Potential circadian rhythm-related pathogenic genes in diabetic nephropathy: a multi-omics Mendelian randomization study

Diabetic nephropathy (DN), a microvascular complication of diabetes, is a leading global cause of end-stage renal disease (ESRD) [1], affecting 30%-40% of diabetic patients and accounting for about one-third of chronic kidney disease (CKD) cases [2]. Despite its high prevalence, current diagnostic markers, such as albuminuria, estimated glomerular filtration rate (eGFR), lack sensitivity [3]. On the therapeutic front, current treatments primarily target hyperglycemia, hypertension, and dyslipidemia but fail to address specific molecular mechanisms underlying DN progression [4]. Consequently, there is an urgent need not only for DN-specific biomarkers and mechanism-based therapeutics, but also for a deeper understanding of how these genes influence key clinical manifestations and systemic metabolic regulation. Such integrative insight will bolster both early detection strategies and the development of precision interventions.

Recent studies have investigated the connection between DN and circadian rhythm genes [5]. The circadian rhythm is an endogenous oscillator that coordinates diverse biological processes, ensuring synchronization between internal physiology and the external environment. It plays a crucial role in regulating metabolic processes, including glucose and lipid metabolism, and its disruption has been linked to metabolic disorders such as type 2 diabetes, obesity, and cardiovascular diseases [6]. Circadian rhythm disruptions may also contribute to DN through pathways involving metabolic dysregulation, inflammation, and oxidative stress. Studies have highlighted the role of circadian rhythms in maintaining normal renal function and suggested potential links to DN, emphasizing the need for further clinical and mechanistic exploration [7]. These approaches are susceptible to confounding and reverse causality, making it challenging to determine if circadian pathway alterations causally drive DN or are merely a symptom. Identifying the specific molecular players with a causal link is essential for developing targeted interventions.

To overcome the limitations of observational studies, such as confounding and reverse causality, we employed summary-data-based Mendelian Randomization (SMR) [8]. SMR integrates genome-wide association study (GWAS) summary statistics with molecular quantitative trait loci (QTL) data to infer causal relationships, while the integrated heterogeneity in dependent instruments (HEIDI) test helps distinguish these signals from linkage or pleiotropic effects [9]. To date, no comprehensive SMR analysis has systematically investigated the causal role of circadian rhythm-related genes in the development of diabetic nephropathy. Therefore, this study aimed to systematically identify circadian rhythm genes associated with DN risk through multi-omics data integration. Our findings offer a more comprehensive framework for understanding and intervening in DN progression.

This study utilized publicly available, aggregated data from studies that had already received ethics approval; therefore, no additional institutional review was required.

In this study, SMR and colocalization analyses identified several circadian rhythm-related genes (ANKIB1, BICC1, GNAI2, KIF11 and PEBP1) showing significant association with the risk of DN. These findings establish a genetic framework linking circadian disruption to DN, offering potential targets for therapeutic intervention.

BICC1, an evolutionarily conserved RNA-binding protein, is crucial for various biological processes, including kidney development and the regulation of circadian rhythms [22]. Functionally, BICC1 regulates mRNA stability and translation, modulating signaling pathways in epithelial polarity and nephron morphogenesis [23]. Experimental studies, such as BICC1 knockout in mice leading to diabetes phenotypes [24], and human genetic studies linking BICC1 variants to CKD outcomes [25], support its potential involvement in renal and metabolic dysfunction. In this study, BICC1 was causally associated and colocalized with DN risk across mQTL and eQTL levels, further supported by replication studies. This suggested that BICC1 might contribute to the occurrence of DN. Notably, BICC1 expression showed positive associations with both eGFR and DN-related traits (ESRD and microalbuminuria). This seemingly paradoxical pattern may reflect the compensatory hyperfiltration phase often observed in early-stage chronic kidney disease, where elevated GFR initially manifests as increased filtration due to glomerular hypertension, however, this sustained glomerular hypertension ultimately drives progression to proteinuria and renal failure [26]. Despite the causal associations between BICC1 expression and DN-related traits, its expression was comparable between kidney tissues from both controls and DN patients. This controversy may stem from the fact that blood eQTL signals often diverge from those observed in kidney tissue, while sample-size constraints and inter-study heterogeneity further dilute statistical power. Moreover, BICC1 may exert stage- or cell-type-specific effects within the diabetic kidney. Notwithstanding the lack of direct tissue validation, convergent MR and colocalization evidence across multiple levels suggest its relevance in DN pathogenesis.

