Rare Variant in Metallothionein 1E Increases the Risk of Type 2 Diabetes in a Chinese Population

Genetic and environmental factors both play a crucial role in the development of type 2 diabetes (T2D) (1–3). Advances in genome-wide association studies have helped identify genetic variants linked to T2D. These studies have also demonstrated significant differences in the genetic background between Chinese and Caucasian populations (4). However, the combined effect of common genetic variants is still unable to explain the total heritability of T2D; the impact of rare variants remains uncertain because of limited study power (5,6). Therefore, future studies should focus on identifying common and rare variants associated with T2D.

Similar to other complex diseases, T2D also has its Mendelian forms, such as maturity-onset diabetes of the young (MODY). MODY is an early-onset nonimmune form of diabetes that exhibits an autosomal-dominant mode of transmission. Rare variants in 13 genes have been established as the cause of MODY. Common variants within these MODY genes have also been demonstrated to increase the risk of later-onset diabetes (7). Next-generation sequencing has enabled the identification of rare or low-frequency variants of MODY genes with strong effects in a tightly restricted subset of the human population (8,9). Based on these findings, MODY and T2D are not distinct conditions but rather opposite ends of a continuum of diabetes subtypes. Currently, the genetic background of ∼30% of families with MODY remains unexplained (called MODYx, a group of seemingly monogenic diabetes) (10).

A powerful and cost-effective strategy for identifying critical variants is to trace the coinheritance of potential disease alleles with diabetes in families with seemingly monogenic diabetes. In this study, we used whole-exome sequencing (WES) in a three-generation family with early-onset diabetes to identify the most potentially causative rare variant cosegregating with diabetes. We also conducted in vitro and in vivo functional studies to deeply explore the contribution of the putative gene to the development of diabetes.

This study was conducted in compliance with the principles of the Declaration of Helsinki. The study protocol for humans was approved by the ethics committees of Peking University People’s Hospital and Peking University Health Science Center (Beijing, China), and animal work was approved by the ethics committee of Peking University People’s Hospital. Written informed consent was obtained from all study participants.

A three-generation family with diabetes (named family 1) was recruited in 2012. The proband and her family members are citizens of Beijing. Diabetes was diagnosed according to the 1999 World Health Organization (WHO) criteria. Normal glucose tolerance was defined using WHO criteria based on a 75-g oral glucose tolerance test. Seven participants (III-1, III-2, II-1, II-2, II-3, I-1, and I-2) from family 1 underwent WES.

A total of 96 unrelated participants with T2D were recruited. These patients were diagnosed with T2D before age 35 years and lacked the classic presentation of T1D or tested positive for T1D-related autoantibodies. Next-generation sequencing (cat. no. 5190-4857; Agilent SureselectXT2 Custom 0.5–2.9 Mb library) using the Illumina HiSeq2500 was used to screen variants in all the exons of the metallothionein 1E (MT1E) gene in S1.

A total of 849 patients with common T2D were used to identify additional carriers of the potentially causal variant (MT1E p.C36Y) using the MassARRAY iPLEX system (MassARRAY Compact Analyzer; Sequenom, San Diego, CA), a high-throughput genotyping method. All of these participants lacked the classic presentation of T1D and lacked positive results for T1D-related autoantibodies.

We recruited participants from a community-based cohort formed between 2012 and 2013, the details of which have been previously published (11). According to the 2011 WHO criteria, diabetes defined as fasting plasma glucose ≥7 mmol/L and/or 2-h plasma glucose ≥11.1 mmol/L, HbA_1c ≥6.5%, or previously diagnosed diabetes according to a questionnaire. Prediabetes was defined as fasting glucose ≥6.1 and <7.1 mmol/L, 2-h plasma glucose ≥7.8 and <11.1 mmol/L, or HbA_1c ≥6.0 and <6.5%, and normoglycemia was defined as fasting glucose <6.1 mmol/L and HbA_1c <6.0%. In total, 3,350 participants age 18 to 75 years were identified, including 545 diagnosed with diabetes based on an oral glucose tolerance test, 797 with prediabetes, and 2,008 with normal glucose tolerance. Because C36Y is located at exon 3 of MT1E, exon 3 was amplified by PCR, and participants underwent genotyping by Sanger sequencing.

