Role of the Neutral Amino Acid Transporter SLC7A10 in Adipocyte Lipid Storage, Obesity, and Insulin Resistance

Adipocyte hypertrophy in both subcutaneous (SC) and visceral white adipose tissue (AT) is strongly associated with whole-body insulin resistance, with or without obesity and AT inflammation (1,2), and with fatty liver, dyslipidemia, impaired mitochondrial function, and reduced insulin-stimulated glucose uptake in adipocytes (3). Experimental impairment of mitochondrial respiration and increased reactive oxygen species (ROS) generation in adipocytes reduce adipocyte insulin sensitivity (4), and the extent of mitochondrial dysfunction determines the severity of insulin resistance and type 2 diabetes (5). However, the molecular mechanisms that promote adipocyte hypertrophy and insulin resistance remain incompletely understood, and we urgently need new potential treatment targets.

SLC7A10, also known as alanine-serine-cysteine transporter-1 (ASC-1), has sodium-independent activity and high affinity for the small neutral amino acids (AAs) glycine, l-alanine, l-threonine, l-cysteine, l-serine, and d-serine (6,7). SLC7A10 is highly expressed in certain regions of the brain and is being explored as a therapeutic target in neuropsychiatric disorders (e.g., schizophrenia) (8).

A previous report showed selective expression of SLC7A10 in white but not beige or brown adipocytes, with fivefold higher expression in AT compared with the highest expressing parts of the brain, and diminished expression in other tissues (9). However, the possible role of SLC7A10 in metabolic regulation has not been explored.

While AAs known to be transported by SLC7A10 in the brain, e.g., serine, glycine, and cysteine, have central roles in one-carbon metabolism, the methionine cycle, glutathione synthesis, and redox balance (6,7,10,11), the AAs carried by SLC7A10 in adipocytes and the consequent metabolic effects are unknown. By transcriptome screens and interrogation of several human obesity cohorts along with functional experiments, we here uncover SLC7A10 as an important novel regulator of core metabolic functions in white adipocytes, providing new insight into the development of obesity and insulin resistance.

Anthropometric data are summarized in Table 1 (seven cohorts).

Whole AT was homogenized or fractionated into isolated adipocytes or stromal vascular fraction (SVF), and RNA was purified, as described previously (12). Global gene expression in whole AT was measured by microarrays as described previously for the biliopancreatic diversion (BDP)-Fat cohort (13), Sib Pair cohort (14), very low calorie diet (VLCD) study (baseline data) (15) and RIKEN cohort (16). Quantitative PCR (qPCR) was performed with SYBR Green dye following cDNA synthesis with a high-capacity cDNA reverse transcription kit (Applied Biosystems) (Supplementary Table 1).

Human liposuction aspirates from the abdomen and flanks were collected with informed consent from donors undergoing cosmetic surgery at Aleris medical center and Plastikkirurg1. The donors comprised 18 women and 1 man between 21 and 68 years of age (mean ± SD age 46.3 ± 12.4 years) and with BMI between 24.3 and 32.8 kg/m² (27.9 ± 2.7 kg/m²), all free of diabetes and otherwise healthy (Supplementary Table 2).

The SVF from SC AT was isolated as previously described (17) with some modifications. Briefly, Krebs-Ringer phosphate buffer containing Liberase Blendzyme 3 (Roche) and DNase was added to the liposuction aspirate. Following a 1-h incubation at 37°C, the digested fat tissue was filtered, washed with 0.9% NaCl, and centrifugated. Red blood cells were lysed (NH₄Cl [155 mmol/L], K₂HPO₄ [5.7 mmol/L], and EDTA [0.1 mmol/L]). Preadipocytes were seeded and cultured in GlutaMAX DMEM (Thermo Fisher Scientific) supplemented with 10% FBS and 50 μg/mL gentamicin (Sigma-Aldrich) and grown at 37°C with 5% CO₂. The following day, differentiation of primary human adipose cells (human adipose stromal cells [hASCs]) was induced by supplementing of the culture medium with 33 μmol/L biotin, 1 nmol/L triiodothyronine, 17 μmol/L dl-pantothenate, 10 μg/mL transferrin, 66 nmol/L insulin, 100 nmol/L cortisol, 15 mmol/L HEPES, and 10 μmol/L rosiglitazone. Rosiglitazone was discontinued after 6 days, and terminal differentiation was defined at 12–13 days.

3T3-L1 mouse preadipocytes were cultured and differentiated as described previously (18).

