Progression of Type 1 Diabetes: Circulating MicroRNA Expression Profiles Changes from Preclinical to Overt Disease

Objectives To evaluate the potential biological involvement of miRNA expression in the immune response and beta cell function in T1D. Methods We screened 377 serum miRNAs of 110 subjects divided into four groups: healthy individuals (control group) and patients at different stages of T1D progression, from the initial immunological manifestation presenting islet autoantibodies (AbP group) until partial and strong beta cell damage in the recent (recent T1D group) and long-term T1D, with 2 to 5 years of disease (T1D 2-5y group). Results The results revealed 69 differentially expressed miRNAs (DEMs) in relation to controls. Several miRNAs were correlated with islet autoantibodies (IA2A, GADA, and Znt8A), age, and C-peptide levels, mainly from AbP, and recent T1D groups pointing these miRNAs as relevant to T1D pathogenesis and progression. Several miRNAs were related to metabolic derangements, inflammatory pathways, and several other autoimmune diseases. Pathway analysis of putative DEM targets revealed an enrichment in pathways related to metabolic syndrome, inflammatory response, apoptosis and insulin signaling pathways, metabolic derangements, and decreased immunomodulation. One of the miRNAs' gene targets was DYRK2 (dual-specificity tyrosine-phosphorylation-regulated kinase 2), which is an autoantigen targeted by an antibody in T1D. ROC curve analysis showed hsa-miR-16 and hsa-miR-200a-3p with AUCs greater than for glucose levels, with discriminating power for T1D prediction greater than glucose levels. Conclusions/Interpretation. Our data suggests a potential influence of DEMs on disease progression from the initial autoimmune lesion up to severe beta cell dysfunction and the role of miRNAs hsa-miR-16 and hsa-miR-200a-3p as biomarkers of T1D progression.


Introduction
miRNAs are small noncoding RNAs functioning as posttranscriptional regulators of gene expression, affecting cell proliferation, differentiation, apoptosis, metabolism, and immunity. They can be released actively by cells or during tissue damage and have been used as biomarkers of destruction or regeneration of beta cells and of altered immunological activity, revealing mechanisms underlying the pathophysiology of type 1 diabetes [1][2][3]. A previous work from our group revealed the potential role of circulating miR-101 in its pathogenesis [4]. However, there are still contradictory data comparing miRNA profiles from individuals at different stages of diabetes, from preclinical to recent and long-duration T1D. Furthermore, the same miRNAs were shown to be up-or downregulated in the same phase of the disease and to be protective or at risk for diabetes. There is no clear definition about the effects of age, diabetes duration, and glucose levels, considering that metabolic derangements caused by glucolipotoxicity and inflammatory cytokines can change the miRNA milieu and interfere with the results [5][6][7]. Here we investigated whether there is a differential profile of serum miRNAs at different stages of T1D, which could suggest their participation in its pathogenesis considering these variables. We covered the phases of the highly active autoimmune process and those subject to glucolipotoxicity effects, e.g., from the first autoimmune manifestations (islet autoantibodies) without diabetes to recent and long-duration T1D. Additional information can come from the admixed populations of Brazil, expressing different frequencies of genetic markers of T1D [8].

Methods
The study was approved by the Ethical Committee of Hospital das Clinicas, Faculdade de Medicina, Universidade de São Paulo (Cappesq 11601), and followed the guidelines of the Declaration of Helsinki. Informed consent was obtained from all patients. Collected serum samples were stored at −80°C until use. Hemolyzed samples were excluded.

Experimental Design.
We screened 377 serum miRNAs of 110 subjects divided into four groups according to ADA criteria [9]: individuals with islet autoantibodies without diabetes (AbP group; n = 25), newly diagnosed patients with T1D with duration ≤6 months (recent T1D group; n = 30), patients with T1D with 2 to 5 years of duration (T1D 2-5y group; n = 26), and islet autoantibody negative healthy individuals (control group; n = 29). Exclusion criteria comprised other types of diabetes, use of medications except insulin, a febrile state within 10 days prior to blood collection, and individuals with liver, kidney, thyroid, and inflammatory/ autoimmune diseases.
Demographic characteristics, such as age, self-reported skin color, and sex were similar between groups (Table 1). Patients with T1D (both recent and lasting 2-5 years) were similar to each other and differed from the control group by higher values of glucose, HbA1c, islet autoantibody, and lower C-peptide levels (p < 0:05). AbP group presented intermediate characteristics (lower HbA1c and IA2A and higher C-peptide levels than both groups with T1D) and differed from the controls by higher IAA and GADA levels. HLA high-risk alleles for diabetes (DR3/DR4; DQ2/DQ8) were less frequent in the control group (p < 0:05).

