Author response: Identification of a weight loss-associated causal eQTL in MTIF3 and the effects of MTIF3 deficiency on human adipocyte function

Abstract

Article Figures and data Abstract Editor's evaluation Introduction Results Discussion Materials and methods Appendix 1 Data availability References Decision letter Author response Article and author information Metrics Abstract Genetic variation at the MTIF3 (Mitochondrial Translational Initiation Factor 3) locus has been robustly associated with obesity in humans, but the functional basis behind this association is not known. Here, we applied luciferase reporter assay to map potential functional variants in the haplotype block tagged by rs1885988 and used CRISPR-Cas9 to edit the potential functional variants to confirm the regulatory effects on MTIF3 expression. We further conducted functional studies on MTIF3-deficient differentiated human white adipocyte cell line (hWAs-iCas9), generated through inducible expression of CRISPR-Cas9 combined with delivery of synthetic MTIF3-targeting guide RNA. We demonstrate that rs67785913-centered DNA fragment (in LD with rs1885988, r2 > 0.8) enhances transcription in a luciferase reporter assay, and CRISPR-Cas9-edited rs67785913 CTCT cells show significantly higher MTIF3 expression than rs67785913 CT cells. Perturbed MTIF3 expression led to reduced mitochondrial respiration and endogenous fatty acid oxidation, as well as altered expression of mitochondrial DNA-encoded genes and proteins, and disturbed mitochondrial OXPHOS complex assembly. Furthermore, after glucose restriction, the MTIF3 knockout cells retained more triglycerides than control cells. This study demonstrates an adipocyte function-specific role of MTIF3, which originates in the maintenance of mitochondrial function, providing potential explanations for why MTIF3 genetic variation at rs67785913 is associated with body corpulence and response to weight loss interventions. Editor's evaluation In this study, Huang et al. perform detailed functional genomics assays in cultured adipocytes to provide mechanistic insight underlying an important obesity GWAS locus. These studies not only demonstrate allele-specific effects of MITF3 as a potential causal gene for variations in rs67785913 and rs1885988 alleles, but further provide a foundational framework from bridging GWAS associations to actionable pathways. The study strengths include genetic manipulation followed by detailed biochemical characterization to mimic and test the impacts of association, where future studies potentially involving in vivo characterizations could further inform the metabolic consequences of these observations. https://doi.org/10.7554/eLife.84168.sa0 Decision letter Reviews on Sciety eLife's review process Introduction Over 650 million people are obese and often suffer from metabolic abnormalities, including dyslipidemia, type 2 diabetes, and hypertension (Adams et al., 2006; May et al., 2020). It is widely believed that obesity results from an interplay between genetic and environmental factors (Thomas, 2010), but the biological mechanisms behind these interactions are poorly understood. Genetic variation (rs12016871) at MTIF3 (encoding the Mitochondrial Translation Initiation Factor 3 protein [Kuzmenko et al., 2014]) has been robustly associated with body mass index (BMI) in humans (Locke et al., 2015). Several subsequent studies have linked MTIF3 genetic variation with the response to weight loss interventions, including diet, exercise, and bariatric surgery (Papandonatos et al., 2015; Rasmussen-Torvik et al., 2015), and with weight-related effects of habitual diet (Nettleton et al., 2015). For example, analyses in two of the world’s largest randomized controlled weight loss trials (Diabetes Prevention Program [DPP] and Look AHEAD) found that homozygous minor allele carriers (rs1885988) were slightly more prone to weight gain in the control arm, yet achieved significantly greater weight loss at 12-month post-randomization and retained lost weight longer (18–36 months) than major allele carriers (Papandonatos et al., 2015). Elsewhere, the same locus has been associated with greater and more sustained weight loss following bariatric surgery (Rasmussen-Torvik