mtDNA Haplogroup N9a Increases the Risk of Type 2 Diabetes by Altering Mitochondrial Function and Intracellular Mitochondrial Signals

Millions of people worldwide live with diabetes, and >90% of these people have been diagnosed with type 2 diabetes (T2D) (1). Although the molecular mechanisms underlying T2D remain incompletely elucidated, deregulation of mitochondrial oxidative phosphorylation (OXPHOS) is widely accepted as one of the major causes of T2D and insulin resistance (2). Diminished OXPHOS function might contribute generally to insulin resistance through elevated generation of reactive oxygen species (ROS) production, a major regulatory signal in T2D-related insulin receptor signaling and inflammation (3,4).

The OXPHOS pathway comprises five complexes, of which four complexes are dually regulated by nuclear DNA (nDNA) and mitochondrial DNA (mtDNA); thus, as expected, variants in both nuclear and mitochondrial genomes have been associated with T2D (5,6). An mtDNA haplogroup is a specific mtDNA genetic background defined by variants in human mtDNA (i.e., single nucleotide polymorphisms [SNPs]) that are inherited during long-term evolution. Initial evidence indicated that mtDNA haplogroups influence cellular respiration and ROS production, which implied the importance of mtDNA haplogroups in the regulation of mitochondrial function (7). Subsequently, mtDNA haplogroups were shown to play a pathophysiological role in rats with T2D (8) and regulate physical performance in mice (9). Shortly thereafter, diagnostic SNPs of two human macro haplogroups, M and N, were shown to alter mitochondrial matrix pH and intracellular calcium dynamics (10), and mtDNA haplogroups have thus far been reported to regulate mtDNA replication and transcriptional efficiency (11), the activity of respiratory chain complex (RCC) I, and the assembly dynamics of multiple RCCs (12,13). Recently, to comprehensively elucidate the mechanisms underlying the roles of mtDNA haplogroups in diseases such as T2D, mitochondrial retrograde signaling has been frequently analyzed using microarray or RNA sequencing technologies (9,14). This approach has yielded clues regarding disease-causing factors such as alterations in gene methylation status and shifts in metabolic pathways in diseases like T2D (14,15).

Mitochondrial haplogroups play a critical role in both mitochondrial function and mitochondria-mediated signaling pathways; accordingly, mtDNA haplogroups have been suggested to be involved in a series of metabolic diseases such as metabolic syndrome, obesity, T2D, and T2D-associated complications in distinct populations (16–19). However, certain contradictory observations remain unresolved, particularly in studies related to T2D. As in the case of studies reporting varying disease phenotypes associated with disorders such as Alzheimer disease and Parkinson disease, several pitfalls in the T2D studies might account for a few of the contradictions (20,21). However, differences in the nuclear genetic background might be responsible for most of the discrepancies, which have been found in populations from both Asia (16–19) and Europe (16,22–24). One potential underlying reason is that the nuclear genetic background might contribute to the difference in the functional performance of mitochondria, which has been referred to in a recent study as mitochondrial–nuclear coevolution (24). Another reason is that the environment can also contribute to divergent responses of the same mtDNA in human diseases (25). Nevertheless, how mtDNA haplogroups affect T2D is currently unknown. For example, NDUFC2 has been recognized to influence the disease susceptibility of mtDNA haplogroup HV in T2D, but whether and how haplogroup HV itself affects the susceptibility remains unresolved (23,24).

In this study, we conducted a large-scale case-controlled study to validate the effect of mtDNA haplogroup N9a on the pathogenesis of T2D in the Han Chinese population. Because nuclear gene expression is unfailingly altered when mitochondrial function is affected in haplogroup N9a, we analyzed the mitochondrial retrograde signaling in two N9a cybrids and three cybrids other than N9a. Last, we tested the effect of alterations in this retrograde signaling on diabetes by using a cellular model.

In this study, 1,295 unrelated patients (mean ± SD age 60.34 ± 12.839 years; median 60, range 17–93) with T2D were recruited at the The First Affiliated Hospital of Wenzhou Medical University (Zhejiang, China) from March 2009 to December 2017. T2D was diagnosed according to the China Medical Nutrition Therapy Guideline For Diabetes (26). A total of 974 geographically matched and sex-matched control participants (mean ± SD age 53.76 ± 15.919 years; median 54, range 17–89) with no history of T2D were also recruited at the same hospital (at its physical examination center). The participants without diabetes were people in whom the fasting plasma glucose concentration was <6.1 mmol/L and blood glycosylated hemoglobin (HbA_1c) level was <6.2% (44 mmol/mol). We have described 675 patients and 649 control subjects used here in a previous study (27). Informed consent was obtained from all participants under protocols approved by the ethics committee of Wenzhou Medical University. All experimental methods were performed in accordance with approved guidelines of Wenzhou Medical University.