ANKIB1 (ankyrin repeat and IBR domain containing 1) is an E3 ubiquitin ligase known for its roles in inflammation and the regulation of circadian rhythms [27]. By degrading or modifying clock-associated molecules, it may modulate circadian-controlled metabolic and inflammatory responses [27]. While prior evidence specifically linking ANKIB1 to DN is limited, our multi-omics analysis revealed a causal association. Specifically, genetically predicted hypomethylation at the cg01198111 locus was associated with increased ANKIB1 expression, and this genetically-driven increased expression was causally linked to an elevated risk of DN. Moreover, the expression of ANKIB1 was associated with higher eGFR and reduced risk for ESRD, implying its potential protective effect. Therefore, ANKIB1 may play a multifaceted role in DN, in that its influence may be stage-dependent, shifting from renoprotective to neutral or even adverse as disease progresses. On the other hand, the blood-based eQTL signal may reflect a systemic compensatory up-regulation rather than true renal expression. Thus, future studies may focus on the dynamic role of ANKIB1 in DN progression and provide novel genetic support for its involvement.

We found a causal link between higher PEBP1 protein levels and reduced DN risk. PEBP1 is a multifunctional regulatory protein known to inhibit the Raf-1/MEK/ERK pathway and play roles in metabolic and inflammatory processes [28]. PEBP1 also modulates circadian regulation through kinase signaling pathways that intersect with core clock components [29]. By influencing MAPK-mediated synchronization of peripheral clocks with metabolic cues, PEBP1 may regulate circadian-controlled metabolic responses in renal cells under hyperglycemic conditions. Emerging evidence highlights its significance in metabolic and inflammatory pathways, particularly in diabetic nephropathy. For example, PEBP1 expression was significantly downregulated in renal tissues in experimental DN models, correlating with disease severity [30]. The downregulation of PEBP1 could contribute to chronic inflammation in renal tissues, as its downregulation could be accompanied by increased NF-κB activity and elevated urinary protein excretion and glomerulosclerosis [30]. Mechanistically, PEBP1 could contribute to the maintenance of mitochondrial function by enhancing eIF2α phosphorylation under hyperglycemia, which promotes selective translation of stress-responsive genes like ATF4 [31]. In PEBP1-deficient renal cells, this adaptive mechanism is impaired, leading to accelerated cell death [31]. Our findings showed the levels of PEBP1 were negatively associated with DN risk. Furthermore, decreased PEBP1 expression in kidney tissues from human DN patients supports its potential protective role and therapeutic relevance. Future research should investigate the specific protective mechanisms of PEBP1 in DN and explore its potential as a therapeutic target.

