The identification of the target genes and pathways followed a specific selection criterion, based mainly on bioinformatics-based selection and other criteria such as Pathway analysis, Relevance to the disease of interest, Functionality, Cost-effectiveness, and Accessibility. These criteria help to ensure that the target genes and pathways selected for the RNA PCR gene expression panel analysis are relevant, well-established, and feasible to measure, which can increase the confidence in the results and the impact of the study.

We used bioinformatics analysis to select candidate mRNA genes which are related to Type 2 DM, insulin resistance signaling, STimulator of Interferon Gene (STING), and NLR signaling pathways from public microarray databases, e.g., National Center of Biotechnology Information GEO (https://www.ncbi.nlm.nih.gov/geo/, available May 2022), Gene atlas expression database (https://www.ebi.ac.uk/gxa/home), and KEGG map (https://www.genome.jp/kegg/) to choose a pathway related to insulin resistance in T2DM and identify the target effector. Search was confined to Homo sapiens species and p-value ≤0.05 was considered statistically significant. Among 156 differentially expressed genes and FDR-adjusted p-value <0.05, We have retrieved a set TMEM173 and CHUK mRNAs according to the following criteria: (i) Genes which are dysregulated in T2DM and insulin resistance molecular pathways; (ii) Genes which are expressed in tissues of interest, e.g., skeletal muscle, adipose tissue and also in blood samples for easy extraction and less invasiveness; (iii) Both genes are related to STING and NLR signaling pathway involved in innate immunity and chronic sterile inflammation; (iv). Gene ontology (GO) enrichment of TMEM173 and CHUK mRNAs were performed using Enrichr (http://amp.pharm.mssm.edu/Enrichr) with p < 0.05 set as the cutoff criterion that verified their correlation to positive regulation of interferon production and kappa beta phosphorylation. TMEM173 and CHUK mRNAs were imported into the Search Tool for the retrieval of interacting genes (STRING; version 11.0; http://stringdb.org) online database for protein–protein interaction (PPI) network building to ensure their link to insulin resistance pathways and previously known genes in T2DM through STRING interaction network. PPI enrichment p-value was 1.6E-05 with combined score 0.943 for CHUK and TMEM173, 0.847. Afterwards, we predicted upstream key miRNAs and lncRNAs using miRWalk 3.0 (http://mirwalk.umm.uni-heidelberg.de/) and DIANA Tools database, which integrate the target prediction of both TargetScan * MiRBase. miRNAs and lncRNAs interaction was predicted by using miRWalk 2.0; miRNA: ncRNA target tool. We found that hsa-miR (-611, -5192, and -1976) can interact with both selected mRNAs with score >0.9 at CDS binding sites. Finally, RP4-605O3.4 and AC074117.2 were identified to be interacting with the chosen miR (-611, -5192, and -1976) and which was confirmed by Clustal Omega tool of The European Bioinformatics Institute based on high complementarity rate between them. Thus, finally, the (TMEM173 and CHUK) - (miR-611, -5192 and-1976) - (RP4-605O3.4 and AC074117.2) regulatory panel was constructed. Detailed bioinformatics are shown in Figure 1 and Supplementary Figures S1–9 .

The present study includes 160 study subjects, classified into 66 cases with type 2 diabetes mellitus, 49 were prediabetics and 45 healthy volunteers. There were no significant differences as regards sex, age, and smoking between the three investigated groups (p > 0.05). On the other hand, there were significant differences among the three studied groups concerning family history of T2DM (p < 0.001), fasting and 2-h post-prandial blood glucose levels (p < 0.001), HbA1c, fasting insulin level (p < 0.001), BMI (p < 0.001), systolic and diastolic BP (p < 0.001), lipid profile including total Cholesterol, LDLc, HDLc, and Triglycerides (p < 0.001), and Alb/Cr. ratio (p < 0.001) as shown in Supplementary Table S1 .

