Skip to main content
Download PDF
Download ePub

Plasma pQTL and brain eQTL integration identifies PNKP as a therapeutic target and reveals mechanistic insights into migraine pathophysiology

The Journal of Headache and Pain volume 25, Article number: 202 (2024) Cite this article

Abstract

Background

Migraine is a prevalent neurological disorder affecting 14.1% of the global population. Despite advances in genetic research, further investigation is needed to identify therapeutic targets and better understand its mechanisms. In this study, we aimed to identify drug targets and explore the relationships between gene expression, protein levels, and migraine pathophysiology.

Methods

We utilized cis-pQTL data from deCODE Genetics, combined with migraine GWAS data from the GERA + UKB cohort as the discovery cohort and the FinnGen R10 cohort as the replication cohort. SMR and MR analyses identified migraine-associated protein loci. Brain eQTL data from GTEx v8 and BrainMeta v2 were used to explore causal relationships between gene expression, protein levels, and migraine risk. Mediation analysis assessed the role of metabolites, and PheWAS evaluated potential side effects.

Results

Four loci were identified: PNKP, MRVI1, CALCB, and INPP5B. PNKP and MRVI1 showed a high level of evidence and opposing effects at the gene and protein levels. PNKP gene expression in certain brain regions was protective against migraine, while its plasma protein levels were positively associated with migraine risk. MRVI1 showed protective effects at the protein level but had the opposite effect at the gene expression level. Mediation analysis revealed that the glutamate to pyruvate ratio and 3-CMPFP mediated PNKP’s effects on migraine. PheWAS indicated associations between PNKP and body composition traits, suggesting drug safety considerations.

Conclusion

PNKP and MRVI1 exhibit dual mechanisms of action at the gene and protein levels, potentially involving distinct mechanistic pathways. Among them, PNKP emerges as a promising drug target for migraine treatment, supported by multi-layered validation.

Peer Review reports

Background

Migraine affects approximately 14.1% of the global population, making it one of the most prevalent neurological disorders, with its incidence continuing to rise [1]. According to the Global Burden of Disease Study 2019, migraine has emerged as the second leading cause of disability worldwide [2]. Due to the heterogeneous nature of symptoms and the limitations of current diagnostic criteria, many patients are unable to receive timely and accurate diagnoses, significantly impacting their quality of life and treatment outcomes [3]. This also imposes a substantial economic and social burden on healthcare systems [4].

In recent years, large-scale genome-wide association studies (GWAS) have substantially advanced our understanding of the genetics of migraine. Through GWAS, researchers have identified several genetic loci associated with migraine susceptibility. For example, a 2022 study identified 123 single nucleotide polymorphism (SNP) risk loci, and a 2023 study further identified 12 subtype-specific loci [5, 6]. These discoveries have enhanced our ability to understand and predict the genetic contributions to migraine, providing critical insights into targeted therapies. The identification of loci such as CALCA and CALCB validated the genetic mechanisms underlying the known drug target, calcitonin gene-related peptide (CGRP) [7]. Currently, CGRP-targeted therapies have proven effective, particularly offering new options for patients unresponsive to traditional treatments [8]. However, the efficacy of targeting new loci remains limited in some populations, suggesting the need for further exploration and identification of additional therapeutic targets to enable more personalized and precise treatment strategies [9].

To better understand and utilize SNP loci identified through GWAS, we rely on quantitative trait loci (QTL) studies to elucidate the impact of genetic variation on various biological traits, such as gene expression (ie: eQTL), protein levels (ie: pQTL), and metabolite concentrations. Additionally, Mendelian randomization (MR) analysis is employed to confirm the causal relationships between genes, proteins, or metabolites and diseases. This method uses genetic variation as an instrumental variable, avoiding confounding factors inherent in traditional observational studies, allowing for more accurate inferences about the causal effects of gene or protein expression on disease. This approach is crucial for identifying potential therapeutic targets more precisely [10].

To date, several studies using QTL and MR analyses have uncovered potential migraine-related targets, such as GSTM4, PHACTR1, and TSPAN2, but most have focused on peripheral tissues such as blood and cerebrospinal fluid, lacking direct validation in brain tissue [11, 12]. In contrast, brain tissue studies could more directly reveal the central nervous regulatory mechanisms underlying migraine. Furthermore, these studies primarily focus on therapeutic targets, with insufficient exploration of potential biomarkers and regulatory factors.

In this study, we employed summary-based MR (SMR) analysis to integrate migraine GWAS, plasma pQTL, and brain tissue eQTL data, enabling precise localization of genetic variants associated with migraine risk. We further validated the causal relationships between these loci and disease risk through multi-layered verification analyses. Additionally, we explored potential metabolites as biomarkers using mediation analysis and assessed the possible side effects of the identified targets through phenome-wide association studies (PheWAS). We aimed to systematically elucidate the genetic and biological mechanisms of migraine and provide robust data support for the design of personalized therapeutic strategies and the development of new drugs.

Methods

Study design

The overall design of our study is illustrated in Fig. 1. Since proteins are the ultimate expression products of genes and serve as the direct executors of function, this study first utilized cis-pQTL data from deCODE Genetics to implement a two-stage SMR method, consisting of a discovery and a replication phase, to identify protein loci associated with migraine. Specifically, the GERA + UKB GWAS meta-analysis dataset was used as the primary discovery dataset, while the FinnGen R10 dataset served as the replication cohort for further validation. The Heterogeneity in Dependent Instruments (HEIDI) test and colocalization analysis were conducted to validate the causal varaint. Next, based on the positive proteins identified through the above process, we utilized cis-eQTL data to explore the causal relationship between gene expression levels in brain tissue and migraine. To further ensure the robustness and directionality of these associations, we performed MR analysis on the positive loci, as well as reverse MR analysis on the proteome. In order to integrate the results from these two different layers of genetic regulation, we employed a two-step mediation analysis to investigate whether gene expression influences migraine pathogenesis by regulating protein levels. For loci with higher levels of evidence, we conducted further mediation analyses to explore whether the associated proteins exert their effects on migraine through certain metabolites or modifiable risk factors. Finally, we performed PheWAS to assess other phenotypes linked to these proteins, thereby providing insights into potential side effects associated with their roles as therapeutic targets.

Fig. 1
figure 1

Flowchart of the study design

Full size image

Datasets

pQTL summary statistics

The data from deCODE Genetics used plasma proteomics technology in a large-scale GWAS. This study analyzed genetic and phenotypic information from 35,559 Icelanders, measuring 4,907 proteins using the SomaScan version 4 multiplex nucleotide testing technique. Researchers identified 18,084 preliminary pQTL associations, covering approximately 94% of the proteins tested. These pQTLs are classified into two categories: 1,881 are cis-regulatory variants located in the same gene expression region, and 16,203 are in distant regulatory (trans) regions. These associations were adjusted for a 1.3% false discovery rate, ensuring the reliability of the statistical results through multiple testing corrections [13].

eQTL summary statistics

The cis-eQTL summary data were sourced from GTEx v8 and BrainMeta v2. The GTEx v8 dataset includes samples from 948 donors, predominantly of European ancestry, and provides RNA sequencing data from 54 tissues, including 13 brain tissues, identifying cis-eQTLs within a 1 Mb region near each gene [14]. The BrainMeta v2 dataset covers eQTL mapping results from 2,865 cortical samples from 2,443 individuals of European ancestry, mapping cis-eQTLs within a 2 Mb range near each gene [15].

Migraine summary statistics

Summary data for migraine were obtained from two major studies. The first study utilized meta-GWAS data from the Genetic Epidemiology Research on Adult Health and Aging (GERA) cohort and the UK Biobank (UKB). The database integrates genotypic and phenotypic data from two large cohorts, encompassing a total of 554,569 participants, including 28,852 migraine cases and 525,717 controls. Approximately 85% of the participants are of European ancestry, with 77.98% of the cases and 53.22% of the controls being female [16]. The second set of data comes from FinnGen Release 10, which includes 20,908 migraine patients and 312,803 controls averaging 39.54 years in age. Migraine cases were identified using the criteria from the International Classification of Diseases, Tenth Revision (ICD-10), supplemented by medication purchase records and prescription codes, such as ATC codes [17].