et al., 2015). Mtif3 loss in the mouse results in cardiomyopathy owing to impaired translation initiation from mitochondrial mRNAs and uncoordinated assembly of OXPHOS complexes in heart and skeletal muscle (Rudler et al., 2019). In the human hepatocyte-like HepG2 cell line, MTIF3 loss decreases the translation of the mitochondrial-encoded ATP synthase membrane subunit 6 (ATP6) mRNA without affecting cellular proliferation (Chicherin et al., 2020). No human genomic mutations leading to total MTIF3 deficiency have been reported, but the studies outlined above suggest that MTIF3 may influence obesity predisposition and weight loss potential by modulating mitochondrial function; thus, MTIF3 may play a key role in adipose tissue metabolic homeostasis, as adipocyte mitochondria not only provide ATP, but also impact adipocyte-specific biological processes such as adipogenesis, lipid metabolism, thermogenesis, and regulation of whole-body energy homeostasis (Gregoire et al., 1998; Boudina and Graham, 2014). In this study, we aimed to experimentally dissect the molecular mechanisms that could underlie the correlation between MTIF3 genetic variation and weight loss intervention outcomes. We hypothesized that among the common genetic variants in MTIF3, one (or more) is causal for altered MTIF3 expression. Secondly, we hypothesized that MTIF3 content in human white adipocytes influences adipocyte-specific, obesity-related traits under basal and perturbed metabolic conditions. For the latter, we used glucose restriction to mimic the effects of in vivo lifestyle interventions focused on energy restriction and expenditure. Results rs67785913 is a regulatory variant for MTIF3 expression The MTIF3 rs1885988 C allele is associated with enhanced weight loss and weight retention following lifestyle intervention trials in DPP and Look AHEAD cohorts (Papandonatos et al., 2015). In the GTEx database, the rs1885988 associates with an eQTL in subcutaneous fat (Figure 1A), with C allele carriers having significantly higher MTIF3 expression (normalized effect size: 0.15, p = 0.0000032). Figure 1 with 1 supplement see all Download asset Open asset Identification of rs67785913 as a causal cis-eQTL for MTIF3. (A) Violin plot of MTIF3 expression in subcutaneous adipose tissue for rs1885988 from Genotype-Tissue Expression (GTEx) Project eQTL. (B) Same as in (A), but for rs67785913. (C) Representative Sanger sequencing traces of rs67785913 CTCT/CTCT and CT/CT clones obtained after CRISPR/Cas9-mediated allele editing and single-cell cloning. (D) Normalized Z-score plot of luciferase reporter assays using vectors carrying different DNA fragments of the MTIF3 gene cloned into pGL4.23 luciferase reporter vector. Hypothesis testing was performed by comparing the transcriptional enhancer activity of each of the 12 vectors (F1–12) to the empty vector (minP). All data were plotted as mean ± standard deviation (SD), n = 4 independent experiments, p values are presented in each graph; ordinary one-way analysis of variance (ANOVA) was used for statistical analysis. (E) Relative MTIF3 expression (mRNA) in rs67785913 allele-edited cells 2 days before, at, or 2 days post-differentiation induction (day −2, 0, and 2, respectively). n = 3 clonal populations for CTCT/CTCT genotype, n = 5 clonal populations for CT/CT genotype, error bars show SD. (F) as in (E), but for GTF3A (mRNA) expression. Two-tailed Student’s t-test was used; p values are presented in each graph. To experimentally validate and fine map the potential causal DNA variation in the haplotype block tagged by rs1885988, we looked up all tightly linked (r2 > 0.8) single-nucleotide polymorphisms (SNPs) in HaploReg database v4.1 (Ward and Kellis, 2012). We then PCR-amplified and cloned 12 DNA fragments from that haploblock, altogether comprising the linked SNP loci, into luciferase reporter plasmids. As shown in Figure 1D, by comparing the luciferase signals with minimal promoter (minP) construct, only one DNA fragment (F11), encompassing the rs67785913 locus, could enhance luciferase transcription. Coincidentally, the rs67785913 also shows an eQTL effect on MTIF3 expression in subcutaneous adipose tissue in GTEx database, with the major CT allele associated with lower