Diabetologia

, Volume 58, Issue 9, pp 1978–1992

Circulating microRNAs and diabetes: potential applications in medical practice

  • Juliette Raffort
  • Charlotte Hinault
  • Olivier Dumortier
  • Emmanuel Van Obberghen
Review

DOI: 10.1007/s00125-015-3680-y

Cite this article as:
Raffort, J., Hinault, C., Dumortier, O. et al. Diabetologia (2015) 58: 1978. doi:10.1007/s00125-015-3680-y

Abstract

The explosive increase in the worldwide prevalence of diabetes over recent years has transformed the disease into a major public health concern. While diabetes can be screened for and diagnosed by reliable biological tests based on blood glucose levels, by and large there are no means of detecting at-risk patients or of following diabetic complications. The recent discovery that microRNAs are not only chief intracellular players in many biological processes, including insulin secretion and action, but are also circulating, has put them in the limelight as possible biological markers. Here we discuss the potential role of circulating microRNAs as biomarkers in the context of diabetes and its associated complications.

Keywords

Biomarkers Circulating microRNAs Diabetes Human studies Review 

Abbreviations

FPG

Fasting plasma glucose

GDM

Gestational diabetes mellitus

IFG

Impaired fasting glucose

IGT

Impaired glucose tolerance

miRNA

microRNA

PAC

Proangiogenic cells

Introduction

The worldwide prevalence of diabetes has drastically increased over the past 50 years and the disease is expected to affect 592 million people by 2035 making diabetes a major public health concern [1]. Diabetes is a heterogeneous disease resulting from a variety of mechanisms leading to chronic hyperglycaemia [2, 3]. The disease can be classified in general categories including type 1, type 2, gestational diabetes mellitus (GDM) and other specific types associated with disorders affecting the liver, the pancreas or endocrine glands such as the adrenal glands. While the molecular and cellular mechanisms underlying the disease are not totally understood, the role of microRNAs (miRNAs) in diabetes pathophysiology has recently emerged. miRNAs form a class of small endogenous non-coding RNAs that act as post-transcriptional regulators of gene expression [4]. Hence, miRNAs impact on physiological processes, including cell proliferation, differentiation, apoptosis and metabolism [4, 5, 6]. Consequently, their dysregulation appears to be involved in various disorders, including metabolic diseases such as diabetes (for reviews, see [5, 7]).

Interestingly, miRNAs have been detected in several biological fluids, including blood, urine, saliva, breast milk, tears, seminal, amniotic and cerebrospinal fluids [8]. While the origins of extracellular miRNAs are not totally clear, with some investigators reporting that they are the result of cellular degradation and others reporting them to be selectively secreted, it is understood that their relative compositions and concentrations vary between these fluids. Interestingly, extracellular miRNAs are very stable even under harsh conditions [9, 10]. The stability of circulating miRNAs may be explained, at least in part, by their association with proteins such as Argonaute-2 [11], HDL [12] and Nucleophosmin-1 [13] or their packaging in vesicles (e.g. apoptotic bodies, microvesicles, exosomes) [14, 15]. Owing to their stability, circulating miRNAs could be used as biomarkers, and this possibility is currently being explored in a wide range of pathologies, including diabetes [5, 16, 17]. Indeed, they can be easily collected using non-invasive procedures. Moreover, their levels can be measured by quantitative RT-PCR (qRT-PCR), which allows easy, rapid, sensitive and specific detection and quantification [18].

The aim of our review is to summarise the knowledge on circulating miRNAs in diabetes, and to analyse their potential relevance as biomarkers in medical practice. In view of the different studies, current limitations and perspectives are discussed.

Circulating miRNAs as biomarkers of the diabetic condition

In general, the diagnosis of diabetes is defined as fasting plasma glucose (FPG) >7.0 mmol/l (>126 mg/dl) or as a 2 h plasma glucose value after a 75 g oral glucose tolerance test (OGTT) >11.1 mmol/l (>200 mg/dl) [19]. The use of HbA1c reflecting the average glycaemia over a 3 month period, has also been recommended to diagnose diabetes, with a threshold of 6.5% [19]. While the above-mentioned methods are the gold standards for the biological diagnosis of diabetes, they have several shortcomings, two of which are major ones. First, the direct or indirect measurement of increased glycaemia reflects the settled disease situation. Moreover, these methods do not reflect the presence or absence of diabetic complications. Therefore, the identification of novel biomarkers that could predict the development of diabetes and identify diabetic complications, would represent a leap forward in terms of diabetes care.

In this context, the question arises of whether circulating miRNAs could be such biomarkers. The observation of the deregulation of miRNA profiles in different biological fluids of diabetic patients compared with healthy controls was the first evidence supporting the idea that miRNAs could be biomarkers of the diabetic disease (Table 1). Indeed, a differential miRNA profile between type 2 diabetic patients and healthy controls was found in whole blood [20, 21], serum [22, 23, 24, 25, 26], plasma [27, 28, 29, 30, 31, 32, 33] and plasma exosomes [34]. Studies revealed a different miRNA pattern in serum of type 1 diabetic patients compared with healthy individuals [35], as well as in plasma [36, 37] and urine [36]. Furthermore, the miRNA content of different types of blood cells were also deregulated in type 1 and type 2 diabetic patients compared with controls [38, 39, 40, 41, 42]. Finally, analysis of circulating miRNAs in women with gestational diabetes mellitus has revealed that in both serum and plasma some miRNAs have an altered profile compared with those in control women [43, 44].
Table 1

Summary of studies on circulating miRNAs in diabetes and related diseases

Disease

Population

Aim of the study

Sample

Main data

Reference

T2D

80 T2D

80 Controls

Explore miRNA profiles

Plasma

• Differential miRNA expression between T2D patients and controls: miR-20b, -21, -24, -15a, -126, -191, -97, -223, -320, -486, -28-3p

• Differential miRNA expression before manifestation of diabetes: miR-15a, -29b, -126, -223, -28-3p

Zampetaki et al, 2010 [27]

