Identification and characterization of novel PAX4 variants in patients with suspected MODY9

WHAT IS ALREADY KNOWN ON THIS TOPIC Pathogenic variants in the PAX4 gene are associated with maturity-onset diabetes of the young type 9 (MODY9), but many identified variants remain of uncertain significance due to limited functional data. Prior research has primarily focused on the impairment of its molecular function as a transcription factor.WHAT THIS STUDY ADDS This study identifies three novel PAX4 variants and provides experimental evidence that the novel p.Leu12Pro and p.Arg163Pro mutants lead to reduced protein levels through enhanced proteasomal degradation, establishing a new mechanism of β-cell dysfunction in MODY9.HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY The results underscore the importance of assessing protein-level consequences when interpreting PAX4 variants and support the incorporation of functional assays into genetic diagnostic workflows for monogenic diabetes, particularly in cases with atypical inheritance patterns.Introduction Maturity-onset diabetes of the young (MODY) is a monogenic form of diabetes characterized by early-onset hyperglycemia and autosomal dominant inheritance. 1 2 Although it accounts for only a small percentage of all diabetes cases,3–6 its true prevalence is likely underestimated due to frequent misdiagnosis.7 To date, 14 MODY subtypes have been identified, with MODY2 and MODY3 representing the most prevalent forms.8 Among these, MODY type 9 (MODY9) is a rarer subtype defined by pathogenic variants in the PAX4 gene.9 10PAX4 encodes a paired-box transcription factor essential for pancreatic endocrine development, particularly for establishing and maintaining β-cell identity.11–13 The PAX4 protein contains a conserved paired domain and a homeodomain, both of which mediate DNA binding and thereby regulate transcriptional repression of target genes.14 Structural studies of other PAX family members have demonstrated that these domains can also participate in protein-protein interactions, suggesting potential roles beyond direct DNA binding.15 16 During pancreatic development in mice, Pax4 is primarily expressed in endocrine progenitors and plays a critical role in directing β-cell lineage commitment by repressing the α-cell determinant Arx.17 18 In adult islets, its sustained expression in a β-cell subpopulation is essential for their survival and stress-induced compensatory proliferation.19Increasing evidence suggests that PAX4 deficiency can impair β-cell function. A key demonstration of this comes from Lau et al,20 who showed that human induced pluripotent stem cells harboring either PAX4 pathogenic variants or a complete knockout could still differentiate into endocrine cells but exhibited abnormal polyhormonal expression and reduced insulin secretion capacity. Established mechanisms for PAX4-related β-cell dysfunction center on the impairment of transcriptional repression of key target genes, such as insulin and glucagon.9 21–23 Pathogenic variants are thought to compromise this function primarily by reducing DNA-binding affinity9 21 23 or interfering with competitive inhibition of other factors such as PAX6.22Although PAX4 variants have been implicated in monogenic diabetes, the pathogenicity of many reported variants remains uncertain. This uncertainty, together with frequent incomplete penetrance, has made PAX4 a controversial MODY gene,24 25 motivating us to investigate how PAX4 variants affect protein expression and function.In this study, we identified three novel PAX4 variants in unrelated patients with early-onset diabetes and performed clinical, genetic, in silico, and in vitro functional analyses to investigate their pathogenic potential. In contrast to prior studies primarily addressing impairment of transcriptional function, we focused on the impact of these variants on the expression of PAX4. Our findings reveal that novel PAX4 variants can reduce protein abundance via enhanced proteasomal degradation, providing new insights into the molecular mechanisms contributing to β-cell dysfunction in MODY9.Methods Clinical data collection The study included probands and their family members recruited from the department of endocrinology at Shandong Provincial Hospital. Inclusion criteria for MODY screening were as follows: (1) age at diabetes diagnosis<35 years and (2) a family history of diabetes spanning at least two successive generations. Comprehensive clinical data were collected, and peripheral blood samples were