For KIF11 and GNAI2, our findings revealed a key discrepancy between systemic and local tissue effects. KIF11 is a microtubule motor protein essential for mitotic spindle formation and cell cycle progression [32]. While previous studies link KIF11 to increased type 2 diabetes risk and more severe DN progression, our genetic analysis revealed a protective association [33]. Previous research has revealed that increased expression of the KIF11 gene was associated with more severe progression of DN [34,35]. By contrast, our findings revealed that KIF11 was negatively associated with DN at m/eQTLs across both discovery and replication dataset, with its expression negatively associated with ESRD risk, fasting glucose and Hb1Ac levels. However, its expression was elevated in kidney tissues from DN patients. This discrepancy between blood eQTL data and transcriptome could be attributed to the tissue-specific expression and role of KIF11. Similarly, GNAI2 encodes a Gαi subunit involved in GPCR-mediated signaling, which is implicated in other diabetic complications like retinopathy and atherosclerosis[36,37]. Previous experimental research has shown that elevated expression of GNAI2 is associated with increased apoptosis of retinal ganglion cells and exacerbated development of atherosclerosis [38]. Despite a lack of direct evidence in DN, our analysis identified a genetically predicted protective role for GNAI2 in blood against DN risk. Yet again, this contrasted with its increased expression in DN kidney tissue. We propose that this tissue-specific discordance does not invalidate the causal interpretation but rather reflects fundamentally different biological dimensions through two non-exclusive mechanisms. First, the elevated expression in DN tissues likely represents a reactive compensatory up-regulation rather than a primary pathogenic driver. While our SMR evidence suggests that higher lifelong baseline expression of these genes exerts a protective effect against DN onset, their increased levels in diseased tissues may signify the kidney’s adaptive response to hyperglycemia-induced stress. Given that KIF11 has been implicated in regulating cell cycle progression [32], its up-regulation in DN tissues may represent a compensatory response aimed at promoting cell-cycle entry and proliferation to support the repair of injured tubular epithelial cells or podocytes. Similarly, given that GNAI2 has been implicated in inflammatory responses to hyperglycemic stress [38], its up-regulation in DN tissues may reflect a compensatory response induced by oxidative stress and inflammatory cytokines, potentially activating anti-inflammatory feedback mechanisms to maintain podocyte integrity. In this context, the blood-based genetic signal may capture systemic protective capacity against the initial injury, whereas increased local tissue expression may reflect a failed or exhausted compensatory response representing a late-stage reparative attempt within the pathological microenvironment. Second, this inconsistency may reflect a temporal hierarchy in causal inference. Genetic evidence provides insight into causal precedence based on lifelong baseline expression, whereas tissue transcriptomics captures a synchronized snapshot of an established disease state and therefore mainly represents associative patterns. Given the chronic nature of DN, the pathological milieu characterized by hyperosmolar stress and persistent inflammation can substantially remodel the gene expression landscape. Consequently, the divergence between SMR findings and observational transcriptomic results may highlight the dynamic and stage-specific roles of these genes as the disease progresses from initial susceptibility to chronic maladaptation. These hypotheses, while biologically plausible based on the known roles of KIF11 in cell cycle and GNAI2 in stress response, remain speculative and warrant further longitudinal single-cell profiling and lineage-tracing studies to fully dissect their functional transitions across disease progression. Nevertheless, the robust genetic links of KIF11 and GNAI2 to DN and its related metabolic traits establish them as pivotal nodes in pathogenesis, warranting focused mechanistic investigation into their context-dependent roles.

Based on our integrative network analyses, we next uncovered IRF2 as the potential key transcriptional factor regulating the expression of BICC1, ANKIB1, KIF11, PEBP1 and GNAI2. To bridge the gap between database-driven predictions and disease-specific biological relevance, we further validated these regulatory links using clinical transcriptomic data from the GSE96804↗ dataset. Our correlation analysis revealed a robust positive association between IRF2 and GNAI2 in DN patients. The high significance of this correlation in human renal tissues provides strong evidence that IRF2 may indeed act as a proximal regulator of GNAI2 under diabetic conditions. Additionally, the marginal correlation trends observed for BICC1 and PEBP1 further suggest a potential, albeit perhaps more complex, regulatory influence. Although the correlations for KIF11 and ANKIB1 did not reach statistical significance, potentially due to the inherent noise in clinical samples or cell-type-specific regulatory dynamics, the overall concordance, particularly the strength of the IRF2-GNAI2 axis, underscores the validity of our TF-target predictions. Given its established roles in inflammation and proliferation, IRF2 appears to act as a crucial regulatory node that translates upstream circadian disruption into downstream transcriptional reprogramming that collectively drives DN progression.