expression than the minor CTCT allele (normalized effect size: −0.16, p = 3.0 × 10−8) (Figure 1B). To demonstrate an allele-specific regulatory effect on MTIF3 expression, we then used CRISPR-Cas9 to substitute the major CT for the minor CTCT allele at the rs67785913 locus in the pre-adipocyte hWAs cell line. Due to rather low CRISPR editing efficiency of that locus, we needed to genotype over 700 single-cell clones to obtain five CT/CT and three CTCT/CTCT clones without random indels, as confirmed by Sanger sequencing (Figure 1C). We then examined MTIF3 expression in these clones at pre- and post-adipogenic differentiation induction, and found rs67785913 CTCT/CTCT to confer higher MTIF3 expression at all time points (Figure 1E), although without apparent change on adipogenic differentiation markers (Figure 1—figure supplement 1). As rs67785913 also correlates with an altered GTF3A expression in other tissues (e.g., muscle, lung), we also detected, but found no apparent difference in GTF3A expression in rs67785913-edited cells (Figure 1F). Generating inducible Cas9-expressing pre-adipocyte cell line (hWas-iCas9) Next, we intended to use the rs67785913-edited cells in functional genomics experiments to examine the phenotypic consequences of the eQTL. To conduct meaningful studies of gene × environment interaction, it is desirable to use similarly differentiated cells with comparable baselines (e.g., similar triglyceride or mitochondrial content). Unfortunately, marginally different passage numbers between control and experimental groups can confound adipogenic differentiation. This problem can originate during single-cell cloning to create genetic knockouts/knockins, and became apparent with our rs67785913 allele-edited cells. While the mean values of adipogenic markers were similar in both rs67785913 genotypes (Figure 1—figure supplement 1), the variation between clones of the same genotype precluded the use of these cells in gene × environment studies. To circumvent this, we instead established an inducible Cas9-expressing pre-adipocyte cell line that allowed us to knockout MTIF3 after completed adipogenic differentiation. As illustrated in Figure 2, we co-transfected hWAs with two plasmids: one encoding piggyBac transposase, and the other carrying piggyBac transposon-flanked doxycycline-inducible Cas9 and constitutively expressed puromycin resistance genes. In this setup, piggyBac transposase drives the integration of the piggyBac-transposon-flanked genes, and transgenic cells are then selected and expanded in puromycin-supplemented culture medium. We have thus obtained an hWAs cell line with doxycycline-inducible Cas9 expression and maintained adipogenic differentiation capacity (henceforth called hWAs-iCas9). The Cas9-expressing differentiated cells could then be transfected with relatively low molecular weight synthetic single guide RNAs (sgRNAs) that complex with intracellularly expressed Cas9 and target the gene exon of interest to generate random indels (in essence, gene knockouts). We used this method here to determine the functional role of MTIF3 in adipocyte biology. Figure 2 Download asset Open asset The workflow of establishing hWAs-iCas9 cell line and its application in studying MTIF3 and environment interactions in vitro. Generation of MTIF3 knockout in hWAs-iCas9 mature adipocytes To investigate the role of MTIF3 in human adipocyte development and energy metabolism we generated stable MTIF3 knockouts in differentiated hWAs-iCas9 adipocytes. We designed Cas9-specific sgRNA to generate random indels in the exon expressed in all three MTIF3 protein-encoding transcripts (Figure 3A) and obtained a >80% reduction in MTIF3 protein levels in every experiment, as assessed by western blotting (Figure 3B–D, Figure 3—figure supplement 1, and Figure 4A, B). To assess off-target effects of CRISPR-Cas9, we also performed T7EI assays on PCR-amplified top 5 predicted off-target sites and did not observe any detectable off-targeting (data are not shown). Figure 3 with 1 supplement see all Download asset Open asset MTIF3 perturbation in mature adipocytes does not affect adipocyte-specific protein expression or total triglyceride content. (A) An illustration of Cas9-specific single guide