Adipose Mitochondrial Complex I Deficiency Modulates Inflammation and Glucose Homeostasis in a Sex-Dependent Manner

Abstract

Mitochondrial dysfunction in adipose tissue has been associated with type 2 diabetes, but it is unclear whether it is a cause or the consequence. Mitochondrial complex I is a major site of reactive oxygen species generation and a therapeutic target. Here we report that genetic deletion of the complex I subunit Ndufs4 specifically in adipose tissue results in an increased propensity to develop diet-induced weight gain, glucose intolerance, and elevated levels of fat inflammatory genes. This outcome is apparent in young males but not in young females, suggesting that females are relatively protected from the adverse consequences of adipose mitochondrial dysfunction for metabolic health. Mutant mice of both sexes exhibit defects in brown adipose tissue thermogenesis. Fibroblast growth factor 21 (FGF21) signaling in adipose tissue is selectively blunted in male mutant mice relative to wild-type littermates, consistent with sex-dependent regulation of its autocrine/paracrine action in adipocytes. Together, these findings support that adipocyte-specific mitochondrial dysfunction is sufficient to induce tissue inflammation and can cause systemic glucose abnormalities in male mice.

Mitochondria in adipose tissue are involved in key metabolic processes, such as lipolysis, lipogenesis, and adaptive thermogenesis (1, 2). A number of studies have reported reduced levels of mitochondrial amount or function in adipose tissues of patients with type 2 diabetes (3-5). This includes downregulation of oxidative phosphorylation genes independently of obesity, reduced mtDNA content, and increased reactive oxygen species (ROS) production and oxidative damage in adipose tissues from subjects with type 2 diabetes. Rodent models of diabetes, including db/db mice, exhibit comparable mitochondrial phenotypes (6, 7). However, it has not been established whether these mitochondrial abnormalities occur as the result of diabetes or are themselves involved in its pathogenesis. Adipocyte-specific deletion of mitochondrial genes in mice may be useful in distinguishing between causation and correlation in this context. Previous studies modeled adipose mitochondrial dysfunction by genetic disruption of mitochondrial transcription factor A (TFAM) in mice (8, 9). Targeting TFAM using the aP2 promoter decreased mtDNA copy number and complex I and complex IV activity in adipose tissue. Yet these mice were protected from diet-induced obesity and insulin resistance, likely due to incomplete recombination in some adipose depots and off-target recombination in non-adipose tissues (8, 9). When the more efficient and adipocyte-specific adiponectin promoter was used to delete TFAM, widespread adipocyte death occurred, resulting in a lipodystrophic syndrome and systemic insulin resistance (9). Thus, neither of these models faithfully recapitulates the metabolic alterations that usually accompany type 2 diabetes.

Mitochondrial complex I consists of 14 central subunits and some 30 accessory subunits and is the largest complex in the respiratory chain (10). Complex I contributes to the formation of mitochondrial ROS and impaired complex I function is often implicated in neurological disorders and metabolic abnormalities. Multiple complex I genes are reduced in visceral fat from type 2 diabetes patients (3) and impaired complex I activity in leukocytes has been reported in type 2 diabetes and insulin resistance (11, 12). We reasoned that deletion of a single mitochondrial complex subunit is less likely to cause adipocyte death and may provide a more appropriate model. Therefore, we targeted the complex I subunit Ndufs4, which is implicated in the assembly or stability of the complex. Global Ndufs4 knockout (KO) mouse is the most extensively characterized mouse model of complex I deficiency and leads to progressive neurological deficits and premature death at about 7 weeks of age (13). We report here the metabolic consequences of deleting Ndufs4 specifically in adipocytes using the adiponectin promoter. Ablation of adipose tissue Ndufs4 produces no apparent phenotype under standard conditions, but it severely impairs cold adaptation, consistent with a thermogenic defect. Significantly, adipose-specific Ndufs4 KO mice exhibit increased susceptibility to diet-induced obesity and insulin resistance in a sex-dependent manner. Our data demonstrate that a monogenic defect involving a single mitochondrial complex subunit in adipocytes can contribute to systemic insulin resistance. Furthermore, sexual dimorphism is observed in this context.

Methods

Mice and Diets

All animal experiments were approved by the UC Davis Institutional Animal Care and Use Committee (IACUC) and were conducted with accepted standards of humane animal care as outlined in the Code of Ethics of the Endocrine Society. Floxed Ndufs4 mice (B6.129S4-Ndufs4^tm1Rpa/J, RRID:IMSR_JAX:026963; https://scicrunch.org/resolver/RRID:%20IMSR_JAX:026963) and adiponectin-Cre mice (B6;FVB-Tg(Adipoq-cre)1Evdr/J, RRID:IMSR_JAX:010803; https://scicrunch.org/resolver/RRID:%20IMSR_JAX:010803) were from the Jackson Laboratory. Floxed littermates without the Adipo-Cre transgene were used as controls. Unless otherwise noted, 8- to 10-week-old mice were used for experiments. Mice were maintained at 22 °C and fed a standard chow diet (Envigo Teklad Global Rodent Diets) or a high-fat diet with 60 kcal% fat (Research Diets D12492).