In summary, these prioritized genes potentially represent a diverse functional spectrum linking circadian rhythms to DN pathogenesis. It is hypothesized that BICC1 and ANKIB1 may modulate the speed and amplitude of the circadian clock through post-transcriptional and post-translational modifications, thereby potentially influencing renal epithelial polarity and inflammatory homeostasis in a stage- or context-dependent manner. PEBP1 likely acts as a metabolic-circadian integrator that could protect the kidney by stabilizing mitochondrial function, regulating MAPK-mediated peripheral clock synchronization, and suppressing NF-κB-mediated inflammation under hyperglycemic stress. In contrast, KIF11 and GNAI2 appear to govern the transition between initial injury protection and late-stage compensatory repair, suggesting their involvement in cell-cycle progression and G-protein signaling. Collectively, these genes, potentially coordinated by the key transcription factor IRF2, may form a regulatory network that translates systemic circadian desynchrony into localized renal pathological processes.

Moreover, molecular docking analysis nominated candidate small-molecule binders of the key proteins, including calycosin, nitrofural and nobiletin. Notably, nobiletin has been known as a circadian modulator, underscoring its therapeutic potential and reinforcing the rationale for circadian-related genes in DN [39]. Detailed inspection of the binding pose indicates that nitrofural is deeply anchored within the hydrophobic pocket of GNAI2. The stability of this complex is reinforced by a network of specific interactions, most notably the hydrogen bonds and hydrophobic contacts with key residues (ARG150, SER189, VAL276 and ILE232). The involvement of these conserved residues suggests that nitrofural may interfere with the conformational transitions of GNAI2 required for its coupling with downstream effectors. Previous studies have highlighted that the S-nitrosylation of GNAI2 at Cys66 is a critical driver of diabetes-accelerated atherosclerosis via the CXCR5/Hippo/YAP axis [38]. Our docking data provide a structural rationale for the potential therapeutic efficacy of nitrofural. By occupying the GNAI2 binding pocket, nitrofural might sterically hinder the pathological S-nitrosylation process or disrupt the aberrant SNO-GNAI2/CXCR5 coupling, thereby restoring Hippo pathway activity and alleviating endothelial inflammation.

A major strength of this study is the use of the multi-omics SMR framework, which leverages genetic variation to infer causal relationships, providing stronger evidence than traditional observational studies. Identifying circadian rhythm-related molecular traits with putatively causal links to DN risk offers a robust genetic basis for developing novel therapeutic strategies and personalized medicine approaches. We acknowledge that the definition of DN varies slightly across different GWAS consortia. While most cohorts rely on ICD coding and electronic health records, variations in diagnostic sensitivity, such as biopsy-confirmed cases versus those based on billing codes, may introduce phenotype heterogeneity. Nevertheless, the high consistency of our SMR results across the discovery and replication datasets indicates that the identified circadian genes represent core drivers of diabetic kidney injury regardless of specific clinical sub-definitions. This cross-dataset validation mitigates the risk that our findings are artifacts of a single phenotypic classification system.