Genomic DNA from peripheral blood was extracted using a standard SDS lysis protocol. Complete mtDNA sequence was Sanger sequenced for 347 T2D patients and 383 control subjects by using 24 previously reported pairs of mtDNA primers (28). The mtDNA of 235 T2D patients and 141 control subjects was completely sequenced previously (27); for all other study participants, Sanger sequencing was performed using two pairs of mtDNA primers (Supplementary Table 1). SNPs of each participant were identified by comparing the obtained sequences with the revised Cambridge Reference Sequence by using CodonCode Aligner 3.0.1 (CodonCode Corporation, Centerville, VA). We used the HaploGrep program (http://haplogrep.uibk.ac.at/) to annotate the mtDNA haplogroup for the cases where the mtDNA was completely sequenced. For all other study participants, mtDNA haplogroup was assigned by comparing the target SNPs from the D-loop, ND3, and ND4L with the diagnostic SNPs of the most up-to-date Chinese mtDNA haplogroup tree (29).

Two N9a haplogroups (N9a1 and N9a10a) were used to exclude the effect of private SNPs (such as mt.13214) in the terminal clades of the mtDNA tree (Supplementary Fig. 1). As control haplogroups, we included haplogroups G (G3 in this study) and D4 (D4j in this study), both of which were not positively associated with metabolic diseases in previous work and were evenly distributed in this study among T2D patients (haplogroup G, 2.5%; haplogroup D4, 10.9%) and control subjects (haplogroup G, 2.2%; haplogroup D4, 11.4%). Moreover, haplogroup Y (Y1 in this study) (0.8% in T2D patients; 0.6% in control subjects), which forms a neighboring clade of haplogroup N9a, was included as additional control haplogroup to exclude the potential phenotypic effects produced by haplogroup N9-defining SNPs such as mt.5417 (Supplementary Fig. 1). These three control haplogroups, G3, D4j, and Y1, are referred to as non-N9a haplogroups in this study.

By using the standard protocol (30), transmitochondrial cybrids were generated through the fusion of mtDNA-lacking ρ0 human osteosarcoma 143B cells with platelets of haplogroup N9a1, N9a10a, G3a2, Y1, or D4j obtained from five volunteers. Platelets containing blood was collected from the volunteers when they were 22 years old during physical examination before their enrollment in the graduate school of Wenzhou Medical University. The transformant cybrid clones were cultured in high-glucose DMEM (Thermo Fisher Scientific, Waltham, MA) containing 10% Cosmic Calf Serum (Sigma, St. Louis, MO). Pathogenic mtDNA mutations and cross contamination during single-clone selection were ruled out through Sanger sequencing of the mitochondrial genome in both the five volunteers and the cybrid cells during culture (Supplementary Table 1).

Both mtDNA content and the mRNA levels of 13 mitochondrially encoded OXPHOS subunits were determined using the 2^(–ΔΔCT) method as previously described (31). Briefly, genomic DNA and total RNA were extracted using standard protocols, and the total RNA was then treated with DNase and reverse-transcribed using random 6-mers primers (Takara Biotechnology, Dalian, China). Quantitative real-time PCR was performed using primers targeted to mtDNA, mtRNA, a subset of inflammatory genes, and related nuclear housekeeping genes on a StepOne Real-Time PCR System (Thermo Fisher Scientific) by using SYBR Green qPCR Master Mix (Takara Biotechnology). All primers used in these analyses are listed in Supplementary Table 2.

Mitochondria from cultured cells were isolated as previously described (30). The enzymatic activity of four RCCs was measured in the mitochondria of cybrids as described (30). The RCC enzymatic activity in each case was normalized against that of citrate synthase, a mitochondrial matrix marker enzyme.

