Spatial regulation of UBXD8 and p97/VCP controls ATGL-mediated lipid droplet turnover

In unstimulated HeLa cells, endogenous UBXD8 was largely restricted to the ER, but was redistributed to LDs after treatment with the lipogenic unsaturated fatty acid oleate (Fig. 1A). In contrast, overexpressed S-peptide–tagged UBXD8 (UBXD8-S) was strongly localized to LDs irrespective of the presence of oleate (Fig. 1B). These findings indicate that the ER membrane has a limited capacity to retain UBXD8, possibly owing to a limiting ER-retention factor, and that excess UBXD8 partitions by default to LDs. We considered the possibility that UBAC2, an ER-localized protein that we recently identified as a UBXD8 interactor (10), could serve such a role in restricting UBXD8 trafficking. UBAC2 is an ER-localized protein with six potential transmembrane domains that identify it as a member of the rhomboid pseudoprotease family (10, 16). Unlike UBXD8, UBAC2 partitioned with dense fractions on sucrose gradient fractionation of cell lysates (Fig. S1A) and remained strictly localized to the ER after oleate treatment (Fig. S1 B and C).

If UBAC2 functions as a limiting ER-retention factor for UBXD8, then it should be possible to reverse the predominant LD localization of overexpressed UBXD8 by increasing the abundance of UBAC2. Indeed, we found that overexpression of UBAC2-myc completely suppressed both the localization (Fig. 1C) and biochemical partitioning (Fig. 1D) of overexpressed UBXD8 to LDs. This effect was specific for UBXD8, in that UBAC2 overexpression did not affect the trafficking of other integral membrane LD-associated proteins (Fig. S2A), and that overexpression of other ER-resident integral membrane proteins found in complex with UBXD8 (10) did not influence the partitioning of UBXD8 (Fig. S2B). Knockdown of endogenous UBAC2 (Fig. 1E) caused a fraction of endogenous UBXD8 to redistribute into LD-enriched fractions (Fig. 1F) and to LDs (Fig. 1G), indicating that the effects of UBAC2 expression on UBXD8 trafficking are not simply artifacts of overexpression, but reflect a bona fide interaction that contributes to the trafficking of UBXD8. Similar partitioning of endogenous UBXD8 between the ER and LDs and ER retention by UBAC2 was observed in Huh7 hepatocytes (Fig. S3 A–E). In contrast, UBXD8 was not detected in LDs in 3T3-L1 adipocytes (Fig. S3F) (17), and its expression was not induced on differentiation (Fig. S3G). These data demonstrate that UBAC2 levels can be experimentally manipulated to selectively control the UBXD8 levels in LDs irrespective of cellular metabolic status.

Overexpression of UBXD8-S led to striking increases in both the size and number of LDs (Fig. 2 A and B), an effect that could be related to elevated UBXD8 levels in the ER, LDs, or both compartments. However, our findings that coexpression of UBAC2 with UBXD8 completely suppressed LD enhancement (Fig. 2 A and B) and that knockdown of endogenous UBAC2 also increased cellular LD content (Fig. 2C) argue strongly that the observed effects on LD stores are due to LD-localized UBXD8. An increase in LD content was not observed on overexpression of UBXD8(∆UBX)-S (Fig. 2 A and B), indicating that UBXD8’s effects on LD size and abundance require its UBX domain. Indeed, we found that recruitment of endogenous p97/VCP to LDs was dependent on the UBX domain of UBXD8 (Fig. 2D), and that this recruitment was suppressed by coexpression of UBAC2 (Fig. S4A), suggesting that UBXD8 is the principal p97/VCP binding site on LDs. A role for p97/VCP in LD regulation is also supported by the observation that expression of a dominant-negative p97/VCP variant (18) strongly suppressed the ability of oleate to increase LD content (Fig. 2E). These data indicate that UBXD8 increases LD size and abundance through recruitment of p97/VCP ATPase activity to the LD surface.

