). Candidates for downstream validation were selected based on ranking by enrichment magnitude and statistical significance. Co-immunoprecipitation Immunoprecipitation was performed as previously described27. HEK293T cells that stably expressed TurboID-V5-tagged TMEM120A were cultured in 6-well plates and transfected with the following plasmids: pcDNA5FRT_CHP1::2xHA pcDNA5FRT_GPAT4::2xHA pcDNA5FRT_Venus::2xHA pcDNA3.1_CHP1::FLAG pcDNA3.1_GPAT4::FLAG pcDNA3.1_Venus::FLAG Cells in each well were transfected with a total of 2 µg plasmid DNA using Lipofectamine 2000 (Life Technologies). After incubation for 6–8 h, fresh medium was added with either 400 μM OA or ethanol (control). The cells were harvested after 20 h, and whole-cell lysates were prepared using MAPK buffer and protease inhibitors. Immunoprecipitation was performed by incubating the lysates with either anti-HA-conjugated magnetic beads (Thermo Scientific, 88837) or anti-FLAG-conjugated magnetic beads (Sigma, M8823) for 30 min at room temperature. For co-immunoprecipitation (CoIP) of endogenous proteins, both undifferentiated and differentiated OP9 cells, including knock-in lines, were lysed using the same MAPK buffer. The anti-HA and anti-FLAG beads were washed three times with TBST and PBS, respectively, after which the protein complexes were eluted by boiling in 1× SDS reducing buffer prior to SDS-PAGE and Western blotting analysis. In vitro acyltransferase assay A GPAT4 assay was performed according to a previously published method with modifications73. HEK293T cells were cultured in 6-well plates and transfected with the following plasmids at a ratio of TMEM120A:GPAT4:CHP1 = 1:4.5:4.5: pcDNA5FRT_GPAT4::2xHA pcDNA5FRT_GPAT4 (H248A)::2xHA pcDNA5FRT_GPAT4 (R292A)::2xHA pcDNA5FRT_V5::TMEM120A pcDNA3.1_CHP1::FLAG The cells in each well were transfected with a total of 2 µg plasmid DNA using PEI Max reagent (Polysciences, 24765-100). After incubation for 48 h, cells were washed twice with DPBS and then homogenized by sonication (10 s on, 5 s off, for a total of 30 s at 100% power) in 0.5 mL buffer A (20 mM Tris-HCl, pH 7.5, 1 mM EDTA, 250 mM sucrose) supplemented with protease inhibitors, as described above. Homogenates were centrifuged at 600 × g for 5 min to remove large debris and nuclei, and the resulting supernatant was further centrifuged at 10,300 × g for 10 min to remove the mitochondrial fraction. The microsomal fraction was isolated by ultracentrifugation of the supernatant at 100,000 × g for 1 h, and the resulting pellet was resuspended in buffer B (20 mM Tris-HCl, pH 7.5, 1 mM EDTA, 250 mM sucrose, 150 mM NaCl, 0.2% Tween 20) by gentle pipetting. The protein concentration was determined using the BCA assay, and for the subsequent GPAT4 enzymatic assay and Western blot analysis, 20 µg microsomal protein was used. The GPAT4 enzymatic assay was performed in a total volume of 100 µL containing 75 mM Tris-HCl (pH 7.5), 4 mM MgCl2, 1 mM DTT, 4 mM NaF, 50 µg/mL fatty acid-free BSA, 80 µM glycerol-3-phosphate (G3P) (Macklin, G881258), and 20 µM NBD-palmitoyl-CoA (Avanti, 810705). Reactions were incubated in the dark at 25 °C for 30 min with occasional shaking. The reaction was terminated by the addition of 180 µL methanol, followed by 600 µL MTBE, and the mixture was vortexed, after which 50 µL water was added, and the mixture was vortexed again. The samples were then centrifuged at 13,200 × g for 15 min at 4 °C to separate the aqueous and organic phases. The upper organic phase was collected, dried under a stream of nitrogen gas, and the lipids were resuspended in 30 µL chloroform/methanol (80:20, v/v). Thin-layer chromatography (TLC) was developed using a solvent system of chloroform/methanol/water (65:25:4, v/v/v). The TLC chamber was equilibrated with the mobile phase vapor for 15 min before plate development. Samples (10 µL each) were spotted using a metal pipette and allowed to dry for 5 min before running the plate. After development, the TLC plate (Whatman) was dried for 20 min in the dark, and the fluorescent lipid signals were visualized and imaged using a ChemiDoc Go imaging system in GelGreen mode with the blue tray setting. Antibodies The following primary antibodies were used for the Western blotting and immunoprecipitation experiments: rabbit anti-V5 (ProteinTech 14440-1-AP, Lot # 00169983), rabbit anti-HA (ProteinTech 51064-2-AP, Lot # 00179344), rabbit anti-FLAG (ProteinTech 20543-1-AP, Lot # 00195106), mouse anti-FLAG M2 antibody (Sigma F1804, Lot # 080M6035), and rabbit anti-AGPAT6 (ProteinTech 16762-1-AP, Lot # 00112879). All these antibodies were diluted to 1:1000 prior to use. A peroxidase-conjugated streptavidin (Invitrogen S911, Lot # 1809361) was also used, at a dilution of 1:3000. Statistics and reproducibility The number of biological replicates (n) is specified in the corresponding figure legends. No statistical method was used to predetermine sample size. In C. elegans experiments, stage-matched animals of each strain were randomly selected for data acquisition. In the case of in vitro experiments, individual wells were randomly assigned to different experimental conditions. Investigators were blinded to group allocation during the assessment of phenotypic outcomes, including lipid droplet size quantification, brood size quantification, cell death assays, and OCR measurement. Data are expressed as the mean ± standard deviation (SD), as noted in the legends. Statistical analyses were performed using GraphPad Prism 10 software (GraphPad Software, USA). For comparisons between two groups, an unpaired, two-tailed Student’s t test was used. For comparisons involving one independent variable with multiple groups, a one-way ANOVA was applied. For comparisons involving two independent variables, a two-way ANOVA was utilized. Post hoc testing following ANOVA was conducted using Tukey’s test for multiple comparisons or Dunnett’s test for comparisons against a control group to determine statistical significance. ns, not significant; ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001. Uncropped and unprocessed scans for blots are provided in the Source data file. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Reporting Summary Transparent Peer Review File Source data Source Data Acknowledgements We thank Yifei Qiu for preliminary data, Zhenfeng Liu and Sean McKinney for discussion, Tom Rapoport and Sarah Ho for comments on the manuscript, Pui Shuen Wong at the HKUST Bioscience Central Research Facility for lipidomic analysis, the Molecular Biology core at the Stowers Institute for Medical Research for whole-genome sequencing. H.Y.M. discloses support for the research of this work from the HKUST Center for Aging Science [grant number Z1043] and the Research Grants Council of Hong Kong [grant number T13-602/21N]. Author contributions Y.L. was responsible for Figs. 1A–E, 2, 3A–G, J, K, 4A–D, 5, S1A–E, S1G, S2K, S3, S4A–D, and S5. S.H. was responsible for Figs. 1F–J, 4E–H, 7, 8, 9F, S2A–J, and S4E–H. S.H., L.S.C. were responsible for Figs. 6, 9C, and S6. L.S.C. was responsible for Figs. 3I, 9B, E, and S1I. X.L., J.Q. were responsible for developing the SRS system and Figs. 1A, C, and 2. N.T., T.H.C. contributed reagents and analysis for LC-MS/MS, as shown in Fig. 6F. X.Z., Y.X. were responsible for Figure S4C, D. K.J. was responsible for Figs. 3H, 9A, D, and S1H. X.Y. was responsible for Fig. S1F. Y.C.T. contributed to lipid analysis. N.X. conducted genetic screens, genetic mapping and molecular cloning of C. elegans mutants. H.Y.M. supervised the project and wrote the paper with input from Y.L., S.H. Peer review Peer review information Nature Communications thanks the anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available. Data availability All data supporting the findings of this study are available within the article and the Supplementary Information file. Confocal and SRS images have been deposited in Zenodo and are publicly accessible at (10.5281/zenodo.19150821). Relative and quantitative lipidomic data have been deposited to the ProteomeXchange Consortium via the MassIVE partner repository with dataset identifiers MSV000101166 and MSV000101150, respectively. TurboID-based TMEM120A/B proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with dataset identifier PXD075982. Source data are provided with this paper. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. These authors contributed equally: Yan Li, Siwei Huang. Supplementary information The online version contains Supplementary material available at 10.1038/s41467-026-72786-x. References Associated Data Supplementary Materials Supplementary Information Reporting Summary Transparent Peer Review File Source Data Data Availability Statement All data supporting the findings of this study are available within the article and the Supplementary Information file. Confocal and SRS images have been deposited in Zenodo and are publicly accessible at (10.5281/zenodo.19150821). Relative and quantitative lipidomic data have been deposited to the ProteomeXchange Consortium via the MassIVE partner repository with dataset identifiers MSV000101166 and MSV000101150, respectively. TurboID-based TMEM120A/B proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with dataset identifier PXD075982. Source data are provided with this paper.