Transgenic mice were developed on a C57BL/6 background using the PiggyBac micromanipulation technique (Cyagen Biosciences, Guangdong, China). Purified plasmids containing wild-type (WT) MT1E (hMT1E[NM_175617.3]) or mutated MT1E p.C36Y were injected into the zygotes along with a piggyback helper, which was subsequently transplanted into the fallopian tubes of female mice. MT1E was incorporated into the zygote genome. This genome could then be passed on to the offspring over successive generations. Transgenic mice with MT1E p.C36C (MT) and MT1E p.C36Y (MU) mutations were housed with their WT littermates. Both lines were bred and maintained under controlled light and temperature conditions in a specific pathogen-free area of the animal housing facility at Peking University People’s Hospital. Male mice age 8 to 10 weeks were given a high-fat diet (HFD; 60% of kilocalories from fat; Beijing Hfk Bioscience Co., Ltd, Beijing, China) for 30 weeks. An intraperitoneal glucose/insulin tolerance test was used to assess glucose homeostasis in the animals. Fasting serum was collected for biochemistry tests, and ELISA was performed for insulin. mRNA and protein levels of genes involved in glucose and fat metabolism in the liver, kidneys, and gonadal fat of C56BL/6 mice were assessed. RNA sequencing was used to explore the underlying molecular mechanisms in MU and MT mice (Supplementary Methods).

For human studies, normally distributed continuous variables are presented as means ± SDs, and nonnormally distributed variables are presented as medians (interquartile ranges). Categorical data are presented as numbers and ratios. The Student t test was used to compare the means of the quantitative traits if applicable. Categorical variables were compared between groups by the χ² or Fisher exact test. P < 0.05 was considered statistically significant. Statistical analyses were performed with SPSS (version 20.0; (Chicago, IL).

For animal studies, data are expressed as means ± SEMs. Two-way ANOVA followed by a post hoc (Sidak) test was performed to compare group differences and interactions between genotypes and diets. Significance was set at P < 0.05. Figures and statistical analyses were performed in GraphPad Prism (version 8.0).

Details are provided in the Supplementary Material.

The data that support the findings of this study are available from corresponding author L.J. upon reasonable request.

The study design is depicted in Supplementary Fig. 1. Four of five members of a family with diabetes had an age at onset ≤40 years (details described in Supplementary Table 1). Using sequencing data filtering, 26 rare variants were identified (Supplementary Table 2) that cosegregated with diabetes in family 1. Four variants located on chrX (CCNB3, DMD, TEX11, and XK) were obviously incompatible with the genetic model of this family and were excluded. Ten rare variants (Supplementary Table 3) were predicted to be harmful by at least two programs (ProVean, Sift, and Polyphen2). The frequency of the variants ranged from 0 to 0.000004 in the 1000 Genomes and GnomAD data sets. Of these, p.C36Y (rs766233163) in MT1E was thought to be the most potentially causative variant because the MT1E gene is closely related to metal ion binding and islet β-cell and adipocyte function (12–14). The frequency of MT1E p.C36Y was 0 in 1000 Genomes, 0.000052 in GnomAD, and 0.00203 in the Chinese mBiobank (www.mbiobank.com) data sets. Additionally, the spatial structure of MT1E p.C36Y differed from that of MT1E p.C36C (Fig. 1D), as predicted by Αlphafold (Google DeepMind, New York, NY). The variant site (36th amino acid) is conserved across species, and the mutated cysteine residue is considered to be involved in metal binding, as predicted by the RCSB Protein Data Bank (https://www.rcsb.org) (Fig. 1E and Supplementary Fig. 2C). Therefore, we selected MT1E as a candidate diabetes-related gene.

Supplementary Table 4 lists the characteristics of S1. Two patients were identified with V49I (rs201582313; 0.0002 in 1000 Genomes and 0.000258 in GnomAD data sets), which was predicted to be benign by bioinformatics tools, including Provean, polyphen2, and SIFT. One individual was a carrier of the MT1E p.C36Y variant; his family members were subsequently recruited (family 2). This pedigree included five individuals who were diagnosed with diabetes (as depicted in Fig. 1B). Sanger sequencing confirmed that II-2 and II-3 with diabetes also carried MT1E p.C36Y. The characteristics of this family are listed in Supplementary Table 5.