BMS-466442 (AOBIOUS), referred to as SLC7A10 inhibitor 1 (19), and Lu AE00527, referred to as SLC7A10 inhibitor 2 (20), were used to inhibit SLC7A10 function in in vitro cell culture experiments at a standard final concentration of 10 μmol/L. The latter inhibitor was provided by H. Lundbeck A/S (Valby, Denmark).

Total RNA from human and mouse cell cultures was isolated with RNeasy kit (QIAGEN) and quality checked by QIAxpert spectrophotometer (QIAGEN) prior to cDNA synthesis with 200 ng RNA input with use of a high-capacity cDNA reverse transcription kit (Applied Biosystems). cDNA was analyzed by LightCycler 480 (Roche) quantitative real-time PCR with SYBR Green dye (Roche) and target primers (Supplementary Table 1). Relative mRNA expression was determined by standard curves and normalized to a reference gene (HPRT or Rps13). Prior to RNA sequencing (RNA-seq), samples were DNase treated and RNA integrity number (>9) was confirmed by Bioanalyzer (Agilent). cDNA libraries were generated with a TruSeq Stranded mRNA Library Prep kit and sequenced by Illumina HiSeq 4000. Reads were mapped against the human genome (GrCH38) with use of HISAT (version 2.1.0), tabulated by featureCounts (version 1.5.2), and analyzed with DESeq2 (version 1.22.2).

Coexpression analysis was performed based on global gene expression data for human adipocytes isolated directly from biopsies of 12 lean and 12 obese patients (ADIPO cohort). Pearson correlation coefficients were calculated for correlations between SLC7A10 mRNA and all other transcripts globally (∼47,000 probes, including 21,000 individual genes) across the 24 patients. Genes with correlation coefficients β > 0.65 or β < −0.65 were analyzed with the PANTHER Gene List Analysis tool (http://www.pantherdb.org/) for performance of a statistical overrepresentation test (binomial statistics, Bonferroni corrected for multiple testing). RNA-seq results from hASCs were analyzed by PANTHER (v.14.0), using Fisher exact test (21).

Transfection with mouse Slc7a10 expression plasmid was performed with the transfection reagent TransIT-L1, following the manufacturer’s protocol (Mirus Bio LCC). Slc7a10 plasmid (2.0 µg) or empty vector was used per milliliter of medium (Supplementary Table 3).

Cells were lysed in radioimmunoprecipitation assay buffer (Thermo Fisher Scientific) containing EDTA-free protease inhibitor cocktail (Roche) and PhosphoStop (Sigma-Aldrich). Protein content in the lysates was determined with DC Protein Assay (Bio-Rad Laboratories), loaded onto 4–20% TGX Gels (Bio-Rad Laboratories) and subjected to SDS-PAGE prior to transfer to nitrocellulose membrane. Membranes were developed with Femto solution (Thermo Fisher Scientific), and target protein amount relative to endogenous control was quantified by densitometry.

Extracellular (medium) AA concentrations were assayed by gas chromatography–tandem mass spectrometry (GC-MS/MS) (Agilent) (22). Flux was calculated based on AA concentration in unconditioned and conditioned medium.

AA uptake was measured in sodium-free assay buffer, as described previously (19), with use of unlabeled controls (d- or l-serine, final concentration 10 mmol/L) and stock solutions of d-[³H]-serine (15 Ci/mmol) and l-[¹⁴C]-serine (100 mCi/mmol) (PerkinElmer). Sample radioactivity was measured as 5-min averaged counts per minute (CPM) (QuantaSmart software).

Cellular respiration was measured by the Seahorse XF Cell Mito Stress Test kit and XF96 Analyzer (Agilent). Preadipocytes were seeded in gelatin-coated (0.1% w/v) microplates and differentiated. Cells were washed and treated in XF base medium supplemented with l-glutamine (2 mmol/L), sodium pyruvate (2 mmol/L), and glucose (10 mmol/L) and incubated in a CO₂-free incubator for 1 h. Oxygen consumption rate (OCR) data were normalized to numbers of cells per well, measured by Hoechst staining.

Lipid accumulation was assessed by oil red O (ORO) staining as described previously (23).

Adipocytes were washed with PBS and incubated overnight in glucose-reduced DMEM with or without treatment. Subsequently, cells were starved in glucose-free medium with or without treatment for 2.5 h. Insulin (final concentration 10 nmol/L) was added to indicated wells for 30 min prior to the assay. Deoxy-d-[¹⁴C]-glucose (57.7 mCi/mmol) (PerkinElmer) was added for 30 min and incubated at 37°C. Cells were placed on ice and washed, and lysates were collected in Ultima Gold fluid cartridges (PerkinElmer). Isotope retention in lysate was measured as 5-min averaged CPM (QuantaSmart software) in a scintillation counter.