RNA (sgRNA)-binding site in the exon expressed in all three MTIF3 protein-encoding transcripts. (B) Representative Sanger sequencing of control and knockout hWAs mature adipocytes. (C) Immunoblots of adipocyte markers in scrambled control and MTIF3 knockout adipocytes, n = 5 independent experiments. (D) Quantitative analysis of MTIF3 band densities in (C). (E) Quantitative analysis of ACC, FABP4, and FAS band densities in (C). (F) Representative Oil-red O staining images of control and MTIF3 knockout in hWAs mature adipocytes. Scale bar is 200 µm. (G) Total triglyceride content in scrambled control (SC) and MTIF3 knockout (KO) cells. n = 3 independent experiments. Error bars show standard deviation in all plots. Statistical analysis was performed using two-tailed Student’s t-test, p values are presented in each graph. Uncropped blot images for (C) and raw.scn data files can be found in Figure 3—source data 1. Figure 3—source data 1 Raw data files for western blots shown in Figure 3C. https://cdn.elifesciences.org/articles/84168/elife-84168-fig3-data1-v2.zip Download elife-84168-fig3-data1-v2.zip Figure 4 Download asset Open asset MTIF3 perturbation in mature adipocytes disrupts mitochondrial gene expression and OXPHOS complex assembly. (A) Immunoblots of mitochondrial genome-encoded proteins in scrambled control and MTIF3 knockout adipocytes. (B) Quantitative analysis of band densities in (A). (C) qPCR for mitochondrial gene expression in scrambled control and MTIF3 knockout adipocytes, n = 5 independent experiments. (D) Relative mitochondrial DNA content in scrambled control and MTIF3 knockout adipocytes, n = 5 independent experiments. (E) Immunoblots of mitochondrial OXPHOS complex assembly after Blue Native-PAGE electrophoresis, n = 4 independent experiments. (F) Quantitative analysis of band densities in (E). Error bars show standard deviation in all plots. Statistical analysis was performed using two-tailed Student’s t-test, p values are presented in each graph. Uncropped blot images for (A) and raw.scn data files can be found in Figure 4—source data 1. Uncropped blot images for (E) and raw.scn data files can be found in Figure 4—source data 2. Figure 4—source data 1 Raw data files for western blots shown in Figure 4A. https://cdn.elifesciences.org/articles/84168/elife-84168-fig4-data1-v2.zip Download elife-84168-fig4-data1-v2.zip Figure 4—source data 2 Raw data files for western blots shown in Figure 4E. https://cdn.elifesciences.org/articles/84168/elife-84168-fig4-data2-v2.zip Download elife-84168-fig4-data2-v2.zip MTIF3 knockout in mature adipocytes does not affect adipogenic marker or lipid content Although the MTIF3 knockout in hWAs-iCas9 cells was generated after the cells were differentiated, we wanted to ensure the genetic perturbation did not affect adipogenic markers or triglyceride content, as that could confound results from downstream functional studies. Incidentally, we observed that the quantities of the adipogenic markers, including ACC, FABP4, and FAS were comparable in control and MTIF3 knockout cells (Figure 3C, E; see also Figure 3—figure supplement 1). Similarly, there were no apparent differences in Oil-red O or total triglyceride content (Figure 3F, G). MTIF3 knockout disrupts mitochondrial DNA-encoding gene and protein expression, mitochondrial content, as well as mitochondrial OXPHOS assembly in hWAs-iCas9 adipocytes MTIF3 is a mitochondrial translation initiation factor; thus, we examined the effects of MTIF3 ablation on differentiated hWAs adipocyte mitochondrial respiration chain. Assessed by western blotting, the MTIF3 knockout cells had significantly decreased COX II (subunit of OXPHOS complex IV) and ND2 (subunit of OXPHOS complex I), trending decrease of CYTB (subunit of OXPHOS complex III), and unchanged ATP8 (subunit of OXPHOS complex V) content (Figure 4A , B). Moreover, using qPCR, we observed an altered expression of several mitochondrial DNA-encoding genes. Specifically, MTIF3 deficiency led to higher expression of MT-ND1, MT-ND2, a trending increase of MT-ND4, and lower expression of MT-ND3, and MT-CO3 (Figure 4C). In addition, we also found significantly reduced mitochondrial content in MTIF3 knockout adipocytes (Figure 4D). Taken together, the