Proteins were extracted using RIPA lysis buffer (Cell Signaling Technology, Danvers, MA) supplemented with a protease-inhibitor cocktail (Sigma-Aldrich). Proteins separated using SDS-PAGE were blotted with these antibodies: anti-VDAC, anti-JNK 1/2, anti–phospho-JNK 1/2, anti-p38, anti–phospho-p38 (Thr389), anti-ERK1/2, anti–phospho-ERK (Thr202/Tyr204), anti–NF-κB, anti–phospho-NF-κB, anti-SRC, anti–phospho-SRC, anti-MEK1/2, anti–phospho-MEK1/2, anti-AMPK, and anti–phospho-AMPK (all from Cell Signaling Technology; 1:1,000); anti-TOMM20, anti-SDHA, anti-RXRA, anti–POLY-γ, anti-TFAM, anti-NRF1, anti-AFG3L2, anti-ClpP, anti-ClpX, anti-HSP60, anti-PINK1, anti-DRP1, anti–phospho-DRP1, anti-OPA1, anti-MFN1, and anti-MFN2 (all from Abcam, Cambridge, MA; 1:1,000); and anti–β-actin (1:5,000), anti-SOD2 (1:1,000), anti-TFAM (1:1,000), and anti-GRP75 (1:2,000) (all from Santa Cruz Biotechnology, Santa Cruz, CA).

Endogenous oxygen consumption in intact cells was determined using a Seahorse XF24 Extracellular Flux Analyzer (Seahorse Bioscience, North Billerica, MA) as described in our previous study (2). Briefly, 4 × 10⁴ cells were seeded into 24-well Seahorse plates together with 250 µL of growth medium 1 day before experiments. Oxygen consumption rate (OCR) was determined with and without the inclusion of 1 μmol/L oligomycin. Extracellular acidification rate (ECAR) was determined by sequentially injecting 10 mmol/L glucose, 1 μmol/L oligomycin, and 50 mmol/L 2-deoxy-d-glucose by using the Seahorse system.

Mitochondrial membrane potential (MMP), total ATP content, and NAD⁺/NADH ratio were determined using the cationic fluorescent redistribution dye tetramethylrhodamine, methyl ester (Thermo Fisher Scientific), an ATP measurement kit (Thermo Fisher Scientific), and an NAD⁺/NADH ratio assay kit (Abcam), respectively (32). Mitochondrial and cytoplasmic ROS production was measured using MitoSOX and carboxy-DCFDA (both from Thermo Fisher Scientific), respectively. Briefly, cells in 12-well plates were treated with MitoSOX (5 μmol/L) or carboxy-DCFDA (40 µg/mL) for <1 h at 37°C in the dark and then washed with HBSS and analyzed immediately using a fluorescence microscope (Eclipse Ti-E, Nikon Eclipse Ti-S; Nikon Instruments, Inc., Tokyo, Japan). At least five regions were quantitatively analyzed for each cybrid to generate the average fluorescence intensity in one independent experiment by using ImageJ (Bethesda, MD).

Cells were incubated with 500 nmol/L MitoTracker Red (Thermo Fisher Scientific) for 30 min and fixed for 15 min with 4% paraformaldehyde at room temperature. The cells were then permeabilized with 0.2% Triton X-100 (Sigma), stained with DAPI (Thermo Fisher Scientific), and examined using an Olympus imaging system (Olympus FV1000; Olympus, Melville, NY). Mitochondrial length and complexity were quantified by measuring the form factor and aspect ratio, respectively.

Total RNA was isolated from three biological triplicates of each group of cybrids by using an RNeasy Mini extraction kit (Qiagen, Valencia, CA), and mRNA from 20 μg of the total RNA was purified using poly-T–attached magnetic beads. After fragmenting the mRNA, first-strand cDNA was synthesized and then sequenced using an Illumina HiSeq 2000 platform (Illumina, San Diego, CA) as described previously (32).

To obtain high-quality reads, reads containing adaptor sequences and poly-N and low-quality reads were removed from the raw data. Reference-genome and gene-model annotation files were downloaded from genome websites directly. The reference genome was built using STAR, and paired-end high-quality reads were aligned to the reference genome by using STAR (v2.5.1b). HTSeq v0.6.0 was used to count the reads mapped to each gene, after which the fragments per kilobase million of each gene was calculated based on the length of the gene and the count of the reads mapped to the gene. Differential expression analysis under two conditions (three biological replicates per condition) was performed using the DESeq2 R package. DESeq2 provides statistical routines for determining differential expression in digital gene expression data by using a model based on negative binomial distribution. The resulting P values were adjusted using the Benjamini and Hochberg approach for controlling the false discovery rate (33). Genes found using DESeq2 that feature an adjusted P < 0.05 were regarded as differentially expressed genes (DEGs). To identify the genes that differed between the two N9a and three non-N9a cybrids, we respectively compared N9a1 and N9a10a with non-N9a cybrids (G3a2, Y1, and D4j), and the genes that overlapped in both comparisons were confirmed as the final DEGs. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway and gene ontology (GO) biological performance enrichment analyses of DEGs were implemented using the clusterProfiler R package. GO and KEGG pathway terms featuring P values of <0.05 were regarded as significantly enriched among DEGs.