T2D

30 T2D

30 T2D-susceptible individuals (IFG)

30 controls

Explore expression of 5 miRNAs and investigate their potential as early predictors of diabetes in susceptible individuals

Plasma

• Differential miRNA expression between T2D patients and controls: miR-126

• Differential miRNA expression between susceptible individuals and controls: miR-126

Zhang et al, 2013 [30]

T2D

90 newly diagnosed T2D

90 controls

Explore miR-146a levels

Plasma

• Differential miRNA expression between T2D patients and controls: miR-146a

Rong et al, 2013 [29]

T2D

19 Iraqis with T2D

14 Swedes with T2D

65 Iraqis (controls)

54 Swedes (controls)

Investigate ethnic-specific expression of 14 miRNAs

Plasma

• Differential miRNA expression between T2D patients and controls: miR-24, -29b

• Dysregulation of miR-144 associated with T2D in Swedes, not in Iraqis

Wang et al, 2014 [28]

T2D

100 T2D from Kazak population

100 T2D from Han population

Study the possible genetic differences between Kazak and Han population

Plasma

• Differential miRNA expression between Kazak T2D and Han T2D: miR-375

• Differences in the methylation of promoter region of the miR-375 gene between Kazak T2D and Han T2D

Chang et al, 2014 [84]

T2D

1. 48 T2D, 45 controls

2. 17 metformin-treated T2D; 18 placebo T2D

3. 7 healthy volunteers

Investigate miRNA profile and its response to changes in insulin sensitivity

Plasma

1. • Differential miRNA expression between T2D patients and controls: miR-140-5p, -142-3p, -222, -423-5p, -125b, -192, -195, -130b, -532-5p, -126

2. • Modification of miRNA expression by metformin: miR-192, 140-5p, 222

3. • Modification of miRNA expression by molecules inducing insulin resistance: miR-222, 140-5p

Ortega et al, 2014 [31]

T2D

100 T2D

100 controls

Explore the changes in miR-375 levels and the methylation status of CpG within the promoter of the miR-375 gene

Plasma

• Differential miRNA expression between T2D patients and controls: miR-375

• Hypomethylation of 4 of the 8 analysed CpG units within the promoter of the miR-375 gene in T2D patients

Sun et al, 2014 [33]

T2D

23 T2D

12 T2D with peripheral arterial disease

26 T2D with peripheral arterial disease and chronic wounds

20 controls

Explore influence of impaired wound healing on miRNA signature and its functional importance

Plasma

• Differential miRNA expression between T2D and controls: miR-191, -200b

• Differential miRNA expression between T2D with peripheral arterial disease and chronic wounds and T2D: miR-191, -200b

• Correlation between miR-191, -200b and C-reactive protein and cytokine levels in T2D patients

Dangwal et al, 2015 [32]

T2D

18 treatment-naive T2D

12 controls

Study miRNA expression and explore its influence on glycaemic control

Plasma exosomes

• Differential miRNA expression between T2D patients and controls: miR-326, -let-7a, -let-7f

• Negative correlation between miR-326 and its putative target adiponectin

• Modification of miRNA expression by glucose-lowering treatment: miR-let-7a, -let-7f

Santovito et al, 2014 [34]

T2D

18 newly diagnosed T2D

19 prediabetic (IFG and/or IGT)

19 T2D-susceptible individuals with normal glucose tolerance

Investigate the clinical significance of 7 diabetes-related miRNAs

Serum

• Differential miRNA expression between T2D patients and controls: miR-9, -29a, -30d, -34a, -124a, -146a, -375

• No significant difference in miRNA expression between prediabetic and normal glucose tolerance group

Kong et al, 2011 [22]

T2D

160 newly diagnosed T2D

82 IGT

75 IFG

138 controls

Investigate whether miR-126 is associated with T2D and prediabetes

Serum

• Differential miRNA expression between T2D patients and controls: miR-126

• Differential miRNA expression between prediabetic patients and controls: miR-126

• Modification of miRNA expression by glucose-lowering treatment: miR-126

Liu et al, 2014 [23]

T2D

24 T2D

20 prediabetic (IFG and/or IGT)

20 controls

Explore the miRNA expression profiles of T2D and prediabetes

Serum

• Differential miRNA expression between T2D patients and controls: miR-23a, -let-7i, -486, -96, -186, -191, -192, -146a

• Differential miRNA expression between prediabetic patients and controls: miR-23a

Yang et al, 2014 [25]

T2D

1. 20 newly diagnosed T2D

21 IGT

22 IFG

29 controls

2. 3 IGT

3 IFG

12 controls

Explore the profile of circulating miRNAs in prediabetic patients

Serum

1. • Differential miRNA expression between T2D patients and controls: miR-191, -139-5p, -21

• Differential miRNA expression between prediabetic patients and controls: miR-191, -15b, -128, -125a-5p, -150, -192, -193b

• Differential miRNA expression in prediabetic but not in diabetic patients: miR-192, -193b

2. • Modification of miRNA expression by lifestyle intervention in prediabetic patients: miR-192, -193b

Parrizas et al, 2014 [26]

T2D

13 T2D

20 obese

16 obese + T2D

20 controls

Compare miRNA expression between obese, non-obese diabetic and obese diabetic patients

Serum

• Differential miRNA expression between controls, diabetic and obese diabetic patients: miR-138, -376a

Pescador et al, 2013 [24]

T2D

T2D with coronary stenting:

36 controls

36 treated with pioglitazone for 9 months

Study the effects of pioglitazone in neointimal hyperplasia after coronary stenting and the modification in circulating miRNAs

Serum

• Lower coronary neointimal volume in the pioglitazone group

• Differential miRNA expression in the pioglitazone group vs controls: miR-24

Hong et al, 2015 [60]

T2D

108 T2D + diabetic nephropathy

104 T2D

62 controls

Investigate differential gene expression profiles in diabetic nephropathy

Whole blood

• Differential miRNA expression between T2D patients and controls: miR-let-7a

• Differential miRNA expression between T2D with diabetic nephropathy and T2D: miR-let-7a