Plasma metabolite data

The plasma metabolite data were derived from the Canadian Longitudinal Study on Aging (CLSA), encompassing 1,458 plasma metabolites measured in 8,299 European-ancestry participants. These data were obtained using ultra-performance liquid chromatography-tandem mass spectrometry (UPLC-MS/MS) and underwent stringent quality control and standardization processes [18].

Cerebrospinal fluid (CSF) metabolite data

The cerebrospinal fluid metabolite data were sourced from the Wisconsin Alzheimer’s Disease Research Center (WADRC) and the Wisconsin Registry for Alzheimer’s Prevention (WRAP) longitudinal cohort studies. These data were analyzed using high-resolution mass spectrometry and covered 338 CSF metabolites. To enhance the generalizability and accuracy of the analysis, only cognitively healthy participants of European ancestry were included, with a final sample of 291 individuals, the average age being approximately 60 years, and two-thirds of the participants were female [19].

Modifiable risk factor data

The modifiable risk factor data were obtained from the public database of the Medical Research Council Integrative Epidemiology Unit (MRC-IEU). Modifiable risk factors refer to those factors that can be managed or reduced through lifestyle changes, medical interventions, or environmental adjustments, and are commonly associated with chronic diseases and health outcomes. The risk factors included in this study were glycine levels, smoking status, waist-to-hip ratio adjusted for BMI, creatinine levels, serum uric acid levels, waist circumference, adult BMI sex-combined, low-density lipoprotein (LDL) cholesterol, HbA1C, alcohol consumption, triglycerides, total cholesterol, high-density lipoprotein (HDL) cholesterol, fasting insulin, fasting glucose, and vigorous physical activity. The basic information of the datasets is shown in Table S1.

PheWAS phenotype data

The phenotype data were sourced from six major databases, most of which were based on genetic and phenotypic information from the UKB and were analyzed in depth by different research teams to provide extensive genetic association data. The MRC-IEU provided 2,514 phenotypes based on analyses in European populations. The PheWeb platform combined data from UKB-SAIGE and UKB-TOPMed, providing 1,403 phenotypes, predominantly covering European populations [20, 21]. The Neale Lab dataset published in 2018 included 4,359 phenotypes, all based on European populations, with sex-stratified analyses that included 3,407 male phenotypes and 3,547 female phenotypes. The 2020 Neale Lab PAN-UKB dataset further extended the analysis to multi-ethnic populations, covering 7,228 phenotypes [22]. Additionally, the FinnGen R10 dataset, primarily based on the Finnish population, provided 2,408 phenotypes [17].

cis-pQTL SMR analysis and HEIDI tests

Instrumental variables were selected from the deCODE Genetics database, which contains a total of 4,907 pQTLs. We specifically focused on a subset of 1,429 cis-acting pQTLs, each of which contains at least one cis-pQTL. Each SNP in these selected pQTLs exhibited at least suggestive evidence of association with protein expression levels, with a P-value less than 5 × 10⁻⁸. These selected SNPs were rigorously aligned to ensure consistency and accurate matching between exposure and outcome datasets.

MR relied on three core assumptions: 1) genetic variation must be correlated with exposure; 2) genetic variation must be independent of any confounders; 3) genetic variation must not have a direct effect on the outcome [23]. The SMR method, an extension of the MR concept, was developed to investigate the pleiotropic associations between genetically determined traits such as gene expression, protein abundance, DNA methylation, or metabolite levels (as exposures) and complex traits like disease phenotypes (as outcomes) [24]. In this study, we followed the core principles of MR, using the following formula to calculate the causal relationship:

$$\beta_{pqtl-migraine}=\beta_{SNP-migraine}/\beta_{SNP-pqtl}$$

where βSNP-pqtl represents the effect of an SNP on cis-pQTL, while βSNP-migraine indicates the effect of the same SNP on migraine risk. This formula estimates the potential impact of cis-pQTL on migraine risk.

After completing the initial SMR analysis, we further filtered the results based on beta values and statistical significance to identify the top SNP for each protein—the genetic variant with the greatest impact on protein expression levels. We subsequently applied the False Discovery Rate (FDR) method for multiple hypothesis testing corrections to adjust p-values, and then selected protein loci whose adjusted p-values were less than 0.05 for further analysis [25]. The odds ratio (OR) represents the change in migraine risk with each natural log unit increase in pQTL levels.

Subsequently, to distinguish whether the association between positive loci and migraine susceptibility was due to vertical pleiotropy or linkage disequilibrium (LD), we conducted the HEIDI test. This method tests the heterogeneity of the top 20 associated SNPs that are in LD with the top SNP [24]. Specifically, we assess the differences in effect estimates between pQTL and migraine for these SNPs using the following formula to calculate the difference between the βpqtl-migraine(0) associated with the top SNP and the βpqtl-migraine(i) of other less significant SNPs:

$$di=\beta_{pqtl-migraine(i)}-\beta_{pqtl-migraine(0)}$$

According to the null hypothesis (existence of a single shared causal variant), d should equal zero. If the HEIDI test (PHEIDI ≥ 0.05) is passed, it indicates that we do not reject the hypothesis of a single causal variant, thus ruling out confounding effects from LD.

To validate the primary findings from the GERA + UKB migraine result database, we used the independent dataset FinnGen R10 for replication analysis.

cis-eQTL SMR analysis and HEIDI tests

To further explore the relationships between genes, proteins, and disease, we extended the pQTL analysis by introducing cis-eQTL SMR and HEIDI analyses in brain tissues, focusing on the direct associations at the gene level within the central nervous system. By analyzing the relationship between gene expression and migraine risk, we aimed to identify gene regulatory mechanisms that may drive protein-level changes. Specifically, we obtained cis-eQTL data from the gene regions corresponding to the positive protein loci and used the effect of these eQTLs on gene expression (βSNP-eQTL) along with their effect on migraine risk (βSNP-migraine) to perform SMR analysis to estimate the potential causal effect of gene expression levels on migraine risk. Results with a P-value < 0.05 were considered significant, and HEIDI test results with P-value > 0.05 supported the hypothesis that the association was due to a shared causal variant. This analysis indicated that gene expression in these loci might have a causal relationship with migraine risk. Both SMR and HEIDI tests were conducted using the SMR tool (SMR v1.3.1) [24].

Colocalization analysis

For greater reliability of our results, we performed Bayesian colocalization analysis using the "coloc" package [26]. This analysis investigated whether the identified positive loci and migraine risk share the same genetic variant. Colocalization analysis is based on five hypotheses: (i) absence of any causal variant influencing either protein expression levels or migraine within the genomic region (H0); (ii) a sole causal variant influencing protein expression levels exclusively (H1); (iii) a lone causal variant impacting migraine alone (H2); (iv) two separate causal variants affecting protein expression levels and migraine, respectively (H3); (v) a common causal variant affecting both protein expression levels and migraine (H4) [27]. For each protein locus, SNPs within ± 1000 kb of the pQTL region were included in the analysis. We use default parameters for the prior probabilities: p1 = 1 × 10⁻4 for an SNP being linked to the protein, p2 = 1 × 10⁻4 for an SNP being linked to migraine, and p12 = 1 × 10⁻5 for an SNP being linked to both [27]. The analysis provided posterior probabilities (PP) for the five hypotheses, with PP.H4 > 80% considered strong evidence of colocalization [26]. This further validated the SMR results, confirming their biological significance and offering robust support for further research.

