Validation of miR-423 and U6 as Reference Genes for Circulating microRNA in Pediatric Diabetes Mellitus

Introduction

Type 1 diabetes mellitus (T1DM) is a chronic autoimmune disease affecting 1.1 million children worldwide (American Diabetes Association Professional Practice Committee, 2025; Papapetrou & Swiecicka, 2025). Autoimmune destruction of pancreatic β-cells causes lifelong insulin dependence (Papapetrou & Swiecicka, 2025; Xie et al., 2025). Early detection and monitoring are challenging, and pediatric T1DM carries higher risks of metabolic decompensation, glycemic variability, and long-term complications (Urbano et al., 2023).

MicroRNAs (miRNAs), small RNA molecules with a length of about 22 nucleotides, act as post-transcriptional regulators of gene expression through degradation of mRNA or inhibition of translation (Snowhite et al., 2017; Cho et al., 2025; Santos et al., 2025). The references cite essential functions in β-cells, immune development, and metabolic homeostasis (Eizirik et al., 2020; Santos et al., 2025). Research indicates that miR-1, expressed in cardiac and skeletal muscle, regulates β-cell apoptosis and may be a marker for diabetic cardiomyopathy (Macvanin & Isenovic, 2023; De et al., 2025; Żarek et al., 2025). The inflammatory pathways and immune responses are modulated by miR-155, which regulates the cytokine production in lymphocytes (Gaál, 2024; Koumpis et al., 2024; Zhao & Wang, 2025). Circulating miRNAs are emerging as non-invasive biomarkers for diagnosis, prognosis, and monitoring of glycemic control in paediatric T1DM (Syed et al., 2023).

Despite this, methodological discrepancies, notably in quantitative real-time PCR (qRT-PCR) normalization, affect clinical application (Sekovanić et al., 2025). The stability of Reference gene expression may vary depending on disease state, tissue type, or experimental conditions. A misleading conclusion can be reached by the use of an inappropriate reference gene, which can affect the validation of the biomarker and clinical decisions (Jóźwik et al., 2025; Luigi-Sierra et al., 2025; Sekovanić et al., 2025). The recent studies indicate normalization strategies must be modified to pathological contexts that may affect the expression of housekeeping genes (Jóźwik et al., 2025; Luigi-Sierra et al., 2025).

Information on miRNA biomarkers from the Middle East, especially that from Iraq, is limited. Potential environmental influences such as viral infection and air pollution, as well as socioeconomic conditions, may affect disease phenotype and miRNA expression (Abu-Farha et al., 2021; Cho et al., 2025; Tang et al., 2025). There are also a few comparative studies of normalization strategies within the same cohort, which reduces generalizability and clinical applicability.

The study assesses the presence of circulating miR-1 and miR-155 among T1DM pediatric patients from Kirkuk, Iraq. This paper looks at the role of reference gene selection- miR-423 against U6 snRNA, on diagnostic performance. Further, it explores the associations of miRNAs with glycaemic control (HbA1c). The study shows the way to reliable and reproducible candidate biomarkers in paediatric T1DM.

Materials and Methods

Study Population

This study involved a total of 70 pediatric patients with T1DM and 30 age and sex-matched healthy controls from Kirkuk.  The inclusion criteria specified for the study were an age range of 4 to 20 years with a diagnosis confirmed as per the American Diabetes Association guidelines (American Diabetes Association Professional Practice Committee, 2025). People with other autoimmune diseases, acute infections within four weeks, and chronic kidney or liver disease are excluded. Healthy controls were recruited from the same geographic area, with no personal or family history of diabetes or autoimmune conditions. Data on clinical and demographic factors were obtained, including age, sex, disease duration, current insulin regimen, and last HbA1c. According to pediatric T1DM management guidelines, glycemic control was classified as good (HbA1c <7.0%), moderate (7.0–8.0%), and poor (>8.0%) (de Bock et al., 2024; Hatun et al., 2024; Evans-Molina & Oram, 2025).

