Genetic evidence linking circadian rhythm genes and hypertension: A two-sample Mendelian randomization study

Abstract

Objectives

Circadian rhythms regulate human health, and observational studies have linked their disruption to hypertension; however, confounding factors and reverse causality limit causal interpretation. We aimed to evaluate the existing evidence linking circadian gene sets and hypertension using Mendelian randomization.

Methods

Two-sample Mendelian randomization was performed using genome-wide association study summary-level data from non-overlapping European ancestry samples. Genetic variants of 1276 circadian genes (CGDB) were instrumental variables; the outcome was hypertension genome-wide association study data from the UK Biobank (ieu-b-5144; 463,010 participants). Single nucleotide polymorphisms were selected at p<5 × 10⁻⁸ (mean F-statistic: 276.5). Inverse variance weighting was the primary method used. Benjamini–Hochberg false discovery rate correction was applied across all 1276 genes. Sensitivity analyses were performed using heterogeneity testing, Mendelian randomization–Egger regression, Mendelian Randomization Pleiotropy RESidual Sum and Outlier, Steiger directionality testing, and leave-one-out analyses.

Results

Mendelian randomization identified 39 nominally significant associations (inverse variance weighting, p < 0.05). After false discovery rate correction, four genes reached statistical significance: PHOSPHO1 (odds ratio = 0.988, 95% confidence interval: 0.983–0.993), BAIAP3 (odds ratio = 1.009, 95% confidence interval: 1.005–1.014), OSGIN2 (odds ratio = 0.990, 95% confidence interval: 0.986–0.995), and ZDHHC18 (odds ratio = 1.012, 95% confidence interval: 1.007–1.018). The remaining nominal associations did not survive false discovery rate correction and were classified as exploratory. Functional enrichment implicated circadian rhythm, apoptosis, and glycosaminoglycan biosynthesis pathways. Sensitivity analyses did not reveal significant heterogeneity or pleiotropy.

Conclusions

These findings are hypothesis-generating and require experimental validation. PHOSPHO1, BAIAP3, OSGIN2, and ZDHHC18 demonstrated statistically significant associations with hypertension risk after false discovery rate correction and should be prioritized for follow-up. Other nominal findings should be interpreted as exploratory.

Introduction

Hypertension refers to a clinical syndrome characterized by increased systemic arterial blood pressure, which may be accompanied with functional or organic damage to target organs, such as the heart, brain, and kidneys. Hypertension is the most prevalent chronic non-communicable disease worldwide.^1,2 It is a leading risk factor for cardiovascular disease and mortality. ³ Advances in genome-wide association studies (GWAS) have identified numerous genetic variants associated with blood pressure regulation, offering novel insights into the genetic underpinnings of hypertension. ⁴ The circadian clock is a fundamental biological phenomenon observed across the natural world. ⁵ As an endogenous regulatory system, it is present in nearly all organisms and has evolved to enable adaptation to time-dependent environmental changes.^6,7 This system not only governs the sleep–wake cycle but also profoundly influences key physiological processes such as metabolism, hormone secretion, and cardiovascular function. ⁸ Existing research indicates that multiple cardiovascular functions, including vascular tone, endothelial function, and renin–angiotensin–aldosterone system activity, are finely regulated by the biological clock, thereby maintaining a normal circadian rhythm of blood pressure. ⁹

Recent epidemiological studies have increasingly revealed a close association between circadian rhythm disruption and heightened cardiovascular disease risk. Common modern rhythm-disrupting factors such as shift work and sleep deprivation have been established as independent risk factors for cardiovascular diseases, including hypertension and coronary heart disease.^10–12 At the genetic level, GWAS have identified multiple circadian rhythm-related gene loci significantly associated with blood pressure phenotypes, including nocturnal blood pressure dipping patterns. ¹³ Core circadian genes such as CLOCK, BMAL1, PER, and CRY have also been linked to circadian preference, sleep duration, and other circadian phenotype alterations.^4,14 However, existing studies have significant limitations. First, traditional observational studies struggle to avoid confounding factors and reverse causality, limiting causal interpretation of the relationship between circadian rhythm and hypertension. ¹⁵ Second, although multiple circadian-related genetic loci have been identified, systematic Mendelian randomization (MR) assessments linking circadian rhythm gene sets to hypertension remain limited. ¹⁶ Finally, the underlying biological pathways and molecular mechanisms have yet to be fully elucidated. MR employs genetic variants as instrumental variables (IVs) to simulate the design of randomized controlled trials, relying on three core assumptions:

1. Relevance. The genetic variants are robustly associated with the exposure.

2. Independence. The variants are not associated with confounders.

3. Exclusion restriction. The variants affect the outcome only through the exposure.

This framework can reduce limitations of traditional observational studies and provide evidence relevant to causal inference.^17–19

Accordingly, this study adopted a two-sample MR design, utilizing publicly available pooled GWAS data to evaluate evidence linking 1276 circadian rhythm-related genes to hypertension. The primary objectives of this study were as follows: (a) identify circadian rhythm genes showing evidence of association with hypertension; (b) evaluate the direction and magnitude of their MR-estimated effects on hypertension risk; and (c) elucidate relevant biological pathways through functional enrichment analysis. This research aimed to provide a basis for early risk prediction, personalized treatment strategies, and chronotherapy research in hypertension management.