• Relationship between some single nucleotide polymorphisms in the regulatory region of miR-let-7a-2 gene and risk of diabetic nephropathy

Zhou et al, 2013 [21]

T2D

20 T2D

20 controls

Study the role of miRNAs in PBMCs in relation to inflammation and T2D

PBMCs

• Differential miRNA expression between T2D patients and controls: miR-146a

• Correlation between miR-146a levels and fasting blood glucose, HbA1c and inflammatory signals

Balasubramanyam et al, 2011 [41]

T2D

48 prediabetic

43 non-complicated T2D

36 T2D + coronary heart disease

46 controls

Study miR-103b expression and determine whether it could be a biomarker for the early diagnosis of T2D

Platelets

• Differential miR-103b and gene target SFRP4 expression between prediabetic, non-complicated T2D, coronary heart disease T2D vs controls: miR-103b

• Modification of miR-103b and SFRP4 expression by antiplatelet treatment in controls and prediabetic patients

Luo et al, 2015 [42]

T2D

83 T2D

Explore the potential of miR-29 as biomarker for diabetic nephropathy and atherosclerosis

Urine

• Correlation between urinary albumin excretion rate and miR-29a level

• Correlation between miR-29b and carotid intima–media thickness

Peng et al, 2013 [61]

Metabolic syndrome

50 metabolic syndrome

50 T2D

89 hypercholesterolaemia

30 hypertension

46 controls

Characterise miRNA expression of patients with the metabolic syndrome and compare them with individuals manifesting T2D, hypercholesterolaemia or hypertension

Whole blood Serum exosomes

• Dysregulation of miRNAs in T2D and in patients with the metabolic syndrome: miR-27a, -320a

Karolina et al, 2012 [20]

Metabolic syndrome

71 participants

Investigate the relationship between miRNA profiles and metabolic abnormalities related to the metabolic syndrome

Whole blood

• Correlation between miR-144-5p and glucose levels

• Correlation between miR-1207-5p and HbA1c

• Correlation between miR-484 and metabolites related to insulin resistance

Raitoharju et al, 2014 [48]

T1DM

T2DM

GDM

7 T1D

7 T2D

6 GDM

Explore miRNAs and mRNA expression profiles in different types of diabetes

PBMCs

• miRNAs shared among T1D, T2D, GDM: miR-126, -1307, -142-3p, -142-5p, -144, -199a-5p, -27a, -29b, -342-3p

• Specific miRNAs: 20 identified for T1D, 14 for T2D, 19 for GDM

Collares et al, 2013 [70]

GDM

24 GDM

24 controls

Explore the association between miRNA expression and GDM

Serum

• Differential miRNA expression between GDM women and controls: miR-29a, -222

Zhao et al, 2011 [43]

GDM

10 GDM

10 controls

Identify miRNAs as potential biomarkers for the early diagnosis of GDM

Plasma

• Differential miRNA expression between GDM women and controls: miR-16-5p, -17-5p, -19a-3p, -19b-3p, -20a-5p

Zhu et al, 2015 [44]

T1D

404 newly diagnosed T1D children

151 controls

Identify miRNAs associated with beta cell destruction and regeneration

Serum

• Differential miRNA expression between T1D patients and controls: miR-152, -30a-5p, -181a, -24, -148a, -210, -27a, -29a, -26a, -27b, -25, -200a

• Correlation between miR-25 levels and residual beta cell function and glycaemic control

Nielsen et al, 2012 [35]

T1D

20 T1D

20 controls

Explore miR-21a and miR-93 expression

PBMCs

• Differential miRNA expression between T1D patients and controls: miR-21a, -93

Salas-Perez et al, 2013 [39]

T1D

11 T1D

9 controls

Explore miRNA profiles

PBMCs

• Differential miRNA expression between T1D patients and controls: 44 miRNAs

• Identification of predicted targets of the differentially expressed miRNAs and functional analysis

Takahashi et al, 2014 [40]

T1D

19 T1D

Explore the levels of miR-326 in relation to ongoing islet autoimmunity

Peripheral blood lymphocytes

• Differential miRNA expression between antibodies positive (GAD and/or IA-2) and antibodies negative T1D patients: miR-326

Sebastiani et al, 2011 [49]

T1D

5 T1D

6 controls

Explore the miRNA expression profile in the regulatory T cells of T1D patients and healthy donors

Peripheral blood regulatory T cells

• Differential miRNA expression between T1D patients and controls: miR-510, -342, -191

• Differential miRNA expression between regulatory T cells and conventional T cells

Hezova et al, 2010 [38]

T1D

15 T1D

19 controls

21 T1D + diabetic nephropathy

15 patients with kidney transplant

37 patients with simultaneous pancreas–kidney transplant

Identify miRNAs associated with diabetic nephropathy and microvascular impairment

Plasma

• Differential miRNA expression between T1D patients and controls: miR-25, -27a, -130b, -132, -152, -340, -660, -126, -223

• Differential miRNA expression between T1D + diabetic nephropathy and T1D patients: miR-326, -126,-181a, -223, -574-3p

• Normalisation of some miRNAs after kidney transplantation and after pancreas–kidney transplantation

• Association between levels of some miRNAs and markers of microvascular impairment

Bijkerk et al, 2015 [37]

T1D

68 T1D children

79 controls

Explore 3 miRNAs and investigate their link with cardiovascular and diabetic nephropathy risk factors

Plasma, urine

• Differential miRNA expression between diabetic and controls in plasma and urine: miR-21, -210

• Differential miRNA expression between diabetic and controls in urine but not in plasma: miR-126

• Correlation between urinary miR-126 and HbA1c

Osipova et al, 2014 [36]

T1D

10 T1D without renal disease

10 T1D with nephropathy

10 T1D with intermittent microalbuminuria

10 T1D with persistent microalbuminuria

Investigate if miRNA profiles differ in the different stages of diabetic renal disease

Urine

• Differential expression of 27 miRNAs in different stages of untreated nephropathy

Argyropoulos et al, 2013 [64]