Before measurements, cells were serum starved for 5 h, incubated with 0.1 mmol/L insulin plus 2-deoxy-2-[(7-nitro-2,1,3-benzoxadiazol-4-yl)amino]-d-glucose (2-NBDG) (0.1 μmol/L) in a CO₂ incubator for 30 min, and then washed thrice with cold phosphate-buffered saline. The fluorescence signal was measured after 30 min with excitation and emission at 485 and 535 nm, by using a Varioskan Flash Multimode Reader (Thermo Fisher Scientific). Insulin-stimulated glucose uptake was also measured in with the presence of N-acetyl-l-cysteine (NAC; Sigma) (15 mmol/L, 24 h), nicotinamide (NAM; Sigma) (15 mmol/L, 24 h), U0126 (Selleckchem, Houston, TX) (50 μmol/L, 6 h), C34 (Sigma) (10 μmol/L, 30 min), or TAK-242 (MedChemExpress, Monmouth Junction, NJ) (10 μmol/L, 1 h).

In this case-controlled study, haplogroups featuring a frequency of >5% in either the control participants or T2D patients were analyzed to evaluate the effect of common mtDNA haplogroups on T2D. Haplogroups that featured frequencies of <5% were regarded as “other haplogroups.” Multivariate logistic regression analysis was applied to adjust the risks associated with age, sex, and mtDNA haplogroups in case and control subjects. T2D was a dependent variable, whereas age, sex, and genotype of each mtDNA haplogroup were independent variables. In the case of nine mtDNA haplogroups, dummy coding was applied due to their numerical variability. Bonferroni correction indicating a P value of <0.006 (0.05/9) was considered statistically significant when analyzing these nine haplogroups (with the other haplogroups excluded). The significance of “other haplogroups” was not considered here because the group included multiple mtDNA haplogroups. An independent Student t test was used to evaluate cybrid data, and a null hypothesis was rejected when P < 0.05. All statistical analyses were performed using SPSS 21.0 (IBM, Armonk, NY).

To investigate the relationship between mtDNA haplogroup N9a and T2D in the Chinese population, we performed a large-cohort case-controlled study that included 1,295 T2D patients and 974 geographically matched asymptomatic control participants. The frequency of haplogroup N9a was found to be significantly higher in patients than in control subjects when multivariate logistic regression analysis was performed with adjustment for age, sex, and haplogroups (odds ratio [OR] 1.999 [95% CI 1.229–3.251], P = 0.005) (Table 1). By contrast, a significantly decreased frequency of haplogroup N9a was previously observed in T2D patients from Korea and Japan (Table 1). This finding suggests that distinct populations could present divergent responses in terms of the effect of haplogroup N9a on T2D occurrence. Furthermore, we found that the frequency of haplogroup N9, which comprised haplogroups Y and N9a, was also significantly higher in T2D patients than in control participants (OR 1.967 [95% CI 1.238–3.124], P = 0.004), but the statistical significance was not retained in the case of ethnic Chinese people from Taiwan (Table 1). The distribution of haplogroup Y was similar between T2D patients (0.8%) and control subjects (0.6%).

