The goal of the present study was to understand whether metabolic factors affect the expression of the genes recently implicated in the development of type 2 diabetes for which there was little prior evidence of their potential role(s) in this disease. Although many additional SNPs have been identified in subsequent GWAS and meta-analyses
, we focussed these studies on the genes identified in the first waves of GWAS, as these have been the subject of most follow-up studies to date. Specifically, we examined acute changes in expression of these genes in response to feeding and fasting and longer term changes in the expression of these genes in response to a diet high in fat and sugar, recognized as a critical environmental risk factor for type 2 diabetes.
It has been hypothesized that most of the new genetic variants affect β-cell function, development or survival but not insulin sensitivity
. Consistent with this, we found all of the genes except Adam30 and Cdkn2a were expressed in pancreatic islets. These genes were expressed, however in the transformed β-cell line, MIN6. The expression of all the genes except Lgr5 decreased following incubation of the islets in high glucose concentrations. It can thus be hypothesized that these genes may normally play a beneficial role in islet function, and a reduction in the expression of these genes could contribute to glucotoxic β-cell dysfunction or survival. However, we also found evidence that most of the genes could have potential roles in other metabolically-relevant tissues. Genes affecting insulin sensitivity may be expected to be expressed in peripheral insulin sensitive tissues, such as liver and adipose tissue, and be responsive to metabolic status. Consumption of a high fat diet was associated with a tendency for the expression of several of these genes to be decreased. Similarly, many of the genes were regulated by feeding and fasting. Only the two splice isoforms of Cdkn2a had no evidence of metabolic regulation in any of the other tissues examined.
Jazf1, also known as Tip27, encodes a transcriptional repressor of Nr2c2, an orphan nuclear receptor of the steroid receptor family also known as TR4 and TAK1. Nr2c2 has been reported to modulate apoptosis
[42, 43] and its loss in mice is associated with reduced mitochondrial function and increased oxidative stress, and conversely with reduced adipose tissue inflammation, hepatic steatosis and insulin resistance
[44–46]. Jazf1-mediated alterations in Nr2c2 could thus affect both insulin sensitivity and β-cell function. Genetic variation in Jazf1 has been variably associated with measures of insulin sensitivity and β-cell function
[47–52], and our expression data support roles for this gene in both. Jazf1 was expressed in nearly all tissues examined and its expression in islets was decreased following culture in high glucose-containing media. Consistent with a pathogenic role in islets, it has recently been shown that JAZF1 expression is reduced in individuals with type 2 diabetes or hyperglycemia, and that JAZF1 expression was correlated with insulin secretion
. However, our findings suggest the reduced expression may be a consequence of their hyperglycemia, not the underlying cause. These data are consistent, however, with a role of Jazf1 in further accelerating β-cell dysfunction once individuals develop hyperglycemia or perhaps impaired glucose tolerance. Jazf1 expression in the liver and hypothalamus was decreased in mice fed the high fat diet, with the same tendency in adipose tissue. The GWAS SNPs may affect the expression of Jazf1 in adipose tissue, suggesting that its function in this tissue may be important for its role in type 2 diabetes
. That we also observed changes in the expression of Jazf1 in the brain and hypothalamus, suggests further potentially important sites of action.
Adamts9 is an anti-angiogenic factor known to be expressed in vascular endothelial cells
 and is implicated in endoplasmic reticulum to Golgi transport
. Although some studies have found associations between SNPs in this gene and various measures of insulin sensitivity or secretion
[49, 50], many others have not
[47, 48, 52, 57]. Microvascular structure affects both insulin secretion and sensitivity
. Adamts9 expression tended to be downregulated by high glucose in islets and was decreased in the hypothalamus and liver of high fat diet-fed mice. These data suggest that this gene may play a role in the neural regulation of metabolism in addition to having effects, perhaps on insulin sensitivity, in the liver and also in islets that may be related to its role in vascularization
Hhex encodes a homeobox transcription factor known to be involved in pancreatic and liver development
[60, 61]. SNPs in HHEX have been associated with decreased insulin secretion perhaps due to alterations in vesicle docking
[57, 62–66]. Most studies have failed to find associations between HHEX SNPs and insulin sensitivity
[52, 57, 62–64, 67], although associations with insulin clearance and hepatic insulin sensitivity have been reported
[52, 64]. We found Hhex to be expressed in several tissues besides the pancreas, with evidence of decreased expression in the liver in response to high fat feeding and increased expression in the brain in non-fasted mice. These findings provide potential support for roles of Hhex in metabolism outside the pancreas. The GWAS SNPs associated with type 2 diabetes are located between Hhex and Ide. There was prior evidence for a role of Ide in type 2 diabetes
[21, 22, 25, 26]. We found Ide expression to be increased by high fat feeding in adipose tissue and to be decreased by the incubation of islets in high glucose, consistent with a role of Ide in β-cell function
. Combined, these data suggest potential roles for both Hhex and Ide in type 2 diabetes susceptibility.
