Loss of TMEM55B modulates lipid metabolism through dysregulated lipophagy and mitochondrial function

We also created a Tmem55b knockout (KO) mouse (Supplementary Fig. S1E) and confirmed the absence of hepatic Tmem55b transcript and protein (Fig. 1D, E). Mice were fed a GAN diet starting at 6 weeks old for 12 weeks (Supplementary Fig. S1F). Compared to Tmem55b flox/flox (e.g., wildtype allele, WT) littermates, both male and female Tmem55b KO mice had significantly increased liver fibrosis (p < 0.05) (Fig. 1F, Supplementary Fig. S2A). Male Tmem55b KO mice had increased liver weight (p < 0.01) and trends of greater plasma ALT and hepatocyte ballooning (Supplementary Fig. S2B). Tmem55b KO livers from female mice had 3-fold greater levels of ER and oxidative stress markers (Fig. 1G). Male and female Tmem55b KO mice had increased plasma total cholesterol and non-HDL cholesterol (Supplementary Fig. S3), consistent with previously published effects of Tmem55b knockdown [14].

Transcript differences detected in livers from MASLD cases compared to healthy controls may be a consequence of, rather than a cause of, the disease. In contrast, differences observed in induced pluripotent stem cell-derived hepatocyte-like cells (iPSC-Heps) may reflect genetic contributors to disease. Using RNAseq data (GEO# GSE138312) from a previously described cohort of iPSC-Heps from MASH cases and healthy controls [15], we found that iPSC-Heps from MASH cases had lower levels of PNPLA3 and TM6SF2 (Fig. 1H), which contain loss-of-function alleles that promote MASH development [16, 17]. Similar levels of albumin (ALB), a hepatocyte-specific gene, indicative of equivalent differentiation efficiency, were observed between the two groups. Importantly, we found reduced TMEM55B transcript levels in the MASH iPSC-Heps (P < 0.01, Fig. 1H).

To further examine lipophagy, we tested the effect of TMEM55B knockdown on lysosomes and lipid droplets colocalization. In the control treated cells, the lipid droplet marker perilipin-2 (PLIN2) [20], revealed distinct circles around the periphery of lipid droplets (Fig. 2E). Upon TMEM55B knockdown, PLIN2 levels were increased, the peripheral circular structure was lost, and there was greater colocalization with lysosomes (1.5-fold, p < 0.05, Fig. 2E), indicative of the formation of lipid-filled lysosomes (aka lipolysosomes) [21]. Similar results were observed after an oleate challenge (Supplementary Fig. S4F). We confirmed this increase in lysosomal lipid accumulation (Supplementary Fig. S4G, S4H) by incubating cells with BODIPY-labeled C12-FA or C16-FA and staining with lysosome-associated membrane protein 1 (LAMP1) or LysoTracker. TMEM55B knockdown increased lysosome intensity levels (1.2-fold, p < 0.01) and colocalization of neutral lipids and lysosomes (1.5-fold, p < 0.001, Supplementary Fig. S4G, S4H). Increased lipolysosome formation with TMEM55B knockdown was confirmed in HepG2 cells using electron microscopy (Fig. 2F). We also observed changes in lipolysosome positioning. Lipolysosomes appeared primarily at the cell periphery in control cells (Supplementary Fig. S4G), versus within the perinuclear region after TMEM55B knockdown. Lastly, TMEM55B knockdown led to greater colocalization of C12-FA to LC3B (Fig. 2G). These findings indicate that loss of TMEM55B leads to greater interactions between the lipid droplet, autophagosome, and lysosome, which would typically suggest increased lipophagy. However, as this conclusion contradicts our observations of impaired lysosomal activity and autophagy, we further examined lipophagic flux.

