The genetic structure of pain in depression patients: A genome-wide association study and proteome-wide association study.

BACKGROUND
Comparing with the general population, the pain in depression patients has more complex biological mechanism. We aim to explore the etiological mechanism of pain in depression patients from the perspective of genetics.


METHODS
Utilizing the UK Biobank samples with self-reported depression status or PHQ score ≥10, we conducted genome-wide association studies (GWAS) of seven pain traits (N = 1,133-58,349). Additionally, we used FUSION pipeline to perform proteome-wide association study (PWAS) and transcriptome-wide association study (TWAS) by integrating GWAS summary data with two different proteome reference weights (ROS/MAP and Banner) and Rnaseq gene expression reference weights, respectively.


RESULTS
GWAS identified 3 significant genes associated with different pain traits in depression patients, including TRIOBP (PGWAS = 4.48 × 10-8) for stomach or abdominal pain, SLC9A9(PGWAS = 2.77 × 10-8) for multisite chronic pain (MCP) and ADGRF1 (PGWAS = 1.51 × 10-8) for neck or shoulder pain. In addition, PWAS and TWAS analysis also identified multiple candidate genes associated with different pain traits in depression patients, such as TPRG1L (PPWAS-Banner = 3.38 × 10-2) and SIRPA (PPWAS-Banner = 3.65 × 10-2) for MCP, etc. Notably, when comparing the results of PWAS and TWAS analysis, we found overlapping candidate genes in these pain traits, such as GSTM3 (PPWAS-Banner = 3.38 × 10-2, PTWAS = 6.92 × 10-3) in the stomach or abdominal pain phenotype, ATG7 (PPWAS-Rosmap = 3.15 × 10-2, PTWAS = 2.98 × 10-2) in the MCP, etc. CONCLUSIONS: We identified multiple novel candidate genes for pain traits in depression patients from different perspectives of genetics, which provided novel clues for understanding the genetic mechanisms underlying the pain in depression patients.


Introduction
As one of the most common psychiatric disorders, depression not only causes changes in emotional state (affective disorder), but also usually causes major physical symptoms such as primarily fatigue, pain or sleep disturbances. 1 According to the World Health Organization's latest report, the prevalence of depression is steadily increasing worldwide, with an estimated 322 million people (4.4% of the world's population) living with depression. 2 Furthermore, depression is the single largest contributor to nonfatal health loss globally, accounting for 7.5% of all years of disability in 2015. 2 As a complex disease, depression not only reduces the health-related quality of life of patients, and increase the economic burden of individuals and social medical economic burden, which has become a public health problem of global concern. 3,4 Pain is an unpleasant sensory and emotional experience associated with actual or potential tissue damage, and is an interaction of psychological, emotional, behavioral and social factors. 5 Although pain is not a diagnostic symptom of depression, pain symptoms are common in people with depression and are closely related to depression. Previous studies have found that the prevalence of pain in depression patients was 59.1% (95% CI: 57.7%-60.7%) 6 , and pain occurred more frequently in patients with depression than in those without. 7 A longitudinal study that excluded participants with a history of depression or anxiety showed that baseline pain symptoms were a major risk factor for depression. 8 What's more, previous studies have shown that pain can adversely affect the prognosis and treatment of depression. For example, a systematic review indicates that patients with both pain and depression had reduced physical, mental and social functioning as opposed to those with depression or pain alone. 9 Previous research on between depression and pain symptoms has shown that because of the complex interaction between depression and pain symptoms, multiple common neurobiological mechanisms, genetics, environmental and cognitive factors must be considered when trying to unravel their association. 10 One widely accepted theory holds that depression and painful symptoms follow the same descending pathways of the central nervous system.
That is, in the nerves of the downward pain regulation system, key structures of the brainstem pain regulation circuit (periaqueductal gray(PAG) and rostralventromedial medulla(RVM)) receives many projections from the brain regions involved in the production of emotions (medial prefrontal cortex, insular It is worth noting that several previous studies have emphasized individual differences in susceptibility to depression and pain responses, indicating that genetic factors may have some common roles in the pathogenesis of depression and pain. 12,13 Multivariate models of a twin study estimated 39% (95% CI: 22%-53%) heritability for depression and 25% (95% CI: 9%-41%) heritability for pain. 14 Meng et al. conducted the cross-trait linkage disequilibrium score regression (LD) analysis for different pain phenotypes, and identi ed signi cant and positive genetic associations of multiple pain phenotypes with   depression and neuroticism, revealing that pain and depression have partially overlapping genetic risk factors. 15 Moreover, a genome-wide association study   (GWAS) of multisite chronic pain based on data from the UK Biobank (UKB) cohort identi ed 39 susceptibility genes signi cantly associated with chronic pain, including DCC, MAML3 and FOXP2, with functions implicated neuronal development, neurogenesis and neural connectivity. 16 Summary of previous studies on depression and pain symptoms have found that due to the complex interactions and common genetic risk factors between them, there may be differences in the biological mechanism of pain symptoms between depression patients and the general population. However, few studies have conducted GWAS of pain in depression patients, and the corresponding susceptibility genes are unknown.
Utilizing the UK Biobank cohort, we conducted GWAS of different pain traits in depression patients. Then, the GWAS summary data were further subjected to PWAS and LDSC analysis in order to explore the potential biological mechanism related to pain in depression patients from the perspective of genetics. Finally, gene enrichment analysis was conducted to further reveal the possible biological pathway of genes related to pain phenotype in patients with depression. It is hoped to provide a new direction for the etiological mechanism study of depression and pain symptoms.