Adipocytes were incubated for 30 min with or without 5 μmol/L CM-H2DCFDA as described previously (24). Cells were washed and incubated in Krebs-Ringer bicarbonate buffer or sodium-free assay buffer, with indicated compounds, followed by measurement of fluorescent emission of oxidized H2DCFDA probe (538 nm following 485 nm excitation) with a SpectraMax Gemini EM (Molecular Devices) plate reader at 37°C.

Total glutathione was measured with the GSH/GSSG-Glo Assay (Promega) according to the manufacturer’s protocol. Briefly, cells were lysed with glutathione lysis reagent and incubated in luciferin generation reagent for 30 min at room temperature. Samples were incubated for 15 min following addition of luciferin detection reagent. Luminescence was measured with FLUOstar OPTIMA EM (Thermo Fisher Scientific).

Heterozygous Slc7a10b loss-of-function Danio rerio (zebrafish) were obtained from The Zebrafish Model Organism Database (ZFIN) (genomic feature sa15382) and crossed for obtaining homozygote and wild-type (WT) zebrafish.

Genomic DNA was extracted from caudal fin biopsies of mature zebrafish with a DNeasy kit (QIAGEN). For genotyping, a 298 base pair (bp) region of the Slc7a10b gene containing the intron splice site A→T mutation was amplified by PCR with Platinum Taq High Fidelity DNA Polymerase (Invitrogen) and the flanking primers 5′-TCGCCTACTTCTCCTCCATG-3′ (forward) and 5′-TTCCCAAGTCCTCCTGATGC-3′ (reverse). Samples were subjected to endonuclease digestion with use of the restriction enzyme AccI (New England Biolabs) prior to band separation by agarose gel electrophoresis. Genotypes were determined based on signature fragment digestion of the PCR product where the A→T mutation abolished the AccI recognition cleavage site.

The ZFIN genomic feature sa15382 zebrafish used in this study exhibited a point mutation in the conserved 3′ splice site between exons 6 and 7 in the Slc7a10b gene (ENSDART00000073398.5). This A→T mutation disrupts the dinucleotide splice site, and the intron between exon 6 and 7 is not spliced during maturation of the mRNA. Thus, the mRNA length is increased and the Slc7a10b protein product is inactive. Heterozygous larvae were raised and bred for obtaining zebrafish homozygous for this mutation. Due to the large variation in body weight between male and female zebrafish in a pilot study, the overfeeding study was performed with male zebrafish.

Four-month-old male zebrafish were housed in 3-L tanks (on average 20 per tank) with a recirculating system (Aquatic Habitats; Pentair AES) at 28.5 ± 1°C (mean ± SD) and pH 7.51 ± 0.3 (mean ± SD), with 10% daily water exchange, electrical conductivity 500 ± 50 μs, and a 14-h light and 10-h dark circadian cycle. Zebrafish were fed 8.2 mg Gemma Micro 500 (Skretting) per day, divided over three time points (at 8:00 a.m., noon, and 4:00 p.m.), in addition to freshly hatched Artemia (three drops of a 24-h Artemia culture) once per day. Gemma Micro 500, which the zebrafish were fed under both normal and overfeeding conditions, consisted of fishmeal, lecithin, wheat gluten, zebrafish oil, vitamins and mineral premixes, and betaine, containing 59% (w/w) protein sources and 14% (w/w) lipids (containing 14% n-3 fatty acids).

Three adult zebrafish were held per 1.5-L tank, in total 33 WT and 39 loss-of-function zebrafish. For weight gain, 12.3 mg feed per zebrafish per day (50% more feed than normal) was given for the first 3 weeks and 16.4 mg (100% more feed than normal) for the last 5 weeks of the overfeeding study. The circulation system was turned off for 5 min before and 30 min after each feeding, allowing consumption of all supplied food. Zebrafish were otherwise fed as under normal conditions (described above).

Zebrafish length and weight were recorded at the start and the end of the study, while weight was also measured after 3 and 6 weeks. Since it was not feasible to control the feed intake of each individual fish, we combined recorded data and tissue samples from the three zebrafish in each tank to obtain an average for each tank. Before handling, each zebrafish was sedated with 75–200 mg/L Tricain mesylate (Pharmaq).