Statistical
Analysis. The analysis of miRNAs expression used the Cloud program (Thermo Fisher Scientific, Waltham, MA, USA), an online data analysis software of the Cτ comparative method [10]. The threshold was manually aligned. The chosen criteria for the validation of qRT-PCR reactions were exponential and plateau amplification curves. miRNAs with Ct up to 35 were selected. The relative expression of miRNAs was obtained by the comparative method of Ct (2 -ΔΔCt ), using global normalization and Benjamini and Hochberg's false discovery rate method, and represented as fold change (FC) in relation to controls. Values >1 were considered increased and < 1 decreased. Variable distributions were verified by the Shapiro-Wilk normality test. Numerical variables with parametric and nonparametric distributions were analyzed by ANOVA and Kruskal-Wallis with Tukey's or Dunn's multiple comparisons posttest, respectively. Correlations were performed using the Spearman correlation test. Qualitative variables were compared using chi-square test or Fisher's exact test (statistical package GraphPad Prism, La Jolla, CA, USA). Data were considered significant at p < 0:05. Fisher exact test and the Benjamini and Hochberg's false discovery method were applied to obtain the target pathways in IPA analysis.
2.6. Results of miRNA Profiling. We performed miRNA profiling from 110 subjects divided into four groups (workflow in Figure 1). The number of miRNAs detected in each group sample was similar. The analysis of 135 miRNAs expressed in 20% or more of each group evidenced 69 DEMs across the recent T1D, AbP, and T1D 2-5y groups in comparison to control group. Figure 2 shows a volcano plot representation of the DEMs comparing each group versus control, represented as blue dots for downregulation and red dots for upregulation. The T1D 2-5y group had the highest number    Table 2).
We observed two different profiles. The 18 miRNAs consistently deregulated in AbP or recent T1D groups (13 of them also deregulated in T1D 2-5y) ( Table 2) comprising the cluster A (12 up-and 6 downregulated) were analyzed separately from the miRNAs deregulated only in the T1D 2-5y group (cluster B: 40 up-and 11 downregulated miR-NAs) ( Table 3 and Figure 2).
Fifty-nine (15.6%) of the 377 miRNAs evaluated were not expressed in any group (Suppl.1). Only one miRNA (reclassified as Vault RNA; ncRNA-886) was upregulated in all three groups.

Target Prediction and Pathway Analysis.
In both clusters, most pathways potentially regulated by DEMs were related to cancer, cell growth and metabolism (ErbB, MAPK, Wnt signaling pathway, and endocytosis), insulin production, and axon guidance/neurotrophin signaling pathways. Adherens junction and apoptosis pathways were enriched by DEMs' targets from the T1D 2-5y group.
The deregulated miRNAs were referred to several biological pathways by the miRWalk database, retrieving 206 and 140 pathways modulated by up-and downregulated miR-NAs in cluster A (Suppl.2) and 64 and 46 pathways in cluster B, respectively (Suppl. [3][4]. In both clusters, most pathways potentially controlled by DEMs were related to cancer, growth, metabolism (ErbB, MAPK, Wnt signaling pathway, and endocytosis), insulin, and axon guidance/neurotrophin signaling pathways. Adherens junction and apoptosis pathways were enriched by up-and downregulated DEMs' targets mainly from the T1D 2-5y group. The more frequent DEM gene targets identified by Target Scan tool are presented in Suppl.5-6.
Pathway enrichment analyses were performed using KEGG and IPA software. Figures 4 and 5 list the enriched pathways of the DEMs' targets from both cluster profiles using KEGG tool. Proliferative, metabolic, and immune responses were the top pathways identified for up-and downregulated miRNAs. Pathways related to survival (autophagy) and self-renewal capacity (regulating AbP group (individuals without diabetes expressing islet autoantibody), recent T1D group (newly diagnosed patients with type 1 diabetes with duration ≤6 months), T1D 2-5y group (patients with type 1 diabetes with 2 to 5 years of duration), and health control group, Rq: relative expression represented by fold change in comparison to controls, p value: corrected p value. Targeting and pathway analysis using IPA software identified highly predicted and/or experimentally observed targeting information from 10 DEMs (out of 16) in the AbP group with 359 targets: 1 DEM (out of 4) in the recent T1D group with 198 targets and 45 DEMs (out of 64) in the T1D 2-5y group with 1033 targets.

