The E3 ligase MARCH5 is a PPARγ target gene that regulates mitochondria and metabolism in adipocytes

Abstract

Mitochondrial dynamics refers to the constant remodeling of mitochondrial populations by multiple cellular pathways that help maintain mitochondrial health and function. Disruptions in mitochondrial dynamics often lead to mitochondrial dysfunction, which is frequently associated with disease in rodents and humans. Consistent with this, obesity is associated with reduced mitochondrial function in white adipose tissue, partly via alterations in mitochondrial dynamics. Several proteins, including the E3 ubiquitin ligase membrane-associated RING-CH-type finger 5 (MARCH5), are known to regulate mitochondrial dynamics; however, the role of these proteins in adipocytes has been poorly studied. Here, we show that MARCH5 is regulated by peroxisome proliferator-activated receptor-γ (PPARγ) during adipogenesis and is correlated with fat mass across a panel of genetically diverse mouse strains, in ob/ob mice, and in humans. Furthermore, manipulation of MARCH5 expression in vitro and in vivo alters mitochondrial function, affects cellular metabolism, and leads to differential regulation of several metabolic genes. Thus our data demonstrate an association between mitochondrial dynamics and metabolism that defines MARCH5 as a critical link between these interconnected pathways.

INTRODUCTION

Mitochondria are essential organelles. Although their primary role in mammalian cells is generation of cellular energy, they also have crucial roles in maintaining processes such as calcium signaling, cell growth and division, apoptosis, and cell death (23). Historically, the bulk of studies investigating mitochondrial disease have been performed in tissues of high energy demand, including striated muscles and the brain. However, more recent studies have also demonstrated that manipulation of mitochondrial health and activity in adipose tissue can have profound effects on adipose tissue expansion and differentiation, a process largely regulated by the nuclear receptor peroxisome proliferator-activated receptor-γ (PPARγ) (35). PPARγ is the master regulator of adipogenesis and, consequently, modulates several gene sets necessary for adipocyte function, including those involved in mitochondrial homeostasis (40, 41). Indeed, the thiazolidinediones, or TZDs, such as pioglitazone, which are indicated for type 2 diabetes in humans, are ligands for PPARγ and have been shown to have beneficial effects on adipose tissue expansion and mitochondrial activity in individuals with obesity and type 2 diabetes (6, 22).

Mitochondrial integrity is fundamental to maintaining healthy cellular functions. Indeed, when mitochondria become dysfunctional or damaged, overall cellular health deteriorates, which can induce apoptosis and cell death (18, 27, 30, 46). Mitochondria continually divide and fuse through well-described mechanisms of fission (fragmentation) and fusion (elongation), including the removal of dysfunctional and unhealthy mitochondria through mitophagy, to regulate mitochondrial health (11, 13, 19). These processes, collectively known as mitochondrial dynamics, are part of a highly regulated and coordinated program whereby mitochondria enter a constant state of flux through processes such as biogenesis, fusion, fission, and mitophagy (1, 13). Several key proteins, including dynamin-related protein 1 (Drp1), dynamin-like 120-kDa protein [optic atrophy 1 (OPA1)], and mitofusins 1 and 2 (Mfn1/2), have been identified to regulate mitochondrial fission and fusion (8, 44). Mitochondrial dynamics are also important in maintaining the health of white adipose tissue (WAT), inasmuch as inhibition of autophagy or mitophagy has been shown to result in a severe reduction in WAT mass reminiscent of lipodystrophy (10, 32, 47). Improving mitochondrial health by increasing mitochondrial function has the potential to restore or improve the function of WAT. This was elegantly highlighted by Kusminski and colleagues, who demonstrated substantial improvement in metabolic health in a mouse model of obesity that specifically overexpressed the protein MitoNEET in adipose tissue (12).

Another protein more recently implicated in mitochondrial dynamics is membrane-associated RING-CH-type finger 5 (MARCH5) (3, 20, 37). The MARCH family proteins are RING-type transmembrane E3 ubiquitin ligases that regulate a range of cellular processes by ubiquitinating target proteins and facilitating their degradation (28). MARCH5 is the only member of the MARCH family localized to the mitochondria, where it is bound to the outer mitochondrial membrane, with the active E3 ligase domain protruding into the cytosol. MARCH5 has been implicated in the regulation of a variety of cellular processes, including apoptosis, viral signaling, endoplasmic reticulum-mitochondria binding, and calcium dynamics. However, the most well-documented role for MARCH5 is its ability to ubiquitinate and regulate several mitochondrial dynamics proteins, including Drp1, Mfn1/2, mitochondrial fission 1 protein (Fis1), and 49- and 51-kDa mitochondrial dynamics proteins (Mid49/51) (3, 33, 42, 43). Furthermore, recent studies have demonstrated that MARCH5 is a regulator of hypoxia-induced apoptosis and mitophagy via ubiquitination of the mitochondrial receptor FUN14 domain-containing protein 1 (FUNDC1), demonstrating a further link between MARCH5 and mitophagy pathways (2). Most studies on MARCH5 have been performed in neuronal cells and tissues, and there have been no investigations into the role of MARCH5 in metabolic tissues, including adipose tissue. Thus the current study aimed to investigate the role of MARCH5 in adipose tissue. We provide in vitro and in vivo evidence that MARCH5 expression is enriched in adipose tissue and is regulated by PPARγ and that modulation of MARCH5 in adipocytes regulates energy metabolism.

MATERIALS AND METHODS

Cell culture.

3T3-L1 cells were cultured in DMEM (catalog no. 11965-084, GIBCO) supplemented with 10% newborn calf serum and maintained in 5% CO₂ at 37°C. Cells were differentiated in DMEM supplemented with 10% FBS, 20 nM insulin, 50 nM GW1929, 0.5 mM IBMX, and 1 µM dexamethasone for 48 h and then cultured in DMEM supplemented with 10% FBS and 20 nM insulin for 10 days. Wild-type (WT) and MITOL (MARCH5) knockout (KO) mouse embryonic fibroblasts (MEFs; a gift from Shigeru Yanagi, Tokyo University, Japan) were cultured in DMEM supplemented with 10% FBS and maintained in 5% CO₂ at 37°C. Primary adipocytes were isolated and cultured from inguinal fat pads, as previously described (5), from floxed (fl/fl) or adipose-specific PPARγ-KO (fl/fl aP2-cre) mice. Confluent preadipocytes were induced to differentiate in the medium described above to induce 3T3-L1 adipocytes and imaged or harvested for protein/RNA (see below).

Short-hairpin RNA knockdown and MARCH5 expression constructs.

