CLCC1 governs ER bilayer equilibration to maintain lipid homeostasis

Abstract

Orchestration of lipid production, storage and mobilization is vital for cellular and systemic homeostasis^1,2. Dysfunctional plasma lipid control represents the major risk factor for cardiometabolic diseases—the leading cause of human mortality^3,4. Within the cellular landscape, the endoplasmic reticulum (ER) is the central hub of lipid synthesis and secretion, particularly in metabolically active hepatocytes in the liver or enterocytes in the gut^5,6. Initially assembled in the ER lumen, lipid-ferrying lipoproteins necessitate the cross-membrane transfer of both neutral and phospholipids onto the lumenal apolipoprotein B (APOB), in a poorly defined process^7–10. Here we show that the ER protein CLCC1 regulates cellular lipid partition and, consequently, systemic lipid homeostasis by participating in trans-bilayer equilibration of phospholipids. CLCC1 partners with the phospholipid scramblase TMEM41B^11,12 to recognize imbalanced bilayers and promote lipid scrambling, thereby supporting lipoprotein biogenesis and the subsequent bulk lipid transport. Loss of CLCC1 or TMEM41B leads to the emergence of giant lumenal lipid droplets enclosed by imbalanced ER bilayers and, consequently, accelerated pathogenesis of metabolic-dysfunction-associated liver steatohepatitis. The results reveal that phospholipid scrambling at the ER is essential for establishing a dynamic equilibrium. Considering the requirement of trans-bilayer phospholipid equilibration in numerous biological processes, ranging from catabolic autophagy to viral infection^13–16, we anticipate that future work will elucidate a homeostatic control mechanism intrinsic to ER function in lipid biogenesis and distribution.

Phospholipid scrambling at the endoplasmic reticulum is essential for establishing a dynamic equilibrium to maintain cellular homeostasis.

Main

The cross-bilayer transit of amphipathic phospholipids synthesized asymmetrically on the cytosolic leaflet of the ER bilayer has remained an unaddressed problem until the recent identification of biogenic lipid scramblases, particularly TMEM41B^9,11,12. Notably, deficient phospholipid scrambling due to hepatic TMEM41B inactivation not only blocks neutral lipid loading and the biogenesis of very-low-density lipoproteins (VLDLs), but also triggers massive lipid overproduction and substantially accelerates metabolic-dysfunction-associated liver steatohepatitis (MASH)¹¹. Transmission electron microscopy (TEM) revealed that the innumerable lipid droplets in TMEM41B-deficient hepatocytes were all tightly surrounded by membranes (Fig. 1a and Extended Data Fig. 1a), probably originating from the ER. To further understand these unusual lipid storages, we used cryo-electron tomography (cryo-ET) coupled with high-pressure cryo-fixation (HPF) of hepatic tissues^17,18. As expected^19,20, lipid droplets in wild-type (WT) hepatocytes were observed to be situated in the cytosol (Fig. 1b (left)), often juxtaposed to the ER with an average lumen width of around 50 nm (Fig. 1b (arrowhead)). By contrast, lipid droplets in TMEM41B-deficient hepatocytes were enclosed by the ER bilayer and exhibited an average diameter of around 1 μm (Fig. 1b (right)). Subsequent 3D reconstruction and segmentation further highlighted the enclosure by extensively curved membranes of the giant lipid droplets, occupying much larger surface areas than those of ER tubules (Fig. 1b and Extended Data Fig. 1b,c). Correlative light and electron microscopy (CLEM) experiments further confirmed that ER membranes enclosed the unique lipid droplets in TMEM41B-deficient Huh7 hepatoma cells (Extended Data Fig. 1d). Further cryo-electron tomography (cryo-ET) analysis revealed that these giant lipid droplets were tightly wrapped within the ER lumen by a single ER bilayer characterized by the smooth, ribosome-free appearance, while covering the spheric surface of giant ER-enclosed lipid droplets (geLDs) and therefore adapting curved morphology (Fig. 1c and Extended Data Fig. 1e). Collectively, the ultrastructural analysis revealed an unusual lipid storage, featured by the emergence of geLDs within the ER lumen, accompanied by curved ER bilayers with imbalanced leaflets that appear to contain different amount of phospholipids (Extended Data Fig. 1f).

Fig. 1. CLCC1 or TMEM41B loss triggers geLDs.

a, TEM analysis of livers from wild-type (WT) or Tmem41b liver-specific knockout (LKO) mice. The arrowheads show the ER. LD, lipid droplet. b, Cryo-ET reconstruction of HPF WT or Tmem41b-deficient livers. The arrowhead shows the ER lumen. The arrow indicates the ER lumen filled with a LD. Right, 3D rendering of a geLD. Mito, mitochondria. c, Cryo-ET reconstruction of the ER bilayer encircling the geLD in HPF Tmem41b-deficient liver. Bottom, relative electron intensity (left) along the orange line crossing the ER bilayer (right). d, Silver staining of LDs isolated from WT and Tmem41b-deficient livers, with equal load. e, Quantitative proteomics of LDs from WT and Tmem41b-deficient livers. Red and blue indicates substantially altered ER-associated or cytosolic-LD-associated proteins. f,g, Differential labelling of cLDs and geLDs by the cLD marker PLIN2 (f) and the ER marker CB5 (g). Control and TMEM41B-KO Huh7 cells were infected with adeno-associated viruses (AAVs) expressing GFP–PLIN2 or GFP–CB5 before analysis using confocal microscopy. h, Immunoblot analysis of cLDs from WT liver, geLDs from Tmem41b-deficient liver and LDLs from plasma, with an equal load. i, Model of a geLD accompanied by curved ER bilayers in TMEM41B-deficient hepatocytes. j, The geLD-enriched proteome, TMEM41B interactome and SURF4 interactome were analysed using MS. k, Schematic of the targeted CRISPR–Cas9 screen for geLDs. The diagram was created using BioRender; Wang, X. https://BioRender.com/xtx5els (2025). l, CLCC1 deficiency induces geLDs. CRISPR–Cas9-mediated CLCC1-KO Huh7 cells were infected with AAVs expressing GFP–PLIN2 or GFP–CB5 before microscopy analysis. In a, representative images from five mice for each group are shown. In b and c, representative views from two control mice and three Tmem41b-LKO mice were reconstructed. For d–h and l, representative results of at least three biologically independent replicates are shown. Scale bars, 200 nm (a), 100 nm (b,c) and 5 μm (f,g,l). Gel source data are provided in Supplementary Fig. 1.

Source data

Extended Data Fig. 1. TMEM41B deficiency results in giant ER-enclosed lipid droplets (geLD).

a, TEM images of the livers from WT (left) or Tmem41b LKO (right) mice. Arrowheads indicate the ER. Scale bars, 500 nm. b, 3D rendering of tomogram corresponding to the region in Fig. 1b. Yellow: LD, green: mitochondria, grey: ribosome, and magenta: the ER. Scale bars, 100 nm. c, Cryo-ET reconstruction of HPF Tmem41b-KO liver samples and the 3D rendering of ER tubules. Scale bar, 100 nm. d, Cryo-SEM image of a whole grid with overlaid fluorescence signal of LDs (co-stained with LipidTox) and ER (GFP-CB5) in TMEM41B KO Huh7 cells. e, Cryo-SEM image of the lamella milled at the boxed position in (d) with overlaid fluorescence signal. f, Cryo-TEM overview of the lamella with overlaid fluorescence signal. g, Reconstructed tomogram acquired from the boxed area in (f). h, Cryo-ET reconstruction of “neck” regions where geLDs remain connected with the ER membranes in HPF Tmem41b KO liver samples. Lower left: colour rendering of the zoomed-in view indicating outer and inner layers of the ER membrane where a geLD originates. Scale bar, 100 nm. i, Model depicting the biogenesis and enlargement of geLDs, driven by asymmetric phospholipid buildup on the cytosolic leaflet and up-regulated lipogenesis. The surface monolayer of geLDs remains connected with the lumenal leaflet of the ER (depicted in light blue) in the absence of a “scission factor”. The relative shortage of phospholipids on the lumenal side may recruit unknown proteins to stabilize or compensate such imbalance. For a, n = 5 mice for each genotype. For b-c, and h, representative views of HPF liver samples from 2 control mice and 3 Tmem41b LKO mice. For d-g, representative images of 8 cells from 2 independently experiments were shown.

Biochemical isolation and silver staining revealed distinctive protein compositions in conventional cytosolic lipid droplets (cLDs) purified from wild-type liver and geLDs from TMEM41B-deficient liver (Fig. 1d and Extended Data Fig. 2a,b). Subsequent quantitative mass spectrometry (MS) revealed that marker proteins associated with cLD, notably perilipin 2 (PLIN2)^21,22, were depleted from geLD (Fig. 1e (blue dots)). Instead, these unusual lipid droplets exhibited enrichment of selective ER membrane proteins, such as torsins, sigma-1 receptor, cytochrome b₅ and UGT1A9 (Fig. 1e (red dots)). Consistent with these MS data, confocal microscopy confirmed that PLIN2 delineated the surface of cLDs in wild-type cells, but lacking from geLDs in TMEM41B-deficient cells (Fig. 1f). Conversely, the geLDs were encircled by an ER membrane marker generated from the cytochrome b₅ protein (GFP–CB5), which was absent from cLDs in wild-type cells (Fig. 1g). Immunoblotting further confirmed the absence of cLD markers including PLIN2, HSD17B13 or the lipolytic enzyme ATGL from geLDs and the enrichment of a subset of ER membrane proteins (Fig. 1h (lane 1 versus 2)). FIT2, which is known to regulate lipid partition in yeast yet has a less-pronounced role in mammals^23–25, was not detected. Notably, geLDs also lacked apolipoprotein B (APOB), the principal structural protein of VLDLs (Fig. 1h (lane 2 versus 3)). By contrast, geLDs contained the exchangeable apolipoprotein APOE²⁶.

Extended Data Fig. 2. Theoretical modelling and protein composition of the ER bilayer encircling geLDs.

a, Workflow of LD isolation from mouse liver (left) and images of cLD and geLD after density gradient fractionation (right). b, Gene ontology analysis of differential proteins in geLD compared to cLD. c, General parameters for modelling the ER membranes encircling geLDs. The inner radius was defined as r_i and the outer radius was defined as r_o, respectively. By definition, r_o = r_i + t, where t is the thickness of the surrounding bilayer. d, Algorithm in orange depicting the number of lipids differentially distributed between the lumenal and cytosolic leaflets of the bilayer surrounding a geLD (N_o-N_i) scaling with the geLD radius as 8πr_it/a_lipid, approximately ~ 180r_i, given that estimated values in the literature of t and a_lipid are ~5 nm and ~7 nm² in the eukaryotic ER membrane, respectively. Algorithm in blue depicting the percentage difference in lipid distribution between the lumenal and cytosolic leaflets of the bilayer surrounding a geLD as a reciprocal function of the geLD radius, roughly 2t/r_i, derived from the ratio of (N_o-N_i) and lipid numbers in the lumenal leaflet (N_i), formulated as [8πr_it/a_lipid]/[4πr_i²/a_lipid]. Right, examples estimating the number of differentially distributed lipids and percentage difference in bilayers surrounding geLD with radius of 100 nm (upper) or 5000 nm (lower). Calculations are detailed in the Methods section. e, Thin-layer chromatography (TLC) analysis of lipids extracted from the hepatic ER or geLDs from mice with indicated genotypes. Samples with the same amounts of proteins were subjected to lipid extraction. f, IB analysis of the hepatic ER or geLDs from mice with indicated genotypes. Same quantity of proteins from both samples were loaded. g, Two possible scenarios depicting matched bilayers (upper left) and imbalanced bilayers (upper right) surrounding geLDs. Using cLDs with comparable sizes as the reference for PL (1 N), geLDs with matched bilayers are estimated to contain 3 N PLs, whereas geLDs with imbalanced bilayers would yield less than 3 N PLs. h, Similar size distributions of geLD and cLDs used in the assay, characterized by both confocal microscopy (left) and dynamic light scattering (right). Scale bar, 5 μm. i, Cryo-EM images of isolated cLDs and geLDs from the livers of control (left) and Tmem41b LKO (right) mice, respectively. Scale bar: 50 μm. j, Membrane enclosure of geLDs determined with the protease protection assay. cLD or geLD isolated as in (h) were treated with trypsin in the absence or presence of Triton X-100, followed by IB analysis. Greater than ~85% of TorsinA in the isolated geLDs were protected (n = 3). k, TLC quantification of PLs in the above cLD and geLD samples, using TAG as the internal control. Lower: Quantification of PL signals, with controls arbitrarily set as 1. Data are presented as mean ± s.e.m. n = 3 biologically independent replicates. Statistical significance was determined by One-Sample t test. l, Procedures for fractionating geLD into membrane (M), interspaced (I), and droplet (D) fractions. m, Schematic depicting the geLD structure, corresponding to fractions obtained from (l). n, IB analysis of fractions from (l). For e-f, h, j-k and n representative results of at least 3 biological independent replicates are shown. For i, representative views of LD cryo-EM grids from 2 mice for each genotype. For gel source data, see Supplementary Fig. 1.

