An ethanol-induced loss of the lipid droplet–associated segregase VCP/p97 leads to hepatic steatosis

Lipid droplets (LDs) are the central contributing organelle to hepatic steatosis that afflicts over 30% of the U.S. population. As a result, the cellular processes by which the hepatocyte synthesizes and catabolizes its LD stores are currently of great interest. Importantly, hepatic steatosis or metabolic dysfunction–associated steatotic liver disease condition is exacerbated by chronic ethanol (EtOH) consumption by mechanisms that are poorly defined (Seitz et al., 2018). As the LD proteome consists of many membrane-associated proteins (Bersuker et al., 2018; Casey et al., 2021), we predicted that chronic EtOH-induced changes in this proteome might lead to detrimental changes in how hepatocytes catabolize these key organelles. To test this concept, we performed mass spectrometry analysis of LDs purified from rat livers isolated from animals fed a liquid diet containing 30% EtOH or an isocaloric control diet. From this exercise, we observed that the levels of one important enzyme, termed valosin-containing protein of 97 kDa (VCP/p97), were substantially reduced from the LD surface. The VCP segregase is a “triple ATPase” that plays a key role in cellular protein homeostasis by recognizing and removing misfolded or damaged membrane proteins for subsequent degradation by the proteasome (Braxton and Southworth, 2023; Buchberger et al., 2015; Dantuma et al., 2014; Ferrari et al., 2022).

To provide insights into how this loss of LD-associated VCP/97 could potentiate hepatocellular steatosis, we manipulated VCP levels experimentally. Knockdown of this enzyme resulted in a three- to fivefold increase in cellular LD content, while a liver-specific VCP knockdown markedly increased hepatic steatosis. Importantly, these reductions of LD-associated VCP/p97 by either EtOH exposure, siRNA knockdown, or VCP pharmacological inhibition led to a reciprocal, several-fold increase in the LD-associated hydroxysteroid dehydrogenase (HSD17β13), an enzyme that has been implicated in hepatic steatosis by several groups (Abul-Husn et al., 2018; Ma et al., 2019; Su et al., 2014; Su et al., 2022). Surprisingly, we find that the VCP does not act to remove HSD17β13 from the LD surface for transit to the proteasome as predicted. Instead, VCP appears to act as a “co-chaperone” to present this dehydrogenase to the lysosome for degradation via an interaction with HSC70, as siRNA knockdowns of the LAMP2A lysosome pore protein increase LD-associated HSD17β13 levels. Taken together, these findings provide new insights into the function of both VCP and HSD17β13. How these interactive enzymes might act to regulate hepatocellular lipid levels under steady-state conditions or in response to an EtOH insult is discussed.

The central goals of this study were to determine whether the LD proteome is significantly altered following a chronic EtOH insult and to define the cellular response to these changes. To this end, LDs were isolated from hepatocytes derived from rats on an isocaloric control or EtOH diet after an 8-wk period using standard density gradient centrifugation methods (Schott et al., 2019). A comprehensive mass spectrometry–based proteomic screen was performed on these isolated LDs that was then analyzed for significant protein alterations in LDs from control vs. EtOH-damaged hepatocytes (Fig. 1 A). The protein segregase p97/VCP exhibited one of the greatest changes with a substantial >10-fold decrease in response to the EtOH insult (Fig. 1 A). This decrease was confirmed by western blot analysis of the same isolated LD preparations used for the mass spectrometry analysis (Fig. 1 B) and showed a near-complete loss of LD-associated p97/VCP by EtOH. Importantly, this reduction appeared to be LD-centric as overall cellular levels of VCP remained unchanged in the postnuclear supernatant fractions from which these LDs were derived.

To test whether this observed reduction in LD-associated p97/VCP might affect cellular LD content, siRNA knockdowns were performed over 72 h on isolated primary mouse hepatocytes (Fig. 1, C and D) that were then fixed and stained with Oil Red O (ORO), and the number and total area of LDs were measured. Hepatocytes with reduced p97/VCP displayed a fivefold increase in LD area compared with the nontargeted treated control cells. In support of this finding, a pharmacological inhibition of p97/VCP enzymatic activity was performed using the reversible and well-characterized agent N²,N⁴-dibenzylquinazoline-2,4-diamine (DBeQ) in the AML12 mouse hepatocyte cell line (Fig. 1, E and F) for 24 h prior to fixation and LD staining. A dose-dependent increase in the number and overall LD content was observed (Fig. 1, E and F). Analysis of the isolated LDs from EtOH diet–fed mice demonstrated a highly significant loss of the VCP adaptor UBXD8 (Fig. S1 A). Consistent with the loss of VCP promoting the accumulation of LDs, primary hepatocytes with UBXD8 knockdown also accumulated LDs (Fig. S1, B–D). Further, the expression of a mutant UBXD8 with an altered VCP-interacting domain markedly disrupted the targeting of the ATPase to the LD surface (Fig. S1, E–I), resulting in higher levels of LD content. Thus, acute manipulations of VCP and UBXD8 expression, loss of LD-targeted VCP through the expression of a UBXD8 binding mutant, and loss of VCP enzymatic activity all resulted in a marked increase in LD stores consistent with the premise that the p97/VCP enzyme plays a significant role in regulating hepatocellular lipid levels.