MARCH5-depleted [short-hairpin (sh) MARCH5] and control [luciferase (shLuc)] 3T3-L1 adipocytes were generated using commercially sourced lentiviral particles (MISSION shRNA lentiviral particles, Sigma) expressing shRNAs designed via the Broad Institutes RNAi Consortium database (TRCN0000246413 and SHC002V, respectively). Briefly, 50,000 undifferentiated 3T3-L1 cells were plated and exposed to lentiviruses (multiplicity of infection = 20) in the presence of Polybrene (10 µg/ml) for 24 h and then transferred to normal growth medium (DMEM + 10% newborn calf serum) overnight. On the following day, positive cells stably expressing shRNAs were selected in puromycin (4 µg/ml) for 4 days and then returned to normal growth medium before differentiation. Adenoviruses for human MARCH5 and green fluorescent protein (GFP) were generated using the ViraPower kit (Invitrogen) according to the manufacturer’s instructions. Human MARCH5 cDNA was cloned from human embryonic kidney (HEK-293A) cells. Adenoviruses were titered according to expression of their 3′ V5 tag using an in-house quantitative PCR (qPCR) assay (see Table 1), and equivalent titers were added to differentiated 3T3-L1 cells at day 7. Cells were studied on day 10.

Table 1.

Primers for quantitative PCR

	Primer Sequences
Gene	Forward (5′–3′)	Reverse (5′–3′)
mMfn2	ATTGATCACGGTGCTCTTCC	GTCCTGGACGTCAAAGGGTA
mDnm1L	TGCCTCAGATCGTCGTAGTG	CGTGGACTAGCTGCAGAATG
mPparg	GTGCCAGTTTCGATCCGTAGA	GGCCAGCATCGTGTAGATGA
mPparg2	GTTTTATGCTGTTATGGGTG	GTAATTTCTTGTGAAGTGCTCATAG
mMarch5	TTGGACAGCCGTGACTTATG	AGGGTCAGCTCGCTCCAT
hMARCH5	CAAAAGCCTGTCCATTTGC	CCAGACCTTCTTTATGACCTACAAC
mPgc1a	TGAGGACCGCTAGCAAGTTT	TGAAGTGGTGTAGCGACCAA
mPgc1b	CTGAGTCAAAGTCACTGGCG	GCTCTCGTCCTTCTTCCTCA
mCebpa	TGGACAAGAACAGCAACGAG	GTCACTGGTCAACTCCAGCA
mCd36	TTGTACCTATACTGTGGCTAAATGAGA	CTTGTGTTTTGAACATTTCTGCTT
mLdlr	AGGCTGTGGGCTCCATAGG	TGCGGTCCAGGGTCATCT
mSlc27a1	GACAAGCTGGATCAGGCAAG	GAGGCCACAGAGGCTGTTC
mCptc1c	GTGGACAAGCACCAGGCTCT	TGGACCTGGGTCAGGAAGGG
mCpt2	ACAGTGTGGGCGAGCTTCAG	GGCTGCTGCCAGATACCGTAG
mDgat2	CACTCCAGTGGGTTCCGTGT	TTTGGCCTTGACCCTTCGCT
mBcl2	AGGGTCTTCAGAGACAGCCA	AGTACCTGAACCGGCATCTG
mCytc	CAGCTTCCATTGCGGACAC	CGCTGACAGCATCACCTTTC
mFundc1	ATTGTAATGGGTGGCGTGAC	CATAGCCACTGTGACTGGCA
mFasn	TGCTCCCAGCTGCAGGC	GCCCGGTAGCTCTGGGTGTA
mPnpla2	GAAATTGGGTGACCATCTGC	TGGGTAGGGCCTCACTGTAG
mCrat	AACTGGCTGTCCGAGTGGTG	TGGCAGCAAACCGAAGCTGA
mCrot	GACTTCATGGACGCCCTGGT	ACCTGACGGCCTCCACTGTA
mGfp	CAGGAGCGCACCATCTTCTT	CTTGTGCCCCAGGATGTTG
mCidea	TGCTCTTCTGTATCGCCCAGT	GCCGTGTTAAGGAATCTGCTG
mPrkaca	CCCACCCTTCTTCGCTGACC	GTTCCGCAGCAGGTCCTTCA
mPrkab1	CGGGCATCTCTTGTGACCCA	GCACTGAGCACCATCACTCCA
mPrkab2	TCTCCTTGTACGCCGAACAGC	CCCGCTCGCTGGTAGTGTTT
mFabp4	GGTCGACTTTCCATCCCACTT	TTCGATGAAATCACCGCAGA
mLpl	TTTGTGAAATGCCATGACAAG	CAGATGCTTTCTTCTCTTGTTTGT
mLipe	GCGCTGGAGGAGTGTTTTT	CCGCTCTCCAGTTGAACC
pAD-V5	CCTAACCCTCTCCTCGGTCT	TCTGTCTTTTTATTGCCGTCAT
mPpia	AGCCAAATCCTTTCTCTCCAG	CACCGTGTTCTTCGACATCA
mRplp0	ACCCTGAAGTGCTCGACATC	ATTGATGATGGAGTGTGGCAC

m, Mouse; h, human; p, plasmid/synthetic; Mfn2, mitofusin 2; Dnm1L, dynamin 1-like; Pparg, peroxisome proliferator-activated receptor-γ; March5, membrane-associated RING-CH-type finger 5; Pgc1a and Pgc1b, peroxisome proliferator activated receptor, gamma, coactivator 1 alpha and 1 beta, respectively; Cebpa, CCAAT/enhancer-binding protein; Cd36, cluster of differentiation 36; Ldlr, low-density lipoprotein receptor; Slc27a1, solute carrier family 27; Cptc1c, carnitine palmitoyltransferase 1c; Cpt2, carnitine O-palmitoyltransferase 2; Dgat2, diacylglycerol O-acyltransferase 2; Bcl2, B-cell lymphoma 2; Cytc, cytochrome c; Fundc1, FUN14 domain-containing protein 1; Fasn, fatty acid synthase; Pnpla2, patatin-like phospholipase domain-containing protein 2; Crat, carnitine O-acetyltransferase; Crot, carnitine O-octanoyltransferase; Gfp, green fluorescent protein; Cidea, cell death activator A; Prka, 5′-AMP-activated protein kinase; Fabp4, fatty acid-binding protein 4; Lpl, lipoprotein lipase; Lipe, lipase E; AD-V5, Bovine GH poly and V5 epitope; Ppia, cyclophilin A; Rplp0, ribosomal protein-large-p0.

Hybrid Mouse Diversity Panel, Metabolic Syndrome in Men, and human biopsy studies.

Microarray data generated from WAT of mice in the Hybrid Mouse Diversity Panel (HMDP) were correlated with other transcripts and strain phenotypes, as previously described (25). Correlations were analyzed as a midweight bicorrelation, and all data were significant at a false discovery rate (FDR) of 5%. Metabolic Syndrome in Men (METSIM) data were generated by microarray analysis of adipose tissue biopsies from 770 male participants, as previously described (4). Correlation analysis was performed against 23 cardiometabolic phenotypes at a FDR of 1%. For human samples used in Western blot analysis, subcutaneous adipose tissue biopsies were obtained using standard aseptic technique and local anesthesia (lignocaine). A 0.5- to 1-cm skin incision was made ~5 cm lateral to the navel/umbilicus, and a Bergstrom biopsy needle was passed through the incision to obtain ~1–2 cm³ of subcutaneous adipose tissue under suction. Ice-cold sterile saline was used to rinse blood from all biopsies, the connective tissue was removed, and cleaned adipose tissue was snap-frozen in liquid nitrogen for subsequent storage at −80°C until further analysis. Samples were grouped according to lean/obese and metabolic status and blotted for indicated proteins. Pathway analysis was performed using the Database for Annotation, Visualization, and Integrated Discovery (DAVID v6.8) hosted by the National Institute of Allergy and Infectious Diseases. Briefly, gene sets were entered as a gene list in DAVID. Background adjustment was enabled against the Mus musculus reference gene set, functional annotation clustering was performed with the classification stringency set at medium, and Benjamini-Hochberg-corrected P values are presented.