To determine whether the reduction in mitochondrial fatty acid β-oxidation could be due to impaired mitochondrial function or/and reduced levels of mitochondria, we evaluated the effects of TMEM55B knockdown on mitochondrial membrane potential, essential for the production of ATP [28]. TMEM55B knockdown reduced TMRE (tetramethylrhodamine ethyl ester, a dye that accumulates only in the membrane of active mitochondria) levels by 35% in HepG2 cells (p < 0.05, Fig. 4C). We also observed reduced mitochondrial volume by 61% with 3D-live cell imaging using MitoTracker Deep Red in primary hepatocytes from male Tmem55b KO mice and WT controls (p < 0.0001, Fig. 4D), and decreased mitochondria intensity by 42% in HepG2 cells stained with MitoGreen (p < 0.001, Supplementary Fig. S6E). Using live cell imaging with 3D reconstruction, we observed highly fragmented mitochondrial networks, a sign of mitochondria dysfunction [29], in primary hepatocytes from Tmem55b KO mice (Fig. 4D). Tmem55b KO led to a significantly lower percentage of lysosomes containing mitochondria (p < 0.001, Fig. 4D), while knockdown caused a 28% decrease in LC3B and MitoTracker colocalization (p < 0.05, Supplementary Fig. S6F), indicative of impaired mitophagy. We confirmed that TMEM55B knockdown led to a drastic (>80%) reduction in mitophagy using the MT-mKeima-Red assay to directly visualize mitophagy (p < 0.0001, Fig. 4E, Supplementary Fig. 7).

Mitochondrial damage often leads to oxidative stress, and not surprisingly, TMEM55B knockdown led to significantly higher levels of oxidative stress using a generalized measure of oxidative stress and specific measures of mitochondrial oxidative stress (Supplementary Fig. S6G, S6H). Importantly, palmitate (1 mM) induced significantly higher mitochondria and total cellular oxidative stress Tmem55b KO primary hepatocytes compared to WT (Fig. 4F, Supplementary Fig. S6H, S6I), an effect that was mitigated upon incubation with 20 µM urolithin A, which stimulates mitophagy and maintains mitochondrial function [30, 31]. Together, these findings suggest that loss of TMEM55B results in mitochondrial dysfunction through impaired mitophagy.

The lack of Tmem55b led to hepatic steatosis in vivo (Fig. 1A–C), an effect we confirmed after TMEM55B knock-down in Huh7 cells (Supplementary Fig. S8B), incubated with 1 mM oleic acid and stained cells with Nile Red, a neutral lipid dye. After 24 hours there was 29% greater Nile red staining, 70% larger lipid droplet area per cell (p < 0.01, Fig. 5D), and fewer smaller lipid droplets with TMEM55B knockdown (p < 0.001, Fig. 5D), which is consistent with reports that only smaller lipid droplets can be engulfed by lysosomes through lipophagy [32]. Similar effects were observed in HepG2 cells incubated with BODIPY C12-FA (Supplementary Fig. S8C, S8D) and in Tmem55b KO primary murine hepatocytes, an effect that was enhanced after incubation with 1 mM palmitate and partially rescued with urolithin A (Supplementary Fig. S8E).

Given the role of TMEM55B in lysosome motility [12, 14, 33], which we confirmed here in HepG2 cells (Supplementary Fig. S4C), we tested whether the observed cellular steatosis was dependent on the ability of TMEM55B to modulate lysosome movement. TMEM55B enables lysosome transport by recruiting JIP4, a dynein scaffold, to the lysosome [12]. We created a HepG2 JIP4 knockout (KO) cell line using CRISPR-Cas9, which led to an 86% reduction in JIP4 transcript levels compared to HepG2 cells treated with a non-targeting control (NTC) gRNA (Supplementary Fig. S8F). While TMEM55B knockdown led to increased lipid accumulation in the NTC gRNA-treated cells incubated with BODIPY C12-FA for 24 hours, this increase was not observed in the JIP4-KO cell line (Fig. 5E). A similar lack of lipid accumulation was also seen after TMEM55B/JIP4 siRNA-mediated double knockdown in HepG2 cells using confocal microscope and flow cytometry (Fig. 5F, Supplementary Fig. S8G)