To test whether p97/VCP levels might play a role in regulating hepatic LD content in situ, we modified an animal model developed to reduce p97/VCP in the murine neural system (Wani et al., 2021), targeting the LoxP insertion sites in the murine gene for p97/VCP (VCP f/f) mice, which were injected with AAV8-GFP control or AAV8-TBG-Cre (thyroxin binding globulin promoter) vectors while being maintained on a standard chow diet for 7 days (Fig. S1 E). Western blot analysis of whole liver tissue homogenates from these mice revealed that >85% of VCP was reduced in this 7-day time frame (Fig. S1, G and H). Histological examination of hepatic tissue from these mice revealed a striking >10-fold accumulation of LDs in AAV8-TBG-Cre animals compared with controls. ORO staining provided a morphological confirmation of this LD increase, while a colorimetric triglyceride (glycerol phosphate oxidase) method was also used and demonstrated a similar increase in hepatic triglycerides (Fig, 1, G and H).

As VCP is known to target proteins for removal from protein complexes and membranes to support their clearance by degradation, we predicted that hepatocytes with reduced levels of LD-associated p97/VCP induced by EtOH exposure (Fig. 1 A and Fig. 2 A) might accumulate aberrant levels of LD surface proteins that would alter LD dynamics and catabolism. In addition to changes in VCP, our spectrometric assay identified a substantial sevenfold increase in the LD-associated hydroxysteroid dehydrogenase enzyme HSD17β13 (Fig. 2 A). This EtOH-induced increase was confirmed by western blot analysis (Fig. 2, A and B). This protein has generated significant interest based on the finding that human populations with loss-of-function variants of HSD17β13 have a reduced risk of chronic liver disease and progression from steatosis to steatohepatitis (Abul-Husn et al., 2018). Further, we observed a substantial increase in the levels of hepatocellular LD-associated HSD17β13 in hepatic tissues obtained from patients with alcohol-associated steatohepatitis (Fig. 2 C).

To test the effects of this increased LD association of HSD17β13 on hepatocellular steatosis, a Halo-FLAG–tagged HSD17β13 protein was overexpressed in AML12 hepatocytes for 48 h followed by fixation and ORO staining. As shown in Fig. 2, D and E, a near doubling of baseline LD content was observed (Fig. 2, D and E). Further, the expression of an enzymatically compromised version of the HSD17β13 protein bearing a single proline point mutation (P260S) led to similar LD increases, findings that are consistent with those reported by Ma et al. (2019). These observations suggest that the observed increase in LD content following HSD17β13 protein overexpression was not due to a modified lipid/membrane substrate but likely resulting from either the physical associations with other proteins, or the direct influence of increased HSD17β13 presence on the LD surface. Thus, increased LD-associated HSD17β13 likely leads to an attenuation of LD catabolism resulting in steatosis.

To define further how HSD17β13 leads to impaired lipid catabolism, we assessed LD turnover in cells transfected to express high levels of the exogenous enzyme. To this end, AML12 cells were loaded with 50 µM oleate for 18 h followed by a 24-h treatment with 10 µM concentrations of diglyceride acyltransferase (DGAT1 and DGAT2) inhibitors. The inhibitors lead to a near-complete loss of LD content in vector-transfected cells after the 24-h oleate withdrawal period, in which further TG synthesis is prevented (Fig. 2, F and G). Concomitant with these treatments, cells were transfected to express Halo-FLAG–tagged versions of WT HSD17β13, or the enzymatically dead P260S mutant, to test whether either form might result in retention of LD content due to impaired catabolism. Importantly, the HSD17β13 WT– or mutant–expressing cells retained four to fivefold more LDs compared with the control vector–transfected cells, indicating a catabolic impairment.

To provide a direct test between the correlative changes in VCP levels and in LD-associated HSD17β13 levels described above, we performed a transient knockdown of VCP (72 h) in primary hepatocytes, isolated from four distinct mice, that were then processed for western blot analysis of HSD17β13 levels. In each experiment, a near doubling of the dehydrogenase was observed in hepatocytes with reduced VCP levels (Fig. 3, A and B). In addition, these cells were fixed and stained for LDs and endogenous HSD17β13 and showed a near 50% increase in LD-associated HSD17β13 protein, supportive of the concept that the VCP segregase might regulate the clearance of HSD17β13 from the LD surface (Fig. 3, D and E). To test whether the segregase activity of VCP plays a role in regulating hepatocellular HSD17β13 levels, we again utilized the DBeQ compound that was incubated with primary mouse hepatocytes for 24 h prior to analysis of HSD17β13 levels by western blot. As depicted in Fig. 3, F and G, a corresponding three- to fourfold increase in HSD17β13 levels was observed as the concentration of DBeQ increased. A similar increase in the dehydrogenase was observed to be associated with LDs in the same DBeQ-treated mouse hepatocytes that were fixed and stained (Fig. 3, H and I). These results suggest that an active VCP segregase, rather than the presence of the VCP itself, plays a role in regulating HSD17β13 protein levels on the LD surface.

To pursue this finding further using an in vivo approach, we again utilized our conditional liver knockout (KO) model for VCP and measured changes in HSD17β13 by western blot following a 13-day tail vein injection of AAV8-TBG-Cre vectors in VCP (f/f) animals (Fig. S1 E). A massive increase in liver HSD17β13 protein levels was observed (Fig. 3 C) with no significant change in expression levels of VCP (not shown), further supporting the premise that VCP impacts HSD17β13 protein turnover. A confirmation of the VCP KO was seen in antibody-stained liver tissue sections that displayed near-complete regional loss of the VCP 13 days after vector injection (Fig. S1 F). Further support for a functional, physical interaction between these two proteins was seen in primary mouse hepatocytes expressing a GFP-tagged version of VCP that targeted endogenous HSD17β13 on the perimeter of LDs (Fig. S2 A) while also being capable of co-immunoprecipitating with a FLAG-tagged version of HSD17β13 when overexpressed in HEK293T cells (Fig. S2 B).

