A disintegrin and metalloproteinase (ADAM)-12, is a type-I transmembrane, Zn²⁺-dependent metalloprotease that belongs to the Metzincins superfamily of metalloproteases [1]. It is composed of multiple domains, including an N-terminal signal sequence, a prodomain, a metalloprotease domain, a disintegrin-like domain, a cysteine-rich region, and an epidermal growth factor (EGF)-like domain, followed by a transmembrane domain and a cytoplasmic tail [2]. ADAMs are synthesized as zymogens that are activated through prodomain cleavage by Furin [3]. Their physiological role involves ectodomain processing of membrane-bound proteins resulting in modulation of growth factors’ activity, cytoskeletal reorganization, and cell-cell and cell-extracellular matrix (ECM) contacts through their proteolytic and adhesive potentials [4]. These functions are critical in various biological and pathological processes, including development, inflammation, cancer, and Alzheimer’s disease [5]. ADAM12 may also regulate cellular signalling through its cytoplasmic tail, which contains potential recognition motifs for signalling and adaptor proteins; thus, it may regulate its metalloprotease activity and ADAM12’s subcellular localization [6, 7].

ADAM12 is expressed in various tissues, including adipose tissue [8], where it was shown to regulate adipogenesis using knockout mice models [8–12]. While these mice appear phenotypically normal at birth, approximately 30% of them die during the first week of life [11], and some exhibited abnormalities in interscapular brown adipose tissue and in the skeletal muscles surrounding the brown adipose tissue during embryogenesis. The impairment in brown adipose tissue causes failure of thermoregulation and accounts for the mortality rate in newborns [4]. Conversely, overexpression of ADAM12 increased adipogenesis in mice via multiple pathways, including its proteolytic activity and its interaction with the ECM [10]. Likewise, ADAM12 was found to promote new adipocyte formation in high-fat diet-induced obesity [12], as it is most abundant in preadipocytes that are responsive to adipogenic differentiation [8]. Since ADAM12 can promote adipocyte formation leading to increased fat mass and altered adipose tissue function, it can indirectly influence insulin resistance, which is a key feature in metabolic syndrome and type 2 diabetes (T2D).

Kuwait ranks among the countries with the highest and fastest-growing diabetes burden in the world, with almost 40% of adults affected by prediabetes or T2D as of 2018 [13]. Contributing to this alarming trend, around 80% of Kuwaiti adults were reported to be overweight or obese in 2014 [14], a major risk factor for the development of insulin resistance and metabolic syndrome. This in turn, increases the risk of T2D and cardiovascular diseases development [15, 16].

The high rates of consanguinity in Kuwait and other Arab populations allowed several studies to identify a strong familial and genetic basis for insulin resistance and metabolic syndrome in these populations [17, 18]. Such a population structure increases the prevalence of recessive and familial disorders, including those affecting metabolic traits [19–24]. For example, a recent study identified regions within the TNKS genes associated with T2D, obesity, and hypertension in Kuwaitis [25]. In support of these conclusions, genome-wide association (studies (GWAS) carried out on Kuwaiti cohorts have revealed novel genetic variants associated with metabolic traits that have not been documented in other populations, indicating the existence of distinct, population-specific genetic risk profiles [26].

Moreover, familial genetic studies and GWAS support the genetic basis of insulin resistance and individual components of the metabolic syndrome [27–34]. More specifically, studies have consistently reported that a significant amount of genetic variation affects b-cell function and insulin secretion [35]. The integration of these variations with pancreatic islet epigenomic and transcriptomic datasets enabled the identification of potential effector genes, such as ACSL1 and FAM46C involved in insulin release [36]. Notably, most of the genetic variants associated with b-cells’ dysfunction are mainly located within non-coding regulatory elements, emphasizing the importance of transcription regulatory mechanisms in the genetic framework of T2D susceptibility and supporting the hypothesis that b-cell dysfunction is a key mechanism in T2D pathogenesis [36]. Due to ADAM12's documented role in modulating adipogenesis, we examined its correlation with metabolic syndrome traits in a GWAS conducted at our institute. This study identified four ADAM12 variants, namely ADAM12-K170R rs112264074 [NP_003465.3:p.Lys170Arg], ADAM12-R176W rs140497576 [NP_003465.3:p.Arg176Trp], ADAM12-M662V rs115100580 [NP_003465.3:p.Met662Val], and ADAM12-I908V rs41303603 [NP_003465.3:p.Ile908Val] that associated with various metabolic traits. We also validated these findings in a separate replication study and examined the impact of the identified variants on enzyme kinetics.