UK Biobank Cohort
The UK Biobank is a large prospective cohort study that collected health-related information, including phenotypes and genotypes, from approximately 500,000 participants aged between 40 and 69 years from around the United Kingdom in 2006 and 2010. 17 All participants were asked to report a range of demographic information and health status through questionnaires and interviews, and approved to use their anonymous data for any health-related research.
The UK Biobank provides informed consent to all participants. Our study was approved by the UK Biobank (Application 46478) and health-related records of participants were obtained. 17 Depression Patients from UK Biobank According to the de nition of depression phenotype in the previous study, 18 we selected the depression patients from the participants of UK Biobank cohort.
Brie y, the depression patients were determined based on self-report depression status (UK Biobank data elds: 20002, 20126, and 20544) or Patient Health Question-9 (PHQ) score ≥ 10. For self-reported depression, depression patients were selected based on the code 1286 from ID 20002, code 3,4 or 5 from ID 20126 and code 11 from ID 20544. In order to classify depression patients as closely as possible, we used the Patient Health Question-9 (PHQ) 19 as strict inclusion and exclusion criteria. PHQ is a classi cation scale with an overall score (0-27) that focuses on nine symptoms and signs of depression and is used to screen and measure the severity of depression. Detailed classi cation of depression is presented in the Supplementary Information. Our analysis was limited to white British individuals (UK biobank data domain :21000).

Pain Phenotypes of UK Biobank
According to the speci c pain related questionnaire adapted by the UK Biobank, the pain phenotype was measured by a question: "whether the participant reported having suffered from various types of body pain in the previous 3 months? The options were: 1. headache, 2. facial pain, 3. neck or shoulder pain, 4. back pain, 5. stomach or abdominal pain, 6. hip pain, 7. knee pain, 8. pain all over the body, 9. none of the above. More than one option could be selected (UK Biobank Category ID 100048)." Eight different pain phenotypes were de ned based on pain at different sites reported by participants for the past 3 months or more, and complete de nitions of the eight pain phenotypes can be found in the supplementary information. For each pain phenotype, cases were de ned as people who chose the option of having pain at a speci c site for 3 months or more in response to the above question. The controls were those who chose "none of the above" or "pain in a speci c site did not last 3 months." Due to the small sample size for general pain for 3+ months (Field ID:2956) and facial pains for 3+ months (Field ID:4067), our study excluded these two pain phenotypes. In addition, we de ned multisite chronic pain (MCP) based on the UK Biobank pain phenotype of seven different body sites with a pain score questionnaire (score 0-7). Once participants chose the most recent pain in one or more body sites or all over the body, they were also asked if the pain had been ongoing for three months or more (de ned as chronic pain). Thus, MCP is de ned as the total number of body sites that record chronic pain (lasting at least 3 months) :0 to 7. Those who chose "pain over the body" were excluded from our study because of the difference between widespread pain and localized chronic pain. 16 Genotyping, Imputation and Quality Control of UK Biobank In UK Biobank cohort, a total of 488,377 participants were genotyped by either the Affymetrix UK BiLEVE Axiom array or the UK Biobank Axiom array. 20 Subsequently, these genotypes results were imputed and quality controlled based on the Haplotype Reference Consortium (HRC) reference panel and UK10K project Reference panels. 21 The details of array design, genotyping and quality control procedures are described elsewhere. 22 Then, we deleted those participants who reported discrepancies between genetic gender and self-reported gender, as well as without ethical consent and interpolation data.

