Metabolic changes in type 2 diabetes are reflected in peripheral blood cells, revealing aberrant cytotoxicity, a viral signature, and hypoxia inducible factor activity

The incidence of metabolic syndrome (MetS) and type 2 diabetes (T2DM) with its associated mortality and morbidity is rapidly increasing in the western world, leading to an extensive medical and societal burden. MetS is defined by a complex set of clinical parameters, which all constitute risk factors for the development of T2DM. These risk factors include central obesity, dyslipidemia (raised triglycerides and lowered high-density lipoprotein cholesterol), high blood pressure, and elevated fasting plasma glucose levels [1].

High blood pressure, hyperglycemia and dislypidemia are held responsible for the increased risk for cardiovascular disease in MetS and T2DM patients, including microvascular complications (including retinopathy, nephropathy and neuropathy) and atherosclerosis at the macrovascular level [1–4]. Both high fasting glucose levels and impaired glucose tolerance are associated with increased cardiovascular events [5], while impaired glucose tolerance predicts cardiovascular death [6]. Intensive glucose control, however, only modestly reduces cardiovascular events [3, 6], indicating a more systemic dysregulation in these patients.

T2DM may develop after a stage of insulin resistance in subjects with MetS. However, only approximately one-third of obese, insulin-resistant individuals develop T2DM because of an inability of beta cells to produce sufficient amounts of insulin [7]. Systemic and local activation of the immune system accompanies obesity, and contributes to the development of insulin resistance, T2DM and cardiovascular disease [8, 9]. It is not entirely understood which mechanisms trigger the onset of T2DM in subjects at risk, but inflammation is a critical candidate. Pancreatic inflammation in T2DM has been shown by increased local infiltration of macrophages in the beta cell areas [10, 11]. In animal models for type 2 diabetes, characterization of the increased infiltrating islet-associated macrophages indicated a pro-inflammatory M1 phenotype, with production of IL-1β and TNFα [12]. Adipose tissue may become a source of inflammation as well, as adipocyte hypertrophy is associated with increased macrophage accumulation, which produce proinflammatory mediators such as TNFα and IL-6, in obese individuals [8]. Trials aimed at inhibition of the immune system by blockade of IL-1α and –β signaling (anakinra) and inhibition of NFκB have shown to reduce HbA1c levels suggesting an effect on beta cell function in patients with T2DM [7, 12]. In animal models, the IL-1 receptor antagonist reduced the numbers of macrophages in islet areas and improved insulin sensitivity and beta cell function [12]. Infiltrating circulating immune cells are thus important in the development of islet dysfunction, in inflammatory adipose tissue, and in the development of atherosclerotic plaques leading to macrovascular disease. The consequence of acute hyperglycemia on circulating immune cells is largely unknown, but a proinflammatory role for hyperglycemia has been observed, because an oral glucose tolerance test increases transcript levels for ICAM-1, TNFα, and IL-6 in peripheral white blood cells from MetS subjects, but not from healthy controls without MetS [13].

Thus blocking inflammation improves glucose tolerance in T2DM patients, but whether in vivo hyperglycemia is able to initiate the activation of the immune system, without the influence of obesity, dyslipidemia, and high blood pressure, is not known. Along this line, it is thus also unknown whether the immune response to acute hyperglycemia is different between healthy subjects and MetS subject and T2DM patients.

