MicroRNA Signatures as Future Biomarkers for Diagnosis of Diabetes States

Diabetes results from the inability of pancreatic islets to maintain blood glucose concentrations within a normal physiological range. Clinical features are usually not observed until islets begin to fail and irreversible damage has occurred. Diabetes is generally diagnosed based on elevated glucose, which does not distinguish between type 1 and 2 diabetes. Thus, new diagnostic approaches are needed to detect different modes of diabetes before manifestation of disease. During prediabetes (pre-DM), islets undergo stress and release micro (mi) RNAs. Here, we review studies that have measured and tracked miRNAs in the blood for those with recent-onset or longstanding type 1 diabetes, obesity, pre-diabetes, type 2 diabetes, and gestational diabetes. We summarize the findings on miRNA signatures with the potential to stage progression of different modes of diabetes. Advances in identifying selective biomarker signatures may aid in early detection and classification of diabetic conditions and treatments to prevent and reverse diabetes.


Introduction
Pancreatic islets regulate glucose homeostasis through insulin and glucagon, which are central to key biological processes and energy homeostasis. Loss or impairment of islet function results in dysregulated blood glucose, which gives rise to multiple life-threatening complications including cardiovascular disease, neuropathy, nephropathy, blindness, and stroke. High blood glucose, the most common clinical sign of diabetes, is usually observed only when islet beta cells are already deficient or exhausted. In the case of type 1 diabetes (T1D), clinical symptoms manifest when >80% of islet beta cells are already lost to autoimmunity. During the progression of type 2 diabetes (T2D), islet compensation maintains normoglycemia for years without symptoms before progressing to a state of detectable impaired glucose tolerance (IGT) and impaired fasting glucose (IFG). In addition, during pregnancy, lack of islet compensation leads to gestational diabetes (GDM), which increases risk for T2D. Between 24% and 62% of people with diabetes are unaware of the disease, undiagnosed, and untreated [1], suggesting a large gap in current diagnostic practices. Early detection of islet cell stress preceding loss of islet function would allow for therapeutic interventions to delay or alleviate the onset of diabetes.
However, each of these measures have their challenges in meeting the aforementioned biomarker criteria. For example, insulin cfDNA reflects beta-cell death and is elevated in response to autoimmunity in NOD mice [10], recent-onset T1D [11,12], and islet transplant 24 h after islet infusion [13]. However, insulin cfDNA is not useful in predicting beta-cell death prior to the pathological events. Advanced glycation end-products, including HbA1c, on the other hand, reflect the effects of chronic hyperglycemia, which are useful in monitoring glucose control in diabetic patients. Other markers including oxidative DNA damage markers and advanced oxidation protein products are a result of chronic oxidative stress and implicated in diabetes pathogenesis. However, chronic oxidative stress is also implicated in other diseases, including cancer, Parkinson's disease, and Alzheimer's disease, and thus is non-specific. Branched chain amino acids and short-chain fatty acids are influenced by exercise and dietary changes (acute starvation or low-protein diets), maple syrup urine disease, and possibly hypermetabolic states such as sepsis, injury, or cancer [14].

miRNAs as Biomarkers
miRNAs are specialized short non-coding RNAs (20-22 nt) that inhibit target mRNA translation. Recent research on circulating miRNAs highlights their usefulness as biomarkers of diseases. Under various conditions, cells release miRNAs that are free or in microvesicles that can be taken up by other cell types. These extracellular miRNAs are important mediators of cell-to-cell communication and coordinate biological functions including angiogenesis, tumor cell invasion, and immune response. miRNAs are reliable biomarkers due to their resistance to degradation in circulation and their enrichment in particular tissues, which reveals the source of circulating miRNAs [15]. Modern technologies in nucleic acid amplification, sequencing, and analysis have allowed us to identify miRNAs produced and released by islets under stress conditions. For this review, we have compiled existing RNA sequencing, microarrays and quantitative polymerase chain reaction (qPCR) data from comprehensive studies in humans and present the potential use of miRNA signatures for future use in detecting pre-DM and/or staging progression and mode of diabetes.