Materials and methods

Research design

This study used a two-sample MR design to investigate potential associations between circadian rhythm genes and hypertension. The overall workflow consisted of four sequential phases: (a) data acquisition and processing; (b) IV selection and validation; (c) MR effect estimation using multiple methods; and (d) functional enrichment analysis of genes showing evidence of association. This design follows the established MR framework that utilizes genetic variants as IVs to support causal inference and minimize confounding and reverse causation bias inherent to traditional observational studies. First, data from the pooled GWAS for exposure (1276 circadian rhythm genes identified from the Circadian Gene DataBase (CGDB)) and outcome (hypertension, OpenGWAS dataset ID: ieu-b-5144) were retrieved from public repositories. Second, valid IVs were screened in line with the three core MR assumptions. Third, MR effects were estimated using multiple methods, with inverse variance weighting (IVW) as the primary approach. Finally, functional enrichment analysis was performed on genes showing evidence of association to clarify their biological implications. This study was conducted and reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE)–MR guidelines.

Data sources

This was a two-sample MR study using GWAS summary-level data; the exposure (circadian gene expression) and outcome (hypertension) samples were non-overlapping. Both datasets are derived from European ancestry populations, and no individual-level data were accessed. All the genetic data used in this study were sourced from publicly available GWAS databases. The circadian rhythm gene set comprising 1276 genes was obtained from the CGDB (http://cgdb.biocuckoo.org/), and the complete gene list is provided in Supplementary File 1 (circadian.geneList.txt). Hypertension GWAS summary statistics were acquired from the MRC Integrative Epidemiology Unit (IEU) OpenGWAS database (https://opengwas.io/; dataset ID: ieu-b-5144), comprising 463,010 UK Biobank participants (54,358 cases and 408,652 controls) of European ancestry. From these datasets, we extracted key information, including single nucleotide polymorphism (SNP) identifiers, alleles, effect allele frequencies, effect sizes, standard errors (SEs), and p-values. This study constituted secondary analysis of publicly available summary-level data and did not involve individual-level data; therefore, no additional ethical approval was required. Ethical approval was obtained for the original GWAS from their respective institutional review boards, with written informed consent provided by all participants. The original GWAS (ieu-b-5144) was approved by the UK Biobank Ethics Committee (Approval Number: 11/NW/0382).

IV screening

To ensure the validity of the IVs, screening criteria were established in strict adherence to the core assumptions of MR:

1. Strong association. SNPs were selected that were significantly associated with circadian rhythm genes at the genome-wide significance level (association significance threshold set at p < 5 × 10⁻⁸). F-statistics were calculated for all retained SNPs; weak IVs with F <10 were excluded to mitigate weak instrument bias. The mean F-statistic across all 3174 retained SNPs was 276.5 (range: 29.7–14,552.9), confirming adequate instrument strength.

2. Independence. To exclude linkage disequilibrium (LD) effects, the clustering parameters were set at r² <0.001 with a physical distance threshold >10,000 kb.

3. Confounder exclusion. The GWAS Catalog and PhenoScanner databases were used to exclude SNPs significantly associated with known potential confounders of hypertension. Among the 3174 retained SNPs, 1864 (58.7%) were cis-acting variants located within ±1 megabase (Mb) of the target gene; the remaining 1310 (41.3%) were trans-acting variants.

Notably, SNPs were selected based on their association with circadian gene expression GWAS data and genomic proximity to circadian gene regions, rather than exclusively from experimentally validated cis-eQTL datasets. Results therefore reflect MR-estimated associations involving genetic variation in circadian gene regions and hypertension risk, rather than specifically the effect of gene expression levels per se. This distinction is acknowledged as a key limitation.

MR analysis methods

To ensure the robustness of the results, two-sample MR analysis was performed using R (v4.2.0) and the package ‘TwoSampleMR’ (v0.6.14).^20,21 The IVW method served as the core analysis approach for estimating MR effects. In addition, the Mendelian randomization–Egger regression (MR-Egger) method, weighted median, simple mode, and weighted mode were employed as validation techniques to cross-check the consistency of MR effect estimates. Positive results required the fulfillment of four criteria: (a) IVW, p <0.05; (b) consistent effect direction across at least three supplementary methods; (c) no significant heterogeneity or pleiotropy in sensitivity analyses; and (d) robustness in leave-one-out analysis.

Multiple testing correction

As 1276 circadian rhythm genes were tested simultaneously, multiple testing correction was applied using the Benjamini–Hochberg false discovery rate (FDR) method. An FDR-adjusted p-value <0.05 was used as the significance threshold. For reference, Bonferroni-corrected results (p < 3.92 × 10⁻⁵) are provided in Supplementary Table S1. All 39 genes with nominal IVW (p < 0.05) are reported as exploratory findings; four of these also met the FDR-adjusted threshold and constitute the primary statistically significant results.

Sensitivity analyses

To validate the reliability of MR analysis results, four sensitivity analyses were conducted:

1. Heterogeneity testing. Cochran's Q statistic was calculated using both IVW and MR-Egger methods; p >0.05 indicated no significant heterogeneity. Where heterogeneity was present, a random-effects IVW model was employed.

2. Pleiotropy testing. Horizontal pleiotropy was assessed using the MR-Egger intercept test, with p >0.05 indicating no significant horizontal pleiotropy. Mendelian Randomization Pleiotropy RESidual Sum and Outlier (MR-PRESSO, v1.0) was additionally applied to detect and correct for individual outlier SNPs; of the 39 nominally significant genes, 16 had ≥4 SNPs and were eligible for MR-PRESSO analysis, whereas the remaining 23 had ≤3 SNPs and could not be assessed.