Protein Extraction and Western Blotting

Tissue lysates were prepared by homogenization in RIPA buffer supplemented with protease inhibitor (Roche). Protein samples were separated on 4% to 12% NuPAGE gels (Invitrogen) and transferred to nitrocellulose membrane. The primary antibodies used in this study were anti-Ndufs4 (1:3000, Thermo Fisher Scientific Cat# PA5-21677, RRID:AB_11152796; https://scicrunch.org/resolver/RRID:%20AB_11152796), anti-OXPHOS (1:1000, Abcam Cat# ab110413, RRID:AB_2629281; https://scicrunch.org/resolver/RRID:%20AB_2629281), anti-Fabp4 (1:3000, Santa Cruz Biotechnology Cat# sc-271529, RRID:AB_10650265; https://scicrunch.org/resolver/RRID:%20AB_10650265), and anti-Vinculin (1:2000, Sigma-Aldrich Cat# V9131, RRID:AB_477629; https://scicrunch.org/resolver/RRID:%20AB_477629).

Histology

Tissues from male and female mice were isolated and fixed in 10% neutral buffered formalin (Fisher Scientific) overnight at room temperature. On the following day, tissues were dehydrated and embedded in paraffin. The paraffin sections were stained with hematoxylin and eosin. Microscope images were obtained by Nikon Eclipse 80i under 20× magnification.

8-Hydroxy-2-Deoxyguanosine Enzyme-Linked Immunosorbent Assay

Genomic DNA was extracted from subcutaneous white adipose tissue (scWAT) and brown adipose tissue (BAT) of 21-week-old high-fat diet (HFD)-fed mice using DNeasy Blood & Tissue kit (Qiagen). Five-microgram samples of DNA were digested into mononucleotides and mononucleosides by treating with nuclease P1 (New England Biolabs) and alkaline phosphatase (New England Biolabs), respectively. Digested DNA samples and standards were added to 96-well plates and assayed using the 8-hydroxy-2-deoxyguanosine (8-OH-dG) enzyme-linked immunosorbent assay (ELISA) kit (Abcam Cat# ab201734, RRID:AB_2904542; https://scicrunch.org/resources/Any/search?q=AB_2904542&l=AB_2904542).

Thermoregulation Assay

Eight-week-old female or male wild-type (WT) and KO mice were used for core rectal temperature assessment upon acute cold exposure. Mice were individually housed at 7 °C for 4 hours without food but with free access to water, and body temperature was monitored with a thermometer (Bioseb, Vitrolles, France) every hour. At the end of the cold exposure, surface temperature images were taken by an infrared camera (FLIR) and euthanized for tissue collection.

RNA Extraction and Real-Time Quantitative PCR

Total RNA was extracted from tissues using Trizol (ThermoFisher) and purified with RNA mini spin columns (Epoch Life Science). Reverse transcription was performed using SuperScript IV reverse transcriptase (ThermoFisher). The synthesized cDNA was subjected to quantitative polymerase chain reaction (qPCR) using SYBR Green (Bio-Rad) in a Bio-Rad CFX real-time PCR system (denaturation at 95 °C for 15 seconds, annealing and extension at 58 °C for 30 seconds, 45 cycles). Primer sequences are listed in Table 1.

Table 1.