Little is known about the potential roles of Thada in metabolic disease. It is a gene associated with a common chromosomal breakpoint in thyroid cancers, that may affect cell death receptors
. We found widespread expression of Thada and noted its decreased expression in islets following culture in high glucose. Thada expression was also reduced in response to high fat feeding in both adipose tissue and the hypothalamus. No evidence for association of SNPs in this gene with insulin sensitivity have been found
[48, 49, 52, 57], and while most studies have not found associations with insulin secretion
[47–49, 57], an association with reduced insulin secretion in response to non-nutrient secretagogues and potentially β-cell mass has been reported
. Our data are consistent with a primary role of this gene in the pancreas in determining type 2 diabetes susceptibility, although indicate it may also have effects on the regulation of energy balance and metabolism.
Although Adam30 (A disintegrin and matrix metalloprotease 30) has been reported to be expressed only in the testis
, we found evidence of expression in several metabolically relevant tissues including the brain, adipose tissue, heart and stomach. It was not expressed in islets. We observed a marked reduction of its expression in adipose tissue collected from non-fasting animals, providing a potential site of action whereby this gene may affect metabolism and thus type 2 diabetes risk.
In contrast to the above genes, the expression of other genes were not widely altered aside from within pancreatic islets, consistent with the primary mechanism by which they are associated with diabetes being through alterations in islet biology. The mechanisms by which SNPs at the CAMK1D-CDC123 locus affect diabetes susceptibility are unknown, and it is unclear which of these two genes is affected by the causative genetic variation. Some, but not all, studies have found associations of SNPs at this locus with insulin secretion, while no associations with insulin sensitivity have been found
[47–50, 52, 57]. There is evidence that genetic variation near CAMK1D can affect its expression, at least in lymphocytes
. The expression of both genes was similarly reduced in islets cultured in high glucose, suggesting the possibility that they are under common regulatory control in these cells. We found decreased expression of Camk1d in the hypothalamus of high fat-fed mice and increased expression in other regions of the brain in non-fasted mice, suggesting it may affect the neuronal control of metabolism or islet function. In contrast, no substantial changes in Cdc123 expression in response to feeding and fasting or high fat diet consumption were observed. As the SNPs at this locus are primarily associated with insulin secretion and the expression of both genes in islets was altered, these data cannot distinguish which may be the causative gene.
SNPs within CDKAL1 have been associated with insulin secretion and not insulin sensitivity
[13, 47, 52, 57, 63, 66, 67, 70]. Cdkal1 is a tRNA modification enzyme. Specifically, this protein is a methylthiotransferase that modifies tRNALys, stabilizing interactions between the tRNA and mRNA, decreasing misreading of its cognate codon
. Mice deficient in Cdkal1 have impaired glucose tolerance and insulin secretion, and evidence of β-cell ER stress
[71, 72]. We found that Cdkal1 expression in pancreatic islets was decreased following incubation in high glucose, which could contribute to β-cell dysfunction in settings of hyperglycemia. Interestingly, we also found that Cdkal1 expression was reduced in the fed state in the hypothalamus, suggesting it may have metabolic functions in addition to those in insulin synthesis and secretion.
