Machine learning workflows identify a microRNA signature of insulin transcription in human tissues

Prediction of progression to type 1 diabetes with dynamic biomarkers and risk scores

Identifying biomarkers of functional β-cell loss is an important step in the risk stratification of type 1 diabetes. Genetic risk scores (GRS), generated by profiling an array of single nucleotide polymorphisms, are a widely used type 1 diabetes risk-prediction tool. Type 1 diabetes screening studies have relied on a combination of biochemical (autoantibody) and GRS screening methodologies for identifying individuals at high-risk of type 1 diabetes. A limitation of these screening tools is that the presence of autoantibodies marks the initiation of β-cell loss, and is therefore not the best biomarker of progression to early-stage type 1 diabetes. GRS, on the other hand, represents a static biomarker offering a single risk score over an individual's lifetime. In this Personal View, we explore the challenges and opportunities of static and dynamic biomarkers in the prediction of progression to type 1 diabetes. We discuss future directions wherein newer dynamic risk scores could be used to predict type 1 diabetes risk, assess the efficacy of new and emerging drugs to retard, or prevent type 1 diabetes, and possibly replace or further enhance the predictive ability offered by static biomarkers, such as GRS.

MicroRNA regulation of islet and enteroendocrine peptides: Physiology and therapeutic implications for type 2 diabetes

The pathogenesis of type 2 diabetes (T2D) is associated with dysregulation of glucoregulatory hormones, including both islet and enteroendocrine peptides. Microribonucleic acids (miRNAs) are short noncoding RNA sequences which post transcriptionally inhibit protein synthesis by binding to complementary messenger RNA (mRNA). Essential for normal cell activities, including proliferation and apoptosis, dysregulation of these noncoding RNA molecules have been linked to several diseases, including diabetes, where alterations in miRNA expression within pancreatic islets have been observed. This may occur as a compensatory mechanism to maintain beta-cell mass/function (e.g., downregulation of miR-7), or conversely, lead to further beta-cell demise and disease progression (e.g., upregulation of miR-187). Thus, targeting miRNAs has potential for novel diagnostic and therapeutic applications in T2D. This is reinforced by the success seen to date with miRNA-based therapeutics for other conditions currently in clinical trials. In this review, differential expression of miRNAs in human islets associated with T2D will be discussed along with further consideration of their effects on the production and secretion of islet and incretin hormones. This analysis further unravels the therapeutic potential of miRNAs and offers insights into novel strategies for T2D management.

Reassessing the Abundance of miRNAs in the Human Pancreas and Rodent Cell Lines and Its Implication

miRNAs are critical for pancreas development and function. However, we found that there are discrepancies regarding pancreatic miRNA abundance in published datasets. To obtain a more relevant profile that is closer to the true profile, we profiled small RNAs from human islets cells, acini, and four rodent pancreatic cell lines routinely used in diabetes and pancreatic research using a bias reduction protocol for small RNA sequencing. In contrast to the previous notion that miR-375-3p is the most abundant pancreatic miRNA, we found that miR-148a-3p and miR-7-5p were also abundant in islets. In silico studies using predicted and validated targets of these three miRNAs revealed that they may work cooperatively in endocrine and exocrine cells. Our results also suggest, compared to the most-studied miR-375, that both miR-148a-3p and miR-7-5p may play more critical roles in the human pancreas. Moreover, according to in silico-predicted targets, we found that miR-375-3p had a much broader target spectrum by targeting the coding sequence and the 5' untranslated region, rather than the conventional 3' untranslated region, suggesting additional unexplored roles of miR-375-3p beyond the pancreas. Our study provides a valuable new resource for studying miRNAs in pancreata.

The link between glycemic control measures and eye microvascular complications in a clinical cohort of type 2 diabetes with microRNA-223-3p signature

BACKGROUND

Type 2 diabetes (T2D) is a critical healthcare challenge and priority in Qatar which is listed amongst the top 10 countries in the world, with its prevalence presently at 17% double the global average. MicroRNAs (miRNAs) are implicated in the pathogenesis of (T2D) and long-term microvascular complications including diabetic retinopathy (DR).

METHODS

In this study, a T2D cohort that accurately matches the characteristics of the general population was employed to find microRNA (miRNA) signatures that are correlated with glycemic and β cell function measurements. Targeted miRNA profiling was performed in (471) T2D individuals with or without DR and (491) (non-diabetic) healthy controls from the Qatar Biobank. Discovery analysis identified 20 differentially expressed miRNAs in T2D compared to controls, of which miR-223-3p was significantly upregulated (fold change:5.16, p = 3.6e-02) and positively correlated with glucose and hemoglobin A1c (HbA1c) levels (p-value = 9.88e-04 and 1.64e-05, respectively), but did not show any significant associations with insulin or C-peptide. Accordingly, we performed functional validation using a miR-223-3p mimic (overexpression) under control and hyperglycemia-induced conditions in a zebrafish model.

