MicroRNAs Associated with Metformin Treatment in the Diabetes Prevention Program

The Diabetes Prevention Program (DPP) randomized controlled trial demonstrated that metformin treatment reduced progression to type 2 diabetes (T2D) by 31% compared to placebo in adults with prediabetes. Circulating micro-ribonucleic acids (miRs) are promising biomarkers of T2D risk, but little is known about their associations with metformin regimens for T2D risk reduction. We compared the change in 24 circulating miRs from baseline to 2 years in a subset from DPP metformin intervention (n = 50) and placebo (n = 50) groups using Wilcoxon signed rank tests. Spearman correlations were used to evaluate associations between miR change and baseline clinical characteristics. Multiple linear regression was used to adjust for covariates. The sample was 73% female, 17% Black, 13% Hispanic, and 50 ± 11 years. Participants were obese, normotensive, prediabetic, and dyslipidemic. Change in 12 miR levels from baseline to 2 years was significantly different in the metformin group compared with placebo after adjusting for multiple comparisons: six (let-7c-5p, miR-151a-3p, miR-17-5p, miR-20b-5p, miR-29b-3p, and miR-93-5p) were significantly upregulated and six (miR-130b-3p, miR-22-3p, miR-222-3p, miR-320a-3p, miR-320c, miR-92a-3p) were significantly downregulated in the metformin group. These miRs help to explain how metformin is linked to T2D risk reduction, which may lead to novel biomarkers, therapeutics, and precision health strategies.


Predictors of miR Change after 2 Years
For the 12 miRs with adjusted p values less than 0.05, a multiple regression was applied to account for the in uence of covariates (i.e., age, sex, race and ethnicity, and BMI) on mean change (Table 3).For seven of the adjusted miR models (i.e., let-7c-5p, miR-151a-3p, miR-222-3p, miR-29b-3p, miR-320a-3p, miR-320c, miR-92a-3p), treatment group was the only signi cant independent predictor of miR level.Age was a signi cant independent predictor of miR-130b-3p.Self-identi ed female gender was a signi cant independent predictor of miR-22-3p.Black race was a signi cant independent predictor of both miR-17-5p and miR-22-3p mean change and was of borderline signi cance for miR-20b-5p (adjusted p = 0.056).The "other race and ethnicity" category was a signi cant predictor of both miR-20b-5p and miR-93-5p z-scores in the regression model.

Correlations between Baseline Clinical Characteristics and Changes in MiRs
Signi cant correlations between baseline clinical characteristics and miR change in z-scores are presented in Fig. 2. Baseline fasting blood glucose was inversely associated with mean change in let-7c-5p.Baseline fasting triglycerides were positively correlated with miR-130b-3p and miR-151a-3p.Total cholesterol was positively correlated with miR-20b-5p and negatively correlated with miR-22-3p.High density lipoprotein (HDL) cholesterol was inversely associated with miR-151a-3p, while low density lipoprotein (LDL) cholesterol was positively associated with miR-20b-5p.Systolic blood pressure and hip circumference were both inversely associated with miR-130b-3p, and diastolic blood pressure was inversely associated with miR-222-3p expression.
Table 4 summarizes what is known about the miRs, their response to metformin in our study, signi cant independent predictors or covariates in the multiple regression models, associations with clinical characteristics at baseline, gene targets, and mechanisms of action related to T2D risk.

Discussion
We compared mean change in standardized circulating miR expression levels after two years of metformin therapy versus placebo in adult participants in the Diabetes Prevention Program randomized controlled trial.The change in levels of twelve miRs was signi cantly different between treatment groups, six of which were upregulated and six of which were downregulated.Eight of these miRs are novel and have not previously been described in relation to metformin response (i.e., miR-151a-3p, miR-17-5p, miR-20b-5p, miR-22-3p, miR-320a-3p, miR-320c, miR-92a-3p, and miR-93-5p).Findings from the other four miRs -let7c-5p, miR-130b-3p, miR-222-3p, and miR-29b-3p -are aligned with previously reported literature about metformin response.
Collectively, what is known about this set of miRs to-date suggests that metformin therapy acts via myriad functions, organs, tissues, and biological systems related to the progression from an at-risk state to T2D disease onset, with or without development of related comorbidities.Differentially expressed miRs may regulate neuroin ammation, cellular senescence and aging, glucose and lipid metabolism, insulin signaling pathways, vasculitis and thrombogenesis, and immune function.Though the science is still developing, the ndings from our study point to subgroups who may be of highest risk or receive the best therapeutic bene t from metformin therapy.These miRs provide mechanistic clues for optimal metformin treatment response, advancing our knowledge about precision prevention for T2D.Their diversity underscores the complexity of T2D prevention.