3. Steiger directionality test. Steiger filtering was applied to evaluate whether the direction was more consistent with circadian gene exposure influencing hypertension than with the reverse direction.

4. Leave-one-out analysis. IVW analysis was repeated after sequentially excluding each SNP to evaluate the impact of individual SNPs on overall results. Functional enrichment analyses used clusterProfiler (v4.6.2) and ggplot2 (v4.0.2).^22,23

Sample size considerations

Formal a priori sample size calculations were not performed for this MR study as the analyses used pre-existing publicly available GWAS data with predetermined sample sizes. The hypertension outcome GWAS comprised 463,010 participants, providing substantial statistical power for the detection of the genetic effects of the magnitude typical for complex traits. This approach is standard for two-sample MR studies using summary-level data.

Results

IV screening results

Status of SNP screening. Using the GWAS summary data for circadian rhythm gene sets as exposures, 3174 SNPs were retained after applying genome-wide significance screening (p < 5 × 10⁻⁸), LD clumping (r² < 0.001, distance > 10,000 kilobasepair (kb)), F >10 filtering, and confounder exclusion. The mean F-statistic was 276.5 (range: 29.7–14,552.9). Among these, 58.7% (1864/3174) were cis-acting variants (±1 Mb of the target gene). Detailed information is provided in Supplementary File 2 (exposure.F.csv).

MR core analysis

MR association results. Two-sample MR analysis identified 39 circadian rhythm genes with nominally significant associations with hypertension (IVW, p < 0.05). Among these, 18 genes exhibited odds ratio (OR) >1, suggesting that genetic variation in these circadian gene regions was associated with increased hypertension risk (Figure 1). The full set of 39 genes encompassed both core clock genes such as PER2 and CRY2 as well as additional rhythm-related genes that have received less attention in previous hypertension studies (Table 1, Figure 2). These findings may provide clues regarding the genetic mechanisms by which circadian rhythms regulate blood pressure.

Figure 1.

Forest plot demonstrating IVW effect estimates for 39 circadian rhythm genes associated with hypertension. Genes are listed alphabetically on the y-axis. Horizontal lines represent 95% confidence intervals.

nSNP: number of instrumental SNPs per gene; SNP: single nucleotide polymorphism; IVW: inverse variance weighting.

Table 1.

Mendelian randomization analysis results for circadian rhythm genes and hypertension.