obtained from probands and available family members across three generations. Genomic DNA from these samples was subjected to whole-exome sequencing (WES). Sequencing libraries were prepared using the SeqCap EZ Med Exome Enrichment Kit (Roche NimbleGen, USA) and sequenced on an Illumina HiSeq platform. Data were aligned to the hg19 reference genome, and variants were called using NextGene V.2.3.4 software. The average coverage of the target exome was >100×. All putative pathogenic variants were confirmed by Sanger sequencing.Bioinformatic analysis All analyses of PAX4 variants were performed with reference to transcript NM_001366110.1, encoding the 351-amino acid protein isoform (RefSeq: NP_001353039.1). To comprehensively evaluate the pathogenicity of the identified variants, a suite of bioinformatic analyses was performed. To assess the novelty of the identified variants, their frequencies were queried against several major population genomic databases, including gnomAD (v4.1.0), the 1000 Genomes Project, and the Exome Aggregation Consortium. Variants were classified according to American College of Medical Genetics and Genomics (ACMG) guidelines.26 The spectrum of previously reported PAX4 variants was reviewed using the Human Gene Mutation Database (HGMD). Variant pathogenicity was initially assessed usingProtein Variation Effect Analyzer (PROVEAN) (http://provean.jcvi.org/index.php),27 Sorting Intolerant From Tolerant (SIFT) (https://sift.bii.a-star.edu.sg/),28 and MutationTaster (http://www.mutationtaster.org/).29 Protein functional domains were predicted using the Simple Modular Architecture Research Tool (SMART) (https://smart.embl.de/).30 Sequence conservation of the affected residues was evaluated via multiple sequence alignment of PAX4 sequences retrieved from the HomoloGene database using ESPript V.3.0 (https://espript.ibcp.fr/ESPript/ESPript/index.php)31 to determine evolutionary conservation. Finally, structural models of the wild-type (WT) and mutant PAX4 were predicted by AlphaFold V.2 (https://alphafold.com/)32 and visualized in PyMOL (V.2.6.0a0)33 to assess potential structural consequences.Cell culture MIN6 cells (Beina Biology, China) were maintained in Dulbecco’s Modified Eagle Medium (Gibco, USA) containing 25 mM glucose, supplemented with 10% fetal bovine serum (ExCell Biology, China) and 1% penicillin-streptomycin (MeilunBio, China). Cells were maintained at 37°C in a humidified atmosphere containing 5% CO 2 and 95% air. Cells at 80-90% confluence were passaged using 0.05% trypsin-EDTA (MeilunBio, China).Plasmids construction and transfection WT and variant PAX4 (c.35T>C, c.83delA, and c.488G>C) plasmids were constructed based on the reference transcript. The coding sequences were amplified by PCR using primers containing the desired variants and cloned into the CV702 vector (GeneChem, China) using XhoI and KpnI restriction sites, with all constructs carrying a triple FLAG (DYKDDDDK) tag. The schematic representation of the constructs (online supplemental figure S1) was adapted from a previously published study.34 All plasmid sequences were confirmed by Sanger sequencing. Plasmids were transiently transfected into MIN6 cells using Lipofectamine 3000 (Invitrogen, USA) according to the manufacturer’s instructions. Cells were harvested 48 hours post-transfection for downstream analyses.SP110.1136/bmjdrc-2025-005375.supp1Supplementary dataWestern blot (WB) PAX4 expression was assessed by WB. 48 hours after transfection, total protein was extracted using radioimmunoprecipitation assay (RIPA) buffer supplemented with protease and phosphatase inhibitors (Biocolor, China). Equal amounts of protein were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (Epizyme, China) and then transferred to polyvinylidene fluoride (PVDF) membranes. After blocking with 5% non-fat milk, membranes were incubated with primary antibodies against PAX4 (1:2000; Abclonal, China) and β-Actin (1:7500; Proteintech, China), followed by incubation with horseradish peroxidase (HRP)-conjugated secondary antibodies (1:5000; Proteintech, China). Protein signals were finally detected using an enhanced chemiluminescence (ECL) system (Cytiva, USA).Immunofluorescence (IF) assay The subcellular localization of FLAG-tagged PAX4 was assessed by IF. At 48 hours post-transfection, cells were fixed with 4% paraformaldehyde for 15 min and then permeabilized with 0.1% Triton X-100 in phosphate-buffered saline (PBS) for 10 min at room temperature. After blocking with 5% BSA in PBS for 30 min, cells were