let-7c-5p
In our study, we observed downregulation of let-7c-5p in the metformin group after two years compared with controls.This observation aligns with previous ndings about members of the let-7-5p family of miRNAs, such as let-7b-5p, let-7e-5p, and let-7f-5p, which were also reported to be downregulated after 3-12 months of metformin use.[14,16,[19][20] After adjusting for covariates, treatment group was the only signi cant predictor of let-7c-5p change after 2 years of therapy, however, the mean change was inversely correlated with baseline fasting blood glucose in our participants.This nding suggests that those with the highest blood glucose levels when they begin metformin therapy may see the greatest reduction in let-7c-5p levels over time.
Metformin's bene cial effect may in part be due to let-7c-5p's role in neuroin ammation.[22] A prior study of let-7c-5p in individuals with T2D and mild cognitive impairment (MCI) found elevated exosomal let-7c-5p levels in these people with T2D compared to those with T2D but without MCI.Zhang et al.
Let-7c has been linked to both metformin response and senescence and cellular aging.[25] In at least one mouse model study, metformin decreased cellular senescence by increasing DICER1 protein, lowering P16 and P21 protein levels, decreasing the abundance of in ammatory cytokines and oncogenes that are hallmarks of the senescence-associated secretory phenotype (SASP).[25] Let-7c, along with miR-130b-3p, are two of the miRs identi ed in our study that are considered regulators of cellular senescence and in uenced by metformin therapy.[25][26] miR-130b-3p miR-130b-3p was signi cantly downregulated in the metformin therapy group in our study of at-risk individuals after 2 years.This nding was aligned with previous work by Demirsoy et al. (2018) in which miR-130b-3p was downregulated after 3 months of metformin therapy in people who had already developed T2D.[19] In another study of the participants of indigenous Native American descent, miR-130b-3p levels were upregulated in an increasing pattern based on glucose tolerance -from normal glucose tolerance (lowest miR-130b-5p levels) to impaired glucose tolerance to T2D (highest miR-130b-3p levels).[27] Age was a signi cant independent predictor of miR-130b-3p change, with higher levels associated with older age in years.miR-130b-3p was positively correlated with baseline triglycerides and negatively correlated with baseline systolic blood pressure and hip circumference.Prabu et al. (2015) found links between miR-130b-3p and dyslipidemia in individuals who reported Asian Indian race group with impaired glucose tolerance versus with T2D.[27] Overexpression of miR-130b-3p has been linked to both diet-induced mouse models of diabetes [27] and high fat diets in mice.
[28] We found that triglyceride levels at baseline were signi cantly positively correlated with change in miR-130b-3p after 2 years.Overexpression of miR-130b-3p has been linked to increased triglyceride secretion into the bloodstream.[29] It may be that those individuals with higher levels of triglycerides at baseline are less likely to experience a reduction in miR-130b-3p with metformin therapy.
Elevated miR-130b-3p expression was associated with reduced intracellular pancreatic beta cell ATP levels in both rats and humans with T2D.[30] MiR-130b-3p is also thought to be a regulator of lipid homeostasis.[29] Overexpression of miR-130b in adipose tissue was associated with changed expression of critical genes implicated in lipid metabolism, with microsomal triglyceride transfer protein mRNA and protein expression levels increasing signi cantly in response to increased miR-130b-3p.[29] Ofori et al. (2017) demonstrated negative regulatory effects of miR-130b-3p on pyruvate dehydrogenase E1 alpha (PDHA1) and on glucokinase (GCK) proteins, which are both involved in ATP production.[30] In that study, overexpression of miR-130b-3p in the insulin-secreting INS-1 832/13 cell line resulted in altered dynamics of intracellular ATP/ADP ratio ultimately perturbing fundamental ATP-requiring beta cell processes such as glucose-stimulated insulin secretion, insulin biosynthesis and processing.[30] miR-151a-3p MiR-151a-3p was upregulated in the metformin group after two years in our sample.To the best of our knowledge, no prior studies have reported on the relationship between miR-151a-3p and metformin therapy or its role in T2D.It may be related to atherosclerosis and cardiovascular comorbidity associated with T2D, however, because miR-151 may inhibit the apoptosis of endothelial cells in atherosclerosis and target IL-17A.[31] This may be characterizing complications from, rather than the etiology of, elevated fasting blood glucose and prediabetes.In our sample, miR-151a-3p was positively correlated with baseline fasting triglycerides and inversely correlated with baseline HDL, so its function may be related to cholesterol, but further investigation is needed.