PCR primers used in this study

Name	Foward (5′ to 3′)	Reverse (5′ to 3′)	Purpose
Ndufs4 flox	AGTCAGCAACATTTTGGCAGT	GAGCTTGCCTAGGAGGAGGT	Genotyping
Adipo endo	GGATGTGCCATGTGAGTCTG	ACGGACAGAAGCATTTTCCA	Genotyping
Adipo TG	CTAGGCCACAGAATTGAAAGATCT	GTAGGTGGAAATTCTAGCATCATCC	Genotyping
Pgc1a	GAAAGGGCCAAACAGAGAGA	GTAAATCACACGGCGCTCTT	qPCR
Prdm16	CAGCACGGTGAAGCCATTC	GCGTGCATCCGCTTGTG	qPCR
Ucp1	ACTGCCACACCTCCAGTCATT	CTTTGCCTCACTCAGGATTGG	qPCR
Cidea	TGCTCTTCTGTATCGCCCAGT	GCCGTGTTAAGGAATCTGCTG	qPCR
Cox5a	GGGTCACACGAGACAGATGA	GGAACCAGATCATAGCCAACA	qPCR
Dio2	CAGTGTGGTGCACGTCTCCAATC	TGAACCAAAGTTGACCACCAG	qPCR
Ccl2	AGGTCCCATGTCATGCTTCTGG	CTGCTGCTGGTGATCCTCTTG	qPCR
Tgfb1	CTCCCGTGGCTTCTAGTGC	GCCTTAGTTTGGACAGGATCTG	qPCR
Tnfa	ATGGCCTCCCTCTCATCAGT	TTTGCTACGACGTGGGCTAC	qPCR
Il1b	ATGCCACCTTTTGACAGTGAT	AGCCCTTCATCTTTTGGGGT	qPCR
Fgf21	GAAGCCCACCTGGAGATCAG	CAAAGTGAGGCGATCCATAGAG	qPCR
Fgfr1	CTGAAGGAGGGTCATCGAAT	GTCCAGGTCTTCCACCAACT	qPCR
Klb	CAGAGAAGGAGGAGGTGAGG	CAGCACCTGCCTTAAGTTG	qPCR
Il10	GCTCTTACTGACTGGCATGAG	CGCAGCTCTAGGAGCATGTG	qPCR
Il6	CCCCAATTTCCAATGCTCTCC	GGATGGTCTTGGTCCTTAGCC	qPCR
Cd68	TCCAAGATCCTCCACTGTTG	ATTTGAATTTGGGCTTGGAG	qPCR
Pparg	GAAAGACAACGGACAAATCACC	GGGGGTGATATGTTTGAACTTG	qPCR
Fabp4	AAGGTGAAGAGCATCATAACCCT	TCACGCCTTTCATAACACATTCC	qPCR
Serbp1c	ATCGGCGCGGAAGCTGTCGG	GGGAAGTCACTGTCTTGGTTG	qPCR
Scd1	GCTGGAGTACGTCTGGAGGAA	TCCCGAAGAGGCAGGTGTAG	qPCR
Dgat2	TACTCCAAGCCCATCACCAC	CAGTTCACCTCCAGCACCTC	qPCR
β-Actin	CTAAGGCCAACCGTGAAAAG	ACCAGAGGCATACAGGGACA	qPCR

Glucose Tolerance Test

After 10 weeks on a HFD, 18-week or 24-week-old mice were fasted overnight and received an intraperitoneal injection of glucose. Female mice were given 2 g/kg of glucose. Male mice all weighed over 35 g at the end of the HFD experiment and were given 1 g/kg to avoid an excessive glucose load. Blood glucose was monitored at indicated time points using a glucometer. Fasting serum glucose was measured in the following week after a 6-hour fast.

Serum Biochemical Analysis

After 13 weeks on a HFD, 21-week-old mice were fasted for 6 hours and euthanized. Blood samples were collected immediately afterwards, and the serum separated using Microvette 200 capillary blood collection tube with serum-gel (Sarstedt). Serum was stored at −80 °C until use. Serum insulin and serum fibroblast growth factor 21 (FGF21) were assessed using an insulin ELISA kit (Alpco Diagnostics Cat# 80-INSMSU-E01, RRID:AB_2792981; https://scicrunch.org/resolver/RRID:%20AB_2792981) and Mouse/Rat FGF-21 Quantikine ELISA Kit (R and D Systems Cat# MF2100, RRID:AB_2783730; https://scicrunch.org/resolver/RRID:%20AB_2783730), respectively.

RNA Sequencing

Total RNA samples were extracted from subcutaneous white adipose tissue (scWAT) of 21-week-old HFD-fed mice using Trizol (ThermoFisher) and purified with DNase-treated column (Zymo Research). Sequencing libraries were prepared and high-throughput 3′-Tag-sequencing performed using a HiSeq 4000 system (Illumina) at the UC Davis Genome Center. The RNA-seq data were deposited to the National Center for Biotechnology Information Gene Expression Omnibus under the accession number GSE190645.

Sequencing Data Analysis

Raw sequence reads were preprocessed using HTStream (https://github.com/s4hts/HTStream) which removed technical features such as Illumina adapters, polyAT sequences, reads less than 50 base pairs in length, and regions at the end of reads with an average quality score below 20. Unique molecular indices (UMIs) were also extracted from sequence reads using a custom python script to later be used for PCR deduplication. The remaining reads were then aligned to the GRCm39 mouse genome (accession: GCA_000001635.9) using STAR (14), followed by PCR duplicate removal via UMI-tools (15). A gene counts table for each sample was then generated using featureCounts (16) and the corresponding GENCODE GRCm39 annotation. Using R (17) and the limma voom pipeline (18, 19), differential expression analysis was conducted. This consisted of normalization, statistical testing, and multiple testing correction using the Benjamini-Hochberg procedure (20) to control the false discovery rate. Significantly enriched gene ontology (GO) terms were identified using DAVID bioinformatics resources 6.8 (21, 22) and plotted using R (version 4.0.2).