The increases in LD size and abundance induced by UBXD8 recruitment of p97/VCP to LDs could be related to either the promotion of LD biogenesis or the suppression of LD turnover. To evaluate these two possibilities, we used a “pulse-chase” protocol in which cells were initially depleted of LDs with triacsin C, a drug that blocks TAG synthesis by inhibiting fatty acyl-CoA synthetase (19), and then treated briefly with oleate in the absence of drug to induce synchronous LD production (Fig. 3 A and B). Neither UBXD8-S overexpression nor UBAC2-S overexpression affected LD area (Fig. 3B), indicating that oleate-induced LD biogenesis is unaffected by UBXD8 abundance. To assess the role of UBXD8 in LD degradation, we exposed cells treated as above to a second round of triacsin C to synchronously promote LD degradation (Fig. 3 A and C). Cells overexpressing UBXD8-S displayed a significantly reduced rate of LD turnover (Fig. 3C), whereas cells depleted of endogenous UBXD8 from LDs by overexpression of UBAC2-S demonstrated an increased rate of LD turnover (Fig. 3C). The inhibitory effect of UBXD8-S on LD turnover is not simply related to UBXD8 overexpression, but requires the physical presence of UBXD8 on LDs, as demonstrated by the suppression of the effect of UBXD8 overexpession on LD turnover by simultaneous UBAC2 overexpression (Fig. 3D).

Given that LD volume is composed almost entirely of neutral lipids, the most direct way for overexpressed UBXD8 to slow the decrease in LD content after inhibition of de novo TAG synthesis with triacsin C is by inhibiting the rate of TAG hydrolysis. Indeed, we found that UBXD8 overexpression significantly inhibited the decrease in ¹⁴C-TAG after triacsin C treatment of cells pulse-labeled with ¹⁴C oleate, whereas overexpression of UBAC2 accelerated TAG hydrolysis (Fig. 3E). We conclude that UBXD8 acts in LDs to inhibit TAG hydrolysis.

The most direct mechanism through which UBXD8 can inhibit lipolysis is by inhibiting the activity of ATGL, the lipase that catalyzes the rate-limiting step in TAG hydrolysis (20–22). In support of this, we found that LD content was reduced between threefold and fourfold in oleate-treated cells overexpressing ATGL-GFP compared with GFP-overexpressing controls, and that this effect was largely abrogated by simultaneous overexpression of UBXD8-S (Fig. 4A), suggesting that UBXD8 antagonizes the effect of ATGL in LDs. To determine whether UBXD8 requires ATGL to influence cellular LD content, we analyzed the effect of UBXD8 expression in WT or ATGL^−/− mouse embryonic fibroblasts (MEFs). As observed in HeLa cells, expression of UBXD8 in WT MEFs increased LD content (Fig. 4B). In contrast, UBXD8 expression had no effect on LD content in ATGL^−/− MEFs (Fig. 4B), despite the fact that MEFs lacking this lipase retain the ability to synthesize LDs in response to oleate (Fig. S5). These data support the conclusion that UBXD8 inhibits LD turnover by negatively modulating ATGL.

Immunoprecipitation of UBXD8 complexes from detergent-solubilized LDs coprecipitated ATGL, but not the LD-associated proteins CGI-58 or PLIN3 (also known as TIP47) (Fig. 4C), indicating that endogenous UBXD8 and ATGL interact with each other either directly or indirectly. To confirm this interaction, we used a bimolecular fluorescence complementation assay in which UBXD8 fused to the N terminus of YFP (Yn-UBXD8) was coexpressed with a catalytically inactive ATGL variant (so as not to induce LD degradation) fused to the C terminus of YFP [ATGL(S47A)-Yc]. YFP fluorescence was observed only on coexpression of the two constructs (Fig. 4D), confirming an interaction between UBXD8 and ATGL in intact cells and indicating that the interaction between these two proteins is likely direct.