miR-17-5p
MiR-17-5p is involved in cell proliferation, in ammation, mitochondrial function, and diabetes-related vascular damage.[32] Altered expression of miR-17-5p has previously been associated with T2D and gestational diabetes, obesity, and non-alcoholic fatty liver disease (NAFLD), [32][33] but it has not yet been reported in relation to metformin therapy for these diseases.miR-17-5p was upregulated in our metformin sample compared with placebo.
The direction of change (i.e., upregulated vs downregulated) in previous studies has been mixed depending on the comparison group, the tissue type, and the speci c disease pathology being studied.MiR-17-5p has been found to be upregulated in the blood of people with both T2D and NAFLD compared to people without NAFLD.[34] Upregulation of miR-17-5p also may be linked to the regulation of smooth muscle cell proliferation in the development of cardiovascular complications of T2D.It may play a role in the link between adipose tissue dysfunction and the development of obesity associated disorders including T2D because upregulation has been signi cantly related to lower visceral fat mass, lower circulating parameters of chronic glycemia, and improved insulin sensitivity.[35] Black race was a signi cant independent predictor of miR-17-5p levels after adjusting for covariates in our sample.In a longitudinal study of metformin adherence and bene ts in the overall DPP trial, Black race was associated with a higher incidence of T2D and was associated with lower metformin regimen adherence in the trial.
[36] However, in the present study, participants who identi ed as Black race were not overrepresented in the subset who had developed T2D in either group.Adherence was highest in participants in our sample who identi ed as Black race and lowest in participants who identi ed as Hispanic ethnicity.While miR-17-5p was not assessed, we previously reported differences in other miRs between race and ethnic groups included in the DPP trial.[37] Further studies are needed to validate these differences and determine if they provide mechanistic insights between the construct of race and ethnicity and risk for T2D.

miR-20b-5p
MiR-20b-5p was upregulated in the metformin group in our study after 2 years.In adults with T2D, miR-20b-5p has been found to be upregulated in exosomes of T2D patients, which can be transferred to the cells of the vascular endothelium, where it targets wingless-related integration site family member 9b (wnt9b) signaling to inhibit wound healing and angiogenesis.
[38] In a study of four men with T2D, miR-20b-5p was upregulated compared to age and sex matched men with normal glucose tolerance, but not signi cantly upregulated compared with the men in their study with prediabetes and at high risk for developing