Total RNA Extraction

The TRIzol™ RNA Isolation Kit (Invitrogen, Carlsbad, CA, USA) was used to extract total cell-free RNA, including miRNAs, with slight modifications. In brief, 0.50 mL of TRIzol™ was added to 0.25 mL serum, followed by 15 minutes of incubation at room temperature, and then 0.15 mL chloroform was added.  Centrifugation at 12,000 × g for 15 min at 4 °C was done after vigorous mixing and incubation at −20 °C for 20 min. The aqueous phase was collected, and the RNA was precipitated using isopropanol. The pellet was washed with 75% ethanol, air-dried, and resuspended in 40 µL of RNA dissolving solution for 15 minutes at 60 °C. Nothing, first RNAse-free condition was carried out, and RNA was maintained at −80°C till cDNA synthesis.

Primer Design and Validation

Primers specific to miR-1, miR-155, and reference genes (U6 and miR-423) were designed from miRBase (Release 22.1) and NCBI GenBank. The stem-loop RT primers are designed to increase specificity for mature miRNAs (~22 nucleotides). The lyophilized primers were dissolved in nuclease-free water to generate a stock (100 µM) and working (10 µM) solutions. These solutions were stored at -20 °C.

Complementary DNA (cDNA) Synthesis

Synthesizing the first strand of cDNA was done using the ProtoScript® First Strand cDNA Synthesis Kit from New England BioLabs (Ipswich, MA, USA), which was optimized for small RNAs. Reverse transcription was done using M-MuLV Reverse Transcriptase, prepared as instructed.

Quantitative Real-Time PCR (qRT-PCR)

Performed qRT-PCR using the Luna® Universal qPCR Master Mix with SYBR® Green I on the Rotor-Gene Q system (Qiagen, Hilden, Germany). Segregated reactions were prepared for target miRNAs and reference genes. Each of the samples was run in technical duplicates, and relative expression was determined using the 2^−ΔΔCq method using the U6 or miR-423 normalization. Melt curve analysis confirmed the specificity of amplicons.

Data Analysis and Normalization

ΔCq values were calculated as Cq(target miRNA) − Cq(reference gene), and ΔΔCq as ΔCq(patient) − mean ΔCq(control). Fold change was determined as 2^−ΔΔCq. Reference gene stability was assessed using the coefficient of variation (CV), and independent samples t-tests or Mann-Whitney U tests were applied to confirm no differential expression between groups.

Statistical Analysis

Analyses were performed using SPSS 28.0. Normality was assessed with the Shapiro-Wilk test. Normally distributed data were expressed as mean ± SD; non-normal data as median (IQR). Mann-Whitney U tests were used for miRNA comparisons, with effect sizes calculated via rank-biserial correlation. ROC curve analysis evaluated diagnostic performance, including AUC, sensitivity, specificity, and optimal cut-off (Youden index), with 95% CI determined by the DeLong method. Associations with HbA1c were assessed using Spearman correlation, t-tests, Mann-Whitney U tests, or Fisher’s exact test as appropriate. Effect sizes were reported using Cohen’s d (parametric) or rank-biserial r (non-parametric). Bonferroni correction was applied for multiple comparisons (Babaei et al., 2023).

Results and Discussion

To evaluate the circulating miR-1 and miR-155 profiles in pediatric Type 1 Diabetes Mellitus (T1DM)-70 patients (n=70), a comprehensive microRNA expression analysis was conducted, and miR-155 was among the differentially expressed miRNAs. In all quantitative determinations, RT-qPCR employed two independent reference genes (miR-423 and U6) for methodological robustness (Babaei et al., 2023; Cahyaningsih et al., 2023; Doddapanen et al., 2024; Kovalchuk et al., 2024; Shaji et al., 2024; Singar, 2024).