T1D

12 T1D with persistent microalbuminuria and normal renal function

12 T1D with normoalbuminuria

10 controls

Explore miRNA expression from T1D with and without incipient diabetic nephropathy

Urine exosomes

• Differential miRNA expression between T1D patients with microalbuminuria and T1D with normoalbuminuria: miR-155, -424, -130a, -145

Barutta et al, 2013 [65]

Limb ischaemia

Mouse model of limb ischaemia: T1D mice induced by streptozotocin, controls

Humans: 11 diabetic patients, 11 controls

Explore mechanisms involved in diabetes-induced impairment of endothelial function and reparative neovascularisation after limb ischaemia

Mice: limb muscles

Humans: plasma, limb muscles

• Mice: deregulation of miR-503 in ischaemic limb muscles and endothelial cells of diabetic mice; correction of impairment of postischaemic angiogenesis and blood flow recovery by local inhibition of miR-503

• Human: deregulation of miR-503 expression in plasma and muscle of diabetic patients with critical ischaemia

Caporali et al, 2011 [58]

Limb ischaemia

20 patients with CLI

122 CLI + T2D patients

43 controls

Mouse model

Study miRNA expression in proangiogenic cells from CLI patients and investigate circulating miR-15a and -16 as potential biomarkers

Serum, plasma, PAC from PBMCs

• Deregulation of miRNA expression in CLI PAC: miR-15a, 16

• Influence of miR-15a and -16 expression on PAC survival and migration

• Deregulation of miR-15a, 16 in serum of CLI patients with and without diabetes

Spinetti et al, 2013 [59]

Chronic kidney diseases

56 patients with chronic kidney diseases (17 with diabetic nephrosclerosis)

Explore urinary miRNAs in chronic kidney diseases

Urinary sediment

• Differential miRNA expression between diabetic glomerulosclerosis and other groups of kidney diseases: miR-15

• Correlation between miRNA expression and proteinuria, renal function, rate of glomerular filtration decline, risk of progression to dialysis-dependent renal failure

Szeto et al, 2012 [63]

Diabetic retinopathy

90 patients with PDR

90 patients with NPDR

20 controls

Investigate the relationship between serum miRNA and the development of diabetic retinopathy

Serum

• Differential miRNA expression between patients with PDR and those with NPDR: miR-21, -181c, -1179

Qing et al, 2014 [66]

Diabetic retinopathy

4 patients with macular hole

4 patients with PDR

Compare the miRNA expression profiles between patients with macular hole and those with PDR

Vitreous humour, serum

• Differential miRNA expression in the vitreous between patients with macular hole and those with PDR: miR-15a, -320a, -320b, -93, -29a, -423-5p

Hirota et al, 2015 [67]

CLI, critical limb ischaemia; NPDR, non-proliferative diabetic retinopathy; PAC, proangiogenic cells; PBMCs, peripheral blood mononuclear cells; PDR, proliferative diabetic retinopathy; SFRP4, secreted frizzled-related protein 4; T1D, type 1 diabetes; T2D, type 2 diabetes

Circulating miRNAs as early predictive biomarkers of diabetes

In addition to its use for the diagnosis of diabetes, blood glucose measurement also allows the detection of impaired fasting glucose (IFG; FPG 5.6–6.9 mmol/l [100–125 mg/dl]) and impaired glucose tolerance (IGT; 2 h plasma glucose value after OGTT 7.8–11.0 mmol/l [140–199 mg/dl]) [19]. IFG and IGT correspond to a prediabetic state and as such represent harbingers of future diabetes and its complications. An HbA1c range of 5.7–6.4% categorises individuals as being prediabetic [19]. However, FPG, OGTT and HbA1c measurements have some limitations in regards to predicting diabetes. Indeed, glycaemia monitoring does not allow the identification of individuals who are susceptible to developing diabetes when glucose homeostasis is still normal. Discovering new biomarkers to detect prediabetic individuals and individuals at high risk of becoming diabetic is relevant for clinical practice, as several reports indicate that intervention strategies can delay diabetes onset in individuals at high risk of the disease [45, 46].

Interestingly, some reports show that the miRNA profile in plasma or serum is deregulated in the prediabetic state before the development of overt type 2 diabetes. Indeed, type 2 diabetic and prediabetic individuals have a differential profile of circulating miRNAs compared with controls [23, 25, 30, 42]. Interestingly, for some miRNAs, such as miR-126 and miR-23a, the difference was significantly greater in type 2 diabetic than in prediabetic individuals, suggesting that their levels could be correlated with the worsening of the glycaemic status [23, 25]. Furthermore, in one study miR-192 and miR-193b were reported to be deregulated in the serum of prediabetic individuals but not in that of diabetic individuals, suggesting that prediabetic, diabetic and healthy individuals have distinct profiles [26]. Finally, altered miRNA expression (miR-15a, -29b, -126, -223, -28-3p) has been observed in serum from normoglycaemic individuals who developed type 2 diabetes over a 10 year period [27]. While additional investigations are clearly needed to strengthen these findings, the results are promising, as circulating miRNAs could be early predictive biomarkers of type 2 diabetes and could help to distinguish healthy individuals from prediabetic and diabetic patients.

To the best of our knowledge, there has been no report to date identifying circulating miRNAs as early predictive biomarkers of type 1 prediabetic individuals. In contrast, an analysis of circulating miRNAs in GDM before diagnosis revealed the dysregulation of some in the serum and plasma of women who went on to develop GDM compared with the controls [43, 44]. These results suggest that circulating miRNAs could be used as biomarkers for the early diagnosis or as a predictive factor for GDM. Hence, the measurement of circulating miRNAs could facilitate GDM detection, allowing early interventions to reduce maternal and fetal complications.