incubated overnight at 4°C with an anti-FLAG primary antibody (1:500; Sigma-Aldrich, USA). After five washes with phosphate-buffered saline with tween (PBST), a fluorescent secondary antibody (Alexa Fluor 555, 1:1000; Invitrogen, USA) was applied for 1 hour at room temperature. Samples were mounted in VECTASHIELD Antifade Mounting Medium with 4',6-diamidino-2-phenylindole (DAPI) (Vectorlabs, USA) and imaged using a confocal laser scanning microscope (Nikon, Japan).Reverse transcription quantitative real-time PCR (RT-qPCR) Total RNA was extracted using RNA isolater Total RNA Extraction Reagent (Vazyme, China). Subsequently, 1 µg of total RNA was reverse transcribed into cDNA using Hifair V One-Step RT-gDNA Digestion SuperMix for qPCR (Yeasen, China). The resulting cDNA was then diluted 1:2 and subjected to RT-qPCR using qPCR SYBR Green Master Mix (Yeasen, China) in a 10 µL reaction volume with 5 µM primers. Gene expression was normalized to β-Actin and calculated via the 2–ΔΔCt method.35 Primer sequences (listed in online supplemental table S1) were obtained from the PrimerBank database or a previous publication.36Statistical analysis Results are presented as mean±SD. Statistical analyses were performed using GraphPad Prism V.9.0. After confirming homogeneity of variance, group comparisons were conducted using one-way Analysis of Variance (ANOVA) followed by Dunnett’s post hoc test for comparisons against the WT control group. Statistical significance was defined as p<0.05.Results Identification of three novel PAX4 variants in patients with diabetesFamily 1: the proband in family 1 was diagnosed with diabetes in early adulthood. On presentation, diabetes-associated autoantibodies were negative, with fasting C-peptide (1.87 ng/ml, reference range 1.1–4.4 ng/ml) and an hemoglobin A1c (HbA1c) of 5.7%. The patient has maintained stable glycemic control with lifestyle intervention alone. The family history revealed multiple relatives with diabetes across two generations, including a parent, uncle, and aunt. WES identified a heterozygous PAX4 variant (c.83delA; p.Gln28ArgfsTer6), predicted to either produce a truncated protein or trigger nonsense-mediated decay (NMD, a common mechanism for eliminating transcripts harboring premature termination codons37). Sanger sequencing confirmed the variant, which was also detected in an asymptomatic parent and child.Family 2: the proband in family 2 was a pediatric patient diagnosed with diabetes. On presentation, autoantibodies were negative, with a normal fasting C-peptide level (3.09 ng/ml) and an HbA1c of 5.5%. The patient is currently receiving insulin therapy. Two paternal relatives were also diagnosed with diabetes. WES identified a heterozygous PAX4 variant (c.488G>C; p.Arg163Pro), predicted to result in a missense substitution. Sanger sequencing confirmed the variant, which was also detected in an unaffected parent.Family 3: The proband in family 3 was a pediatric patient with a history of multiple episodes of diabetic ketoacidosis. At admission, HbA1c was markedly elevated (14.0%), and C-peptide was severely reduced (0.65 ng/ml), with negative autoimmune markers. The patient continues to receive insulin therapy. The family history revealed diabetes in the father, both paternal grandparents, and one maternal grandparent, while the other parent had gestational diabetes. WES identified a heterozygous PAX4 variant (c.35T>C; p.Leu12Pro), predicted to result in a missense substitution. Sanger sequencing confirmed the variant, which was also detected in the affected parent.The pedigrees of the three families, showing variant segregation patterns, are presented in figure 1.Figure 1Pedigrees of three probands with diabetes carrying PAX4 variants. Squares and circles represent males and females, respectively. Black-filled symbols denote affected individuals. Gray-filled symbols denote asymptomatic carriers. The proband is indicated by an arrow. Representative Sanger sequencing results are shown below each pedigree. The red arrow highlights the variant position, and the green arrow indicates the reference (wild-type) allele.Bioinformatic analysis To evaluate the novelty of the PAX4 variants identified in our research, we first queried major population databases. All three variants were absent, supporting their rarity. To characterize the current spectrum of PAX4 variants, we systematically reviewed HGMD. According to HGMD, over 30 variants in the PAX4 gene have been reported to date. However, not all of these variants have a clearly defined