As VCP often targets proteins for proteasomal degradation, we assessed whether the HSD17β13 protein is targeted for degradation by the proteasome through treatments of primary mouse hepatocytes with the widely used proteasome inhibitory drug MG132 (10 µM), for 6 h. Surprisingly, MG132 treatments did not affect endogenous HSD17β13 protein levels, suggesting that its turnover was regulated by a pathway not associated with proteasome degradation (Fig. 4 A). In contrast to the proteasome inhibitory drug, treatment with the lysosomal inhibitor chloroquine (CQ) at 50 and 100 µM doses for 24 h increased the HSD17β13 protein levels in primary mouse hepatocytes more than fivefold (Fig. 4, A and B). Combined, these results suggest that the HSD17β13 protein is degraded in the lysosome and its levels are not regulated through proteasome-based clearance as predicted. Consistent with the increase in whole-cell HSD17β13 protein levels following CQ treatment, a substantial twofold increase in LD-localized levels of HSD17β13 was seen in mouse hepatocytes that were stained for endogenous HSD17β13 protein following lysosome inhibition (Fig. 4, C and D). The association of the HSD17β13 protein with lysosomes was confirmed by immunofluorescence of mouse hepatocytes where endogenous HSD17β13 protein is seen within LAMP1-labeled lysosomes, an association that dramatically increased following inhibition with CQ (Fig. S2 C).

From the findings described above, we concluded that the HSD17β13 protein is not being degraded via the proteasome but instead is reduced by lysosomal degradation. Potentially, the VCP segregase activity could be utilized to make the HSD17β13 protein available for a targeted insertion into the lysosomal lumen by a specific autophagic process termed chaperone-mediated autophagy (CMA). In this autophagic process, a cytoplasmic chaperone protein directs a client target protein into the lysosomal lumen via a multimeric LAMP2-mediated pore. This chaperone (HSC70) has been shown to interact with a specific motif of modest primary fidelity (KFERQ) within the confines of the targeted “client” protein (Kaushik and Cuervo, 2018). To assess whether HSD17β13 might undergo degradation via a CMA-mediated process, a transient knockdown of LAMP2A was performed in isolated primary mouse hepatocytes prior to western blot analysis to assess changes in HSD17β13 levels. Importantly, a near twofold increase in HSD17β13 protein levels was observed in cells with LAMP2A knockdown (Fig. 4, E and F). As interactions of HSC70 with target proteins are an important first step of the CMA process, interactions between the chaperone HSC70 and its putative client protein HSD17β13 were examined. To this end, a FLAG pull-down experiment was performed from HEK293T cells expressing HSD17β13-FLAG that was then subjected to western blot analysis of co-associated HSC70. From this approach, a marked association between the chaperone and client proteins is observed (Fig. 4 H). To define the domain within HSD17β13 that might associate with HSC70, a search for KFERQ motifs in the HSD17β13 protein was first performed using a KFERQ finder program V0.8 (https://rshine.einsteinmed.edu) (Kirchner et al., 2019) that identified a putative CMA-like motif (¹⁰⁶QVKKE¹¹⁰) in the human HSD17β13 sequence. As a CMA mutant version of HSD17β13 (¹⁰⁶AAKKE¹¹⁰) expressed in HEK293T cells retained a capacity to interact with HSC70 (not shown), a series of truncation mutants were generated that identified an additional interaction KFERQ motif in the last 50aa of the HSD17β13 C terminus (Fig. 4, G and H). This ²⁷⁵ERASA²⁷⁹ domain was subsequently mutated to ²⁷⁵AAAAA²⁷⁹, then expressed in Hep3B cells, and tested for alterations in HSC70 interactions. This secondary mutation alone abolished HSC70 binding and, importantly, led to decreased protein turnover (Fig. 4, I–L). These findings further support the premise that the HSD17β13 protein associates with its HSC70 chaperone through a KFERQ domain (²⁷⁵ERASA²⁷⁹) and is cleared from the cell through a CMA-based lysosomal degradation process.

To further define how VCP function participates in the CMA-based clearance of HSD17β13, the effects of reduced VCP segregase activity on the observed HSC70 interactions were assessed through treating AML12 cells with 7.5 µM uM DBeQ inhibitory drug for 24 h prior to an HSD17β13-FLAG pull-down and western blot for HSC70 as described above (Fig. 4, H and I). As depicted in Fig. 4, M and N, the DBeQ treatment resulted in a 60–70% reduction of the interaction between the HSC chaperone and its client protein. In contrast, this interaction was increased modestly in cells treated with the CQ lysosome inhibitor. Further, a targeted knockdown of VCP reduced the interaction of HSD17β13-halo-FLAG with the chaperone HSC70 (Fig. 4, O and P). Taken together, these findings provide support for a VCP-directed, lysosome-based, CMA degradation of the HSD17β13 protein.