Considering that in the ER, UBXD8 is thought to recruit p97/VCP to facilitate the extraction and degradation of ERAD substrates, a plausible mechanism by which UBXD8 can negatively regulate ATGL activity is to promote its extraction from LDs and its degradation by the Ub-proteasome system (UPS) (5, 6, 10–12). Indeed, we found that ATGL is a short-lived (t_1/2 ∼45 min) protein that it is strongly stabilized (t_1/2 >12 h) by the proteasome inhibitor MG132 (Fig. 5 A and B) and, surprisingly, by oleate (Fig. 5 C and D). Stabilization of ATGL by oleate is not related to a general impairment of the UPS, given that degradation of the UPS substrates GFPu (cytosolic) and TCRα-GFP (TCR subunit α, ER) were unaffected by the presence of oleate (Fig. S6 A and B). Surprisingly, the steady-state abundance (Fig. 5C) and half-life of ATGL (Fig. 5D), as well as the association of ATGL with LDs (Fig. S6C), were unaffected by overexpression of UBXD8 or UBAC2, arguing against a role for UBXD8 in controlling the extraction, turnover, or abundance of ATGL.

An alternative mechanism by which UBXD8 might inhibit ATGL is that, through its association with p97/VCP, UBXD8 could promote the dissociation of ATGL from its endogenous activator CGI-58 (23), a mechanism consistent with the known segregase function of p97/VCP (13). Antibodies against endogenous UBXD8, but not CGI-58, largely depleted ATGL from detergent-solubilized LDs (Fig. 6A), indicating that the majority of ATGL in LDs isolated from oleate-treated cells is associated with UBXD8, not with CGI-58. In contrast, the amount of endogenous ATGL captured with CGI-58 antibodies was substantially increased by overexpression of UBAC2-S (Fig. 6B), a condition that restricts all detectable endogenous UBXD8 to the ER (Fig. S6D).

Finally, we used a bimolecular fluorescence complementation assay of split YFP-tagged ATGL and CGI-58 to directly interrogate the interaction between ATGL and CGI-58 (Fig. 6C). Whereas overexpression of UBXD8-S resulted in a small but significant decrease in fluorescence, UBAC2-S overexpression led to strongly enhanced fluorescence (Fig. 6D). Our data support the conclusion that UBXD8, together with p97/VCP, inhibits ATGL by promoting CGI-58 dissociation.

As described previously (27), after 16–24 h of treatment with 200 µM oleate, cells were incubated in hypotonic lysis buffer [20 mM Tris⋅HCl (pH 7.4), 1 mM EDTA] for 10 min, dounce-homogenized, and centrifuged at 1,000 × g for 10 min. The postnuclear supernatant was transferred to a new tube, adjusted to a final concentration of 20% (wt/vol) sucrose, and overlaid with 4 mL of hypotonic lysis buffer containing 5% (wt/vol) sucrose and then with 4 mL of hypotonic lysis buffer. After centrifugation at 28,000 × g for 30 min, the buoyant LD-enriched fraction was collected using a Beckman tube slicer, and the remaining fractions were collected by pipetting. Equivalent percentages of each fraction were measured by immunoblot analysis. Under these conditions, quantification of immunoblots indicate that ∼75% of the ER proteins employed as ER markers, UBAC2 and Derlin-2, sedimented and were present in the pellet fraction (Fig. S1B). Analysis of the pellet fraction underestimated the amount of ER-localized proteins by ∼25%.

Cells grown on poly-L-lysine–coated coverslips were fixed in 4% (wt/vol) paraformaldehyde, immunostained, and analyzed using a Zeiss Axiovert 200M fluorescence microscope. LDs were stained with BODIPY 493/503 (Invitrogen). ImageJ 1.42q was used to quantify LD number, size, and area. These experiments and analyses are described in more detail in SI Methods.

LD-enriched fractions were solubilized in a final buffer containing 20 mM Tris⋅HCl (pH 7.4), 150 mM NaCl, and 1% (vol/vol) Triton X-100 supplemented with Complete Protease Inhibitor Mixture (Roche). Endogenous immunoprecipitations were performed by incubation with antibody immobilized on agarose beads using the Direct Immunoprecipitation Kit (Thermo Fisher Scientific). Samples were separated by SDS/PAGE and evaluated by immunoblot analysis.

Bimolecular fluorescence complementation in HEK293 cells transiently transfected with split YFP constructs fused to the N terminus of YFP (Yn) or the C terminus of YFP (Yc), including ATGL-Yc, Yn-UBXD8, or Yn-CGI-58, was analyzed using a BD Biosciences LSRII flow cytometer, as described in SI Methods.