A discovery GWAS was conducted as part of the Kuwait Obesity Genome Project (KOGP) performed at Dasman Diabetes Institute (DDI) [37, 38]. Briefly, this study included 1298 randomly selected Arab adults from the six governorates of Kuwait. The participants were recruited according to procedures authorized by the scientific and ethics advisory boards at DDI. Multiple anthropometric (body mass index (BMI) and waist circumference) and metabolic biomarkers including fasting blood glucose (FBG), Haemoglobin A1c (HbA1c), total cholesterol (TC), triglycerides (TGs), high-density lipoprotein (HDL), low-density lipoprotein (LDL), systolic and diastolic blood pressure (SBP and DBP) were ascertained in the study in relation to the assessed variants. Resultant variants that were significantly associated with biomarkers relevant to metabolic syndrome traits were assessed in a replication study. The replication study participants were selected from the Kuwait Diabetes Epidemiology Program (KDEP) at DDI following the guidelines of the scientific and advisory boards at DDI. These participants were randomly selected to represent the adults of Kuwait using a stratified random sampling technique from a computerized register of the Public Authority of Civil Information, a depository of personal information on Kuwaiti citizens and expatriates [39].

The discovery phase inclusion criteria adhered to established methodological standards used in large-scale GWAS including verification of Arab ethnicity by a questionnaire about parental lineage, which considered parental lineages up to the third generation to mitigate population stratification effects as recommended by Petersen et al. [40]. On the other hand, the replication study allowed the inclusion of Arab, South Asian, and Southeast Asian ancestries. In accordance with standard GWAS methodology our exclusion criteria for both the discovery and the replication studies included age < 18 years, the presence of limiting serious complications of diabetes, mental illness, or cognitive limitation. Furthermore, pregnant women and patients undergoing weight reduction medication/surgery or fitness programs were also excluded [39, 41, 42]. Before blood sample collection, the patients fasted overnight and signed an informed consent form. Simultaneously, their SBP and DBP were measured, and the nationality and ethnicity of each participant were confirmed using a detailed questionnaire. Participants were classified as having T2D based on self-reported diagnosis or in accordance with American Diabetes Association (ADA) criteria using FBG ≥ 7 mmol/l; 126 mg/dL, and HbA1c ≥ 6.5% measurements, while healthy controls were free of diabetes. Other illnesses, including cardiovascular complications, were verified and recorded. The clinical assays conducted for trait outcome measurements were performed at a College of American Pathologists (CAP) accredited laboratory at DDI [38].

All protocols and procedures conducted in this study were reviewed and approved by DDI Ethical Review Committee, which adheres to the guidelines of the Declaration of Helsinki and the US Federal Policy for the Protection of Human Subjects (ERB/ERC numbers: RA HM 2010-005 for KOGP and RA HM 2010-004 for KDEP). Participants were categorized into healthy control and T2D patient groups based on clinical results. This classification was preserved for the examination of genotype associations.

The discovery GWAS was conducted on the Illumina Human Omni-Express and the iSCAN system (Illumina) using the Infinium®HD ultra genotyping assay method. Genotype calling was performed after pooling the intensity data of all genotyped samples.

A prospective power analysis was conducted using the discovery GWAS frequencies for the four variants with the assumption that the diabetic cohort would have higher frequencies than the healthy controls in the replication cohort. We found that a sample size of 400 in each group has a 90% power to detect an increase of 0.03 with a significance level (alpha) of 0.05 (two-tailed) for ADAM12-K170R, ADAM12-R176W, and ADAM12-M662V. As for ADAM12-I908V, a sample size of 400 in each group has a 90% power to detect an increase of 0.05 with a significance level (alpha) of 0.05 (two-tailed).