Statistical Analysis

Statistical analyses were performed using GraphPad Prism (version 8.2) and Microsoft Excel. Two-tailed Student’s t test was used to determine the statistical significance between 2 independent data points. For time series comparison, a 2-way analysis of variance (ANOVA) followed by Fisher’s least significant difference test was used. Sex-genotype interaction was determined by a 2-way ANOVA followed by Tukey post hoc test. Statistical significance was considered at P ≤ 0.05 (in Figures, * P ≤ 0.05, ** P ≤ 0.01, *** P ≤ 0.001).

Results

Deletion of Ndufs4 in Adipose Tissue Selectively Reduces Mitochondrial Complex I Protein Levels

To determine the consequences of selectively ablating Ndufs4 in adipocytes, we obtained floxed Ndufs4 mice with loxP sites flanking exon 2 and bred with the adiponectin-Cre (Adipo-Cre) mice to generate Adipo-Ndufs4 KO mice (Fig. 1A). We confirmed the deletion of Ndufs4 in BAT, where it is highly expressed, as well as in scWAT from the inguinal fat pad and visceral white adipose tissue (vWAT) from the perigonadal (epididymal or periovarian) fat pad (Fig. 1B). The Ndufs4 protein level in the brain was not affected, as expected. Loss of Ndufs4 reduced the level of Ndufb8, another complex I subunit, consistent with a reduction in other complex I proteins besides Ndufs4 itself (Fig. 1C and 1D). However, Sdhb, MtCo1, Uqcrc2, and Atp5a, which were selected as representative complex II, III, IV, and V subunits, were unaffected. These changes were consistently observed in both male and female mice and in BAT, scWAT, and vWAT. There were no differences between the Adipo-Ndufs4 KO and the wild-type (WT) mice in the gross or histological appearance of the adipose depots (Fig. 1E and 1F). Thus, unlike the Adipo-TFAM KO mice, the Adipo-Ndufs4 KO mice do not develop severe lipodystrophy (Fig. 1E and 1F) or a global reduction in the oxidative phosphorylation (OXPHOS) machinery. We next examined whether Ndufs4 deficiency causes alterations in oxidative stress by measuring 8-hydroxy-2-deoxyguanosine (8-OH-dG) levels in scWAT and BAT from female and male WT and Adipo-Ndufs4 KO mice. No significant differences in tissue oxidative damage were detected between WT and Adipo-Ndufs4 KO mice for either sex, although female scWAT showed lower 8-OH-dG levels than male scWAT (P = 0.03) (Fig. 1G).

Figure 1.

Adipocyte-specific deletion of Ndufs4 selectively reduces complex I proteins in adipose tissue. (a) Generation of the Adipo-Ndufs4 KO mice. (b) Confirmation of Ndufs4 deletion in adipose tissues by immunoblotting with an Ndufs4 antibody. 4-week-old female and male mice were used. (c, d) Representative mitochondrial complex subunits in BAT, scWAT, and vWAT from (c) 9-month-old female and (d) 11-month-old male mice. Only the complex I subunit is decreased in the KO vs WT littermates, while the complex II-V subunits are unchanged. (e) Gross appearance of BAT, scWAT, and vWAT from 6-week-old female and male mice. (f) Histological appearance (20×) after hematoxylin/eosin staining of BAT, scWAT, and vWAT from 6-week-old female and male mice. Scale bar = 100 μm. (G) Tissue oxidative stress assessed by measurement of 8-OH-dG levels in DNA extracted from BAT and scWAT of 21-week-old HFD-fed female and male mice. Bar graphs indicate mean ± SD. P value was obtained by Student’s t test.

Both Male and Female Adipo-Ndufs4 KO Mice Have Severe Cold Intolerance and Reduced Thermogenic Gene Expression

Mitochondria in brown adipocytes contain the inner mitochondrial membrane protein uncoupling protein 1 (UCP1), which promotes heat production by uncoupling the electron transport chain from ATP synthesis (2). The respiratory chain itself is thus critical for supporting adaptive thermogenesis in BAT. Based on these considerations, we examined if the Adipo-Ndufs4 KO mice have impaired cold adaptation. When acutely exposed to cold temperatures (7 ^oC), the Adipo-Ndufs4 KO mice were unable to maintain their core body temperature and became hypothermic over a period of several hours (Fig. 2A and 2B). Infrared thermal imaging confirmed lower surface temperatures in the KO mice, both males and females. These results confirm the presence of a functional defect with mitochondrial thermogenic respiration in Ndufs4 deficiency. In the KO BAT, we also observed a significant reduction in the expression of several key genes involved in thermogenesis, such as Ucp1, Cox5a, Cidea, and Dio2 (Fig. 2C and 2D), suggesting that loss of Ndufs4 can affect aspects of the thermogenic gene program in the nucleus. Similar patterns were noted in both male and female mice.