Journal of Immunology Research
After excluding cancer-related pathways, we identified a total of 366, 320, and 425 enriched canonical pathways related to DEMs' targets from the AbP, recent T1D, and T1D 2-5y groups, respectively (Suppl.7). The top forty canonical pathways most enriched with DEMs' targets identified by IPA analysis for each one of the three groups com-pared to control were the PTEN (phosphatase and tensin homolog deleted from chromosome ten), aryl hydrocarbon receptor, STAT3 (signal transducer and activator of transcription 3), epithelial-mesenchymal transition, and senescence pathways ( Figure 6). Pathways related to cell proliferation and immune response, e.g., cyclins/cell cycle regulation, and Interleukins IL-7 and IL-8, were overrepresented in the AbP and recent T1D groups. NF-KB (nuclear factor kappa B), IL-6, IL-10, acute phase response, and glucocorticoid receptor signaling pathways were overrepresented in the T1D 2-5y group. miRNAs from all groups were associated with a great number of pathways related to growth (IGF-1, FGF, HGF, EGF, ErbB, JAK/STAT, and PI3K/AKT signaling) and cell division (cyclins/cell cycle regulation). Inflammatory and defense pathways (ILK, neuregulin, dendritic cell maturation, toll-like receptor signaling, and Th1/Th2 activation pathways) were modulated.  Apoptosis, virus response, and neuronal development/repair (neuroinflammation, PEDF, neuregulin, and axonal guidance signaling pathways) were also affected. In general, the DEMs from the T1D 2-5y group modulated most genes from all pathways, with a greater magnitude than the other groups, many of them expressing an innate and adaptive immune responses.