Linkage Disequilibrium Score Regression (LDSC) Analysis
According to the LDSC software(v1.0.0,https://github.com/bulik/ldsc), we used the cross-trait LDSC method to calculate the genetic correlation between different pain phenotypes and seven mental disorders 24 , including Alzheimer's disease(AD) 25 , anxiety 26 , attention-de cit hyperactivity disorder(ADHD) 27 , autism spectrum disorder(ASD) 28 , bipolar disorder(BD) 29 , major depressive disorder(MDD) 30 and post-traumatic stress disorder(PTSD) 31 . The GWAS data of these seven psychiatric disorders were obtained from the published articles by the Psychiatric Genomics Consortium (PGC) (https://www.med.unc.edu/pgc/download-results/). The LDSC analysis is not affected by sample overlap and is fast, making it a reliable and powerful method to calculate genetic correlations between different phenotypes using GWAS summary-level results data. 24,32 Proteome-Wide Association Study (PWAS) We separately integrated two different reference human brain proteomes(the ROS/MAP brain proteome reference weights 33 and Banner brain proteome reference weights 34 ) with our GWAS summary of pain in depression patients for PWAS using FUSION pipeline, respectively.Speci cally, FUSION pipeline was used to calculate PWAS association statistics (PWAS Z-score), and then combined with the pre-calculated brain proteome reference weight (Z-score × proteome weight) to evaluate the effects of signi cant SNP in the pain GWAS of depression patients on the protein abundance. Finally, FUSION identi es the candidate genes associated with pain in patients with depression that act by regulating the abundance of proteins in the brain. To control the potential effect of multiple testing on study results, the permutation-based P-value (named PERM.PV) was calculated for each gene by FUSION. The permutation-based Pvalue (P PWAS ) <0.05 was used as signi cance threshold in our PWAS analysis.
Two different reference human brain proteomics used in our study were from Religious Orders Study and Rush Memory 33 and Aging Project (ROS/MAP) and the Banner Sun Health Research Institute (Banner) 34 . According to previous study of human brain proteome 35 , proteome was sequenced and combined with signi cant SNPS in GWAS database to identify cis-regulated proteins that were signi cantly related to genetic variation. Detailed information on proteome sequencing, quality control, and standardization can be obtained in the study by Wingo et al. 35 After analysis and quality control, the ROS/MAP proteome (N=376) identi ed a total of 1,475 proteins that showed a signi cant cis-association with genetic variation. The Banner proteome (N=152) identi ed 1,139 proteins that showed a signi cant cis-association with genetic variation.

Functional exploration of identi ed candidate genes
We used Metascape's online analytical tool (http://metascape.org) to explore the functions of genes identi ed by our GWAS and PWAS, including Gene Ontology (GO), pathway analysis. GO was used for gene function classi cation and annotation, Pathway Analysis was used to analyze a speci c classical metabolic or regulatory network of genes, proteins or molecules. All genes in the genome are used as the enrichment background. The terms with P-value<0.05, minimum count of 3, and enrichment factor>1.5 were collected, then group them into clusters based on the similarity of their members.