Figure 2.

Adipo-Ndufs4 KO mice have impaired cold adaptation. (a, b) Body temperature of (a) 8-week-old female (WT, n = 5, KO, n = 3) and (b) 8-week-old male mice (WT, n = 3; KO, n = 5) during cold exposure at 7 ^oC. At the end of the cold exposure, surface temperature images were taken using an infrared camera (FLIR). (c, d) Thermogenic gene expression in BAT from (c) 8-week-old female (WT, n = 5, KO, n = 3) and (d) 8-week-old male mice (WT, n = 3; KO, n = 3) following 4 hours of cold exposure at 7 ^oC. Bar graphs indicate mean ± SD. P value was obtained by Student’s t test (* P ≤ 0.05, ** P ≤ 0.01, *** P ≤ 0.001). The primer sequences are listed in Table 1.

Deletion of Adipocyte Ndufs4 Increases Susceptibility to High-Fat Diet-Induced Obesity and Defective Glucose Disposal Only in Male Mice

We next investigated if ablation of Ndufs4 affects the metabolic response to a high-fat diet (HFD). Female mice lacking adipocyte Ndufs4 showed no differences in body weight or glucose metabolism from WT control mice whether on regular diet or HFD (Fig. 3A and 3C-3E). Male Adipo-Ndufs4 KO mice did not differ from WT in body weight on a chow diet but showed an accelerated weight gain on a HFD (Fig. 3B). In addition, HFD-fed male Adipo-Ndufs4 KO mice had worse glucose tolerance and higher fasting plasma glucose and insulin levels (Fig. 3C-3E), demonstrating that defective mitochondrial function in adipocytes can indeed impair systemic glucose metabolism. A 2-way ANOVA confirmed sex-dependence of the effect of the Ndufs4 genotype on the fasting glucose and insulin levels (Tukey post hoc P < 0.005 and P < 0.01, respectively).

Figure 3.

HFD feeding reveals increased susceptibility to obesity and glucose intolerance in male Adipo-Ndufs4 KO mice, but not in female mice. (a, b) Body weight and food intake in (a) female (WT, n = 12; KO, n = 7) and (b) male mice (WT, n = 8; KO, n = 11) fed a standard chow or a high-fat diet. (c) Glucose tolerance test after 10 weeks of HFD feeding in female (WT, n = 7; KO, n = 7) and male mice (WT, n = 8; KO, n = 7), all at 18 weeks of age. (d) Serum glucose levels after a 6-hour fast in female (WT, n = 7; KO, n = 7) and male mice (WT, n = 8; KO, n = 7) after 11 weeks of a HFD. (e) Fasting serum insulin levels in female (WT, n = 9; KO, n = 7) and male mice (WT, n = 5; KO, n = 4) after 13 weeks of a HFD. (f) Body weight and food intake in weight-matched female mice (WT, n = 7; KO n = 7) fed a HFD. (g) Glucose tolerance test in weight-matched female mice (WT, n = 7; KO, n = 5) after 10 weeks of a HFD. (h) Tissue weights from HFD-fed weight-matched female (WT, n = 7; KO, n = 6) and male mice (WT, n = 7; KO, n = 6), after 11 weeks of a HFD. Inguinal, perigonadal, mesenteric, and retroperitoneal fat pads and BAT were pooled for combined adipose tissue weight, which is also shown as a percentage of total body weight. * P ≤ 0.05, ** P ≤ 0.01.

The female mice weighed substantially less than the male mice at the beginning of the HFD treatment, which may affect the body weight curve. Thus, to further evaluate the sex-dependence, we performed a HFD study using older female mice to match the weight of the males. The weight-matched female Adipo-Ndufs4 KO mice exhibited no difference in body weight or glucose tolerance from WT (Fig. 3F and 3G). When the weights of individual adipose depots and other organs were examined at the end of the HFD treatment, inguinal fat pad, BAT, and liver weights were higher in male KO mice compared with WT mice, but female mice showed no such differences (Fig. 3H). No weight differences between WT and KO were seen with the heart and gastrocnemius muscles from animals of either sex. Relative to the total body weight, the percentage of combined adipose tissues, which included inguinal, perigonadal, mesenteric, retroperitoneal, and brown fat pads (23), was increased in male KO mice, indicating greater adiposity in these animals (Fig. 3I). The livers in male KO mice were larger in size than WT livers. We measured the expression of selected genes related to inflammation or lipid metabolism by qPCR and did not observe significant differences (Fig. 4A and 4B). Upon microscopic examination, livers from HFD-fed male KO mice exhibited increased steatosis compared with those from WT mice (Fig. 4C).

