Summary
In this study, we conducted proteomic two-sample MR to identify putatively causal associations between genetically predicted levels of plasma proteins and several mental disorder traits. Performed with the largest and most recent GWAS for the exposure and outcome variables of interest, the MR analyses revealed 188 associations between genetically predicted plasma protein levels and schizophrenia, 86 with bipolar disorder, 32 with opioid use disorder, 30 with major depressive disorder and one with depression treatment resistance. We performed colocalisation analyses to reduce the risk of horizontal pleiotropy on our results. After colocalisation, we found evidence to support 12 of the schizophrenia associations, 11 bipolar associations, six opioid use disorder associations, three associations with major depressive disorder and one association with treatment resistance in depression.
No associations were found between genetically predicted levels of plasma proteins and panic disorder, obsessive compulsive disorder (OCD), treatment resistance in schizophrenia, or the antidepressant response trait rate of improvement; these traits had lower GWAS sample sizes than the traits where we detected associations, and – supported by the fact that we found strong correlations between the number of hits (both MR, and colocalising) and psychiatric outcome instrument strength, as represented in Figure 6 – we speculate that the lack of associations in the current study is likely a reflection of reduced statistical power, rather than any inherent difference in biology. We anticipate associations will be detected in future for these traits as GWAS sample sizes increase.
Building upon existing observational and experimental data, the proteins we identified as associated with mental disorder may provide insights into underlying mechanisms and may help prioritise plasma protein targets for investigation as being potentially therapeutically relevant. Circulating protein levels may themselves be targets for pharmaceutical intervention or could in some cases be mediating factors through which modifiable environmental risk factors operate to alter illness risk39. Although detailed discussion of each protein is not in the scope of this report, below we discuss a selection of proteins highlighted by our study as potentially involved in mental disorder pathogenesis.
A number of findings implicate immune system related plasma protein involvement in the pathogenesis of mental disorders. Our MR and colocalisation analyses found associations that passed our evidential threshold between BTN2A1 (Butyrophilin subfamily 2 member A1) protein and three psychiatric traits: schizophrenia, bipolar disorder and major depressive disorder. We also found MR evidence for association with the related protein BTN3A2 (Butyrophilin subfamily 3 member A2) encoded protein, though this was not corroborated by colocalisation. BTN2A1 had already been found as significantly associated with schizophrenia in a smaller proteome wide association study40. Further, the rs13195401 variant in BTN2A1 was previously shown to be in high LD with known schizophrenia susceptibility SNPs, which are inherited with the illness and assumed to include the schizophrenia vulnerability allele that partially explains schizophrenia risk41.
BTN2A1 and BTN3A2 are located on chromosome 6p21.3 within the human leukocyte antigen (HLA) region. They are transmembrane glycoproteins and are structurally related to MHC class I molecules, which are cell surface recognition elements expressed on somatic cells. These proteins have known roles in lipid, fatty-acid, and sterol metabolism. Butyrophilin-like proteins have also been implicated in modulating T cell responses and have been associated with certain immune-related conditions, although their exact roles in the immune system are still the subject of ongoing research42. Another member of the butyrophilin family, BTN3A3, was found to be associated with schizophrenia in a recent proteomic MR performed by Dang et al43. Given that butyrophilin genes are located in the HLA region, it is important to note that the complex pattern of linkage disequilibrium present in this region can obscure the interpretation of findings. The issue of population stratification in this region can further complicate analyses, and the presence of pleiotropic effects, in which a single genetic variant impacts multiple traits or diseases, may lead to spurious associations. Our method using only lead cis-quantitative trait variants as genetic instruments, and combination of Mendelian randomization and colocalisation mitigates to an extent against linkage disequilibrium association errors, but the results should be viewed in the context of the genetic complexity of this region, and that whilst colocalisation mitigates against short-range linkage disequilibrium association errors, it does not address long-range linkage disequilibrium. Variants in the MHC region have been shown to carry the highest risk of schizophrenia even after accounting for the complexity outlined above22,44, and our results further support the importance of this region, potentially relating to immune function, in the aetiology of mental disorders.