RESULTS

Over-expression of miR-223-3p alone was associated with significantly higher glucose (42.7 mg/dL, n = 75 vs 38.7 mg/dL, n = 75, p = 0.02) and degenerated retinal vasculature, and altered retinal morphology involving changes in the ganglion cell layer and inner and outer nuclear layers. Assessment of retinal angiogenesis revealed significant upregulation in the expression of vascular endothelial growth factor and its receptors, including kinase insert domain receptor. Further, the pancreatic markers, pancreatic and duodenal homeobox 1, and the insulin gene expressions were upregulated in the miR-223-3p group.

CONCLUSION

Our zebrafish model validates a novel correlation between miR-223-3p and DR development. Targeting miR-223-3p in T2D patients may serve as a promising therapeutic strategy to control DR in at-risk individuals.

Circulating microRNAs from early childhood and adolescence are associated with pre-diabetes at 18 years of age in women from the PMNS cohort

With type 2 diabetes presenting at younger ages, there is a growing need to identify biomarkers of future glucose intolerance. A high (20%) prevalence of glucose intolerance at 18 years was seen in women from the Pune Maternal Nutrition Study (PMNS) birth cohort. We investigated the potential of circulating microRNAs in risk stratification for future pre-diabetes in these women. Here, we provide preliminary longitudinal analyses of circulating microRNAs in normal glucose tolerant (NGT@18y, N = 10) and glucose intolerant (N = 8) women (ADA criteria) at 6, 12 and 17 years of their age using discovery analysis (OpenArray™ platform). Machine-learning workflows involving Lasso with bootstrapping/leave-one-out cross-validation identified microRNAs associated with glucose intolerance at 18 years of age. Several microRNAs, including miR-212-3p, miR-30e-3p and miR-638, stratified glucose-intolerant women from NGT at childhood. Our results suggest that circulating microRNAs, longitudinally assessed over 17 years of life, are dynamic biomarkers associated with and predictive of pre-diabetes at 18 years of age. Validation of these findings in males and remaining participants from the PMNS birth cohort will provide a unique opportunity to study novel epigenetic mechanisms in the life-course progression of glucose intolerance and enhance current clinical risk prediction of pre-diabetes and progression to type 2 diabetes.

The microRNA-29 family - role in metabolism and metabolic disease

The microRNA-29a family members miR-29a-3p, miR-29b-3p and miR-29c-3p are ubiquitously expressed and consistently increased in various tissues and cell types in conditions of metabolic disease; obesity, insulin resistance and type 2 diabetes. In pancreatic beta cells, miR-29a is required for normal exocytosis, but increased levels are associated with impaired beta cell function. Similarly, in liver miR-29 species are higher in models of insulin resistance and type 2 diabetes, and either knock-out or depletion using a microRNA inhibitor improves hepatic insulin resistance. In skeletal muscle, miR-29 upregulation is associated with insulin resistance and altered substrate oxidation, and similarly, in adipocytes over-expression of miR-29a leads to insulin resistance. Blocking miR-29a using nucleic acid antisense therapeutics show promising results in preclinical animal models of obesity and type 2 diabetes, although the widespread expression pattern of miR-29 family members complicates the exploration of single target tissues. However, in fibrotic diseases, such as in late complications of diabetes and metabolic disease (diabetic kidney disease, non-alcoholic steatohepatitis), miR-29 expression is suppressed by TGFβ allowing increased extracellular matrix collagen to form. In the clinical setting circulating levels of miR-29a and miR-29b are consistently increased in type 2 diabetes and in gestational diabetes, and are also possible prognostic markers for deterioration of glucose tolerance. In conclusion, miR-29 plays an essential role in various organs relevant to intermediary metabolism and its upregulation contribute to impaired glucose metabolism, while it suppresses fibrosis development. Thus, a correct balance of miR-29a levels seems important for cellular and organ homeostasis in metabolism.

Human pancreatic islet miRNA-mRNA networks of altered miRNAs due to glycemic status

MicroRNAs (miRNAs) are short non-coding RNAs that regulate gene expression via mRNA targeting, playing important roles in the pancreatic islets. We aimed to identify molecular pathways and genomic regulatory regions associated with altered miRNA expression due to glycemic status, which could contribute to the development of type 2 diabetes (T2D). To this end, miRNAs were identified by a combination of differential miRNA expression and correlation analysis in human islet samples from donors with normal and elevated blood glucose levels. Analysis and clustering of highly correlated, experimentally validated gene targets of these miRNAs revealed two islet-specific clusters, which were associated with key aspects of islet functions and included a high number of T2D-related genes. Finally, cis-eQTLs and public GWAS data integration uncovered suggestive genomic signals of association with insulin secretion and T2D. The miRNA-driven network-based approach presented in this study contributes to a better understanding of impaired insulin secretion in T2D pathogenesis.