Circulating miRNAs as biomarkers of the disease process

Insulin resistance associated with type 2 diabetes is not evaluated by biomarkers as only indirect tools are available. For example, HOMA estimates steady-state beta cell function and insulin sensitivity based on algorithms that use fasting glycaemia and fasting insulinaemia measurements (www.dtu.ox.ac.uk/homacalculator/). Even if several studies have revealed a metabolic signature of insulin resistance in blood [47], no ‘simple’ nor bona fide biomarker of insulin resistance is currently available and systematically used in routine medical practice. Identifying and estimating the degree of insulin resistance is relevant as this phenomenon occurs very early in the pathogenesis of type 2 diabetes. Moreover, it could help in deciding the most appropriate approach to combat the development of overt diabetes.

Several studies have investigated circulating miRNAs of potential interest as biomarkers of insulin resistance. One study revealed that the deregulation of plasma miR-140-5p and miR-222 in type 2 diabetic patients was partially reversed after 3 months of metformin treatment, which improved insulin sensitivity as evidenced by a decrease in fasting glycaemia and HbA1c [31]. Interestingly, in healthy male volunteers, the plasma level of these two miRNAs appears to be altered after insulin infusion followed by intralipid and heparin mixture infusion, which is known to induce insulin resistance [31]. A second study reported a correlation between levels of circulating miR-484 and metabolites related to insulin resistance, which points to the potential benefit of circulating miRNAs as biomarkers of insulin resistance [48].

The possibility that circulating miRNAs could represent biomarkers of the disease process itself has also been investigated for type 1 diabetes. Levels of miR-326 were measured in peripheral blood lymphocytes from type 1 diabetic patients [49]. Note that these miRNAs were extracted from whole blood samples and were in fact associated with blood cells. Interestingly, the miR-326 level was increased in type 1 diabetic patients who tested positive for islet antibodies vs those who were antibody negative. Thus, increased miR-326 levels appear to be linked to ongoing islet autoimmunity and could be a candidate biomarker for the autoimmune process in type 1 diabetes.

Pancreatic failure is the common feature of all forms of diabetes, although the mechanisms involved are different depending on the diabetes type. Interestingly, an association between miR-25 levels and residual beta cell function was identified 3 months after diabetes onset in type 1 diabetic patients, suggesting that circulating miRNAs could reflect pancreatic endocrine function [35]. Among the miRNAs expressed in the endocrine pancreas, miR-375 has one of the highest levels of expression both in humans and rodents. In animal studies, miR-375 seems to play a complex role in beta cell function and physiology [5, 7]. Under physiological conditions its overexpression leads to reduced beta cell growth [50], while in stressed beta cells its decreased expression appears to counteract the beta cell hyperplasia [51]. Its altered expression in islets and/or pancreas has been observed in animal models with a predisposition to type 2 diabetes [52] and overt diabetes [7]. While miR-375 expression is not significantly augmented in human islets from glucose intolerant donors [53], increased levels have been observed in the pancreas [54] and blood of type 2 diabetic patients [22, 33]. Given that miRNAs have been described as being responsive to cellular and extracellular stress and to be used by cells to adjust changes in their environment, it is conceivable that, under conditions of harsh stress, such as type 2 diabetes, miR-375 is increased, while under circumstances of mild stress, such as glucose intolerance, its expression remains in the normal range. Its altered expression thus corresponds to beta cell failure. However, since miR-375 is highly expressed in both alpha and beta cells, it cannot be used to accurately estimate beta cell mass. Therefore, a beta cell-specific biomarker is needed to develop appropriate prevention strategies.

Circulating miRNAs as biomarkers of therapeutic follow-up

The biological evaluation of the efficacy of the treatment of diabetes is largely based on glycaemia monitoring and measurement of HbA1c [55]. Reflecting average glycaemia over several months, HbA1c is a relevant marker of long-term treatment efficacy. However, it is technically difficult to measure and has several limitations. First, the normal range may vary depending on the patient’s ethnic origin [56]. Second, when interpreting HbA1c levels confounders that may modify the results, such as the presence of haemoglobinopathies, need to be taken into account. Third, in situations of abnormal erythrocyte turnover, such as anaemia owing to haemolysis, transfusion or recent blood loss, the HbA1c value is not reliable and should therefore not be assessed [55]. Finally, given the worldwide pressures on healthcare systems, its cost precludes it from becoming a routine screening test in the near future. Therefore, novel biomarkers to evaluate the effectiveness of treatment on glucose homeostasis would be a great asset.

Several studies have reported a link between the level of circulating miRNAs and glycaemic control. In one study a correlation was found between miR-25 measured at 1 month after type 1 diabetes diagnosis and the HbA1c levels 3 months after onset [35]. In another cohort, consisting of 68 type 1 diabetic patients treated with insulin for at least 6 months, urinary miR-126 was lower in type 1 diabetic patients compared with controls and was inversely correlated with HbA1c levels [36]. In a third study, blood levels of miR-144-5p and miR-1207-5p reflected glucose and HbA1c levels, respectively [48].

In addition, according to other observations, the profile of circulating miRNAs changes upon treatment. Treatment, including lifestyle adjustments and medications, can partially normalise the deregulated miRNA profile observed in type 2 diabetic patients, as demonstrated by the modification of exosomal levels of miR-let-7a, miR-let-7f [34], miR-126 [23], miR-140-5p, miR-222 and miR-192 [31]. Surgical interventions such as kidney transplantation or simultaneous pancreas–kidney transplantation appear to partially normalise the deregulated miRNA profile observed in type 1 diabetic patients [37]. Furthermore, a partial normalisation of deregulated miRNAs (miR-126, -192, -193b) was also observed after lifestyle interventions in prediabetic type 2 patients [23, 26], and this was concomitant with an improvement of glycaemic control [26]. These results suggest that miRNAs could be potential biomarkers of treatment efficacy. In healthy volunteers, changes in miR-222 levels have been reported after an i.v. infusion of insulin [31]. Hence, diabetes treatment can modify the miRNA pattern, not only in type 2 diabetic and prediabetic individuals, but also in healthy people.

Circulating miRNAs as biomarkers to assess diabetic complications

Diabetes is associated with microvascular complications including nephropathy, retinopathy, and neuropathy, and with macrovascular complications such as atherosclerotic cardiovascular diseases. Both lead to increased risks of morbidity and mortality.