T2D. [38]
Black race was borderline signi cant (p = 0.056) independent predictor of miR-20b-5p change.Like miR-122, miR-20b-5p has been associated with NAFLD in people with T2D.[34] MiRMiR-20b-5p in skeletal muscle may impair glucose metabolism, downregulate gene sets associated with immune function (including interferon-α and interferon-γ response, tumor necrosis factor-α, IL-2-STAT5, and IL6-janus kinase (JAK)-STAT3 signaling pathways), and upregulate gene sets associated with cholesterol homeostasis, fatty acid metabolism, and mammalian target of rapamycin (mTOR) signaling pathways.[39] Overexpression of miR-20b-5p in human skeletal muscle cells suppressed expression of Ak strain transforming interacting protein (AKTIP), signal transducer and activator of transcription 3 (STAT3), and glycogen synthase and impaired insulin signaling [39] miR-22-3p One of the most abundant miRs in the liver, miR-22-3p is involved in the regulation of lipid and glucose metabolism with differential effects in speci c organs.[40][41][42] Its role in T2D may be complex because both overexpression and underexpression have been linked to pathophysiological effects on T2D incidence, in ammation, obesity, NAFLD, and liver steatosis.[40][41][42] In the present study, miR-22-3p was signi cantly downregulated in metformin group after 2 years compared to placebo.MiR-22-3p has not previously been described in relation to metformin therapy, but it has previously been studied in people who responded to thiazolidinedione treatment for insulin resistance compared to those who did not respond (Flowers et al., 2015).[43] In that study by Flowers et al. (2015), miR-22-3p expression levels were higher at baseline in insulin resistant individuals compared to insulin sensitive participants, and higher at baseline in those who responded to the thiazolidinedione treatment compared to those who did not respond, though they did not compare the change in miR over time after therapy.[43] Both female gender and Black race were signi cant, independent predictors of miR-22-3p change after 2 years in our study.Sex-based differences in gene expression have been observed in both healthy mice and mice with T2D, [44] but they did not report on known targets of miR-22-3p.Research in both human and mouse models suggests that expression of miRs in males and females may be either consistently, differentially, or oppositely expressed in speci c diseases (i.e., one sex is upregulated and the other sex is downregulated).[32,45] More work is needed to understand sex-biased expression of the miRs in patients with T2D, and to differentiate the effects of biological sex from self-reported gender related to miR expression and risk for T2D.
A previous study of monocytes from individuals from varying race and ethnic groups found miR-22-3p to be inversely correlated with HDL cholesterol levels and positively associated with BMI, but those associations were not observed in our sample.[41] Gene targets of miR-22-3p are associated with cholesterol metabolism, such as stearoyl-CoA desaturase gene (SCD) which encodes stearoyl-CoA desaturase.[41,46] miR-222-3p MiR-222-3p was downregulated in the metformin group compared with placebo after 2 years, and the miR change was inversely associated with baseline diastolic blood pressure in our study.MiRMiR-222-3p has previously been described as downregulated in response to metformin therapy [17][18], and it was found to play a role in the therapeutic effects of the T2D drug pioglitazone in a mouse model, acting on skeletal muscle tissue through a mechanism unrelated to the drug's known PPARγ action.[32,47] Mechanistic investigations have pointed to the involvement of miR-222-3p levels in T2D, supported by its targeting of genes like O-6-Methylguanine-DNA Methyltransferase (MGMT), protein phosphatase 2 regulatory subunit B-α isoform (PPP2R2A), and Reversion Inducing Cysteine Rich Protein With Kazal Motifs (RECK) associated with T2D pathology.[32,[48][49] miR-29b-3p Circulating miR-29b has been observed to be upregulated in adults with T2D and may be a potential prognostic marker of diabetes risk.[50] Its role in T2D may be in mediating brosis, human insulin gene transcription, and circadian rhythms in β cells by targeting the period circadian regulator 3 (Per3) gene.[50] MiR-29b-3p was signi cantly upregulated after 2 years of metformin treatment in our sample of at-risk individuals.
Metformin, an endogenous competing RNA of H19 Imprinted Maternally Expressed Transcript (H19) gene, reduced expression of H19 in animal models.[51] Both matrix metalloproteinase (MMP)-2 and MMP-9 are possible targets of miR-29b-3p.Chen et al. ( 2019) speculated that metformin may also reduce the expression of MMPs via the H19/miR-29b-3p signaling pathways.[51] miR-320a-3p MiR-320a-3p has not previously been described with metformin therapy.MiR-320a-3p was downregulated in the metformin group compared with placebo in our sample.In another pharmacologic study, miR-320a-3p was increased in people with insulin resistance verse insulin sensitive, and this miR was inversely associated with response to thiazolidinediones, perhaps based on the overall positive association with insulin resistance and greater room for a therapeutic response.[43] There is some evidence for a role in T2D-related complications (i.e., miR-320a-3p has been associated with diabetic retinopathy, [52] cardiotoxicity, [53] and differences based on social determinants (i.e., miR-320a was inversely related to glycemic impairment in South Asians.[54] miR-320c MiR-320c not previously described with metformin therapy.It was signi cantly upregulated in the metformin group after 2 years compared with placebo in our sample.MiR-320c has been studied for its role in diabetic nephropathy pathogenesis due to its actions on the PI3K/AKT pathway.[55] PTEN is a gene target of miR-320c.Sun et al. (2022) observed that the function of miR-320c was reversed by down-regulation of PTEN, and that inhibition of miR-320c could alleviate toxicity of HK-2 cells induced by hyperglycemia.[55] It has also been linked to β cell damage in mouse models of type 1 diabetes.[44] miR-92a-3p MiRMiR-92a-3p was downregulated in the metformin group versus placebo, and our study is describing miR-92a-3p for the rst time in relation to metformin response.Upregulated in brous plaques of early-stage atherosclerotic lesions.
[56] Like miR-320a-3p, miR-92a-3p has been associated with diabetic retinopathy.[52] Signi cantly downregulated in extracellular vesicles in response to pioglitazone treatment relative to placebo in people with T2D.[57] These results were unchanged after inclusion of metformin as a cofactor in their analysis model.[57] miR-93-5p In previous studies of patients with T2D, miR-93-5p expression levels were signi cantly lower than in healthy controls.[58] In the present study, after 2 years of metformin therapy, participants in the metformin group had signi cantly upregulated miR-93-5p compared to placebo.This change may re ect a normalization effect in response to the metformin for this miR.
MiR-93 may negatively regulate glucose transporter 4 (GLUT4) which is a dominant regulator of whole-body glucose homeostasis.[59] Like let-7c, miR-93 may also regulate ERBB4 and NRG4 gene expression.[24] ERBB4 is involved in carbohydrate and lipid metabolism pathways while NRG4 is associated with cardiac pathology.[24] Particularly for miR-93-5p, the presence of single nucleotide polymorphisms (SNPs) in the seed region may affect the interaction between the two genes, so the presence of variants could lead to either gain or loss of function and should be studied further as a potential therapeutic target.[24] Limitations and Future Directions This sample was limited to 100 randomly selected participants from the overall DPP study with miR expression levels from baseline and 2 years available.
The ndings should be validated with prospective samples large enough to support detailed subgroup analyses to fully understand the best candidates for metformin prevention therapy.
Minimal evidence is available about how miRs change in response to metformin therapy of varying durations.Further repeated measures, longitudinal studies are needed to optimize treatment response and to fully understand the mechanisms that support prevention of T2D occurrence after metformin therapy.Our study does not have a measure of when normal glucose was achieved and how that might affect miR expression levels in the metformin group.Future studies may also consider comparing miR expression levels in relation to varying metformin doses or regimens, monitoring time to achieve normal fasting blood glucose or hemoglobin A1C levels.
Our study did not investigate the effect of time on miR expression levels in both placebo and treatment group, and the role that metformin has, if any, on inhibiting cellular aging.Some literature suggests that there are anti-aging bene ts of metformin in healthy individuals, but the available evidence suggests that its effects on lifespan primarily result from its anti-hyperglycemic action, enhancing insulin sensitivity, reduction of oxidative stress and protective effects on the endothelium and vascular function. [60] As miR expression re ects how an individual is adapting to its environment, data about social determinants and behavioral and environmental factors would help to provide context to the biological ndings.This is particularly important to understanding differences in miR expression by self-identi ed race, as that re ects a social construct as opposed to a biological construct.[37] As seen in this study and others, [61] lack of controlling for genetic ancestry, discretely from the social constructs of race and ethnicity, may be one reason for the lack of validation of relationships between miRs and risk for T2D.[37,61] Methods The study was a secondary analysis of a subset of participants (N = 100) from the metformin and placebo arms of the DPP who agreed to additional longitudinal blood collection.All participants in the DPP provided informed consent and all experiments were conducted in alignment with the Declaration of Helsinki.Institutional review board (IRB) approval for the secondary analysis was provided by the University of California, San Francisco IRB.Participant data and biospecimens from the baseline and enrollment visits and the year-2 study visit were included.