MR analysis of positive loci and reverse MR analysis

After completing SMR analyses for two different layers of gene regulation and identifying loci significantly associated with migraine, we further conducted two-sample MR analyses on the protein and gene expression levels of these specific loci using the "TwoSampleMR" package. This step aimed to validate the findings and ensure robustness. The selection criteria for instrumental variables were as follows: each SNP had to show a significant association with either protein or gene expression levels (P-value < 5 × 10⁻⁸), and the F-statistic had to exceed 10, confirming the robustness of the instrumental variable. In addition, we applied an r2 threshold < 0.01 and a 5,000 kb window to ensure the independence of SNPs. For analyses with more than one instrumental variable, we primarily used the inverse variance weighted (IVW) method to derive causal effect estimates. The IVW method aggregates the effects of individual SNPs and weights them by the inverse of their variance, providing an overall causal estimate. This method is most valuable when all instrumental variables are valid, as it typically provides the most accurate estimate [28]. We also used the Q statistic to assess the heterogeneity of the genetic instruments and evaluate the consistency of the instruments. Additionally, we performed other analyses, including MR-Egger, weighted median, simple mode, and weighted mode methods. MR-Egger's intercept test allows for a non-zero intercept, which helps detect and adjust for horizontal pleiotropy [29]. When there was only one instrumental variable, we used the Wald ratio method to estimate the causal effect [30].

Since proteins are expression products of disease genes and may be regulated by disease states, we performed reverse MR analyses to clarify the causal direction between migraine and specific protein expression levels and to rule out any potential reverse causality. A more lenient P-value threshold of 5 × 10⁻⁶ was applied to select instrumental variables associated with migraine, retaining as many SNPs as possible to enhance the detection of causal relationships. We calculated the F-statistic for each SNP to assess the robustness of the instrumental variables, ensuring that all selected SNPs had an F-statistic greater than 10. Subsequently, we performed LD clumping with an r2 threshold < 0.01 and a 5,000 kb window to ensure SNP independence. We then applied robust statistical methods for reverse MR analysis, including the IVW, MR-Egger, weighted median, weighted mode, and simple mode methods. Combining these approaches ensured a comprehensive evaluation of and exclusion of reverse causality.

Multi-omics integration and mediation analysis

We integrated all the above results and ranked candidate gene loci based on different validation analyses. The loci were categorized into three tiers, all of which required evidence of a causal association with migraine at both the protein and gene levels through SMR analysis: 1) Tier 1 loci passed all validation analyses (including MR analysis, HEIDI test, and colocalization analysis), with a PPH4 > 0.9; 2) Tier 2 loci passed most validation analyses (including some MR analyses, HEIDI test, and colocalization analysis), with a PPH4 > 0.9; 3) Tier 3 loci passed partial validation analyses (including some MR and HEIDI tests) but did not pass colocalization analysis.

Next, to further understand the regulatory mechanisms at different gene levels, we first explored the causal relationship between gene expression and protein levels through MR analysis and investigated the potential mechanisms by which gene expression indirectly affects migraine via protein levels using mediation analysis. In the first step, we assessed the causal effect of gene expression on protein levels (β1 value) using MR analysis. In the second step, we used the causal effect of protein levels on migraine outcomes (β2 value) and the product method (β1 × β2) to calculate the indirect effect of each mediator, and the significance of the mediation effect was determined using the Delta method [31]. Similarly, to gain a more comprehensive understanding of the mechanism by which positive proteins affect migraine, We applied the same two-step approach to explore whether potential mediators (e.g., plasma and cerebrospinal fluid metabolites or modifiable risk factors) play a role in the causal relationship between positive proteins and migraine. The MR analysis was primarily based on the IVW random effects model, and we used the MR-Egger model to detect and adjust for potential horizontal pleiotropy while using Cochran's Q test to assess heterogeneity among instrumental variables. For MR analyses with only one SNP, we used the Wald ratio method to estimate causal effects. All MR analyses were performed using the “TwoSampleMR” package.

PheWAS analysis

We further conducted an MR-PheWAS analysis to systematically investigate potential causal associations between these positive protein targets and other traits to assess their safety in drug development [32]. Instrumental variables for the PheWAS analysis were based on the SNPs significantly associated with migraine from the previous MR analysis, representing the causal effects of specific protein targets. First, we screened phenotype data from six major databases with sufficiently large sample sizes from European populations to ensure the reliability of the analysis results. Next, we extracted phenotype data that showed significant associations with our SNP set of interest, covering a wide range of traits related to various biomarkers and clinical characteristics. During the analysis, we primarily used the IVW method to evaluate the causal relationship between each target and multiple traits, with FDR correction methods applied to control for false-positive rates and ensure the robustness of the results. We then used the Q statistic to test for heterogeneity, ultimately identifying phenotypes that were significantly associated with the target, with adjusted P-values below 0.05 and no evidence of heterogeneity. This analysis provided new insights into complex gene-trait relationships while offering data to support the safety assessment of these targets.

Results

Plasma protein expression and migraine

In this study, we extracted 1,429 genes containing cis-pQTLs from the deCODE database (Table S2) and identified four genes (MRVI1, PNKP, CALCB, and INPP5B) with significant causal associations with migraine susceptibility through SMR analysis in the discovery cohort. Subsequently, we successfully validated the MRVI1 locus and identified a novel locus, AIF1, in the replication cohort. All loci passed the HEIDI test, but only MRVI1 and PNKP passed the coloc analysis, suggesting that these loci may imply a true causative effect on migraine risk.

In the discovery cohort, SMR analysis identified four gene loci linked to the risk of migraine. Variants in MRVI1 (OR = 0.620, 95% CI: 0.514–0.749, P_adj = 9.051 × 10–4, P_HEIDI = 0.910) and CALCB (OR = 0.828, 95% CI: 0.765–0.895, P_adj = 1.547 × 10–3, P_HEIDI = 0.396) were linked to a lower risk of migraine, whereas variants in PNKP (OR = 1.588, 95% CI: 1.295–1.947, P_adj = 4.180 × 10–3, P_HEIDI = 0.833) and INPP5B (OR = 1.769, 95% CI: 1.335–2.345, P_adj = 2.595 × 10–2, P_HEIDI = 0.999) were linked to a higher risk of migraine. In the replication phase, the MRVI1 locus (OR = 0.568, 95% CI: 0.459–0.704, P_adj = 3.473 × 10–4, P_HEIDI = 0.977) was identified, consistent with the results from the discovery cohort. Additionally, a novel locus, AIF1 (OR = 1.400, 95% CI: 1.196 to 1.639, P_adj = 2.149 × 10–2, P_HEIDI = 0.506), was identified and linked to a increased risk of migraine. In both the discovery and replication cohorts, all identified positive loci passed the HEIDI test (Table S3, Figs. 2, 3).

Fig. 2
figure 2

Forest plot of the causal relationship between protein loci and migraine risk

Full size image
Fig. 3
figure 3

Volcano plot showing results from the two-stage (discovery and replication) SMR analysis

Full size image

Additionally, colocalization analysis provided strong evidence of colocalization between migraine risk and the protein loci MRVI1 (PP.H4.abf = 99%) and PNKP (PP.H4.abf = 96.2%) that passed both SMR and HEIDI tests. However, CALCB (PP.H4.abf = 6.24%, PP.H3.abf = 93.7%) and INPP5B (PP.H4.abf = 0.0674%, PP.H3.abf = 95.5%) showed no evidence of colocalization with migraine. In the replication phase, MRVI1 (PP.H4.abf = 99.4%) again passed the colocalization analysis. These consistent results not only reinforced the evidence supporting MRVI1 as a potential protective factor for migraine but also highlighted its consistency across different populations and genetic backgrounds. However, AIF1 (PP.H4.abf = 0.0841%, PP.H3.abf = 93.1%) did not show evidence of colocalization with migraine. This discrepancy underscores the need to use multiple analytical methods to fully understand gene-phenotype associations. Colocalization plots for various protein loci can be found in Fig. 4.