The aim of this study was to define how EtOH-induced changes in the LD surface proteome could exacerbate hepatic steatosis. Our mass spectrometry observations revealed a near-total loss of LD-associated VCP/p97 (Fig. 1). From this initial observation, the role of this triple ATPase in regulating the accumulation and catabolism of LDs in hepatocytes was interrogated. We have found that reduction of hepatocellular VCP levels or pharmacological inhibition resulted in profound increases in LD content (Fig. 1). Most exciting to us was the strong causative effect between VCP perturbations and the substantial increase in levels of the HSD17β13 enzyme (Fig. 3). This increase mimicked that observed by our mass spec findings from EtOH-damaged LDs. We were surprised to observe such a dramatic increase (10–12-fold) in LD content and TAG levels within the livers of the VCP KO mice (Fig. 1). This increase was remarkably high despite the fact that the AAV8-based KO differed between adjacent cells (Fig. S1). Thus, the observed steatotic response might have been more severe had the KO extended to all of the hepatocytes within each affected liver.

Our findings focused on HSD17β13 (Figs. 2, 3, and 4) suggest that VCP plays a role in removing the dehydrogenase from LDs, not by mediating transport to the proteasome as we predicted, but by acting as a co-chaperone with HSC70. This action then directs the segregated HSD17β13 from the LD surface through the lysosomal pore and into the lysosomal lumen for subsequent degradation (Figs. 4 and S2). This CMA process appears to be key toward removal of HSD17β13 from the LD surface. This premise extended from our initial surprise that treatment of cells with the proteasome inhibitor MG132 had no effect on cellular HSD17β13 levels, while increasing concentrations of the lysosome inhibitor CQ induced a five- to eightfold increase (Fig. 4).

Our findings support a central role of the VCP-HSD17β13 interaction within a specific autophagic class, that is, CMA. These findings include, first, our previous report that lipophagy in the hepatocyte is largely carried out by a direct fusion and engulfment of the LD by the lysosome, a process termed “microlipophagy” (Schulze et al., 2020). Thus, VCP function toward the formation of a nascent, canonical, autophagosome may not be required. Second, a targeted knockdown of LAMP2A, a CMA pore protein, increases HSD17β13 levels by >50% (Fig. 4, E and F). Third, a deletion of a putative, canonical KFERQ CMA domain within the dehydrogenase reduces its interaction with the CMA chaperone HSC70 (Fig. 4, G–J), and fourth, inhibition or knockdown of the VCP ATPase reduces the interaction between HSC70 and its client protein HSD17β13 (Fig. 4, M and N). Of interest is a careful study demonstrating a VCP-centric, but autophagy-independent, regulation of LD content in nonhepatocyte cell lines that requires the VCP adaptor protein (UBXD8). In these cells, the adaptor mediates VCP binding to LDs where it reduces the activity of the adipose triglyceride lipase (ATGL) to increase LD size (Olzmann et al., 2013). Thus, in this system VCP activity would promote steatosis rather than preventing it.

How then does an EtOH insult lead to increased LD-associated HSD17β13 and what is the beneficial outcome of VCP acting to “segregate” HSD17β13 from the LD surface into the lysosome? VCP is known to interact with more than a dozen different UBX adaptors, which are likely to provide a specificity toward the distinct cellular processes mentioned above. Our findings that a 3 aa residue alteration in one of these adaptors (UBXD8) results in a drastic loss of VCP from the LD surface (Fig. S1), with a corresponding increase in LDs, are consistent with this premise. It is also possible then that EtOH exposure could reduce LD-VCP levels by disrupting an adaptor–VCP complex to the LD phospholipid monolayer, again by an undefined action. Finally, while this enzyme has been implicated in hepatocellular steatosis by several groups, the precise mechanism by which HSD17β13 contributes to this process remains unclear, though its increased presence may negatively regulate ATGL activity as recently proposed (Su et al., 2022). While humans expressing a truncated mutant form of HSD17β13 show some protection from steatohepatitis (Abul-Husn et al., 2018), studies in cultured cells, or in mice, show either an increase (Su et al., 2014; Su et al., 2022; Wang et al., 2022), no change (Abul-Husn et al., 2018; ; Ma et al., 2019; Ma et al., 2021), or decrease (Wang et al., 2022) in steatosis following manipulations of HSD17β13 levels. Further, the expression of enzymatically compromised allele variants in HepG2 cells did not change LD content (Abul-Husn et al., 2018). Thus, not all of the current studies of HSD17β13 lead to a confluent understanding on the function of this enzyme. It will be important to define how the EtOH-induced accumulation of HSD17β by a VCP dissociation potentiates hepatocellular steatosis.

Cell lines were purchased from ATCC. AML12 cells (Cat# CRL-2254) were grown in DMEM/F12K (Gibco/Cat#11330-032; Thermo Fisher Scientific) supplemented with 1% insulin–transferrin–selenium (ITS), 10% fetal bovine serum (FBS). Hep3B cells (Cat# HB-8064) were cultured in MEM (Cat# 10-010CV; Corning) media supplemented with 10% FBS. HEK293 cells (Cat# CRL-1573) were cultured in DMEM supplied with 10% FBS. All the cells were maintained at 37°C in water-jacketed incubators.

Primary cultures of isolated hepatocytes from male and female C57BL/6J mice at 2–4 mo of age were cultured in Williams E media (Cat# 12551-032; Gibco) supplemented with 5% FBS, 1% ITS (Cat# 41400-45; Gibco).