Source data

The combination of cryo-ET, microscopy and biochemical characterizations collectively revealed the emergence of the unique geLD storage, accompanied by curved ER bilayers with imbalance between the cytosolic versus the lumenal leaflets (Fig. 1i), due to defective phospholipid scrambling¹¹. Consistent with this notion, mathematical modelling revealed that geLD expansion could accommodate a greater absolute number of imbalanced phospholipids within the surrounding ER bilayer, while reducing the percentage of imbalance (Extended Data Fig. 2c,d), together providing a potential adaptive response to cope with sustained deficiency in lipid scrambling. Consistent with the modelling, analysis of bilayer phospholipids and characterization of the enriched proteins both confirmed a large induction of curved bilayers enclosing geLDs in TMEM41B-deficient liver, accounting for a substantial portion of the ER membranes (Extended Data Fig. 2e,f).

The massive induction of geLDs, accompanied by the pronounced alterations ER bilayers, prompted us to further investigate a potential shortfall of phospholipids on the lumenal leaflet and the geLD surface (Extended Data Fig. 2g (top right)). Using cLDs of comparable sizes as a reference, we reasoned that fully matched ER bilayers and the geLD surface—comprising a total of three phospholipid monolayers surrounding geLDs as observed by cryo-ET (Fig. 1c)—would contain around three times phospholipids (around 3N) relative to cLDs (1N). Notably, thin-layer chromatography (TLC) quantification of biochemically isolated geLDs and cLDs revealed that geLDs contained only around 1.5N phospholipids, despite having a similar size distribution to reference cLDs and largely retained the enwrapping ER bilayers (Extended Data Fig. 2h–k). The data therefore suggested that there is a phospholipid shortfall on the lumenal/geLD monolayers, supporting an asymmetric phospholipid distribution between ER leaflets caused by scrambling defects. Further fractionation of geLDs into membrane, interspaced and droplet fractions confirmed the enrichment of lumenal amphipathic proteins such as APOE and lumen-facing proteins such as torsins, indicating a selective recruitment of a unique set of functional proteins to the lumenal sites with a shortfall of phospholipids, thereby stabilizing or compensating for the imbalanced phospholipid distribution (Extended Data Fig. 2l–n).

This massive emergence of structurally altered leaflets prompted us to hypothesize that geLDs may in turn functionally enrich selective factors promoting ER bilayer equilibration, coincided with efficient phospholipid scrambling and also lipoprotein secretion. To test this hypothesis, we compiled the geLD enriched proteome with the interactomes associated with the TMEM41B scramblase¹¹ and the SURF4 cargo receptor²⁷ (Fig. 1j), resulting in a short list of proteins led by the poorly understood ER membrane protein CLCC1, which was highly enriched in membrane fractions surrounding geLDs (Extended Data Fig. 2m). Subsequently, we initiated a targeted CRISPR-mediated screen focusing on the hits identified in Fig. 1j, guided by the characteristics of geLD (Fig. 1k). Notably, loss of CLCC1 led to induction of geLDs that lacked PLIN2 decoration and instead were encircled by GFP–CB5 (Fig. 1l), recapitulating those resulting from TMEM41B deficiency. Notably, the rough ER marker GFP–Sec61β showed little enrichment around geLDs in TMEM41B- or CLCC1-deficient cells and the ER lumen marker GFP–KEDL exhibited moderate enrichment (Extended Data Fig. 3a,b), whereas APOE–GFP displayed marked enrichment to delineate the surface of geLDs (Extended Data Fig. 3c). Moreover, the cytosolic-targeted APOE mutant did not localize to the geLDs (Extended Data Fig. 3d), and markers of ER tubules or ER sheets did not decorate the geLDs (Extended Data Fig. 3e–h), further implicating the functional importance of these unique structures emerged from defective lipid scrambling at the ER bilayer.

Extended Data Fig. 3. ER bilayers encircling geLDs do not enrich markers of ER tubules or ER sheets.

a, Confocal microscopy of CRISPR/Cas9-mediated control (upper), TMEM41B KO (middle) and CLCC1 KO (lower) Huh7 cells infected with AAVs expressing GFP-Sec61β. Red, LipidTox. Scale bar, 5 μm. b, The same cells in (a) expressing GFP-KDEL. Red, LipidTox. Scale bar, 5 μm. c&d, Lumen-targeted APOE but not cytosol-targeted APOE coats on the surface of geLDs. The same cells in (a) were infected with AAVs expressing GFP-APOE (c) or AAVs expressing GFP-APOE lacking signal peptide (^ΔSPAPOE, d), prior to confocal microscopy. Red, LipidTOX. Scale bar, 5 μm. e&f, Confocal microscopy of the same cells in (a) infected AAVs expressing the ER tubule marker GFP-DP1, and co-stained with LipidTOX and APOE (e) or Plin2 (f). Scale bar, 5 μm. g&h, Confocal microscopy of the same cells in (a) infected AAVs expressing the ER sheet marker GFP-Climp63, and co-stained with LipidTOX and APOE (g) or Plin2 (h). Scale bar, 5 μm. a-h, representative results of 3 biologically independent replicates are shown.

CLCC1 and TMEM41B control bulk lipid export

Consistent with their convergence of cellular function uncovered above, tandem affinity purification revealed biochemical associations between CLCC1 and TMEM41B (Fig. 2a). Accordingly, endogenous CLCC1, but not calnexin, could be isolated from the endogenous TMEM41B immune complex from either hepatic tissues or hepatoma cells (Fig. 2b and Extended Data Fig. 4a). Moreover, genetic analysis based on the recently available Global Lipid Genetics Consortium (GLGC) data²⁸ revealed strong associations between genetic variants in the human CLCC1 gene and plasma lipids in populations (Fig. 2c,d and Extended Data Fig. 4b-e), particularly in atherogenic low-density lipoprotein (LDL) levels (P = 1.69 × 10⁻⁴¹, β = −0.078). Notably, the minor allele is present at a rather low frequency in different ethnicities (less than or close to 0.01; Fig. 2e), indicating possible strong biological effects of CLCC1 on lipid homeostasis.

Fig. 2. CLCC1–TMEM41B controls bulk lipid export.

a, Coomassie blue staining of the CLCC1–TMEM41B complex purified from HEK293F cells expressing CLCC1–Flag and twin-StrepTagII–TMEM41B. Lysates were processed for tandem purification with anti-Flag and Streptactin affinity beads. The arrows indicate CLCC1 (top) and TMEM41B (bottom). b, Co-immunoprecipitation analysis of endogenous TMEM41B and CLCC1 from liver lysates of Rhesus monkeys. c, Regional plot of CLCC1 associated with human plasma LDL-cholesterol levels. Chr., chromosome. d, The association between lead SNP rs149700491 in CLCC1 and human plasma lipid traits (GLGC data). Significantly associated traits are highlighted in red. e, The minor allele frequency of rs149700491 in individuals of European (EUR) and East Asian (EAS) descent from 1000 Genomes (Global). f, Schematic of CRISPR–Cas9-mediated acute hepatic gene inactivation. g, Hepatic CLCC1 inactivation leads to the depletion of TMEM41B and APOB100, analysed by immunoblot liver lysates of mice that received the indicated sgRNAs. h, Circulating triglyceride (TG) levels in control or Clcc1-LKO mice. n = 8 (LacZ sgRNA control), 8 (for each Clcc1-targeting sgRNA) and 8 (Clcc1 sgRNA2 plus CRISPR-resistant Clcc1 cDNA). i, TG measurements in VLDLs, LDLs and high-density lipoproteins (HDLs) fractionated from plasma samples in h. j, Circulating cholesterol levels in mice from h. k, Cholesterol measurement of samples in i. l, Immunblot analysis of plasma from control or Clcc1-LKO mice. m, Hepatic VLDL secretion from control or Clcc1-LKO mice after tyloxapol injection. n = 4 for each group. n, Immunoblot analysis of plasma APOB from m. Each sample contains plasma pooled from two mice. o, Negative staining of VLDL particles from i. Scale bar, 200 nm. Representative results of at least three biologically independent replicates are shown (in a,b,g,i,k,l,n,o). Data are mean ± s.e.m. Statistical significance was determined using unpaired two-tailed Student’s t-tests. Gel source data are provided in Supplementary Fig. 1.

Source data

Extended Data Fig. 4. CLCC1 partners with TMEM41B scramblase in governing bulk lipid secretion.

a, Co-immunoprecipitation (co-IP) of endogenous TMEM41B and CLCC1 in Huh7 cells. Lysates from cells with indicated genotypes were subjected to anti-TMEM41B IP or anti-control IgG IP, prior to SDS-PAGE and IB with the indicated antibodies. Representative of 3 biologically independent replicates was shown. b, Regional plot of CLCC1 associated with plasma cholesterol levels in humans. c, Summary of the GLGC genome-wide association data between the SNP rs77134701 in CLCC1 and plasma lipid traits in humans. In b and c, The P and β values are from GLGC. See Methods section ‘Association analyses of the SNPs with plasma lipid level’ for details. d, Minor allele frequency (MAF) difference at SNP rs77134701 in European (EUR) and East Asian (EAS) descents from gnomAD-Genomes (Global). e, The minor allele of rs77134701 is associated with decreased CLCC1 expression in human skeletal muscles (n = 706 samples from GTEx). Violin plot shows probability density of data at normalized gene expression levels for each genotype, with the centre line representing the median and the box spaning the interquartile range (IQR). The P-value was tested by a covariate-adjusted linear regression model. f, Schematic of Clcc1 sgRNA1 and sgRNA2 targeting sites in the gene. g, The FPKM values of Tmem41b mRNA from RNA sequencing of LacZ sgRNA control (CTL), Clcc1 sgRNA1 and Clcc1 sgRNA2 mouse liver. Data are presented as mean ± s.e.m. n = 4 mice for each group. h, Pulse-chase analysis of APOB degradation in primary hepatocytes from control and Clcc1 LKO mice. The indicated hepatocytes were pulse with 1 mM L-AHA for 30 min, as described in the Methods. At each chase point, total labelled APOB100 were isolated from the cell and medium by anti-APOB IP, and subjected to the click reaction, followed by SDS-PAGE for blotting biotinylated APOB. Lower: Quantification of recovered APOB as the function of the chase time. Data are presented as mean ± s.e.m. n = 3 independently biological replicates. In g-h, statistical significance was determined by unpaired two-tailed Student’s t test. i, IB analysis of APOB in CTL or Clcc1 KO murine primary hepatocytes treated with DMSO, MG132 (20 μM), 3-MA (10 mM) or MG132 (20 μM) combained with 3-MA (10 mM) for 8 h. Representative results of 3 independent replicates are shown. j, Design of CRISPR-resistant Clcc1 cDNA for the rescue experiments. k, IB analysis of plasma fractions separated by FPLC in Fig. 2i. Representative results of 3 independent experiments were shown. For gel source data, see Supplementary Fig. 1.

Source data

The biochemical and genetic data prompted us to investigate the physiological function of CLCC1, focusing on lipoprotein biogenesis and transport in which TMEM41B has an essential role. We therefore used CRISPR-mediated in vivo gene editing²⁹ to acutely inactivate hepatic CLCC1 in mouse models (Fig. 2f and Extended Data Fig. 4f). Notably, immunoblot analysis of liver proteins showed that CLCC1 depletion by two different sgRNAs both led to the downregulation of TMEM41B protein without altering its transcript levels (Fig. 2g and Extended Data Fig. 4g), further indicating physical associations between these two transmembrane proteins. Moreover, APOB100 protein also became depleted (Fig. 2g and Extended Data Fig. 4h,i), a phenomenon that is known to reflect the failure in APOB lipidation and therefore in the subsequent VLDL biogenesis^30–32. Notably, such defects with APOB100 were also observed with hepatic TMEM41B inactivation¹¹, together pointing to an essential role of the trans-bilayer phospholipid supply in the initiating step of VLDL biogenesis.

Notably, CRISPR-mediated inactivation of hepatic CLCC1 led to depletion of plasma triglycerides to near zero in fasted mice (Fig. 2h (column 1 versus 2 and 3)). This substantial lipid depletion could be rescued by reintroduction of sgRNA-resistant CLCC1 (Fig. 2h (column 4) and Extended Data Fig. 4j), demonstrating the specificity of the lipid-lowering effects to CLCC1 deficiency. Profiling of plasma lipids through size-exclusion chromatography also confirmed the depletion of atherogenic lipoproteins including VLDLs and LDLs (Fig. 2i). Similar reductions in plasma cholesterol were also observed (Fig. 2j,k). Consistent with the lipid measurements, circulating apolipoproteins, including APOB, APOA1 and APOE, were also depleted in the plasma of hepatic-CLCC1-deficient mice, while albumin levels remained unaltered (Fig. 2l and Extended Data Fig. 4k). Accordingly, the mutant mice exhibited a substantial reduction in hepatic triglyceride secretion compared with controls after injection of the LPL inhibitor tyloxapol²⁷ (Fig. 2m). Examination of circulating APOB protein levels and visualization of isolated lipoproteins by EM negative staining further confirmed diminished lipoprotein secretion caused by hepatic CLCC1 inactivation (Fig. 2n,o).