Outcome	Exposure	Method	SNP	β	SE	Pval
Hypertension	BAIAP3	MR-Egger	3	0.010795	0.003474	0.198193
Hypertension	BAIAP3	Weighted median	3	0.009477	0.002272	3.02E−05
Hypertension	BAIAP3	Inverse variance weighted	3	0.00917	0.002173	2.45E−05
Hypertension	BAIAP3	Simple mode	3	0.00977	0.004433	0.158352
Hypertension	BAIAP3	Weighted mode	3	0.009488	0.002165	0.048342
Hypertension	SDCCAG8	MR-Egger	3	0.008466	0.00471	0.323204
Hypertension	SDCCAG8	Weighted median	3	0.007508	0.002339	0.00133
Hypertension	SDCCAG8	Inverse variance weighted	3	0.00743	0.002324	0.001386
Hypertension	SDCCAG8	Simple mode	3	0.007981	0.003441	0.146181
Hypertension	SDCCAG8	Weighted mode	3	0.007507	0.002559	0.099203
Hypertension	RELT	MR-Egger	3	−0.00533	0.003974	0.407824
Hypertension	RELT	Weighted median	3	−0.00606	0.002313	0.008827
Hypertension	RELT	Inverse variance weighted	3	−0.00603	0.002204	0.006188
Hypertension	RELT	Simple mode	3	−0.00704	0.004582	0.264083
Hypertension	RELT	Weighted mode	3	−0.00603	0.002277	0.117761
Hypertension	NT5C2	MR-Egger	4	−0.0136	0.007922	0.228259
Hypertension	NT5C2	Weighted median	4	−0.01289	0.003871	0.000872
Hypertension	NT5C2	Inverse variance weighted	4	−0.01116	0.004139	0.007023
Hypertension	NT5C2	Simple mode	4	−0.01219	0.004776	0.083757
Hypertension	NT5C2	Weighted mode	4	−0.01323	0.003781	0.039546
Hypertension	NUCB1	MR-Egger	4	0.006082	0.004085	0.27494
Hypertension	NUCB1	Weighted median	4	0.004866	0.002315	0.035597
Hypertension	NUCB1	Inverse variance weighted	4	0.004472	0.002233	0.0452
Hypertension	NUCB1	Simple mode	4	8.99E−05	0.003999	0.983477
Hypertension	NUCB1	Weighted mode	4	0.005108	0.00254	0.137795
Hypertension	CPD	MR-Egger	5	−0.00368	0.005843	0.573124
Hypertension	CPD	Weighted median	5	−0.00589	0.002061	0.004247
Hypertension	CPD	Inverse variance weighted	5	−0.00528	0.002491	0.033979
Hypertension	CPD	Simple mode	5	−0.00461	0.00364	0.273708
Hypertension	CPD	Weighted mode	5	−0.00593	0.00214	0.050274
Hypertension	ABTB1	MR-Egger	6	−0.0061	0.003225	0.131744
Hypertension	ABTB1	Weighted median	6	−0.00411	0.001827	0.024612
Hypertension	ABTB1	Inverse variance weighted	6	−0.00359	0.001749	0.040307
Hypertension	ABTB1	Simple mode	6	−0.00439	0.004327	0.356852
Hypertension	ABTB1	Weighted mode	6	−0.00423	0.001936	0.080714
Hypertension	PLEK	MR-Egger	3	−0.00216	0.004954	0.73844
Hypertension	PLEK	Weighted median	3	−0.00525	0.002614	0.044529
Hypertension	PLEK	Inverse variance weighted	3	−0.00557	0.002408	0.020642
Hypertension	PLEK	Simple mode	3	−0.00458	0.003959	0.366914
Hypertension	PLEK	Weighted mode	3	−0.00523	0.002438	0.165346
Hypertension	CRY2	MR-Egger	3	−0.00315	0.004138	0.585772
Hypertension	CRY2	Weighted median	3	−0.00519	0.002398	0.030514
Hypertension	CRY2	Inverse variance weighted	3	−0.0054	0.002385	0.023595
Hypertension	CRY2	Simple mode	3	−0.00411	0.004275	0.437429
Hypertension	CRY2	Weighted mode	3	−0.00507	0.002482	0.177555
Hypertension	RTN2	MR-Egger	3	0.000304	0.006611	0.970716
Hypertension	RTN2	Weighted median	3	0.006401	0.003363	0.05694
Hypertension	RTN2	Inverse variance weighted	3	0.006987	0.003247	0.031403
Hypertension	RTN2	Simple mode	3	0.008976	0.005816	0.262759
Hypertension	RTN2	Weighted mode	3	0.005328	0.003929	0.307818
Hypertension	FFAR2	MR-Egger	3	0.002999	0.008476	0.783512
Hypertension	FFAR2	Weighted median	3	0.009576	0.00578	0.097554
Hypertension	FFAR2	Inverse variance weighted	3	0.010256	0.004908	0.036645
Hypertension	FFAR2	Simple mode	3	0.011927	0.007771	0.264599
Hypertension	FFAR2	Weighted mode	3	0.008063	0.006357	0.332348
Hypertension	MRPL35	MR-Egger	3	0.011924	0.029423	0.754881
Hypertension	MRPL35	Weighted median	3	0.00819	0.00222	0.000225
Hypertension	MRPL35	Inverse variance weighted	3	0.007072	0.002762	0.010465
Hypertension	MRPL35	Simple mode	3	0.007925	0.003836	0.174829
Hypertension	MRPL35	Weighted mode	3	0.008941	0.002273	0.058944
Hypertension	PER2	MR-Egger	5	0.002327	0.028665	0.940419
Hypertension	PER2	Weighted median	5	0.015242	0.006246	0.014681
Hypertension	PER2	Inverse variance weighted	5	0.009733	0.004845	0.044569
Hypertension	PER2	Simple mode	5	0.016692	0.008706	0.127672
Hypertension	PER2	Weighted mode	5	0.016631	0.008787	0.131343
Hypertension	DYSF	MR-Egger	3	−0.00736	0.006668	0.468821
Hypertension	DYSF	Weighted median	3	−0.00608	0.00264	0.02121
Hypertension	DYSF	Inverse variance weighted	3	−0.00555	0.002627	0.034579
Hypertension	DYSF	Simple mode	3	−0.00739	0.003787	0.190446
Hypertension	DYSF	Weighted mode	3	−0.00622	0.002897	0.164735
Hypertension	ANKRD36	MR-Egger	3	0.013376	0.00717	0.313246
Hypertension	ANKRD36	Weighted median	3	0.011438	0.00355	0.001274
Hypertension	ANKRD36	Inverse variance weighted	3	0.011479	0.003415	0.000775
Hypertension	ANKRD36	Simple mode	3	0.010264	0.005181	0.186112
Hypertension	ANKRD36	Weighted mode	3	0.011161	0.003816	0.099716
Hypertension	HLX	MR-Egger	3	−0.00479	0.015211	0.805907
Hypertension	HLX	Weighted median	3	−0.01265	0.004785	0.008209
Hypertension	HLX	Inverse variance weighted	3	−0.01069	0.004401	0.015117
Hypertension	HLX	Simple mode	3	−0.01262	0.007301	0.225958
Hypertension	HLX	Weighted mode	3	−0.01422	0.005538	0.12408
Hypertension	LRPPRC	MR-Egger	3	−0.00423	0.011056	0.767244
Hypertension	LRPPRC	Weighted median	3	−0.00669	0.002015	0.0009
Hypertension	LRPPRC	Inverse variance weighted	3	−0.00679	0.002231	0.002346
Hypertension	LRPPRC	Simple mode	3	−0.00488	0.004042	0.350829
Hypertension	LRPPRC	Weighted mode	3	−0.0067	0.00198	0.077259
Hypertension	TSPAN3	MR-Egger	6	0.012419	0.008069	0.198629
Hypertension	TSPAN3	Weighted median	6	0.003889	0.001817	0.032318
Hypertension	TSPAN3	Inverse variance weighted	6	0.003425	0.001613	0.033748
Hypertension	TSPAN3	Simple mode	6	0.003199	0.002456	0.249498
Hypertension	TSPAN3	Weighted mode	6	0.003923	0.001903	0.094181
Hypertension	ITGAX	MR-Egger	10	−0.00439	0.002831	0.159364
Hypertension	ITGAX	Weighted median	10	−0.00465	0.001877	0.013284
Hypertension	ITGAX	Inverse variance weighted	10	−0.00499	0.001543	0.001223
Hypertension	ITGAX	Simple mode	10	−0.00393	0.002791	0.192155
Hypertension	ITGAX	Weighted mode	10	−0.00502	0.00188	0.025587
Hypertension	MYO1F	MR-Egger	3	−0.01204	0.006944	0.332986
Hypertension	MYO1F	Weighted median	3	−0.01104	0.004447	0.013071
Hypertension	MYO1F	Inverse variance weighted	3	−0.01052	0.004096	0.010227