Male Wistar rats were purchased from Charles River Laboratory, and pair-fed control- and EtOH-containing Lieber-DeCarli diets for 8 wk as previously detailed (Schulze et al., 2017). Hepatocytes were isolated at the termination of the feeding period in the laboratory of Dr. Carol Casey and transported to Rochester, MN, overnight on ice. C57BL/6J WT mice were purchased from Jackson Laboratory (Cat# 0006664). VCP floxed C57BL/6J mice were provided by the laboratory of Dr. Conrad Weihl (Washington University, St. Louis, MO, USA) and generated as previously described (Wani et al., 2021). Genotypes were determined by PCR amplification of genomic DNA using primers: (FWD):5′-GGTCTTGTTGTAGAGCCCTGTTCTGTAG-3′ and (REV): 5′-CCT​GGG​ACA​AGG​ACA​CCA​CGC​TGT​TAT​C-3′. Mixed-sex cohorts of littermates were randomly assigned and used at 12–13 wk of age for 7-day conditional KO studies and 7–8 wk of age for the 13-day conditional KO study. Mice were maintained on a standard chow diet, and all animal protocols were approved by Institutional Animal Care and Use Committees in a manner consistent with the guidelines of the National Institutes of Health, United States Department of Agriculture, and Association for Assessment and Accreditation of Laboratory Animal Care International.

We used both male and female mice in our studies. Male rats were used for the EtOH diet model to avoid variabilities introduced from female hormone fluctuations during the estrous cycle. These hormonal fluctuations are reported to enhance EtOH-induced liver damage in both clinical and animal models of alcohol-associated liver injury.

Unless otherwise noted, chemicals were purchased from Sigma-Aldrich including cycloheximide (Cat# C7698), CQ (Cat# C7698), MG132 (Cat# 7449), and DGAT1 and DGAT2 inhibitors (Cat# PZ0207 and #PZ0233, respectively). The DBeQ inhibitor was purchased from Selleckchem (Cat# S7199). Commercially available antibodies were obtained as follows: FLAG epitopes (Cat# 2368S, #8146S; Cell Signaling), LAMP1 (Cat# 1D4B; DSHB), vinculin and actin (Cat# V9131 and #A2066; Sigma-Aldrich), HSD17b13 and adipophilin/PLIN2 (Cat# LS-C334590 and # LS-B3121; LS Bio), VCP (Cat# 10736-1; Proteintech), LAMP2A (Cat# ab18528; Abcam), LC3B (Cat# NB600-1384; Novus Biologicals), HSC70 (Cat# 7298; Santa Cruz Biotechnology). Secondary antibodies conjugated to Alexa Fluor probes and HRP were purchased from Thermo Fisher Scientific (Cat# A11001, A11005, A11006, A11008, A11012, G21234, G21040).

The AAV8-TBG-GFP/iCre vectors were provided by Vector Biosystems (Cat# VB1743 and VB1724). Mutagenesis was performed using KOD-Plus-Mutagenesis Kit (Cat# SMK-101; Toyobo) and the following primers: HSD17β13-Δ1-100 (FWD) 5′-TAT​CGC​TCT​CTA​AAT​CAG​GTG​AAG​AAA​GAA​GTG​GGT-3′, (REV) 5′-CAT​GGT​GGC​GGA​TCC​GAG​CTC​G-3′; HSD17β13-Δ101-200 (FWD) 5′-TCA​GAA​CTT​CAG​GCC​TTG​GGA​AAA​ACT​GG-3′, (REV) 5′-GAT​CTC​TTC​TCT​GTT​GCT​GCA​GTC​TAC​CAC​ATA​C-3′; HSD17β13-Δ201-300 (FWD) 5′-GAT​TAC​AAG​GAT​GAC​GAC​GAT​AAG​TGA​TAA​ACC​CG-3′, (REV) 5′-TGT​CAG​ACC​TCT​GTG​AAA​GCC​AAC​AGC-3′; HSD17β13-(270-271-272)Amu (FWD) 5′-GCG​GCG​GCT​CTT​CCT​GAA​CGC​GCC​TCA​GCG-3′, (REV) 5′-TAG​TCT​CAG​AAA​GAT​ATT​GAT​ATA​CGA​TGG​AAC​AAA​AAT​CAT​TTT​CTT​ATT​GGT​AA-3′; HSD17β13-(275-276-278)Amu (FWD) 5′-GCA​GCC​GCC​GCA​GCG​ATT​TTA​AAT​CGT​ATG​CAG​AAT​AT-3′, (REV) 5′-AGG​AAG​AAA​CTT​CTG​TAG​TCT​CAG​AAA​GAT​ATT​GAT​ATA​CGA​TGG-3′. The siRNA duplexes used in knockdown studies were purchased from Horizon Discovery (Cat# M-057592-01-0010, #M059036-01-0005 and #D-001810-01-20).

Hepatocytes were isolated from 2–4-mo-old C57BL/6J male and female mice. In brief, an incision was made in the abdominal cavity of anesthetized animals to expose the inferior vena cava, liver, and portal vein. Perfusion was started by placing a catheter (Cat# 387103; 20G BD Angiocath Autoguard) in the inferior vena cava, connected to a peristaltic pump, and KRH I buffer (NaCl 115 mM, HEPES 20 mM, KCl 5 mM, 1 mM KH₂PO_4, 0.5 mM EGTA) was pumped into the liver at the rate of 5 ml per minute. After approximately a minute, the portal vein was cut and KRH I buffer was replaced with KRH II buffer (NaCl 115 mM, HEPES 20 mM, KCl 5 mM, KH₂PO_4, 1 mM CaCl_2, 0.04% collagenase), using ∼40 ml to digest the liver. Following digestion, the liver was carefully excised and transferred to a 10-cm petri plate. Liver cells were dissociated in cold William’s E medium, and the cell suspension was passed through a 100-µM strainer, and then, hepatocytes were pelleted at 50 × g for 2 min. Cells were washed again in William’s E medium, and live hepatocytes were isolated using a Percoll density gradient centrifugation at 200 × g for 10 min and plated onto collagen-coated plates or coverslips.