The characteristics of S2 are summarized in Supplementary Table 6. One patient with diabetes carrying MT1E p.C36Y was identified, and his family members were recruited (family 3) (pedigrees shown in Fig. 1C, and details listed in Supplementary Table 7). The proband’s mother, elder sister, and elder brother had been diagnosed with T2D. They were all confirmed to carry the MT1E p.C36Y variant by Sanger sequencing.

In the Pinggu cohort, we identified seven participants with the MT1E p.C36Y variant, eight with the p.G40D variant (rs728184174), and 14 with the p.V49I variant (rs201582313). The frequency of MT1E p.G40D was 0.0375 in 1000 Genomes and 0.060291 in GnomAD data sets.

Among the seven carriers of p.C36Y, four were diagnosed with diabetes and three with prediabetes. The MT1E p.C36Y allele frequency was 0.26% in participants with impaired glucose regulation (IGR), including those with diabetes and prediabetes, and 0% in participants with normoglycemia (χ² = 4.888; P = 0.027). There was no difference in p.G40D frequency between participants with IGR (0.19%) and normoglycemia (0.12%; χ² = 0.084; P = 0.772). There was also no difference in the p.V49I frequency between participants with IGR (0.34%) and normoglycemia (0.20%; χ² = 0.973; P = 0.324). Details of all the participants carrying all variants are listed in Table 1.

Compared with carriers of MT1E p.C36C, carriers of MT1E p.C36Y demonstrated increased HbA_1c, BMI, and postchallenge glucose and insulin levels, while exhibiting similar levels of serum triglycerides (TGs) and cholesterol. Considering that the MT1E p.C36Y variant was present in most of the members diagnosed with diabetes in these three pedigree groups and was associated with abnormal glycemic levels and obesity in the community, it was hypothesized that this variant greatly increases genetic susceptibility to T2D and metabolic syndrome, and in vivo functional studies were conducted to verify this hypothesis.

MT1E mRNA was upregulated in the liver and adipose tissue of MU mice fed a control diet (CD), and MT1E protein levels were mainly upregulated in the liver in MU mice fed with either diet (Supplementary Fig. 3A–H). MU mice fed an HFD exhibited increased MT1 mRNA levels in the fat tissue, whereas MU mice fed the CD demonstrated increased MT2 transcripts in the liver (Supplementary Fig. 3I–N). MU mice gained more weight than their littermates after 23 weeks of HFD feeding (Fig. 2A). Similar to humans with MT1E p.C36Y, who had elevated postchallenge glucose, glucose levels in MU mice were higher at 90 min after the intraperitoneal glucose challenge at baseline (Fig. 2B). The fasting glucose levels of MU mice were higher than those of their WT littermates at 24 (Fig. 2C) and 30 weeks (Fig. 2D). However, there were no differences in fasting serum insulin, serum TG, or serum HDL cholesterol levels among the different genotypes. MU mice fed an HFD had increased plasma ALT levels and liver TG content compared with their littermates (Fig. 2E–J).

The transcripts related to glucose metabolism were similar between MU and WT mice (Supplementary Fig. 3O and P). However, MU mice had increased transcript levels of hepatic CD36 and protein expression of FAS and peroxisome proliferator–activated receptors (PPARs) in the liver (Fig. 2L–Q). In adipose tissue, genes involved in lipid metabolism, including PPARs and FASN, were consistent between MU and WT mice (Supplementary Fig. 3Q–S).