CLCC1 links phospholipid imbalance to scrambling

The essential role of CLCC1 in lipoprotein-mediated lipid homeostasis led to determine its underlying molecular mechanism in lipid control. Notably, AlphaFold3-assisted prediction³³ showed that CLCC1 may exist in high-molecular-mass oligomeric forms that assemble into a ring-like configuration (Fig. 3a,b). We next performed blue native PAGE to analyse CLCC1, and observed an induction of CLCC1 oligomers in TMEM41B-deficient cells (Fig. 3c). Notably, CLCC1 oligomers were enriched in ER membranes encircling geLDs (Fig. 3c (right two lanes)), indicating a preferred localization to the imbalanced ER bilayers induced by scramblase deficiency. Consistent with this notion, confocal microscopy revealed that endogenous CLCC1 exhibited a diffusive localization throughout the ER in wild-type cells (Fig. 3d (top)), whereas, in TMEM41B-deficient cells, CLCC1 was recruited and enriched on the curved ER membranes delineating geLDs (Fig. 3d (bottom)). To further characterize the spatial recruitment of CLCC1 to the geLD-encircling bilayers, we generated a CLCC1 lacking its lumenal domain (Extended Data Fig. 5a,b). Notably, this mutant did not undergo oligomerization or decorate geLDs (Extended Data Fig. 5c–e); unlike wild-type CLCC1, the mutant CLCC1 was unable to suppress geLDs when reintroduced into CLCC1-deficient cells (Extended Data Fig. 5f,g).

Fig. 3. CLCC1 links phospholipid imbalance to scrambling.

a,b, AlphaFold3-predicted structures of monomer (a) and hexadecamer (b) of human CLCC1 (amino acids 90–360). Blue to red, N to C terminus of CLCC1. Top view, cytosol to lumen perspective. c, Blue native PAGE analysis of CLCC1 oligomerization in membranes from WT, Tmem41b-deficient and Clcc1-deficient livers and in geLDs from Tmem41b-deficient liver. d, Confocal microscopy analysis of control and TMEM41B-KO Huh7 cells co-stained with anti-CLCC1 antibodies and LipidTox, showing CLCC1 recognizing ER bilayers encircling geLDs. e, Working model of CLCC1-dependent TMEM41B recruitment to imbalanced leaflets to catalyse bilayer equilibration. f, CLCC1-dependent recruitment of TMEM41B to the ER membranes encircling geLDs. Control and CLCC1-KO cells generated in TMEM41B-deficient Huh7 cells were infected with AAV-Flag-TMEM41B, and co-stained with anti-CLCC1 and anti-Flag antibodies and LipidTox. g, Quantification of geLD-associated Flag–TMEM41B signals from f. n = 96 and 214 LDs from 10 CLCC1-intact and 14 CLCC1-KO cells, respectively. h, CLCC1 inactivation impairs lipid scrambling at the ER in vivo. Control, CLCC1-KO, TMEM41B-KO and CLCC1/TMEM41B-double-knockout (DKO) Huh7 cells were incubated with alkyne-choline, then fixed, digitonin-permeabilized and subjected to click chemistry to label alkyne-PC at the cytosolic leaflet of ER. i, Quantification of labelled PC signals in h. n = 30, 27, 21 and 20 for control, CLCC1-KO, TMEM41B-KO and DKO cells, respectively. j, In vitro lipid-scrambling assay using NBD-PC-labelled liposomes reconstituted with buffer, TMEM41B in different doses or recombinant CLCC1–TMEM41B complex, showing CLCC1 promoting TMEM41B-mediated lipid scrambling. ‘FI(t)/FI(0)’ denotes ratio of the fluorescent intensity at time ‘t’ compared with the initial fluorescent intensity at time 0. PPR, protein-to-phospholipid molar ratio. Data are representative of at least three biologically independent replicates (c,d,f,h,j). Scale bars, 5 μm (d,f,h). The violin plots (g,i) show the probability density of different values, and the lines inside present the upper quartile, median and lower quartile. Statistical significance was determined using unpaired two-tailed Student’s t-tests. Gel source data are provided in Supplementary Fig. 1.

Source data

Extended Data Fig. 5. CLCC1 recognizes curved ER leaflets in an oligomerization-dependent manner.

a, AlphaFold3-predicted structure of murine CLCC1 dimer (aa50-165), with two alpha-helices cross-linked by interchain disulfide bonds (C67, C81), and residues 91-165 clamped together mainly based on hydrophobic interactions. b, Schematics of different CLCC1 constructs. FL, full length. ΔNTD, amino 19-165 required for CLCC1 dimerization deleted. c, CLCC1 lacking NTD (ΔNTD) fails to oligomerize. Huh7 cells were infected with AAVs expressing CLCC1-FLAG or ^ΔNTDCLCC1-FLAG, prior to blue native PAGE (upper) or SDS-PAGE (lower) analysis. d, Non-reducing PAGE analysis of CLCC1-FLAG or ^ΔNTDCLCC1-FLAG in samples as in (c). e, CLCC1 lacking NTD (ΔNTD) fails to localize to the ER encircling geLDs. TMEM41B-deficient Huh7 cells infected with AAVs expressing CLCC1-FLAG or ^ΔNTDCLCC1-FLAG, were co-stained with anti-FLAG antibodies and LipidTox, prior to confocal microscopy. Scale bar, 5 μm. f, Re-Introduction of CLCC1 FL but not ^ΔNTDCLCC1 suppresses geLDs in CLCC1-deficient Huh7 cells. CLCC1-deficient Huh7 cells infected with AAVs expressing CLCC1-FLAG or ^ΔNTDCLCC1-FLAG, were co-stained with anti-FLAG antibodies, anti-Plin2 antibodies and LipidTox, prior to confocal microscopy. Scale bar, 5 μm. g, IB analysis of the same cell sets as in (e). For b-f, representative of results of at least 3 biologically independent replicates are shown. For gel source data, see Supplementary Fig. 1.

The above biochemical and localization data led us to test whether CLCC1 is required for targeting the scramblase to imbalanced ER bilayers and subsequently equilibrating such asymmetry (Fig. 3e). We therefore transiently reconstituted Flag-tagged TMEM41B into cells deficient of the scramblase. Indeed, Flag-tagged TMEM41B was recruited to the curved membranes encircling geLDs in the presence of cellular CLCC1 (Fig. 3f (bottom)). By contrast, such recruitment did not take place after CRISPR-mediated inactivation of CLCC1 in either TMEM41B-deficient (Fig. 3f,g) or wild-type (Extended Data Fig. 6a) Huh7 cells.

Extended Data Fig. 6. CLCC1 regulates TMEM41B scramblase activity.

a, Little enrichment of endogenous TMEM41B around geLD in CLCC1 deficient cells. WT (upper), CLCC1 KO (middle) and TMEM41B KO (lower) Huh7 cells were co-stained with anti-TMEM41B antibody and LipidTox, prior to confocal microscopy. Scale bar, 5 μm. Representative results of at least 3 biologically independent replicates are shown. b, Schematic diagram of ER lipid scrambling assay in cells. Upper: a workflow of metabolic labelling and click-chemistry to detect newly-synthesized phosphatidylcholine (PC) in cells. Lower: Cells are metabolically labelled with alkyne-choline, allowing the synthesis of alkyne-PC at the cytosolic leaflet of the ER. In scrambling intact cells, alkyne-PC would be shuttled across the ER bilayer to the inner leaflet. Cell membranes are selectively permeabilized with Digitonin, without disrupting the ER bilayer. This procedure restricts ‘click-chemistry’ reagents to only label the alkyne-PC on the outer leaflet, but not the inner leaflet. When scrambling activity is impaired in cells, an accumulation of alkyne-PC occurs on the outer leaflet, resulting in enhanced fluorescent signal detection. c, Similar uptake of Alkyne choline among cells with indicated genotypes. Cells were incubated with alkyne-choline for 10 mins, and collect for metabolites extraction. The extracts were subjected to the click-reaction, followed by fluorescent TLC analysis. d, Fluorescent TLC analysisof total alkyne-PC in the CTL, CLCC1 KO and TMEM41B KO Huh7 cells. e, Re-Introduction of CLCC1 FL but not ^ΔNTDCLCC1 rescues the ER lipid scrambling defects in CLCC1-deficient Huh7 cells. Cells with indicated genotypes were infected with AAVs expressing CLCC1-FLAG or ^ΔNTDCLCC1-FLAG, and subjected to the ER lipid scrambling assay as in (b). Scale bar, 5 μm. f, Quantitative analysis of metabolically labelled PC signals in (e). Violin plots show probability density of different labelled PC signals at the cytosolic leaflet, and lines inside from top to down present the upper quartile, median, and lower quartile. n = 26 for CTL cells, 25 for CTL cells introduced with CLCC1 FL, 25 for CTL cells introduced with ^ΔNTDCLCC1, n = 30 for CLCC1 KO cells, 27 for CLCC1 KO cells introduced with CLCC1 FL, and 25 for CLCC1 KO cells introduced with ^ΔNTDCLCC1. g, Similar uptake efficiency of Alkyne choline among the cell sets as in (e). h, Fluorescent TLC analysis of total alkyne-PC in the same cell sets as in (e). In c, d, g and h, n = 3 biologically independent replicates, and data are presented as mean ± s.e.m. In c-d and f-h, Statistical significance was determined by unpaired two-tailed Student’s t test. For gel source data, see Supplementary Fig. 1.

Source data

To test whether the previously described spatial regulation by CLCC1 may impact phospholipid equilibration at the ER membrane, we used a cell-based chemical biology assay³⁴. In this assay, defective in vivo phospholipid scrambling across the ER bilayer would result in an accumulation of alkyne-labelled phosphatidylcholine (PC) on the cytosolic side of the ER membrane in cells, reflected by a leaflet-selective increase in fluorophore-conjugated PC signals after click reaction in cells permeabilized with digitonin (Extended Data Fig. 6b), while total alkyne-PC on both leaflets of the ER could be assessed in cells permeabilized with Triton X-100¹¹. Loss of the TMEM41B scramblase in cells caused cytosolic leaflet PC accumulation (Fig. 3h), as expected. Notably, CLCC1 inactivation also led to PC accumulation, albeit to a lesser extent than TMEM41B, despite similar uptake of alkyne-labelled choline and slightly reduced total alkyne-PC compared with that in control cells detected by biochemical assays (Extended Data Fig. 6c,d). Combined deficiency of CLCC1 and TMEM41B did not further exacerbate scrambling defects caused by TMEM41B inactivation alone (quantified in Fig. 3i). Although the current technology enables tracing only the leaflet-specific accumulation of PC, accounting for around 50–70% of total phospholipids, the data pointed to a regulatory role of CLCC1 in scramblase-mediated bilayer equilibration. Reintroduction of wild-type CLCC1, but not the CLCC1 mutant lacking its lumenal domain, rescued the alkyne-PC accumulation in cells (Extended Data Fig. 6e–h), further supporting a role of CLCC1 in regulating lipid scrambling at the ER bilayer. Notably, microscopy experiments detected little alteration in markers of the Golgi (MANII) or plasma membrane (wheat germ agglutinin, WGA), or components of COPII (SEC31A) and COPI (ARF1) coats in either TMEM41B- or CLCC1-deficient cells, indicating that general membrane trafficking originating from the ER remained functional (Extended Data Fig. 7a–d).

Extended Data Fig. 7. The in vitro scramblase assay and leakage detection.

a-d, Confocal microscopy of CRISPR/Cas9-mediated control (upper), TMEM41B KO (middle) and CLCC1 KO (lower) Huh7 cells co-stained with LipidTox and MANII-GFP (a), WGA (b), anti-SEC31A antibody (c), or anti-ARF1antibody (d). Scale bar, 5 μm. e, Schematic diagram of liposome-based lipid scramblase assay in vitro. NBD-PC is symmetrically distributed across both leaflets of liposomes. Upon treatment with dithionite, the NBD-PC on the outer leaflet is converted to non-fluorescent ABD-PC, while the inner leaflet remains unaltered due to the impermeability of dithionite, resulting in approximately a 50% reduction in overall fluorescence. With the activation of scramblase, there is facilitated translocation of NBD-PC between leaflets, leading to a further diminution of fluorescence intensity over time. NBD-PC, 7-nitrobenz-2-oxa-1,3-diazol-4-yl -PC. ABD-PC, 7-amino-2,1,3-benzoxadiazol-4-yl-PC. f, CLCC1 alone lacks detectable phospholipids scramblase activity in vitro. Protein-free liposomes and CLCC1 liposomes were prepared for in vitro lipid scrambling, and CLCC1-liposomes were indistinguishable from protein-free liposomes in corresponding conditions. g, Schematic diagram of floating assay. h, IB analysis of TMEM41B protein levels incorporated into liposomes for the indicated proteoliposomes used in Fig. 3j, following flotation-mediated isolation. Numbers below indicate the ratios of TMEM41B signal intensity in the top fraction versus the combines signals in top plus bottom fractions, reflecting the incorporation efficiency. i, Schematic diagram of liposome-based leakiness assay. During the assembly process, NBD-glucose is encapsulated within liposomes. If the liposome remains intact, adding dithionite to the external buffer would not quench the NBD-glucose signal, as dithionite cannot permeate the membrane. However, in compromised liposomes, dithionite penetrates and interacts with NBD-glucose, resulting in fluorescence quenching. j, NBD-Glucose leakiness assay demonstrating the intactness of protein-free liposomes, 0.3 nM TMEM41B-liposomes, CLCC1-liposomes and CLCC1-TMEM41B complex liposomes. k, Cryo-EM of protein-free liposomes (left) and proteo-liposomes containing TMEM41B (middle) or the CLCC1-TMEM41B complex (right). For a-d, f, h, and j-k, representative results of at least 3 biologically independent replicates are shown. For gel source data, see Supplementary Fig. 1.