General population characteristics
In our study, we analyzed seven different pain phenotypes in two depression samples, de ned by self-reported depression status and PHQ score ≥ 10. Table 1 details the general characteristics of study samples. Abbreviations: SD, age was described as mean ± standard deviation (SD), PHQ score, patient health questionnaire (PHQ) is used to describe the depression.

Functional exploration results
After pathways and processes enrichment analysis of the signi cant genes identi ed by GWAS and PWAS, a total of 56 GO terms were obtained. These GO terms are mainly involved in sulfur compound metabolic process, regulation of actin lament organization, tRNA metabolic process, signal release, chemical synaptic transmission et al. The biological processes or functional pathways of the 16 clusters detected in the enrichment analysis were summarized by the Bubble plot, and the detailed results were shown in Figure 4.

Discussions
Previous studies on pain symptoms have suggested that pain mainly involves peripheral, spinal and central effects, including various factors that affect nociceptive perception, in ammatory response, and the neural pathways of pain centers and pain signals in the brain. 36, 37 In addition, a large number of studies have found that there are complex interactions and common genetic risk factors between depression and pain symptoms. 12,38 Such as, some studies have found that depression can affect changes in multiple brain regions, most commonly in areas such as prefrontal cortex(PFC), anterior cingulate cortex (ACC), hippocampus and amygdala, where gray matter volume loss and alterations in activity similar to those that occur in chronic pain. 38, 39 Neuroimaging studies have also revealed the close relationship between brain regions (somatosensory, IC ACC, PFC and thalamus) involved in the integration of pain sensation and emotions and the regions regulated by depression. 40 These common in uencing factors may cause differences in pain susceptibility genes between depression patients and the general population. However, to our knowledge, studies exploring the genetic mechanisms of pain in patients with depression are limited, and the corresponding susceptibility genes remain unclear. Therefore, we conducted GWAS studies on different pain phenotypes in patients with depression, and identi ed three candidate genes associated with different pain phenotypes. These candidate genes not only have multiple functions affecting psychiatric disorders, nervous system development and neurogenesis, but it is noteworthy that some genes are also involved in physiological processes affecting pain symptoms, such as ADGRF1 identi ed for neck or shoulder pain, SLC9A9 gene for MCP and TRIOBP for stomach or abdominal pain.
ADGRF1 (adhesion G protein-coupled receptors) is a group of speci c target receptors for synaptic amides that are highly expressed in neural stem cells and the brain during development. 41 Its ligand, N-Docosahexaenoylamine (Synaptamide), is an endocannabinoid metabolite. This signaling pathway can not only by in uencing the neurogenic and synaptic genes expression, induction of neurogenesis of hippocampus and cortex neurons and synapses, and the neuronal differentiation of neural stem cells, can also ring through cyclic adenosine monophosphate (cAMP) to transmit anti-in ammatory signals, inhibiting the expression of pro-in ammatory genes and pro-in ammatory cytokines. 41,42 Because of these biological functions, synaptamide play a role in improving cognitive function and neuronal plasticity in neuropathic pain. 43 What's more, Park et al. demonstrated the immunomodulatory effects of ADGRF1 in the brain and in the periphery, and jointly promoted the anti-neuroin ammatory effects of synaptamide in systemic in ammatory conditions. These results suggest that ADGRF1 mediated inhibition of innate immune cell activation may be a novel therapeutic strategy for controlling brain and/or peripheral in ammation and related diseases. 44 SLC9A9(solute carrier family 9, member A9) mainly encodes sodium/proton exchanger 9(NHE9), It is expressed in many brain tissues and modulates luminal pH of circulating endosomes, which are important organelles for synaptic transmission and plasticity. 45 And SLC9A9 pathway is associated with long-term potentiation, which underlies cognitive functions that are frequently disrupted in learning and memory, ADHD, and ASD. 46 Previous studies have found that changes in SLC9A9 gene expression and protein function are associated with a variety of human diseases, such as ADHD, ASD et al. 47 In addition, a proteinprotein interaction network pathway analysis showed that SLC9A9 is also involved in oxidative stress, nociception and other functions. 46 Since there are multiple overlapping factors between the functions involved in ADGRF1 and SLC9A9 and the in uencing factors of pain symptoms, we speculate that two genes are associated with the pain in patients with depression by in uencing the factors of pain symptoms. Further studies are still needed to verify our speculation.
TRIOBP gene primarily encodes TRIO and F-actin binding proteins, which are involved in neural tissue development and control actin cytoskeleton organization, cell motility and cell growth. 48 Among them, several studies have implicated that TRIOBP-1 is involved in schizophrenia by forming protein aggregates in the brain. 49 At present, studies of TRIIOBP gene are limited, and further studies are needed to con rm its possible etiological association with pain symptoms.