MT mice demonstrated upregulated MT1E transcript expression in the liver, kidneys, and fat tissues and increased MT1E protein levels in adipose tissue when fed the CD (Supplementary Fig. 4A–H). The levels of MT1 and MT2 in the liver, fat, and kidneys did not differ between MT mice and their WT littermates (Supplementary Fig. 4I–N). Unlike MU mice, MT mice demonstrated lower body weight from the third week compared with their WT littermates (Fig. 3A). MT mice had lower glucose levels from 30 to 120 min after glucose challenge at the 16th week of HFD feeding (Fig. 3B). They also had better insulin sensitivity than their WT littermates at week 17 of HFD feeding (Fig. 3C). At week 30, serum glucose and TG levels were reduced in MT mice fed the CD. In addition, serum ALT and AST levels were reduced in MT mice fed an HFD. There was no difference in the fasting serum insulin, HDLC or hepatic TG content between MT mice and their WT littermates (Fig. 3D–J).

No alterations were observed in genes involved in glucose or TG metabolism in the liver of MT mice compared with that in WT mice. However, there was a genotypic difference in CD36 transcript levels in the liver (Supplementary Fig. 4O–T). In adipose tissue, PPARα transcripts were upregulated in MT mice fed the CD, with a trend toward upregulation of PPARα protein expression. However, there was no significant difference in PPAR or FAS transcripts between MT mice and their WT controls (Fig. 3K–O).

Our study suggests that a rare variant of MT1E may result in enhanced susceptibility to developing T2D in humans. MT1E p.C36Y was located in a highly conserved region of MT1E and clustered in almost all members with diabetes of the three families with diabetes included in this study. In a randomly sampled community cohort, this variant was associated with hyperglycemia and obesity. Forced expression of the MT1E variant recapitulated the human phenotype, whereas forced expression of WT MT1E alleviated high blood glucose levels, weight gain, and hypertriglyceridemia in mice.

In this study, we observed that MT1E p.C36Y cosegregated with diabetes in families 1 and 3. However, in family 2, only the proband’s father and aunt with diabetes were carriers of MT1E p.C36Y, and his cousin with diabetes did not carry this variant. This could be attributed to genetic heterogeneity (additional possible causative genes) or phenocopy, a phenomenon that was also observed in previous studies of MODY (15). This cousin had a BMI of 44.9 kg/m², which may also lead to early-onset diabetes. Unlike those of common variants, association studies of rare variants of T2D usually require a larger sample size, at the expense of cost. Because of the heterogeneity of T2D and the lack of precise phenotype measures, it was difficult to observe ideal cosegregation of causative variants in the families with T2D than it was in the families with monogenic diabetes. Rare causative variants are more likely to be enriched in affected members of families with diabetes; therefore, MT1E could possibly be a candidate gene associated with T2D. We obtained further evidence through in vivo functional studies, which indicated that MT1E p.C36Y greatly increases the risk of T2D. Therefore, this strategy of combining genetic analysis with functional studies resulted in the discovery of new genes related to diabetes. A more precise phenotype definition beyond hyperglycemia is required to identify the real cosegregation of a genetic variant with diabetes subtypes, followed by the identification of putative candidate genes for diabetes.

In fact, several population-specific low-frequency genetic variants, such as p.Gly319Ser of HNF1A in the Oji-Cree population, p.Glu508Lys of HNF1A in the Latino population, and p.Arg1420His of ABCC8 in the southwest American Indian population (8,9,16), have been reported to increase the risk of diabetes with larger effect sizes than those of common variants. We noted that the Y allele frequency of MT1E p.C36Y in the Chinese mBiobank database (0.00203) was significantly higher than that in Caucasians in the 1000Genome (0), GnomAD (0.000052), and EXAC databases (0.000049). This finding suggests that MT1E p.C36Y may play a greater role in increasing the risk of T2D in the Chinese population.

Structural analyses and predictions suggested that MT1E p.C36Y may affect the binding of metal ions and lipid metabolism. MTs are intracellular proteins that are abundant in cysteine and involved in various functions, such as heavy metal detoxification, copper and zinc homeostasis, reactive oxygen species defense, DNA protection, cell survival, angiogenesis, apoptosis, and proliferation (17). These proteins, primarily MT1 and MT2, consist of 61 to 68 amino acids, 20 of which are conserved cysteines (Supplementary Fig. 2) (18). The 36th amino acid, cysteine, was conserved across different species (Fig. 1E) and different isoforms of MT1 (Supplementary Fig. 2). Twenty cysteine residues in the MT1E sequence coordinate with metal ions to form a tetrahedral structure. The binding of metal ions to MT1 generates distinct α and β domain structures. The α domain, located at the C-tail, binds four metal ions (Fig. 1E and Supplementary Fig. 2C). The p.C36Y mutation, which replaces a cysteine residue at position 36 with a tyrosine, is located in the α domain; it may affect metal ion binding (19) (Supplementary Fig. 2C). RNA sequencing analysis revealed changes in the expression of ion-binding genes (CYP29, CYP37, and CYP50) in MU and MT mice (Supplementary Fig. 5). This result suggests that MT1E p.C36Y may alter ion-binding affinity and stability.