Therefore we performed genome-wide expression profiling to achieve an unbiased analysis of the effect of hyperglycemia and T2DM on circulating immune cells. Peripheral blood cells express a large proportion (approximately 80 %) of the genes encoded in the human genome [14]. In addition, the changes in the expression levels of individual genes reflects changes in the (micro)environment of peripheral blood cells and may also reflect organ-specific changes to the same milieu [14], thus serving as a diagnostic tool for a diseased state. Previous studies using peripheral blood profiling revealed that in T2DM patients the mitochondrial oxidative phosphorylation pathway was affected compared to younger lean controls [15]. This finding is remarkable as a reduced expression of oxidative phosphorylation genes was previously established in muscle tissue from T2DM patients [16]. The immune system is affected by metabolic changes in various organs, and vice versa, the immune system influences metabolic parameters such as insulin resistance. Because of this cross-talk the immune system plays a central role in metabolic changes. To be able to obtain a global and unbiased overview of the changes in the immune system, in response to hyperglycemia and T2DM, we studied the alterations in gene expression in whole blood in T2DM patients in comparison with age-, sex- blood pressure- and BMI- matched MetS subjects and lean controls during fasting conditions and after a well-controlled hyperglycemic clamp. We reasoned that this approach may lead to an improved understanding of the underlying pathology of T2DM and may give leads for medical interventions. The comparison of our data with expression data sets from other studies under various experimental conditions, using pathway analysis, allowed us to identify novel pathological pathways.

Peripheral blood cells in T2DM are chronically exposed to abnormally high levels of glucose, due to the beta cell dysfunction in the presence of insulin resistance.

After imposing acute hyperglycemia in vivo in control individuals by applying a 70 min hyperglycemic clamp, we observed a markedly reduced expression of cytotoxicity-related genes. In T2DM patients the expression levels of these cytotoxic genes were already low, and were hardly affected by the hyperglycemic clamp. MetS subjects showed an intermediate response. We thus provide evidence that the reduced expression of cytotoxic genes is a direct consequence of high glucose levels. We did not observe a difference in baseline cytotoxic gene expression between age- and BMI-matched obese controls and MetS subjects, indicating that age and obesity may contribute to the low expression of cytotoxicity genes. In mice, cytotoxic gene expression declines with old age which is correlated with a reduced cytolytic capacity of cytotoxic T cells [36]. For humans, it has been reported that the functional capacity of virus-specific cytotoxic CD8⁺ T cells diminishes after the age of sixty [37]. High glucose levels may therefore accelerate the decline of functional CD8⁺ cells in MetS and especially T2DM patients (mean age was 59–60 years). This decline may influence the risk of cardiovascular disease, as experimental injury-induced arterial neointima formation is reduced by the most lytic fraction of CD8⁺ cells, due to lysis of vascular smooth muscle cells [38, 39], while CD8⁺ cells also mediate the athero-protective effect of immunization with an ApoB-100 peptide [39].

Our pathway analysis confirmed a low expression of effector cytotoxic cells in T2DM patients compared to MetS subjects, most likely indicating a reduced number of circulating cytotoxic cells, as was observed in azathioprine-treated Crohn’s Disease patients [35]. A reduced lytic cell activity, may lead to an increased vulnerability for viral infections and tumor development in the host. CD8⁺ cytotoxic T cells and NK cells are required for anti-viral and anti-tumor responses through their lytic activity (mediated by granzymes, granulysin and perforin) and non-lytic mechanisms through release of anti-viral cytokines [40, 41]. Thus a low activity of cytotoxic cells, indicated by the reduced expression of these genes in T2DM patients may cause an impaired response to cancer and viral infections. The incidence of cancer is indeed higher in subjects with MetS and T2DM [42]. For hepatocellular carcinoma in chronic hepatitis C patients the increased incidence in T2DM patients is also related to high concentrations of glycosylated hemoglobin (HbA1c), corresponding with our finding that hyperglycemia reduces cytotoxic genes. Active viral infections may also be more common in diabetes, suggested by the increased incidence of cytomegalovirus in atherosclerotic lesions in arteries from patients with T2DM [43], a virus which usually reactivates in immunocompromised hosts. The detrimental effect of elevated glucose levels on anti-viral responses has been previously observed in type I diabetes patients, which showed a decreased in vivo T cell response to influenza antigen [44] and decreased in vitro cytotoxic T cell response to influenza vaccination [45]. Both phenomena were associated with high concentrations of HbA1c, while B cell responses were normal.