Metabolic flux assays.

All metabolic flux analyses were performed using the Seahorse XFe96 extracellular flux analyzer (Agilent, Santa Clara, CA). 3T3-L1 cells were plated in a 96-well Seahorse V3 polystyrene plate postdifferentiation for analyses at 1.5 × 10⁴ cells/well. Cells were incubated overnight at 37°C in 5% CO₂, and before the assay the cells were washed with oxygen consumption rate (OCR) assay medium (Seahorse XF base medium supplemented with 25 mM glucose, 1 mM glutamine, and 1 mM sodium pyruvate) or extracellular acidification rate (ECAR) assay medium (Seahorse XF base medium supplemented with 2 mM glutamine). Cells were then equilibrated in respective assay medium (175 µl/well) and incubated at 37°C in the absence of CO₂ for 30–60 min. The assay protocol consisted of repeat cycles of 3 min of mixing, 2 min of waiting, and 3 min of measuring, with OCR and ECAR measured simultaneously. Basal energetics were established after four of these initial cycles followed by a mitochondrial stress test or a glycolysis stress test. The mitochondrial stress test consists of sequential injections of the following compounds and three subsequent repeat measurements of OCR after each: the ATP synthase inhibitor oligomycin (1 µM), the proton ionophore carbonyl cyanide-4-(trifluoromethoxy) phenylhydrazone (FCCP, 1 µM), and the mitochondrial complex III and complex I inhibitors antimycin A (1 µM) and rotenone (1 µM), respectively. The glycolysis stress test measures ECAR and consists of sequential injections of the following compounds after four basal measurements: glucose (10 mM), oligomycin (1 µM), and 2-deoxy-glucose (50 mM), from which glycolysis and glycolytic capacity can be determined, inasmuch as ECAR is a proxy measure of glycolysis. Glycolytic capacity is the sum of the glycolytic rate and reserve glycolytic capacity. All analyses were performed with 8–23 replicates for each independent experiment. At the completion of each assay, cells were lysed, and protein concentration was determined using the Bradford method (Bio-Rad, Sydney, Australia) according to the manufacturer’s instructions.

Glucose uptake assay.

shLuc and shMARCH5 3T3-L1 cells were grown in 12-well plates to day 7 postdifferentiation, and glucose uptake was determined as previously described (7).

Oil red O.

shLuc and shMARCH5 3T3-L1 cells were grown in 12-well plates to day 10 postdifferentiation to allow for lipid loading. Cells were stained with oil red O, as previously described (39), and imaged using an inverted microscope (Olympus IX71).

SDS-PAGE and immunoblotting.

Cells were harvested and lysed in radioimmunoprecipitation assay buffer supplemented with protease and phosphatase inhibitors. Matched protein quantities were separated by SDS-PAGE and transferred to polyvinylidene difluoride membranes. Membranes were blocked in 3% skim milk for 2 h and then incubated with the appropriate primary antibody overnight at 4°C. After incubation, the membranes were probed with their respective horseradish peroxidase-conjugated secondary antibodies (anti-mouse/rabbit) in 3% skim milk for 2 h at room temperature and then visualized with chemiluminescence (Pierce). The following primary antibodies were used: PPARγ (1:2,000 dilution; Cell Signaling Technology), Mfn2 (1:2,000 dilution; Cell Signaling Technology), Drp1 (1:2,000 dilution; Cell Signaling Technology), oxidative phosphorylation (OXPHOS, 1:2,000 dilution; Mitosciences), porin (1:2,000 dilution; Mitosciences), histone-lysine-N-methyltransferase (SETDB1, 1:2,000 dilution; Abcam), pan 14-3-3 (1:5,000 dilution; Santa Cruz Biotechnology), and MARCH5/MITOL (1:5,000 dilution; a gift from S. Yanagi). Densitometry analyses were performed using the Bio-Rad Image Lab software, and all quantification results were normalized to their respective loading control (14-3-3). Approximated molecular weights of proteins were determined from a coresolved molecular weight standard (catalog no. 1610374, Bio-Rad).

Quantitative PCR and chromatin immunoprecipitation-sequencing analysis.

RNA was isolated from cells and tissues using RNAzol reagent and isopropanol precipitation. Moloney murine leukemia virus reverse transcriptase (Invitrogen) was used according to the manufacturer’s instructions to generate 1 µg of cDNA from RNA. Quantitative PCR was performed on 20 ng of cDNA using the SYBR green method on a real-time PCR system (ABI 7500) using primer sets as outlined in Table 1. Quantification of a given gene using qPCR was expressed in one of two ways: 1) relative mRNA expression compared with control was calculated after normalization to a housekeeping gene, cyclophilin A (Ppia) or ribosomal protein-large-p0 (Rplp0), using the ΔC_T method (where C_T is cycle threshold); or 2) gene expression was quantified as “normalized RNA quantity” by interpolation of adjusted C_T values from a log₁₀ dilution curve previously generated for a given gene and primer set. Primers were designed to span exon-exon junctions and tested for specificity using Basic Local Alignment Search Tool [BLAST; National Center for Biotechnology Information (NCBI)]. Amplification of a single amplicon was estimated from melt curve analysis, ensuring that only a single peak with an expected temperature dissociation profile was observed. Analysis of PPARγ (GSM340799), retinoic acid receptor-α (RXRα; GSM340805), and CCAAT enhancer-binding protein-α (C/EBPα; GSM678392) occupancy in proximity to the March5 gene in 3T3-L1 adipocytes after 6 days of differentiation was completed using Integrated Genome Browser and publicly available chromatin immunoprecipitation-sequencing (ChIP-seq) data sets deposited through NCBI (21, 29).

Confocal microscopy.

Cells were grown on glass coverslips and then stained with 150 nM MitoTracker deep red (Cell Signaling Technology) for 30 min. Cells were washed three times in 1× PBS, fixed in 4% paraformaldehyde in PBS for 30 min, and washed again three times in 1× PBS. Cells were visualized on a confocal microscope (Nikon A1r). Mitochondrial networking was quantified using the Mitochondrial Network Analysis (MiNA) macro plugin tool for Fiji ImageJ. Mitochondrial network size was calculated by quantification of the average number of mitochondrial branches in three different sections of individual cells (3 cells per group).

Animal experiments.