Fig. 4
figure 4

A-F represent the colocalization scatter plots and Manhattan plots for the protein loci MRVI1 (discovery and replication phases), PNKP, CALCB, INPP5B, and AIF1. In each panel, the left plot compares the -log10(p) values of each single nucleotide polymorphism (SNP) for migraine risk and protein expression, with the color indicating the linkage disequilibrium (r2) with the lead SNP (purple diamond). The right plots show the genomic positions on the x-axis and the -log10(p) values for SNPs from the migraine GWAS (top) and the protein expression study (bottom) on the y-axis

Full size image

Furthermore, the results of MR analyses indicated causal relationships between PNKP (OR = 1.502, 95% CI: 1.247–1.808, P_value = 1.790 × 10⁻5) and CALCB (OR = 0.843, 95% CI: 0.757–0.940, P_value = 2.115 × 10⁻3) with migraine risk. The weighted median and weighted mode methods further supported these significant associations. The MR-Egger intercept did not show significant horizontal pleiotropy, and Cochran's Q test did not detect any heterogeneity. No significant results were observed for other proteins (detailed data in Table S4).

Finally, to determine the direction of the causal relationship between pQTL and migraine risk, we performed reverse MR analysis. The analysis showed that for all protein expression loci, including those identified in the primary and validation analyses, the results of the IVW method, MR Egger, weighted median, simple mode, and weighted mode methods were consistent, with no significant evidence of reverse causality (detailed data in Table S5).

Brain tissue gene expression and migraine

To gain a more comprehensive and direct understanding of how gene expression regulation affects disease outcomes, we used cis-eQTL data from various brain tissue databases for multi-level validation. Using SMR analysis and the HEIDI test in the GTEx v8 database, we found significant associations between gene loci (PNKP, MRVI1, CALCB and INPP5B) and migraine risk in certain brain tissues. Specifically, PNKP gene expression in multiple brain tissues, including the anterior cingulate cortex (BA24), caudate nucleus, cerebral cortex, frontal cortex (BA9), hippocampus, nucleus accumbens, and putamen, was significantly negatively associated with migraine risk. Conversely, MRVI1 gene expression in the anterior cingulate cortex (BA24), caudate nucleus, and substantia nigra was positively associated with migraine risk. Furthermore, INPP5B gene expression in the anterior cingulate cortex (BA24), caudate nucleus, hippocampus, hypothalamus, and nucleus accumbens was significantly negatively associated with migraine risk. CALCB gene expression in the cerebellum was also significantly negatively associated with migraine risk. In the BrainMeta v2 cortical database, we again validated the association of PNKP and MRVI1/IRAG1 with migraine, with consistent effect directions. Additionally, colocalization analysis showed that all PNKP and MRVI1 loci passed coloc analysis (PP.H4 > 90%), while CALCB and INPP5B loci did not pass coloc analysis(detailed data available in Table S6).

To further validate the robustness of the results, we conducted additional MR analyses. The results confirmed that, except for some loci where effective SNPs could not be selected due to issues during data alignment, the expression of the PNKP gene in the anterior cingulate cortex (BA24), cerebral cortex, nucleus accumbens, and putamen, as well as the expression of the MRVI1 gene in the anterior cingulate cortex (BA24) and substantia nigra, were significantly associated with migraine risk, with effect directions consistent with previous findings. Additionally, the expression of the CALCB and INPP5B genes in various brain regions was also found to be significantly associated with migraine risk, with consistent effect directions (detailed data available in Table S7).

Integration of multi-omics evidence

By combining the results from two different levels of gene regulation, we classified the identified gene loci based on evidence strength. PNKP passed all analyses and was classified as Tier 1. MRVI1, which did not pass MR analysis at the protein level, was classified as Tier 2. CALCB and INPP5B, which failed both colocalization analysis and MR analysis at the protein level, were classified as Tier 3.

Given the observation that the effects of PNKP, MRVI1, and INPP5B loci on migraine risk were opposite at the gene expression and protein expression levels, we used MR analysis to assess the causal relationship between gene expression and protein levels. Several brain tissues showed significant causal associations between gene expression and plasma protein levels. Specifically, PNKP gene expression in the anterior cingulate cortex (BA24), cerebral cortex, nucleus accumbens, and putamen was negatively associated with plasma protein levels, as was MRVI1 gene expression in the anterior cingulate cortex (BA24) and substantia nigra. INPP5B gene expression in the anterior cingulate cortex (BA24), hippocampus, hypothalamus, and nucleus accumbens was negatively associated with plasma protein levels, while CALCB gene expression in the cerebellum was positively associated with plasma protein levels. These results suggest that these loci may indirectly influence migraine pathogenesis by regulating plasma protein levels at the gene level. Further mediation analysis confirmed that PNKP gene expression in the anterior cingulate cortex (BA24), nucleus accumbens, cerebral cortex, and putamen reduces migraine risk by lowering plasma protein levels, with mediation accounting for over 65% of the effect (detailed data in Table S8, S9).

Follow-up analysis of the PNKP locus

Given the close relationship between changes in plasma and CSF metabolites and migraine attacks, and considering that certain controllable risk factors are often associated with chronic diseases, we conducted two-step mediation analyses to explore whether the positive protein loci identified influence migraine risk through metabolites or controllable risk factors. The IVW method showed that the glutamate to pyruvate ratio and 3-carboxy-4-methyl-5-pentyl-2-furanpropanoate (3-CMPFP) played a significant mediating role in the causal effect of PNKP on migraine (P_value = 4.60 × 10⁻2), with a mediation effect value (β1 × β2) of 0.02 (detailed data in Table S10).

Finally, we used MR-PheWAS analysis to investigate the potential side effects of the PNKP protein locus as a therapeutic target for migraine. The SNPs were selected from the previous MR analysis of protein loci and migraine. The results showed significant associations between PNKP and multiple phenotypes across two phenotype databases. In the MRC-IEU database, PNKP was significantly associated with 22 different phenotypes, primarily related to body composition (e.g., lean body mass, impedance of the whole body and limbs), metabolic rate, and specific clinical conditions (e.g., atrial fibrillation). Due to heterogeneity (Q_Pval < 0.05), two phenotypes related to trunk lean mass were excluded from the final analysis. Of the remaining 20 phenotypes, PNKP showed a negative effect on phenotypes related to body impedance (e.g., whole-body and limb impedance), suggesting that higher PNKP expression may be associated with lower impedance (i.e., higher muscle mass and lower fat content). PNKP also showed positive effects on lean body mass, predicted mass, and basal metabolic rate in the whole body and limbs. In the Neale Lab database with sex-stratified data, PNKP was significantly associated with 6 phenotypes in females, primarily showing a negative effect on whole-body and arm impedance and a positive effect on lean mass and predicted mass in the arms. In males, PNKP was significantly associated with four phenotypes, similarly showing negative associations with whole-body and arm impedance and positive associations with behavioral traits (e.g., frequency of mobile phone use and alcohol intake). Notably, in both the MRC-IEU and Neale Lab databases, PNKP was significantly associated with migraine phenotypes, directly confirming the core focus of our study. Figure 5 presents the Manhattan plot illustrating the associations between genetic variants at the PNKP locus and multiple phenotypes in the MR-PheWAS analysis.

Fig. 5
figure 5

Manhattan plot for the MR-PheWAS analysis of PNKP protein locus. Panel (A) shows results from the MRC-IEU database, while panels (B) and (C) show results from the Neale Lab database for females and males, respectively. In panel (A), two phenotypes—Trunk fat-free mass and Trunk predicted mass—were excluded due to heterogeneity. The x-axis represents phenotype categories, and the y-axis represents -log10(P) values, where higher points indicate smaller P-values. The red dashed line indicates the FDR-corrected significance threshold (P_adj < 0.05)

Full size image

Discussion

In this study, through the integration of multi-omics data from plasma pQTL and brain tissue eQTL analyses, we identified four gene loci that may be significantly associated with migraine: PNKP, MRVI1, CALCB, and INPP5B. Among these, PNKP showed the most compelling evidence, having been validated at both the plasma protein level and gene expression across multiple brain tissues, classified as Tier 1 evidence. Similarly, MRVI1 displayed significant associations with migraine across two distinct regulatory levels, though it did not pass protein-level MR analysis, it was validated in the replication cohort, warranting Tier 2 evidence. CALCB and INPP5B were identified as being associated with migraine at both the protein and gene expression levels, but only passed partial validation, providing moderate support consistent with Tier 3 evidence. For the PNKP locus, we observed that gene expression in multiple brain regions may influence migraine risk through regulation of plasma protein levels. Additionally, certain plasma metabolites may mediate the causal relationship between PNKP protein levels and migraine risk. Finally, MR-PheWAS analysis provided insight into potential side effects associated with PNKP as a therapeutic target.