To analyze the effect of mtDNA haplogroups on the regulation of mitochondrial function, we determined the mtDNA content and the RNA level of mtDNA-encoded OXPHOS subunits in two N9a and three non-N9a cybrids. The mtDNA content in non-N9a cybrids was ∼30% higher than that in N9a cybrids (Fig. 1A). The lower mtDNA content measured in N9a cybrids was not because of the presence of fewer mitochondria, and the mitochondrial mass was roughly the same in the N9a and non-N9a cybrids (Fig. 1B and E). Next, examination of the RNA level of mtDNA-encoded OXPHOS subunits revealed that the RNA levels of ATP8, ND1, ND5, and CO1 were higher in all non-N9a cybrids than in N9a cybrids (Fig. 1C). This result suggested superior mitochondrial function in non-N9a cybrids than in N9a cybrids (34), and to test this possibility, we examined the activity of three RCCs containing mtDNA-encoded subunits. After normalization of the RCC activities relative to citrate synthase activity, we found that the activities of complexes I and IV were significantly higher in non-N9a cybrids than in N9a cybrids, whereas the activity of complex III did not differ (Fig. 1D and E). Although the activity of certain RCCs was diminished in N9a cybrids, transcriptomic analysis performed using RNA sequencing technology revealed that N9a cybrids exhibited increased mRNA levels of most nDNA-encoded OXPHOS subunits as compared with non-N9a cybrids (Fig. 1F), and, notably, the mRNA levels of complex II subunits were not significantly affected, particularly those of the subunits SDHA and SDHB (Fig. 1F). The observed pattern of mtDNA-encoded OXPHOS subunits was not reliable in the experiment performed here using RNA sequencing, as the length of poly-A tails varied among distinct mitochondrial genes during RNA capture. However, the ratio of nDNA-encoded subunits to mtDNA-encoded subunits was found to be higher in N9a cybrids than in non-N9a cybrids when we compared the nDNA-encoded gene expression determined from the transcriptome analysis with the mtDNA-encoded gene expression measured using quantitative real-time PCR analysis (Fig. 1C and F). Furthermore, the level of NRF1, an essential transcription factor for nuclear genes required for respiration, was higher in N9a cybrids than in non-N9a cybrids (Fig. 1G), which suggests that a retrograde signaling machinery might be involved in compensatory protection of mitochondrial function in N9a cybrids (35). Together with this mitonuclear imbalance of OXPHOS subunits, the detection of a higher level of the mitochondrial quality-control protein ClpP in N9a cybrids than in non-N9a cybrids supported the notion that N9a cybrids exhibit increased mitochondrial unfolded protein response (mtUPR) as compared with non-N9a cybrids (Fig. 1H). However, we found that other mtUPR proteins did not differ between the N9a and non-N9a cybrids (Fig. 1H), and thus the mtUPR level in N9a cybrids might be limited.

Next, we measured mitochondrial respiratory profiles by using a Seahorse XF24 Extracellular Flux Analyzer. A respiration assay of the cybrids revealed that both intracellular respiration and proton leakage were significantly higher in non-N9a cells than in N9a cells (Fig. 2A). The measured ratio of coupled-to-uncoupled respiration indicated that the coupling efficiency did not differ significantly between N9a and non-N9a cells (Fig. 2B), and an increased ECAR and OCR/ECAR ratio confirmed that glycolytic function was lower but mitochondrial function was higher in non-N9a cells than in N9a cells (Fig. 2C and D), which suggested distinct mitochondrial retrograde signaling pathways between N9a and non-N9a cells. Accordingly, the mean values of MMP and total ATP content in non-N9a cells were ∼50% higher than those measured for N9a cells (Fig. 2E and F). Notably, total ATP content and mtDNA content in peripheral blood mononuclear cells (PBMCs) from non-N9a haplogroup study participants were significantly higher than those in PBMCs from N9a haplogroup participants (Fig. 2G and H), which suggested that mitochondrial function was distinctly affected by N9a and non-N9a haplogroups. Furthermore, we analyzed mitochondrial fragmentation by measuring the form factor (an index of mitochondrial branching) and the aspect ratio (an index of mitochondrial branch length) of single mitochondrion in the case of these haplogroups (Fig. 2I); a low degree of mitochondrial fragmentation is indicated by high values of aspect ratio and form factor, parameters that represent increased mitochondrial length/width and branching, respectively. Our results revealed a lower percentage of mitochondrial fragmentation in non-N9a cells than in N9a cells (Fig. 2J and K). Last, we examined mitochondrial fission/fusion proteins and found that the level of the long form of OPA1, which is considered to promote mitochondrial fusion, was higher in non-N9a cells than in N9a cells (Fig. 2L). Collectively, our results demonstrated that N9a cybrids exhibit diminished mitochondrial function relative to non-N9a cybrids.