All animal experiments were approved by the Alfred Medical Research and Education Precinct Animal Ethics committee (E/1618/2016/B). MARCH5 and control adeno-associated virus (AAV6) vectors were generated by the Baker Institute AAV core, as previously described (17). Four-week-old male C57BL/6J mice were sourced from the Alfred Medical Research and Education Precinct Animal Centre and randomly selected for injection under isoflurane anesthesia with 50 µl of AAV solution at 1 × 10¹¹ vector genomes (vg) in PBS into contralateral inguinal regions (MARCH5 or control vector on one of either side). The mice were then fed a high-fat diet (HFD, 43% energy from fat; Specialty Feeds catalog no. SF04-001) for 4 wk, with access to food and water ad libitum. Tissues from ob/ob mice were collected and analyzed as previously described (5). All mice were housed at 22°C on a 12:12-h light-dark cycle.

Statistical analysis.

All animal and laboratory data are expressed as means ± SE. Human characteristic data from Western blot analysis are presented as means (SD). All statistical analyses were performed using PRISM7 software. Groups in cell culture experiments were compared by paired Student’s t-test. Animal studies were analyzed by two-way ANOVA. Human characteristics data were analyzed by two-tailed heteroscedastic t-test. P < 0.05 was considered statistically significant. Details for midweight bicorrelation analysis for HMDP and METSIM data sets can be found in their respective publications and at https://systems.genetics.ucla.edu/.

RESULTS

MARCH5 expression is reduced in obesity and correlates with PPARγ expression.

Initially, we aimed to determine if MARCH5 expression was correlated with measures of adiposity and obesity in mice and humans. Thus we analyzed data from a collection of genetically diverse mouse strains, known as the HMDP, that have demonstrated utility in identifying genetic drivers of complex traits (25, 26, 31). We explored data from the HMDP microarray expression profiles to analyze March5 transcript expression in WAT across 101 strains of HFD-fed HMDP mice (25). The heat map in Fig. 1A, demonstrates that March5 mRNA expression is strongly negatively correlated with several indices of adiposity, including fat mass, percent body fat, and weight of multiple fat pads, including retroperitoneal, gonadal, subcutaneous (femoral), and mesenteric (Fig. 1, A and B). These data also confirm that the majority of adiposity measurements in the HMDP strains positively correlate with each other and with indexes of metabolic dysregulation (plasma glucose and insulin levels), accurately replicating known phenotypes associated with metabolic syndrome. Consistent with MARCH5 being a mitochondrial protein and reduced in the setting of increased adiposity, we also reveal that several other mitochondrial genes [dynamin 1-like (Dnm1L), mitochondrial fission factor (Mff), Opa1, mitochondrial Rho GTPase 1 (Rhot1), and Mfn1] were negatively correlated with fat mass in the HMDP (Table 2). Additionally, to validate this interaction in an experimental model of obesity, we analyzed March5 expression in WAT of male C57BL/6J and ob/ob mice, which further demonstrated that March5 expression was reduced in the setting of increased adiposity (Fig. 1C).

Fig. 1.

Membrane-associated RING finger (C3HC4) 5 (March5) expression is negatively correlated with obesity phenotypes in rodents and humans. A: heat map depicts correlations between March5 mRNA expression and adiposity/metabolism traits from mouse strains of the Hybrid Mouse Diversity Panel (HMDP, n = 101 strains). MRI, magnetic resonance imaging; BF, background backfill; FFA, free fatty acid; Chol, cholesterol; TG, triglycerides; FFP, femoral (subcutaneous) fat pad; GFP, gonadal fat pad; MFP, mesenteric fat pad; RFP, retroperitoneal fat pad; %, tissue weight as percentage of total body weight. B: dot plot and linear regression of white adipose tissue (WAT) March5 mRNA expression with fat mass as determined by EchoMRI in 101 strains of the HMDP (each dot represents an individual strain). C: March5 gene expression in WAT from wild-type (WT) and ob/ob mice, normalized to cyclophilin A (Ppia) gene expression. AU, arbitrary units. D: dot plot and linear regression of WAT March5 mRNA expression with fat mass in 770 men from the Metabolic Syndrome in Men (METSIM) study (each dot represents 1 individual). E: Western blots for MARCH5, peroxisome proliferator-activated receptor-γ (PPARγ), and 14-3-3 (loading control) from subcutaneous WAT biopsies from lean normal-glucose-tolerant (NGT) and obese impaired-fasting-blood-glucose (iFBG) men. F and G: densitometric quantification of MARCH5 and PPARγ protein levels, normalized to 14-3-3. Values are means ± SE. *P < 0.05 vs. WT or lean.

Table 2.

Midweight bicorrelation analysis of other mitochondrial transcripts negatively correlated with fat mass in HMDP (mouse) and METSIM (human) databases

Gene	r	P Value
	Mouse
Dnm1L‡	−0.323	1.42E−03
Mff‡	−0.49	1.01E−05
Opa1‡	−0.269	8.50E−03
Rhot1‡	−0.353	4.52E−04
Mfn1‡	−0.501	2.36E−07
Pparg*	−0.309	2.32E−03
Lep†	0.708	1.07E−15
	Human
DNM1L‡	No human probe
MFF‡	−0.204	4.01E−03
OPA1‡	−0.217	2.13E−03
RHOT1‡	−0.428	3.07E−10
MFN1‡	−0.203	4.05E−03
PPARG*	−0.453	2.04E−11
LEP†	0.383	2.51E−08

HMDP, Hybrid Mouse Diversity Panel; METSIM, Metabolic Syndrome in Men; r, bicorrelation; Dnm1L/DNM1L, dynamin 1-like; Mff/MFF, mitochondrial fission factor; Opa1/OPA1, optic atrophy 1; Rhot1/RHOT1, Rho GTPase 1; Mfn1/ MFN1, mitofusin 1; Pparg/PPARG, peroxisome proliferator-activated receptor-γ; Lep/LEP, leptin.

^*Similar negative correlation with Pparg/PPARG expression.

^†Expected highly positive correlation with Lep/LEP expression.

^‡Mitochondrial transcripts negatively correlated.

Given this correlation between March5 and fat mass in rodent models, we next sought to determine whether March5 expression was also decreased in association with increasing adiposity in humans. Subsequently, we analyzed March5 expression in abdominal subcutaneous adipose tissue biopsies from participants in the METSIM study (770 individuals across a range of obesity indexes) (4). These data demonstrate a significant negative correlation of March5 with fat mass (r = −0.202, P = 9.75e⁻⁹), consistent with data from the HMDP strains and confirming that March5 expression is negatively correlated with adiposity in both mice and humans (Fig. 1, B and D). Finally, to validate correlation data from the human METSIM study, we performed Western blot analysis on subcutaneous WAT biopsies collected from a small cohort of age-matched male participants across lean and obese groups (52.8 ± 10.8 and 54.2 ± 8.4 yr old, respectively). These data demonstrate a lower abundance of PPARγ and MARCH5 protein in WAT from the obese group (body mass index = 33.0 ± 4.2 kg/m²) with subsequently impaired fasting blood glucose levels (9.9 ± 3.3 mmol/l) than in lean individuals (body mass index = 25.3 ± 3.3 kg/m²) with normal glucose tolerance (5.0 ± 0.4 mmol/l), confirming that MARCH5 is reduced in the setting of obesity and metabolic dysregulation, further suggesting that pathways important to adipocyte expansion are also downregulated in obesity (Fig. 1, E–G). Collectively, these data demonstrate that MARCH5 expression has a strong negative relationship with fat mass in rodents and humans and that pathways associated with adipogenesis (PPARγ) and mitochondrial function are coordinately downregulated in WAT in the setting of obesity.