above data suggest MTIF3 knockout disrupts mitochondrial DNA-encoding gene and protein expression. Next, we hypothesized the above observations could have originated from the insufficient MTIF3 supply during OXPHOS complex assembly (a role previously ascribed to MTIF3 [Rudler et al., 2019]). To test this, we used Blue Native-PAGE to examine OXPHOS complexes in mitochondria isolated from MTIF3 knockout adipocytes. MTIF3 deficiency led to decreased complex III2/IV2 and IV1, and a trending decreased complex V/III2 + IV1 assembly. In contrast, OXPHOS complex II assembly was significantly increased in MTIF3 knockout cells (Figure 4E, F). Interestingly, we also observed faster-migrating undefined bands in MTIF3 knockout adipocytes (Figure 4E), which could be single chain proteins, or mistranslation or degradation products. Lastly, The OXPHOS complex I + III2 + IVn appeared to be less abundant in MTIF3 knockout mitochondria, although the bands appeared more diffuse and could not be quantified (Figure 4E). MTIF3 knockout affects mitochondrial respiration in hWAs-iCas9 adipocytes Having established the role of MTIF3 in adipocyte mitochondria OXPHOS complex assembly, and in mitochondrial gene expression, we then investigated the mitochondrial function in MTIF3-ablated differentiated hWAs adipocytes using Seahorse Mito Stress Test. Additionally, to avoid potential cofounders caused by the high glucose content in the differentiation medium, we adapted the cells to 1 g/l glucose growth medium for 3 days before running the assay. As shown in Figure 3, MTIF3 knockout cells exhibited lower basal oxygen consumption rate (OCR), as well as lower ATP-forming capacity, the latter estimated by calculating OCR decrease after blocking ATP synthase with oligomycin (Figure 5A–C). MTIF3 knockout cells also showed a trending decrease in maximal respiration OCR (Figure 5D). Furthermore, both MTIF3 knockout and control cells, had comparable proton leak OCR, non-mitochondrial respiration OCR and coupling efficiency (Figure 5E–G). Figure 5 Download asset Open asset Cellular mitochondrial respiration in hWAs adipocytes. (A) The average oxygen consumption rate (OCR) traces during basal respiration, and after addition of oligomycin, FCCP, and rotenone/antimycin A. (B) Basal respiration OCR, n = 4 different cell passages. (C) ATP production OCR, n = 4 different cell passages. (D) Maximal respiration OCR, n = 4 different cell passages. (E) Proton leak OCR, n = 4 different cell passages. (F) Non-mitochondrial respiration OCR, n = 4 different cell passages. (G) Coupling efficiency, n = 4 different cell passages. Error bars show standard deviation. Statistical analyses were performed using paired Student’s t-test in each condition, p values are presented in each graph. MTIF3 knockout affects hWAs-iCas9 adipocyte endogenous fatty acid oxidation Next, we used Seahorse to assess the endogenous fatty acid oxidation in MTIF3 knockout versus control cells treated with etomoxir (an inhibitor of carnitine palmitoyl transferase). We found that MTIF3 ablation mimics the effect of etomoxir on basal endogenous fatty acid oxidation OCR. Furthermore, while etomoxir decreases basal fatty acid oxidation OCR in control cells, it does not markedly decrease it in MTIF3 knockout cells (Figure 6A, B). Figure 6 with 1 supplement see all Download asset Open asset MTIF3 perturbation affects adipocyte fatty acid oxidation. (A) A representative Seahorse oxygen consumption rate (OCR) trace for endogenous fatty acid oxidation assay. MTIF3 knockout and scrambled control adipocytes were treated with or without etomoxir for 15 min before the assay. Following the basal OCR measurement, oligomycin, FCCP (carbonyl cyanide-p-trifluoromethoxyphenylhydrazone), and rotenone + antimycin A were added sequentially to measure the detection of ATP production OCR, maximal respiration OCR and non-mitochondrial respiration OCR. (B) Basal endogenous fatty acid oxidation OCR in scrambled control (SC) and MTIF3 knockout (KO) adipocytes, n = 4 independent experiments. (C) Upper panel: workflow of glucose restriction in differentiated adipocytes; Lower left panel: total triglyceride content in scrambled control (SC) and MTIF3 knockout (KO) adipocytes in 25 mM glucose