After sacrifice, tissue biopsies from three zebrafish in each tank were pooled together and snap frozen in liquid nitrogen. Adipose and liver biopsies were fixed in 4% (v/v) paraformaldehyde in 0.1 mol/L phosphate buffer for 12 h and paraffin embedded after gradual dehydration as described previously (25). Slice sections of 5 μm were stained with hematoxylin-eosin, and adipocyte size was analyzed and quantified with ImageJ open-source software as previously described (18). RNA was isolated from visceral AT (VAT) of WT and Slc7a10b mutant zebrafish and cDNA libraries were generated for RNA-seq in the same way as described for human adipose cultures. Sample reads were mapped against the zebrafish genome (GRCz10) with HISAT (version 2.05), tabulated by featureCounts (version 1.5.2), and analyzed with DESeq2 (version 1.22.1).

Differences between groups in human cohort data were analyzed with paired t test, one- or two-way ANOVA, and are presented as means ± SD. For Pearson correlation and multiple regression analyses, adjustments for BMI and sex are indicated. Differences between groups in cell culture experiments were assessed using with two-tailed unpaired Student t test or one-way ANOVA with Dunnett or Sidak correction for multiple comparisons. Sample data from empirical experiments were assumed to be normally distributed, and results are presented as means ± SD, except for the Seahorse data, which are presented as geometric means ± 95% CI. All data were plotted as box and whisker plots, with Tukey’s method to detect and remove outliers. Statistical significance was calculated in GraphPad or with the R Bioconductor package limma (v3.34.9). Statistical details and the number of biological samples (n) for all experiments are provided in the figure legends. For cell experiments, n annotates the number of parallel wells per treatment. For zebrafish samples, n annotates the number of zebrafish per treatment. For qPCR data, n annotates either the number of patients or the number of wells in a cell experiment.

All human samples analyzed in the current study were obtained with written informed consent, and approval was given by the respective regional ethics committees (REC) (2010/502 and 2010/3405, REC West Norway; Dnr 721-96 and S 172-02, REC in Gothenburg, Sweden; 2009/1881-31/1, 2011/1002-31/1, and 2015/530-32, Committee of Ethics at Karolinska Institutet; and the Ethics Committee of the University of Leipzig, Leipzig, Germany).

Zebrafish were raised and cared for according to the Norwegian Animal Welfare Act guidelines, and all experiments after 5 days postfertilization were approved by the Norwegian Food Safety Authority (FOTS identifier 9199).

Data sets generated and/or analyzed during the current study are available from the corresponding author upon reasonable request. In addition, the RNA-seq data sets in this publication have been deposited in the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) (26) and are accessible through GEO Series accession number GSE135156.

In transcriptome screens of adipose samples from people at the peak of extreme obesity (BPD-Fat cohort [Table 1]), we prioritized candidate genes with concomitant differential expression in two separate disease-relevant comparisons of adipose function: VAT (omental [OM]) and abdominal SC AT in extreme obesity, and before and after profound fat loss (SC AT from the same patients). This combined transcriptome screen identified 65 genes with both depot- and fat loss–dependent expression (fold change ≥1.5 and q value <0.01 cutoffs in both analyses) (Supplementary Table 4). Further supporting a role for many of these genes in insulin resistance, we found strong significant correlations for 27 of the 65 genes (42%) with waist-to-hip ratio (WHR) adjusted for BMI and sex, in a second cohort of 88 people (Sib Pair cohort) (14) (Supplementary Table 4). Among these genes, only five (8%) showed inverse correlations with WHR, four of which displayed highest expression in OM AT and upregulation in SC AT after profound fat loss (CIDEA, SLC7A10, GPD1L, and HOXA5) (Supplementary Table 4). Among the 27 candidates, we additionally prioritized genes with high expression in adipocytes specifically, based on a cohort containing isolated adipocytes and SVF (ADIPO cohort). Calculation of adipocyte–to–SVF expression ratios revealed CIDEA and SLC7A10 as standout candidates with similar expression profiles (Fig. 1A). While previous studies have investigated functional roles of CIDEA in adipocytes (27), there is a paucity of functional data on adipocyte SLC7A10.

qPCR analysis of a larger cohort of people with severe obesity (Western Norway Obesity Biobank [WNOB] cohort) confirmed the adipose depot- and fat loss–dependent expression pattern of SLC7A10 (Fig. 1B). SLC7A10 mRNA was also higher in OM compared with SC samples in both isolated adipocytes and SVF (ADIPO cohort) but with diminished expression in SVF (Fig. 1C). Consistent with increased SC SLC7A10 expression after surgery-induced fat loss (Fig. 1C), SC adipocytes and SVF from lean people showed higher expression than samples from those with obesity (Fig. 1C). Furthermore, insulin-resistant obese (IRO) patients showed lower SLC7A10 mRNA compared with BMI-matched insulin-sensitive obese (ISO) patients (28) in SC as well as OM fat (ISO cohort) (Fig. 1D). Consistently, SC whole tissue SLC7A10 mRNA levels showed strong inverse correlations with HOMA of insulin resistance, triacylglycerol levels, SC adipocyte volume, WHR, and visceral fat volume (Fig. 1E and Supplementary Fig. 1). On the other hand, SC SLC7A10 mRNA showed positive correlations with SC adipocyte number and SC abdominal fat mass with adjustment for BMI and sex (Fig. 1E and Supplementary Fig. 1). Finally, comparing patients with and without type 2 diabetes in the WNOB morbid obesity cohort, we observed decreased SLC7A10 mRNA expression in OM as well as SC postsurgery AT (Fig. 1F).