Source data

We next used the in vitro scrambling assay³⁵ to recapitulate the contribution of CLCC1 in the cross-bilayer shuttling of phospholipids, measured by the quenching of fluorescently labelled phospholipids distributed on both leaflets of liposomes (Extended Data Fig. 7e), incorporating assay conditions designed to capture cholesterol-mediated scramblase inhibition^36,37. Considering the robust in vitro scramblase activity of TMEM41B reported previously^11,12, we initially titrated the dose of the recombinant enzyme. Accordingly, while around 50% of fluorescently labelled phospholipids were protected in the protein-free liposomes, incorporation of high concentrations of recombinant TMEM41B resulted in near-complete fluorescence quenching in the proteoliposomes (Fig. 3j (black versus red lines)). Low doses of TMEM41B led to a modest increase in fluorescence quenching relative to protein-free liposomes, as expected (Fig. 3j (green line)). By contrast, recombinant CLCC1 displayed no detectable scramblase activity in vitro (Extended Data Fig. 7f,g). Notably, incorporation of TMEM41B in a complex with CLCC1 induced a substantially greater reduction in fluorescence compared with TMEM41B alone at similar doses that also resulted in comparable incorporation efficiency into the proteoliposomes (Fig. 3j (blue versus green lines) and Extended Data Fig. 7g,h), consistent with the regulatory role of CLCC1 in trans-bilayer phospholipid shuttling at the ER membrane, as also observed in the cellular scrambling assay (Fig. 3h,i). In all conditions, little leakage or membrane deformation of the liposomes or proteoliposomes was detected (Extended Data Fig. 7i–k). Taken together with the stabilization (Fig. 2g) and the spatial enrichment of TMEM41B by CLCC1 in cells (Fig. 3f,g), the in vitro data further supported an important role of CLCC1 in scramblase-mediated bilayer equilibration.

CLCC1 safeguards liver ER architecture

The fundamental requirement of bilayer equilibration at the ER in biology prompted us to further investigate the contribution of CLCC1 to tissue homeostasis. Notably, CRISPR-mediated hepatic inactivation of CLCC1 resulted in substantially accelerated pathogenesis into MASH as early as at 4 weeks, without additional diet challenge. The mutant livers were characterized by pathologies including whitening appearance, hepatocyte ballooning, fibrosis and immune cell infiltration (Fig. 4a). Importantly, these pathologic defects could all be rescued by the reintroduction of CLCC1, demonstrating the specificity of the gene inactivation. Compared with the controls, the normalized weights of CLCC1-defecient liver became nearly doubled (Fig. 4b). Notably, the mutant liver floated to the surface in water, in contrast to the wild-type or CLCC1-rescued mutant livers, which sunk (Fig. 4c). Accordingly, the mutant liver exhibited substantial accumulation of hepatic lipids (Fig. 4d and Extended Data Fig. 8a) accompanied by severe liver damage, as reflected by elevated liver enzymes and MASH scores in the circulation compared with the controls (Fig. 4e and Extended Data Fig. 8b–g). Consistent with the histological defects, these global and molecular pathologies could all be rescued by CLCC1 reintroduction, further demonstrating the essential role of CLCC1 in liver homeostasis.

Fig. 4. CLCC1 safeguards liver ER architecture.

a, Hepatic CLCC1 inactivation leads to rapid progression into MASH. Livers of mice receiving LacZ-targeting sgRNA (control), Clcc1-targeting sgRNA or Clcc1-targeting sgRNA plus CRSIPR-resistant Clcc1 cDNA were subjected to the indicated pathology assessment. The mice were analysed 4 weeks after AAV delivery. Scale bars, 1 cm (top) and 100 μm (bottom). Representative images from LacZ control (n = 8), Clcc1 sgRNA1 (n = 8), Clcc1 sgRNA2 (n = 8) or Clcc1 sgRNA2 plus Clcc1 cDNA (n = 8) are shown. ORO, Oil Red O. b, Quantification of the liver weight (percentage of body weight) of mice in a. c, The altered density of CLCC1-deficient livers, which could float to the surface in water. d, Quantification of liver TG levels in a. n = 6 for each group. e, Quantification of plasma ALT of indicated mice in a. Data are mean ± s.e.m. Statistical significance was determined using unpaired two-tailed Student’s t-tests. f, TEM images of the hepatic LD and ER from control (left) or Clcc1-LKO (right) mice. Scale bars, 200 nm. g, TEM images of the hepatic Golgi and VLDL from control (left) or Clcc1-LKO (right) mice. Scale bars, 200 nm. Representative images of five independent mice for each group are shown (in f,g). h, Epistatic analysis of Clcc1/Tmem41b and Surf4. Liver samples from the indicated mice with single or double gene deficiency were subjected to the indicated pathology assessments. Analysis was conducted 4 weeks after AAV administration. Scale bars, 1 cm (top) and 100 μm (bottom). Representative images of Tmem41b-LKO (n = 8), Clcc1/Tmem41b-LDKO (n = 8), Surf4-LKO (n = 8) and Surf4/Clcc1-LDKO (n = 8) are shown.

Source data

Extended Data Fig. 8. CLCC1 functions with TMEM41B to maintain tissue homeostasis.

a, Quantification of liver cholesterol ester (CE) levels in Fig. 4a. n = 6 for each group. b, Quantification of plasma AST of mice in Fig. 4a. n = 8 for LacZ control, n = 8 for Clcc1 sgRNA1, n = 8 for Clcc1 sgRNA2, and n = 8 for Clcc1 sgRNA2 plus Clcc1 cDNA. In a-b, data are presented as mean ± s.e.m. c, Quantification of steatosis scores of mice in Fig. 4a. d, Quantification of ballooning scores of mice in Fig. 4a. e, Quantification of inflammation scores of mice in Fig. 4a. f, MASH scores of mice in Fig. 4a. g, Quantification of fibrosis scores of mice in Fig. 4a. In c-g, n = 8 mice for each group, and data are presented as median with error bars indicating interquartile range (IQR). h, IB analysis of Golgi enriched fractions and ER enriched fractions isolated from control mice and Clcc1 LKO mice. PNS, post nuclear supernatants. Representative of 3 biologically independent replicates was shown. i, Quantification of liver weight (% of B.W.) of mice in Fig. 4h. j, Quantification of liver TG levels of mice in Fig. 4h. k, Quantification of liver CE levels of mice in Fig. 4h. l, Quantification of plasma AST of mice in Fig. 4h. m, Quantification of plasma ALT of mice in Fig. 4h. In i-m, data are presented as mean ± s.e.m. In a-b, i-m, statistical significance was determined by unpaired two-tailed Student’s t test. For gel source data, see Supplementary Fig. 1.

Source data

We further performed ultrastructural analysis by TEM. As expected, the ER in wild-type hepatocytes exhibited an elaborate reticulum structure, characterized by a regular lumen width and smooth ends often containing lipid particles of pre-VLDL size (<100 nm in diameter). By contrast, loss of CLCC1 led to the emergence of numerous geLDs (often over 1 μm in diameter), encircled by curved ER membrane (Fig. 4f). Accordingly, VLDLs filled the Golgi apparatus in wild-type hepatocytes, but were completely absent in the Golgi of CLCC1-deficient hepatocytes (Fig. 4g and Extended Data Fig. 8h). These data were consistent with the failure in APOB lipidation observed in Fig. 2, further implicating the pivotal role of CLCC1 in phospholipid scrambling that licences APOB lipidation and VLDL assembly.

Lastly, we designed an epistatic analysis to genetically pinpoint the function of CLCC1 in lipid regulation (Fig. 4h). As previously reported, loss of hepatic TMEM41B in mice led to a rapid emergence of MASH phenotypes, to a slightly greater extent than CLCC1 inactivation alone. However, combined deficiency of both TMEM41B and CLCC1 did not further exacerbate liver pathologies compared with the single deficiency of TMEM41B (Fig. 4h (lane 1 versus 2)). These genetic data strongly suggested that these two factors converge into the same step of lipid scrambling, thereby mediating leaflet equilibration and lipoprotein biogenesis. By contrast, blocking lipoprotein export through inactivation of the cargo receptor SURF4 did not cause overt liver pathogenesis. However, combined deficiency of SURF4 and CLCC1 produced phenotypes similar to CLCC1 deficiency alone (Fig. 4h (lane 3 versus 4)). These data further positioned CLCC1 in lipoprotein biogenesis, before the transport step mediated by the SURF4 receptor. Consistently, quantitative measurements in liver weight, lipid levels and plasma AST/ALT also supported the functional convergence between CLCC1 and the TMEM41B scramblase in vivo (Extended Data Fig. 8i–m), as previously revealed by the molecular investigations in Fig. 3.

CLCC1 counters lipid-induced stress

The preceding studies underscored the severe pathological consequences of sustained leaflet imbalance at the ER, necessitating efficient equilibration through lipid scrambling across the ER bilayer. However, the identification of CLCC1 as a regulatory factor implied that the active mobilization of the lipid-scrambling mechanism orchestrates a dynamic balance, thereby avoiding deleterious progression from a potentially widespread physiological event. Consistent with this notion, loss of hepatic CLCC1 was accompanied by a substantial increase in lipogenic enzymes, including LIPIN1, FASN and ACC1. By contrast, ER chaperones for protein folding such as BIP and calnexin remained unaltered (Fig. 5a). Notably, the lipid-transfer enzyme complex MTP–PDI was also elevated, suggesting increased lipid transfer and therefore biased partition after ER leaflet imbalance^38–40. Indeed, MTP inhibition in CLCC1-deficient cells with pharmacological inhibitors^41,42 redirected neutral lipids from geLDs marked by GFP–CB5 to small cLDs delineated by GFP–PLIN2 (Fig. 5b, quantified in Fig. 5c). Similar lipid redistribution was observed with MTP inhibition in wild-type and TMEM41B-deficient cells (Extended Data Fig. 9a–c). The redistribution of lipids was accompanied by a decrease in cellular APOE and an increase in PLIN2, as detected by both microscopy and biochemical experiments (Extended Data Fig. 9d,e). The data therefore indicate that phospholipid asymmetry between ER leaflets may be coupled with directional lipid transfer, furthering a physiological role of properly regulated bilayer imbalance in lipid regulation.

Fig. 5. CLCC1 counters lipid-induced stress.

a, Liver immunoblotting analysis showing that CLCC1 loss upregulates lipogenic and lipid-transfer enzymes. b, MTP inhibition redirects neutral lipids from geLDs to cLDs in CLCC1-KO Huh7 cells expressing GFP–CB5 or GFP–PLIN2, treated with DMSO or lomitapide (MTPi) for 24 h. c, Quantification of the geLD (CB5⁺) or cLD (PLIN2⁺) area relative to total LDs from b; n = 5–6 cells per condition. d, Schematic of the CLCC1–APOE GFP complementation reporter: GFP11 and GFP1–10 were attached to the N terminus of CLCC1 and C terminus of APOE, respectively; CLCC1 and APOE engagement complements GFP and reconstitutes fluorescence. Cyto, cytosol. e, GFP complementation in TMEM41B-KO or oleic acid (OA)-treated (200 μM, 8 h) WT Huh7 cells expressing the reporter. f, Endogenous CLCC1 relocalizes around a subset of LDs after oleic acid treatment in WT Huh7 cells. TMEM41B-KO or WT Huh7 cells (with or without 200 μM OA for 8 h) were co-stained with anti-CLCC1 antibodies and LipidTox. g, Hepatic CLCC1 and TMEM41B immunoblot analysis in ob/ob versus lean mice. h, Quantification of g; n = 11 lanes per group from 3 experiments. i, Liver histology of ob/ob mice injected with AAV-GFP or AAV-CLCC1 (n = 8 per group, 4 weeks after injection). j,k, Liver TG (j) and plasma ALT (k) levels of mice from i. OE, overexpression. l, The proposed model: leaflet imbalance and its equilibration constitute a homeostatic control mechanism that is intrinsic to ER function in lipid biogenesis and distribution. Data are mean ± s.e.m. Significance was assessed using unpaired two-tailed t-tests (c h,j,k). Scale bars, 5 μm (b,e,f) and 50 μm (i). For a–c and e–g, Representative results from at least three independent experiments are shown (in a–c,e–g). Gel source data are provided in Supplementary Fig. 1.