Methods
To locate relevant articles, we searched PubMed, beginning with the following keywords: ((mir) or (miRNA) or (circulating mir) or (circulating miRNA)) and ((biomarker) or (biomarkers)) and ((diabetes) or (islet) or (beta)). The second search included the following search keywords: ((miRNA [Title/Abstract]) or (microRNA[Title/Abstract])) and ((diabetes[Title/Abstract]) or (obesity) or (prediabetes)) and (biomarker *). We filtered the results for clinical studies, trials, comparative studies, evaluation studies, journal articles, meta-analyses, multicenter studies, reviews, systematic reviews, and validation studies in the date range of 2009 to 2019, including only studies with human subjects. In addition, we excluded studies that used whole-blood miRNA analyses instead of plasma or serum analyses.

miRNAs as Biomarkers in Type 1 Diabetes (T1D)
Table 1 provides circulating miRNA profiles for patients with recent-onset T1D, T1D > 1 year, maturity-onset diabetes of the young, latent autoimmune diabetes in adults, and individuals at risk for T1D (autoantibody positive). Immune cell infiltration is the first event that leads to significant reduction in beta-cell mass and eventual hyperglycemia, followed by changes in blood glucose, autoimmunity, beta-cell mass, beta-cell death, islet compensation, and corresponding circulating miRNA profiles ( Figure 1). For the autoantibody-positive non-diabetic group, all miRNAs identified in 4 studies are provided in Figure 1. For recent-onset and longstanding T1D groups, only those miRNAs that were identified independently in at least 2 studies are provided in the figure. Any correlations of miRNAs to metabolic parameters are provided in Supplementary Table S1. It is evident that distinct groups of miRNAs can be identified for non-diabetic individuals vs. those with T1D.
TrialNet Pathway to Prevention cohort and center recruits qPCR P-14 C-10 [19] * P and C indicate patient and control sample size, respectively. LADA, latent autoimmune diabetes in adults; MODY, maturity-onset diabetes of the young; NC, no change; qPCR, quantitative polymerase chain reaction; seq, sequencing; T1D, type 1 diabetes.

Figure 1.
Circulating miRNA profiles at different stages of type 1 diabetes (T1D) pathophysiology. Changes in autoimmunity, beta-cell mass/death, blood glucose, and islet compensation over the course of T1D pathophysiology are provided as line profiles. Circulating miRNA profiles consistent in at least two studies are provided for T1D (<1 and >1 year). All miRNAs identified in autoantibody-positive non-diabetic children before onset of T1D are provided. Upregulated miRNAs are shown as brown; downregulated miRNAs, green; and upregulated or downregulated, blue. These conflicting observations are not surprising for at least three reasons: (1) the bulk of beta-cell death in T1D precedes clinical diagnosis, and hence any increase in circulating miR-375 is possibly missed at the time of sampling; (2) the degree, extent, and dynamics of beta-cell death differ between individuals; and (3) islet compensatory mechanisms and ongoing autoimmunity may influence circulating levels even after disease diagnosis. For instance, our observations in mice suggest that miR-375 levels are significantly elevated in circulation immediately (3-6 h) after streptozotocin administration, subsiding to undetectable levels within 24 h. In the context of islet transplantation, we observed that miR-375 levels increase markedly during islet autotransplantation in patients receiving total pancreatectomy with islet auto transplantation, but normalized to baseline levels 7 days after transplantation [35]. In mice, we also observed elevated miR-375 levels 24 h after human islet transplantation, indicating islet inflammation and damage in the peritransplant period [35]. Thus, miR-375 is a useful marker for islet cell damage during transplantation. In the context of diabetes prediction, miR-375 can be used alongside other biomarkers and not as a standalone biomarker.
In recent-onset T1D (<5 years), miR-200a-3p and miR-155-5p correlated negatively with HbA1c and IDAA1c levels [31]. This association was lost in patients with more than 5 years of T1D. Alterations in circulating miRNA profiles were also reported in another cohort, where miRNAs with elevated levels at baseline or 1 year after diagnosis were undetectable 4 to 8 years after diagnosis [27]. In fact, as depicted in Figure 1 and mentioned earlier, miRNAs in autoantibody-positive non-diabetic children were identified in independent studies involving recent-onset T1D or children with longstanding T1D (but not consistently in at least 2 studies, thus missing our threshold for inclusion in Figure 1). Nevertheless, these profiles of circulating miRNAs at different stages of T1D pathophysiology (Figure 1), especially those expressed during early stages of autoimmunity, are attractive candidates that should be validated further in international cohorts for diagnostic utility.