AML12 cells were seeded in 10-cm dishes (3 × 10⁶) and transfected the following day with Lipofectamine 2000 (Thermo Fisher Scientific) following the manufacturer’s protocol. Cells were then treated with inhibitors (CQ and DBeQ) for 24-h periods. Alternatively, HEK293 cells were transfected with indicated plasmids and collected for co-immunoprecipitation assays after a 24-h expression period. Cells were lysed in lysis buffer (50 mM Tris-Cl, pH [7.4], 150 mM NaCl, 1 mM EDTA [pH 8.0], 1% Triton X-100, 0.1% IGEPAL, and protease inhibitor) and centrifuged at 10,000 g for 10 min. Protein concentrations were determined by the bicinchoninic acid assay (Pierce/Thermo Fisher Scientific). Cell lysate was precleared by incubation with mouse IgG beads (A0919; Sigma-Aldrich). Equal amounts of protein were subjected to immunoprecipitation reaction using M2 beads (A2220; Sigma-Aldrich) for 3–16 h at 4°C. Samples were washed five times in washing buffer (10 mM Tris-Cl, pH 7.4, 150 mM NaCl, 0.5 mM EDTA, pH 8.0, 0.5% IGEPAL) and eluted in 30 μl of 2 × Laemmli buffer. After elution, 2.0 μl β-ME was added and samples were boiled for 10 min. Eluted samples were subjected to the further analysis by SDS-PAGE and western blotting.

50 mg of liver samples was weighed, and 1 ml chloroform: methanol (2:1) solution was added to samples followed by homogenization using a bead homogenizer. The homogenate was shaken at room temperature for 1 h followed by a centrifugation at 10,000 × g for 5 min. The supernatant containing extract was transferred into a new tube (∼900 μl). 180 μl of double-distilled water was added to extracted samples, vortexed, and centrifuged at 3,000 g for 5 min. The upper phase was discarded, while the clear lower phase containing triglyceride was dried overnight. The following day, the dried extracts were resuspended in 100 μl 100% isopropanol. TAG was quantified using a colorimetric kit from Pointe Scientific (T7532, T7531-STD) according to the manufacturer’s instruction.

Cells were collected and washed with ice-cold 1 × PBS. Cells were lysed in Radioimmunoprecipitation assay (RIPA) lysis buffer (25 mM Tris-Cl, pH 7.4, 150 mM NaCl, 1% NP-40, 1% sodium deoxycholate, 0.1% SDS, and protease inhibitor added freshly) for 20 min, centrifuged at 10,000 × g for 10 min. The clear supernatant containing protein lysate was collected, and concentrations were determined by bicinchoninic acid assay (Pierce/Thermo Fisher Scientific). Protein was resolved on SDS-PAGE/TGX gel and transferred onto polyvinylidene fluoride (PVDF) membranes. Membranes were blocked in 5% skim milk for an hour at room temperature and incubated with the indicated primary antibody overnight at 4°C. Membranes were washed with 1X TBST (Tris-buffered Saline with Tween-20) and incubated in HRP-conjugated secondary antibodies for an hour. Membranes were developed using enhanced chemiluminescence (SuperSignal West PICO, Thermo Fisher Scientific).

For knockdown studies, cells were transfected with RNAiMAX transfection reagent from Thermo Fisher Scientific according to the manufacturer’s protocol. siGENOME pools (Dharmacon/Horizon Discovery) targeting murine VCP/p97, LAMP2A, or nontargeting control siRNA were used. Briefly, RNAiMAX reagent and siRNA were added to Opti-MEM reduced serum medium separately. Contents were mixed after 5 min and incubated for another 20 min at room temperature. After incubation, transfection mixtures were added to cells containing fresh Opti-MEM. 5 h later, fresh growth media were added to cells. Cells were collected, and knockdown efficiency was confirmed 72 h later by western blot assay.

For overexpression studies, cells were transfected with indicated plasmids using Lipofectamine 2000 and 3000 according to the manufacturer’s protocol. Briefly, Lipofectamine reagents and DNA were added to separate tubes containing Opti-MEM and incubated for 5 min at room temperature. Contents of tubes were then mixed and incubated for an additional 20 min at room temperature, and mixture was added to cells containing fresh Opti-MEM reduced serum medium. The cell media were replaced with fresh complete media after 5 h.

Cells were seeded onto coverslips in a 6-well plate and cultured overnight. For primary hepatocytes, coverslips were coated overnight with collagen (type I collagen, rat tail, 354236; Corning) before seeding. LDs were stained with ORO. The formaldehyde-fixed samples were washed in 60% isopropanol for 25 s, followed by 2-min incubation in ORO solution (5 mg/ml in isopropanol), and then washed again with 60% isopropanol for 25 s. ORO staining was analyzed using ImageJ software as described previously using the Auto Local Thresholds, “method = Bernsen radius = 5 parameter_1 = 50 parameter_2 = 0 white, size = 2-Infinity pixel circularity = 0–1.00” (Schott et al., 2019). Cells were fixed for 20 min (2.5% formaldehyde, 0.1 M PIPES, 1 mM EGTA, 3.0 mM MgSO₄) and permeabilized for 2 min using 0.1% Triton X-100. Cells were treated with blocking buffer for 30 min before incubation with indicated primary antibodies for 2 h at 37°C. After washing in DPBS, cells were further incubated with CF405, Alexa 488, and Alexa 592 fluorescently tagged secondary antibodies. The coverslips were washed and mounted on the slides after the incubation with secondary antibody in ProLong Antifade Mountant (Cat# P36934; Thermo Fisher Scientific).