Figure 4.

Livers from male Adipo-Ndufs4 KO mice accumulate more lipid compared to WT mice on a HFD. (a, b) Expression of inflammation and lipid metabolism-related genes in (a) female (WT, n = 5; KO, n = 4) and (b) male (WT, n = 5; KO, n = 4) livers following 11 weeks of a HFD. Bar graphs indicate mean ± SD. P value was obtained by Student’s t test. (C) Histological appearance (20×) after hematoxylin/eosin staining of liver tissues from weight-matched female mice and male mice following 11 weeks of a HFD. Scale bar = 100 μm.

Subacute chronic inflammation in adipose tissue is viewed as an important mechanism of insulin resistance and increased levels of adipose tissue macrophages have been observed in pathological states associated with insulin resistance (24). The mitochondrial electron transport chain has been implicated in the control of innate immunity (25) and global Ndufs4 loss has been reported to increase systemic inflammation (26). With adipocyte-specific Ndufs4 loss, we found that pro-inflammatory markers such as Tgfb1, Tnfa, and Il1b are increased in male scWAT but not in female scWAT (Fig. 5A and 5B), again pointing to sex differences in this context. A 2-way ANOVA demonstrated that the interaction between sex and genotype on the expression of Tgfb1, Tnfa, and I11b reached statistical significance (Tukey post hoc P < 0.05 for the 3 genes). However, no such differences were found in vWAT from perigonadal fat pad.

Figure 5.

HFD-fed male Adipo-Ndufs4 KO mice, but not female KO mice, show elevated inflammatory genes in scWAT compared to littermate controls. (a, b) Relative expression of inflammation-related genes in (a) vWAT and (b) scWAT from female Adipo-Ndufs4 KO mice and WT littermates after 13 weeks of a HFD. n > 3 for each group. Bar graphs indicate mean ± SD. P value was obtained by Student’s t test. The primer sequences are listed in Table 1. (c) Multidimensional scaling (MDS) plot based on global transcriptome analysis of scWAT from 21-week-old HFD-fed male and female WT and Adipo-Ndufs4 KO mice. (d) Dot plot of Gene Ontology (GO) terms enriched among the genes showing a greater KO vs WT differential upregulation in male scWAT than in female scWAT. Based on DAVID bioinformatics resources 6.8, the top 23 GO terms with the fold enrichment score (number of genes associated with a specific GO term among the genes of interest vs a background population of genes) greater than 10 are plotted in order of the gene ratio (percentage of genes linked to a specific GO term among the genes of interest). The color of each dot indicates the FDR-adjusted P value and the size of the dot represents the fold enrichment score for the specific GO term.

To assess global alterations in the transcriptome, we performed RNA sequencing studies using scWAT from male and female WT and Adipo-Ndufs4 KO mice. As shown in the multidimensional scaling (MDS) plot in Fig. 5C, the gene expression profiles of WT and Adipo-Ndufs4 KO scWAT showed clear separation in male mice, but not in female mice, indicating that male scWAT had more differentially regulated genes between WT and KO than female scWAT. We identified 618 genes that showed a higher KO to WT signal ratios in male scWAT than in female scWAT, and 177 genes that had a lower KO to WT signal ratios in males than females. When classified based on gene ontology (GO), inflammation-related genes, such as those involved in lymphocyte activation and cytokine secretion, were prominent in the list of the genes that showed a greater KO vs WT differential upregulation in males than in females (Fig. 5D). This result is consistent with the qPCR data in Fig. 5B.

The hormone fibroblast growth factor-21 (FGF21) has been shown to enhance systemic glucose disposal and has attracted much interest as a possible therapeutic agent (27). Significantly, FGF21 has also been reported to be a key component of the stress response induced by mitochondrial genetic defects and proposed as a biomarker of mitochondrial disease (28, 29). Recent data indicate that FGF21 is regulated differently by sex (30). When age-matched male and female mice were examined after 13 weeks of a HFD, no significant differences in circulating serum FGF21 levels were seen between WT and Adipo-Ndufs4 KO mice, either in males or in females (Fig. 6A). On the other hand, the expression levels of FGF21 and its associated signaling proteins such as the FGF21 receptor (Fgf21r) and the cofactor bKlotho (Klb) were decreased locally in scWAT from male Adipo-Ndufs4 KO compared with WT controls, while no such difference between KO and WT was detected in female mice (Fig. 6B). Sex-dependent effects of the genotype on Fgf21r expression were confirmed by a 2-way ANOVA (Tukey post hoc, P < 0.05). In BAT, male KO mice had lower levels of Fgf21r and Klb compared with WT, whereas female KO mice showed no changes in these genes and an increase in the FGF21 mRNA levels (Fig. 6C). A sex-genotype interaction on Klb expression was found (Tukey post hoc, P < 0.05). vWAT showed no differences between KO and WT, whether male or female (Fig. 6D). These findings raise the possibility that reduced local FGF21 signaling in male KO scWAT and BAT exacerbates insulin resistance on a HFD and helps to explain the sexual dimorphism in the metabolic phenotypes consequent to a monogenic respiratory chain deficiency.