miRNAs as Biomarkers for Pre-Diabetes (Pre-DM) and T2D
Changes in fasting and postprandial glucose, insulin resistance, islet compensation, and circulating miRNA profiles in the pathophysiology of T2D are shown in Figure 2. This figure also lists all miRNAs identified in non-diabetic healthy individuals who proceeded to develop pre-DM or T2D. Notably, miRNAs identified independently in at least 2 studies for obesity, pre-DM, and T2D are included. For an extensive list of all differentially expressed miRNAs, see Tables 2 and 3. Association (positive or negative) of differentially expressed miRNAs with metabolic parameters are provided in Supplementary Table S1. Here, we provide a brief review of these correlation analyses conducted in different ethnic populations.
Case-control studies conducted in healthy populations with follow up for future development of pre-DM or T2D provide insights on circulating miRNA profiles before apparent changes in glucose homeostasis and onset of irreversible loss of beta-cell mass. For example, in healthy children aged 7 years, levels of miR-221, miR-28-3p, miR-142-3p, miR-486-3p, and miR-486-5p can be used for risk estimation and obesity classification [36]. In non-diabetic healthy adults, baseline miR-122, miR-15a, miR-197, miR-320a, miR-423, and miR-486 levels were inversely associated with progression to glycemic impairment at 2.5-year follow-up [37]. In another cohort of non-diabetic healthy adults, miR-320a and miR-486-5p increased the odds while miR-375 decreased the odds of insulin resistance [38]. In the Bruneck study cohort, normoglycemic individuals who developed T2D over 10 years were appropriately classified as diabetics based on baseline levels of miR-15a, miR-126, miR-320, miR-223, and miR-28-3p [39]. Thus, the assessment of miRNA signatures instead of a singular miRNA profile can help in not only accurately classifying DM but also distinguishing DM from other diseases.
Obesity increases the risk for development of T2D/metabolic syndrome and is termed 'diabesity' (obesity-induced diabetes). Although not the focus of our review, we provide a brief overview of miRNAs associated with obesity traits (Figure 2) and related effects on glycemic status. In Caucasian obese children, increased miR-486-5p/miR-486-3p, miR-142-3p and HOMA-IR, together with decreased miR-28-3p, explained about 60% of variance in body mass index [36].
In obese Europeans, miR-144-5p, let-7d, miR-34a, and miR-532-5p strongly predicted insulin resistance [62]. For other associations, see Supplementary Table S1. Interestingly, in patients with metabolic syndrome enrolled in the Practicing Restorative Yoga vs. Stretching for the Metabolic Syndrome (PRYSMS) study, weight loss improved the circulating miRNA profile [59]. In another cohort with obesity, an acute aerobic intervention in a single session further increased circulating levels of miRNAs involved in inflammation [66]. In contrast, in a different cohort participating in the Centro Universitario Ricerca Interdipartimentale Attività Motoria (CURIAMO) trial, a 3-month exercise intervention markedly reduced the levels of miR-146a-5p, also correlating with decreases in total cholesterol and waist circumference [72]. In patients who underwent bariatric surgery, miR-125b, miR-378a, miR-192, miR-629, miR-22-5p, and miR-15a levels were reduced [65]. Similar reductions in circulating miRNAs (although different miRNAs) were observed in an independent cohort after bariatric surgery [74]. Thus, miRNA profiles are helpful in understanding and tracking the efficacy of intervention programs.
As depicted in Figure 2, progressive insulin resistance induces islet compensation that maintains glucose homeostasis. Progression of the disease results in an abnormal glycemic state, characterized by a reduction in beta-cell secretory capacity. This abnormal glycemic state is represented either by IFG or IGT. However, individuals may exhibit both states, thus reflecting a heterogeneous pathogenesis of T2D. Despite impairment in glycemic states, it takes years for patients to progress to overt T2D, and the disease is often undiagnosed due to infrequent/insufficient analysis of both of these parameters. Moreover, distinguishing pre-DM patients from healthy populations has proven difficult due to fluctuations in the tested ranges