Hypertension	MYO1F	Simple mode	3	−0.01316	0.006238	0.16937
Hypertension	MYO1F	Weighted mode	3	−0.01138	0.004635	0.133483
Hypertension	GPA33	MR−Egger	3	−0.01509	0.006438	0.256642
Hypertension	GPA33	Weighted median	3	−0.01005	0.004232	0.017524
Hypertension	GPA33	Inverse variance weighted	3	−0.00969	0.004615	0.03585
Hypertension	GPA33	Simple mode	3	−0.01163	0.006004	0.192251
Hypertension	GPA33	Weighted mode	3	−0.00982	0.004588	0.165669
Hypertension	MCL1	MR-Egger	3	−0.00505	0.007709	0.630921
Hypertension	MCL1	Weighted median	3	−0.00928	0.003025	0.002149
Hypertension	MCL1	Inverse variance weighted	3	−0.00958	0.003036	0.001599
Hypertension	MCL1	Simple mode	3	−0.01168	0.005908	0.186771
Hypertension	MCL1	Weighted mode	3	−0.00898	0.003342	0.115123
Hypertension	PHOSPHO2	MR-Egger	4	−0.01356	0.006673	0.179218
Hypertension	PHOSPHO2	Weighted median	4	−0.00465	0.001937	0.016361
Hypertension	PHOSPHO2	Inverse variance weighted	4	−0.00409	0.001976	0.038515
Hypertension	PHOSPHO2	Simple mode	4	−0.00635	0.003094	0.132481
Hypertension	PHOSPHO2	Weighted mode	4	−0.00465	0.001847	0.086371
Hypertension	ADM	MR-Egger	7	0.005797	0.003654	0.173533
Hypertension	ADM	Weighted median	7	0.005988	0.002157	0.005491
Hypertension	ADM	Inverse variance weighted	7	0.005297	0.001913	0.005629
Hypertension	ADM	Simple mode	7	0.005814	0.003483	0.146169
Hypertension	ADM	Weighted mode	7	0.005762	0.002143	0.036112
Hypertension	BAG3	MR-Egger	4	−0.0092	0.007252	0.3322
Hypertension	BAG3	Weighted median	4	−0.00892	0.00353	0.011549
Hypertension	BAG3	Inverse variance weighted	4	−0.00801	0.003338	0.016369
Hypertension	BAG3	Simple mode	4	−0.00891	0.005744	0.218862
Hypertension	BAG3	Weighted mode	4	−0.00917	0.003764	0.092731
Hypertension	OSGIN2	MR-Egger	4	−0.00632	0.004026	0.257135
Hypertension	OSGIN2	Weighted median	4	−0.00962	0.00238	5.30E−05
Hypertension	OSGIN2	Inverse variance weighted	4	−0.0096	0.002299	2.97E−05
Hypertension	OSGIN2	Simple mode	4	−0.01646	0.005594	0.060388
Hypertension	OSGIN2	Weighted mode	4	−0.00964	0.00248	0.030225
Hypertension	NPAS2	MR-Egger	4	0.006208	0.00538	0.367825
Hypertension	NPAS2	Weighted median	4	0.006562	0.003606	0.068784
Hypertension	NPAS2	Inverse variance weighted	4	0.007107	0.003165	0.024735
Hypertension	NPAS2	Simple mode	4	0.009407	0.004816	0.145767
Hypertension	NPAS2	Weighted mode	4	0.005183	0.004346	0.318709
Hypertension	IQCB1	MR-Egger	5	0.006327	0.00341	0.160524
Hypertension	IQCB1	Weighted median	5	0.004456	0.001393	0.001384
Hypertension	IQCB1	Inverse variance weighted	5	0.004354	0.001353	0.001291
Hypertension	IQCB1	Simple mode	5	0.004917	0.002638	0.13577
Hypertension	IQCB1	Weighted mode	5	0.00452	0.001558	0.044103
Hypertension	TRIB1	MR-Egger	3	0.276442	0.388569	0.606339
Hypertension	TRIB1	Weighted median	3	0.02851	0.00939	0.002397
Hypertension	TRIB1	Inverse variance weighted	3	0.0257	0.009577	0.007284
Hypertension	TRIB1	Simple mode	3	0.032543	0.013491	0.137318
Hypertension	TRIB1	Weighted mode	3	0.034192	0.015403	0.156611
Hypertension	TIGD3	MR-Egger	3	−0.02189	0.015355	0.38938
Hypertension	TIGD3	Weighted median	3	−0.00983	0.004893	0.044665
Hypertension	TIGD3	Inverse variance weighted	3	−0.01016	0.004549	0.025492
Hypertension	TIGD3	Simple mode	3	−0.00201	0.008551	0.835715
Hypertension	TIGD3	Weighted mode	3	−0.01331	0.005476	0.135737
Hypertension	PHOSPHO1	MR-Egger	8	−0.00702	0.006519	0.323055
Hypertension	PHOSPHO1	Weighted median	8	−0.01018	0.003239	0.001675
Hypertension	PHOSPHO1	Inverse variance weighted	8	−0.01223	0.002678	4.97E−06
Hypertension	PHOSPHO1	Simple mode	8	−0.01061	0.005077	0.075012
Hypertension	PHOSPHO1	Weighted mode	8	−0.01004	0.003517	0.024468
Hypertension	CTSF	MR-Egger	3	0.007476	0.004401	0.338698
Hypertension	CTSF	Weighted median	3	0.004212	0.001951	0.030846
Hypertension	CTSF	Inverse variance weighted	3	0.004131	0.001916	0.031097
Hypertension	CTSF	Simple mode	3	0.00095	0.003316	0.801389
Hypertension	CTSF	Weighted mode	3	0.004809	0.002098	0.148913
Hypertension	NR1D2	MR-Egger	3	0.006586	0.007831	0.554836
Hypertension	NR1D2	Weighted median	3	0.009717	0.004442	0.028696
Hypertension	NR1D2	Inverse variance weighted	3	0.010947	0.004101	0.0076
Hypertension	NR1D2	Simple mode	3	0.009239	0.005302	0.223527
Hypertension	NR1D2	Weighted mode	3	0.00963	0.005066	0.197664
Hypertension	SHMT1	MR-Egger	3	0.012874	0.005837	0.271016
Hypertension	SHMT1	Weighted median	3	0.007836	0.002089	0.000177
Hypertension	SHMT1	Inverse variance weighted	3	0.007632	0.002889	0.008253
Hypertension	SHMT1	Simple mode	3	0.005448	0.003838	0.291611
Hypertension	SHMT1	Weighted mode	3	0.008286	0.002283	0.068236
Hypertension	CHST15	MR-Egger	4	−0.00582	0.008119	0.547634
Hypertension	CHST15	Weighted median	4	−0.00918	0.003472	0.008211
Hypertension	CHST15	Inverse variance weighted	4	−0.00954	0.003406	0.005095
Hypertension	CHST15	Simple mode	4	−0.01607	0.006379	0.086276
Hypertension	CHST15	Weighted mode	4	−0.00923	0.003761	0.091424
Hypertension	GNG2	MR-Egger	3	−0.00807	0.006738	0.442979
Hypertension	GNG2	Weighted median	3	−0.01212	0.004034	0.00267
Hypertension	GNG2	Inverse variance weighted	3	−0.0125	0.003889	0.001304
Hypertension	GNG2	Simple mode	3	−0.01574	0.005958	0.118418
Hypertension	GNG2	Weighted mode	3	−0.01123	0.004317	0.121505
Hypertension	H2BC18	MR-Egger	3	0.016698	0.006264	0.228461
Hypertension	H2BC18	Weighted median	3	0.010097	0.004195	0.01608
Hypertension	H2BC18	Inverse variance weighted	3	0.008316	0.003788	0.028129
Hypertension	H2BC18	Simple mode	3	0.011817	0.005821	0.179479
Hypertension	H2BC18	Weighted mode	3	0.011394	0.005015	0.151015
Hypertension	ZDHHC18	MR-Egger	4	0.00686	0.004798	0.289061
Hypertension	ZDHHC18	Weighted median	4	0.011671	0.001947	2.05E−09
Hypertension	ZDHHC18	Inverse variance weighted	4	0.011998	0.002779	1.58E−05
Hypertension	ZDHHC18	Simple mode	4	0.01036	0.003449	0.057511
Hypertension	ZDHHC18	Weighted mode	4	0.011572	0.001989	0.010108
Hypertension	KIR3DL2	MR-Egger	3	−0.00578	0.019986	0.820917
Hypertension	KIR3DL2	Weighted median	3	−0.00702	0.002227	0.00162
Hypertension	KIR3DL2	Inverse variance weighted	3	−0.00691	0.002121	0.001122
Hypertension	KIR3DL2	Simple mode	3	−0.00717	0.003136	0.149504
Hypertension	KIR3DL2	Weighted mode	3	−0.00702	0.002289	0.091857