Lysosomes normally fuse with endosomes or autophagosomes in the perinuclear region [34], after which lysosomal hydrolytic enzymes degrade the cargo [35]. We previously reported that in human hepatoma cell lines incubated with oleic acid, TMEM55B knockdown led to enlarged perinuclear lysosomes [14]. Here, we show that these lysosomes are filled with lipids. Lipolysosome formation could be attributed to increased delivery of lipids to the lysosome, impaired degradative activity within the lysosome, and/or impaired release of lysosomal contents. Loss of TMEM55B led to greater colocalization of lipids with LC3B, consistent with increased macrolipophagy, an autophagosome-mediated process [36], as well as greater colocalization and direct contacts between lipid droplets and lysosomes, indicative of greater microlipophagy, an autophagosome-independent process of direct lipid extrusion from lipid droplets to lysosomes [37]. We also observed reduced lysosomal acid lipase activity, the enzyme that degrades triglycerides into free fatty acids [19], suggesting that impaired lysosomal function also contributes to lipolysosome formation. Our pulse-chase assays demonstrated greater movement of C12-FA-containing particles from the lysosome to the mitochondria. As TMEM55B is critical for the repair of lysosomes [38], this trafficking may be attributed to dysfunctional lysosomes, allowing the release of partially degraded lipid intermediates instead of free fatty acids. These findings suggest that despite increased lipid delivery to the lysosome, impaired lysosomal acid lipase activity upon loss of TMEM55B disrupts lipophagic flux.

Loss of TMEM55B reduced the number and size of mitochondria, generated fragmented mitochondrial networks, decreased mitochondrial membrane potential, and dramatically diminished mitochondrial β-oxidation activity. This is likely due to both increased mitochondrial damage as well as reduced repair. Our findings suggest that TMEM55B knockdown/knockout leads to the delivery of partially degraded lipids (and potentially toxic lipid intermediates) to mitochondria, which can cause mitochondrial damage [39, 40]. These damaged mitochondria cannot be efficiently removed due to the impairment of mitophagy. While the loss of TMEM55B appears to disrupt lipophagy at the point of degradation within the lysosome, mitophagy appears to be inhibited at a much earlier stage based on reduced interactions between mitochondria with autophagosomes or lysosomes. Urolithin A, which promotes mitophagy through the recruitment of autophagosomes to mitochondria [41], was able to rescue the increase in palmitate-induced mitochondrial oxidative stress and lipid accumulation observed in the absence of TMEM55B.

It is notable that TMEM55B knockdown decreases LC3B-mitochondrial colocalization, while simultaneously increasing LC3B colocalization with lipid droplets. Although the mechanism underlying this shift in autophagosome targeting is unclear, TMEM55B depletion has been reported to mimic a fasted state [13]. Under these conditions, cells may prioritize mobilization of free fatty acids from lipid droplets to meet immediate energy demands, leading to preferential autophagosome engagement with lipid droplets over damaged mitochondria. However, since loss of TMEM55B also impairs lipophagic flux, this could hinder recycling of autophagy machinery (including the lysosome) and reduce the availability of autophagosomes to target mitochondria. This possibility is consistent with the known role of TMEM55B as a molecular sensor that links autophagic flux and lysosomal repair during oxidative stress [38].

Our finding that loss of TMEM55B disrupts lipophagic flux and mitophagy and promotes MASLD is consistent with their established roles in MASLD. However, our observations highlight the importance of monitoring the entire lipophagic process, and not discrete measures. As the absence of TMEM55B increases lipid droplet-lysosome colocalization and causes greater delivery of FA-containing particles first to the lysosome and subsequently to the mitochondria, these results could be interpreted as increased lipophagy, which should be protective against MASLD. However, due to the impaired ability of lysosomes to degrade lipid cargo, TMEM55B knockdown/knockout ultimately causes lipid accumulation and promotes MASLD. Notably, excess lipolysosomes have been documented in MASLD patients [21, 42], where lipolysosome formation is positively correlated with the MASLD activity score and fibrosis stage [43]. The impact of TMEM55B loss on MASLD is further promoted by the disruptions in mitophagy, leading to severely disrupted mitochondrial networks, ultimately causing oxidative and ER stress. Stimulating mitophagy was able to reverse FA-induced oxidative stress seen with loss of TMEM55B, underscoring the likelihood that the more severe MASH observed in our murine models of Tmem55b knockdown and KO is attributed to impaired mitochondrial health [44 –46].

TMEM55B impacts cellular processes that are critical for many aspects of human health beyond MASLD. For example, lipolysosomes formation, accumulation of partially degraded lipids, and impaired mitochondrial turnover are features associated with lysosomal lipid storage, neurodegenerative, and atherosclerotic diseases, and TMEM55B-mediated disruptions in autophagy and mitochondrial quality control could contribute to the pathophysiology of these conditions. Given our prior findings that loss of TMEM55B increases plasma non-HDL cholesterol [14, 47], these effects may act synergistically to promote atherosclerosis. Thus, TMEM55B may represent a broader regulatory node linking lipid metabolism, autophagic flux, and organelle homeostasis in diverse pathologic settings.