in clinical parameters and the influence of lifestyle factors. In this regard, even though the timeline of progression from pre-DM to T2D varies, distinct miRNA signatures can be used to predict future development of T2D. In the Bruneck study cohort, miR-126 remained a significant predictor of T2D, with a gradual decrease in levels from controls to those with IFG/IGT to those with T2D. T2D cases can be correctly classified using a panel of miRNAs including miR-15a, miR-126, miR-320, miR-223, and miR-28-3p [39]. In two other cohorts, miR-126 had an inverse association with fasting glucose, HbA1c, and 1 h postprandial glucose [46,56]. In newly diagnosed T2D and pre-DM patients, low levels of serum miR-126 increase the odds ratio for T2D and can distinguish T2D from controls [47]. Pre-DM patients were best distinguished from healthy populations by using a binary random forest classifier based on levels of miR-146a, miR-126, miR-30d, and miR-148a [26]. In T2D patients, miR-486, miR-146b, and miR-15b correlated positively with fasting plasma glucose, and these biomarkers could be used to discriminate T2D patients from healthy controls [67]. Pre-DM patients with higher levels of miR-150 and miR-30a-5p or lower levels of miR-375 and miR-15a were at a higher risk of developing T2D [69]. A miRNA panel consisting of miR-7641-3p, miR-136-5p, miR-490-3p, miR-501-5p, miR-127-5p, miR-4532-5p, miR-483-5p, and miR-210-3p distinguished obese nonprogressors from progressors who developed T2D at 5-year follow-up [77]. Thus, using a panel of miRNAs and other clinical parameters including HbA1c to classify the health status of patients may help in the discovery of the disease before onset.
Use of biomarkers may help not only in predicting future development of disease, but also in monitoring intervention efficiency. For example, in pre-DM patients enrolled in the CORDIOPREV study, elevated miR-150 and miR-130a-5p levels were associated with a decrease in the disposition index (insulin signaling and release); elevated miR-150 levels were associated with a decrease in the insulin sensitivity index and muscle insulin sensitivity index; and elevated miR-375 levels were associated with a decrease in the hepatic insulin resistance index after 4-year follow-up [69]. In another group of pre-DM patients, miR-192 and miR-193b correlated positively with serum triglycerides and fatty liver index. An exercise intervention program significantly reduced these levels over 6 weeks [51]. In patients with metabolic syndrome, after a 3-month weight loss intervention, circulating levels of miR-326, miR-24, miR-425, and miR-652 increased while circulating levels of miR-106b, miR-140, miR-20b, miR-363, miR-486, miR-532, miR-92a, miR-93, and miR-let7c decreased. Among these, changes in miR-146a, miR-151a, miR-23a, miR-181b, miR-181d, miR-21, miR-221, miR-222, miR-223, miR-23, miR-24, and miR-27b levels strongly correlated with weight change over 3 months. Baseline levels of miR-143, miR-145, miR-146, miR-191, miR-221, miR-23a, miR-29a, and miR-584 strongly correlated with weight change after 3 months [59]. A vitamin D-induced increase in miR-152 levels correlated negatively with HbA1c, while a decrease in miR-192 levels correlated positively with fasting glucose in individuals at risk for T2D [63]. In a Chinese cohort with IFG, upregulated levels of miR-144, miR-20a, and let-7b increased the risk for T2D and positively correlated with HOMA-IR, while decreased levels of miR-142 increased the risk for IFG and negatively correlated with HOMA-IR [70].
The ability of miRNA signatures to predict pre-DM or T2D should be tested in various ethnic populations. It is possible that specific miRNA signatures may be driven by ethnicity-related factors (genetic or environmental). For example, in a Swedish cohort, miR-15a, miR-29b, miR-24, miR-126, miR-144, miR-223, miR-191, and miR-486-5p correlated inversely with insulin sensitivity index. In an Iraqi cohort, miR-197 (but not the miRNAs noted in the Swedish cohort) correlated positively with insulin sensitivity index, thus suggesting ethnicity-specific associations [86]. Overall, specific miRNA signatures have the capability to predict future development of T2D. Ethnicity-specific miRNA signatures should be established to improve the predictive potential of these signatures.