Previous studies, utilizing transcriptome-wide association studies (TWAS) and Functional Mapping and Annotation (FUMA)-based gene functional annotation and prioritization of GWAS results, identified a significant association between PNKP gene expression in brain tissue and migraine risk [16, 33]. Additionally, a related MR analysis provided preliminary evidence that plasma PNKP protein levels are positively associated with migraine risk [12]. Our study further expanded the evidence of the relationship between PNKP and migraine by investigating both protein and gene regulatory layers. Our findings suggest that increased PNKP gene expression in multiple brain regions may serves as a protective factor against migraine, while elevated plasma PNKP protein levels are indicative of an increased risk of migraine.

At the gene level, given that PNKP is a key DNA repair protein, we speculate that increased expression of PNKP in brain tissue indicate an enhanced repair capacity [34, 35]. This repair process likely reduces DNA damage, which in turn decreases cytoplasmic immune-activating factors, suppresses the activation of the STING pathway, and may reduce type I interferon release, thereby alleviating oxidative stress and neuroinflammation [36]. Our analysis specifically demonstrates that increased PNKP expression in the anterior cingulate cortex (BA24), nucleus accumbens, cerebral cortex, and putamen is associated with a reduced risk of migraine. The pathophysiology of migraine is complex, involving processes such as neuroinflammation, cerebrovascular responses, oxidative stress, and neuronal damage. These dysregulated mechanisms often lead to functional impairments in the nervous system, and long-term chronic impacts may even result in structural changes in brain tissue [37]. The brain regions identified in our study align with known regions exhibiting functional and structural abnormalities in migraine. The anterior cingulate cortex as a core region for pain regulation and emotional response, where increased activity in migraine patients often leads to heightened pain sensitivity and emotional dysregulation [38]. Furthermore, during acute migraine attacks, the visual cortex shows significantly enhanced activity, with cortical spreading depression (CSD) potentially triggered by hyperexcitability in this region, contributing to visual aura [39]. In contrast, the nucleus accumbens and putamen exhibit reduced activity in migraine patients, both of which are involved in the suppression of dopamine signaling, potentially leading to sustained pain and negative emotional states [40]. These long-term functional disturbances in chronic migraine patients further contribute to cortical thinning and reduced brain volume [41]. In summary, PNKP helps maintain neuronal stability in these brain regions, reducing dysfunction and preventing long-term neurodegeneration and pathological structural changes [35].

At the protein level, our findings are consistent with those of previous MR analysis, although the study did not provide a clear explanation [12]. Through further mediation analysis, we found that a decrease in plasma PNKP protein levels may lower migraine risk by reducing the plasma glutamate to pyruvate ratio or increasing plasma 3-CMPFP levels. Among these, the glutamate to pyruvate ratio, as a downstream effect indicator of plasma PNKP protein levels, may influence migraine risk due to glutamate’s key role as an excitatory neurotransmitter in migraine. Animal studies have shown that pyruvate’s role in “glutamate clearance” exhibits neuroprotective effects, suggesting the potential for similar neuroprotective benefits in migraine. By creating a concentration gradient, lowering glutamate levels in plasma could reduce the neurotoxic effects of glutamate in the nervous system, thereby preventing migraine onset [42, 43]. Additionally, although a direct link between 3-CMPFP and migraine has not been fully established, studies indicate that 3-CMPFP is closely related to lipid and amino acid metabolism. For example, regulating the metabolism of long-chain polyunsaturated fatty acids (n3 and n6) may help suppress pro-inflammatory responses, while regulating amino acids such as glutamine may enhance the production of antioxidants like glutathione, further reducing oxidative stress and thereby lowering migraine risk [44, 45]. Thus, plasma PNKP, by influencing plasma metabolites, emerges as a potential drug target for migraine treatment. However, to avoid disrupting PNKP’s critical role in neuronal repair, drug development could follow the approach used in CGRP antibodies: designing large-molecule antibodies to restrict the therapeutic effect to the peripheral blood system through the natural barrier of the blood–brain barrier, ensuring safety and specificity [46]. Moreover, drug development should consider the risk of metabolic side effects. PheWAS analysis results suggest that PNKP may affect body composition, such as muscle mass and fat metabolism, with more significant effects in females, indicating that sex differences must be considered to ensure precise and safe therapeutic outcomes. Furthermore, downstream metabolites of plasma PNKP, especially the easily detectable glutamate to pyruvate ratio, may serve as biomarkers for monitoring treatment efficacy, enabling personalized management by assessing patient response to therapy.

Our findings suggest that plasma PNKP protein levels and PNKP gene expression in brain tissue may influence the occurrence of migraine through two independent pathways. Although mediation analysis indicates that PNKP gene expression in brain tissue might impact migraine risk via plasma PNKP protein levels, the integration of cross-tissue data requires caution when inferring this causal chain due to potential limitations related to tissue and population differences. Specifically, changes in PNKP gene expression observed in brain tissue may not be fully reflected in plasma protein levels, as this relationship could be influenced by tissue-specific regulatory factors and individual variability. Additionally, the cross-sectional nature of the data limits our ability to comprehensively understand the dynamic and time-dependent regulation of PNKP expression at different levels throughout the progression of migraine. Based on the current observational associations, further experimental validation in clinical and in vitro models is needed to directly assess causal links across tissues and to explore the precise connection between these two pathways.

The causal relationship between the MRVI1 locus and the risk of migraine also warrants attention. Previous GWAS and functional genomics studies (TWAS, PWAS) have shown an association between MRVI1 and migraine, with shared genetic links to vascular diseases such as cervical artery dissection and stroke, suggesting it may influence migraine risk through vascular regulation [16, 33, 47, 48]. Additionally, previous MR studies suggest a negative correlation between MRVI1 protein levels and migraine, while our study also indicates a protective role of plasma MRVI1 protein levels against migraine onset [12]. The IRAG1 protein, encoded by the MRVI1 gene, is believed to maintain intracellular calcium homeostasis by regulating calcium release from the endoplasmic reticulum, thereby preventing calcium overload within cells. By precisely modulating the function of the IP3 receptor, IRAG1 may help prevent excessive neurotransmitter release, which in turn suppresses neuronal hyperexcitability and mitigates neuroinflammatory responses [49, 50]. Furthermore, after phosphorylation, IRAG1 can influence vascular function through the cGMP-PKG signaling pathway, helping to stabilize blood flow and reduce abnormal vasoconstriction, thereby lowering the risk of migraine attacks. At the gene expression level in brain tissues, our analysis revealed a positive correlation between MRVI1 expression in the anterior cingulate cortex, caudate nucleus, and substantia nigra with migraine risk. This is in contrast to the protective role observed at the protein level. To further investigate this discrepancy, we conducted an MR analysis from eQTL to pQTL, which showed that higher MRVI1 gene expression in the anterior cingulate cortex and substantia nigra was negatively correlated with plasma IRAG1 protein levels. In other words, increased expression of MRVI1 in these brain regions may prioritize local regulation of calcium homeostasis and neuroinflammation, subsequently reducing IRAG1 protein levels in plasma. This decrease in protein levels may weaken the protective effects of IRAG1 in the circulatory system against migraines. Thus, MRVI1 exhibits a complex regulatory pattern at both the gene and protein levels. Future research could further explore this intricate regulatory network, particularly the specific mechanisms at play in the anterior cingulate cortex and substantia nigra, to better understand the dual role of MRVI1 in migraine.