All assays were performed in triplicate using 100 ng cDNA on an ABI PRISM 7900 Sequence Detection System using TaqMan and SYBR Green qPCR assays.

Protein expression levels were measured by immunoblot as described [14].

Cell culture and transfection were performed as previously described [14].

For all assays, fluorescence intensity was quantified by the BD LSRFortessa™ Cell Analyzer as the median fluorescence values of 10,000 gated events.

Cells were incubated with 1 mM BODIPY 558/568 C12 (Red C12, D3835, Invitrogen) or 2 mM BODIPY FL C12 (Green, D3822, Invitrogen) in duplicate wells. After 16 hr, one well was collected as timepoint "HR 0" and the other was chased with DMEM and collected after 24 hr as timepoint "HR 24". Mitochondria were labeled with 100 nM MitoTracker Deep Red FM for 30 min, LDs were labeled with 1 µM BODIPY 493/503 for 30 min, and lysosomes were labeled with 75 nM LysoTracker Red DND-99 for 1 hr. Cells were imaged with the Zeiss LSM 710 by confocal microscopy.

Cells were transfected with mKeima-Red-Mito-7 (Plasmid #56018, Addgene) to monitor mitophagy in live cells. Cells were incubated with 0.2 μM LysoTracker (DND-99, Life Technologies) for 15 min, 1 μM BODIPY C12 (Green, Life Technologies) for 30 minutes, 0.2 μM MitoTracker Deep Red (Life Technologies) for 15 minutes, and 5 µM MitoSOX Green for 10 minutes, and imaged on a spinning disk confocal microscope.

Mitochondrial oxygen consumption rate (OCR) was measured on a Seahorse Bioscience XFe96 extracellular flux Analyzer as described [48].

Animal studies were performed with approval from the Institutional Animal Care and Use Committee at the University of California, San Francisco, cared for in accordance to the recommendations of the Panel on Euthanasia of the American Veterinary Medical Association, and housed in AAALAC-accredited facilities. Animal studies comply with the ARRIVE guidelines.

Tmem55b floxed mice were created by inserting loxP sites flanking Tmem55b exons 1 to 6 in a C57BL6/N background. The Tmem55b^fl/+ mice were backcrossed to the C57BL/6 J strain and crossed with Sox-2-Cre mice (Strain #:008454, JAX) to generate Tmem55b^fl/- mice. These mice were intercrossed to generate whole-body Tmem55b knockout (KO) and littermate Tmem55b flox/flox control (referred to in figures as "wildtype allele" or "WT"). We have previously described an ASO-mediated Tmem55b knockdown mouse model [14]. Six-week-old animals were randomly selected for i.p. injection with 25 mg/kg body weight/week of ASO targeting Tmem55b or a non-targeting control (Ionis Pharmaceuticals) and fed a GAN Diet (n = 10/treatment). Mouse liver was embedded in Tissue-Tek or paraffin, sectioned, and stained in Oil Red O solution, H&E, Sirius Red, and Masson Trichrome. Slides were assessed by a pathologist blinded to the sample identity.

RNAseq libraries were prepared from hepatic RNA from mice and sequenced. Sequence fragments were aligned to the mouse GRCm39 genome and GENCODE transcriptome, adjusted for library size, and analyzed using DESeq2 [49]. RNAseq data from iPSC-derived hepatocyte-like cells were downloaded from GEO (GSE138312) from a previously described MASLD cohort [15]. Gene expression count data were adjusted for library size and variance stabilized.

For ex vivo and in vitro experiments, representative results are reported. Data are shown as mean ± standard error of the mean (SEM). Grubb's test for outliers was used to identify statistical outliers. Continuous variables for two groups were compared using Student's t-tests. Continuous variables for more than two groups were compared using one-way analysis of variance (ANOVA) with Tukey's post hoc test. Analyses were performed using GraphPad Prism 7 software (GraphPad Software, Inc. La Jolla, CA, USA). P values < 0.05 were considered statistically significant. No statistical analysis was employed to determine sample sizes. Sample sizes were established according to previous studies or our experimental experience.