miRNA Biomarkers in Gestational Diabetes
Changes in blood glucose, insulin resistance, islet compensation and circulating miRNA profiles in GDM are summarized in Figure 3. Only miRNAs that were identified independently in at least 2 studies are listed. For an extensive list of all differentially expressed miRNAs, see Table 4. The association (positive or negative) of differentially expressed miRNAs with metabolic parameters is provided in Supplementary Table S1.
During pregnancy, insulin resistance imposes increased metabolic demand on islets. Islet adaptation to pregnancy is crucial not only for maintaining glucose homeostasis but also for preventing excessive nutrient flow from mother to fetus across the placenta. Placental lactogens induce beta-cell mass expansion, and islet neogenesis (in humans) has been demonstrated [96]. Lack of islet adaptation to pregnancy results in GDM, the incidence of which is steadily increasing. GDM also increases the risk of future development of T2D in both mother and child. Glucose tolerance is usually tested between 24 and 28 weeks of pregnancy. By this time, blood glucose levels are already significantly elevated and may affect the developing fetus. There is a need for predicting development of GDM earlier in pregnancy, so that intervention strategies can be applied before blood glucose levels increase.   Let-7e-5p, let-7g-5p, miR-100-5p, miR-101-3p, miR-146a-5p, miR-18a-5p, miR-195-5p, miR-222-3p, miR-23b-3p, miR-30b-5p, miR-30c-5p, miR-30d-5p, miR-342-3p, miR-423-5p, miR-92a-3p * P and C indicate patient and control sample size, respectively. GDM indicates gestational diabetes mellitus; qPCR, quantitative polymerase chain reaction; seq, sequencing. Table 4 lists the miRNA profiles identified in various studies at different stages of gestation. Of these miRNA profiles, the most consistently upregulated miRNAs before 20 weeks of pregnancy were miR-16-5p, miR-17-5p, miR-223, miR-210-3p, miR-342-3p, and miR-20a-5p, whereas miR-222 was the most consistently downregulated miRNA (Figure 3). In women 16.1 weeks' pregnant, higher levels of miR-155-5p and miR-21-3p were associated with higher odds for GDM. In women with male offspring (but not female offspring), miR-155-5p, miR-21-3p, miR-146b-5p, miR-223-3p, miR-517-5p, and miR-29a-3p were associated with GDM risk [101]. In pregnant women, miR-30a-5p, miR-130a, and miR-150 were associated with higher pregnancy weight gain, whereas miR-99b, miR-103, mIR-128a, miR-221, miR-324-3p, and miR-652 were associated with lower pregnancy weight gain. In pregnant women who were followed up to the 28th week, levels of miR-16-5p, miR-17-5p, and miR-20a-5p were positively correlated with insulin resistance (HOMA-IR) [100]. In pregnant women with GDM, high levels of miR-330-3p were inversely associated with insulinemia. In contrast, in GDM patients with lower miR-330-3p levels but levels higher than those of control women, insulinemia was higher than in the group with high miR-330-3p levels. Moreover, women with high miR-330-3p levels had an aggressive diabetic phenotype among GDM patients [102]. In pregnant South African women, miR-20a-5p and one or more risk factors were significant predictors of GDM [105]. The predictive potential for these miRNAs should be validated in further studies. Figure 4 depicts common miRNA signatures for those at risk for T1D, with recent-onset T1D, longstanding T1D, obesity, pre-DM, T2D, and GDM. For specific upregulation or downregulation of these common miRNAs in specific stages, see Tables 1-4. miR-148a-3p was common in all stages of T1D, suggesting its potential as an early biomarker specific to T1D pathogenesis. Among those at risk for T1D and with recent-onset T1D, miR-29a-3p, miR-342-3p, and miR-93-3p were common. miR-25 was common in individuals with recent-onset or longstanding T1D. miRNAs common to obesity, pre-DM, and T2D were miR-142, miR-126, and miR-21. miR-375 was common to both pre-DM and T2D patients, suggesting early involvement of this miRNA in the pathogenesis of T2D. miR-342-3p was common to those with GDM, those at risk for T1D, and patients with T1D, whereas miR-210 was shared by individuals with GDM and longstanding T1D. Thus, there appear to be some common signatures, but more studies are needed to make solid connections between various types of diabetes.