Analysis of Half a Billion Datapoints Across Ten Machine-Learning Algorithms Identifies Key Elements Associated With Insulin Transcription in Human Pancreatic Islet Cells

Machine learning (ML)-workflows enable unprejudiced/robust evaluation of complex datasets. Here, we analyzed over 490,000,000 data points to compare 10 different ML-workflows in a large (N=11,652) training dataset of human pancreatic single-cell (sc-)transcriptomes to identify genes associated with the presence or absence of insulin transcript(s). Prediction accuracy/sensitivity of each ML-workflow was tested in a separate validation dataset (N=2,913). Ensemble ML-workflows, in particular Random Forest ML-algorithm delivered high predictive power (AUC=0.83) and sensitivity (0.98), compared to other algorithms. The transcripts identified through these analyses also demonstrated significant correlation with insulin in bulk RNA-seq data from human islets. The top-10 features, (including IAPP, ADCYAP1, LDHA and SST) common to the three Ensemble ML-workflows were significantly dysregulated in scRNA-seq datasets from Ire-1αβ-/- mice that demonstrate dedifferentiation of pancreatic β-cells in a model of type 1 diabetes (T1D) and in pancreatic single cells from individuals with type 2 Diabetes (T2D). Our findings provide direct comparison of ML-workflows in big data analyses, identify key elements associated with insulin transcription and provide workflows for future analyses.

Machine Learning Identifies Cellular and Exosomal MicroRNA Signatures of Lyssavirus Infection in Human Stem Cell-Derived Neurons

Despite being vaccine preventable, rabies (lyssavirus) still has a significant impact on global mortality, disproportionally affecting children under 15 years of age. This neurotropic virus is deft at avoiding the immune system while travelling through neurons to the brain. Until recently, research efforts into the role of non-coding RNAs in rabies pathogenicity and detection have been hampered by a lack of human in vitro neuronal models. Here, we utilized our previously described human stem cell-derived neural model to investigate the effect of lyssavirus infection on microRNA (miRNA) expression in human neural cells and their secreted exosomes. Conventional differential expression analysis identified 25 cellular and 16 exosomal miRNAs that were significantly altered (FDR adjusted P-value <0.05) in response to different lyssavirus strains. Supervised machine learning algorithms determined 6 cellular miRNAs (miR-99b-5p, miR-346, miR-5701, miR-138-2-3p, miR-651-5p, and miR-7977) were indicative of lyssavirus infection (100% accuracy), with the first four miRNAs having previously established roles in neuronal function, or panic and impulsivity-related behaviors. Another 4-miRNA signatures in exosomes (miR-25-3p, miR-26b-5p, miR-218-5p, miR-598-3p) can independently predict lyssavirus infected cells with >99% accuracy. Identification of these robust lyssavirus miRNA signatures offers further insight into neural lineage responses to infection and provides a foundation for utilizing exosome miRNAs in the development of next-generation molecular diagnostics for rabies.

Manipulating cellular microRNAs and analyzing high-dimensional gene expression data using machine learning workflows

MicroRNAs (miRNAs) are elements of the gene regulatory network and manipulating their abundance is essential toward elucidating their role in patho-physiological conditions. We present a detailed workflow that identifies important miRNAs using a machine learning algorithm. We then provide optimized techniques to validate the identified miRNAs through over-expression/loss-of-function studies. Overall, these protocols apply to any field in biology where high-dimensional data are produced.

For complete details on the use and execution of this protocol, please refer to Wong et al. (2021a).

•

LASSO and bootstrapping identify important miRNAs associated with gene of interest

•

Generating puromycin resistant, doxycycline inducible, miRNA overexpressing cells

•

Transient miRNA knockdown using LNA inhibitors in human primary cells

•

Optimized reagent concentrations and cell densities for miRNA manipulation

Molecular Mechanisms of Nutrient-Mediated Regulation of MicroRNAs in Pancreatic β-cells

Pancreatic β-cells within the islets of Langerhans respond to rising blood glucose levels by secreting insulin that stimulates glucose uptake by peripheral tissues to maintain whole body energy homeostasis. To different extents, failure of β-cell function and/or β-cell loss contribute to the development of Type 1 and Type 2 diabetes. Chronically elevated glycaemia and high circulating free fatty acids, as often seen in obese diabetics, accelerate β-cell failure and the development of the disease. MiRNAs are essential for endocrine development and for mature pancreatic β-cell function and are dysregulated in diabetes. In this review, we summarize the different molecular mechanisms that control miRNA expression and function, including transcription, stability, posttranscriptional modifications, and interaction with RNA binding proteins and other non-coding RNAs. We also discuss which of these mechanisms are responsible for the nutrient-mediated regulation of the activity of β-cell miRNAs and identify some of the more important knowledge gaps in the field.