Additionally, we identified the relationship between CALCB and INPP5B variants and migraine risk; however, these loci did not pass subsequent colocalization analysis, warranting cautious interpretation. Notably, these loci have been preliminarily linked to migraine in previous GWAS, MR, or TWAS studies [12, 16, 47]. The CALCB gene encodes β-CGRP, a peptide known to play a role in pain perception and vascular function regulation. Although β-CGRP may not have as prominent a role in migraine pathophysiology as α-CGRP, it is still implicated in migraine mechanisms [51]. Our study found a significant association between CALCB expression in the cerebellum and migraine risk, indicating that β-CGRP may play a role in migraine pathogenesis by regulating pain perception and neural functions in the cerebellum, suggesting that the cerebellum could be a key region for CGRP regulation. Previous research has demonstrated that the cerebellum is not only involved in motor control and coordination but also plays important roles in cognitive function and pain perception [52]. This provides a potential perspective on the role of CALCB in migraine, and future studies should explore the specific role of CALCB in the cerebellum in greater depth. There is limited research on the association between the INPP5B locus and migraine, and its biological mechanism may be indirectly related to known migraine pathways. INPP5B encodes a phosphatase involved in phosphoinositide metabolism and may influence migraine occurrence by regulating neuronal excitability and inflammatory responses [53]. eQTL analysis suggests that INPP5B gene expression in regions such as the anterior cingulate cortex, caudate nucleus, nucleus accumbens, hippocampus, and hypothalamus is associated with migraine risk. These brain regions overlap with areas affected by aberrations in PNKP and MRVI1, suggesting the possibility of shared underlying mechanisms. Future research should further investigate the role of these brain regions in neuronal excitability, neuroinflammation, and metabolic regulation to more comprehensively uncover the neurobiological basis of migraine.

Our study employed a comprehensive and rigorous methodological design, incorporating multiple analytical approaches and a multi-cohort design to strengthen the robustness and reliability of the results. SMR analysis utilized cis-pQTLs and cis-eQTLs with larger effect sizes and stronger association significance, which meaningfully enhanced statistical power. At the same time, the use of both plasma protein levels and brain tissue gene expression allowed for cross-validation, potentially providing a more comprehensive biological interpretation. HEIDI tests, colocalization analysis, as well as MR and reverse MR analyses helped to reduced confounding factors, further enhancing the robustness of the findings. Mediation analysis deepened our understanding of how genetic variants influence disease risk through metabolic pathways, while PheWAS contributed to validating target specificity and helped identify potential off-target effects.

However, several limitations should be considered. First, the study focused on individuals of European ancestry, which limits the generalizability of these findings. Future studies will need to validate the results in more diverse populations to ensure broad applicability. Second, due to the lack of subgroup GWAS data, we were unable to conduct stratified analyses, which may have led to missing certain specific genetic effects related to migraine with aura or distinctions between episodic and chronic migraine types. Future research should incorporate more comprehensive data resources to enable detailed stratified analyses and further refine our understanding of the genetic mechanisms underlying migraine. Third, this study relies on bioinformatic and statistical analyses, and currently, no direct experimental evidence exists between the target and the disease. While MR analysis provides valuable insights into potential causal relationships from real-world data, the complexity and variability inherent in such data also pose limitations, making it challenging to control for all confounding variables. Therefore, further validation is needed through experimental studies, including wet-lab functional experiments and large-scale epidemiological studies. Subsequent animal and cellular studies would be especially valuable in strengthening causal inferences and providing robust support for the clinical application of these findings.

In conclusion, this study provides substantial evidence and new insights into the potential causal relationship between PNKP and MRVI1 variants and migraine risk, offering foundational data to support further exploration of PNKP as a potential therapeutic target.

Data availability

The summary statistics for the combined multiethnic meta-analysis (GERA + UKB) GWAS can be downloaded from the NHGRI-EBI GWAS Catalog using study accession number GCST90000016 at the following link: https://www.ebi.ac.uk/gwas/downloads/summary-statistics. The summary data from the FinnGen GWAS can be accessed at https://www.finngen.fi/en/access_results. The pQTL summary data from deCODE Genetics can be accessed at http://www.decode.com/summarydata/. The eQTL summary data from the BrainMeta v2 and GTEx v8 datasets can be accessed at https://yanglab.westlake.edu.cn/software/smr/#eQTLsummarydata and https://gtexportal.org/home/downloads/adult-gtex/qtl, respectively. The summary statistics for plasma metabolite data can be accessed at the following link: https://www.ebi.ac.uk/gwas/, with accession numbers for European GWASs being GCST90199621-90201020. The Cerebrospinal Fluid (CSF) Metabolite Data full GWAS meta-analysis summary statistics may be accessed at ftp://ftp.biostat.wisc.edu/pub/lu_group/Projects/MWAS/. The Modifiable Risk Factors Data were sourced from the MRC Integrative Epidemiology Unit (IEU) and are available via the following link: https://gwas.mrcieu.ac.uk/. PheWAS phenotype data were obtained from six major databases, which can be accessed via the following links: MRC-IEU (https://gwas.mrcieu.ac.uk/phewas/), FinnGen r10 (https://www.finngen.fi/en/access_results), UKB-SAIGE (https://pheweb.org/UKB-SAIGE/), UKB-TOPMed (https://pheweb.org/UKB-TOPMed/), Neale Lab (http://www.nealelab.is/uk-biobank), and Neale Lab (Pan UKB) (https://pan.ukbb.broadinstitute.org/downloads).

Abbreviations

GWAS:

Genome-wide association studies

TWAS:

Transcriptome-wide association studies

PWAS:

Proteome-wide association studies

SNP:

Single nucleotide polymorphism

CGRP:

Calcitonin gene-related peptide

QTL:

Quantitative trait loci

MR:

Mendelian randomization

SMR:

Summary-based MR

PheWAS:

Phenome-wide association studies

HEIDI:

Heterogeneity in Dependent Instruments

GERA:

Genetic Epidemiology Research on Adult Health and Aging

UKB:

UK Biobank

ICD-10:

International Classification of Diseases, Tenth Revision

CLSA:

Canadian Longitudinal Study on Aging

UPLC-MS/MS:

Ultra-Performance Liquid Chromatography-Tandem Mass Spectrometry

WADRC:

Wisconsin Alzheimer’s Disease Research Center

WRAP:

Wisconsin Registry for Alzheimer’s Prevention

CSF:

Cerebrospinal fluid

MRC-IEU:

Medical Research Council Integrative Epidemiology Unit

LDL:

Low-density lipoprotein

HDL:

High-density lipoprotein

SAIGE:

Scalable and Accurate Implementation of Generalized mixed model

TOPMed:

Trans-Omics for Precision Medicine

FDR:

False Discovery Rate

OR:

Odds ratio

PP:

Posterior probabilities

IVW:

Inverse Variance Weighte

LD:

Linkage disequilibrium

3-CMPFP:

3-Carboxy-4-methyl-5-pentyl-2-furanpropanoate

IRAG1:

IP3 Receptor-Associated cGMP Kinase Substrate

References

  1. Safiri S, Pourfathi H, Eagan A, Mansournia MA, Khodayari MT, Sullman MJM et al (2022) Global, regional, and national burden of migraine in 204 countries and territories, 1990 to 2019. Pain 163(2):e293–e309

    Article  CAS  PubMed  Google Scholar 

  2. Burch RC, Buse DC, Lipton RB (2019) Migraine: Epidemiology, Burden, and Comorbidity. Neurol Clin 37(4):631–649

    Article  PubMed  Google Scholar 

  3. Ashina M, Terwindt GM, Al-Karagholi MA, de Boer I, Lee MJ, Hay DL et al (2021) Migraine: disease characterisation, biomarkers, and precision medicine. Lancet 397(10283):1496–1504

    Article  CAS  PubMed  Google Scholar 

  4. Ashina M, Katsarava Z, Do TP, Buse DC, Pozo-Rosich P, Özge A et al (2021) Migraine: epidemiology and systems of care. Lancet 397(10283):1485–1495