The replication phase genotyping was conducted by targeted genotyping of the selected ADAM12 single nucleotide polymorphisms (SNPs), including ADAM12-K170R (rs112264074 [NP_003465.3:p.Lys170Arg]), ADAM12-R176W (rs140497576 [NP_003465.3:p.Arg176Trp]), ADAM12-M662V (rs115100580 [NP_003465.3:p.Met662Val]), and ADAM12-I908V (rs41303603 [NP_003465.3:p.Ile908Val]). For simplicity, the variant amino acid change will be used throughout the remaining text. These missense variants were prioritized for further investigation based on their associations with metabolic syndrome and T2D-related traits, as well as their predicted functional consequences and biological significance to metabolic regulation pathways. The TaqMan® SNP genotyping assays (Applied Biosystems, MA, USA) were run on a QuantStudio 5 real-time polymerase chain reaction (PCR) system (Applied Biosystems, MA, USA). Genotyping PCR reactions contained 20 ng of genomic DNA, 10µl of 2X TaqMan Genotyping master mix, 1µl of 20X TaqMan SNP genotyping assay, and nuclease-free water to make up the volume to 20µl. The thermal cycling conditions included a pre-PCR read at 60°C for 1 minute followed by polymerase activation at 95°C for 10 minutes and 40 cycles of denaturation at 95°C for 15s and annealing/extension at 60°C for 1 minute. Finally, a post-PCR read was performed at 60°C for 1 minute, and genotype calls were determined. Sanger sequencing of PCR products for selected cases of homozygotes and heterozygotes was conducted to confirm the obtained genotypes. Sequencing reactions were conducted using the BigDye terminator cycle sequencing kit v.3.1 (Applied Biosystems, MA, USA) according to the manufacturer’s instructions on an ABI PRISM 3730xl genetic analyser (Applied Biosystems, MA, USA).

The ADAM12 variants identified were mapped, modelled, and analysed using PyMOL modelling software (The PyMOL Molecular Graphics System, Version 3.0 Schrödinger, LLC). To examine the impact of the genetic variants on ADAM12 stability and flexibility, the DynaMut web server (https://biosig.lab.uq.edu.au/dynamut/ ) was used. Both analyses were based on the predicted crystal structure (UniProt accession ID O43184).

Human-tagged ORF clone plasmid encoding ADAM12 was purchased from OriGene Technologies (RG212501, MD, USA). Site-directed mutagenesis of studied SNPs was conducted using the QuickChange II site-directed Mutagenesis kit (Agilent Technologies, CA, USA) according to the manufacturer’s instructions. Sanger sequencing was conducted to verify successful incorporation of the intended nucleotide substitutions. The sequencing reactions were conducted using the BigDye terminator cycle sequencing v.3.1 kit (Applied Biosystems, MA, USA) according to the manufacturer’s instructions on an ABI PRISM 3730 xl genetic analyser (Applied Biosystems, MA, USA). The primers used to generate the nucleotide substitutions and to verify them by sequencing are listed in Table 1.

HEK293 cells were purchased from ATCC (CRL-1573, VA, USA). The cells were grown in Dulbecco’s modified Eagle’s medium (DMEM) (Invitrogen Corporation, CA, USA) supplemented with 10% fetal bovine serum (FBS) (Invitrogen Corporation, CA, USA) and 100 U/ml of penicillin and 100 µg/ml streptomycin (Gibco, CA, USA) and incubated at 37°C in a humidified atmosphere containing 5% CO₂. The cells (2x10⁵ or 5x10⁴) were seeded into 6- or 96-well plates, respectively, for 24 hours prior to transfection. HEK293 cells in 6- or 96-well plates were transfected with 2 µg or 200 ng, respectively, of wild-type (WT) or variant constructs using lipofectamine 3000 reagent (DNA:lipofectamine ratio 1:2) (Invitrogen, MA, USA) in an antibiotic-free medium. Twenty-four hours after transfection, the medium was changed to complete DMEM medium, in which cells were grown for another 24 hours before being harvested for protein analysis. Overexpression of WT and variant proteins was confirmed using western blotting.

For western blot analysis, the cells were harvested and lysed in an extraction buffer containing 0.5% NP-40 and 2% CHAPS in phosphate-buffered saline (PBS) (Gibco, CA, USA). The lysates were centrifuged at 4°C for 15 minutes at 13,000 rpm. The protein concentration in the supernatant was quantified using the Bradford assay (Thermo Fischer Scientific, Driesch, Germany). Then, 5 µg of total protein was resolved on 8% denaturing polyacrylamide gel and transferred to a nitrocellulose membrane (EMD Millipore Corporation, MA, USA). Membranes were stained with Ponceau S stain to verify proteins transfer, then it was cut into two pieces to allow probing for multiple targets separately (Figure S1). After that, the membranes were blocked using 5% fat-free milk and blotted with the corresponding primary and horseradish peroxidase (HRP)-linked secondary antibodies. The primary antibodies used were b-actin (ab3700, Abcam, MA, USA), and ADAM12 (ab223745, Abcam, MA, USA). Chemiluminescent detection of HRP activity was done using a chemiluminescent substrate (Thermo Scientific, MA, USA) according to the manufacturer’s instructions, and the membranes were visualized using a ChemiDoc Imaging system and analysed by Image Lab software (Bio-Rad, CA, USA).