Human MT1E and adjacent genomic regions are associated with phenotypes other than T2D, including HDL, cholesterol, TGs, and BMI, suggesting that common variants within this area may be involved in lipid metabolism, cholesterol homeostasis, and obesity (Supplementary Fig. 2D). The relationship between MT1E and T2D was considered to be specific to the variant p.C36Y. This study also predicted the lack of binding of hsa-miR-4668-3p, which is involved in hepatic fat metabolism (20), to the mutated MT1E sequences (Supplementary Table 8). This suggests the possibility that the variant disrupts the binding of hsa-miR-4668-3p to MT1E, which may affect the metabolic phenotype.

Our in vivo results indicate the involvement of MT1E in lipid metabolism and its association with the human phenotypes of hyperglycemia and obesity. MT1E p.C36Y may contribute to the development of fatty liver, as evidenced by the increased serum ALT levels, liver steatosis, and expression of hepatic gene expression related to TG accumulation. PPARγ has been demonstrated to play a significant role in adipogenesis and liver steatosis (21). Additionally, CD36 is involved in fat accumulation (22) and insulin resistance (23). Our results are consistent with previous studies that have shown that MT1 regulates Zn and Cu homeostasis and protects against oxidative stress (24). The phenotype of MT1E p.C36Y was consistent with that of MT1/2-knockout mice, which exhibited increased fat deposition, obesity, insulin resistance, hyperleptinemia (14,25–27), and hyperglycemia (27) after a high-fat challenge. MT1E protein expression was mainly upregulated in the liver rather than in other tissues, suggesting that the forced expression of hepatic MT1E pC.36Y may have driven the phenotype of MU mice (Supplementary Fig. 3A–H). Furthermore, our results indicate that the observed phenotype was likely due to the MT1E p.C36Y variant rather than the downregulation of intrinsic mouse MTs, because we observed increased rather than reduced hepatic mouse MT1 levels in MU mice (Supplementary Fig. 3I–N).

Forced MT1E p.C36C expression in mice promoted a metabolically protective profile. This suggests that MT1E could be a therapeutic target for further drug development. The liver phenotype of MT mice was not as pronounced as that of MU mice, although ALT and AST were reduced in MT mice. However, serum TG levels were decreased, and lipid metabolic genes were reduced in fat tissue. The disparity in the gene expression levels between the liver and fat tissue can be attributed to the increased abundance of MT1E transcripts in adipose tissue. MT1E protein expression was mainly upregulated in adipose tissue and kidneys (Supplementary Fig. 4A–H). Because the kidneys were less involved in glucose and lipid metabolism, we speculate that the forced expression of MT1E pC.36C in adipose tissue may have driven the phenotype of MT mice. The intrinsic expression of MT1 and MT2 transcripts was not upregulated in MT mice (Supplementary Fig. 4I–N); therefore, the forced expression of MT1E could have been the potential mechanism of action, which could be used in the development of a novel antidiabetic drug with the added benefits of weight loss and resolution of dyslipidemia.

There is growing evidence that MT1E plays an important role in T2D development (28). The mRNA levels of MT1E were increased in the islets of patients with T2D (12). Forced MT1E expression or MTs may protect against β-cell dysfunction (29,30). However, there is also evidence that isolated islets from MT1-overexpressing mice displayed impaired glucose-stimulated insulin secretion (12). In fact, we observed that INS-1 cells transfected with MT1E p.C36Y grew slower and were arrested in the G1 and S phases, with increased apoptosis, under high glucose and lipid concentrations. In contrast, cells transfected with MT1E p.C36C showed the opposite phenotype (Supplementary Fig. 6A–I). Our human studies did not reveal any clinical β-cell deficiency in the probands or the MT1E p.C36Y carriers. In particular, in S3, carriers of MT1E p.C36Y showed even higher postchallenge insulin levels. This result suggests that the variant carriers were prone to insulin resistance rather than insulin deficiency. Therefore, we did not explore the β-cell protective effect of MT1E in detail.