To determine a potential causal role for SLC7A10 in the regulation of fat storage and adipose metabolism, we obtained zebrafish containing a splice site (loss-of-function) mutation in intron 6 of Slc7a10 isoform b. After 2 months of overfeeding, the knockout zebrafish had on average gained 38% more body weight than their WT counterparts (Fig. 2A). Assessing AT morphology, we observed on average 49% larger adipocytes in the loss-of-function zebrafish compared with WT (Fig. 2B and C). Histological analyses of liver, which may reveal altered morphology related to impaired fatty acid oxidation (i.e., fatty liver), showed no apparent differences in cell size and lipid droplet formation between the Slc7a10b loss-of-function and WT zebrafish (Fig. 2C).

We consequently explored mechanisms of the fat accretion by studying direct effects of altered SLC7A10 function in adipocytes. In differentiated mouse 3T3-L1 adipocytes and primary hASCs, SLC7A10 showed a marked increased mRNA and protein expression (Fig. 3A–D). To study the consequence of decreased SLC7A10 function in adipocytes, we inhibited SLC7A10 using the selective inhibitors BMS-466442 (SLC7A10 inhibitor 1) (19,29) or Lu AE00527 (SLC7A10 inhibitor 2) (20) during differentiation and observed increased lipid accumulation in 3T3-L1s (Fig. 3E) as well as hASCs (Fig. 3F) in comparison with controls.

To systematically explore SLC7A10-dependent metabolic processes, we first correlated mRNA levels of SLC7A10 and other genes in a global transcriptome analysis of adipocytes isolated directly from human SC fat biopsies (ADIPO cohort). By performing gene ontology (GO) analysis, we found that genes antiexpressed to SLC7A10 (113 unique genes [Supplementary Table 5]) mapped primarily to cellular/developmental processes and protein transport (Fig. 4A) and genes coexpressed (323 unique genes [Supplementary Table 5]) mapped to lipid metabolic processes and oxidative phosphorylation (Fig. 4A). Moreover, RNA-seq in SLC7A10 inhibitor–treated primary human adipocytes revealed a profound transcriptome effect (Supplementary Fig. 2A–C). Among 862 significantly downregulated genes, GO analysis revealed an enrichment of genes involved in biological processes such as immune response, inflammation, extracellular matrix organization, and cell differentiation (Supplementary Fig. 2D). Among 1,113 significantly upregulated genes, there were striking enrichments for the isopentenyl diphosphate biosynthetic process (26 fold), NADPH regeneration (26 fold), pentose phosphate shunt (26 fold), triglyceride biosynthetic process (9.3 fold), and glutathione metabolic processes (8.2 fold) (Supplementary Fig. 2D), indicating mechanisms that fueled lipid storage. Additionally, by pathway analysis we found a marked enrichment of genes involved in ATP synthesis (16 fold), tricarboxylic acid cycle (13 fold), and cholesterol biosynthesis (12 fold) among upregulated genes, and in, for example, glycolysis and angiogenesis for the downregulated genes (Fig. 4B and C).