    Article  PubMed  Google Scholar 

  5. Hautakangas H, Winsvold BS, Ruotsalainen SE, Bjornsdottir G, Harder AVE, Kogelman LJA et al (2022) Genome-wide analysis of 102,084 migraine cases identifies 123 risk loci and subtype-specific risk alleles. Nat Genet 54(2):152–160

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  6. Bjornsdottir G, Chalmer MA, Stefansdottir L, Skuladottir AT, Einarsson G, Andresdottir M et al (2023) Rare variants with large effects provide functional insights into the pathology of migraine subtypes, with and without aura. Nat Genet 55(11):1843–1853

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  7. Fila M, Sobczuk A, Pawlowska E, Blasiak J (2022) Epigenetic Connection of the Calcitonin Gene-Related Peptide and Its Potential in Migraine. Int J Mol Sci 23(11):6151

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  8. Alpuente A, Torres-Ferrus M, Terwindt GM (2023) Preventive CGRP-targeted therapies for chronic migraine with and without medication-overuse headache. Cephalalgia 43(3):03331024221150235

    Article  Google Scholar 

  9. Scuteri D, Corasaniti MT, Tonin P, Nicotera P, Bagetta G (2021) Role of CGRP pathway polymorphisms in migraine: a systematic review and impact on CGRP mAbs migraine therapy. J Headache Pain 22(1):87

    Article  PubMed  PubMed Central  Google Scholar 

  10. Porcu E, Rüeger S, Lepik K, Santoni FA, Reymond A, Kutalik Z (2019) Mendelian randomization integrating GWAS and eQTL data reveals genetic determinants of complex and clinical traits. Nat Commun 10(1):3300

    Article  PubMed  PubMed Central  Google Scholar 

  11. Sun X, Chen B, Qi Y, Wei M, Chen W, Wu X et al (2024) Multi-omics Mendelian randomization integrating GWAS, eQTL and pQTL data revealed GSTM4 as a potential drug target for migraine. J Headache Pain 25(1):117

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  12. Niu PP, Zhang R, Zhang C, Li S, Li YS (2024) Identifying novel proteins for migraine by integrating proteomes from blood and CSF with genome-wide association data. CNS Neurosci Ther 30(6):e14817

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  13. Ferkingstad E, Sulem P, Atlason BA, Sveinbjornsson G, Magnusson MI, Styrmisdottir EL et al (2021) Large-scale integration of the plasma proteome with genetics and disease. Nat Genet 53(12):1712–1721

    Article  CAS  PubMed  Google Scholar 

  14. GTEx Consortium (2020) The GTEx Consortium atlas of genetic regulatory effects across human tissues. Science 369(6509):1318–1330

  15. Qi T, Wu Y, Fang H, Zhang F, Liu S, Zeng J et al (2022) Genetic control of RNA splicing and its distinct role in complex trait variation. Nat Genet 54(9):1355–1363

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  16. Choquet H, Yin J, Jacobson AS, Horton BH, Hoffmann TJ, Jorgenson E et al (2021) New and sex-specific migraine susceptibility loci identified from a multiethnic genome-wide meta-analysis. Commun Biol 4(1):864

    Article  PubMed  PubMed Central  Google Scholar 

  17. Kurki MI, Karjalainen J, Palta P, Sipilä TP, Kristiansson K, Donner KM et al (2023) FinnGen provides genetic insights from a well-phenotyped isolated population. Nature 613(7944):508–518

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  18. Chen Y, Lu T, Pettersson-Kymmer U, Stewart ID, Butler-Laporte G, Nakanishi T et al (2023) Genomic atlas of the plasma metabolome prioritizes metabolites implicated in human diseases. Nat Genet 55(1):44–53

    Article  PubMed  PubMed Central  Google Scholar 

  19. Panyard DJ, Kim KM, Darst BF, Deming YK, Zhong X, Wu Y et al (2021) Cerebrospinal fluid metabolomics identifies 19 brain-related phenotype associations. Commun Biol 4(1):63

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  20. Su WM, Gu XJ, Dou M, Duan QQ, Jiang Z, Yin KF et al (2023) Systematic druggable genome-wide Mendelian randomisation identifies therapeutic targets for Alzheimer’s disease. J Neurol Neurosurg Psychiatry 94(11):954–961

    Article  PubMed  Google Scholar 

  21. Taliun D, Harris DN, Kessler MD, Carlson J, Szpiech ZA, Torres R et al (2021) Sequencing of 53,831 diverse genomes from the NHLBI TOPMed Program. Nature 590(7845):290–299

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  22. Karczewski KJ, Gupta R, Kanai M, Lu W, Tsuo K, Wang Y, et al (2024) Pan-UK Biobank GWAS improves discovery, analysis of genetic architecture, and resolution into ancestry-enriched effects. medRxiv 2024.03.13.24303864

  23. Sanderson E, Glymour MM, Holmes MV, Kang H, Morrison J, Munafò MR et al (2022) Mendelian randomization. Nat Rev Methods Primers 2:Article number 92

  24. Wu Y, Zeng J, Zhang F, Zhu Z, Qi T, Zheng Z et al (2018) Integrative analysis of omics summary data reveals putative mechanisms underlying complex traits. Nat Commun 9(1):918

    Article  PubMed  PubMed Central  Google Scholar 

  25. Benjamini Y, Hochberg Y (2018) Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. J Roy Stat Soc: Ser B (Methodol) 57(1):289–300

    Article  Google Scholar 

  26. Giambartolomei C, Vukcevic D, Schadt EE, Franke L, Hingorani AD, Wallace C et al (2014) Bayesian test for colocalisation between pairs of genetic association studies using summary statistics. PLoS Genet 10(5):e1004383

    Article  PubMed  PubMed Central  Google Scholar 

  27. Sun J, Zhao J, Jiang F, Wang L, Xiao Q, Han F et al (2023) Identification of novel protein biomarkers and drug targets for colorectal cancer by integrating human plasma proteome with genome. Genome Med 15(1):75

    Article  PubMed  PubMed Central  Google Scholar 

  28. Hemani G, Zheng J, Elsworth B, Wade KH, Haberland V, Baird D et al (2018) The MR-Base platform supports systematic causal inference across the human phenome. Elife 7:e34408

    Article  PubMed  PubMed Central  Google Scholar 

  29. Bowden J, Davey Smith G, Burgess S (2015) Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression. Int J Epidemiol 44(2):512–525

    Article  PubMed  PubMed Central  Google Scholar 

  30. Burgess S, Small DS, Thompson SG (2017) A review of instrumental variable estimators for Mendelian randomization. Stat Methods Med Res 26(5):2333–2355

    Article  PubMed  Google Scholar 

  31. VanderWeele TJ (2016) Mediation Analysis: A Practitioner’s Guide. Annu Rev Public Health 37:17–32

    Article  PubMed  Google Scholar 

  32. Millard LA, Davies NM, Timpson NJ, Tilling K, Flach PA, Davey SG (2015) MR-PheWAS: hypothesis prioritization among potential causal effects of body mass index on many outcomes, using Mendelian randomization. Sci Rep 5:16645

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  33. Meyers TJ, Yin J, Herrera VA, Pressman AR, Hoffmann TJ, Schaefer C et al (2023) Transcriptome-wide association study identifies novel candidate susceptibility genes for migraine. HGG Adv 4(3):100211

    CAS  PubMed  PubMed Central  Google Scholar 

  34. Kalasova I, Hailstone R, Bublitz J, Bogantes J, Hofmann W, Leal A et al (2020) Pathological mutations in PNKP trigger defects in DNA single-strand break repair but not DNA double-strand break repair. Nucleic Acids Res 48(12):6672–6684

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  35. Fila M, Jablkowska A, Pawlowska E, Blasiak J (2023) DNA Damage and Repair in Migraine: Oxidative Stress and Beyond. Neuroscientist 29(3):277–286