Source data

Extended Data Fig. 9. MTP inhibition redirects neutral lipids from geLD to cytosolic LD in TMEM41B or CLCC1 deficient cells.

a, Confocal microscopy of CRISPR/-Cas9-mediated LacZ targeting Huh7 cells (CTL) infected with AAVs expressing GFP-CB5 (left) or GFP-Plin2 (right) with DMSO (upper) or 50 nM of MTP inhibitor lomitapide (lower) treatments for 24 h. Cells were stained with LipidTox prior to confocal microscopy. Scale bar, 5 μm. b, Confocal microscopy of CRISPR/-Cas9-mediated TMEM41B KO Huh7 cells with treatments and staining procedures as in (a). For a and b, Representative results of 3 biologically independent experiments were shown. Scale bar, 5 μm. c, Quantification of the geLD/total LD area ratio (CB5 positive) and cLD/total LD area ratio (Plin2 positive) from (b). Data are presented as mean ± s.e.m. n = 5 cells for CB5 and 6 for Plin2 in both DMSO or MTPi group. Statistical significance was determined by unpaired two-tailed Student’s t test. d, Confocal microscopy of the indicated Huh7 cells treated with DMSO, or 50 nM of the MTP inhibitor lomitapide for 72 h, followed by co-staining with LipidTox, anti-APOE antibody and anti-Plin2 antibody. Scale bar, 5 μm. Representative results of 3 biologically independent replicates are shown. e, IB analysis of cell lysates from indicated Huh7 cells treated with the MTP inhibitor lomitapide (50 nM) or implitapide (5 μM) for 72 h. Representative results of 3 biologically independent replicates are shown. For gel source data, see Supplementary Fig. 1.

Source data

To further detect the general presence of leaflet imbalance and biased lipid partition, we designed a complementation-based reporter pair (Fig. 5d). In brief, the non-fluorescent GFP fragments were fused to CLCC1 and the exchangeable apolipoprotein APOE (CLCC1–GFP11 and APOE–GFP1–10), respectively. Neither protein of the pair emitted fluorescence when expressed alone (Extended Data Fig. 10a). However, when co-introduced into TMEM41B-deficient cells, the pair efficiently complemented fluorescent GFP, delineating the geLD surface (Fig. 5e (top)), reflecting the spatial engagement of sufficient CLCC1 and APOE at sites of imbalanced leaflets. Notably, lipogenic treatment such as oleic acid also increased the GFP signal complemented by the reporter pair in wild-type cells (Fig. 5e (middle and bottom)). Moreover, oleic acid treatment relocated endogenous CLCC1 to encircle a small number of lipid droplets (Fig. 5f (middle and bottom)), which lacked PLIN2 but colocalized with GFP–CB5 and APOE–GFP (Extended Data Fig. 10b–d). These data indicated potentially widespread leaflet imbalance at the ER during trans-bilayer lipid shuttle, furthering that lipid scrambling represent an actively regulated biological process to maintain a dynamic equilibrium that deserves future investigation.

Extended Data Fig. 10. CLCC1 mitigates metabolic dysfunction caused by membrane stress.

a, Neither GFP11-CLCC1 or APOE-GFP1-10 could emit fluorescence in cells. Confocal microscopy of TMEM41B KO cells infected with AAVs expressing either GFP11-CLCC1 or APOE-GFP1-10 alone. Scale bar, 5 μm. b, The absence of Plin2 on CLCC1 encircled LDs induced by OA treatment. WT Huh7 cells infected with AAVs expressing GFP-Plin2, were treated with control (upper) or OA (200 μM) (lower) for 8 h, followed by co-staining with anti-CLCC1 antibody and LipidTox, prior to confocal microscopy analysis. Scale bar, 5 μm. c, OA treatment induced-colocalization of CLCC1 and CB5 surrounding LDs. WT Huh7 cells infected with AAVs expressing GFP-CB5 were treated as in (b). Scale bar, 5 μm. d, OA treatment induced-colocalization of CLCC1 and APOE surrounding LDs. WT Huh7 cells infected with AAVs expressing APOE-GFP were treated as in (d). Scale bar, 5 μm. e, IB analysis of hepatic CLCC1 and TMEM41B protein levels in the mice fed with chow diet (CD) or high fat diet (HFD). f, Quantification of IB signal in (e). n = 11 IB lanes for each group collectively from 3 biologically independent replicates. Data are presented as mean ± s.e.m. g, Quantification of liver cholesterol levels of mice in Fig. 5i. GFP, n = 6; CLCC1, n = 6. Data are presented as mean ± s.e.m. h, Quantification of plasma AST of mice in Fig. 5i. GFP, n = 8; CLCC1, n = 8. Data are presented as mean ± s.e.m. i, IB analysis of liver samples from mice in Fig. 5i. In f-h, statistical significance was determined by unpaired two-tailed Student’s t test. For a-d, e and i, representative results of at least 3 biologically independent replicates are shown. For gel source data, see Supplementary Fig. 1.

Source data

To further investigate the general presence of ER leaflet imbalance and the metabolic defects resulted from sustained bilayer disequilibrium, we employed high-fat-diet (HFD)-induced or genetical predisposed obese mice (ob/ob mice). Notably, hepatic levels of CLCC1 protein were downregulated by around 50% in both HFD (Extended Data Fig. 10e,f) or ob/ob mice (Fig. 5g,h). Consistent with the coinciding loss of TMEM41B protein after Clcc1 inactivation observed in Fig. 2g, downregulation of TMEM41B protein levels was also observed (Fig. 5g,h), further implicating ER leaflet imbalance in the metabolic defects in these obese models. Notably, hepatic expression of CLCC1 alleviated lipid accumulation in ob/ob mice, as revealed by haematoxylin and eosin (H&E), Oil Red O staining and hepatic lipid measurements (Fig. 5i,j and Extended Data Fig. 10g). Compared with the GFP control, CLCC1 expression also ameliorated liver damages, as reflected by decreased liver enzymes leaked into the plasma (Fig. 5k and Extended Data Fig. 10h). Proteins levels of TMEM41B were also elevated by CLCC1 introduction, coincided with reduced lipogenic enzymes such as LIPIN1 (Extended Data Fig. 10i). In conclusion, our study identified the poorly characterized ER membrane protein CLCC1 as a critical regulatory factor in scramblase-catalysed bilayer equilibration at the ER, thereby counteracting widespread leaflet imbalance intrinsic to normal ER physiology in lipid biogenesis and distribution. Inadequate lipid scrambling, or sustained ER bilayer disequilibrium, triggers severe cellular and systemic pathologies despite normal proteostasis (Fig. 5l). Notably, an independent study from the Olzmann group also identified a selective role of CLCC1 in lipid flux control and nuclear envelope organization⁴³. Considering the emergence of TMEM41B, CLCC1 and APOE in various diseases, spanning viral, neurodegenerative and cardiometabolic conditions^14,44–47, future investigations may identify additional regulators to establish a common mechanism in lipid-based ER homeostatic control in both health and disease.

Methods

Animals

All animal housing and use procedures were approved by the Institutional Animal Care and Use Committees of Peking University, an AAALAC accredited laboratory animal facility. All mice used in the experiments were bred on the C57BL/6J background. Surf4 ^fl/fl mice were generated and maintained as previously described²⁷. Tmem41b ^fl/fl mice (GemPharmatech, T026750) were generated by GemPharmatech on the C57BL/6J background with loxP sites flanking exons 2 and 5, using CRISPR–Cas9 genome-editing technology. Cre-dependent spCas9 knock-in (mice were purchased from the Jackson Lab (026556)²⁹. Clcc1-LKO mice were obtained by intravenous injection of spCas9 mice with AAV8 carrying TBG-cre and Clcc1-targeting sgRNA. Surf4^fl/fl mice and Tmem41b^fl/fl mice were bred with spCas9 knock-in mice to generate Surf4^fl/flspCas9 mice and Tmem41b^fl/flspCas9 mice, respectively. A list of the primers for the genotypes is provided in Supplementary Table 1. C57BL/6J mice were provided by Peking University at 6 weeks of age, and ob/ob mice (GemPharmatech, T001461) were purchased from GemPharmatech at 4 weeks of age. Mice were housed under standardized conditions, including a temperature of approximately 22 °C, a 12 h–12 h light–dark cycle and a humidity of 40–60%. Mice had free access to food and water unless otherwise stated. Male mice aged 6–16 weeks were used in all experiments. Mice were randomly assigned to different experimental groups. For fasting studies, all experiments were performed at 09:00 on the second day, with fasting starting at 17:00 on the first day. Liver samples of male macaque monkeys were acquired from the AAALAC accredited Nonhuman Primate Research Center, Peking University.

Cell culture

HEK293T cells (ATCC, CRL-3216), Huh7 cells (JCRB Cell Bank, JCRB0403) and HEK293F cells (Thermo Fisher Scientific, R79007) were obtained from ATCC, JCRB and Thermo Fisher Scientific, respectively. HEK293F suspension cells were cultured in SMM 293T-I medium (Sino Biological, M293T1), supplemented with 0.5% penicillin–streptomycin (Caisson, PSL01) and 1% FBS (VisTech, SE100-011) and maintained at 37 °C in a 5% CO₂ environment, with an optimal spinner speed of approximately 100 rpm to 130 rpm. The other cell lines were cultured in Dulbecco’s modified Eagle’s medium (DMEM) (HyClone, SH30022.01B) supplemented with 1% penicillin–streptomycin and 10% FBS, under the same temperature and CO₂ conditions. Transfections were performed with polyethyleneimine (PEI) (Polysciences, 23966-1) according to the manufacturer’s protocol.

DNA vector construction

The sgRNAs were designed using the Benchling platform (https://benchling.com/) to optimize editing efficiency and minimize the unintended off-target effects. The sgRNA sequences are listed in the Supplementary Table 2. Oligonucleotides were cloned into either the pX602-AAV-Cre sgRNA backbone for CRISPR–Cas9-induced acute gene knockout in mouse liver or the pLentiCRISPR V2 (Addgene, 52961) and pLentiGuide Blast for genome editing in cell lines, according to established protocols described previously⁴⁸. The pX602-AAV-Cre sgRNA construct was derived from the pX602 vector (Addgene, 61593) in our previous study⁴⁹. The pLentiGuide Blast construct was derived from the pLentiGuide Puro vector (Addgene, 52963), in which blasticidin S deaminase was cloned between the BsiWI and MluI restriction sites to replace puromycin N-acetyltransferase.

pAAV-TBG-mCLCC1-StrepTagII-Flag-StrepTagII was generated from pAAV-TBG-EGFP (Addgene, 105535) by replacing the GFP sequence with the mouse Clcc1 cDNA. For rescue experiments, the PAM region of mouse Clcc1 cDNA was synonymously mutated as annotated in Extended Data Fig. 3c.

For fluorescence imaging, mEGFP–PLIN2, mEGFP–CB5 (residues 90 to 124 at the C-terminal end of cytochrome b₅ type A, CYB5A), mEGFP–Sec61β, IgH signal peptide–mEGFP–KDEL, APOE–mEGFP, ApoE–GFP1-10 and GFP11–mCLCC1 (GFP11 was inserted into the N terminus of mCLCC1, followed by the signal peptide) were cloned between the KpnI and HindIII restriction sites to substitute for enhanced GFP (eGFP) in the pAAV-CAG-EGFP vector (Addgene, 51502).

For protein purification, CLCC1–Flag, twin StrepTagII–TMEM41B and Flag–TMEM41B were cloned into pKH3 (Addgene, 12555) between the EcoRI and XbaI restriction sites.

Recombinant AAV production and delivery

AAV packaging and purification were performed as previously described⁴⁹. For the mouse experiment, AAV shuttle plasmids, Rep/Cap (2/8) plasmids and helper plasmids were transfected into HEK293T cells using PEI. At 60 h after transfection, cells were collected and virus was purified, quantified by both Coomassie Blue R250 staining and quantitative PCR (qPCR). AAV-2/8 was administered by tail vein injection. To achieve acute inactivation of the hepatic Clcc1 gene in 6-week-old spCas9 knock-in mice, each mouse was administered with pX602-AAV-Cre sgRNA at a viral genome copy number of 4 × 10¹¹. For rescue experiments, each mouse was simultaneously injected with 4 × 10¹¹ viral genome copies of pX602-AAV-Cre-sgRNA and 5 × 10¹⁰ viral genome copies of AAV-TBG-mCLCC1-StrepTagII-Flag-StrepTagII.

For AAV delivery into a cultured cell line, pRep/Cap (2/8) was replaced by pAAV-DJ for AAV-DJ preparation. AAV-DJ with 1 × 10¹⁰ genome copies number was used per well in six-well plate to delivery indicated gene into cultured cells. Subsequent experiments were conducted 24 h after AAV-DJ infection, unless otherwise specified.

Lentiviral packaging and knockout cell construction

HEK293T cells were used for lentivirus packaging. In brief, the lentivirus shuttle plasmids, psPAX2 (Addgene, 12260) and pMD2.G (Addgene, 12259) plasmids were introduced into the cells using PEI according to the protocol provided by the manufacturer. Then, 48 h after transfection, the medium containing the lentivirus was collected and subsequently added to the wild-type Huh7 cells. Transduced cells were selected by antibiotic at 72 h after infection.