Global gene expression in VAT from Slc7a10b loss-of-function and WT zebrafish was also assayed by RNA-seq. We visualization of RNA-seq reads we identified the expected mutation in the splice site of exon 6 (Supplementary Fig. 3A). Of note, mutants exhibited increased mRNA levels of the defective Slc7a10b (Supplementary Fig. 3B), compensatory to the impaired splicing and function. Before further analysis, we removed outliers using multidimensional scaling plots (Supplementary Fig. 3C) and excluded samples that differed significantly from the others in the expression of immediate-early stress-responsive genes (Supplementary Fig. 3D). RNA-seq revealed 1,736 differentially expressed genes in the Slc7a10b loss-of-function zebrafish, including 880 upregulated and 856 downregulated transcripts. The loss of function caused a particularly striking upregulation of Urahb, which encodes an enzyme that regulates degradation of uric acid to (S)-allantoin (Supplementary Fig. 4A and B), a product of purine nucleotide degradation and a marker of oxidative stress in most nonhuman mammals. Consistently, metabolism of urate, purine nucleobase, and hydrogen peroxide (H₂O₂) were among the most affected biological processes in the zebrafish VAT, together with oxygen transport, AA metabolism, lipoprotein remodeling, and lipid and citrate transport (Supplementary Fig. 4B). The RNA-seq analysis for SLC7A10 inhibition in the differentiating hASCs largely supported an effect on these processes, including steroid biosynthesis and oxidation-reduction process (Supplementary Fig. 5A). A total of 121 of the 216 GO terms identified in the zebrafish data set overlapped with the human data set. While genes in some terms showed directionality opposite that of expression, several of the most significant terms in zebrafish showed the same directionality in human cells (Supplementary Fig. 5A). From the 444 differentially expressed genes in the zebrafish data set, 26 genes overlapped with the human data set, of which 17 were regulated in the same direction. Among these were SLC7A10 (reflecting a compensatory upregulation), SCD (a marker of nutritionally regulated lipid storage), HSD17B10 (an isoleucine-catabolizing enzyme), PKM and PC (related to pyruvate metabolism and glyceroneogenesis which provides glycerol for lipid storage), and CPT1A (rate limiting for mitochondrial lipid β-oxidation) (Supplementary Fig. 5B).

In accordance with the transcriptome changes indicating effects on mitochondrial function, SLC7A10 inhibition for 24 h decreased basal respiration, ATP synthesis, maximal consumption rate, and spare respiratory capacity by up to 50% in murine (Supplementary Fig. 6A and B) and primary human adipocytes (Fig. 5A), with detectable effects after 2 h inhibition. Exposure to SLC7A10 inhibitor 2 (Lu) reproduced the suppression of maximal respiration and spare respiratory capacity (Fig. 5B). Conversely, overexpressing Slc7a10 (Supplementary Fig. 6C) in murine fat cells increased these measures along with basal respiration and ATP synthesis (Fig. 5C).

To examine the mechanism by which SLC7A10 modulates adipocyte metabolism, we tested the effect of SLC7A10 inhibition on the flux of neutral AAs, some of which are direct precursors of glutathione (e.g., cysteine, glycine, and serine) (6,7). SLC7A10 impairment strongly increased medium concentrations of serine from around day 8 of human adipocyte differentiation, while concentrations of other SLC7A10-linked neutral AAs showed only minor effects (Fig. 6A and Supplementary Fig. 7A). We confirmed the reduction in serine influx in response to SLC7A10 inhibition in cultured adipocytes from four independent donors (Supplementary Fig. 7B). Additionally, with use of radiolabeled AAs in sodium-free conditions, SLC7A10 inhibition reduced uptake of d-serine (Fig. 6B and Supplementary Fig. 7C) and l-serine in adipocytes (Supplementary Fig. 7C and D). The primary effect on serine transport is consistent with an important role for this AA in lipid synthesis, antioxidant regeneration, tricarboxylic acid cycle, glycolysis, and oxidative phosphorylation, in part because serine serves as a key methyl donor that controls biosynthesis and regeneration of ATP, NADPH, purines, glutathione, and other molecules through one-carbon metabolism (11) (Fig. 6C).

The effects of SLC7A10 inhibition on several genes in NADPH- and glutathione-related metabolism (Fig. 6D) prompted us to examine whether SLC7A10 affects cellular glutathione levels. SLC7A10 impairment decreased total glutathione levels detected after only 15–45 min inhibition in murine and human adipocytes (Fig. 6E and F). The decrease was confirmed with inhibitor 2, albeit not significant in the human adipocytes (Fig. 6E and F), while SLC7A10 overexpression in 3T3-L1 adipocytes increased total glutathione levels (Fig. 6G). Consistently, intracellular ROS levels increased progressively after 20 min of SLC7A10 inhibition in 3T3-L1 adipocytes (Fig. 6H) and after 60 min in mature human adipocytes (Fig. 6I), while SLC7A10 overexpression reduced ROS generation (Fig. 6J). Interestingly, when treating SLC7A10-inhibited adipocytes with the ROS scavenger N-acetyl-l-cysteine (Nac), we observed a 50–70% reduction in lipid accumulation (Fig. 6K), indicating that ROS generation may have partially mediated the lipid-storing effect of reduced SLC7A10 activity. On the other hand, this partial reversal of SLC7A10 inhibitor–dependent lipid accumulation by Nac was not clear upon prolonged stimulation with insulin, which increased lipid accumulation to a degree similar to that with SLC7A10 inhibition (Fig. 6K). These data suggest that the lipid-storing effects of SLC7A10 impairment might at least partially involve increased levels of ROS, whereas the effects of insulin may largely occur independent of ROS generation.