Discussion
We identified miRNA dysregulation during T1D evolution. By comparing patients at different stages of disease progression with healthy individuals, from the initial immunological manifestation (one to three autoantibodies in AbP group) until partial and strong beta cell damage in the recent T1D and T1D 2-5y groups, we have found 69 differentially expressed miRNAs. These miRNAs have predicted targets related to immune regulation, metabolism, glucose homeostasis, cell proliferative/survival mechanisms, and beta cell function. Dysregulated expression of miRNAs pointed to possible mechanisms underlying the pathophysiology of T1D. The AbP group could also provide the effects of miRNA dysregulation and pathways predicted to be activated during the onset of islet autoimmunity that are unre-lated to blood glucose levels. The most enriched pathways potentially regulated by DEMs were related to immune cell activation, inflammation, and apoptosis.
The miR-200a-3p was the most highly expressed miRNA in AbP and recent T1D groups positively correlated to IA2A and GADA levels, suggesting a robust association with T1D pathogenesis. miR-200a was previously reported to be highly expressed in beta cells [11], being associated with their damage and apoptosis in vitro [12]. We hypothesize that its decrease to values similar to controls in the T1D 2-5y group could be linked to the scarcity of beta cells in this group or to hyperglycemia, although its expression was not correlated with glucose or HbA1c levels. Mechanistically, it was described as an antiapoptotic and stress-resistance miRNA, with targets that include the beta cell chaperone Dnajc3 and the caspase inhibitor XIAP and positively controls the activation of the tumor suppressor Trp53 [13].
Other important target of this miRNA is the thioredoxin-interacting protein (TXNIP). Its proapoptotic and diabetogenic function prevents beta cell function via induction of miR-200a in vitro. The TXNIP/miR-200/Zeb1/ E-cadherin signaling pathway links miR-200 to beta cell apoptosis and diabetes and links TXNIP to inhibition of epithelial-mesenchymal transition (EMT), a process involved in beta cell expansion [12], predicting its decline [14]. Catabolic and oxidizing reactions related to NAD generation, antioxidant mechanisms, and synthesis of nucleotides, suggested by canonical pathways associated with miR-200a-3p, are probably related to cell lesion and repair (Figure 7).
Besides miR-200a-3p, other upregulated miRNAs (miR-181a and miR-323) were also described as inhibitors of EMT [14], which could worsen beta cell function precociously, in the phase of ongoing autoimmunity (AbP) or in recent T1D group. In accordance, the EMT pathway was an enriched pathway in all three groups ( Figure 6).
Among cluster A downregulated miRNAs, miR-16-5p was the unique miRNA downregulated in all three groups in comparison to controls. It seems to favor Treg induction [22,25], which is highly expressed in beta cells, and negatively regulates the protein Ptch1 (protein patched homolog 1) involved in the inhibition of beta cell proliferation [30]. miR-16 seems to protect from high glucose-induced pancre-atic beta cell apoptosis by targeting CXCL10 [31] and from immune aggression as it was negatively correlated with ZnT8A levels in our cohort (r = −0:358; p = 0:044). The most enriched pathways for miR-16 predicted targets were related to cell cycle regulation and division, checkpoint control, and DNA repair (Figure 7).
The insulin resistance due to autoimmune aggression and the release of inflammatory cytokines might function decreasing the levels and the protective role of miR-16 [32]. miR-16 was downregulated already in the preclinical phase of diabetes, whereas hyperglycemia, which was correlated negatively with miR-16 levels (r = −0:289, p = 0:003), probably influenced its great decline in the T1D 2-5y group. Other five miRNAs downregulated only in the AbP group are involved in immune regulation (miR-195-5p and miR-590-5p) [33], cell proliferation, insulin transcription (miR-376a-3p and miR-19a-3p) [34,35], and residual beta cell function: miR-25 [7]. Their downregulation could potentially prevent beta cell recovery and inflammation resolution and is mirroring the autoimmune aggression in progress, unrelated to glucose levels. Therefore, parts of upregulated genes, including those in the AbP group, reveal a profile toward inflammation, apoptosis, and commitment of beta cell and insulin secretion,   evidencing cluster A DEMs as relevant to autoantibody development, decreased C-peptide levels, T1D pathogenesis, and progression. Few miRNAs were associated with immunomodulation and anti-inflammatory effects. The expression of most miRNAs of the cluster A followed a similar pattern in the three groups, although not always with statistical significance. The worsening of beta cell function, dysregulation of glucose levels, and insulin resistance probably accounted for the progressive upregulated trend of the expression of these miRNAs toward the T1D 2-5y group, particularly for miR-518b and miR-323a-3p, interfering negatively with beta cell expansion [14] and insulin secretion [36]. The exceptions were miR-200a-3p and miR-25-3p, associated with islet autoantibodies, in which the response of the T1D 2-5y group was opposite of the AbP and recent T1D groups, supporting its role in immunological aggression.
Thus, the altered expression of miRNAs from cluster A may be implicated in the cycle of damage regeneration of beta cells, inflammation, and metabolic disorders, contemplating insulin resistance and later, hyperglycemia. Cluster B (composed of DEMs expressed only in the T1D 2-5y group; n = 51) conferred similar results. Most upregulated miRNAs were present in human pancreatic beta cells [11].
Nc886 (pre-miR-886), upregulated in the three groups, seems to activate the protein kinase RNA-activated; an interferon-inducible kinase maybe related to defense against viruses [39].
The analysis using miRWalk and KEGG databases reinforced the relevance of both clusters in the autoimmune process. The potential biological pathways enriched by DEMs targeted oncogenic/proliferative and metabolic pathways, processes related to cell differentiation, migration, survival, apoptosis, neural development, insulin signaling, and immune system pathways. Most of these pathways are essential for proper lymphocyte development and function, and their enrichment, which was greater in cluster A, may be a reflection of the autoimmune activity.