However, other viral infections may be easily overlooked, because many-, including reovirus infections, commonly cause only mild symptoms or follow a subclinical course.

Our pathway analysis indicated that the expression profile of T2DM patients contained a viral signature of the Reoviridae family comprising reovirus [26], and Rotavirus [27]. The reovirus signature is derived from in vitro infection of human embryonic kidney cells, while the Rotavirus signature is derived from PBMC from Rotavirus- infected children. Nevertheless, the reovirus signature is more significant. The other virus-regulated gene sets included in our analysis; Influenza-, forty CMV-, two HPV- and one HCV-regulated sets of genes were not significantly associated with T2DM, indicating a certain level of specificity for reovirus. Reovirus is a double-stranded RNA virus that is believed to cause mild infections of the upper respiratory and gastrointestinal tract of humans, while Rotavirus (RV) is another ubiquitous double-stranded RNA virus, which may also cause gastroenteritis. Interestingly, reoviruses are known to experimentally induce type 1 diabetes, by inducing insulitis, reduction in insulin content of the pancreas and abnormal glucose tolerance [46, 47]. Likewise, Rotavirus infection is associated with a faster progression of diabetes in diabetes-prone mice [48] and with pancreatic apoptosis and hyperglycemia in nondiabetes-prone mice [49]. Previous studies already pointed towards a role for viruses in development human T1DM. For T1DM it is hypothesized that viruses contribute to beta cell destruction, in particular enteroviruses are strong candidates [50], because there is a clinically significant association between enterovirus infection and T1DM. In accordance with this hypothesis, enterovirus (echovirus) is able to infect and destroy human beta cells in vitro [51]. Additional evidence is provided by the fact that in more than half of the human T1DM patients, beta cells show clear enteroviral capsid protein vp1 expression, indicative for enteroviral infection, while only 6 % of controls are marginally positive. Interestingly, also 10 out of 25 T2DM patients were positive for enteroviral capsid protein vp1 expression, suggesting that viral infections may contribute to beta cell loss and insulin dependency in T2DM as well [52]. Our results support the hypothesis, using an unbiased approach, that viral infections (potentially reovirus), may contribute to development of insulin-dependent diabetes in T2DM patients.

The fact that the reovirus signature was present in T2DM but not in MetS subjects is in line with the observation that inflammation in general contributes to the development of T2DM [10–12], and supports the finding that inhibition of inflammation (by blockade of IL-1 signaling) improves HbA1c levels in T2DM patients without changing insulin sensitivity, but not in non-diabetic subjects with MetS [53]. Inhibition of inflammation therefore most likely has beneficial effects on insulin production in T2DM patients only.

Apart from the viral signature in T2DM, systemic inflammation was also indicated by the increased transcript expression of S100A12/ENRAGE, corresponding with the reported increased serum levels of S100A12/EN-RAGE in T2DM patients [54], CD164, a HIF-1α-response gene [55], encoding a sialomucin expressed on peripheral blood monocyte, involved in adhesion [56], and TLR1. Interestingly, mRNA encoding for the appetite hormone leptin was also increased in T2DM vs MetS, which is normally expressed in adipose tissue, and at higher levels per gram tissue from obese subjects compared to lean individuals [57, 58].