Diabetes Prevention Program
The Diabetes Prevention Program was a longitudinal, three-arm randomized controlled trial study that compared an intensive lifestyle intervention, metformin, and placebo.[2,62] The study design, inclusion and exclusion criteria, and sample recruitment and characteristics have been described previously.
[63] For the metformin group, metformin therapy was administered orally at 850 mg per dose taken once per day for the rst month and then increased to 850 mg orally twice per day for the remainder of the study.Exceptions were made if gastrointestinal symptoms interfered with tolerance, in which case the drug was titrated over a longer period than one month.Placebo was given once per day for the placebo group for the rst month, and then increased to twice per day for the duration of the study.Adherence to the medication regimen for both placebo and the metformin groups was assessed quarterly with pill counts and structured interviews.The proportion of participants who took at least 80 percent of the prescribed dose of the study medication was 77 percent in the placebo group and 72 percent in the metformin group over the entire study period (p < 0.001).
The study described in this article used existing demographic (i.e., age, sex, race and ethnicity, body mass index (BMI)) and baseline clinical data (i.e., total cholesterol, triglycerides, low density lipoprotein cholesterol (LDL), high density lipoprotein cholesterol (HDL), fasting blood glucose, blood pressure, fasting blood glucose, fasting insulin, and metformin regimen adherence) from the DPP trial and banked biospecimens from the National Institute of Diabetes, Digestive Disease, and Kidney Diseases biorepository.[36-37]