HOMA-IR was used as a valid surrogate indicator of insulin resistance and was calculated according to the equation: Fasting insulin (μU/L) x fasting glucose (nmol/L)/22.5 (39), while HOMA-B was calculated as follows: (20x insulin in mIU/ml)/(glucose in mmol/L – 3.5) and it has been suggested to be an effective indicator of B-cell function (40). Both formulas measure HOMA-IR, although they do so from a different perspective. We found significant differences between the three studied groups concerning HOMA-IR (p = <0.001), which was significantly higher in T2DM group compared to prediabetic and healthy control groups; it was also higher in the prediabetic group compared to the healthy control group which indicates different degrees of insulin sensitivity between the three groups.

There was high positive significant correlation between both mRNAs TMEM173 and CHUK and the three miRNAs: hsa-miR (-611, -5192, and -1976) while it was a negative correlation with the two lncRNAs: RP4-605O3.4 and AC074117.2 among all the studied groups and also among the prediabetic and the healthy control group which suggests that the lncRNAs work as a ceRNA competing with the miRNAs for the mRNA pool in the molecular pathogenesis of IR, prediabetes, and T2DM ( Table S2-S4 ).

There was also a positive significant correlation between both mRNAs TMEM173 and CHUK, the three miRNAs; hsa-miR (-611, -5192, and -1976) and the different clinicopathological factors except HOMA-B, which was negatively correlated, while the correlation was negative with the two lncRNAs; RP4-605O3.4 and AC074117.2 and the different clinicopathological factors but it was positive correlation with HOMA-B ( Table S4 ). Spearman correlation was also done between the STING/NOD/IR related RNA associated panel among prediabetic and T2DM groups, there were significant positive correlations between the expression of TMEM173 and CHUK mRNAs, hsa-miR (-611 and -1976), and negative correlation with RP4-605O3.4 lncRNA. To evaluate the diagnostic potential of the investigated IR panel, Spearman correlation of the clinical traits such as: HbA1c and HOMA-IR, with the levels of expression of individual RNA was done for better modeling of how these RNA levels could explain the pathogenesis of IR/T2DM and to reveal the usefulness of these RNAs as diagnostic markers ( Figures 2 – 4 and Tables S2–S4 ).

The heatmap (https://biit.cs.ut.ee/clustvis/) is generated using the web tool ClustVis representing hierarchal clustering of the associations between serum RNA expression and the selected clinical traits (41). It was a standardization with a mean at 0 and a standard deviation at 1. The differences were set as |FC| > 1.2 and p-value <0.05. Gene expression levels were illustrated as Z-scores (blue to red). Individuals are in rows, while transcripts are in columns. Blue color indicates lower expression while red color indicates high expression. Light brown color indicates downregulation while dark red color indicates upregulation ( Figure 5 ).

The level of expression of the investigated RNA panel was assessed in terms of fold-change (RQ) values in the three investigated groups (T2DM group, prediabetes group, and healthy control group) aiming to confirm the retrieved bioinformatics data. TMEM173 and CHUK mRNAs, hsa-miR (-611, -5192, and -1976) miRNAs expression levels increased gradually from the control group to the pre-diabetic group to reach the highest levels in the T2DM group (p < 0.001), while RP4-605O3.4 and AC074117.2 lncRNAs expression levels decreased gradually from the control group to the pre-diabetic group to reach the lowest levels in the T2DM group (p < 0.001). Thus, the pattern of expression of the investigated STING/NOD/IR RNA associated panel was discriminative between prediabetes and T2DM groups compared to healthy controls and between prediabetes and T2DM patients ( Table 1 and Figure 6 ).

ROC curve analysis was performed for the investigated STING/NOD/IR RNA associated panel to discriminate between healthy controls and the prediabetes group and to determine the best cutoff values. As regards TMEM173 and CHUK mRNAs, the ideal cutoff values were ≥1.80 (AUC=0.935) and ≥2.05 (AUC = 0.959), sensitivities were 95.9% and 93.9%, specificities were 86.7% and 93.3%, respectively. For serum hsa-miR (-611, -5192, and -1976) the best cutoff values were ≥1.018 (AUC = 0.985), ≥0.1654 (AUC = 0.992) and ≥0.0165 (AUC = 0.803), sensitivities were 100%, 98.0%, and 81.6%, specificities were 95.6%, 95.6%, and 73.3% respectively. Concerning serum RP4-605O3.4 and AC074117.2 lncRNAs the best cutoff values were ≤95 (AUC = 0.931) and ≤97.5 (AUC = 0.836), sensitivities were 95.9% and 85.7% specificities were 82.2% and 73.3%, respectively. Thus, the selected STING/NOD/IR RNA panel was associated with the prediabetes stage. Indeed, recognizing patients at risk for insulin resistance and prediabetes will assist clinicians to determine the optimal management strategy before developing T2DM ( Table 2 and Figures 7 , 8 ). Interestingly, the chosen RNAs could be used by clinicians to identify prediabetic patients especially those with family history of T2DM during routinely applied checking up.