The calibration fluorescent peptide (5`-FAM-Pro-Leu-Arg-Arg-Thr-Leu-Ser-Val-Ala-Ala-OH) was purchased from CPC Scientific (Sunnyvale, CA, USA), while the ADAM12 fluorogenic substrate PEPDAB005 (Dabcyl-LAQAPhe(homo)RSK(5-FAM)-NH₂) peptide was purchased from BioZyme Inc. (St. Joseph, MO, USA). The calibration fluorogenic peptide was used at different concentrations (0–10 µM) to generate a fluorescent calibration curve. Cleavage assays of PEPDAB005 were performed in 96-well plates with 10 µg of the total protein lysate supernatant from cells transfected with either the WT or variant ADAM12 in a reaction buffer containing 1 mM CaCl₂, 10 mM NaCl, 20 mM Tris pH 8.0, and 0.5% Brij-35. Protein extracts from untransfected cells or cells transfected with a GFP-expressing plasmid were used as negative controls. Reactions were conducted at 37°C, and fluorescence emission was measured for a total of 3 hours at 5-minute intervals using a multimode microplate reader (Synergy H4 Hybrid, Bio Tek Instruments, VT, USA) with excitation and emission wavelengths of 485 nm and 530 nm, respectively. Western blot analysis was used to normalize enzyme activities and Lineweaver-Burk plot enzyme kinetic model was used to determine the maximum velocity (Vmax) and the Michaelis-Menten constant (Km) [43].

Clinical and demographic data were analysed using student’s t-test or the non-parametric Mann-Whitney test. Hardy-Weinberg equilibrium was assessed in the healthy control cohort for the four variants using expected genotype frequencies by χ²-test and linear regression analysis was performed to determine the allele effect size (b coefficient). The associations were considered statistically significant at p < 0.05. To correct for multiple comparisons, a Bonferroni correction was applied, adjusting the significance threshold to p < 0.0045. The ADAM12 activity assays were conducted using two biological replicates and the data are presented as mean ± standard deviation. Significance was tested using the student t-test and P-values ≤0.05 were considered statistically significant. Biological replicates refer to data obtained from different transfection reactions of cultured cells. All statistical analyses were performed using SPSS v.25 (IBM, NY, USA).

A cohort comprised of 1298 Arab participants that included healthy controls (n = 697) and T2D patients (n = 601) was used to conduct a discovery GWAS as part of the KOGP study at DDI (Table 2). This study showed a significant association of four ADAM12 variants, namely ADAM12-K170R, ADAM12-R176W, ADAM12-M662V, and ADAM12-I908V, with several metabolic traits (Table 3), including FBG, HbA1c, SBP, DBP, and lipids including TC, HDL, and LDL. The location of the variants on ADAM12 gene and protein are illustrated in (Fig. 1A).

The ADAM12 variants’ association with metabolic syndrome traits/T2D identified in the discovery cohort was verified in a separate cohort from the KDEP study, which included healthy controls (n = 409) and T2D patients (n = 450). This cohort consisted of adults of Arab, South Asian and Southeast Asian ethnicities in Kuwait. A summary of the demographic and clinical data of the replication cohort is summarized in Table 4. The minor allele frequencies (MAFs) of ADAM12-K170R and ADAM12-M662V variants were higher (0.031 and 0.044, respectively) in the replication cohort compared to those detected in the discovery cohort (0.003436 and 0.006873, respectively), as well as compared to the total MAFs reported in GnomAd database (0.001106, and 0.002337, respectively) (GnomAd v2.1.1). The MAF of the ADAM12-I908V variant, on the other hand, was lower in the replication cohort (0.018) compared to the discovery cohort (0.029), and the total MAF reported in GnomAd database (0.02375) (GnomAd v2.1.1). In contrast, the minor allele of the ADAM12-R176W variant was not detected among participants in the replication cohort. All variants were in Hardy-Weinberg equilibrium in the healthy cohort (P > 0.05).