Finally, we tested whether reduced SLC7A10 activity also affects insulin-stimulated glucose uptake. Inhibition of SLC7A10 diminished insulin-stimulated glucose uptake in mouse and human adipocytes (Fig. 7A and B), supporting that SLC7A10 directly modulates adipocyte insulin sensitivity.

We here identified SLC7A10 as a novel facilitator of serine uptake in adipocytes, and that this function may buffer against oxidative stress, lipid accumulation, insulin resistance and dyslipidemia. Our data show that pharmacological inhibition of SLC7A10 in adipocytes decreases glutathione levels (within minutes), increases ROS generation (within an hour), decreases mitochondrial respiratory capacity (within hours) and promotes lipid accumulation (within days). SLC7A10 inhibition also decreases insulin-stimulated glucose uptake. Furthermore, SLC7A10 overexpression showed inverse effects, suggesting that SLC7A10 activation may improve important metabolic functions in adipocytes, potentially counteracting development of obesity and insulin resistance. The overfeeding experiments in zebrafish support that functional impairment of SLC7A10 increases body weight and adipocyte size in vivo.

Our clinical cohort data reveal consistent inverse correlations between adipose SLC7A10 mRNA expression and several key features of insulin resistance, including WHR, adipocyte hypertrophy, visceral fat mass, triacylglycerol, and HOMA of insulin resistance after adjustment for BMI and sex, further underscored by increased SLC7A10 mRNA in people with insulin sensitive compared with insulin resistant obesity. SC adipose SLC7A10 was previously shown to have a strong heritable expression (h² = 0.79), to be lower in people carrying type 2 diabetes risk variants in the KLF14 locus, and to correlate negatively with metabolic traits linked to disease risk (30).

The potential clinical relevance of SLC7A10 in adipocyte metabolism is further supported by previously unconnected lines of evidence from other studies. Firstly, a metabolomics study found reduced uptake of serine in VAT from people with severe obesity compared with nonobese participants (31). Secondly, total glutathione levels are higher in OM than SC AT of lean individuals, and altered glutathione synthesis in adipocytes affects insulin sensitivity (32,33). Moreover, total glutathione levels are reduced in AT of people with obesity compared with lean people (34), in line with the pattern of SLC7A10 expression reported here. Thirdly, SC AT in obesity and type 2 diabetes exhibits increased mitochondrial ROS levels, e.g., H₂O₂, combined with reduced expression of antioxidant enzymes (35). A recent report showed 46% higher H₂O₂ levels in visceral fat of men with central obesity compared with lean men and positive correlations of the adipose H₂O₂ concentrations with insulin resistance (36). Importantly, recent studies suggest that oxidative stress in adipocytes is not only a consequence of metabolic disease but also a cause (37) and that elevated intracellular ROS levels in adipocytes might contribute to adipocyte dysfunction, increased fat storage, and insulin resistance (24,38). Taken together, our study points to SLC7A10 as a potential candidate for therapeutic intervention to mitigate oxidative stress and unhealthy lipid storage in adipocytes.

In line with our experimental data linking reduced SLC7A10 function to increased lipid accumulation via decreased glutathione levels and elevated ROS, glutathione depletion has been found to promote adipogenesis in 3T3-L1 adipocytes (39). Although a recent study in 3T3-L1 adipocytes found that long-term treatment with the ROS scavenger Nac increased ROS levels (40), others found that Nac treatment decreased ROS levels (as expected), while increasing oxygen consumption, decreasing body fat in mice in vivo (41), and inhibiting insulin-stimulated lipid accumulation in 3T3-L1 adipocytes (42), in line with our data. ROS can modulate intracellular signaling and a transient increase in ROS levels can promote adipocyte differentiation (38,43), while sustained elevation of cellular ROS levels has been linked to adipocyte lipid storage (24), also observed in microorganisms (44). Our data show a clear positive relationship between ROS levels and lipid accumulation, in contrast to a recent study in mice where increased mitochondrial levels of the H₂O₂-hydrolyzing enzyme catalase were associated with reduced ROS and increased adiposity, adipocyte size, and adipose glyceroneogenic and lipogenic gene expression (45). Another study also found reduced body weight with increased ROS levels in AT with aging (46). A possible explanation for these inconsistent results might be the specific metabolic contexts and distinct effects of specific sources of ROS on glyceroneogenesis and lipid accumulation, which requires further investigation.