    Article  CAS  PubMed  Google Scholar 

  36. Kate WD, Fanta M, Weinfeld M (2024) Loss of the DNA repair protein, polynucleotide kinase/phosphatase, activates the type 1 interferon response independent of ionizing radiation. Nucleic Acids Res 52(16):9630–9653

    Article  PubMed  PubMed Central  Google Scholar 

  37. Noseda R, Burstein R (2013) Migraine pathophysiology: anatomy of the trigeminovascular pathway and associated neurological symptoms, cortical spreading depression, sensitization, and modulation of pain. Pain 154(Suppl 1):S44-53

    Article  CAS  PubMed  Google Scholar 

  38. Ou Y, Ni X, Gao X, Yu Y, Zhang Y, Wang Y et al (2024) Structural and functional changes of anterior cingulate cortex subregions in migraine without aura: relationships with pain sensation and pain emotion. Cereb Cortex 34(2):bhae040

    Article  PubMed  PubMed Central  Google Scholar 

  39. Karsan N, Silva E, Goadsby PJ (2023) Evaluating migraine with typical aura with neuroimaging. Front Hum Neurosci 17:1112790

    Article  PubMed  PubMed Central  Google Scholar 

  40. Kim DJ, Jassar H, Lim M, Nascimento TD, DaSilva AF (2021) Dopaminergic Regulation of Reward System Connectivity Underpins Pain and Emotional Suffering in Migraine. J Pain Res 14:631–643

    Article  PubMed  PubMed Central  Google Scholar 

  41. Lai KL, Niddam DM, Fuh JL, Chen WT, Wu JC, Wang SJ (2020) Cortical morphological changes in chronic migraine in a Taiwanese cohort: Surface- and voxel-based analyses. Cephalalgia 40(6):575–585

    Article  PubMed  Google Scholar 

  42. Gruenbaum BF, Kutz R, Zlotnik A, Boyko M (2020) Blood glutamate scavenging as a novel glutamate-based therapeutic approach for post-stroke depression. Ther Adv Psychopharmacol 10:2045125320903951

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  43. Frank D, Kuts R, Tsenter P, Gruenbaum BF, Grinshpun Y, Zvenigorodsky V et al (2019) The effect of pyruvate on the development and progression of post-stroke depression: A new therapeutic approach. Neuropharmacology 155:173–184

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  44. Prince N, Begum S, Mínguez-Alarcón L, Génard-Walton M, Huang M, Soeteman DI et al (2023) Plasma concentrations of per- and polyfluoroalkyl substances are associated with perturbations in lipid and amino acid metabolism. Chemosphere 324:138228

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  45. Gusdon AM, Savarraj JPJ, Feng D, Starkman A, Li G, Bodanapally U et al (2024) Identification of metabolites associated with preserved muscle volume after aneurysmal subarachnoid hemorrhage due to high protein supplementation and neuromuscular electrical stimulation. Sci Rep 14(1):15071

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  46. Oliveira R, Gil-Gouveia R, Puledda F (2024) CGRP-targeted medication in chronic migraine - systematic review. J Headache Pain 25(1):51

    Article  PubMed  PubMed Central  Google Scholar 

  47. Li SJ, Shi JJ, Mao CY, Zhang C, Xu YF, Fan Y et al (2023) Identifying causal genes for migraine by integrating the proteome and transcriptome. J Headache Pain 24(1):111

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  48. Daghlas I, Sargurupremraj M, Danning R, Gormley P, Malik R, Amouyel P et al (2022) Migraine, Stroke, and Cervical Arterial Dissection: Shared Genetics for a Triad of Brain Disorders With Vascular Involvement. Neurol Genet 8(1):e653

    Article  PubMed  PubMed Central  Google Scholar 

  49. Prüschenk S, Majer M, Schlossmann J (2023) Novel Functional Features of cGMP Substrate Proteins IRAG1 and IRAG2. Int J Mol Sci. 24(12):9837

    Article  PubMed  PubMed Central  Google Scholar 

  50. Ali S, Solano AS, Gonzales AL, Thakore P, Krishnan V, Yamasaki E et al (2021) Nitric Oxide Signals Through IRAG to Inhibit TRPM4 Channels and Dilate Cerebral Arteries. Function (Oxf) 2(6):zqab051

    Article  PubMed  Google Scholar 

  51. Wattiez AS, Sowers LP, Russo AF (2020) Calcitonin gene-related peptide (CGRP): role in migraine pathophysiology and therapeutic targeting. Expert Opin Ther Targets 24(2):91–100

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  52. D’Mello AM, Gabrieli JDE, Nee DE (2020) Evidence for Hierarchical Cognitive Control in the Human Cerebellum. Curr Biol 30(10):1881–92.e3

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  53. Mills SJ, Silvander C, Cozier G, Trésaugues L, Nordlund P, Potter BV (2016) Crystal Structures of Type-II Inositol Polyphosphate 5-Phosphatase INPP5B with Synthetic Inositol Polyphosphate Surrogates Reveal New Mechanistic Insights for the Inositol 5-Phosphatase Family. Biochemistry 55(9):1384–1397

    Article  CAS  PubMed  Google Scholar 

Download references

Acknowledgements

We appreciate all the participants and institutions from the Genetic Epidemiology Research on Adult Health and Aging (GERA) cohort, UK Biobank, FinnGen cohort, deCODE Genetics, the GTEx consortium, the Yang Lab at Westlake University, the Canadian Longitudinal Study on Aging (CLSA), the Wisconsin Alzheimer’s Disease Research Center (WADRC), the Wisconsin Registry for Alzheimer’s Prevention (WRAP), the MRC Integrative Epidemiology Unit (IEU), UKB-SAIGE, UKB-TOPMed, Neale Lab, and the PheWeb platform for providing the data.

Funding

This research was funded by the General Health Project of Zhejiang Province (2024KY130) and the Research Project of Zhejiang Chinese Medical University (2022FSYYZZ01).

Author information

Authors and Affiliations

  1. Department of Radiology, The First Affiliated Hospital of Zhejiang Chinese Medical University (Zhejiang Provincial Hospital of Chinese Medicine), Hangzhou, China

    Jiafei Lou, Miaoqian Tu, Maosheng Xu, Zhijian Cao & Wenwen Song

  2. The First School of Clinical Medicine, Zhejiang Chinese Medical University, Hangzhou, China

    Jiafei Lou, Miaoqian Tu, Maosheng Xu, Zhijian Cao & Wenwen Song

Authors
  1. Jiafei Lou
    View author publications

    You can also search for this author inPubMed Google Scholar

  2. Miaoqian Tu
    View author publications

    You can also search for this author inPubMed Google Scholar

  3. Maosheng Xu
    View author publications

    You can also search for this author inPubMed Google Scholar

  4. Zhijian Cao
    View author publications

    You can also search for this author inPubMed Google Scholar

  5. Wenwen Song
    View author publications

    You can also search for this author inPubMed Google Scholar

Contributions

ZJC and WWS proposed the idea. JFL and MQT participated in the study design. JFL collected and organized the summary data, performed the statistical analysis, interpreted the data, and drafted the initial manuscript. JFL, WWS, and MSX revised and reviewed the manuscript. All authors read and approved the final submitted version.

Corresponding authors

Correspondence to Zhijian Cao or Wenwen Song.

Ethics declarations

Ethics approval and consent to participate

This MR study utilized de-identified, summary-level data from both publicly available GWAS, large-scale pQTL, and eQTL databases. All necessary informed consents and ethical approvals were obtained in the original studies from which these data were sourced. Therefore, this study did not require additional ethical approval.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Lou, J., Tu, M., Xu, M. et al. Plasma pQTL and brain eQTL integration identifies PNKP as a therapeutic target and reveals mechanistic insights into migraine pathophysiology. J Headache Pain 25, 202 (2024). https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s10194-024-01922-z

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s10194-024-01922-z

Keywords

The Journal of Headache and Pain

ISSN: 1129-2377

Contact us