In medical practice, specific reliable biomarkers to predict and detect diabetic cardiovascular complications are lacking. Indeed, global cardiovascular risk can be estimated by a score index that takes into account age, sex, family history of cardiovascular disease, smoking and the presence of hypertension or dyslipidaemia. Moreover, the detection of cardiovascular problems is mainly based on clinical symptoms and medical tests, which in general provide late markers of the disease process. Thus, the identification of early and more specific biomarkers is eagerly awaited to allow optimal treatment and to combat progression of the complications.

Several studies have recommended plasma miRNAs as potential biomarkers of peripheral artery complications in type 2 diabetic patients [27, 32]. Among them, miR-126 is expressed in endothelial cells and hence is enriched in highly vascularised tissues such as the heart. Interestingly, miR-126 plays a role in the processes of angiogenesis and inflammation, and hence has been proposed as an intercellular mediator between endothelial cells and vascular smooth muscle cells [57]. Studies focusing on limb ischaemia have reported increased expression of miR-503 in the limb muscles and in the plasma of diabetic patients with critical ischaemia, suggesting that miR-503 could be a biomarker of ongoing ischaemia in diabetic patients [58]. In addition, miR-15a and miR-16 have been found to be increased in the serum and in proangiogenic cells of patients with critical limb ischaemia [59]. Levels of these two markers were found to be correlated with the risk of amputation after restenosis at 12 months post-revascularisation in type 2 diabetic patients with critical limb ischaemia. Levels of miR-15a were also correlated with the risk of post-revascularisation restenosis in those patients. Besides their potential use as biomarkers of peripheral artery disease and critical limb ischaemia, circulating miRNAs have also been investigated in atherosclerosis-related diseases. For example, miR-24 has been identified as a potential predictor of neointimal hyperplasia after coronary stenting in type 2 diabetics [60]. Furthermore, analysing urinary miRNAs in type 2 diabetic patients, miR-29b levels were found to correlate significantly with carotid intima–media thickness, pointing to the potential interest of this miRNA as a biomarker for atherosclerosis [61].

In addition to cardiovascular complications, nephropathy is another major diabetes-related public health concern, as diabetes is the leading cause of end-stage renal disease in the Western world. In general, detection of renal complications in diabetic patients is based on urinary albumin excretion measurements. However, these measurements vary under certain circumstances such as infection, fever, pronounced hyperglycaemia, severe hypertension or after physical exercise [55]. Moreover, a substantial percentage of diabetic patients have decreased glomerular filtration, reflecting kidney impairment in the absence of increased urine albumin excretion [62]. As yet, no biomarker in the urine is available to predict the decline in kidney function in those patients.

Assessment of miRNA expression in the urinary sediment of patients with chronic kidney diseases revealed a significant difference in miR-15 level between patients with diabetic glomerulosclerosis and those with other renal diseases [63]. The levels of 27 miRNAs vary significantly at different stages of untreated nephropathy in the urine of patients with type 1 diabetes [64]. In addition, a differential miRNA level in urine exosomes was found between type 1 diabetic patients with microalbuminuria and type 1 diabetic patients with normoalbuminuria [65]. While most of the investigations related to markers for renal complications have been based on urinary samples, studies on blood samples have identified an alteration of miRNA levels in type 1 and type 2 diabetic patients with diabetic nephropathy [21, 37]. Thus, the findings to date suggest that circulating miRNAs could help to diagnose and/or to monitor diabetic renal complications.

Among the diabetic complications, retinopathy carries a heavy social and economic burden as it is the leading cause of adult blindness in the Western world. HbA1c is a biomarker currently used by clinicians as it has a strong predictive value for diabetic retinopathy [19]. However, some patients develop eye complications despite satisfactory glycaemic control. Thus, new biomarkers would be useful to improve the detection of these complications. Interestingly, a study revealed three serum miRNAs (miR-21, -181c, -1179) that are deregulated in patients with proliferative diabetic retinopathy vs those with non-proliferative retinopathy suggesting that circulating miRs could be non-invasive biomarkers for the detection and/or the follow-up of retinopathy [66]. Furthermore, in a small cohort, a higher level of several miRNAs related to angiogenesis (miR-15a, -320a, -320b, -93, -29a) was found in vitreous humour of patients with proliferative diabetic retinopathy compared with patients with macular hole, with some of these miRNAs also deregulated in the serum [67].

To the best of our knowledge, no reports concerning the relevance of circulating miRNAs in diabetic neuropathy have been published to date. As intracellular miRNAs appear to be deregulated in this diabetic complication, it would be worthwhile to investigate whether circulating miRNAs could be potential biomarkers of this disease [68, 69].

Current limits to the use of circulating miRNAs in clinical practice

Diversity of study design

While promising results have been gathered concerning the potential use of circulating miRNAs as biomarkers, the studies conducted to date need to be critically evaluated at the level of the patient populations analysed and the methodology used.

First, the investigations concern different diabetic populations, with some analysing type 1 patients and others analysing type 2 or GDM patients (Table 1). Only a few miRNAs appear to show a comparable change in levels in type 2 and type 1 diabetic patients compared with controls. For example, miR-29a is increased in the serum of both type 2 [22] and type 1 [35] diabetic patients. In contrast, several groups have reported a decrease in the plasma level of miR-126 in type 2 diabetes [27, 30, 31], but no difference or an increase in type 1 diabetes [36, 37]. Comparison of several studies reveals that the majority of the deregulated miRNAs in type 1 diabetes are different from those deregulated in type 2 diabetes. This suggests that there is a distinct pattern of circulating miRNAs in type 1 diabetes vs type 2 diabetes. Indeed, a study comparing miRNA profiles between type 1 diabetes, type 2 diabetes and GDM in peripheral blood mononuclear cells confirmed divergent profiles in the different forms of diabetes despite the fact that a few miRNAs are shared among the three groups [70].