For perimeter analysis, the ratio was derived from the mean fluorescence intensity obtained for endogenous, stained HSD17B13 in a defined 0.4 µm perimeter of LDs in VCP knockdown, DBeQ, and CQ-treated primary mouse hepatocytes, relative to control cells from the same experimental trial. This analysis was performed over eight fields of cells/condition in separate trails, and the average ratio of LD perimeter intensity was displayed in the graph. The perimeter of the LDs was defined in this analysis through an automated script using Fiji (ImageJ) software and automatic thresholding using Bernsen’s parameters (radius of 10; parameter_1 = 25).

Images were taken using Zeiss LSM 980 or LSM 780 confocal microscopes with Plan-Apochromat 63× oil 1.4 numerical aperture (NA) or Plan-Apochromat 40x oil 1.4 (NA) objective lenses, respectively, controlled by Zen software. Images were processed using Adobe Photoshop (Adobe Systems Incorporated) and Fiji.

Surgically resected liver tissue generated at the Mayo Clinic was collected for research with informed consent, approved by the Mayo Clinic Institutional Review Board, and deidentified. Liver tissue collected from mice for histological analysis was washed in sterile PBS and fixed in a 10% buffered formalin phosphate solution (Cat# SF100-4; Fisher Chemical) for 24 h or alternatively directly embedded in O.C.T compound (4583; Tissue-Tek) for use in ORO staining. Processing of fixed mouse liver tissue, sectioning, and hematoxylin and eosin staining were performed by the Mayo Clinic Biomaterials and Histomorphometry Core Laboratory. Histological staining of mouse liver and human patient samples was performed using Cell and Tissue Staining Kit (Cat# CTS005; R&D Systems). ORO staining of O.C.T.-embedded tissue was performed as previously described (Rozeveld et al., 2020). Briefly, liver sections were washed in 60% isopropanol for 30 s, stained for 15 min in an ORO solution (5 mg/ml in isopropanol), followed by destaining in 60% isopropanol for 30 s, deionized water rinses, and hematoxylin counterstain.

LDs were isolated from rat liver–derived hepatocytes using OptiPrep density gradient centrifugation techniques as described previously (Schott et al., 2019). Briefly, Dulbecco's Phosphate Buffered Saline (DPBS)-rinsed cells were lysed in a hypotonic lysis medium (40 mM Tris-HCl, pH 7.4, 2 µM EDTA, 20 mM NaF, cOmplete protease inhibitor) by incubating on ice for 10 min. before Dounce homogenization. Nuclei were removed from the homogenate by a 10-min. centrifugation at 1,500 × g at 4°C. Postnuclear supernatants were adjusted to 30% (iodixanol; wt/vol) and layered with discontinuous steps from 25% to 0% for separation over 30 min. at 4°C using a SW55Ti rotor at 17,200 rpm (36,000 × g). The top gradient containing floating fat/LDs was collected and spun in microfuge tubes at 4°C for 10 min at 13,000 × g to remove excess buffer and enrich LDs, which were supplemented with SDS sample buffer prior to gel electrophoresis and proteomic analysis.

SDS-PAGE gel bands are prepared for mass spectrometry analysis using the following procedures. Colloidal blue–stained gel bands are destained in 50% acetonitrile/50 mM Tris, pH 8.1, until clear, and the proteins were reduced with 50 mM TCEP/50 mM Tris, pH 8.1, at 55°C for 30 min, followed by alkylation using 25 mM iodoacetamide/50 mM Tris, pH 8.1, at room temperature for 30 min in the dark. Proteins are digested in situ with 0.15ug trypsin (Promega Corporation) in 25 mM Tris, pH 8.1/0.0002% Zwittergent 3–16, at 37°C overnight, followed by peptide extraction with 2% trifluoroacetic acid and acetonitrile. The pooled extracts are concentrated, and the proteins are identified by nano-flow liquid chromatography electrospray tandem mass spectrometry using a Thermo Fisher Scientific Q Exactive Mass Spectrometer (Thermo Fisher Scientific) coupled to a Thermo Ultimate 3000 RSLCnano HPLC system. The digest peptide mixture is loaded onto a 250 nl OPTI-PAK trap (Optimize Technologies) custom packed with Michrom Magic C18 solid phase (Michrom Bioresources). Chromatography is performed using 0.2% formic acid in both the A solvent (98% water/2% acetonitrile) and B solvent (80% acetonitrile/10% isopropanol/10% water), and a 2%B to 45%B gradient over 90 min at 325 nl/min through a 100 µm × 35 cm PicoFrit column hand-packed with Agilent Poroshell C18 2.7 um solid phase. The Q Exactive mass spectrometer setup was an FT full scan from 340 to 1,800 m/z at resolution 70,000 (at 200 m/z), followed by HCD MS/MS scans on the top 15 ions at resolution 17,500 with the NCE set to 27. The MS1 AGC target is set to 3e6, and the MS2 target is set to 2e5 with max ion inject times of 60 ms for each. Dynamic exclusion places selected ions on an exclusion list for 30 s.