MARCH5 expression correlates with PPARγ and is a likely PPARγ target gene.

To investigate whether MARCH5 and PPARγ expression are coordinately regulated, as suggested by data in Fig. 1, we again probed microarray data sets from WAT in the HMDP mouse strains. Figure 2A plots transcripts in WAT that positively correlated (r > 0.4) with Pparg gene expression across all strains of the HMDP mice. Several well-described PPARγ target genes were strongly positively correlated with Pparg expression in this analysis (red dots), including 3-hydroxyisobutyrate dehydrogenase (Hibadh), peroxisome biogenesis factor 13 (Pex13), oxysterol-binding protein like 11 (OsbpL11), and β-Klotho. Interestingly, the transcript most strongly correlated with Pparg was March5 (blue circle and arrow, Fig. 2A). Based on known PPARγ target genes, these data demonstrate that correlation analysis, which implicated March5 as a PPARγ target gene, is a useful approach to identify PPARγ-regulated genes. To provide further evidence, we performed pathway enrichment analysis of all transcripts that positively correlated with Pparg at a midweight bicorrelation cutoff of r > 0.5, equating to 173 individual genes (see Supplemental Table S1, available online at the Journal website). This analysis displayed strong enrichment for pathways including mitochondria, peroxisome, and lipid/fatty acid metabolism (Fig. 2B), consistent with pathways classically regulated by PPARγ and supporting the notion that this gene set is indeed reflective of PPARγ transcriptional activity.

Fig. 2.

Membrane-associated RING finger (C3HC4) 5 (MARCH5) is highly expressed in adipose tissue and is regulated by peroxisome proliferator-activated receptor-γ (PPARγ). A: dot plot depicting genes significantly (5% false discovery rate) positively correlated (bicorrelation r > 0.4) with PPARγ expression in white adipose tissue (WAT) of Hybrid Mouse Diversity Panel (HMDP) strains; red dots, previously validated PPARγ target genes; blue dot, March5. Hibadh, 3-hydroxyisobutyrate dehydrogenase; OsbpL11, oxysterol-binding protein-like 11; Pex13, peroxisomal biogenesis factor 13. B: pathway enrichment analysis of genes correlated (bicorrelation r > 0.5) with PPARγ in WAT from strains of the HMDP. C: Western blots for MARCH5 and 14-3-3 (loading control) in WAT and brown adipose tissue (BAT) from C57BL/6J mice. D: gene expression analysis for March5 mRNA in various tissues of C57BL/6J mice. Hrt, heart; Quad, quadriceps; Sol, soleus; Liv, liver. Values are means ± SE; n = 5. E: chromatin immunoprecipitation-sequencing analysis of PPARγ, retinoic acid receptor-α (RXRα), and CCAAT enhancer-binding protein-α (C/EBPα) at the March5 locus in 3T3-L1 adipocytes. F: photomicrographs of predifferentiated (Pre-Diff) and differentiated (day 7 Diff) primary adipocytes isolated from WT (PPARγ floxed) and PPARγ knockout [PPARγ floxed + aP2-cre (PPARγ-KO)] C57BL/6J mice. G: gene expression analysis for March5, normalized to cyclophilin A (Ppia), in cells shown in photomicrographs in F. PPg-KO, PPARγ-KO. Values [arbitrary units (AU)] are means ± SE. *P < 0.05 vs. non-diff WT.

Considering that March5 was strongly correlated with PPARγ expression in WAT, we further investigated its tissue distribution in mice. MARCH5 protein levels were robustly detected in brown adipose tissue (BAT) and gonadal WAT in C57BL/6J mice, as shown by Western blot analysis (Fig. 2C). Furthermore, qPCR analysis of several tissues from male WT mice demonstrated that March5 expression was enriched in WAT and BAT compared with other tissues such as heart, skeletal muscle, and liver (Fig. 2D), consistent with March5 being a transcriptional target of PPARγ. To directly test whether March5 was a transcriptional target of PPARγ and regulated by the adipogenic transcriptional program, we analyzed existing PPARγ, RXRα, and C/EBPα ChIP-seq data sets from differentiating adipocytes and investigated whether these proadipogenic transcription factors were enriched on the March5 locus (14, 16). These data confirm that all three transcription factors occupied the March5 promoter in close proximity to the transcriptional start site (Fig. 2E), confirming that PPARγ, RXRα, and C/EBPα cooperatively regulate March5 expression, similar to other adipogenic genes (16). These findings are strengthened by findings from primary adipocytes isolated from WT and adipocyte-specific PPARγ-KO mice [PPARγ floxed (WT) and PPARγ floxed + aP2 Cre (PPARγ-KO), respectively; Fig. 2, F and G]. These data confirm that WT adipocytes differentiated proficiently; however, consistent with the known role of PPARγ in adipogenesis, the ability of PPARγ-KO adipocytes to differentiate was severely impeded. In WT adipocytes, March5 expression was upregulated upon differentiation; however, this effect was completely blunted in differentiated PPARγ-KO cells and was also lower in undifferentiated PPARγ-KO cells. Thus the data presented above provide compelling evidence that March5 is transcriptionally regulated by PPARγ and the adipogenic program and is subsequently highly enriched in adipose cells and tissue in vitro and in vivo, respectively.

Mitochondrial proteins including MARCH5 are upregulated during adipocyte differentiation.

Because MARCH5 has been demonstrated to influence proteins that regulate mitochondrial dynamics, including Mfn2 and Drp1 (9, 16, 20), we investigated this pathway in our own experimental models. Accordingly, we subjected immortalized MEFs isolated from WT and MARCH5-KO mice to Western blot analysis and demonstrated that loss of MARCH5 in these cells promotes an increase in Mfn2 protein abundance (Fig. 3A). This is consistent with previous findings that Mfn2 is ubiquitinated by MARCH5 and subsequently degraded, thereby promoting an environment that would induce mitochondrial fusion (20, 33).

Fig. 3.

Peroxisome proliferator-activated receptor-γ (PPARγ) and mitochondrial protein expression, including membrane-associated RING finger (C3HC4) 5 (MARCH5) are increased during adipocyte differentiation. A: Western blots and densitometric quantification of MARCH5 and mitofusin2 (Mfn2) from wild-type (WT) and MARCH5 knockout (KO) mouse embryonic fibroblasts (MEFs), adjusted for loading control (14-3-3). B and C: Western blot and densitometry analyses of adipogenic and mitochondrial proteins from predifferentiated through day 10-postdifferentiated 3T3-L1 adipocytes. Values are means ± SE; n = 60,000 cells/well. *P < 0.05. Drp1, dynamin-related protein 1; C-II, C-III, C-IV, and C-V, complexes II–V of the mitochondrial electron transport chain; DMI, dexamethasone/methyl-3-isobutylxanthine (IBMX)/insulin; Mito oxphos, mitochondrial oxidative phosphorylation; SETDB1, histone-lysine-N-methyltransferase.