effect on the protein sequence. As illustrated in figure 2, both previously reported and novel amino acid-altering variants are predominantly clustered within the paired domain.Figure 2Schematic of the PAX4 domain structure with variant locations. The positions of previously reported (black) and novel (red) variants are shown.Mapping the novel variants onto the PAX4 protein sequence revealed that the p.Leu12Pro variant lies within the paired domain. In contrast, the p.Arg163Pro variant is located outside any known structural domain. Although the c.83delA transcript is a likely target for NMD, if escape occurs, the variant is predicted to produce a truncated protein (p.Gln28ArgfsTer6) lacking extensive functional sequence.To further assess the potential functional impact of the three variants, we performed in silico prediction and evolutionary conservation analyses. As shown in online supplemental table S2, among the three PAX4 variants analyzed, c.83delA (p.Gln28ArgfsTer6) was predicted to be pathogenic, whereas the other two variants showed opposite predictions: c.35T>C (p.Leu12Pro) was consistently deleterious, and c.488G>C (p.Arg163Pro) was uniformly benign. Structural models generated with AlphaFold V.2 and visualized in PyMOL showed altered side-chain orientations for p.Leu12Pro and p.Arg163Pro, and a pronounced folding disruption for p.Gln28ArgfsTer6 (figure 3A). Multiple sequence alignment across five vertebrate species (human, mouse, rat, rhesus monkey, chimpanzee) demonstrated strict conservation of the Leu12 and Arg163, supporting the functional importance of these sites (figure 3B).Figure 3Structural models and evolutionary conservation analysis of novel PAX4 variants. (A) Predicted structures of wild-type and mutant PAX4. (B) Multiple sequence alignment across five species. The boxed residues indicate variant sites; shading reflects conservation.Under the ACMG guidelines, the three variants received distinct classifications. The c.83delA (p.Gln28ArgfsTer6) variant was initially evaluated using the PVS1 criterion, which applies to null variants in genes where loss of function is an established disease mechanism. However, gnomAD data indicated a low loss-of-function intolerance score for PAX4 (pLI=0), prompting the downgrade of PVS1 to ‘Moderate’ strength. In the absence of additional supporting evidence, the combination of PVS1_Moderate and PM2 (absence from population databases) led to a final classification of variant of uncertain significance (VUS). In contrast, the c.35T>C (p.Leu12Pro) and c.488G>C (p.Arg163Pro) variants were classified as VUS, although based on different evidence. The c.35T>C (p.Leu12Pro) variant met the criteria for PM2, PP1 (cosegregation with disease), and PP3 (multiple computational predictions of deleteriousness), whereas the c.488G>C (p.Arg163Pro) variant was supported solely by PM2.These findings provided a preliminary basis for functional validation and clinical interpretation, highlighting the need for subsequent experimental verification of the predicted pathogenic effects.Functional characterization of novel PAX4 variantsPAX4 variant expression is decreased in MIN6 cellsTo investigate the pathogenic mechanisms of the novel PAX4 variants, we transfected MIN6 cells with plasmids expressing either WT or mutant PAX4 and analyzed their expression by WB (figure 4A). Endogenous PAX4 expression in MIN6 cells was nearly undetectable, allowing us to attribute detected signals to exogenous constructs. Compared with WT PAX4, both p.Leu12Pro and p.Arg163Pro exhibited markedly reduced expression levels in whole-cell lysates. In contrast, the predicted molecular weight of p.Gln28ArgfsTer6 is ~2 kDa, and it was undetectable by conventional WB.Figure 4Functional characterization of wild-type (WT) and mutant PAX4 in MIN6 cells. (A) Representative western blot (WB) image of PAX4 protein levels. Whole-cell lysates from MIN6 cells transfected with the indicated plasmids were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and immunoblotted with antibodies against PAX4 and β-Actin (loading control). (B) Subcellular localization analyzed by immunofluorescence. Cells were immunostained with an anti-FLAG (DYKDDDDK) antibody (green). Nuclei were counterstained with 4',6-diamidino-2-phenylindole (DAPI) (blue). Images were acquired using a confocal laser scanning microscope. Scale bar = 5 µm. (C) Reverse transcription quantitative real-time PCR (RT-qPCR) analysis of PAX4 mRNA levels. Data were normalized to β-Actin. Values are presented as mean±SD (n=4). *p<0.05, **p<0.01. (D) Representative WB image of mutant PAX4 expression mediated by proteasome and lysosome inhibitors. MIN6 cells transfected with the indicated plasmids were treated with 10 µM MG132 (proteasome inhibitor) or 50 µM chloroquine (CQ, lysosome inhibitor) for 6 hours prior to harvesting. Whole-cell lysates were separated by SDS-PAGE and immunoblotted for PAX4 and β-Actin (loading control). Abbreviations: EV, empty vector; MERGE, merged image.Subcellular localization of missense mutants remains unchanged Since PAX4 is a transcription factor that functions in the nucleus, we next performed IF to assess whether the variants altered the subcellular localization of PAX4. The results showed that WT PAX4 was uniformly distributed in the nucleus, and that both p.Arg163Pro and p.Leu12Pro were localized to the nucleus, similar to WT PAX4, indicating that their subcellular localization remained unaffected. However, p.Gln28ArgfsTer6 was not detectable ( figure 4B).Missense mutants are degraded by the proteasome pathway To elucidate the mechanisms underlying the reduced expression of the missense mutants, we investigated whether transcriptional regulation contributed to the observed protein expression changes. As shown in figure 4C, RT-qPCR analysis revealed that mRNA levels of p.Arg163Pro were significantly lower than those of WT, suggesting that reduced transcription may partially explain the decreased protein abundance. In contrast, p.Leu12Pro exhibited significantly increased mRNA levels compared with WT, indicating that transcriptional downregulation is not the primary cause of reduced protein levels in this case. These results suggest that transcriptional changes alone cannot fully account for the differential expression of the mutants.Given this, we further investigated whether post-translational degradation pathways were involved. Aberrant protein degradation in cells typically occurs via the proteasome or autophagy-lysosome pathways.38 To explore this, MIN6 cells were transfected with each variant. At 48 hours post-transfection, the cells were treated with 10 µM MG132 (a proteasome inhibitor; Sigma-Aldrich, USA) or 50 µM chloroquine (CQ, an autophagy-lysosome inhibitor; Sigma-Aldrich, USA) for 6 hours, using a protocol adapted from previous studies.39–42 Both inhibitors were dissolved in dimethyl sulfoxide (DMSO) to ensure consistent solvent conditions across treatments. DMSO was used as a vehicle control. Treatment with MG132 markedly restored the expression levels of both missense mutants, while CQ treatment had no significant effect, indicating that the ubiquitin-proteasome system predominantly mediates their degradation rather than the autophagy-lysosome pathway (figure 4D).Discussion While PAX4 is crucial for β-cell function and has been implicated in MODY9, the pathogenic mechanisms of most reported variants remain poorly understood. Here, we investigated three novel PAX4 variants (c.83delA, p.Gln28ArgfsTer6; c.35T>C, p.Leu12Pro; and c.488G>C, p.Arg163Pro) through an integrated clinical and experimental approach. Our findings reveal that these variants compromise β-cell function by reducing PAX4 protein levels through distinct molecular mechanisms.Clinically, all three probands exhibited early-onset hyperglycemia and a family history of diabetes. However, the observed inheritance patterns were not consistent with a classic autosomal dominant mode. In probands 1 and 2, the variant was inherited from an asymptomatic parent. Only in proband 3 were the genetic findings consistent with the family history. This incomplete penetrance is consistent with reports of PAX4 as a low-penetrance gene.24 The genotype-phenotype discordance in probands 1 and 2 suggests that the penetrance of PAX4 variants may be influenced by additional factors, a notion supported by the high prevalence of type 2 diabetes in their extended families. Notably, an exome-chip analysis identified an association between an Asian-specific PAX4 variant (p.Arg200His) and type 2 diabetes.43 Collectively, these findings support a model in which PAX4 variants contribute to diabetes risk additively or synergistically with other genetic and environmental risk factors.To definitively evaluate the intrinsic pathogenicity of these variants, we performed in vitro functional assays. Our data provide strong functional evidence to reassess their initial ACMG classifications. For the c.83delA (p.Gln28ArgfsTer6) variant, complete loss of protein expression fulfills the PS3 criterion (damaging functional effect). When