Fine-tuning of mitochondrial function can activate diverse retrograde signaling pathways in the nucleus by affecting the levels of mitochondrial signaling mediators (33). Because redox signaling pathways play a major role in cellular physiology, we measured the mitochondrial redox signal, the ROS level, and the NAD⁺/NADH ratio in N9a and non-N9a cybrids. N9a cybrids generated more ROS than non-N9a cybrids did, as determined using either the cytosolic ROS probe carboxy-DCFDA (Fig. 3A and B) or the mitochondrial ROS probe MitoSOX (Fig. 3C and D), but mitochondrial antioxidant activity did not differ between the cybrids, as revealed by measurement of the antioxidant protein SOD2 (Fig. 1B). Furthermore, the NAD⁺/NADH ratio was lower in N9a cybrids than in non-N9a cybrids (Fig. 3E). These results suggested that N9a and non-N9a cybrids feature distinct mitochondrial retrograde signaling profiles.

To uncover the differences between the transcriptome of N9a and non-N9a cybrids, we used high-throughput RNA sequencing for transcriptomic profiling of the two N9a cybrids and three non-N9a cybrids. Our analysis revealed 826 statistically significant DEGs between the two N9a cybrids and three non-N9a cybrids, of which 604 and 222 genes were upregulated and downregulated, respectively, in both N9a cybrids as compared with the levels in three non-N9a cybrids; moreover, among these DEGs, 52 genes encoded transcription factors and were related to signaling pathways such as the ERK1/2 pathway (Fig. 4A and Supplementary Data Set 1) (full access to the data set is available upon request to the author). We next performed both GO and KEGG pathway enrichment analyses to determine the contribution of these DEGs in biological performance. As shown in Fig. 4B and Supplementary Data Set 2 (full access to the data set is available upon request to the author), multiple mitochondrial OXPHOS-related pathways and three signal transduction pathways (Wnt, ERK1/2, and p38-MAPK) were presented in all 79 GO biological processes that showed significantly different enrichment between the two N9a and three non-N9a cybrids. KEGG enrichment analysis further revealed that the two N9a and three non-N9a cybrids exhibit distinct performance in terms of metabolic regulation and pathways related to metabolic diseases such as nonalcoholic fatty liver disease (Fig. 4C). The statistical significance of most pathways was retained when we excluded the DEGs encoding OXPHOS subunits (Fig. 4D and E and Supplementary Data Set 2), which indicated that mitochondrial retrograde signaling contributed substantially to the difference in biological performance between the two N9a and three non-N9a cybrids. Notably, although only 90 DEGs remained after we applied the criterion of fold-change >2 (Fig. 4F), the difference in biological performance between N9a and non-N9a cybrids was retained in several aspects (Fig. 4G and H).

We sought to functionally assess how and through which pathway mitochondrial retrograde signaling influences T2D susceptibility in N9a and non-N9a cybrids. Including the candidate ERK1/2 pathway, we tested seven pathways that are commonly associated with mitochondrial retrograde signaling pathway (33). As expected, ERK1/2 phosphorylation levels differed between the two N9a and three non-N9a cybrids (Fig. 5A), and p38 phosphorylation was lower in non-N9a cybrids than in N9a cybrids (Fig. 5A). To examine how redox signals affect p38 and ERK1/2, we treated N9a cells, which exhibited higher ROS generation and a lower NAD⁺/NADH ratio than non-N9a cells, with NAC and NAM to reduce ROS generation and increase the NAD⁺/NADH ratio, respectively. Our results showed that only ERK1/2 phosphorylation was affected by the mitochondrial redox signals in our cybrids (Fig. 5B and C). ERK1/2 activation has been associated with the expression of inflammation (36) and with inflammation-induced insulin resistance (37). Therefore, we measured the mRNA levels of 38 inflammation-related genes, which revealed that mRNA levels of four genes, IL13, TLR4, CSF3, and CCL3, were lower in non-N9a cybrids than in N9a cybrids (Fig. 5D). Because TLR4 expression is closely associated with ERK1/2 activation (36), we tested whether TLR4 is downregulated upon ERK1/2 inhibition. Treatment of cells with either U0126, a specific ERK1/2 inhibitor, or NAM, an effective antioxidant, caused a significant decrease in the mRNA level of TLR4 (Fig. 5E and F). Physiologically, non-N9a cells exhibited higher insulin-stimulated glucose uptake than N9a cells (Fig. 5G), whereas administration of the two antioxidants, NAC and NAM, upregulated insulin-stimulated glucose uptake in N9a cells (Fig. 5H). Furthermore, the upregulation of insulin-stimulated glucose uptake in N9a cells was mimicked when ERK1/2 phosphorylation was inhibited in N9a cells through U0126 treatment (Fig. 5H). Last, blockage of TLR4 signaling by using two TLR4 inhibitors increased the insulin-stimulated glucose uptake (Fig. 5I). Our results support the proposal that regardless of the other signaling pathways that might regulate insulin sensitivity and cellular glucose uptake, mitochondrial redox signal–mediated ERK1/2 phosphorylation contributes to the insulin-stimulated glucose uptake, at least partially through TLR4 activation.