Among the candidate genes associated with different pain phenotypes identi ed by PWAS, several candidate genes, such as P2RX7 and SIRPA, are associated with biological mechanisms that in uence pain symptoms or participate in the pain process. P2RX7 (purinergic receptor P2X 7) is a non-selective cationic channel activated by extracellular ATP. It is mainly expressed in the peripheral and central nervous system and immune system. Activation of P2RX7 not only contributes to pro-in ammatory response to injury or bacterial invasion and mediates apoptosis, but also plays a role in chronic in ammation and neuropathic pain. 50 Such as, an animal study revealed that P2RX7 may promote pain modulation through its effect on peripheral tissue damage and altered central nervous system processing in clinical pain states. 51 Kambur's study reinforces this evidence and suggests that P2RX7 gene and genetic variants may be involved in regulating human pain sensitivity. 50 In addition, previous studies have found that the P2RX7 variant is associated with a higher risk of psychiatric disorders such as bipolar disorder and depression, and that the P2RX7 receptor is involved in psychosis related pathways, such as synaptic plasticity, neurotransmission, and immune regulation. 52 It is worth noting that SIRPA gene was overlapping genes in two PWAS studies with different reference weights. SIRPA gene mainly encodes proteins that are members of the signal-regulatory-protein (SIRP) family, which are involved in signal transduction mediated by various growth factor receptors. Previous studies have found that cleavage of signal-regulatory-protein α (SIRPα) is related to enhanced in ammatory signaling, 53 CD47 is a demonstrated ligand for this receptor protein, and its signaling pathway, CD47-SIRPα, is involved in the regulation of immune homeostasis and neuronal networks. 54 What's more, Haiyue Zhang et al revealed the importance of CD47 and SIRPα in the neuroin ammatory process of central nervous system diseases. 55 Overall, P2RX7 has been shown to be directly involved in the regulation of chronic pain, while SIRPA is mainly involved in the biological process of in uencing factors of pain symptoms, so it can be speculated that they may in uence pain symptoms by regulating and participating in in ammatory responses. Further studies are needed to con rm our speculation. In addition, among the numerous candidate genes identi ed in the PWAS study, many genes are also associated with psychiatric disorders, nervous system development, and neuronal conduction. For example, PPM1F 56 and GSTM3 57 et al.
In the calculation results of genetic correlation with different psychiatric disorders, we found that ADHD, and PTSD were signi cantly positively correlated with different pain phenotypes, re ecting the possible common genetic structure. It is particularly noteworthy that the genetic positive correlation between ADHD and MCP phenotype was consistent with the signi cant association gene (SLC9A9) identi ed in the GWAS of MCP. SLC9A9 gene pathway is associated with long-term potentiation, which is the basis of cognitive functions that are frequently disrupted in ADHD. 46 This further indicates that ADHD and MCP phenotype share the common genetic structure, and also re ects pleiotropy, in which speci c genetic alleles may increase the risk of both phenotypes. 58 In addition, enrichment analysis revealed that signi cant genes were associated with multiple biological processes and functional pathways. Some of these related biological processes and functional pathways can directly or indirectly affect pain symptoms. For example, sulfur compound metabolism process, previous studies have found that sulfur compounds (such as hydrogen sulphide) can relieve pain symptoms and in ammation 59 , so signi cant genes may regulate the pain of patients with depression by participating in sulfur compound metabolism process, What's more, some studies have also found that signal release and chemical synaptic transmission are involved in the biological process of pain symptoms 60 , and the biological relationship between other pathways and pain symptoms needs to be further studied.
Overall, one advantage of our study is the use of a large sample size from the UK Biobank, which can eliminate statistical noise by overcoming potential confounding factors such as selection biases and heterogeneity. In addition, compared with previous pain GWAS studies, our research focused more on multiple pain phenotypes in patients with depression. Based on GWAS, PWAS study and LDSC analysis were further used to explore the potential biological mechanisms related to pain in patients with depression from the perspective of protein expression level and genetics.
Nevertheless, it is worth noting that our research has some limitations. First, all data in our research were derived from the UK Biobank, and the research participants were limited to depression patients of European descent. When applying the research results to different populations, the impact of genetic backgrounds differences on the results should be considered, Secondly, because the phenotyping of pain in the UK Biobank was based on a speci c nonstandard pain-related questionnaire, it may result in pain phenotypes being broadly de ned and not ltered by other potentially relevant information about the nature, duration or intensity of pain. Finally, genes signi cantly associated with pain symptoms were identi ed in our study, and while previous studies have suggested that they play a role in neurological function and disease, there is no direct evidence that they are involved in pain symptoms, and further research is needed to con rm this.
In summary, through GWAS and PWAS analysis, we found several susceptibility genes signi cantly associated with pain phenotype in patients with depression, What's more, LDSC analysis found that there was a signi cant genetic correlation between different pain phenotypes and ADHD and PTSD, suggesting that there may be a common genetic structure. These ndings contribute to the understanding of the biological mechanism of pain phenotype in patients with depression and provide new clues for the study of its pathogenesis. Figure 1 Manhattan plot of different pain phenotypes in patients with depression. (a) Multisite chronic pain, a SNP allele was found to be signi cantly associated with MCP (b) Stomach or abdominal pain phenotype, four independent SNP alleles were signi cantly associated with stomach pain or abdominal pain phenotype.