Our study provides evidence from an integrated spectrum of family, cohort, and animal studies, supporting the role of MT1E p.C36Y in diabetes susceptibility. However, this study has the following limitations: 1) although we screened the whole exome of participants from family 1, we were unable to exclude the possible effects of other genetic variants and their interactions, which may have been underestimated or undetected, and 2) we used transgenic mice with unstable expression of MT1E p.C36Y or MT1E p.C36C. To compensate for this limitation and better explain the phenotype of our mouse model, we measured the expression of MT1E in two tissues that mainly affect glucose and lipid metabolism (liver and fat) and found that the expression of proteins varied across tissues and diets (Supplementary Figs. 3A–H and 4,A–H). C36Y mRNA may simply be more strongly expressed in some settings and tissues as a result of a difference in chromosomal localization between the WT and mutant transgenic mice, an impact on mRNA stability (e.g., because of altered mature form of miRNA binding, protein translation, or posttranscriptional regulation). The instability of the model may affect our interpretation of the causal relationship between MT1E variants and diabetes. An alternative strategy is the use of CRISPR/Cas9 to generate humanized MT1E mice. Nevertheless, replacing mouse MT1 with MT1E may not be an ideal humanized mouse model, because mice do not have any other isoform of MT1 (3). We did not determine the detailed mechanism of how MT1E p.C36C and MT1E pC.36Y affect hyperglycemia, obesity, or lipid metabolism. Altered food intake or energy expenditure might have contributed to the obesity of the transgenic mice, as evidenced by the increased body weight and food intake in MT-null mice (14). Nevertheless, we could not directly measure these parameters. Although the seemingly phenotypic differences between the WT and variant carriers were likely driven chiefly by women in humans, we did not investigate the impact of different genotypes in female mice. In addition, we are unable to provide data from other hyperglycemia-related tissues, including muscle and islet, regarding expression in WT and MU or MT mice. The RNA sequencing results suggest that the mechanism in MT and MU mice may be involved in steroid metabolism but in different ways. Further investigation is required to understand the mechanisms underlying this difference and how WT and mutated MT1E affect metabolism.

In conclusion, we identified a rare variant, MT1E p.C36Y, that may increase susceptibility to T2D, obesity, and metabolic syndrome. MT1E could be a candidate therapeutic target in T2D.

Acknowledgments. The authors thank all participants for their cooperation; Hong Lian, Chaochao Yang, and the medical staff of the Department of Endocrinology and Diabetes Metabolism for their collaboration; and Dr. Xiangjun He and Xin Yu in the unit of flow cytometry from the central laboratory of Peking University People’s Hospital for their support. The authors acknowledge Elsevier Language Editing Services for manuscript proofreading.

Funding. This work was supported by the Research Key Project of the Ministry of Science and Technology of China (2016YFC1304901), the National High-Technology Research and Development Program of China (2012AA02A509), and the Natural Science Foundation of Beijing Municipality (7142163).

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

Author Contributions. X.Zo. performed the in vivo study and wrote the manuscript. M.H. and X.Hu. performed the in vivo study and contributed to the data analysis of the cohort study. L.Z. performed the in vitro study and revised the manuscript. M.L. and X.G. contributed to the in vivo study. J.C., Y.Lu., and X.C. contributed to participant recruitment for the clinical cohorts. L.M. and N.L. contributed to the in vitro study. Y. Li. and X.Zh. assembled the community cohort. Y.S. contributed to the sample analysis of the in vivo data. X.Ha. designed the study, conducted the genetic study, contributed to the in vivo and in vitro studies, and wrote the manuscript. L.J. designed the study and revised the manuscript. X.Ha. and L.J. 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.