For control, TMEM41B-KO and CLCC1-KO Huh7 cell construction, pLentiCRISPR V2 lentivirus containing LacZ-targeting sgRNA, human TMEM41B targeting sgRNA and human CLCC1 targeting sgRNA (Supplementary Table 2) were generated and infected into Huh7 cells, respectively. Then, 2 days after transfection, cells were selected with puromycin (Sigma-Aldrich, P8833). For TMEM41B and CLCC1 double-deficient Huh7 cells, TMEM41B-KO Huh7 cells were infected with pLentiGuide Blast lentivirus containing control or human CLCC1 targeting sgRNA, and further selected with blasticidin (Wako, 022-18713).

Plasma characterization and FPLC analysis

Blood samples were collected from the tail tips of mice fasted for 16 h using a heparinized capillary. Plasma was separated by centrifugation at 6,000 rpm for 10 min at 4 °C. Triglyceride, total cholesterol, ALT/GPT and AST/GOT levels were determined using specific commercial kits (Sigma-Aldrich, TR0100; 000180, 1000000010 and 1000000020 from Zhongsheng Beikong, respectively) according to the manufacturer’s instructions. Absorbance values were measured using a Thermo Fisher Scientific Varioskan LUX multimode microplate reader controlled by Thermo Fisher Scientific SkanIt Software 3.3 for Microplate Readers. For fast-protein liquid chromatography (FPLC) analysis, pooled plasma samples from the same treatment group were fractionated using Superose 6 increase columns. The fractions were collected at a flow rate of 0.5 ml min⁻¹ for subsequent measurements of cholesterol and triglyceride levels.

VLDL secretion assay

Tyloxapol (Sigma-Aldrich, T8761) was administered to mice at a dose of 50 mg per kg body weight after a 16 h fast. Blood samples were collected at 1, 2 and 4 h after injection, and plasma triglyceride levels were analysed as previously described.

Histology

Tissue samples were collected and preserved in 4% paraformaldehyde (PFA) in PBS (Leagene, DF0135). Tissue embedding, sectioning and H&E staining were performed by the Pathology Center of Peking University or Beijing ZKWB-Bio Biotechnology. For Oil Red O staining, tissues were embedded in OCT compound and rapidly frozen. Cryosections at a thickness of 8 μm were cut and stained with Oil Red O according to the manufacturer’s instructions. Immunohistochemistry was performed on paraffin-embedded liver sections. Sections were deparaffinized and rehydrated, followed by antigen retrieval. To prevent non-specific binding, the sections were blocked with 10% goat serum for 1 h and then incubated overnight at 4 °C with primary antibodies diluted in blocking buffer. The sections were then exposed to secondary antibodies for 2 h at room temperature. Finally, they were visualized by DAB (3,3′-diaminobenzidine) staining.

Histological quantifications were performed according to the standard NAFLD activity scoring and fibrosis scoring system (https://tpis.upmc.com/), in a double-blinded manner. In brief, the histological features of metabolic-dysfunction-associated steatotic liver disease (MAFLD) and MASH were grouped into four broad categories: steatosis, lobular inflammation, hepatocyte ballooning and fibrosis. Oil Red O-stained histology images were used to score steatosis, H&E-stained histology images were used to score lobular inflammation and hepatocyte ballooning, and Masson-trichrome-stained histology images were used to score fibrosis. The MASH score represents the sum of scores for steatosis, lobular inflammation and hepatocyte ballooning.

Quantification of hepatic triglycerides, cholesterol and cholesterol esters

Liver samples were quantified and homogenized in PBS. Lipids were extracted from the homogenates according to the established protocols of the modified Bligh–Dyer method. In brief, the homogenates were vigorously mixed with a chloroform–methanol mixture (2:1). After centrifugation, the organic phase was carefully collected and concentrated using a rotary evaporator under vacuum. The lipid extract obtained was reconstituted in a solution of 15% Triton X-100 (Sigma-Aldrich) in double-distilled H₂O. Triglycerides levels were measured using commercially available kits (Zhongsheng Beikong, 100000220, for triglycerides and free glycerol; Sigma-Aldrich, F6428, for free glycerol) according to the manufacturers’ protocols. The total cholesterol levels were assessed using a commercially available kit (Zhongsheng Beikong, 100000180) according to the manufacturer’s instructions. Quantification of cholesteryl esters was measured using the Amplex Red Cholesterol and Cholesteryl Ester Test Kit (Beyotime Biotechnology, S0211M).

LD isolation and identification

The procedure of LD isolation was modified from a previously published protocol⁵⁰. Before LD dissection, mice were fasted for 4 h. After anaesthesia, the liver was perfused through the portal vein to remove blood. The liver was cut into approximately 1 mm³ pieces in buffer A (25 mM tricine pH 7.6, 250 mM sucrose) plus 0.2 mM PMSF and protease inhibitors (Roche, 4693132001) and homogenized ten times on ice with a Dounce type glass-Teflon homogenizer. The lysate was centrifuged twice at 500g for 10 min at 4 °C to remove debris. The postnuclear supernatant (PNS, 10 ml) was then transferred to a SW40 Ti centrifuge tube, and 2 ml buffer B (20 mM HEPES pH 7.4, 100 mM KCl and 2 mM MgCl₂) was added to the PNS. The tubes were loaded into a Beckman SW40 rotor and centrifuged at 300,000g for 2 h at 4 °C. After the ultracentrifugation step, the white band of LD floating on the top of the solution was carefully collected into a new SW40 Ti centrifuge tube filled with 12 ml buffer B and centrifuged at 300,000g for 30 min to remove contaminating organelles. This was repeated for three times to wash the isolated LD until no more pellets could be observed. The size of purified lipid droplets was analysed using dynamic light scattering (DLS). For estimating the membrane integrity, the isolated LD samples were treated with 0.005% trypsin in the absence or presence of 1% Triton X-100 for 30 min. For isolation of ER membrane-encapsulated geLDs, purified geLDs were treated with a hypotonic solution (1 mM Tris-HCl, pH 8.8) for 30 min on ice, to induce osmotic shock and rupture the surrounding ER membrane. Subsequent centrifugation at 100,000g for 1 h at 4 °C was performed to obtain the top fraction as the pure droplets (D), the supernatant fraction as interspace and/or peripheral-associated contents (I) and the pellet fraction as membranes (M).

LD proteins were precipitated with acetone at −80 °C overnight, centrifuged at 15,000g for 10 min, washed with ice-cold acetone, dried with the lid open and then solubilized in RIPA buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1% Nonidet P-40, 1% sodium deoxycholate, 0.1% SDS, 1 mM EDTA, 10% glycerol with protease inhibitor) supplemented with 2% SDS. Protein concentrations were measured using the BCA Protein Assay Kit (Thermo-Pierce, 23227). Silver staining of LD proteins was performed using a commercial kit according to the manufacturer’s protocol (Thermo Fisher Scientific, 24600). Proteomics analysis was performed on the Thermo Fisher Scientific Q-Exactive HF-X mass spectrometer, and raw data were processed using Proteome Discoverer software v.2.5 (Thermo Fisher Scientific), according to a previously published procedure⁴⁹. The proteins with more than one unique peptide were further analysed. The criterion for identifying differentially expressed proteins was a fold change greater than 10. Enrichment analysis of upregulated or downregulated different expressed proteins was performed separately using Cluster Profiler R package (v.4.10.0). Venn analysis was performed using the VennDiagram R package (v.1.7.3). The significant GO terms of the cellular component were visualized using the ggplot2 R package (v.3.4.4).

Isolation of subcellular organelles

Mice were fasted for 16 h overnight. After anaesthesia, the liver was perfused and homogenized as described in the ‘LD isolation and identification’ section. The lysates were centrifuged twice at 3,000g for 10 min at 4 °C to remove debris and nuclei (PNS). For analysis of APOB distribution in the early secretory pathway, an eleven-step OptiPrep gradient of 10–30% (top to bottom, 2.5% increments) was used in a SW40 Ti centrifuge tube, as a previously decribed⁵¹. The PNS (1 ml, 1 mg ml⁻¹) were then transferred to the top of the tube, followed by centrifugation at 100,000g for 4 h at 4 °C using a Beckman SW40 rotor. After ultracentrifugation, fractions (0.5 ml) were collected from top to bottom and immunoblotted for Golgi markers and ER markers. Fractions enriched with the Golgi or the ER were collected for further immunoblot analysis. For analysis of the lipid and protein composition of the ER, PNS were subjected to sequential centrifugation at 9,000g for 40 min, and 21,000g for 60 min. The 21,000g pellets were designated as the ER fractions for further analysis.

Immunoblot analysis

Liver lysates were prepared using RIPA buffer supplemented with protease inhibitors. Protein concentrations were measured using the BCA Protein Assay Kit. 1× SDS–PAGE sample buffer was added. Equal amounts of protein were loaded onto 3% to 15% Tris-acetate SDS–PAGE gels and transferred overnight at 4 °C to nitrocellulose membranes (Cytiva, 10600006). The membranes were then blocked with 5% milk in TBST (20 mM Tris pH 7.4, 150 mM NaCl and 0.1% Tween-20) for 1 h at room temperature. Primary antibodies were diluted in TBST containing 5% milk and 0.02% Na₃N, incubated with membranes overnight at 4 °C and washed three times with TBST at room temperature, for 15 min each time. HRP-conjugated goat secondary antibodies were diluted in TBST containing 5% milk and incubated for 1 h at room temperature, followed by three washes. Membranes were visualized after exposure to ECL substrate (Thermo Fisher Scientific, 34578) using Amersham Imager 600/800 systems (Cytiva) and western blot images were acquired and analysed using Amersham Imager 600/800 Analysis Software (v.3.0, Cytiva).

The following primary antibodies were used for IB: rabbit polyclonal anti-PLIN2 (Cell Signaling Technology, 45535, 1:1,000), rabbit polyclonal anti-HSD17B13 (ABclonal, A6256, 1:1,000), rabbit polyclonal anti-ATGL (Cell Signaling Technology, 2138, 1:1,000), rabbit polyclonal anti-SigmaR1 (Proteintech, 15168-1-AP, 1:1,000), rabbit polyclonal anti-calnexin (Proteintech, 10427-2-AP, 1:1,000), rabbit polyclonal anti-APOB (Proteintech, 20578-1-AP, 1:1,000), rabbit polyclonal anti-ApoE (Fitzgerald, 10R-10633, 1:1,000), mouse monoclonal anti APOE (Cell Signaling technology, 74417, 1:1,000), rabbit polyclonal anti-APOA1 (Fitzgerald, 70R-15769, 1:1,000), rabbit polyclonal anti-LIPIN1 (Proteintech, 27026-1-AP, 1:1,000), rabbit polyclonal anti-FASN (Proteintech, 10624-2-AP, 1:1,000), rabbit polyclonal anti-ACC1 (Proteintech, 21923-1-AP, 1:1,000), mouse monoclonal anti-MTP (Santa Cruz Biotechnology, sc-135994, 1:1,000), mouse monoclonal anti-PDI (BD transduction Laboratories, 610946, 1:1,000), rabbit polyclonal anti-BIP (Proteintech, 11587-1-AP, 1:1,000), rabbit polyclonal anti-tubulin (Proteintech, 11224-1-AP, 1:1,000), rabbit polyclonal anti-torsin A/DYT1 (Abcam, ab34540, 1:1,000), rabbit polyclonal anti-CES3 (Proteintech, 14587-1-AP, 1:1,000), rabbit polyclonal anti-REEP5 (Proteintech, 14643-1-AP, 1:1,000), mouse monoclonal anti-Flag (M2) (Sigma-Aldrich, F1804, 1:500). rabbit anti-GM130 antibody (Proteintech, 11308-1-AP, 1:1,000), rabbit polyclonal antibody against the N-terminal epitope of human TMEM41B (residues 1–109) were produced by the Proteintech Group (www.ptglab.com), the rabbit anti-serum was collected and purified by NHS beads conjugated with His-tagged human TMEM41B (residues 1–109), followed by washing with PBS containing 0.15% Triton X-100, eluted with 50 mM glycine (pH 2.5) and neutralized with Tris-HCl pH 8.0. Rabbit polyclonal antibodies against the C-terminal epitope of mouse CLCC1 (residues 355–539) was a gift from Y. Jia.

The following secondary antibodies were used: goat anti-mouse IgG (H+L) secondary antibody (Thermo Fisher Scientific, 31430, 1:10,000) and goat anti-rabbit IgG (H+L) secondary antibody (Thermo Fisher Scientific, 31460, 1:10,000).

Blue Native PAGE

Blue native PAGE analysis of CLCC1 was performed according to a protocol described previously⁵². In brief, livers from mice of the indicated genotype were disrupted in buffer C (25 mM Tris pH 7.4, 250 mM sucrose, 1 mM EDTA and protease inhibitors) and homogenized ten times on ice using a Dounce type glass-Teflon homogenizer. The lysates were centrifuged at 1,000g for 10 min at 4 °C to remove nuclei and unbroken cells. The supernatants were further centrifuged at 100,000g for 30 min to isolate membrane fractions. The membranes from the indicated mice and geLDs from Tmem41b-LKO mice were incubated with buffer D (25 mM Tris pH 7.4, 50 mM NaCl, 1% n-dodecyl-β-d-maltoside (DDM, Qisong biological, QS81007015) and protease) for 2 h at 4 °C, followed by centrifugation at 100,000g for 30 min. The concentrations of the supernatants were measured using the BCA Protein Assay Kit. 5% Coomassie Brilliant Blue G-250 dye was added at a final dye to detergent ratio of 1/8 (g/g). Equal amounts of protein were loaded on 4–16% blue native PAGE gel, transferred overnight at 4 °C to PVDF membranes (Cytiva, 10600023), followed by immunoblot analysis.