conditions; Lower right panel: triglyceride content in adipocytes cultured in glucose-restricted conditions (5, 3, and 1 mM) relative to adipocytes cultured in 25 mM glucose, n = 4 independent experiments. (D) Z-score-normalized data for glycerol release in scrambled control and MTIF3 knockout adipocytes under basal, insulin-stimulated, and isoproterenol-stimulated conditions, n = 4 independent experiments. (E) qPCR for mitochondrial and adipocyte-related gene expression in scrambled control and MTIF3 knockout adipocytes. Error bars show standard deviation in all plots. Statistical analysis was performed using two-tailed Student’s t-test, p values are presented in each graph. MTIF3 knockout affects hWAs-iCas9 adipocyte triglyceride content after glucose restriction challenge To mimic the interactions between MTIF3 content and dietary intervention on weight change, we generated hypertrophic control and MTIF3 knockout hWAs-iCas9 adipocytes and then used glucose-limited medium, not supplemented with free fatty acids (FFAs), to mimic energy restriction in vivo (schematic shown in Figure 6C). Triglyceride content decreased both in control and MTIF3 knockout cells after 3 days of different levels of glucose restriction when compared with 25 mM glucose medium. Interestingly, a more extensive decrease in triglyceride content occurred in control cells cultured in 1 mM glucose medium (p = 0.053), and a similar trending decrease, albeit with higher coefficient of variation, occurred in 3 and 5 mM glucose medium (Figure 6C). MTIF3 knockout does not affect lipolysis-mediated glycerol release in hWAs adipocytes Owing to the effects of MTIF3 ablation on triglyceride content and on fatty acid oxidation, described above, we then examined the effects of MTIF3 knockout on lipolysis. We measured basal, insulin-attenuated, and isoproterenol-stimulated glycerol release in differentiated hWAs cells. As shown in Figure 6D, in all three conditions, glycerol release in control and MTIF3 knockout cells was comparable. In addition, basal glycerol release in glucose-restricted conditions was similar (Figure 6—figure supplement 1), and significantly reduced in low glucose versus high glucose assay medium. MTIF3 knockout affects mitochondrial function- and fatty acid oxidation-related gene expression Next, we examined how MTIF3 ablation affects the gene expression programmes pertinent to mitochondrial function, fatty acid oxidation, lipolysis and lipid catabolism. As shown in Figure 6E, MTIF3 knockout cells had decreased expression of the mitochondria-related MT-CO1, and the fatty acid oxidation-related ACADM and ACAT1, but unchanged expression of other genes involved in mitochondrial function and lipid metabolism (TFAM, TOMM20, PRDM16, CPT1B, ABHD5, PNPLA2, and ACACB). MTIF3 knockout results in glucose level-depending alterations in metabolism Considering the observed effect of MTIF3 ablation on mitochondrial function and fatty acid oxidation, we assessed the metabolite profile in MTIF3 knockout cells. Using combined GC/MS and LC/MS metabolite profiling resulted in relative quantification of 110 metabolites. First, we analyzed metabolite profiles at a global level using PCA. The score plot reveals a clear systematic difference in the metabolite profile between cells in 25 mM glucose versus in glucose restriction (Figure 7A). Interestingly, differences between MTIF3 knockout and control cells at 25 mM glucose are observed along principal component 1 (PC1), whereas differences between genotypes at glucose restriction are observed along PC2, suggesting the effect of MTIF3 ablation to depend on the calorie level. Next, to identify alterations in metabolite levels underlying this differential response, we analyzed data using orthogonal projections to latent structures discriminant analysis (OPLS-DA) separately at 25 mM glucose condition (two components R2 = 0.82, Q2 = 0.66) and at glucose restriction (two components, R2 = 0.95, Q2 = 0.52). These analyses revealed systematic differences between genotypes at both growth conditions (Figure 7 B, C). Next, to examine whether the differences between genotype depended on growth condition, we combined the correlations from the two OPLS-DA models into a shared and unique