A second limitation concerns the nature of the samples. Two types of biological fluid were generally investigated (blood and urine) in most studies, as these fluids are easily collected by non-invasive methods and provide a sufficient volume for analysis. Depending on the study, when blood samples were used, analyses were usually performed on plasma or serum, but rarely on whole blood or on blood cells (Table 1). Even though, according to the manufacturers, miRNA profiles obtained from serum and plasma should be comparable, one group of investigators reported higher concentrations of circulating miRNAs in serum than in plasma from the same individual, suggesting that some miRNAs are lost during plasma preparation [71]. In addition, blood cells had more detectable miRNAs species than either serum or plasma. Therefore, haemolysis should be carefully prevented when preparing plasma or serum.

An additional issue with confounding effects relates to the nature of the circulating miRNAs. Indeed, circulating miRNAs travel in different forms of packaging, including vesicle-like exosomes, microparticles or apoptotic bodies, or they are associated with proteins. It is possible to extract all the miRNAs in the sample or to specifically extract miRNAs associated with particular vesicles or proteins. The large majority of the studies extract all the miRNAs, whereas some investigate only those associated with exosomes [20, 34, 65]. However, the profile of extracellular miRNAs associated with vesicles is different from that of vesicle-free miRNAs [13]. Hence, it is important to take into account the packaging of circulating miRNAs as this will modify the observed pattern.

Another issue to consider is that RNA isolation and miRNA profiling with qRT-PCR are performed using different commercial kits having distinct specifications. The method of RNA isolation can affect the miRNA profile depending on the protocol chosen for either total RNA or specific miRNA purification [72]. The most commonly used miRNA profiling assays are from Applied Biosystems (Taqman technology) and Exiqon (LNA technology). Both approaches are designed to enhance qRT-PCR specificity and sensitivity. Taqman technology combines miRNA-specific RT and PCR primers containing a fluorophore and a quencher for the qPCR. Exiqon combines a universal RT with locked nucleotide acid (LNA)-enhanced PCR primers, a class of high-affinity RNA analogues in which the ribose ring is ‘locked’ in the ideal conformation for Watson–Crick binding. Examining the miRNA level from plasma and serum from four individuals, one study compared the miRNA profiles obtained with these technologies [71]. For 67 commonly detectable miRNAs in plasma and serum, a low correlation between the two qRT-PCR methods was found. A second study revealed that the two methods yielded significantly different copy number estimations of some miRNAs, even though the efficiency appears to be similar [73]. To the best of our knowledge, at present, no consensus has been reached regarding which assay should be preferentially used. Thus, depending on the technique applied, the results may vary and this should be taken into account when analysing the results.

Another issue linked to the miRNA profiling technology is that at present there is no well-established housekeeping RNA for data normalisation of circulating miRNAs, therefore variable references are utilised. Urine measurements apply exogenous controls such as miR-39, Uni SP3 or endogenous controls such as RNA U6, while studies based on blood utilise exogenous controls such as miR-39 or endogenous controls such as RNU 6b, RNU 48, RNA U6, miR-16, miR-19b, miR-30c, miR-103, miR-106a, miR-191, miR-146a, miR-221, miR-223, miR-238, miR-423-3p, miR-425, miR-454, or miRNAs belonging to the let-7 family. Although these miRNAs have been selected for their relative stability, the levels of some may vary in diabetic populations such as miR-let-7a [21, 34] and miR-let-7f [34], miR-let-7i [25], miR-191 [27, 38], miR-223 [27] or miR-146a [22, 29, 41]. Given these shortcomings, a consensus for data normalisation is urgently needed based, for example, on the geometric mean of a common set of miRNAs.

Because of the heterogeneity in study design, the comparison of the results of the different reports should take into account that divergent results may be explained by differences in the techniques applied and the populations analysed. Unfortunately, only a few studies can be compared on the basis of having similar diabetic populations, samples, RNA extraction methods and miRNAs investigated. Among the reports on the same circulating miRNAs, some show similar results. For example, miR-126 was found by several authors to be decreased in plasma from type 2 diabetic patients [27, 30, 31]. At the same time, observations made in some studies were not totally reproducible in others. Although it was reported that plasma levels of miR-15a and miR-223 were significantly different between type 2 diabetic patients and controls [27], this finding was not replicated by other investigators [30]. Remarkably, a meta-analysis including 38 studies in human and animal models comparing miRNA profiles between type 2 diabetics and non-diabetics identified 40 dysregulated miRNAs among different tissues, including blood. Of these, eight miRNAs emerged as potential circulating biomarkers of type 2 diabetes [74]. In summary, the investigated populations and the methodologies used in different reports suffer from important dissimilarities. These need to be standardised to allow the study of circulating miRNAs. Even though there are recent reports that seem to address this issue, to the best of our knowledge no global consensus exists at this time [72, 75, 76].

Potential confounding factors

Besides the impact of the methodology on the observations made, there are some recently revealed confounding factors that should be taken into account when interpreting miRNA profiling results (ESM Table 1). Several studies have identified age-related modifications in circulating miRNA levels [77, 78, 79, 80, 81, 82]. Notably, among these, miR-29b [27, 28, 70], miR-130b [31], miR-222 [31], miR-375 [22], miR-21 [26, 27, 36] and miR-126 [23, 27, 30, 31] were found to be deregulated in diabetic patients. These results underline that age could induce a bias, which has to be dealt with.

In the same way, sex should be taken in consideration, as indicated by a study of a cohort of 102 individuals with and without the metabolic syndrome where the profile of circulating miRNAs was found to be different only in women with the metabolic syndrome [83]. Another potential confounding factor is linked to ethnicity. Indeed, several reports revealed ethnicity-related plasma miRNA levels in diabetic populations [28, 84].