Associations of ADAM12-K170R, and ADAM12-M662V detected in the discovery cohort were not confirmed in the replication cohort (Table S1), however, new associations were detected (Table 5). ADAM12-K170R was found to be associated with waist:hip ratio in T2D cohort (p = 0.044) suggesting a potential but modest association. ADAM12-M662V was found to be associated with HDL in both healthy control (p = 0.012) and T2D (p = 0.017) cohorts suggesting a moderately strong and potentially meaningful effect, which sustained after adjustment for confounding factors, including age, sex, and BMI (β = 0.098; 95%CI: 0.024–0.172; p = 0.009) and also after adjusting for ethnicity (β = 0.080; 95%CI: 0.014–0.145; p = 0.018) (Table S2). In contrast to the significant association of ADAM12-M662Vwith TC:HDL ratio (p = 0.005), HbA1c level showed a borderline association (p = 0.044) in the healthy control cohort. Also, ADAM12-M662V association with waist:hip ratio was revealed after adjustment for the confounding factors mentioned above (β = 0.447; 95%CI:0.120–0.774; p = 0.007) (Table S2). While most of these associations reached nominal significance ( p < 0.05), Bonferroni correction was applied to account for multiple testing (adjusted threshold: p < 0.0045). After correction, only the association between ADAM12-M662V and HDL and TC:HDL remained statistically significant, supporting the robustness of these findings. In addition, the association between ADAM12-I908V and FBG identified in the discovery cohort was confirmed in healthy control and T2D cohorts, with a stronger association observed in the healthy controls (p = 0.017) compared to the T2D group (p = 0.044). This association was sustained after adjustment for confounding factors (β = -0.007; 95%CI:-0.014- (-0.001); p = 0.02) and ethnicity (β = -0.009 ; 95%CI: -0.015 - -0.003; p = 0.004) (Table S3). Also, ADAM12-I908V showed a new association with HbA1c in the replication cohort (β = -0.010; 95%CI: -0.019–0; p = 0.04). When analysing healthy controls and T2D patients separately, ADAM12-I908V showed a significant association with LDL (p = 0.001) and TC (p = 0.006) in the T2D cohort, which remained statistically significant after Bonferroni correction. Further scrutiny of genotypes found three participants (0.35%) carried dual variants in the ADAM12 gene, specifically the ADAM12-K170R and ADAM12-M662V variants’in heterozygosity. Similarly, another three participants exhibited double-variants of the ADAM12-K170R and ADAM12-I908V alleles, while only a single participant (0.12%) carried the ADAM12-M662V and ADAM12-I908V variants. All variants were detected in heterozygosity; no homozygous individuals were detected for any of the assessed ADAM12 variants.

The variants detected in the ADAM12 gene, namely ADAM12-K170R, ADAM12-R176W, ADAM12-M662V, and ADAM12-I908V are missense variants that cause the substitution of an amino acid in the prodomain (170R, 176W), EGF-like domain (662V) and the cytoplasmic (908V) tail (Fig. 1A). The detected variants 170R, 662V and 908V affect amino acid residues that are conserved across species (Fig. 1B, C, and D), whereas the variant 176W affects an amino acid that is not conserved across species (Fig. 1B).

Pymol and DynaMut software assessed how these variants impact the ADAM12 structure’s stability and flexibility (Fig. 2). The assessment predicts that 170R, 662V, and 908V variants destabilize the protein and increase its flexibility, while the 176W variant stabilizes the protein and decreases its flexibility (Fig. 2A). The molecular mechanisms causing the observed changes in ADAM12 variants’ stability and flexibility were investigated by analysing the intramolecular interactions within the WT and variant protein structures. This analysis revealed that 170R variant may introduce new interactions with glutamic acid (Glu150 within the prodomain and Glu374 of the metalloprotease domain) residues that are absent in the WT (Fig. 2B). Similarly, the 176W variant stabilizes the proteins’ structure and possibly introduces new intramolecular interactions with nearby glycine and tyrosine residues [Gly385 and Tyr386] (Fig. 2C). In contrast, 662V variant may reduce the intramolecular interactions providing the enzyme with increased flexibility (Fig. 2D). Finally, the structural analysis of 908V reveals the interaction of 908V with valine and proline [V899 and P900] that are in close proximity (Fig. 2E). Enzyme Activity and Kinetic Parameters of the ADAM12 variants