structures plot (Figure 7D). These analyses revealed levels of intermediates in cytosolic metabolic pathways connected to the glycolysis, such as glycerate 3-phosphate, glycerol 2-phosphate, UDP-N-acetylglucosamine, and ribose 5-phosphate, to be lower in MTIF3 knockout cells at both glucose concentrations. Interestingly, levels of fatty acids, ranging from 9 to 17 carbons and including several odd-chain fatty acids, were lower in the MTIF3 knockout cells only at 25 mM glucose condition. At glucose-restricted conditions, levels of both essential and non-essential amino acids were lower in the knockout. Finally, we analyzed data using two-way analysis of variance (ANOVA), incorporating glucose concentration and genotype, thereby providing information on effects at the individual metabolite level. These analyses revealed 18 and 20 significantly different metabolites between control and MTIF3 knockout cells at 25 mM glucose condition and glucose-restricted conditions, respectively (q <0.05). These included ribose 5-phosphate, glycerate 3-phosphate, glycerol 2-phosphate, and glycerol 3-phosphate (Figure 7E). Figure 7 Download asset Open asset Mass spectrometry-based metabolomics data for control (SC) and MTIF3 knockout (KO) cells in 25 mM glucose (NF, normal feeding) and 5 mM glucose (GR, glucose-restricted) conditions. (A) Principal component analysis (PCA) score plot displaying the discrimination between MTIF3 knockout and control cells in normal and glucose-restricted conditions (PC1: 28%, PC2: 19%). (B) Orthogonal projections to latent structures discriminant analysis (OPLS-DA) score plot showing classification of MTIF3 knockout and control cells in 25 mM glucose condition. (C) OPLS-DA score plots showing classification of MTIF3 knockout and control cells in glucose-restricted condition. (D) Shared and unique structures (SUS) plot, based on OPLS-DA models in (B, C), showing glucose concentration-dependent differences between MTIF3 knockout and control cells. (E) Box plots showing the abundance of some of the significantly altered metabolites in normal and MTIF3 knockout cells in normal and glucose-restricted conditions. Statistical analysis was performed using two-way analysis of variance (ANOVA) test, p values are presented in each graph. Discussion Excessive weight gain caused by dietary excess, and its effects on adipocyte lipid metabolism, can cause life-threatening disease (Appleton et al., 2013; Denis and Obin, 2013; Chu et al., 2017; Lotta et al., 2018). Findings from clinical trials (Papandonatos et al., 2015), a bariatric surgery case series (Rasmussen-Torvik et al., 2015) and epidemiological cohorts (Nettleton et al., 2015) showed the MTIF3 variation modulates weight loss-promoting exposures on body weight (see also UK Biobank analysis in Supplementary file 1b). Here, we validated MTIF3 rs1885988 C allele correlates with higher MTIF3 expression in subcutaneous fat tissue, and our in vitro luciferase reporter assay and CRISPR-Cas9 genome editing results revealed that the tightly linked rs67785913 variant is likely to be the actual eQTL for MTIF3 expression (Figure 1). Pre-adipocyte cell lines have been used extensively for studying adipogenic differentiation (Ruiz-Ojeda et al., 2016), but their application for lipid metabolism studies has been limited, especially in the context of gene–environment interaction. This is largely due to the variation of differentiation capacity across cell passages (Poulos et al., 2010) and differential genetic effects on adipocyte differentiation (Kamble et al., 2020), which can alter baseline phenotypes in differentiated cells. Therefore, we established an inducible Cas9-expressing human pre-adipocyte cell line (hWAs-iCas9), which enabled us to generate gene knockout of interest in differentiated adipocytes, thus circumventing these limitations (Figure 2). Using the inducible knockout cell model, we then tested interactions between MTIF3 and environmental changes (lifestyle mimetics). Our data reveal that MTIF3 deficiency mediated disrupted mitochondrial respiration, probably as a consequence of decreased OXPHOS complex assembly. This led to perturbed cellular functions, including reduced fatty acid oxidation and a trending increase in intracellular