Immunoprecipitation analysis

For immunoprecipitation of endogenous TMEM41B, cells were washed twice with ice-cold PBS and incubated with lysis buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1% MNG, 1 mM EDTA, 10% glycerol with protease inhibitor) for 20 min on ice. Liver samples from Rhesus monkeys were homogenized in buffer C (25 mM Tris pH 7.4, 250 mM sucrose, 1 mM EDTA and protease inhibitors) on ice using a Dounce type glass-Teflon homogenizer, followed by centrifugation at 1,000g for 10 min at 4 °C to remove nuclei and unbroken cells. The supernatants were further centrifuged at 100,000g for 30 min to isolate membrane fractions. The membrane fractions were lysed in buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1% MNG, 1 mM EDTA, 10% glycerol with protease inhibitor) for 60 min on ice, followed by centrifugation at 100,000g for 30 min at 4 °C to collect the supernatants. Supernatant concentrations were adjusted based on measurements taken with the BCA Protein Assay Kit. Subsequently, the supernatants were incubated with anti-TMEM41B antibodies with end-over-end rotation for 2 h at 4 °C, followed by incubation with rProtein A beads 4FF (Smart-life sciences, SA015025), which were continuously rotated for 4 h at 4 °C. The beads were washed five times with wash buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1% MNG and 10% glycerol). For immunoblot analysis, proteins were further eluted from the beads with elution buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1% MNG, 1 mM EDTA, 10% glycerol, protease inhibitor with 0.4 mg ml⁻¹ TMEM41B (residues 1–51) peptides). After gel separation, proteins were transferred to NC membrane for immunoblotting.

Metabolic labelling of nascently synthesized APOB

Mouse primary hepatocytes were isolated from male mice liver of the indicated mouse genotypes, and cultured in collagen-coated 10 cm culture plates in high-glucose DMEM containing 10% FBS and 1% penicillin–streptomycin. Cell were washed twice with PBS, and the medium was then switched to methionine-free DMEM (Thermo Fisher Scientific, 21013-024), and cultured for an additional hour. To start the metabolic labelling, cells were next cultured in methionine-free DMEM containing 1 mM l-azidohomoalanine hydrochloride (l-AHA) (MedChemExpress, 942518-29-8) for 30 min. The chase period was initiated by washing cells twice with PBS before being cultured in high-glucose DMEM for subsequent analysis. At 10, 60 and 120 min, 20% NP-40 was added to the medium to reach a final concentration of 1%. APOB IP was performed using a rabbit anti-mouse APOB antibody (Meridian, K23300R). The immunoprecipitated APOB proteins were then incubated with 20 μM alkyne-biotin (MedChemExpress, HY-138749), 200 μM TBTA (MedChemExpress, HY-116677), 1 mM CuSO₄ and 1 mM TCEP (Thermo Fisher Scientific, 20491) in PBS at room temperature for 3 h for labelling biotin to l-AHA-pulsed APOB through the click reaction. The samples were subsequently used for SDS–PAGE analysis to evaluate the remaining newly synthesized APOB by detecting biotin through HRP-conjugated streptavidin (Proteintech, SA00001-0, 1:2,000).

Protein expression and purification

HEK293F cells were transfected with CLCC1-Flag, Flag-TMEM41B or CLCC1-Flag + twin StrepTagII-TMEM41B plasmids. Cells were grown for 48 h, and collected by centrifugation at 1,000g for 10 min at 4 °C, then washed with ice-cold TBS buffer. Cell pellets were resuspended in buffer E (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1% DDM, 1 mM EDTA, 10% glycerol with protease inhibitor). After incubation for 2 h, the cell lysates was centrifuged at 100,000g for 30 min, and the supernatant was incubated with prewashed anti-DYDDDDK affinity beads (Smart-lifesciences, SA042500) for 2 h at 4 °C, the beads were washed with buffer E (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 0.025% DDM and 10% glycerol). The protein was eluted with buffer E containing 0.4 mg ml⁻¹ Flag peptides. For tandem purification of the CLCC1–TMEM41B complex, the flag-peptide-eluted products were further incubated with Streptactin beads 4FF (Smart-lifesciences, SA053500) for 2 h at 4 °C, and the slurry was washed and eluted with buffer E containing 10 mM desthiobiotin (Sigma-Aldrich, 71610-M). Proteins were loaded onto Amicon 0.5 ml concentrators (10 kDa cut-offs, SEP, UFC501008), concentrated, buffer replaced with buffer E and quantified by Coomassie Blue R250 staining using BSA standards.

Immunofluorescence and confocal microscopy

Cells were plated into 35 mm glass-bottom dishes (Cellvis, 150680) for subsequent treatment and staining. For immunofluorescence, cells were fixed with 4% PFA in PBS for 10 min at room temperature. After fixation, cells were washed three times with PBS, then permeabilized with immunostaining permeabilization buffer with Saponin (Beyotime, P0095) for 20 min and washed three times with PBS. After blocking with 2% BSA in PBS for 1 h, the cells were incubated with primary antibodies in blocking solution overnight at 4 °C. The cells were then washed three times with PBS. The cells were then incubated with fluorescent-labelled secondary antibodies for 1 h at room temperature and washed three times with PBS. For staining, lipid droplets were visualized with HCS LipidTox Red (Thermo Fisher Scientific, H34476, 1:1,000) or HCS LipidTox Deep Red (Thermo Fisher Scientific, H34477, 1:1,000) for 20 min, and immersed in PBS or complete cell culture medium for imaging. For plasma membrane (PM) staining, PM were visualized with WGA–Alexa Fluor 488 (Thermo Fisher Scientific, W11261, 2 μg ml⁻¹) for 10 min, and immersed in phenol-red-free DMEM (Gibco, 21063029) supplemented with 10% FBS for imaging.

The following primary antibodies were used for immunofluorescence: mouse monoclonal anti-Flag (M2) (Sigma-Aldrich, F1804, 1:500); rabbit polyclonal anti-PLIN2 (Proteintech, 15294-1-AP, 1:200); mouse monoclonal anti ApoE (Cell Signaling Technology, 74417, 1:800); rabbit polyclonal anti-SEC31A (Proteintech, 17913-1-AP, 1:300); rabbit polyclonal anti-ARF1 (Proteintech, 10790-1-AP, 1:300), rabbit polyclonal antibody against the C-terminal epitope of human CLCC1 (residues 355–539, 1:500; a gift from Y. Jia); rabbit polyclonal anti-TMEM41B (1:20), purified from rabbit serum against TMEM41B (residues 1–109) using the N-terminal 1–20 amino acid peptide; goat anti-rabbit IgG (H+L) cross-adsorbed secondary antibody, Alexa Fluor 488 (Thermo Fisher Scientific, A-11008, 1:1,000); goat anti-rabbit IgG (H+L) cross-adsorbed secondary antibody, Alexa Fluor 568 (Thermo Fisher Scientific, A-11011, 1:1,000); goat anti-rabbit IgG (H+L) cross-adsorbed secondary antibody, Alexa Fluor 568 (Thermo Fisher Scientific, A-11031, 1:1,000). Images were acquired on the ZEISS 900 confocal microscope with Airyscan2 (Carl Zeiss) controlled by ZEISS ZEN v.3.2 (blue edition) and processed with Fiji (ImageJ distribution, v.2.1.0; based on ImageJ, NIH; https://imagej.net/software/fiji/).

Image analysis

To analyse the TMEM41B signal enriched around LDs, images were imported and processed using Fiji (ImageJ distribution, v.2.1.0; based on ImageJ, NIH; https://imagej.net/software/fiji/). LD regions of interest (ROIs) were detected using ‘Threshold-Analyse particle’ and the mean intensity of TMEM41B at LD ROIs was measured. The mean intensity of TMEM41B at five cytosol ROIs was then randomly measured. The ratio of mean intensity of LD ROIs to cytosol ROIs was defined as TMEM41B enrichment around LDs.

To quantify the percentages of LDs with PLIN2 or CB5 signals, images were imported into and processed using Fiji (ImageJ distribution, v.2.1.0; based on ImageJ, NIH; https://imagej.net/software/fiji/). LDs with PLIN2 or CB5 were detected and collected using ‘Threshold-Analyse particles (include holes)’ using the corresponding channel. Total LDs were detected using ‘Threshold-Analyse particles’ using the LD channel. The LD number or area with PLIN2 or CB5 signal divided by the total LD number or area is shown as the percentage of LDs with PLIN2 or CB5 signal.

Association analyses of SNPs and plasma lipid levels

GWAS summary statistics of the CLCC1 gene and plasma lipid levels were obtained from the GLGC dataset²⁸ and performed using LocusZoom (http://locuszoom.org/). The allele frequency in humans of the rs149700491 single-nucleotide polymorphism (SNP) was obtained from 1000 Genome Phase 3.

In vitro lipid scrambling assay and NBD-glucose leakiness assay

POPC (Sigma-Aldrich, 42773), POPG (Sigma-Aldrich, 840457P) and NBD-PC (Sigma-Aldrich, 810144P-1MG) were solubilized in chloroform, at a molecular ratio 180:19:1, the lipids were dried under an argon or nitrogen stream and the flask was transfected to a vacuum desiccator at room temperature for at least 3 h. The dried lipid film was reconstituted in buffer F (50 mM HEPES pH 7.4, 200 mM NaCl,) at a concentration of 5.25 mM lipid. This lipid solution was incubated in a water bath at 37 °C for 10 min, processed for ten freeze–thaw cycles and extruded 30 times through a 400 nm polycarbonate filter (Avanti, 610020). Liposomes were destabilized with 7 mM DDM at room temperature for 2 h. The following samples were added: (1) detergent buffer control; (2) detergent-solubilized protein preparations containing 50 nM (low PPR) or 500 nM (high PPR) TMEM41B; or (3) 50 nM TMEM41B in a complex with CLCC1. Individually purified CLCC1 comparable to its concentration in the CLCC1–TMEM41B complex was used. After incubation for 1 h, Bio-beads (BioRad, 1528920) were used to remove detergent according to the manufacturer’s protocol, followed by 11 extrusions through a 200 nm Avanti polycarbonate filter to minimize multilamellar. The PPR in the 50 nM TMEM41B case was approximately 1:100,000 (~3–5 proteins per vesicle on average). Examination of protein incorporation was performed using a liposome flotation assay, followed by SDS–PAGE analysis of liposome-associated proteins. The scramblase assay was performed in a cuvette equipped with a magnetic stirrer at room temperature, with a 2 ml reaction volume (diluted 20 times, with proteoliposomes containing around 250 μM lipid and 2.5 nM TMEM41B in the case of low PPR). Fluorescence signals were monitored (excitation at 460 nm, emission at 538 nm) using the Synergy H1 Hybrid Multi-Mode Reader (BioTek) controlled by Gen5 software (v.3.10, BioTek) at 1 s intervals. To assess lipid scrambling, NBD fluorescence was monitored over time after the addition of dithionite (Sigma-Aldrich, 71699) to a final concentration of 10 mM. Subsequently, additional Triton X-100 (0.5%) was then added to disrupt the liposomes, allowing for a thorough reduction of all NBD fluorescence.

The NBD-glucose leakage assay was performed in a manner similar to the lipid scramblase assay described above, except that NBD-PC was omitted from the liposome composition. Instead, NBD-glucose (Abcam, ab146200, 20 μM) was added to the buffer during the destabilization step. Subsequently, the liposomes were dialyzed to remove unincorporated NBD-glucose, and were then subjected to the leakage assay.

Click labelling with alkyne-choline

CRISPR–Cas9-mediated control, TMEM41B-KO, CLCC1-KO and TMEM41B/CLCC1-DKO Huh7 cells were seeded onto glass coverslips in six-well plates at 25% density, and cultured with 100 μM alkyne-choline (CONFLUORE, BCP-44) for 18 h. Cells were fixed with 4% PFA in PBS for 10 min at room temperature and washed three times with PBS. The cells were then permeabilized with 25 μg ml⁻¹ digitonin for 15 min and washed three times with PBS. Cells were incubated with 10 μM 5-TAMRA azide (Confluore, BDR-22), BTTAA (Confluore, BDJ-4) CuSO₄ complex (50 μM CuSO₄, BTTAA/CuSO4 6:1, mol/mol) and 2.5 mM sodium ascorbate (Aladdin, S105024) in PBS at room temperature for 1 h, and washed three times with PBS. After the click reaction, the coverslips were mounted onto glass slides and examined under the ZEISS 900 confocal microscope with the Airyscan2 module (Carl Zeiss). The fluorescence intensity was quantified using ImageJ Fiji (NIH) and analysed using Prism (GraphPad).