SNP: single nucleotide polymorphism; SE: standard error; Pval: P-value.

Figure 2.

Scatter plot demonstrating odds ratios (OR) for 39 circadian rhythm genes with nominal MR associations with hypertension, ranked from highest to lowest OR.

MR: Mendelian randomization.

After Benjamini–Hochberg FDR correction across all 1276 tested genes, four genes demonstrated statistically significant associations (FDR-adjusted, p < 0.05). PHOSPHO1 showed the most significant result (β = −0.0122, SE = 0.0027, IVW p = 4.97 × 10⁻⁶, FDR-adjusted p = 0.0031, OR = 0.988, 95% CI: 0.983–0.993). ZDHHC18 (β = +0.0120, SE = 0.0028, IVW p = 1.58 × 10⁻⁵, FDR-adjusted p = 0.0046, OR = 1.012, 95% CI: 1.007–1.018), BAIAP3 (β = +0.0092, SE = 0.0022, IVW p = 2.45 × 10⁻⁵, FDR-adjusted p = 0.0046, OR = 1.009, 95% CI: 1.005–1.014), and OSGIN2 (β = −0.0096, SE = 0.0023, IVW p = 2.97 × 10⁻⁵, FDR-adjusted p = 0.0046, OR = 0.990, 95% CI: 0.986–0.995) also met the FDR threshold. No gene met the Bonferroni-corrected threshold (p < 3.92 × 10⁻⁵). These four genes were treated as the primary FDR-supported findings; the other nominal associations were retained as exploratory. Steiger directionality filtering supported an exposure-to-outcome direction for all 617 tested genes (steiger_dir = TRUE throughout), reducing the likelihood of reverse causation in these analyses.