We also determined whether MARCH5 was regulated in adipocytes independently of other mitochondrial proteins or whether it was part of a global mitochondrial change in differentiating adipocytes, as suggested above. We immunoblotted proteins from WT 3T3-L1 adipocytes at different stages of adipocyte differentiation (predifferentiation and days 0, 1, 3, 6, and 10). Figure 3, B and C, confirms a substantial increase in MARCH5, Drp1, and Mfn2 in adipocytes from day 3 of differentiation, which coincided with an increase in PPARγ₂ expression [the adipocyte-specific variant (Fig. 3B, top band)]. Furthermore, other mitochondrial proteins, including proteins of the electron transport chain (complex II–V) and porin [voltage-dependent anion-selective channel protein 1 (Vdac1)], were also increased at this time point. These findings are consistent with the notion that PPARγ and adipocyte differentiation promote an upregulation of genes related to increasing mitochondrial activity, including an increase in MARCH5. Because we show that MARCH5 is increased in differentiating adipocytes and that MARCH5 deletion leads to an increased abundance of Mfn2, we wanted to determine if MARCH5 could play a key role in regulating adipocyte mitochondrial fusion, metabolism, and lipid storage.

Manipulation of MARCH5 abundance in 3T3-L1 adipocytes regulates mitochondrial respiration and glycolysis.

To investigate the functional effect of MARCH5 in differentiating adipocytes, we used lentiviral-delivered shRNAs to generate stable MARCH5 knockdown (shMARCH5) and control knockdown (shLuc) 3T3-L1 adipocytes. Figure 4 confirms an average 50% knockdown of March5 at various stages of differentiation (Fig. 4A), with ∼80% reduction in the early stages of differentiation, the critical time point at which we demonstrate that MARCH5 sharply increases in WT cells. Figure 4 also shows that depletion of MARCH5 in 3T3-L1 adipocytes had no effect on expression of genes involved in adipocyte differentiation and mitochondrial dynamics, including Pparg (Fig. 4B), Dnm1L (Fig. 4C), and Mfn2 (Fig. 4D). Although we observed no change in Dnm1L and Mfn2 gene expression, we and others have demonstrated that MARCH5 is an upstream protein regulator of the outer mitochondrial fusion protein Mfn2. To determine if depletion of MARCH5 in 3T3-L1 adipocytes modulated mitochondrial fusion, we used confocal microscopy to study MitoTracker deep red-stained MARCH5 knockdown and control cells. Mitochondrial populations in MARCH5 knockdown adipocytes were elongated and networked compared with fragmented mitochondria in control cells (Fig. 4E); these characteristics were significant upon quantification (Fig. 4F), confirming that depletion of MARCH5 in adipocytes promoted a pro-fusion phenotype. Moreover, Western blotting for Mfn2 was also elevated in March5-depleted cells (Fig. 4G), providing further evidence of a pro-fusion phenotype in these cells.

Fig. 4.

Membrane-associated RING finger (C3HC4) 5 (MARCH5) modulation in 3T3-L1 adipocytes alters metabolic gene expression and cellular metabolism. A–D: quantitative PCR analysis of gene expression of control [short-hairpin luciferase (shLuc)] and MARCH5 knockdown [KD (shMARCH5)] 3T3-L1 adipocytes, presented as fold change from control. E: pseudo-colored confocal image of mitochondrial morphology in MitoTracker deep red-stained control (shLuc) and MARCH5-KD adipocytes (left). Original magnification ×600. Right: enlargements of cell areas contained in white boxes at left. F: quantification of mitochondrial networks in control (shLuc) and MARCH5-KD adipocytes, measured as mean number of mitochondrial branches using MiNA and Fiji Image software. G: Western blots for mitofusin 2 (Mfn2), as a measure of mitochondrial fusion and MARCH5 activity, in control (shLuc) and MARCH5-KD adipocytes at day 0 and day 6 postdifferentiation. H: oil red O-stained images of neutral lipid deposition in MARCH5-KD and control (shLuc) adipocytes at low (×40) and high (×200) magnification. I: quantification of neutral lipid staining from oil red O imaging. J: glucose uptake in MARCH5-KD and control (shLUC) 3T3-L1 adipocytes. CPM, counts per minute. K: pyruvate dehydrogenase kinase (Pdk4) gene expression in MARCH5-KD and control (shLUC) 3T3-L1 adipocytes throughout differentiation (day −3 to day 6). L and M: Seahorse flux analyzer respirometry data from MARCH5-KD and control (shLuc) adipocytes. Oxygen consumption rate (OCR), including basal, carbonyl cyanide-4-(trifluoromethoxy) phenylhydrazone (FCCP)-uncoupled (UCR), and ATP-linked respiration (L) and extracellular acidification rate (ECAR) as a proxy measure of glycolysis and glycolytic capacity (M) were measured. N and O: gene expression analysis from differentiated 3T3-L1 cells infected with adenoviruses (pAD) for green fluorescent protein (Gfp) and March5. P: Pdk4 gene expression in MARCH5-overexpressing and control (GFP) 3T3-L1 adipocytes. Q and R: Seahorse analysis of OCR and ECAR in control (GFP) and MARCH5-overexpressing cells. Values are means ± SE; n = 8. *P < 0.05 vs. control (shLuc or GFP transgenic).

Because of the tight associations between mitochondrial function and energy metabolism, we investigated whether substrate storage and utilization were altered in MARCH5 knockdown and control adipocytes. Although MARCH5 depletion promoted mitochondrial fusion, oil red O staining for neutral lipid abundance did not show a significant difference in the MARCH5-depleted cells (Fig. 4, H and I). Furthermore, we also demonstrated nearly twofold higher basal and insulin-stimulated glucose uptake [5,168 ± 715 vs. 9,124 ± 1,814 counts per minute (cpm)/mg and 10,293 ± 1,127 vs. 17,191 ± 612 cpm/mg, respectively, P < 0.05) in MARCH5 knockdown cells, suggesting that depletion of MARCH5 led to an increased reliance on carbohydrate metabolism (Fig. 4J).

Finally, because of the effects on mitochondrial fusion and apparent preference for carbohydrate metabolism, we investigated metabolic parameters and mitochondrial function using the Seahorse flux analyzer. The Seahorse apparatus provides measures of cellular OCR (a surrogate readout of oxidative phosphorylation) and ECAR (a surrogate readout of glycolysis). Our data demonstrate that MARCH5 knockdown in adipocytes increased basal mitochondrial respiration, as indicated by an increased basal OCR, which coincided with an increase in ATP-linked respiration (Fig. 4L). In addition, rates of glycolysis were increased in MARCH5 knockdown adipocytes (Fig. 4M) compared with control cells, further suggesting that MARCH5 depletion modulates glucose metabolism and, subsequently, increases mitochondrial metabolism. This effect on glycolysis was also supported by qPCR data demonstrating that depletion of MARCH5 increases expression of the glycolytic enzyme pyruvate dehydrogenase kinase (Pdk4; Fig. 4K), consistent with increases in ECAR measured by the Seahorse device.