Molecular Data Collection
Blood was collected by venipuncture into vacutainers with heparin.[37] The Fireplex Multiplex Circulating MicroRNA Assay (Abcam, MA) was used for direct quanti cation of 58 miRs from plasma collected with heparin.miRs were hybridized to complementary oligonucleotides covalently attached to encoded hydrogel microparticles.The bound target was ligated to oligonucleotide adapter sequences that serve as universal polymerase chain reaction priming sites.The miR-adapter hybrid models were then denatured from the particles, and reverse transcription polymerase chain reaction was performed using a uorescent forward primer.Once ampli ed, the uorescent target was rehybridized to the original capture particles and scanned on an EMD Millipore Guava 6HT ow cytometer (Merck KGaA Darmstadt, Germany).
Expression of individual miRs was normalized using the set of miR probes (i.e., miR-92a-3p, mir-93-5p, miR-17-5p) identi ed by the geNorm algorithm.Prior to analysis, a lower limit of quanti cation (LLOQ) equal to two times the minimal detectable difference for each miR was applied to the samples.Only miRs detected in ≥ 90% of the samples after ltering of sub-LLOQ values were retained for comparison, then converted to z-scores to account for batch effects between data collection timepoints.This resulted in 35 miRs that were retained for the analysis.

Data Analysis
The sample was characterized using descriptive statistics, and chi-squared or Wilcoxon signed rank test for independent samples were used to evaluate baseline categorical and continuous demographics, clinical characteristics, and baseline miRs between groups.Mean change in levels of 24 circulating miRs from baseline to 2 years were compared between the metformin intervention group (n = 50) and the placebo group (n = 50) using Wilcoxon signed rank test for independent samples, and then adjusted for multiple comparisons.The Benjamini-Hochberg False Discovery Rate (FDR) method was used to adjust for multiple comparisons (FDR < 0.05).Spearman's rank correlation was used to assess for relationships between the change in miRs from baseline to 2 years and the baseline continuous clinical variables, adjusted for multiple comparisons.Multiple linear regression was used to assess the in uence of covariates (age, sex, race and ethnicity, and BMI) on differences in mean change by study group.

Conclusions
This study investigated the effects of a two-year metformin therapy versus placebo on the levels of miR expression in adult participants from the DPP randomized controlled trial.Changes in the expression of twelve speci c miRs were identi ed, of which six were upregulated and six were downregulated in response to metformin treatment.Among these miRs, nine were previously not known to be associated with metformin therapy response.The ndings suggested that metformin's mechanisms of action are complex, involving a range of functions, tissues, and biological systems related to the transition from prediabetes to T2D, along with its associated comorbidities.
This study suggests that different miRs are involved in various aspects of metformin's therapeutic effects on T2D risk, possibly indicating subgroups of individuals who could bene t the most from metformin treatment.These effects extend far beyond the glucose-lowering bene ts and the AMPK pathway.However, the study has limitations, including the small sample size and lack of detailed social determinant and environmental data.Further research with larger samples and considering factors such as dose variability and duration of therapy would provide a more comprehensive understanding of miR responses to metformin treatment and its role in T2D prevention.

Figure 1 Box
Figure 1 Box Plots of Mean Change in MicroRNA Expression Level by Treatment Group for MicroRNAs Signi cant After Adjustment for Multiple Comparisons Note: MiR-93a--5p was borderline statistically signi cant after adjusting for multiple comparisons (FDR=0.059).

Figure 2 Correlation
Figure 2 Correlation plot featuring only miRNAs signi cant after adjustment for multiple comparisons and baseline clinical characteristics.Cells with p > 0.05 are omitted.Color coding represents the spearman's rank correlation coe cient.

Table 1
Demographics and Clinical Characteristics in the Parent Study Overall and by Study Group *Self-Reported; HDL = High Density Lipoprotein; LDL = Low Density Lipoprotein

Table 2 Median
MicroRNA Expression Levels at Baseline, Year 2, and Median 2-Year Change in miRNA z-Scores by Treatment Group

Table 3
Multivariate Linear Regression of Mean Change in microRNA z-Score by Treatment Group, Adjusted for Baseline Age, Sex, Race and Ethnicity, and Body Ma

Table 4
Summary of the 12 Differentially Expressed miRs, Response to Metformin Therapy, Demographic and Clinical Associations, Target Genes and Proteins, and Summary of Mechanisms in Type 2 Diabetes