The results for neural cell adhesion molecule NCAM1 in bipolar disorder from the present study may provide additional evidence for the role of immune mediated pathology in mental disorder, as it is a member of the immunoglobulin super family and is expressed on natural killer cells. We found that NCAM1 was associated with major depressive disorder, schizophrenia and bipolar disorder in MR, but only the bipolar disorder result was additionally supported by colocalisation. Variants of the gene encoding NCAM1 have been associated with schizophrenia and bipolar disorder45,46, and NCAM1 knockout and transgenic mice display changes in behaviour related to neuropsychiatric disorder47,48. A recent study demonstrated that anti-NCAM1 autoantibodies in patients with schizophrenia induced schizophrenia-related behaviour and inhibited the formation of synapses in mice, concluding that these antibody proteins may be a potential therapeutic target49. Our finding that CD40 is associated with schizophrenia also implicates immune processes in psychiatric illness aetiology as it is a transmembrane protein found on antigen-presenting cells, which is required for activation of adaptive immune cells. There were other immune related MR associations, such as those between interleukins 20 and 33 and schizophrenia, though these did not receive additional support from colocalisation.
Further, we found that decreased ESAM, an endothelial cell-selective adhesion molecule (ESAM) predominantly expressed at endothelial tight junctions and on platelets50, was associated with greater risk of schizophrenia, with colocalisation support. Its main function involves modulating blood-brain barrier (BBB) permeability and facilitating neutrophil extravasation in the inflammatory response51,52. Notably, junctional adhesion molecules like ESAM play a role in the regulation of tight junctions and the transmigration of leukocytes across the endothelial barrier53. De novo mutations in the ESAM gene have been associated with schizophrenia via exome sequencing and co-expression networks54,55. The association between genetically predicted ESAM plasma levels and schizophrenia was also found in a recent proteomic MR study by Lu et al.: using genetic instruments from 6 genome-wide proteomic studies of primarily European cohorts (n = 28,461, including the Fenland, Framingham Heart, INTERVAL, AGES-Reykjavik, IMPROVE, and KORA F4 studies), they demonstrated that an increase in circulating ESAM levels might be associated with increased schizophrenia risk. They noted, however, that it is unclear whether decreased serum ESAM represents increased or decreased functioning ESAM at tight junctions, an important detail for selecting the best therapeutic target17. Evidence of association with schizophrenia from 2 MR studies implicates ESAM as a target worthy of further investigation.
The FES and FURIN genes are located in the same genomic region, and appear to share regulatory elements56; consequently, it is hard to interpret signals coming from nearby SNPs as indicative of regulation of one, or the other, or both. However, we found associations with either or both genes and multiple psychiatric traits (Table 2).
Fps/Fes, a tyrosine-protein kinase encoded by the FES gene, has a role in membrane and vesicular trafficking that has yet to be fully characterized. Our study implicated this protein through associations with opioid use disorder and bipolar disorder using MR and colocalisation and with schizophrenia and major depressive disorder through MR. Current evidence suggests that it colocalizes with the Golgi apparatus and Rab proteins involved in vesicular transport between subcellular compartments, indicating its potential function in vesicular trafficking regulation57.
Furin, a proprotein convertase, is responsible for converting various inactive precursor proteins into their active forms. We found evidence for the association between FURIN and schizophrenia from MR, and between FURIN and opioid use disorder from MR and colocalisation. Previous studies have highlighted its involvement in multiple neurological and psychiatric processes and condition. FURIN mRNA expression levels were noted to be decreased in the dorsolateral prefrontal cortex of post-mortem schizophrenia patients58. One study demonstrated that the rs17514846 variant in FURIN is implicated in brain-derived neurotrophic factor (BDNF)-related pathways and metabolism59. A decrease in FURIN levels was also shown to potentially disrupt the maturation of other neurotrophins like NGF and NT360. This deficit in neurotrophic support can potentially lead to neural maldevelopment, altered neural plasticity, and neurodegeneration61. It would be beneficial to further investigate the relationship between FURIN deregulation and its effect on the pathology of schizophrenia.