Descriptive Statistics of MicroRNA Expression

Stability of miR-1 levels was observed in T1DM patients, regardless of normalization methodology.   Patients showed a significant increase in miR-423 compared to controls after normalization. There was a 2.7-fold change. In U6 normalization, patients showed a mean fold-change of 3.37±2.79 (median: 2.59) versus controls 1.04±0.41. MiR-423 and U6 ΔCt values in patients (22.34±1.11 and 11.98±1.17) were significantly lower than those in controls (23.49±0.82 and 12.97±0.85).

miR-155 presented patterns reliant on reference genes. After miR-423 normalization of their samples, the patients manifested an obvious downregulation (mean: 0.69±0.47, median: 0.60) versus the controls (1.01±0.41, median: 0.96). In contrast, U6 normalization showed that patients (1.41±1.01, median 1.13) had much greater expression than controls (0.99±0.41, median 0.92). This illustrates the strong influence of reference genes on a quantitative interpretation (Aksoy & Akaydin, 2024; Ha & Hang, 2024; Hima et al., 2024; Jegede, 2024; Karthikeyan et al., 2024).

Reference Gene Validation and Stability Assessment

Essential requirements for accurate assessment of microRNA expression through quantitative RT-PCR include selection and validation of reference genes. The stability characteristics of miR-423 showed a coefficient of variation of 4.86–4.87% in control and patient groups. The mean Ct values in controls and T1DM were 25.83±1.26 and 25.28±1.23, respectively, with standard deviations being 1.3 cycles or less in all cases. The low inter-sample variability of miR-423 indicates its expression does not change a lot regardless of disease state, making it a candidate reference gene by this criterion. U6 snRNA presented satisfactory yet lower stability with coefficient of variation values 7.05–7.14%. The mean Ct values in the control vs patient group were 14.45±1.02 vs 14.27±1.02, respectively. U6 has often been used as a reference gene in microRNA studies; however, its high variability in the current cohort indicates it could be more susceptible to T1DM-associated physiological changes. The slightly higher variability seen with U6 highlights the value of using multiple reference genes to triangulate expression findings and improve result reliability, as illustrated in Table 1.

Table 1. Stability Evaluation of Reference Genes (miR-423 and U6)

Statistical Comparison of MicroRNA Expression between Groups

According to the results of the Mann-Whitney U tests, when normalized to miR-423, the median fold change of miR-1 was 2.30 in T1DM patients versus 1.02 in healthy controls, as shown in Table 2. Importantly, T1DM patients showed significant upregulation of miR-1 compared to healthy controls (U=1176.5, Z=−2.795, p=0.005, r=0.236).  ΔCt analysis supported these findings with median values of 22.42 in patients and 23.42 in controls (U=1182.5, Z=−2.775, p=0.006, r=0.234).

Statistical evidence was stronger for miR-1 upregulation by U6 normalization. Patients had a 2.59-fold change in Z-advanced than controls, with r=0.267. ΔCt comparisons also showed highly significant differences being median of 11.86 in patients versus 12.97 in controls (U=1311.5, Z=−3.389, p=0.001, r=0.286). The repeatability of the results for both reference genes and different expression measures signifies confirmation for the genuine dysregulation of miR-1 in children with type 1 diabetes mellitus (T1DM) (Aksoy & Akaydin, 2024; Cachón-Rodríguez et al., 2024; Ha & Hang, 2024; Jegede, 2024; Ahmed & Rajasekar, 2025; Drissi et al., 2025; Rajadurai & Govindaraju, 2025; Altaie et al., 2026).

miR-155 showed drastically reference-dependent outcomes. When miR-423 was normalized, differential expression revealed statistically significant downregulation in T1DM patients with a median fold change of 0.60 (IQR, 0.52–0.76) compared with 0.96 (IQR, 0.87–1.09) in controls (U=1280.0, Z=−3.215, p=0.001, r=0.271). The ΔCt analysis further confirmed this, showing a higher median of 23.29 in patients than in controls (22.79) (U=1288.0, Z=−3.256, p=0.001, r=0.275). On the other hand, for U6, this effect was no longer significant, with a median fold change of 1.13 in patients and 0.92 in controls (U=987.5, Z=−1.279, p=0.201, r=0.108). The above difference in normalization methods indicates that the apparent up-regulation of miR-155 in T1DM depends on reference gene choice.