Declarations
(c) Neck or shoulder pain phenotype, three independent SNP alleles were signi cantly associated with Neck or shoulder pain phenotype. Biobank Genomic coordinates are displayed along the X-axis, with the negative logarithm of the association P value for each SNP displayed on the Y-axis, meaning that each dot on the Manhattan plot signi es a SNP. The red line indicates the P-value threshold for genome-wide signi cance (P GWAS < 5 × 10 −8 ) while the blue line indicates P-value threshold for suggestive signi cance (P GWAS < 1 × 10 −5 ).

Figure 2
PWAS results of Banner reference weights for different pain phenotypes in patients with depression. (a) Multisite chronic pain, 17 candidate genes were found to be signi cantly associated with MCP. (b) Stomach or abdominal pain phenotype, 16 candidate genes were signi cantly associated with stomach pain or abdominal pain phenotype. (c) Neck or shoulder pain phenotype, 8 candidate genes were signi cantly associated with Neck or shoulder pain phenotype. The results of the PWAS study are shown through the Manhattan plot. These candidate genes may contribute to the biological mechanisms of pain in depressed patients through their cis-regulated brain protein abundance. The red line indicates the P-value threshold for Proteome-wide signi cance (P < 0.05).

Figure 3
PWAS results of Rosmap reference weights for different pain phenotypes in patients with depression. (a) Multisite chronic pain, 22 candidate genes were found to be signi cantly associated with MCP. (b) Stomach or abdominal pain phenotype, 15 candidate genes were signi cantly associated with stomach pain or abdominal pain phenotype. (c) Neck or shoulder pain phenotype, 14 candidate genes were signi cantly associated with Neck or shoulder pain phenotype. The results of the PWAS study are shown through the Manhattan plot. These candidate genes may contribute to the biological mechanisms of pain in depressed patients through their cis-regulated brain protein abundance. The red line indicates the P-value threshold for Proteome-wide signi cance (P < 0.05).

Figure 4
Bubble plot of enrichment analysis. This bubble plot showed biological processes or functional pathways of the 16 clusters in the enrichment analysis results.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. TheSupportingInformation.docx