In light of the results observed for MARCH5 depletion in adipocytes, we sought to determine whether reciprocal effects would be observed following overexpression of MARCH5. We infected differentiated 3T3-L1 cells with adenoviruses that express MARCH5 or a control protein (GFP). After adenoviral transduction, robust expression of Gfp and March5 mRNA (Fig. 4, N and O) was demonstrated at day 7 postdifferentiation (1 ± 0.1 vs. 0 ± 0.0 fold change and 2 ± 1.5 vs. 72 ± 5 fold change, respectively, relative to Gfp expression, P < 0.05). Using qPCR and the Seahorse analyzer, we revealed reduced Pdk4, basal respiration, and glycolysis in MARCH5-overexpressing differentiated 3T3-L1 cells compared with GFP-overexpressing cells; however, these MARCH5-overexpressing differentiated 3T3-L1 cells retained the capacity to increase mitochondrial respiration following FCCP-stimulated mitochondrial uncoupling (Fig. 4, P–R). This is in contrast to MARCH5 knockdown, which promoted an increase in basal respiration and glycolysis compared with control cells (Fig. 4, L and M). Thus these results indicate that manipulation of MARCH5 expression may reciprocally alter metabolic function, which coincides with a change in mitochondrial fusion, providing evidence that manipulation of MARCH5 alters cellular respiration.

MARCH5 overexpression in vivo alters expression of metabolic genes.

To validate the ability of MARCH5 to regulate adipocyte metabolism, we performed in vivo experiments in 4-wk-old male C57BL/6J mice (n = 6/group). We injected MARCH5- or control-recombinant adeno-associated viruses (rAAV6, 1e11 vg) subcutaneously into contralateral (opposing) inguinal fat pads of the same mouse (control on the left side, MARCH5 on the right side). Mice were allowed to recover and, subsequently, housed for 4 wk, during which they were fed a HFD to promote adipose tissue expansion (data not shown); then they were culled, and inguinal fat pads were extracted for analysis. Gene expression analysis of the subcutaneous fat pads confirmed overexpression of March5 (6.7 ± 3.8 vs. 7,145 ± 3,095 fold change relative to control, P < 0.05) and also demonstrated that MARCH5 overexpression in adipose tissue led to increased expression of genes associated with adipose tissue expansion and mitochondrial biogenesis, including PPAR-γ coactivators Ppargc1a and Ppargc1b and Pparg2 (Fig. 5, A and B). In addition, MARCH5-overexpressing fat pads exhibited increased expression of mitochondrial dynamics genes, including Mfn1, Mfn2, and Dnm1L (Fig. 5C). There was no change in expression of indexes of apoptosis, including cytochrome c (Cytc), Fundc1, and B-cell lymphoma 2 (Bcl2) (Fig. 5D). MARCH5 overexpression also induced an increase in genes involved in lipid and peroxisome metabolism, including fatty acid synthase (Fasn), cluster of differentiation 36 (Cd36), patatin-like phospholipase domain-containing protein 2 (Pnpla2), carnitine O-palmitoyltransferase 2 (Cpt2), carnitine O-acetyltransferase (Crat), and carnitine O-octanoyltransferase (Crot), and reduced expression of genes that promote lipid catabolism, including the 5′-AMP-activated protein kinase subunits Prkab1 and Prkab2 (Fig. 5E). In contrast to MARCH5 overexpression, we observed reductions in Fasn expression in our MARCH5 knockdown 3T3-L1 cells (data not shown), consistent with Fasn being a potentially critical gene in the MARCH5 phenotype. Collectively, these data are consistent with the hypothesis that MARCH5 regulates mitochondrial dynamics and metabolism in adipose tissue in vivo, similar to 3T3-L1 adipocytes.

Fig. 5.

Membrane-associated RING finger (C3HC4) 5 (MARCH5) overexpression in white adipose tissue from C57BL/6J mice alters expression of metabolic and mitochondrial genes. Four-week-old C57BL/6J mice received subcutaneous contralateral injections of control [multiple cloning site (MCS)] or MARCH5 adeno-associated viruses (recombinant AAV6) into developing inguinal fat pads and fed a high-fat diet for 4 wk. A–E: expression in excised control and MARCH5-AAV subcutaneous fat pads of the same animal for MARCH5, adipogenic genes [peroxisome proliferative-activated receptor-γ coactivators (Ppargc1a and Ppargc1b), CCAAT/enhancer-binding protein-α (Cebpa), peroxisome proliferative-activated receptor-γ₂ (Pparg2), and cell death activator A (Cidea)], mitochondrial fusion and fission genes [mitofusins 1 and 2 (Mfn1 and Mfn2) and dynamin 1-like (Dnm1L)], apoptosis genes [cytochrome c-1 (Cyc1), FUN14 domain-containing protein 1 (Fundc1), and B-cell lymphoma 2 (Bcl2)], and lipid metabolism [low-density lipoprotein receptor (Ldlr), solute carrier family 27 (Slc27a1), fatty acid synthase (Fasn), cluster of differentiation 36 (Cd36), phospholipase domain-containing protein 2 (Pnpla2), carnitine O-palmitoyltransferase (Cpt1c and Cpt2), carnitine acetyltransferase (Crat), carnitine O-octanoyltransferase (Crot), diacylglycerol O-acyltransferase 2 (Dgat2), protein kinase cAMP-activated catalytic subunit α (Prkaca), fatty acid-binding protein 4 (Fabp4), lipoprotein lipase (Lpl), and lipase E (Lipe)]. Values are means ± SE; n = 8. *P < 0.05 vs. control.

Overall, our data demonstrate that MARCH5 expression regulates mitochondrial metabolism in adipocytes in vitro and in vivo, likely due to its role in modulating mitochondrial dynamics and substrate utilization in the mitochondria.

DISCUSSION

Maintenance of an appropriate ability to manipulate mitochondrial dynamics and, thus, function is critical for overall cellular health. MARCH5 is an outer mitochondrial membrane protein that has been shown to regulate several processes involved in maintaining mitochondrial dynamics, including fusion, fission, and apoptosis (34). Furthermore, several studies over the past 5–10 yr have demonstrated an important role of mitochondrial activity in adipose tissue health and function. Here, we show that MARCH5 is highly enriched in WAT and BAT, suggesting that MARCH5 is critical to adipocyte function.