Lifestyle can also influence the circulating miRNA pattern. In small cohorts of patients, some studies illustrate the impact of nutrition such as zinc or 25-hydroxyvitamin D levels on circulating miRNAs [85, 86]. These observations suggest that nutrition can be a confounding factor, which is difficult to evaluate. Indeed, the long-term impact of diet on the circulating miRNA profile is cumbersome to consider in studies investigating miRNAs in disease states. Another lifestyle factor that can potentially affect the results is physical exercise. Several studies have reported that acute physical activity and prolonged exercise training programmes can modify the circulating miRNA profile [87]. Furthermore, smoking appears to alter the pattern of circulating miRNAs [88, 89, 90, 91], and of the miRNAs affected, miR-29b was identified to be deregulated in diabetic patients [27, 28, 70]. Environmental exposure to air pollutants and chemicals seems to modify circulating miRNAs [92], inducing a bias that is difficult to estimate.

Finally, the patterns of circulating miRNAs in diabetes can be changed by a wide range of associated conditions such as cancer [93], and cardiovascular, neurodegenerative, metabolic [72] and autoimmune diseases [94].

The potential confounding factors discussed here should urge clinical investigators to exercise caution when designing the controls, and they should be obliged to take these factors into account in their analyses. By using the appropriate questionnaires before the clinical study, most confounding factors, such as age, sex, ethnic origin, comorbidity, lifestyle (including smoking or physical exercise) can be revealed. Therefore, these factors should be ruled out using exclusion criteria or controlled for by stratifying the population or using multivariate analysis. However, the impact of other factors, such as nutrition or environmental exposure, appear to be much more challenging to evaluate or to detect, and would need further characterisation to reach a consensus for including them.

While the effects of confounding factors can, at least in part, be limited when running a clinical study, the question of the specificity of circulating miRNAs as a biomarker of diabetes remains. As demonstrated by previous studies, an miRNA pattern can be confounded by an associated disease, which is a current limit for the use of circulating miRNAs as biomarker for multi-disease patients [95]. Therefore, it might be useful not to focus on changes in the level of one particular miRNA, but rather to look for a disease-related signature of a series of miRNAs. Interestingly, novel miRNAs have been identified, many of which are human specific and tissue specific, offering a promising opportunity to identify such biomarkers [96]. Finally, while this is a burgeoning area of research, the recent discovery of the high tissue specificity of long non-coding RNAs [97] and their identification in biological fluids such as plasma could offer new perspectives [98].

Conclusions

While the field is still in its infancy, circulating miRNAs appear to offer various potential applications in medical practice as biomarkers of diabetes and associated complications. Different applications can be envisioned, such as prediction of the disease, detection and monitoring of its complications, evaluation of treatment efficacy and the characterisation of pathogenic traits such as insulin resistance for type 2 diabetes and the autoimmune process for type 1 diabetes. However, the observations from the currently available studies are difficult to compare because of the heterogeneity of the technical methods used and the small and heterogeneous cohorts investigated. As the impact of these shortcomings is not precisely evaluated and not considered in most studies, a top priority is to establish a consensus to standardise the methodology used and to improve reproducibility. While improvements in the standardisation of preanalytical and analytical methods are needed and additional studies on larger cohorts are required before circulating miRNAs can be introduced into medical practice, miRNAs appear to be bona fide biomarkers for diabetes in the future. Even if the measurement of circulating miRNAs currently has limitations, miRNAs are likely to become useful in the near future. Indeed, one can anticipate that they will complement existing biomarkers, especially in the context of the early identification of individuals at high risk to develop diabetes, and in the detection and the follow-up of diabetic complications.

More than being simple markers, which vary depending on physiological or pathological processes, recent observations suggest a functional role of circulating miRNAs. Indeed, circulating miRNAs contained in vesicles or associated with HDL can be transferred into recipient cells and thus could be involved in cell-to-cell communication [12, 14]. To date, a series of studies has investigated the role of circulating miRNAs as intercellular mediators in diabetic conditions, pointing to possible new mechanisms in diabetes pathogenesis [99, 100, 101]. It is clear that the precise potential of circulating miRNAs in complex and multifactorial diseases such as diabetes requires further confirmation both in terms of a biomarker and its involvement in disease mechanisms. A considerable amount of research efforts need to be invested to determine whether circulating miRNAs profiles are of use in routine hospital practice, but given the potential advantages, the challenge is undoubtedly worth it.

Acknowledgements

The authors would like to thank E. Corsetti (University Nice Sophia Antipolis, INSERM, U1081, CNRS, UMR7284, IRCAN, Nice, France) and F. Lareyre (University Nice Sophia Antipolis, Nice, France) for their constructive discussions of the manuscript.

Funding

The authors are supported by INSERM, Université Nice Sophia Antipolis, Conseil Régional PACA, Conseil Général des Alpes-Maritimes, Aviesan/AstraZeneca, ‘Diabetes and the vessel wall injury’ programme by the Agence Nationale de la Recherche (ANR) through ANR-RPV12004AAA, ‘Investments for the Future’ LABEX SIGNALIFE #ANR-11-LABX-0028-01 and European Foundation for the Study of Diabetes (EFSD/Lilly; European Diabetes Research Programme). We apologise for being unable to discuss all the relevant papers due to space limitations.

Duality of interest

The authors declare that there is no duality of interest associated with the manuscript.

Contribution statement

All authors confirm they contributed to the intellectual content of this article and approved the final version.

Supplementary material

125_2015_3680_MOESM1_ESM.pdf (38 kb)
ESM Table 1(PDF 37 kb)

Copyright information

© Springer-Verlag Berlin Heidelberg 2015

Authors and Affiliations

  • Juliette Raffort
    • 1
    • 2
    • 3
    • 4
  • Charlotte Hinault
    • 1
    • 2
    • 3
    • 4
  • Olivier Dumortier
    • 1
    • 2
    • 3
  • Emmanuel Van Obberghen
    • 1
    • 2
    • 3
    • 4
  1. 1.University of Nice-Sophia Antipolis, Medical SchoolNice Cedex 2France
  2. 2.INSERM, U1081, Institute for Research on Cancer and Aging of Nice (IRCAN), Aging and Diabetes TeamNiceFrance
  3. 3.CNRS, UMR7284, IRCANNiceFrance
  4. 4.Clinical Chemistry LaboratoryUniversity HospitalNiceFrance