For alkyne choline uptake analysis, cells in 10 cm dishes at 80–90% density were incubated with 100 μM alkyne-choline for 10 min. Cellular alkyne-choline was extracted by methanol. For total alkyne PC analysis, cells in 10 cm dishes at 80–90% density were incubated with 100 μM alkyne-choline for 12 h. Lipids were extracted by chloroform:methanol (2:1) as previously described¹¹. The extracts were dried under a stream of nitrogen gas and dissolved in organic solvents, followed by a CuBF4-catalysed click reaction with 3-azido-7-hydroxycoumarin (Jena Bioscience, CLK-FA047-1) in ethanol at 55 °C for 4 h, as previously described⁵³. Then, 30 μl of this mixture was added to the lipids and incubated at 55 °C for 4 h. After centrifugation, the mixture was loaded onto 20 cm silica TLC plates and developed in chloroform:methanol:water:acetic acid (65:25:4:1) for 4–5 cm. The plates were treated with 4% N,N-diisopropylethylamine (Sigma-Aldrich, D125806) in hexanes, dried and imaged for fluorescence (excitation, 460–490 nm; emission, 518–546 nm) using ChemiDoc MP imager (BioRad), with analysis performed in Fiji ImageJ.

EM samples preparation and images

Mice were euthanized and perfused systematically with 0.1 M sodium phosphate buffer (pH 7.4), followed by a pre-fixation solution containing 2.5% glutaraldehyde and 0.8% PFA. The liver tissues were then surgically excised, fixed in the pre-fixation solution for 2 h at room temperature and further dissected into smaller sections (0.2 × 0.3 × 0.5 mm³). These samples underwent an additional overnight fixation at 4 °C in the same pre-fixing solution. After rinsing in phosphate buffer, the tissues were immersed in 0.1 M imidazole in 0.1 M sodium phosphate buffer for 30 min and then post-fixed in 2% osmium tetroxide in 0.1 M sodium phosphate buffer. After thorough rinsing with ultrapure water, the tissues were stained with 1% uranyl acetate at 4 °C overnight. The samples were dehydrated through a gradient acetone series, then embedded in epoxy resin and polymerized at 60 °C for 24 h. The prepared samples were sectioned at a thickness of approximately 60 nm using the Leica EM UC7 system, placed onto copper grids and imaged using the FEI Tecnai G2 20 Twin electron microscope equipped with a CMOS camera (EMSIS) controlled by Radius satellite software v.2.7. Examination of liver tissue structure was performed using a double-blinded method.

Cryo-ET sample preparation

High-pressure freezing

The mouse liver samples were vitrified using a previous established protocol¹⁷, with some modifications. In brief, the abdomen of each mouse was dissected under anaesthesia. A small piece of liver tissue was meticulously excised using a scalpel and placed onto an EM grid (Beijing XXBR, Cu microgrid, T10012, 200 mesh). The grid was coated with an additional carbon layer (around 10 nm thick) and freshly glow-discharged using a Model 950 Advanced Plasma System (Gatan) before use. The grid was previously placed into a 6 mm aluminium carrier (200 nm in depth). After filling the carrier with 2-methylpentane (Sigma-Aldrich, M65807), a 6 mm sapphire disc was immediately placed on top of the carrier. The assembly of the sandwich carrier was promptly inserted into the high-pressure cryostat (Leica EM ICE) for vitrification. The vitrified sandwich assembly was then transferred to a liquid ethane and propane mixture (ethane:propane = 36.9%:63.1%) at −170 °C, allowing 2-methylpentane to dissolve. The resulting grid was then transferred to liquid nitrogen for storage until milling.

The cryo-lift-out and lamella thinning

A cryo-FIB/SEM (Aquilos 2, Thermo Fisher Scientific) was used to prepare lamella, using an adapted serial lift-out method^18,54. In brief, the grid was clipped into an Autogrid (Thermo Fisher Scientific) before being loaded into the FIB chamber through the sample transfer rod. To minimize ion-beam damage and enhance the electrical conductivity, the sample was coated with a layer of Pt using a gas-injection system (GIS) for 2 min, followed by sputtering at 30 mA for 15 s. Next, four trenches were milled around the region of interest with the ion beam to expose the sample block. This block was subsequently attached to the cryo-needle and transferred to the receiver grid (copper, 100/400 mesh, Beijing XXBR, G100/400), where it was serially sectioned to several lamellae. These lamellae were securely attached to the grid in turn by redeposition method. Finally, each lamella was finely milled to a target thickness of approximal 200 nm, which was then used for subsequent tomographic data acquisition.

Cryo-ET data acquisition

Lamellae from wild-type mice were loaded onto the 300 kV cryo-transmission electron microscope (Titan Krios G3, Thermo Fisher Scientific) equipped with a K3 camera and a Gatan energy filter. Automatic tomographic tilt series acquisition was performed using SerialEM software⁵⁵. Images were acquired at a magnification of ×64,000 (pixel size, 1.37 Å) in TIFF format using super-resolution mode, resulting in ten frames per image. The defocus was set from −4.0 to −6.0 μm. The acquisition was performed from −50° to +50° (with respect to the pre-tilt angle) in 2° increments, using the dose symmetric scheme⁵⁶. The dose rate was set to 13 e⁻ px⁻¹ s⁻¹ and total dose was limited to 120 e⁻ Å⁻² per tilt series and the total dose was limited to 120 e⁻ Å⁻² per tilt series.

Lamellae from Tmem41b-LKO mice were loaded onto the 300 kV cryo-transmission electron microscope (Titan Krios G4, Thermo Fisher Scientific) with a Falcon 4i camera and a Thermo Fisher Selectris X energy filter. Automatic tomographic tilt-series acquisition was performed using Tomography v.5.12.0 (Thermo Fisher Scientific). Images were acquired at a magnification of ×42,000 (pixel size, 3.00 Å) in EER format. The defocus was set from −2.5 to −5.0 μm. The acquisition was performed from −50° to +50° (with respect to the pre-tilt angle) in 2° increments, using the dose symmetric scheme⁵⁶. The dose rate was set to 8 e⁻ px⁻¹ s⁻¹.

Correlative cryo-light microscopy and cryo-EM

The ERs in Huh7 cells were labelled by stably transfecting meGFP–CB5, while LDs were stained using LipidTox Red. Cells were cultured on gold grids (Quantifoil R2/1, 200 mesh) at 37 °C in a 5% CO₂ environment before vitrification. These grids were precoated with an additional carbon layer (~10 nm thick) and glow-discharged for 60 s using the Model 950 Advanced Plasma System (GATAN) before cell seeding.

Vitrification was performed when the cell confluency reached approximately 50%. The culture medium was replaced with a medium containing 10% glycerol before being plunge-frozen into a mixture of liquid ethane and propane (ethane:propane = 36.9%:63.1%) using Vitrobot Mark IV (Thermo Scientific) at a chamber temperature of 37°C. The vitrified grids were stored in liquid nitrogen until cryo-CLEM imaging and cryo-FIB milling.

Grids were clipped into Autogrids (Thermo Fisher Scientific) before being loaded onto the cryo-stage of the Thunder Imager EM Cryo-CLEM (Leica) equipped with a ×50/0.9 NA objective. To locate ERs and LDs, overview images, GFP signals and Texas Red signals on the grid were acquired through the reflection channels and the corresponding fluorescence channel, respectively. The overlay images were then exported into FEI MAPS software to correlate fluorescence microscopy images with scanning EM images, which guided subsequent FIB milling.

For FIB milling, the grid was transferred to the FIB chamber, where it was first coated with a Pt layer using GIS for 30 s, followed by sputtering at 30 mA for 15 s to minimize ion beam damage and enhance electrical conductivity. Alignment of fluorescence and scanning EM images was performed before milling. Regions containing fluorescence signals were milled using an 1 nA ion beam to produce lamellae with an initial thickness of 1.5 μm. The current was then progressively reduced, and a 30 pA ion beam was used for final polishing, resulting in lamellae with a final thickness of 150 nm. The prepared lamellae were stored in liquid nitrogen for subsequent data acquisition. The lipid droplets and ER network occupy a large volume within the Huh7 cells, making it highly probable that the structures of interest were retained within the thinned lamella.

Tomogram reconstruction and membrane segmentation

TOMOMAN⁵⁷ and TOM toolbox⁵⁸ were used as general tools in image processing. Initially, for TIFF format files, all frames of each tilt were motion corrected using MotionCor2 software⁵⁹. While for EER format files, 20 or 23 frames (depend on the specific EER frames) with a dose of approximal 0.10 e/Ų per frame were rendered before motion correction. Subsequently, each tilt series was aligned using patch-tracking method in IMOD software⁶⁰, and then reconstructed using back projection method to obtain a tomogram. For visualization, all tomograms were rescaled before further processing. Specifically, the wild-type tomograms were binned by a factor of 8, while tomograms from Tmem41b-KO sample used a binning factor of 4 or 6 depending on the desired resolution. Then a deconvolution filter was applied to all tomograms to further improve the contrast⁶¹.

Membrane segmentation was firstly performed using MemBrain-Seg⁶², and then manually polished with Amira 3D 2022.2 (Thermo Fisher Scientific). Lipid droplets were manually segmented with Amira. For the distance between the ER membrane and lipid droplet, each pixel of the lipid droplet edge segmentation was selected and its closest distance to ER membrane segmentation was defined as the distance. ChimeraX (v.1.7.1)⁶³ was used for the final rendering.

Cryo-EM sample preparation

For observation of purified LDs, cryo-EM holey carbon grids (Quantifoil, R2/1, Au, 200 mesh) were glow-discharged before plunge freezing using Vitrobot Mark IV (Thermo Scientific) with 100% humidity and a temperature of 6 °C. Aliquots of around 4 μl of the purified LD samples were applied to the grids, and blotted with a force of 0 and time of 5 s. All of the grids were observed using the Talos Arctica or Glacios2 microscope operated at an accelerating voltage of 200 kV.

Calculating the number imbalance between the cytoplasmic and lumenal leaflets

The number of phospholipid molecules composing a membrane layer was calculated as the ratio of the membrane surface area to the cross-sectional area of a phospholipid molecule’s phosphate group (a_lipid). The number of phospholipid molecules in the cytosolic leaflet of the ER (N_i) is 4πr_o²/a_lipid; and, the number in the lumenal leaflet of the ER (N_o) is 4πr_i²/a_lipid, where r_i is the inside radius of the encapsulating LD and the outer radius (r_o) is the inside radius + t (the thickness of the ER bilayer membrane). As an assumption, t and a_lipid are typically 5 nm and 0.7 nm² in the eukaryotic ER membrane, respectively. The geometric model derivation shows that the difference in the number of phospholipids between the inner and outer membranes (N_o − N_i) is 4πt × (2r_i +t)/a_lipid. As t is much smaller than the r_i of geLDs, the formula of N_o − N_i can be simplified to 8πtr_i²/a_lipid ≈ 180r_i. The imbalance percentage (N_o − N_i)/N_o = 2t/r_i, ~ 10/r_i (assume t = 5 nm).

CLCC1 structure prediction

The amino acid sequence 91–360 of human CLCC1 was used to predict the monomer and hexamer structure using AlphaFold3³³.

Statistics and reproducibility

All experimental results are presented as mean ± s.e.m. unless otherwise noted in the figure legends. Sample sizes were not predetermined by statistical methods. Statistical analysis was performed with GraphPad Prism 9. Statistical significance was determined using Student’s t-tests or one-way ANOVA with Tukey’s post hoc test, as indicated in the figure legends. Results were considered significant if P < 0.05. Statistical test P values are labelled in the relevant figures. Experimental results shown are representative of at least three biologically independent experiments. For Fig. 1b,c and Extended Data Fig. 1c,e, a total of 40 and 197 tomograms was collected from 2 control mice and 3 Tmem41b-LKO mice, respectively, with at least two independent experiments performed per mouse, all yielding consistent results. Mouse experiments were randomized. Imaging and histology were evaluated in a blinded manner.

Reporting summary

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

Online content

Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at 10.1038/s41586-026-10161-y.

Author contributions

X.-W.C., P.L., L.X. and Q.G. conceived the project, designed experimental strategies and analysed the data. X.-W.C. and P.L. supervised the study. X.-W.C. and X.W. generated the figures and wrote the manuscript with comments from all of the authors. L.W., Yawei Wang, D.H. and Q.L. conducted the mouse experiments and analysed the data. J.W., Q.L. and L.L. collected and analysed the imaging data. L.W., Yonglun Wang, R.W. and Y.Y. conducted the biochemical experiments and analysed the data. Y.Y. and X.X. performed phospholipid-scrambling assays. J.Y. and W.D. conducted the cryo-EM and analysed the data. Y.H. performed the bioinformatic analysis. Y.Z. performed TEM. Y.S. performed the structure prediction. S.L. and F.-J.C. carried out the MAFLD scoring. X.Z. provided monkey tissue samples.

Peer review

Peer review information

Nature thanks Edward Fisher, Anant Menon and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available.

Data availability

All data supporting the findings of this study are provided in Article and its Supplementary Information. Representative tomograms obtained in this study have been deposited at the Electron Microscopy Data Bank (EMDB) under accession numbers EMD-66721 (WT) and EMD-66722 (TMEM41b-LKO). Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Extended data

is available for this paper at 10.1038/s41586-026-10161-y.

Supplementary information

The online version contains supplementary material available at 10.1038/s41586-026-10161-y.