combined with PM2, this evidence provides sufficient support for reclassification from VUS to likely pathogenic (LP). Similarly, the c.35T>C (p.Leu12Pro) and c.488G>C (p.Arg163Pro) variants resulted in markedly reduced protein levels due to enhanced proteasomal degradation, also supporting PS3. In conjunction with PM2, and with additional PP1 and PP3 evidence for the c.35T>C (p.Leu12Pro) variant, these data collectively fulfill the criteria for classification as LP.Beyond their implications for clinical classification, our findings also reveal novel insights into the molecular mechanisms underlying PAX4 dysfunction in MODY9. While previous studies have shown that certain PAX4 mutants (eg, p.Arg200His, p.Arg172Trp) do not significantly alter protein expression and primarily act through impaired transcriptional activity,10 23 our work provides the first direct evidence that some variants lead to reduced protein abundance. Enhanced proteasomal degradation appears to be the primary mechanism limiting the accumulation of dysfunctional PAX4 mutants, thereby protecting cellular homeostasis.38 Our findings suggest that haploinsufficiency of PAX4 may compromise β-cell differentiation or impair the adaptive capacity of mature β-cells. This post-translational regulatory process has been largely overlooked. We propose that the p.Leu12Pro and p.Arg163Pro variants destabilize the structure of PAX4, leading to local conformational perturbations that expose hydrophobic residues or degradation motifs, which are subsequently recognized by the ubiquitin-proteasome system.44 45 This proposed mechanism is illustrated in figure 5.Figure 5Schematic model of proposed pathogenic mechanisms for novel PAX4 variants. Under physiological conditions, wild-type PAX4 localizes to the nucleus and represses the transcription of target genes (eg, insulin, glucagon). The novel variants lead to reduced protein abundance, primarily through enhanced degradation via the ubiquitin-proteasome pathway, potentially resulting in β-cell dysfunction and diabetes pathogenesis.Based on the functional link between these novel PAX4 variants and β-cell dysfunction via proteasomal degradation, we next evaluated how these findings supported the molecular diagnosis for each proband. The genetic testing and family analysis results of proband 3 were consistent with the classic genetic pattern of MODY, thus confirming the diagnosis of MODY9. However, in probands 1 and 2, the presence of unaffected variant carriers indicated an inheritance pattern inconsistent with classical MODY, precluding a definitive diagnosis of MODY9. Nevertheless, the functional validation results strongly supported the association of these variants with pancreatic β-cell dysfunction, suggesting that they might be pathogenic for MODY9. For PAX4 variant carriers without disease, long-term follow-up is helpful for assessing their risk of diabetes and clarifying the pathogenicity and genetic pattern of the variants.In conclusion, this study expands the spectrum of PAX4 variants and underscores the importance of integrating genetic and functional analyses in MODY diagnosis. Our findings suggest that impaired protein expression, due to transcript loss or enhanced proteasomal degradation, may represent a key pathogenic mechanism. These insights pave the way for future genotype-phenotype correlation studies in monogenic diabetes.This study has several limitations. First, it relied solely on a single-cell model without validation in more relevant systems, such as primary pancreatic islets. Second, it lacks in vivo validation from animal models to confirm the impact on whole-body glucose metabolism. Third, the downstream molecular mechanisms by which these variants impair β-cell function remain unclear. Future studies should expand cellular models, develop animal models, and explore the molecular mechanisms underlying PAX4-related dysfunction. These steps will help clarify the role of the PAX4 gene in diabetes and its potential in precision medicine.Conclusion This study identified three novel PAX4 variants in three suspected MODY patients and their families. Our in vitro functional experiments provide functional evidence supporting the pathogenicity of c.83delA (p.Gln28ArgfsTer6), c.488G>C (p.Arg163Pro), and c.35T>C (p.Leu12Pro), and elucidate their molecular mechanisms of pathogenesis. This study expands the spectrum of PAX4 variants and underscores the importance of integrating genetic and functional data for accurate genetic diagnosis in monogenic diabetes.