Since the detected ADAM12 variants are in the prodomain, the EGF-like domain, and the cytoplasmic tail, we evaluated the impact of these variants on ADAM12’s proteolytic function. ADAM12 WT and variants were generated through site-directed mutagenesis and expressed separately in HEK-293 cells. Western blotting confirmed that these proteins were overexpressed in the HEK-293 cell lysates (Fig. 3A). Enzymatic assays using the fluorogenic substrate PEPDAB005 revealed that ADAM12 variants 176W and 662V demonstrated a significantly increased proteolytic activity compared to ADAM12 WT (176W 55.4±7.2%; p = 0.013 and 662V 93.5±25.7; p = 0.045; Fig. 3B). No significant change in proteolytic activity was observed using 170R and 908V variants’ cell lysates’ supernatants. To exclude any potential interference of the intracellular environment with enzyme activity, HEK-293 cells seeded on a 96-well black plate were transfected with ADAM12 WT and variants as previously described. Enzyme activity was measured in intact cells, allowing the assessment of the enzyme activity present at the cell surface. The results showed that 170R significantly reduced the proteolytic activity of ADAM12 at the cell surface (18.9%; p = 0.018). Both ADAM12 662V and 908V variants showed a statistically significant increase in proteolytic activity, which was very slight for the 662V variant (0.5%; p = 0.006), while ADAM12 908V variant exhibited a much higher increase in activity at the cell surface (42.5%; p = 0.028). No significant change in the proteolytic activity of ADAM12 variant 176W was observed at the cell surface (Fig. 3C). Further, the effect of ADAM12 variants on its proteolytic activity was studied by determining the kinetic parameters, Vmax and Km. As illustrated in Fig. 3D&E, ADAM12 WT and variants followed the Michaelis-Menten equation, so the Lineweaver-Burk plot was used to calculate Vmax and Km values. Only 662V variant exhibited a significant reduction in Vmax value by 33.6% (p < 0.01) compared to the WT (Fig. 3E). Whereas the Km values of 176W and 662V variants were significantly reduced by 41.1% (p < 0.001) and 76.05% (p < 0.001) respectively, compared to WT (Fig. 3E).

ADAM12 plays an important role in various cellular functions and is associated with different pathological conditions [44]. Several GWA studies have identified loci associated with the risk of developing T2D [34, 45, 46]. However, most of the identified loci are associated with insulin secretion and B-cell function [34–36]. Our discovery GWAS identified the association of four novel rare/low-frequency (MAF ≤0.5%) ADAM12 variants, namely ADAM12-K170R, ADAM12-R176W, ADAM12-M662V, and ADAM12-I908V with different metabolic syndrome and T2D traits including FBG, HbA1c, lipids including TC, HDL and LDL, and hypertension (DBP and SBP). However, these findings were not confirmed in our replication study except for the association of ADAM12-I908V with FBG. Nonetheless, our replication study found new associations with other traits, including FBG, HbA1c, waist:hip ratio, and lipids (TC, HDL, LDL, and TC:HDL ratio). This study reveals an unprecedented identification of associations between ADAM12 variants and metabolic characteristics. Due to their rarity or low frequency, these variants are excluded from the genotyping arrays utilized in global GWA studies. However, the analysis of the GWAS catalogue, which presents GWAS findings from global studies, revealed associations between common variants of the ADAM12 gene and metabolic traits. Specifically, rs10901513 (an intergenic variant FANK1-ADAM12) with visceral fat in men, rs11244839 (an intergenic variant RNA5SP328-ADAM12) with BMI in 0.5-1.5-year-old children, rs7920091 (among others) with height, and rs1551678 associated with thiazide-induced adverse metabolic effects in hypertensive patients (affecting fasting blood glucose measurement). The findings from our research on associations with T2D traits, together with those from global GWA studies on related traits, validate the role of ADAM12 in metabolic processes. Despite the differences between the ethnic origins of the discovery and replication cohorts and their lifestyles, adjustment for ethnicity did not affect the observed associations between ADAM12 variants and the studied traits. However, the differences in MAFs of some of the reported ADAM12 variants in the replication cohort, compared to the discovery cohort, may still reflect underlying population differences [34]. Additionally, the biological significance and effect size of the identified variants may be more specific to the Arab ethnicity of the discovery cohort, which was diluted in the replication cohort as it had different ethnicities.

The identified ADAM12 variants result in amino acid changes within the prodomain (170R, 176W), the EGF-like domain (662V), and the cytoplasmic tail (908V). Thus, we expected that protein structure and function may be affected. The type of amino acid change in the two prodomain variants, 170R and 176W may have contributed to the difference in proteolytic activity in cell lysates and intact cells. The conserved change of lysine to arginine in the 170R variant did not affect the proteolytic activity in cell lysates but was decreased at the cell surface of intact cells. Although 170R variant possibly introduces new intramolecular interaction patterns, as observed in our structural analysis, these alterations did not significantly affect the enzyme’s proteolytic activity in cell lysates. Hence, the increased stability introduced by this alteration may not influence the active site architecture or substrate accessibility. However, these alterations could affect the prodomain’s function. It is established that the prodomain acts as an intramolecular chaperone facilitating proper protein folding [5] and regulating ADAMs’ intracellular traffic [47, 48]. Therefore, the increased stability of 170R variant due to the potential intramolecular interactions introduced between R170, E150 within the prodomain, and E374 of the metalloprotease domain, possibly hindered the translocation of ADAM12 to the plasma membrane, which was reflected by a reduced proteolytic activity at the cell surface.