Table 2. Mann-Whitney U Test Results Comparing miRNA Expression Between Groups

Diagnostic Performance Assessment via ROC Curve Analysis

ROC curve analyses were conducted to evaluate the potential clinical utility of miR-1 and miR-155 as diagnostic biomarkers for pediatric T1DM, as shown in Table 3.

Table 3. ROC Curve Analysis Results for Diagnostic Performance Assessment

miR-1 demonstrated moderate-to-good discriminative ability across all normalization strategies. When normalized to miR-423, fold change analysis yielded an AUC of 0.631 (95% CI: 0.541–0.722, p=0.006) with an optimal cutoff of 1.463, providing 67.1% sensitivity and 60.5% specificity. ΔCt analysis produced a comparable AUC of 0.632 (95% CI: 0.541–0.722, p=0.006). U6 normalization substantially enhanced diagnostic performance, yielding an AUC of 0.700 (95% CI: 0.615–0.785, p<0.001) for fold change—an 8.96% relative increase compared to miR-423 normalization. The optimal cutoff of 1.516 provided 74.3% sensitivity and 64.5% specificity. ΔCt analysis with U6 normalization achieved the strongest performance metrics (AUC=0.717, 95% CI: 0.634–0.799, p<0.001), with a cutoff of 12.42 yielding 75.7% sensitivity and 64.5% specificity. The consistent moderate-to-good AUC values (~0.70) indicate that miR-1 alone is insufficient for definitive diagnostic classification but shows promise as a component of multi-marker diagnostic panels in Figure 1a.

miR-155 exhibited markedly inferior and reference-dependent diagnostic performance. When normalized to miR-423, fold change AUC was 0.665 (95% CI: 0.576–0.754, p=0.001) with an optimal cutoff of 0.760 providing 68.6% sensitivity and 58.1% specificity. However, U6 normalization drastically diminished diagnostic performance, yielding a non-significant AUC of 0.564 for fold change (95% CI: 0.469–0.660, p=0.179) and 0.588 for ΔCt (95% CI: 0.493–0.682, p=0.064). These results align with the non-significant Mann-Whitney comparisons observed with U6 normalization in Figure 1b.

Correlation Analysis of qPCR Expression Metrics

Pearson correlation analyses validated internal consistency and mathematical relationships between ΔCt, ΔΔCt, and fold change values, as shown in Table 4. Across all datasets, ΔCt and ΔΔCt values exhibited strong positive correlations (r=0.875 to 0.962, all p<0.001), confirming the expected mathematical relationship. These correlations were strong (r²>0.76), which proved the internal consistency of the quantification. The fold change values were found to correlate moderately to strongly with ΔCt (r=−0.403 to −0.716, all p<0.001). This was expected, and can be explained with the knowledge that the 2^(-ΔΔCt) transformation inherently has an inverse exponential relationship. In a comparable manner, both fold change and ΔΔCt displayed moderate-to-strong negative correlations (r=−0.443 to −0.756, all p<0.001), which confirmed their mathematical expectations. Normalizing to U6 produced stronger correlations than normalizing to miR-423 for both microRNAs, suggesting consistency.

Table 4. Pearson Correlation Analysis Between qPCR Expression Variables

Association between Circulating miR-1 Expression and Glycemic Control

The relationships of circulating miR-1 expression (U6-normalized) and glycemic control parameters in the T1DM patient cohort (n=70) were complex Table 5. HbA1C values were between 6.40%-11.50%, mean – 7.66±0.92% and median – 7.40%. No significant linear association was detected between the miR-1 fold change and HbA1c levels (ρ=−0.069, p=0.569). In the same way, the Pearson correlation produced a weak, non-significant negative correlation (r=−0.095, p=0.435).

Table 5. Comparative Analysis of Glycemic Control Parameters by miR-1 Regulation Status