The GWAS have identified SNPs at the Cdkn2a-Cdkn2b locus. The Cdkn2a (cyclin dependent kinase inhibitor 2a) gene has two alternative splice isoforms that encode distinct proteins, Cdkn2a and Arf. The Arf isoform is generated by the use of an upstream alternative first exon. Both are involved in cell cycle control. We found expression of the Arf isoform to be very low in chow-fed mice and not detectable in most tissues from high fat diet-fed mice. We found higher levels of expression of this gene in islets, and potential downregulation of its expression by high glucose concentrations. In contrast, the Cdkn2a isoform was found in a limited number of tissues and not in islets. These data suggest that Arf might be the relevant isoform of Cdkn2a, affecting diabetes by affecting β-cell mass. Given the loss of expression of this gene in high fat diet-fed mice it is tempting to speculate that this may be a mechanism affecting high fat diet-induced metabolic dysfunction. The expression of Cdkn2b was decreased in adipose tissue of high fat-fed mice. As this gene encodes a cell cycle inhibitor, this reduced expression may reflect increased proliferation of adipocyte precursors or perhaps infiltrating inflammatory cells. Interestingly, Cdkn2b expression was also decreased in islets incubated in high glucose, consistent with a role in the regulation of β-cell mass and glucose-induced β-cell proliferation
. Thus, as with the Hhex-Ide locus, these studies cannot distinguish whether Cdkn2a or Cdkn2b is the causative gene, and actually suggest a role for both in the development of type 2 diabetes.
Ext2 is a glycosyltranferase involved in the synthesis of heparin sulphate, and mutations in this gene are associated with abnormal bone growths (exostoses)
. This gene may also be involved in neural development
. The association between SNPs in this gene and type 2 diabetes has not been as well replicated
[10, 12, 13, 62]. We found increased expression of this gene in brain, suggesting a possible site of action as to where this gene could affect diabetes risk.
Lgr5 is a seven transmembrane receptor and a member of the rhodopsin family
. It is a marker of mitotically active intestinal stem cells and potentiates Wnt/β-catenin signalling
[76, 77]. This is the only gene for which we did not find a significant decrease in expression in islets cultured in high glucose, although this certainly does not preclude it from having a role in pancreatic and β-cell development. Its expression was increased in the brains of non-fasted mice, suggesting another potential site of action through which it may mediate type 2 diabetes susceptibility.
Tspan8, also known as Co-029, is a cell surface protein implicated in pancreatic, colon and liver tumors and their metastasis, possibly through interaction with integrins
. Although some studies have found associations between SNPs in this gene and insulin sensitivity or secretion
[47, 57], others have not
[48–50, 52]. Loss of this gene is associated with decreased body weight, although there were no detectable effects on glucose tolerance or insulin sensitivity
. In contrast to that study which did not detect expression of this gene in mouse pancreas
, we found expression of this gene in isolated pancreatic islets and suggested regulation of its expression by glucose. Tspan8 expression was significantly decreased in brains of fed compared to fasted chow-fed mice, suggesting it may also have a role in the neural control of metabolism.
In summary, we have identified nutritional regulation of many of the newly found type 2 diabetes-associated genes. As these studies were performed with a relatively small number of samples, it should be noted that smaller changes in expression may also exist that we had insufficient power to detect. These data provide support for the involvement of these newly identified type 2 diabetes susceptibility genes in β-cell function and also suggest potential roles for many of them in peripheral tissues, notably in the brain and hypothalamus, highlighting the potential importance of neuronal regulation of metabolism and islet function to type 2 diabetes
[38–41]. Our study also highlights the tissue-specific regulation of these genes (changes in one or more tissues where the gene is expressed but not in all tissues), suggesting that the SNPs identified in the GWAS studies may need to be examined in the appropriate tissues and under several metabolic contexts
. Indeed, recent studies aimed at identifying genetic variants that affect gene expression (eQTLs) have found varying effects of these SNPs on gene expression in different tissues, particularly for SNPs located within not between genes, and notably that the SNPs were more associated with expression of diabetes-associated genes in metabolically relevant tissues such as liver, adipose and muscle than in lymphocytes, which are sometimes used as a surrogate because they are easily accessible
[80–82]. The abundant regulation of these genes by nutritional status found in our study also suggests there are likely gene-diet interactions involving these SNPs
 that may be a complicating factor in future human studies to assess the functional implications of the associated SNPs.