Exploratory diagnostics for a nominal association. For transparency, diagnostic plots were retained for one non-FDR-supported nominal association. The association with TRIB1 did not remain statistically significant after FDR correction (FDR-adjusted p = 0.213), whereas the associations of PHOSPHO1, BAIAP3, OSGIN2, and ZDHHC18 constituted the primary FDR-supported findings. The IVW estimate for the association with TRIB1 was nominally positive (β = 0.0257, p = 0.007); however, the MR-Egger estimate was nonsignificant and directionally different from the IVW estimate. This result is therefore presented only as a hypothesis-generating observation requiring independent replication.

The forest plot (Figure 3(b)) presents Wald ratio estimates for each of the three TRIB1 instrumental SNPs (rs4604455, rs2385094, and rs2594836). The plot documents the SNP-level pattern underlying this nominal association and should not be interpreted as evidence of an FDR-supported finding.

Figure 3.

Exploratory diagnostic plots for the TRIB1 association that did not remain statistically significant after FDR correction.

FDR: false discovery rate.

Gene effect size ranking. The scatter plot of gene rankings based on ORs (Figure 2) shows that all nominal associations had modest effect sizes (OR range = 0.988–1.026), consistent with the polygenic architecture of hypertension. Among protective associations, GNG2 (OR = 0.988; IVW, p = 0.001) exhibited the lowest OR across all 39 genes. The core clock gene PER2 was associated with an increased risk of hypertension (β = +0.0097; OR = 1.010; IVW, p = 0.045), whereas CRY2 showed a protective association (OR = 0.995; IVW, p = 0.024).

Sensitivity analyses

Heterogeneity test. Cochran's Q test employing the MR-Egger and IVW methods was used to assess heterogeneity. In all analyses, the corresponding p-values for the Q test were >0.05, indicating no significant heterogeneity among SNPs (Supplementary Table 2). For the diagnostic example shown in Figure 3, IVW (Q = 4.370 (df = 2, p = 0.112)) and MR-Egger (Q = 3.085 (df = 1, p = 0.079)) were both nonsignificant.

Pleiotropy test (MR-Egger). Horizontal pleiotropy was examined via the intercept term in the MR-Egger regression. The intercept terms for all genes differed nonsignificantly from zero (p > 0.05), indicating a low likelihood of results being biased due to horizontal pleiotropy. This enhances the credibility of the MR findings (Supplementary Table 3). For the four FDR-significant genes, the MR-Egger intercepts were as follows: PHOSPHO1 (Egger intercept = −0.000880, SE = 0.001001, p = 0.413), BAIAP3 (intercept = −0.000533, SE = 0.000889, p = 0.656), OSGIN2 (intercept = −0.001003, SE = 0.001011, p = 0.426), and ZDHHC18 (intercept = +0.002712, SE = 0.002148, p = 0.334). All intercepts differed nonsignificantly from zero.

MR-PRESSO and Steiger directionality. MR-PRESSO analysis was performed for the 16 nominally significant genes with ≥4 SNPs. No significant outlier SNPs were detected in any of these genes (all global test p > 0.05). Among the four FDR-significant genes, BAIAP3 (nSNP = 3) was ineligible for MR-PRESSO; OSGIN2 (global p = 0.739), PHOSPHO1 (global p = 0.436), and ZDHHC18 (global P = 0.398) showed no significant outliers. The remaining 23 nominally significant genes had ≤3 SNPs and could not be assessed using MR-PRESSO. Steiger directionality filtering supported an exposure-to-outcome direction for all 617 tested genes (steiger_dir = TRUE throughout), reducing the likelihood of reverse causation in these analyses.

Leave-one-out test. The funnel plot (Figure 4(a)) plots SNP effect estimates on the x-axis against precision (reciprocal of SE) on the y-axis. The scatter of points is broadly symmetrical around the MR-Egger fitted line, indicating no apparent horizontal pleiotropy in this exploratory diagnostic example. The leave-one-out method was employed to assess the robustness of MR analysis (Figure 4(b)). For the diagnostic example shown in Figure 4, sequential removal of each SNP did not materially alter the pooled estimate. This diagnostic example should be interpreted in the context of its nominal, non-FDR-supported status.

Figure 4.

Exploratory MR diagnostics for a TRIB1 association that did not remain significant after FDR correction.

FDR: false discovery rate.

Functional enrichment analysis

KEGG pathway enrichment. KEGG pathway enrichment analysis revealed that the 39 genes showing nominal MR associations were enriched across multiple pathways closely associated with circadian rhythm regulation and cardiovascular function (Figure 5(a)). Among these, the circadian rhythm pathway exhibited the highest gene count and represented the most significantly enriched pathway, followed by circadian entrainment, apoptosis, biosynthesis of cofactors, and glycosaminoglycan biosynthesis–chondroitin sulfate/dermatan sulfate pathways. The p-value gradient (deeper blue indicates lower p-values) demonstrates the statistical significance of these pathway enrichments, suggesting that circadian rhythm genes participate in the molecular mechanisms of blood pressure regulation by modulating biological processes such as circadian rhythm maintenance, apoptosis, and extracellular matrix synthesis.