Previously, a study showed that haplogroup N9a is associated with diminished T2D occurrence (i.e., N9a acts a “protective factor”) in both Japanese and Korean patients (18). However, this reported effect of N9a in the case of Japanese T2D patients has been challenged (19). In Taiwan, haplogroup B4, but not haplogroup N9, was found to be associated with T2D (17). Recently, we found that mtDNA haplogroup N9a was marginally associated with an increased occurrence of T2D and significantly associated with diabetic nephropathy incidence (27). Here, to evaluate the causal role of haplogroup N9a in T2D, we conducted another large-scale case-controlled study, which confirmed that haplogroup N9a could serve as a risk factor against T2D incidence in China (Table 1). Such conflicting reports on mtDNA population variants in common diseases have been highlighted by previously (38) and are also known to be common in the case of other diseases such as Leber hereditary optic neuropathy (39). Several major factors might contribute to the distinct reported effects of mtDNA lineage on human diseases, including inappropriate research design and statistical performance (20,21), divergent nuclear genetic backgrounds (24), and different environment factors (25). In the investigation of haplogroup N9a/N9 and T2D, distinct inclusion criteria used for T2D patients might also contribute to the conflicting conclusions reached regarding the relationship between haplogroup N9a/N9 and T2D (17,18). Here, we did not set a cutoff value for the age of the T2D patients, but the patients included in other studies were aged >40 or >30 years old (17,18). We do not believe that the use of age as an inclusion criterion affected the results here because only 17 of the 1,295 patients were <30 years old and none of them were genotyped as N9a; by comparison, 94 control participants were <30 years old, with 8 genotyped as N9a. The T2D risk presented by haplogroup N9a would be even higher than that we have reported if these patients and control subjects were excluded. Moreover, although a limited amount of lifestyle information and clinical data are available for the study participants, it is unclear whether other factors such as smoking contributed to the distinct effects of haplogroup N9a on T2D occurrence. The only recognized difference that could affect the contribution of these factors might be the inclusion criterion “HbA_1c” for the control participants (HbA_1c <6.2% [44 mmol/mol] in this study, <5.6% [38 mmol/mol] in Japan/Korea, <6.0% [2 mmol/mol] in Taiwan). Overall, we speculate that environment factors might contribute to the divergent responses of the same mtDNA in healthy humans because Chinese populations were studied both by us and by Liou et al. (17).

Cytoplasmic hybrid technology is widely used for investigating the effects of distinct mtDNA haplogroups on cellular physiological conditions, including insulin sensitivity (32,40,41). The mtDNA-lacking ρ0 human osteosarcoma 143B cells represent the most accepted cellular model for studying how mtDNA haplogroups influence cellular functions. Although the use of a disease-related cell model as the nuclear donor is the optimal method to uncover the pathogenic role of specific mtDNA haplogroups, 143B cells are widely used in the study of Parkinson disease (42), T2D (40), Alzheimer disease (41), and Leber hereditary optic neuropathy (43). Thus, as in other studies (3,40), we used 143B cells to evaluate how mtDNA haplogroup N9a affects insulin sensitivity. Furthermore, 143B cells express GLUT4, which can translocate to the plasma membrane upon insulin stimulation, and by performing glucose uptake experiments, we obtained data supporting the view that relative to N9a cells, the three non-N9a cybrids are more sensitive to insulin and exhibit higher levels of insulin-stimulated glucose uptake (Fig. 5G–I). Our analysis of mitochondrial function by using PBMCs obtained from N9a and non-N9a control participants further indicated that haplogroup N9a could affect mitochondrial function in disparate global populations (Fig. 2G and H). Moreover, distinct nuclear genetic backgrounds, such as the presence of NDUFC2 polymorphisms, might act as secondary genetic modifiers that enhance or reduce the effect of mtDNA haplogroup N9a in T2D (24).