We did see support from both MR and colocalisation for the association between midkine (MDK) and schizophrenia. Midkine (MDK) is a heparin-binding growth factor that has also been termed neurite growth-promoting factor 2 (NEGF2), and has been shown to be essential for the survival of mouse mesencephalic neurons, specifically dopaminergic neurons, in culture62. MDK-deficient mice exhibit postnatal hippocampal developmental delays, leading to working memory deficits and heightened anxiety63. Notably, a hypodopaminergic state in the striatum of these mice links MDK to dopamine-related disorders, suggesting potential relevance to conditions like schizophrenia and Parkinson’s disease62. This work implicates a potential contribution of MDK to neuropsychiatric pathology via brain-mediated pathways, not blood-mediated pathways. In the current study, we only examined genetically predicted plasma protein levels, but we note that genetic markers of plasma proteins may also associate with protein levels in other tissues. Though pathophysiological mechanisms are not confined to the brain, parsing potential differences between proteins in circulation and proteins in the brain may be important as casually intervening on protein levels or activity in the blood may not affect all tissues equally. Tissue-specific proteomic MR can be helpful in probing these issues, though is dependent on the quality and results of Mendelian randomisation studies should be triangulated with other evidence when considering work towards intervention for illness prevention and/or treatment. We saw MR evidence for association with schizophrenia for several proteins previously implicated in schizophrenia biology whose putative site of action are in the brain, including GRIN2B, a glutamate ionotropic receptor NMDA subunit, and ERBB4, a tyrosine-protein kinase that acts a cell surface receptor for neuregulins and EGF family members. There was also MR support for associations between the calcium channel protein CACNA1C and opioid use disorder, and between CACNB3 and schizophrenia. However, CACNA1C, CACNB3, GRIN2B, ERBB4 and EGF did not receive support from colocalisation analysis.
Strengths and limitations
Here we utilized the most up-to-date GWAS of the plasma proteome and psychiatric disorder traits at the time of analysis, and considered multiple psychiatric disorders, as well as treatment-related psychiatric traits, in this analysis. Psychiatric disorders can be comorbid, and exhibit overlap in aetiology and presentation. As a result, some may share similar targets. By looking at multiple disorders, it is possible to observe unanticipated cross-diagnostic patterns. Although proteomic multiplexing technology has rapidly advanced, large scale GWAS data only exist on a minority of the entire plasma proteome, and our study only investigated associations with proteins assayed by the OLINK platform used in the UK Biobank plasma protein exposure GWAS. Furthermore, of the proteins that are measured, assay may vary in sensitivity with consequences for the precision of GWAS effect estimates. Restricting analyses to cis-pQTLs and using the lead-SNP of the cis-pQTLs mitigates against horizontal pleiotropy but does not eliminate it. Corroborating findings with colocalisation reduces the likelihood of associations due to horizontal pleiotropy, though simply because a Mendelian randomisation association does not receive additional support from colocalisation does not necessarily indicate that the association is not causal. We anticipate that studies investigating plasma protein exposures and mental disorder outcomes will play a role in understanding the aetiology and treatment of mental disorders, but these methods should be complemented by other approaches exploring the mechanistic underpinnings of the identified proteins and their role in the psychiatric pathophysiology. Additional functional studies and target validation may provide further insights into the complex genetic basis of psychiatric conditions and potentially open avenues for future therapeutic targets. It is also important to evaluate genetic and bioinformatic findings against other types of evidence, including from observational studies and clinical trials.