It is possible that inhibition of SLC7A10 and the concomitant increase in ROS levels promoted lipid storage in our study, at least in part, by reducing mitochondrial respiratory capacity. A recent study in SC and visceral human adipose–derived stem cells linked high ROS generation to decreased mitochondrial respiration (47), and increased ROS generation in epididymal fat has been shown to precede lowered mitochondrial biogenesis in nutritionally challenged mice (48). Additionally, elevated mitochondrial and extracellular ROS concentrations have been shown to inhibit mitochondrial respiration and to cause mitochondrial dysfunction in cultured 3T3-L1 and primary rat adipocytes (24,41). Recent studies also showed that ROS can impair insulin-dependent glucose uptake (4,49), in line with the SLC7A10-dependent phenotype we observed.

While our data indicate that altered serine uptake mediated the lipid-storing effects of impaired SLC7A10 function, effects of serine on adipocyte metabolism and mechanisms regulating adipocyte serine flux are largely unknown. Serine is vital in maintenance of mitochondrial respiration (50,51), and both imported and de novo synthesized serine play a role in protein, lipid, and purine metabolism (11). In mouse embryonic fibroblasts lacking the first enzyme in de novo serine synthesis, external l-serine depletion increased formation of specific sphingolipids (52), and serine supplementation in mice reduced hepatic ROS levels, ameliorating alcoholic fatty liver by supporting glutathione levels (53,54).

Our study has limitations. The SLC7A10 inhibitors BMS-466442 and Lu AE00527 have been used to study functions of SLC7A10 in the brain and show high selectivity (19,20,29,55). However, we cannot entirely rule out nonspecific effects, even though overexpression of SLC7A10 showed inverse effects compared with loss of SLC7A10 function in adipocyte cultures. Previous studies have shown the utility of zebrafish for investigating AT biology and the dynamics of obesity and type 2 diabetes development (56,57), and they share common obesity-related pathophysiological pathways with mammals (58). Nonetheless, future studies should perform adipocyte-selective Slc7a10 manipulation, e.g., by overexpression in mice, to determine the degree to which maintained Slc7a10 activity can prevent and reverse obesity and systemic insulin resistance. Additionally, further studies are needed for determination of whether loss of SLC7A10 activity directly in visceral fat, where SLC7A10 mRNA is twofold higher than in SC fat, might render this depot particularly vulnerable to adipocyte hypertrophy and metabolic dysfunction.

In conclusion, our study has identified SLC7A10 as a novel gene involved in the regulation of adipocyte energy metabolism, ROS generation, and lipid accumulation, implicating novel adipocyte pathways linked to serine transport in obesity and insulin resistance.

Acknowledgments. The authors thank Olivera Bozickovic, Margit Solsvik, Iren Drange Hjellestad, and Novin Balafkan at the University of Bergen and Haukeland University Hospital and Øyvind Reinshol at the Institute of Marine Research (technical assistance); personnel at Voss Hospital, Haugesund Hospital, and Haraldsplass Deaconess Hospital, Bergen, Norway (sample collection); and Per Magne Ueland, University of Bergen, for reviewing the manuscript. The Genomics Core Facility (GCF) at the University of Bergen, which is part of the NorSeq consortium, provided services on microarray and RNA-seq global gene expression profiling.

Funding. This project is supported by the Research Council of Norway (263124/F20), K.G. Jebsen Center for Diabetes Research, Western Norway Regional Health Authority, and the Norwegian Diabetes Association, Norway. GCF is supported in part by major grants from the Norwegian Research Council (grant 245979/F50) and the Trond Mohn Foundation (TMS) (Bergens Forskningsstiftelse). The authors also thank Kenneth Vielsted Christensen at H. Lundbeck A/S for providing the SLC7A10 inhibitor Lu AE00527.

Duality of Interest. This project is also supported by Novo Nordisk Scandinavia AS. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. R.Å.J. and S.N.D. designed the study, carried out experiments, analyzed and interpreted the results, and wrote the manuscript. D.S.P.T. and A.M. helped carry out experiments, analyze and interpret data, and write the manuscript. L.S., L.D., A.W., J.-I.B., and M.S.B. assisted with experiments and data analysis. R.Å.J., E.F., L.M., and S.E. performed the zebrafish feeding experiment. A.M. performed metabolomics analyses. V.V., H.J.N., B.G.N., and C.B. planned and carried out collection of AT and clinical data. M.B., P.J., P.-A.S., M.R., P.A., O.N., and M.C. provided and analyzed cohort data. V.M.S. helped design and support the transcriptome analyses. J.F., J.V.S., G.M., and S.N.D. facilitated the laboratory work and collection and analyses of cohort samples. All authors reviewed and approved the final version of the manuscript. S.N.D. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.