ROC curve analysis was also performed for the investigated STING/NOD/IR RNA associated panel to discriminate prediabetes group and T2DM group and to determine the best cutoff values. For serum TMEM173 and CHUK mRNAs the best cutoff values were ≥ 12.75 (AUC = 0.916) and ≥ 17.74 (AUC = 0.816), sensitivities were 89.4% and 75.8% specificities were 89.8% and 73.5% respectively. As regards serum hsa-miR (-611 and -1976 miRNAs the best cutoff values were ≥ 105.39 (AUC = 0.914) and ≥ 0.88 (AUC = 0.912), sensitivities were 89.4% and 97.0%, specificities were 93.9% and 85.7% respectively. And for serum RP4-605O3.4 lncRNA the best cutoff value was ≤ 0.74 (AUC = 0.957), sensitivity was 92.4% specificity was 98.0%. These results suggest that the selected STING/NOD/IR RNA associated panel could be used as a tool to differentiate between prediabetes and type 2 diabetes cases ( Table 2 and Figures 8 , 9 ).

Finally, ROC curve analysis was performed for the investigated STING/NOD/IR RNA associated panel to discriminate T2DM group and healthy controls. As for TMEM173 and CHUK mRNAs the best cutoff values were ≥4.005 (AUC = 0.979), and ≥ 3.64 (AUC = 0.981), sensitivities were 93.9% and 97.0%, specificities were 97.8% and 97.8%, respectively. For serum hsa-miR (-611, -5192, and -1976) the best cutoff values were ≥1.10 (AUC = 0.986), ≥0.0812 (AUC = 0.994) and ≥0.695 (AUC = 0.957), sensitivities were 97.0%, 98.5%, and 98.5%, specificities were 95.6%, 95.6%, and 93.3%, respectively. While for serum RP4-605O3.4 and AC074117.2 lncRNAs the best cutoff values ≤6.45 (AUC = 0.993) and ≤25.5 (AUC = 0.916), sensitivities were 97.0% and 89.4%, specificities were 97.8% and 84.4% respectively. These findings suggest the potential use of this panel dysregulation is associated with T2DM pathogenesis ( Table 2 and Figures 8 , 10 ).

Patients with T2DM were subdivided into two groups according to their glycemic control. Good glycemic control was indicated by HbA1c <7%, while poor glycemic control was indicated by HbA1c ≥7% (42). There was clinical significance without statistically significant upregulation in the levels of expression of hsa-miR-611 and clinically significant downregulation in the expression of CHUK, hsa-miR (-611, and -5192) and RP4-605O3.4 in the individuals with poor glycemic control compared to individuals with good glycemic control (p = 0.040 and 0.010, respectively) suggesting possible use of these biomarkers together with HbA1c in discriminating glycemic control in T2DM patients. ( Table S5 ).

The degree of insulin resistance was examined in relation to the expression of the insulin resistance associated panel. The prediabetic and T2DM groups were redivided into two groups; group which was insulin resistant (HOMA-IR ≥3.8) and the another was insulin sensitive (HOMA-IR <3.8) (43, 44). TMEM173, hsa-miR (-611 and -1976) and RP4-605O3.4 showed a significant difference in their expression ( Table S6 ), which makes the use of these insulin resistance associated biomarkers together with HOMA-IR possible in differentiating insulin resistant from insulin sensitive patients. Multivariate analysis for prediction of insulin resistance has been used and showed that miR-611 then -1976 then CHUK were significant insulin resistance predictors beside HOMA-IR and HOMA-B ( Table 3 )

The identified STING/NOD/IR RNA associated panel was constructed through the following steps:

Retrieval of a set of candidate mRNA genes (TMEM173 and CHUK) related to insulin resistance signaling pathways (STimulator of Interferon Gene and NLR pathways) from public microarray databases available at National Center of Biotechnology Information GEO (https://www.ncbi.nlm.nih.gov/geo/, accessed in May 2022), Gene atlas expression database; Verification of the expression of the retrieved candidate genes in skeletal muscles and adipose tissues via Gene Cards Human Genes database (https://www.genecards.org/), and via the Human protein atlas database (https://www.proteinatlas.org/) so as to decrease the false discovery rate; gene ontology was checked for TMEM173 and CHUK mRNAs using Enrichr (http://amp.pharm.mssm.edu/Enrichr) at p-value <0.05 which was defined to identify up-regulated and down-regulated genes in GO functional enrichments ( Supplementary Table S 7). Interactions of differently expressed biomarkers in IR and T2DM were analyzed via STRING database available at (https://string-db.org/). Identification of ncRNA (miRNA/lncRNA) genetic axis linked to the retrieved candidate mRNA genes in insulin resistance including hsa-miR (-611, -5192 and -1976) miRNAs from miRWalk (http://mirwalk.umm.uni-heidelberg.de) and miRBASE databases available at (https://www.mirbase.org/). RP4-605O3.4 and AC074117.2 lncRNAs retrieved from DIANA Tools database available at (https://diana.e-ce.uth.gr/lncbasev3); these lncRNA–miRNA interactions were obtained through target gene prediction with confirmation of its expression in the interested tissues through NONCODE database and LNCipedia database available at (http://www.noncode.org/ and https://lncipedia.org/) respectively. Alignment between hsa-miR-611 with RP4-605O3.4 and AC074117.2 lncRNAs through EMBL’s European Bioinformatics Institute database; Claustal omega database available at (https://www.ebi.ac.uk/Tools/msa/clustalo/), Detailed bioinformatics are illustrated in Figure 1 and Supplementary Figures S1–9 .

The current study was approved by the Ain Shams ethical committee, Faculty of Medicine. Participants were admitted to the Endocrinology Unit, Internal Medicine Department, Faculty of Medicine, Ain Shams University in the period between April 2022 and May 2022. There were 160 Egyptian unrelated participants. They were classified into 66 cases with type 2 diabetes mellitus, 49 were prediabetics with impaired glucose tolerance state, and there were 45 healthy volunteers came at the Ain Shams hospital outpatients’ clinics for routine checkups. A written informed consent was received from all the participants after being informed about the study. The diagnosis and classification of the diabetic and prediabetic state was made according to American Diabetes Association criteria 2022 for the diagnosis of type 2 diabetes mellitus and prediabetes (57).

Criteria for diagnosing diabetes and prediabetes was made based on the clinical presentation along with biochemical analyses, including fasting and 2 h post prandial blood glucose level, level of fasting insulin, glycated hemoglobin (HbA1c), lipid profile (total cholesterol, triglycerides, LDLc and HDLc), and albumin/creatinine ratio. HOMA-IR formula was calculated to assess the degree of tissue resistance to insulin. Upon hospital admission blood samples were collected and processed by centrifugation at 4000 rpm for 20 min. Resulting sera were kept in aliquots in a freezer at −80°C for further processing within 2 months.

Total RNA was extracted and purified using miRNEasy extraction kit (Qiagen, Hilden, Germany) from the serum samples according to the manufacturer’s manual. Further assessments of RNA concentration, integrity and purity were performed using Nano-Drop instrument (Thermo Scientific, Waltham, MA, USA). miScript II RT Kit (Qiagen, Germany) was used for the preparation of cDNA libraries for mRNAs, lncRNAs and miRNAs. We added 4 µl of 5x miScript HiFlex Buffer, 2 µl of 10x miScript Nucleics Mix, 2 µl of miScript Reverse Transcriptase Mix and continued with RNase free water to 2 µg of extracted RNA to reach final volume of 20 µl, followed by incubation at 37°C for 60 minutes and at 95°C for 5 minutes using Rotor gene Thermal cycler (Thermo Electron, Waltham, MA, USA) then cDNA was stored at -20°C for further processing by real-time PCR.