However, several limitations should be considered when extending our conclusions to diverse populations and molecular contexts. Firstly, the genomic and phenotypic data utilized in this study were predominantly derived from cohorts of European ancestry. This choice was necessitated by the current availability of resources because European-ancestry datasets provide the largest sample sizes and the most comprehensive multi-omics coverage available to ensure maximal statistical power. Nevertheless, this reliance poses an inherent limitation to the global generalizability of our findings. Given the global prevalence of DN, the genetic architecture and environmental modifiers may vary significantly across different ethnic groups. Therefore, future studies incorporating large-scale omics data from non-European populations including those of East Asian and African as well as Latin American descent are essential to perform trans-ethnic validation and confirm the universal transferability of these molecular insights. Secondly, although multi-omics data were integrated, no significant overlap of associations was observed between the eQTL and pQTL levels. This phenomenon may be attributed to several factors. One possible explanation is the difference in statistical power, as currently available pQTL datasets generally have smaller sample sizes than large-scale eQTL resources, potentially leading to false negatives where biologically relevant protein-level associations fail to reach stringent significance thresholds. Another explanation is the inherently complex and non-linear relationship between transcripts and proteins. In the context of diseases, processes such as mRNA stability, translational regulation, and post-translational modifications may partially decouple mRNA expression from protein abundance. From a multi-omics perspective, our analysis aims to identify potential regulatory nodes across molecular layers, and genes identified at the eQTL level may still represent plausible candidates for further functional validation even if they have not yet reached statistical significance in current pQTL datasets. Overall, this discordance underscores the complexity of the regulatory landscape and highlights the need for larger proteomic cohorts in future studies. Thirdly, the discrepancies observed between blood-based QTL analysis and kidney tissue validation for genes such as GNAI2 and KIF11 underscore the need for further validation. While these observations are biologically plausible given the known roles of KIF11 in cell cycle and GNAI2 in stress responses, they remain speculative. Future studies will first employ longitudinal single-cell profiling and lineage-tracing approaches to dissect the functional transitions of these genes across disease progression, followed by experimental validation using perturbation in renal cell lines and diabetic mouse models to confirm their impact on kidney injury. Fourthly, the molecular docking analysis does not account for pharmacokinetic properties such as bioavailability, toxicity, or in vivo efficacy. Accordingly, the identified compounds should be considered preliminary lead candidates requiring further optimization and experimental validation, including biochemical assays such as surface plasmon resonance binding tests and IC₅₀ measurements to confirm binding affinity and inhibitory activity. Finally, the choice of colocalization thresholds warrants consideration. While a PPH4 > 0.8 is often used for high-confidence causal inference, we utilized a more inclusive threshold of 0.5 to facilitate exploratory multi-omics integration. The primary advantage of this approach is that it avoids the loss of biologically relevant signals in pQTL and mQTL datasets, which inherently possess lower statistical power than transcriptomic resources. By combining this with the HEIDI test for pleiotropy and linkage disequilibrium filtering, we maintain a balanced discovery pipeline. However, we acknowledge that a threshold of 0.5 provides suggestive rather than definitive evidence of colocalization. Consequently, the candidates identified through this approach should be interpreted as prioritized targets requiring further experimental validation to confirm their functional roles in DN pathology.

While this study provides a robust genetic framework linking circadian regulators to diabetic nephropathy, several future directions are warranted to translate these findings into clinical utility. First, functional validation remains the immediate priority. We plan to perform CRISPR/Cas9 or siRNA mediated knockdown or overexpression of BICC1, KIF11, and GNAI2 in high-glucose-induced podocytes and proximal tubule cells, as well as in db/db mice, to assess their effects on apoptosis, fibrosis markers, and core circadian clock components. Concurrently, we will integrate single-cell RNA sequencing and spatial transcriptomics in db/db mice to characterize the spatiotemporal expression patterns and in vivo functions of these genes in renal tissues. Second, mechanistic deepening will be pursued. Transcriptional regulatory validation will be conducted using ChIP-seq and luciferase reporter assays to confirm whether transcription factors such as IRF2 directly drive GNAI2 and other candidate genes expression. We will also investigate whether core circadian components, including CLOCK and BMAL1, modulate these pathogenic mediators, aiming to delineate the full molecular chain linking circadian disruption, gene dysregulation, and renal injury. Third, cross-ethnic validation is essential, given that current datasets are predominantly of European ancestry. We aim to replicate these findings in large East Asian and African ancestry cohorts as multi-omics datasets become available, ensuring the generalizability of the identified genetic markers. Fourth, preclinical and pharmacological optimization will focus on candidate compounds such as nitrofural. Structure-activity relationship studies will be conducted to enhance renal specificity and minimize toxicity. Complementary biochemical assays, such as binding affinity measurements and inhibitory activity tests, together with in vivo efficacy evaluations, will help translate computational predictions into experimental validation. Finally, clinical translation will explore the potential of these genes as early diagnostic biomarkers by assessing their expression or methylation levels in blood samples, and will investigate the feasibility of chronotherapy by optimizing drug administration timing according to circadian gene expression patterns to achieve precision medicine in diabetic nephropathy.

In conclusion, our study provides compelling genetic evidence for the causal involvement of specific circadian rhythm-related molecular factors in DN pathogenesis. These findings enhance our understanding of the complex interplay between circadian disruption and DN and identify promising targets for future research aimed at developing effective interventions. Functional validation studies in relevant models are essential next steps to translate these findings into clinical applications.