Figure 5.

Top 15 enriched (a) KEGG pathways and (b) combined GO terms (content fully displayed).

KEGG: Kyoto Encyclopedia of Genes and Genomes; GO: Gene Ontology.

Gene Ontology (GO) functional enrichment. GO functional enrichment analysis revealed the regulatory functions of the associated gene set at the biological process level, with primary enrichment observed in negative regulation of circadian rhythm, circadian regulation of gene expression, circadian rhythm, regulation of neurotransmitter transport, and deoxyribonucleoside monophosphate metabolic process (Figure 5(b)). The gene counts and p-value gradients for each biological process indicate statistically significant enrichment. These results suggest regulatory roles of the associated genes in biological rhythm regulation and metabolic processes and offer insights into how they may contribute to the molecular basis of hypertension.

Discussion

Biological significance of principal findings

Using circadian gene annotations from the CGDB and hypertension GWAS summary data from OpenGWAS, this two-sample MR study evaluated genetic evidence linking circadian rhythm-related genes to hypertension. Thirty-nine genes showed nominal evidence of association (IVW, p < 0.05), including core clock genes such as PER2 and CRY2. ²⁴ After FDR correction across all tested genes, PHOSPHO1, BAIAP3, OSGIN2, and ZDHHC18 remained statistically significant. These four genes are therefore the principal findings of the study and should be prioritized for replication and functional validation.

The FDR-supported genes point to several plausible biological themes. PHOSPHO1, which showed the strongest corrected statistical evidence, encodes a phosphatase involved in phosphate metabolism and mineralization and may be relevant to vascular calcification or arterial stiffness pathways. BAIAP3, a Munc13-family protein involved in dense-core vesicle biology, may relate to neuroendocrine or autonomic regulation of blood pressure. ²⁵ OSGIN2 is implicated in oxidative stress responses, a process closely connected with endothelial function and vascular tone. ZDHHC18, a predicted palmitoyltransferase, may affect membrane receptor trafficking, G-protein signaling, or ion-channel function. These interpretations should be viewed as mechanistic hypotheses, not established pathways.

The association with TRIB1 did not remain statistically significant after FDR correction (FDR-adjusted p = 0.213). It is therefore retained only as an exploratory candidate and not as a primary finding. Previous studies have linked TRIB1 to cardiometabolic biology; however, these findings should be considered background support for future work rather than confirmation of a causal role in the present analysis.^26–29

Functional enrichment analysis suggested that the associated gene set acts through interconnected biological processes, including circadian rhythm regulation, apoptosis and cofactor biosynthesis, and extracellular matrix-related pathways such as glycosaminoglycan biosynthesis. These pathway-level results support the biological plausibility of the MR findings while remaining hypothesis-generating.

Comparison with existing research

Previous studies support. These findings complement previous observational and experimental evidence. Epidemiological studies have linked shift work, sleep deprivation, and jet lag with increased hypertension risk.^30,31 A UK Biobank analysis involving more than 400,000 participants also reported associations between shift work, poor sleep, elevated blood pressure, and systemic inflammation. ³² Experimental studies showing altered blood pressure rhythms in clock gene mutant or knockout models further support the biological relevance of circadian pathways.^33,34 The present MR analysis adds genetic evidence to this broader literature and avoids claims of definitive mechanism.

Research innovation. The study extends previous work by evaluating a broad circadian gene set rather than focusing on individual clock genes. The use of two-sample MR, multiple sensitivity analyses, and FDR correction provides a transparent framework for prioritizing candidate genes. Importantly, the findings do not support clinical application at this stage. Current hypertension guidelines permit antihypertensive medication at any time of day, and the TIME trial found no significant difference in major cardiovascular outcomes between evening and morning dosing in unselected patients. ³⁵ Genotype-guided chronotherapy therefore remains a hypothesis that requires prospective validation.^36,37

Strengths and limitations

The main strengths of this study include the use of large-scale public GWAS data, multiple MR estimators, ³⁸ complementary sensitivity analyses, and transparent reporting in accordance with the STROBE-MR guidelines. These features reduce, but do not eliminate, concerns about confounding, reverse causation, and analytical instability.

Certain study limitations should also be noted. First, the GWAS datasets were primarily based on European ancestry populations; therefore, generalizability to other populations requires validation. Second, the instruments were not restricted to experimentally validated cis-expression quantitative trait loci (cis-eQTLs); the results therefore reflect genetic variation in circadian gene regions rather than confirmed expression-level effects. Third, 41.3% of retained SNPs were trans-acting variants, which may carry greater pleiotropy risk, although sensitivity analyses did not detect significant horizontal pleiotropy. Fourth, the effect sizes were modest, and the study cannot define molecular mechanisms, gene-environment interactions, or hypertension subtypes. Experimental and clinical validation is required before any translational application.

Conclusion

This two-sample MR analysis identified genetic evidence linking circadian rhythm-related genes to hypertension risk. Among 39 nominally associated genes, PHOSPHO1, BAIAP3, OSGIN2, and ZDHHC18 remained statistically significant after FDR correction and should be regarded as the primary findings. The remaining nominal associations, including that with TRIB1, should be interpreted as exploratory.

Future studies should prioritize functional validation of the four FDR-supported genes, replication in non-European populations, and prospective evaluation of whether circadian genetic information can improve hypertension risk stratification or inform chronotherapy hypotheses.