Figure 6.

Differential regulation of FGF21 signaling in male and female Adipo-Ndufs4 KO mice. (a) Serum FGF21 levels in 21-week-old male and female mice after 13 weeks of a HFD. n > 7 for each group. (b, c, d) Fgf21, Fgfr1, and Klb mRNA levels in (b) scWAT, (c) BAT, and (d) vWAT from 21-week-old male and female mice after 13 weeks of a HFD. n > 3 for each group. Bar graphs indicate mean ± SD. P value was obtained by Student’s t test (* P ≤ 0.05, ** P ≤ 0.01). The primer sequences are listed in Table 1.

Discussion

Abnormal mitochondrial amount or function has been implicated in the pathogenesis of various metabolic and degenerative disorders as well as aging itself (31, 32). With regards to type 2 diabetes and insulin resistance, defects in skeletal muscle mitochondria have been extensively investigated using both in vivo and ex vivo methodologies (33-37). However, there is no consensus on whether mitochondrial dysfunction is a cause or the consequence of insulin resistance, and some have even questioned whether muscle mitochondrial dysfunction is an essential feature of type 2 diabetes (38, 39). Adipose tissue mitochondria have been less extensively studied, in part because white adipocytes contain relatively fewer and smaller mitochondria in comparison with muscle cells (40). Nevertheless, mitochondria are critically involved in many important adipocyte functions such as lipolysis and the re-esterification of fatty acids to triglycerides (41, 42) as well as thermogenesis. Reduced white adipose mitochondrial content or function have been reported in multiple studies of type 2 diabetes patients and animal models (3-7). As with skeletal muscle mitochondria, it has not been resolved whether mitochondrial abnormalities in adipose tissue contribute to impaired glucose homeostasis or they merely result from diabetes. Our data presented here suggest that deletion of a single complex I subunit in adipose tissue can bring about tissue inflammation and systemic glucose intolerance, and young males are more sensitive to this outcome than females.

Targeting Ndufs4 specifically in adipose tissue allowed us to genetically define the effect of adipocyte mitochondrial dysfunction on systemic physiology. Unlike with the adipocyte-specific TFAM null mice (9), the metabolic abnormalities in our model are not secondary to lipodystrophy. Male Adipo-Ndufs4 KO mice developed increased adiposity on a HFD compared to littermate controls, as evidenced by greater fat tissue weights relative to the total body weight. The livers from these animals showed increased lipid accumulation, which may reflect caloric excess from defective diet-induced thermogenesis and energy expenditure concomitant with a limited capacity to take up and store energy in adipose tissues. Lastly, diet-induced obese Adipo-Ndufs4 KO mice were relatively glucose intolerant, and this manifested in a sex-dependent manner.

The adverse metabolic consequences of adipocyte Ndufs4 deficiency were accompanied by elevated levels of inflammatory marker genes in scWAT. A transcriptome analysis of scWAT revealed widespread increases in inflammation-related gene expression that were more apparent among males. Ablation of Ndufs4 in global KO mice has previously been linked to activation of systemic inflammation (26), attributed to loss of Ndufs4 activity in immune cells such as macrophages triggering innate immunity via toll-like receptor 4/2 (TLR4/2) signaling. Our data indicate that a selective loss of Ndufs4 activity in adipocytes is sufficient to enhance local tissue inflammation under certain circumstances. While adipocytes are known to interact closely with immune cells, they are not bona fide immune cells themselves and do not express TLR4/2 at significant levels. It appears likely that paracrine and endocrine mechanisms are key mediators of the local and systemic effects associated with adipocyte Ndufs4 deficiency.