Tandem mass spectra were extracted by msconvert (version 3.0.4019; ProteoWizard), and all MS/MS samples were analyzed using Mascot (Matrix Science; version 2.4.0) and X! Tandem (The GPM, https://thegpm.org; version CYCLONE [2010.12.01.1]). Mascot and X! Tandem were set up to search the UniProt rat reference protein database from January 2016, including a decoy reverse database and assuming the digestion enzyme full trypsin. Mascot and X! Tandem were searched with a fragment ion mass tolerance of 0.10 Da and a parent ion tolerance of 10.0 ppm. Oxidation of methionine and iodoacetamide derivative of cysteine were specified as variable modifications.

Scaffold (version Scaffold_4.2.1, Proteome Software, Inc.) is used to validate MS/MS-based peptide and protein identifications. Peptide identifications are accepted if they can be established at >95.0% probability as specified by the PeptideProphet algorithm (Keller et al., 2002). Protein identifications are accepted if they can be established at >95.0% probability and contain at least two identified peptides. Protein probabilities are assigned by the Protein Prophet algorithm (Nesvizhskii et al., 2003). Proteins that contained similar peptides and cannot be differentiated based on MS/MS analysis alone are grouped to satisfy the principles of parsimony.

We employed a label-free peptide MS1 intensity–based method to identify differentially expressed proteins between experimental groups. To identify and quantify protein groups, we employed a label-free peptide MS1 intensity–based method. Raw data quality was assessed using quality control metrics from the Swift proteomic data processing pipeline (Ayers-Ringler et al., 2016). The raw data files were processed using MaxQuant (version 1.5.1) to generate a list of protein groups and their corresponding intensities. MaxQuant was configured to use a composite mouse protein sequence database, incorporating the UniProt mouse reference proteome and common contaminants such as trypsin, keratin, and wool. Reversed protein sequences were included to estimate false discovery rates (FDRs). The software was set to use a 20 ppm m/z tolerance for precursors and fragments, identifying semi-tryptic peptides and considering variable modifications such as carbamidomethylation of cysteine, oxidation of methionine, n-terminal pyroglutamic acid formation, and protein n-terminal acetylation. MaxQuant aligned runs and matched features across multiple sample runs, filtering peptide and protein identifications at a 2% FDR and reporting protein group intensities.

Differential expression analysis was conducted using an in-house R script. Protein group intensities for each sample were log₂-transformed and normalized using the quantile method. The normalized intensities from two groups of samples were modeled using a Gaussian-linked generalized linear model. An ANOVA test was then applied to identify differentially expressed protein groups between experimental pairs. P values were corrected for multiple comparisons using the Benjamini–Hochberg–Yekutieli procedure. Protein groups with an FDR < 0.05 and an absolute log₂ fold change of at least 0.5 were considered significantly differentially expressed and selected for subsequent pathway analysis. This rigorous normalization and statistical approach ensured accurate identification of significant changes in protein expression between experimental conditions.

Differentially expressed proteins identified through bioinformatics analysis were further examined using QIAGEN’s IPA. Only proteins with intensity readings from all three technical replicates were included. Normalized ratios, P values, and FDRs were uploaded to IPA along with corresponding SwissProt/UniProt identifiers. IPA was configured to analyze proteins with a log₂ fold change of ≥1.25, a corrected P value <0.05, and an FDR (q-value) of 0.001. The IPA Knowledge Base was restricted to select tissues and cells primarily from the nervous and immune systems to focus on relevant biological pathways. IPA utilized right-tailed Fisher’s exact test to determine the statistical significance of the pathway associations. Results were presented with significance (P value), ratio (number of proteins from the dataset/total known proteins in the pathway), and z-score (standard deviations above or below the mean). This approach provided a robust framework for interpreting the functional implications of the differentially expressed proteins, offering insights into the biological pathways and networks most impacted by the experimental conditions.

All statistical analysis was done using GraphPad Prism software (v.10). Graphs represent the mean + SEM or SD as noted in the legend; experimental results represent at least three independent biological replicates unless otherwise noted (refer to figure legends for sample size). P values were calculated using Student’s t test or a two-way ANOVA as noted in Fig. 4, K and L. Image analysis and immunoblot quantitation were performed using ImageJ software.

All studies involving the use of animals were performed in accordance with protocols approved by the Institutional Animal Care and Use Committees at the Mayo Clinic and University of Nebraska Medical Center. Deidentified patient specimens were obtained through an active protocol approved by the Mayo Clinic Institutional Review Board.

Fig. S1 demonstrates that the VCP loss is associated with an increase in LD content. Fig. S2 shows p97/VCP and HSD17b13 interacting with each other.

All data are available from the corresponding author upon reasonable request.

This work was supported by National Institute on Alcohol Abuse and Alcoholism and National Institute of Diabetes and Digestive and Kidney Diseases grants AA020735 (to C.A. Casey and M.A. McNiven) and DK044650 (to M.A. McNiven).

Author contributions: S. Sen: conceptualization, data curation, formal analysis, investigation, methodology, project administration, visualization, and writing—original draft, review, and editing. S. Weller: conceptualization, data curation, formal analysis, investigation, methodology, project administration, visualization, and writing—original draft, review, and editing. R.J. Schulze: conceptualization, formal analysis, investigation, methodology, and writing—review and editing. D. Ding: investigation. C.A. Casey: funding acquisition, investigation, resources, and writing—review and editing. C. Weihl: methodology, resources, and writing—original draft, review, and editing. M.A. McNiven: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, software, supervision, validation, visualization, and writing—original draft, review, and editing.