Moreover, the ADAM prodomain maintains the enzyme in its latent form by a cysteine switch mechanism [5, 49]. The activation mechanism involves cleavage of the prodomain by the furin enzyme [50]. Therefore, the non-conserved amino acid change of arginine to tryptophan in the R176W variant may have increased the proteolytic activity of this variant in cell lysates by disrupting the prodomain-active site’s interactions while paradoxically increasing the overall protein stability. In the WT, the enzyme is maintained in an autoinhibited state through the interactions of C179 within the prodomain with the catalytic Zn²⁺ ion [51]. In contrast, the R176W variant introduces a bulky, hydrophobic tryptophan side chain [the indole ring] near the C179. The indole ring of the tryptophan may form van der Waals repulsive forces with neighbouring residues [G385, Y386], which possibly destabilize the local prodomain conformation and disrupt the inhibitory contact with the active site. This conformational alteration possibly increases substrate accessibility to the catalytic Zn²⁺ site consistent with the reduced Km value [7, 51, 52]. Interestingly, while this variant exhibits locally disruptive interactions, Dynamut analysis revealed increased overall protein structural stability. This apparent paradox may be explained by compensatory stabilizing interactions that cause local structural rearrangements, reducing steric hindrance and promoting better hydrophobic packing within adjacent regions of the protein core. This variant may also reduce the translocation of ADAM12 to the plasma membrane, reflected by a non-significant change in activity within the intact cells, despite the increased activity in cell lysates. Moreover, the M662V variant that occurs within the EGF-like domain showed a potential increase in flexibility and proteolytic activity. Although the function of the EGF-like domain is not fully understood [50], it may affect the conformational state of the enzyme and its ability to interact with the substrate. Especially that the reduced intramolecular interactions observed in structural analyses may affect the secondary structure of ADAM12, which plays a vital role in substrate recognition [53]; hence, increasing the flexibility of the protein may allow easier access of the substrate to the active site.

The increased proteolytic activity of the cytoplasmic tail variant, I908V at the cell surface of intact cells compared to cell lysates’ supernatants, may be caused by cytoplasmic tail modifications or interactions with different proteins leading to its translocation to the plasma membrane. Because the tail of ADAM proteins contains binding motifs for SH3-domain containing proteins and potential phosphorylation sites, it may modulate their function or provide binding sites for SH3-domain containing proteins [53–55]. The majority of ADAMs are located in the Golgi apparatus under steady-state conditions [56, 57], with only a minority found at the plasma membrane [58]. The translocation to the plasma membrane is induced by the phosphorylation of ADAM12 by PKCƐ [10, 59], whereas endocytosis of ADAMs depends on the binding of PACSIN3 to the SH3-binding domain [60]. Structural analysis suggests that, in the WT, I908 is possibly involved in an interaction with A906. However, the substitution of Ile with Val at position 908 of the variant possibly allows the interaction of V908 with V899 and P900, creating a local loop-like conformation, which exposes nearby serine and tyrosine [S902, Y903, Y905] residues, making them more accessible for potential phosphorylation. This conformational change may facilitate easier access for kinases or adaptor proteins such as Tks5/FISH [61], PI3K, and actinin-1/2 [7], thereby increasing the possibility of phosphorylation-dependent translocation to the plasma membrane. Hence, this structural rearrangement may partially elucidate the increased surface proteolytic activity observed for the I908V variant.