In many rodent models, the risk of developing diet-induced diabetes is greater in males than females (43). Besides this sex-dependent relationship between diet and glucose homeostasis, our data also demonstrate sex-dependent effects of Ndufs4 deficiency on systemic glucose metabolism. Because sexually mature male mice weigh more than female mice and body weight affects the predisposition to developing insulin resistance, we also compared older, weight-matched females with males in the context of Ndufs4 deficiency and still saw disparities based on the sex. Previous studies have linked sexual dimorphism in metabolic disease to factors such as higher levels of estrogens and better preservation of pancreatic β-cell function (43, 44) but these are not directly targeted by adipocyte Ndufs4. It has also reported that mitochondria from female rodents have lower reactive oxygen species due to better antioxidant defenses (45). We found that female scWAT had lower levels of tissue oxidative damage than male scWAT but did not detect any differences between WT and KO for either males or females. As these data did not support a central role for redox signaling pathways in producing the Ndufs4 deficiency phenotype, we considered other pathways triggered by mitochondrial stress.

FGF21 is a key downstream target of the mitochondrial stress signaling pathway and is activated in response to mitochondrial genetic defects (28, 29). Several of the pleiotropic effects of FGF21 on systemic metabolism are mediated by direct actions on neurons and adipocytes (46, 47). Recent data further support that FGF21’s role as an autocrine/paracrine factor in adipose tissue contributes to some key effects such as fat browning (48). In view of these facts, we considered the possibility that sex-dependent regulation of local, paracrine FGF21 signaling by adipose tissue Ndufs4 deficiency can help explain the observed sex differences in the metabolic phenotypes. The acute stimulatory effects of FGF21 on insulin sensitivity require FGF signaling in adipose tissues, especially BAT, since ablating Klb in mice using adiponectin-Cre or UCP1-Cre diminishes these effects (49). BAT can account for a substantial percentage of total glucose clearance and thereby act as a “glucose sink” (50-52), and it is possible that the sex differences in FGF21 signaling in Adipo-Ndufs4 KO BAT give rise to the differences in systemic glucose homeostasis. Other studies suggest that scWAT is a more important target of FGF21 activity than vWAT (53). Given that FGF21 signaling has been implicated in ameliorating inflammation (54, 55), blunted FGF21 action in male Adipo-Ndufs4 KO scWAT may play a role in increased local inflammation and predisposition to insulin resistance. Thus, differences in FGF21 signaling in various adipose depots can partly account for the distinct sex-dependent metabolic phenotypes in Ndufs4 deficiency (Fig. 7). However, why FGF21 signaling is differentially regulated in male and female models of Ndufs4 deficiency in the first place remains to be elucidated. FGF21 has been reported to enhance mitochondrial function in neurons and adipocytes (56, 57) and is an important component of the stress responses induced by mitochondrial dysfunction in skeletal muscle (28, 29). Sex differences in mitochondrial signaling and function are likely ultimately responsible for the differential regulation of FGF21 action in male and female Adipo-Ndufs4 KO mice (58).

Figure 7.

Schematic diagram of the sex-dependent effects of adipocyte Ndufs4 deficiency on systemic glucose homeostasis. Created with BioRender.com.

In summary, our studies show that genetic ablation of a mitochondrial complex I subunit in adipocytes causes increased local inflammation and impairs whole-body glucose metabolism, supporting a causal role for mitochondrial dysfunction in the etiology of type 2 diabetes. Moreover, we find that males are more susceptible to these effects. One limitation of these studies is the duration of the HFD experiment, which may not have been sufficiently long to bring out differences between WT and Adipo-Ndufs4 KO mice in females. While we have suggested FGF21 as a possible contributor to the sex-dependent metabolic phenotypes, more definitive evidence would involve a functional rescue experiment. Because the circulating FGF21 levels do not differ between WT and Adipo-Ndufs4 KO mice, other adipocyte-produced signaling molecules that target non-adipose tissues are likely at play as well. Further studies are needed to clarify the exact mechanisms that underlie the sexual dimorphism in systemic glucose metabolism associated with Ndufs4 deficiency.

Glossary

Abbreviations

8-OH-dG

8-hydroxy-2-deoxyguanosine

ANOVA

analysis of variance

BAT

brown adipose tissue

ELISA

enzyme-linked immunosorbent assay

FGF21

fibroblast growth factor 21

GO

gene ontology

HFD

high-fat diet

KO

knockout

Ndufs4

nuclear-encoded accessory subunit of mitochondrial respiratory chain NADH dehydrogenase complex I (18-KD subunit)

qPCR

quantitative polymerase chain reaction

ROS

reactive oxygen species

scWAT

subcutaneous white adipose tissue

TFAM

mitochondrial transcription factor A

UCP1

uncoupling protein 1

vWAT

visceral white adipose tissue

WT

wild-type

Financial Support

This work was supported by grants from the Cystic Fibrosis Foundation to J.C.Y. and the National Institute of Diabetes and Digestive and Kidney Diseases Award No. R01DK121035 to K.K.R.

Disclosures

The authors have nothing to disclose.

Data Availability

Some or all data generated or analyzed during this study are included in this published article or in the data repositories listed in References.