The levels of the identified TMEM173 and CHUK mRNAs in serum were measured using QuantiTect SYBR Green PCR kit (Cat. no. 204143, Qiagen, Hilden, Germany) with specific primers for Hs_TMEM173_1_SG QuantiTect Primer Assay (NM_198282) and Hs_CHUK_1_SG QuantiTect Primer Assay (NM_001278) with 500 ng of cDNA added from the undiluted RT reaction to the individual PCR tube, alongside with Hs_GAPDH_1_SG QuantiTect Primer Assay (NM_002046) as the reference gene following the manufacturer’s protocol. Relative expression levels for RP4-605O3.4 and AC074117.2 lncRNAs were analyzed by QuantiNova Probe PCR Kit (Cat. no. 208252; Qiagen, Helman Germany) and QuantiNova LNA Probe PCR Assay for Human RP4-605O3.4 (ENST00000548468) and AC074117.2 (ENST00000417130) supplied by Qiagen. HS_HGDC_2467750 QuantiNova LNA PCR Probe (GeneGlobe ID: UPFH1126608) was used as Reference Assay following the manufacturer’s protocol.

The relative expression levels of hsa-miR (-611, -5192 and -1976) miRNAs in serum were investigated by a miScript SYBRGreen PCR Kit (Cat. no. 218073; Qiagen, Helman Germany), a miScript universal primer and a miRNA‐specific forward primers of (Hs_miR-611 miScript Primer Assay) (Accession: MIMAT0003279) for hsa-miR-611, (Hs_miR-5192 miScript Primer Assay) (Accession: MIMAT0021123) for hsa-miR-5192 and (Hs_miR-1976 miScript Primer Assay) (Accession: MIMAT0009451) for hsa-miR-1976 with RNU‐6 was used as an internal reference control gene. All PCR primers were obtained from Qiagen, and the details of the primers used are supplied in Supplementary Table S8 . The PCR program cycling conditions were adjusted according to the type of the measured RNA by the real time cycler Applied Biosystems 7500 (Cat. no. 4351105) according to the manufacturer’s protocol. Tenfold serial dilutions of cDNA (100 ng to 1 pg) were analyzed in duplicate to create a standard curve with at least five points using the PCR Kit with specific primers. We calculate the log of each sample dilution followed by getting the slope of the regression between the log values and the average Ct values. A PCR efficiency of 96% over six logs of template dilution was achieved, and as little as 1 pg of transcript was sensitively detected.

To ensure reproducibility of the assay, dilutions were prepared fresh and PCR reactions were carried out in triplicate experiments. The limit of detection was defined as the lowest dilution of cDNA in which all three replicates resulted in positive amplification. Threshold detection by using Applied Biosystems 7500 software v2.3 (51) that verify all the samples within geometric phase so delta Ct between any two samples remain constant. Applicants specifically were ensured by melting curve analysis. Negative non template control containing master mix and molecular grade water has been used to check any contamination. Melting curve was used to assess primer specificity. The 2−ΔΔCt technique was used to measure the expression of the IR specific RNA‐based candidate genes panel (58). genes were used as an internal control to normalize the raw data of the samples and compare these results to a reference sample. In this study, appropriate standardization strategies were carried out to recognize any experimental error introduced at any stage during extraction and processing of the RNA according to MIQE guidelines (59) ( Table S9 ).

The software package of statistical analysis version 24 (SPSS24, IBM, Chicago, IL, USA) was used to statistically analyze the output data. Quantitative variables were analyzed using the median and the interquartile range (IQR) for the non-parametric data, and the mean ± SD and standard error were used for the raw symmetrically distributed numerical data. One-way ANOVAs and Mann–Whitney U test were also applied as appropriate. chi-Square tests were used to evaluate qualitative variables in number and percentage. Spearman correlation tests were applied to assess the correlations between quantitative variables. The Receiver Operating Characteristic (ROC) curve was used to evaluate the predictive value of the IR associated panel among investigated groups, in order to estimate the best cut-off points for different parameters with optimal sensitivities and specificities. Multivariate analysis was performed to determine the predictive power of different biomarkers for Insulin resistance and prediabetes. A p-value <0.05 was considered statistically significant.