Although ADAM12 knockout mice exhibit normal glucose tolerance, the increased activity of the identified ADAM12 variants may contribute to the development of metabolic syndrome/T2D-associated traits such as obesity (Figure S2) [44]. While direct evidence associating ADAM12 to insulin resistance is lacking, several mechanistic links are plausible. Evidence supports the role of ADAM12 in the development of high-fat diet-induced obesity by stimulating new adipocyte formation (hyperplasia) rather than by enhancing differentiation and/or maturation (hypertrophy) of preexisting adipocytes via mechanisms involving its proteolytic activity and its interaction with b1 integrins [8, 10–12]. In addition, Coles et al., reported that the two most affected signalling pathways by ADAM12 knockdown included IGF/IGFBP/mTOR growth pathway and PPARγ signalling pathway, both of which are critical regulators of adipogenesis and insulin sensitivity. ADAM12 can cleave IGFBP3 and IGFBP5 [9]; therefore, increased ADAM12 activity may increase the bioavailability of IGF and stimulate preadipocyte proliferation. Hence, the increased ADAM12 variants’ proteolytic activity may enhance IGF bioavailability and subsequently stimulate preadipocyte proliferation. Therefore, the increased proteolytic activity of ADAM12 variants identified in this study may promote white adipose tissue (WAT) mass expansion. Besides, ADAM12 can remodel the cytoskeleton via b1 integrin and alter the dense fibronectin-rich matrix, allowing the flat preadipocytes to take a round shape during differentiation since fibronectin can be cleaved by ADAM12 [62]. The increase in WAT mass could act as a double-edged sword, while initial hyperplasia may transiently maintain insulin sensitivity, prolonged adipose tissue expansion often leads to hypertrophy, adipose tissue dysfunction, chronic low-grade inflammation, and ultimately, systemic insulin resistance development [63, 64]. However, while these mechanistic links are biologically plausible, direct evidence associating increased ADAM12 activity to insulin resistance and metabolic syndrome traits remains limited. Thus, the association of ADAM12 with metabolic disorders remains exploratory. To address this knowledge gap, further in vitro and in vivo studies that integrate functional metabolic phenotyping that assess metabolic syndrome and T2D traits in models overexpressing hyperactive ADAM12 variants are needed to elucidate how the increased ADAM12 variants’ activity reported herein may contribute to insulin resistance and T2D development. Moreover, evaluating how hyperactive ADAM12 may affect adipose tissue remodelling, inflammation, and insulin sensitivity may aid in elucidating its role in T2D pathogenesis. In vitro experiments studying the role of ADAM12-driven IGF/IGFBP axis modulation in adipocyte insulin sensitivity and glucose metabolism may also be of value. Altogether, these approaches will explain how the increased ADAM12 variants activity reported herein may contribute to the pathophysiology of insulin resistance and T2D.

There are several limitations to our study including the relatively small sample size in the two cohorts. However, despite the relatively small sizes of our discovery and replication cohorts, the differences in the ethnic origins of the participants, and their demographic and clinical characteristics we were able to identify rare ADAM12 variants and confirm their association with several metabolic traits relevant to T2D. Another limitation is that the replication results may have been influenced by the heterogeneity of the tested cohorts; with the replication cohort including multiple ethnicities, while the discovery cohort being limited to individuals of Arab descent, through differences in allele frequencies and genetic backgrounds. In addition, the limited sample sizes within individual subgroups impeded ethnicity-stratified analyses or adjustment for population structure, which would have been valuable, and their consideration remains important for future studies. Furthermore, detailed clinical information regarding diabetes duration, treatment, and end-organ complications was not available, which represents another limitation of this study. In addition, the absence of direct or surrogate insulin resistance measurements, such as Homeostatic Model Assessment of Insulin Resistance, HOMA-IR, which would provide a more accurate assessment of T2D status, may added an additional limitation. The use of HEK293 cells in this study, although advantageous for initial enzymatic characterization due to their high transfection efficiency and minimal endogenous ADAM12 activity, lack the physiological context of adipose or muscle cells. Therefore, further in vitro and in vivo studies using more relevant cellular models, such as primary adipocytes, or myoblasts are required to confirm their role in metabolic syndrome and T2D pathophysiology.

Our GWAS suggested the association of four ADAM12 gene variants (namely ADAM12-K170R, ADAM12-R176W, ADAM12-M662V, and ADAM12-I908V) with metabolic syndrome/T2D traits. Using a replication cohort, the association of ADAM12-I908V with FBG was confirmed, and more associations of the variants with different metabolic traits were found, confirming the potential role of ADAM12 variants in metabolic syndrome/T2D development. The functional study revealed a variant-specific difference in the activity of ADAM12 in cell lysates and the cell surface, suggesting that dysregulation of ADAM12 activity at the cell surface may play a role in the development of T2D. However, the exact mechanisms and tissues involved are not fully elucidated yet. Therefore, further studies are required to enable targeting ADAM12 for the potential treatment of these conditions.