Conflicting reports have been published on the effect of mtDNA haplogroup on T2D (16–19,22–24,44), and N9a is a haplogroup regarding which incongruent findings have been reported in Asian populations (17–19); therefore, it is necessary to comprehensively elucidate the biological role of mtDNA haplogroup N9a in the development of T2D. Previously, we showed that impaired mitochondrial function and increased ROS levels played a critical role in T2D (2). Accordingly, we detected lower mitochondrial function in N9a cybrids than in non-N9a cybrids (D4, G3, and Y1 cybrids) and confirmed that the mitochondrial redox status differed significantly between N9a and non-N9a cybrids by measuring mitochondrial ROS levels and the NAD⁺/NADH ratio (45–47). Notably, we found that in N9a cells, the mtDNA content was higher than that in non-N9a cells (Figs. 1A and 3H). N9a cells generated higher levels of ROS than non-N9a cells did, but ROS scavenging did not lead to upregulation of the mtDNA content in N9a cybrids (Supplementary Fig. 2), and the levels mtDNA replication related proteins did not differ between N9a and non-N9a cybrids; these findings suggest that specific SNPs in haplogroup N9a might affect the mtDNA replication process as previously described (11). In this scenario, the distinct mtDNA replication capacities of N9a and non-N9a cybrids might contribute to the difference in mitochondrial function (48). Here, we did not detect any TFAM-binding diagnostic SNPs in N9a cells, but currently we cannot exclude the possibility that the binding abilities of mtDNA replication–related proteins differ between N9a and other haplogroups.

The association of mtDNA haplogroup with degenerative disease such as Parkinson disease could be due to not only a decline in RCC activity (41), but also the subsequent difference in nuclear signaling pathways caused by the disparity in mitochondrial function (33,41). In a previous study, mtDNA haplogroup–responsive retrograde signaling pathways were linked to the insulin pathway (3). Here, we noted that the insulin pathway was less active, as indicated by phospho-IRS1 (Y896) levels, and the fold increase in plasma membrane GLUT4 was considerably lower in N9a cybrids than in D4 cybrids (3). Therefore, it appears highly likely that haplogroup N9a represents a risk factor for T2D. In this study, we adopted a widely used quantitative RNA sequencing technology to identify mitochondrial retrograde signaling pathways, which could uncover the mechanism underlying the effect of N9a in T2D. By performing transcriptome analysis, we determined that nuclear-encoded OXPHOS gene expression was higher in N9a cybrids than in non-N9a cybrids, which likely arises as a compensatory effect for mitochondrial redox stress as a result of the reduction in mtRNA levels in N9a cybrids (49). We then obtained further evidence indicating that N9a might be involved in T2D: most identified changes in biological performance were related to metabolic regulation (Fig. 4C). In another study, microarray analysis used for gene expression profiling yielded data indicating that N9a probably does not represent a protective factor in the case of T2D (14). Here, by targeting candidate signaling pathways, we demonstrated that mitochondrial redox signal–mediated ERK1/2 phosphorylation/activation, a pathway that has been frequently related to insulin sensitivity (50), contributes to the differences in insulin-stimulated glucose uptake between N9a and non-N9a cells. Furthermore, we found that in response to ERK1/2 overactivation, TLR4 activation was increased and insulin-stimulated glucose uptake was decreased in N9a cells. However, several questions remain unanswered, such as whether and how other signaling pathways (e.g., Wnt pathway) are regulated by mitochondrial function, and thus further investigation required to completely reveal the underlying role of N9a in T2D.

In summary, we have presented the most comprehensive analysis to date of mitochondrial function, mitochondrial retrograde signaling, and insulin-stimulated glucose uptake in the study of mtDNA haplogroups in relation to T2D. Our findings support a positive association between the mtDNA haplogroup N9a and T2D and further demonstrate that N9a cells exhibit an altered redox status, which might contribute to an increased risk of T2D through mitochondrial retrograde signaling pathways such as those involving ERK1/2 activation.

Acknowledgments. The authors thank the members of J.L.’s laboratory for valuable discussions on this work.

Funding. This work was supported by grants from the Chinese National Science Foundation (31671486 and 31501156), Zhejiang Provincial Natural Science Foundation of China (LY15H060007), and Specialized Research Fund for the Doctoral Program of Higher Education (20133321110001).

Duality of Interest. No potential conflicts of interests relevant to this article were reported.

Author Contributions. H.F. and J.L. designed the study. H.F., N.H., Q.Z., B.W., H.Z., Q.F., L.S., X.C., and F.S. produced the data. H.F., N.H., Q.Z., B.W., and J.L. analyzed the data. H.F. and J.L. wrote the manuscript. H.F. and J.L. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.