In search of transcription factors that may have caused the altered expression profile in T2DM patients, we identified highly enriched transcription factor binding sites for HIF1 and ARNT in genes with an aberrant expression compared to MetS patients (and lean controls). HIF1 is a heterodimer consisting of HIF-1α and constitutively expressed ARNT (also termed HIF-1β) and regulates the response to hypoxia and growth factors by controlling energy metabolism [59]. Indeed, gene ontology analysis confirmed that metabolic processes were affected in T2DM patients compared to circulating cells from MetS subjects. Affected metabolic processes have been demonstrated before in blood cells from T2DM versus MetS, or versus nondiabetic controls [60, 61], although the subjects were not matched for BMI as in our study. It may not be surprising to identify metabolic processes in metabolically disturbed patients, however it does indicate that these metabolic changes are reflected in the gene expression profile of peripheral blood cells and can thus be monitored. Interestingly, HIF-1α activity, experimentally induced upon intermittent hypoxia, increases plasma triglycerides through upregulation of sterol regulatory element binding protein (SREBP)-1 activity in the liver [62]. These results correspond to our findings that T2DM patients show higher HIF-1α activity, associated with higher plasma triglyceride levels. Furthermore, the genes showing higher transcript levels in T2DM patients not only contained binding sites for HIF-1α and ARNT, but also for SREBP-1, thus reminiscent of the molecular switches during hypoxia in the liver.

The previously described reduced expression of genes involved in electron transport chain and oxidative phosphorylation in PBMC from T2DM vs young lean controls [15] is consistent with increased HIF-1α activity in T2DM. We confirmed the reduced expression of electron transport chain and oxidative phosphorylation genes in T2DM when we compared these patients to young lean controls (Additional file 1B). It is quite remarkable that the analysis of only three controls and six T2DM patients results in such highly significant pathways which are consistent with previously published data.

It is not quite clear what exactly caused the increased activity of HIF-1α in T2DM patients. In normal physiology, hypoxia regulates HIF-1α activity on the protein level by preventing its degradation by an oxygen-dependent mechanism, reviewed in [63]. However, it has been suggested before that hyperglycemia mimics the effects of hypoxia, in terms of an increased cytosolic ratio of NADH/NAD+, referred to as pseudo hypoxia in diabetes patients [64]. Reactive oxygen species (ROS) may also stabilize HIF-1α under normoxic conditions, resulting in increased glycolytic metabolism and increased NADH/NAD+ ratio [65]. High dietary fat intake and obesity causes increased release of skeletal muscle mitochondrial ROS species (H2O2) accompanied by a change in redox environment to a more oxidized state (generally referred to as oxidative stress) and insulin resistance in mice and humans [66] Targeting the altered redox state in mice by neutralization of H2O2 by an antioxidant or by overexpression of catalase in mitochondria of muscle cells prevents high fat diet-induced insulin resistance [66]. The two-fold increased expression of catalase in circulating cells from T2DM patients compared to MetS suggests that a compensatory mechanism for increased oxidative damage is activated in T2DM. Therefore, it might very well be that in T2DM patients, chronic hyperglycemia and oxidative stress contributes to the HIF-1α- regulated metabolic change in peripheral blood cells. Indeed, prolonged high glucose levels in vivo induces the expression of the HIF-1α-responsive gene VEGF and vascular dysfunction in rats [67]. Interestingly, adipocyte-specific disruption of ARNT or disruption of HIF-1α in mice results in similar metabolic phenotypes. In both mouse models the high fat diet-induced abnormalities are reduced, including diminished fat formation, protection from obesity and insulin resistance [68], suggesting a role for HIF1 in the pathogenesis of obesity and insulin resistance.

The gene expression profile in peripheral blood seems to partially mimic the profile of adipocytes, with respect to HIF-1α activity and leptin expression (both associated with obesity), and to partially mimic the response in the liver, reflected by the enriched binding sites for SREBP-1 normally activated by HIF-1α in the liver. These findings are in line with the suggestion that altered gene expression in specific organs, is reflected in peripheral blood cells [14].

Using an unbiased approach we have revealed that a hyperglycemic state affects the cytotoxic immune capacity. The reduced cytotoxic potential may render the T2DM patients less capable to clear viral infections. This suggestion is supported by the identification of a reovirus expression signature in T2DM patients. It is tempting to speculate that viral infections in T2DM fuel systemic inflammatory processes and may even lead to infection of pancreatic beta cells, as has been shown in experimental animals, and thus contribute to the induction of insulitis and insulin-dependent diabetes.