Challenges in Establishing miRNA Biomarkers
Despite careful analysis of a number of studies, our literature review showed only a handful of consistent miRNA signatures. It is possible to miss important miRNA signatures because of technical differences in sample handling, miRNA measurement, and data analysis. As for differences in serum and plasma, the contribution of platelets to miRNA content should be addressed. Circulating miRNAs should

Tissue of Origin and Functional Significance
The tissue of origin of circulating miRNAs is important to establish islet stress-specific signatures. However, such determinations are difficult to implement as miRNAs have diverse roles in different tissues in a context-dependent manner and thus are expressed in multiple tissues with differing expression patterns. For example, miR-375 is abundantly expressed in islets but is predominantly expressed in the pituitary gland. Although islets are important for glucose and energy homeostasis, coordinated actions of other organs including the brain, liver, adipose tissue, and muscle are central to regulation of whole-body metabolism. From our list of consistently up-or down-regulated miRNAs, we observed that miR-150-5p, identified in the circulation in non-diabetic individuals with autoantibodies, obesity, or T2D, was the only miRNA with predominant expression in the pancreas [109]. Nevertheless, irrespective of tissue source, miRNA signatures specific to early stages of diabetes are important for distinguishing diabetes-prone individuals from the healthy population. The functional significance of circulating miRNAs is unclear, with many reports suggesting roles in communication between different tissues. However, circulating miRNAs may also be a result of a specific pathological event in the course of disease. For example, reduced miR-126 in circulation in pre-DM and T2D signifies endothelial dysfunction (caused by hyperglycemia exposure) and correlates with subclinical and manifest peripheral artery disease [39]. Further investigations are needed to delineate functional roles of circulating miRNA signatures in diabetes.

Challenges in Establishing miRNA Biomarkers
Despite careful analysis of a number of studies, our literature review showed only a handful of consistent miRNA signatures. It is possible to miss important miRNA signatures because of technical differences in sample handling, miRNA measurement, and data analysis. As for differences in serum and plasma, the contribution of platelets to miRNA content should be addressed. Circulating miRNAs should ideally be cell free and reflect a true pathological state. Some of the studies reported here used sample pools for analyses which is not ideal for any analyses. Sample pools do not provide information on variability, with assumptions that all individuals in a cohort are identical in their disease state. Sample size is an important factor in determining true miRNA signatures for a particular disease state. We included studies with both small and large sample sizes in our analyses because it is important to establish consistency in miRNA signatures across independent centers. For example, miR-148a was elevated in a cohort of 16 newly diagnosed T1D patients  but this observation was also seen in large cohorts of 275 recent onset T1D patients (Nielsen et al., 2012). Thus including all studies in our analyses give us the opportunity to optimize and tailor future investigations to gaps in literature. In addition, in the studies we analyzed, normalization strategies varied from using endogenous miRNA controls to exogenous spike-in controls. While endogenous housekeeping controls are standard for gene expression analysis using qPCR, one needs to be careful before applying the same controls in measuring circulating miRNAs. For example, one of the studies used miR-191 as an endogenous control for normalization [43], but later studies confirmed that pre-DM and obesity may alter the circulating levels of miR-191 [37,64,78]. On the other hand, exogenous spike-in controls are used for handling technical differences in miRNA extraction and cDNA conversion. In addition, currently existing commercial kits recommend normalizing data to sample input (plasma or serum volume) rather than RNA concentration for cDNA input (as is the norm for qPCR). This is mainly because of difficulties in determining miRNA concentrations. Even though small RNA concentrations can be determined, miRNA fraction of the small RNAs may vary between samples. Such differences in small RNA input for cDNA conversion may introduce bias in the analysis. In our experience, we observed that in normal healthy conditions, the miRNA fraction of circulating small RNAs is lower and hence normalizing input for cDNA to sample volume may introduce bias in data analysis. Small RNA sequencing avoids this bias by normalizing the expression to number of reads instead of the sample input. However, this approach is not practical, as small RNA sequencing is too expensive to perform on a routine basis. Absolute quantification of miRNA concentrations using synthetic miRNA mimics can help overcome these problems. Absolute quantification will also help in comparing different data sets and in correlation analysis. To conclude, miRNA expression methodology should be standardized for possible clinical applications.

Future Directions and Conclusions
We have provided comprehensive information on the potential use of miRNA signatures selective for T1D, pre-DM, T2D, and GDM. Although islet-specific miRNAs are of special interest for identifying islet stress or damage, miRNA signatures not specific to islets may also be important in understanding the metabolic disarray involving multiple tissues in diabetes. Consistent miRNA signatures specific to different stages, identified especially in those studies involving non-diabetic individuals who developed pre-DM or T1D at follow-up, should be validated independently. In future studies, influences on circulating miRNA signatures by gender, age and other factors (other diseases, medications, lifestyle factors) should be investigated. These studies will help establish miRNA signatures that can be used clinically to predict diabetes in the general population. Further, amassing data samples, standardizing miRNA detection technologies, and tracking and validating correlations with disease states will improve their predictive and diagnostic efficacy for developing strategies of therapeutic intervention.