Interestingly, these pathways were enriched by up-and downregulated miRNAs from both clusters, although through different targets.      Pathway enrichment analyses were performed using KEGG and IPA software. Figures 4 and 5 list the enriched pathways of the DEMs' targets from both cluster profiles using KEGG tool. Proliferative, metabolic, and immune responses were the top pathways identified for up-and downregulated miRNAs. Pathways related to survival (autophagy in cluster A) and self-renewal capacity (regulating pluripotency of stem cells in cluster B) were less modulated by downregulated miR-NAs, perhaps in an attempt to contain the self-aggression.
Genes related to B and T cell differentiation/function were targeted by upregulated miRNAs in cluster A (HIPK1) and B (NAA50, NFAT5, OTUB7B) and by downregulated miRNAs in clusters A (BRWD1) and B (ABL2, FOSL2, and PDE7A).
One of the most significantly enriched canonical pathways in the three groups ( Figure 6), pathogenetically relevant in T1D, was the senescence pathway, a state of cellular arrest associated with inflammatory cytokines, growth factors, and matrix metalloproteinases triggered by several damaging factors present in the three groups (like immune attack and metabolic derangements). In the same direction, the STAT3 pathway, regulating Th17 cell differentiation and suppressing Treg generation [41], and the PTEN pathway, involved in T helper follicular cell precursor induction, autoantibody generation and triggering of islet autoimmunity [42,43], and the regulation of EMT [14], all have their role in inflammation and reduction of beta cell function/survival. The IL-7 and IL-8 inflammatory signaling pathways were highly represented in the AbP and recent T1D groups whereas IL-6, NF-KB, and acute phase response pathways in the T1D 2-5y group.
On the opposite direction, there was the aryl hydrocarbon receptor pathway, enriched in all three groups. It modulates the development and functionality of immune cells and suppresses the expression of inflammatory cytokines during diabetes development [44]. DEMs from the T1D 2-5y group also target anti-inflammatory pathways like IL-10 A and glucocorticoid receptor signaling.
In general, DEMs from all groups target mechanisms related to growth, cell cycle regulation, apoptosis, inflammation, defense, and neuronal pathways, whereas few of them favored defense against an autoimmune offensive. The commitment of these pathways increased progressively from AbP to the recent T1D and then to T1D 2-5y group, which regulated the largest number of targets in most pathways, many of them expressing innate and adaptive immune responses. The metabolic dysregulation probably influenced these results.
Of note, similar pathways were enriched by both up-and downregulated miRNAs, triggered probably by different causal hits. Other possibility is the intense metabolic derangement due to glucose/lipotoxicity and inflammatory cytokines, eliciting miRNAs from inflammatory pathways. This can also explain some inconsistent results in the literature, where the same miRNA predisposed or protected from T1D or other autoimmune diseases. Diabetes duration, age of patients, and glucose levels could have acted in miRNA deregulation during disease progression and taken part in these inconsistencies, pointing to the importance of these variables in disease determinations. This can be further evidenced when considering the 69 miRNAs from clusters A to B, where 26 DEMs correlated with glycemic status, 19 with autoantibodies levels and ongoing islet autoimmunity, and 20 miRNAs correlated negatively with age. T1D duration influenced positively and negatively the levels of 19 miRNAs and their association with autoantibody titers, which are known to decrease with time. However, these results should be considered carefully, considering the sample size and low serum RNA yield.
Although the serum miRNA profile might not mirror the situation in the affected pancreas, it is notable that most of the DEMs of our cohort are miRNAs enriched in beta cells, suggesting they may correlate with the severity of beta cell injury. Ongoing destruction of beta cells could result in diffusion of islet-enriched miRNAs into serum, as observed during the autoimmune attack in animal models of diabetes [45]. Furthermore, the negative correlation of several upregulated miRNAs expression with C-peptide levels as well as with age, considering that the immune attack is usually more intense in the youngest, reinforces our results. Even more, previous reports of the participation of most miRNAs in the pathogenesis of T1D and other autoimmune manifestations also point to their role in the autoimmune process. Their expression in the different phases of the autoimmune lesion, starting with positivity to islet autoantibodies up to severe beta cell dysfunction, evidencing the influence of age, duration of diabetes, and glycemic control on their expression brings relevant information and suggests new immunological and metabolic influences.
In conclusion, our data suggested the potential role of miRNAs favoring the preponderance of pathways compromising beta cell function throughout diabetes progression like increased apoptosis, inhibition of EMT, impaired TREG function, inflammatory pathways like STAT3 and PTEN, and senescence probably being indirect signs of islet autoimmunity and metabolic derangements due to gluco-lipid toxicity. The higher discriminating power for T1D prediction for miR-200a-3p and miR-16-5p, distinguishing patients from the different stages of T1D, suggested both miRNAs as potential biomarkers and targets for beta cell recovery.

Data Availability
All data generated or analyzed during this study are included in this published article (and its supplementary information files, Supplementary Tables 1 to 7). 13 Journal of Immunology Research

Disclosure
The funders had no role in the design of the study; in the collection, analyses, or interpretation of the data; in the writing of the manuscript; or in the decision to publish the results.

Conflicts of Interest
The authors declare no conflict of interest.