Given that MARCH5 is a mitochondria-associated protein, it is reasonable to assume that it might be highly expressed in tissues that are high in mitochondria. However, our data do not support this notion, particularly with respect to the low expression in skeletal muscle (which is high in mitochondria) and high expression in WAT (which is low in mitochondria). Instead, our data provide strong evidence that MARCH5 is regulated by the adipogenic program, consistent with its enriched expression in adipose tissues. Indeed, we observed that MARCH5 expression in adipocytes is likely regulated by the nuclear receptor PPARγ and that experimental manipulation of MARCH5 expression regulates mitochondrial respiration in adipocytes. This study is the first to demonstrate that MARCH5 is regulated by PPARγ during adipocyte differentiation. Importantly, suppression of PPARγ expression in hypertrophic adipocytes is observed in the setting of obesity, and this may explain the reduction in mitochondrial gene expression, including March5. This notion is supported by our findings that March5 expression is negatively correlated with fat mass in rodents and humans and that March5 is directly regulated by PPARγ, as shown by ChIP-seq analysis. It is also possible that C/EBPα, along with PPARγ, plays a role in regulating March5 expression due to its proximity at the March5 promoter. Adipogenic genes are known to be regulated by both C/EBPα and PPARγ during adipogenesis, and C/EBPα and PPARγ often coexist as adipogenic gene promoters in ChIP-seq data sets. We demonstrate that loss of PPARγ is sufficient to prevent adipogenesis and March5 expression; however, we have not performed the same experiments with C/EBPα depletion; thus the precise mechanism of regulation remains to be determined in future studies.

We observed that MARCH5 expression was directly linked with mitochondrial metabolism in 3T3L1 adipocytes, since MARCH5 knockdown increased glycolysis and basal mitochondrial respiration, while MARCH5 overexpression decreased these parameters. MARCH5-overexpressing cells maintained FCCP-uncoupled respiration, which suggests that these mitochondria were still functioning normally when stimulated with uncoupling compounds, such as FCCP. The mechanisms by which MARCH5 manipulates adipocyte metabolism are not known and require further investigation, but they may involve the ability of MARCH5 to promote mitochondrial fusion. Indeed, previous reports show that MARCH5 regulates mitochondrial fusion by ubiquitinating Mfn2, thereby manipulating mitochondrial morphology and altering metabolic homeostasis (3, 9, 20, 24, 33). Our studies using confocal microscopy and Western blotting for Mfn2 show that knockdown of MARCH5 in 3T3-L1 adipocytes promotes mitochondrial fusion compared with control cells. This finding is supported by similar observations of the fusion protein Mfn2 in immortalized MARCH5-KO MEFs, likely also facilitating a pro-fusion phenotype in these cells.

Previous studies have shown that mitochondrial fusion occurs as an adaptation to starvation to promote a more energy-efficient mitochondrial pool, allowing energy production to be maintained in the presence of altered substrate availability (15, 36, 38, 45). Because MARCH5 expression is reduced in obesity, loss of MARCH5 might be a mechanism by which the adipocyte can promote mitochondrial fusion in an attempt to alter metabolic preferences. This may suggest that MARCH5 is part of a coordinated interplay between metabolism, mitochondrial fusion, and subsequent retained capacity to generate energy.

A switch in substrate preference may also promote altered metabolic outputs, where changes in lipid metabolism or β-oxidation could lead to increased basal mitochondrial respiration while maintaining other metabolic parameters such as uncoupled respiration. Consistent with this notion, we observed an increase in glucose uptake in MARCH5-depleted cells, which may account for the increase in glycolysis and basal mitochondrial respiration in these cells. Interestingly, when we overexpress MARCH5 in vivo, there is an increase in the expression of key regulators of lipid metabolism, which further suggests that MARCH5 may play a role in mitochondrial substrate utilization. Increased expression of genes involved in lipid uptake and synthesis from MARCH5-overexpressing subcutaneous fat pads suggests that the alterations in lipid metabolism could be favoring the storage of excess energy. Notably, Crat and Crot gene expression is elevated in vivo when March5 is overexpressed. This is curious, given that March5 is a PPARγ target gene and that PPARγ was discovered because of its role in expanding peroxisomes; thus it is possible that manipulation of MARCH5 somehow impacts peroxisome function. Crat and Crot also play a role in the transport of very-long-chain fatty acids between the peroxisome, cytosol, and mitochondria. Future studies should investigate whether MARCH5 directly degrades mitochondrial transporters, leading to a change not only in mitochondrial fusion, but also substrate utilization. Such a pathway would provide an elegant mechanism for regulation of mitochondrial fusion, substrate utilization, and energy output in response to changes in cellular energy requirements.

As the purpose of this study was to investigate the role of MARCH5 in adipose tissue, we also investigated whether MARCH5 expression was involved in browning of WAT. Although we observed increased expression of MARCH5 protein in BAT compared with WAT in mice, we were unable to show any changes in browning-related genes following MARCH overexpression in vitro or in vivo (data not shown), suggesting that MARCH5 in this setting is not affecting browning of WAT. In the context of previously published data, our findings indicate that MARCH5 likely regulates mitochondrial morphology and dynamics in adipocytes by controlling mitochondrial fusion. Preceding studies have demonstrated that MARCH5 manipulates mitochondrial morphology by regulating the abundance of Mfn2 and Drp1, although there is some conjecture as to whether deletion of MARCH5 leads to a pro-fusion or pro-fission phenotype. Our study is the first to investigate MARCH5 in adipose cells and tissue, and our findings suggest that loss of MARCH5 in adipocytes leads to increased mitochondrial fusion.

Overall, these results demonstrate that MARCH5 expression is negatively associated with fat mass and regulated by PPARγ and that changes in MARCH5 expression can alter mitochondrial and cellular metabolism. These functions likely alter the utilization of lipid, which subsequently impacts glucose metabolism. More broadly speaking, these data provide further evidence that alterations in pathways important for mitochondrial function have a significant impact on adipocyte function. However, whether these pathways can be targeted or manipulated for metabolic benefit in vivo remains unanswered.

GRANTS

Confocal imaging was supported by Monash Micro Imaging (Monash University). This study was supported by an award from the Victorian State Government Operational Infrastructure Support Program (to the Baker Heart and Diabetes Institute) and National Heart, Lung, and Blood Institute Grants HL-121172 and HL-28481 (to A. J. Lusis and M. Civelek). B. G. Drew and A. C. Calkin are supported by National Heart Foundation of Australia Future Leader Fellowships.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

B.G.D. conceived and designed research; S.T.B., S.C.M., Y.L., M.C., C.J.V., P.G., B.A.K., A.L.H., A.J.L., D.C.H., A.C.C., and B.G.D. performed experiments; S.T.B., M.C., C.J.V., D.C.H., A.C.C., and B.G.D. analyzed data; S.T.B., C.J.V., D.C.H., A.C.C., and B.G.D. interpreted results of experiments; S.T.B. and B.G.D. prepared figures; S.T.B. and B.G.D. drafted manuscript; S.T.B., A.C.C., and B.G.D. edited and revised manuscript; S.T.B., S.C.M., Y.L., M.C., C.J.V., P.G., B.A.K., A.L.H., A.J.L., D.C.H., A.C.C., and B.G.D. approved final version of manuscript.