Ethical compliance
This study was approved by the Institutional Review Board (IRB) for the protection of human subjects in Sidra Medicine, Qatar (IRB reference number 1702007592). Informed consent and assent was taken from patients and legal guardians as required. All experiments were performed in accordance with relevant guidelines and regulations.
Patient recruitment
In this prospective study every child with diabetes (aged 0-18 years), attending the diabetes clinics, or admitted as an inpatient in Sidra Medicine (which is the only paediatric diabetes center in Qatar) from 2018-2020 was recruited. Clinical details about birth history, gestational age, ethnicity, age of onset of diabetes, family history, BMI, weight, signs of insulin resistance (acanthosis nigricans) and other system involvement were collected and documented. Peripheral blood samples were collected for complete antibody profiling for-all 4 autoantibodies namely Glutamic Acid Decarboxylase 65 (GAD65), Insulin Autoantibody (IAA), Islet Antigen-2 Auto Antibody (IA-2A) and Zinc Transporter 8 (ZnT8A) were measured and titres recorded, C-peptide, celiac and Thyroid Peroxidase (TPO) antibodies were also measured as well as for DNA extraction. Patients were classified as type 1 diabetes mellitus after consulting the clinical history and antibody assays results, based on ADA guidelines. Patients initially thought to be type 1 diabetes mellitus with all autoantibody negative were reclassified as type 1b diabetes[18]. 100 control patients without diabetes were also recruited and autoantibody testing done for validation.
Antibody assay methodology
Complete antibody profiling of every child with diabetes was performed and titres recorded. This was done at the time of recruitment into the study for all known type 1 diabetes mellitus patients. Newly diagnosed cases were tested at the time of diagnosis before starting insulin treatment.
GAD65-Radioimmunoassay performed. (125) I-labeled recombinant human glutamic acid decarboxylase (GAD65) is incubated with the patient's diluted serum. Antihuman IgG and IgM are then added to form an immunoprecipitate. After washing the precipitated immune complexes, specific antibodies are detected by counting gamma-emission from the pellet's bound (125)I-GAD65[19].
Insulin autoantibody-Radioimmunoassay performed. (125) I-labeled recombinant human insulin is added to the test serum, if antibody is present, it forms a soluble complex with the labeled insulin. Subsequent addition of goat antihuman IgG and IgM precipitates the complex. The amount of radioactivity in the precipitate is proportional to the level of antibody in the serum[20].
IA-2 autoantibody- Radioimmunoassay performed. (125) I-labeled recombinant human IA-2 is added to the test serum, if antibody is present, it forms a soluble complex with the (125) I-labeled IA-2. Subsequent addition of goat antihuman IgG and IgM precipitates the complex. The amount of radioactivity in the precipitate is proportional to the level of antibody in the serum[21].
Zinc Transporter 8 (ZnT8) autoantibody- Enzyme immunoassay. Zinc Transporter 8 (ZnT8) antibodies are principally directed against the C terminal domain of ZnT8. The ZnT8 autoantibody ELISA is based on the bridging principle that employs the ability of divalent ZnT8 autoantibodies to bind to ZnT8 coated on to the plate well with one arm, and to liquid ZnT8-biotin with the other arm. Calibrators or undiluted serum samples in duplicate are added to ZnT- coated plate wells and incubated overnight. ZnT8-biotin is added to each well and plate. After another incubation, aspiration, and wash, streptavidin-peroxidase is added to each well. Another incubation, aspiration, and wash are performed and peroxidase substrate is added. After a final incubation, 0.5mol/L H2S04 stop solution is added to each well. Absorbance is measured at 450nm, blanked against wells containing peroxidase substrate and H2S04 only[22].
Genetic testing methodology
DNA samples were extracted from peripheral blood specimen of all individuals recruited into the study including patients and parents in Sidra Medicine, Qatar. Whole Exome Sequencing was used on a set of control subjects and selected patient samples on the basis of autoantibody status, family history of DM and associated clinical features. The 74 patients and 108 control subjects were sequenced on Illumina HiSeq platform using a 150-base paired-end single-index-read format. Reads in FASTQ files were then mapped to the NCBI human reference genome GRGh37/hg19 using Burrows–Wheeler Aligner (BWA-MEM) version 0.7.8. All subjects underwent variant calling using GATK(v3.6) and annotation was performed using SNPEff[23]. Variants file was normalized and decomposed using vt[24]. Additionally, vcfanno[25] was used to annotate VCF file with extensive available data resources like gnomad, exomes.r2.0.2, gnomad.genomes.r2.0.2.sites, 1K genome, Exac etc. Genomic variants belonging to 49 genes already known to be implicated in type 1 diabetes mellitus (see supplementary table 1) were extracted for patient samples from the multisample VCF file. These variants were further filtered for non-exonic regions using exome bed (Exome-Agilent_V6) and only variants belonging to exome regions were retained for downstream analysis. We looked for the non-synonymous variants which are present in more than 5% in our patient cohort and absent or present in frequency less than 0.1% in public data bases. These variants were considered to be present in a significantly higher frequency in our disease cohort.
Extraction of HLA data from WGS
We used Population Reference Graph (HLA-PRG)[26] to identify HLA alleles from the whole exome sequencing of 182 samples comprising of 74 type 1 diabetes mellitus patients and 108 normal controls.
The accuracy of HLA genotypes was assessed on a family-based approach, using 18 full trios (both parents and at least one offspring), and 4 families with only one single founder.
For disease association analysis, the patient and control cohorts were divided into two groups based on the ethnicity of the sample. The association of HLA alleles with type 1 diabetes mellitus phenotype in each group was tested using fisher’s exact test in R and the enrichment of the alleles in cases or controls was assessed using odds-ratio at a p value cut-off of 0.05.