Advertisement
  • Loading metrics

Investigating the Causal Relationship of C-Reactive Protein with 32 Complex Somatic and Psychiatric Outcomes: A Large-Scale Cross-Consortium Mendelian Randomization Study

  • Bram. P. Prins ,

    bram.prins@sanger.ac.uk (BPP); b.z.alizadeh@umcg.nl (BZA)

    Affiliations Department of Epidemiology, University of Groningen, University Medical Center Groningen, Groningen, The Netherlands, Department of Human Genetics, Wellcome Trust Sanger Institute, Hinxton, United Kingdom

  • Ali Abbasi ,

    Contributed equally to this work with: Ali Abbasi, Anson Wong, Ahmad Vaez

    Affiliations Department of Epidemiology, University of Groningen, University Medical Center Groningen, Groningen, The Netherlands, MRC Epidemiology Unit, University of Cambridge School of Clinical Medicine, Institute of Metabolic Science, Addenbrooke’s Hospital, Cambridge, United Kingdom, Department of Internal Medicine, University of Groningen, University Medical Center Groningen, Groningen, the Netherlands

  • Anson Wong ,

    Contributed equally to this work with: Ali Abbasi, Anson Wong, Ahmad Vaez

    Affiliations Institute of Medical Sciences, University of Toronto, Toronto, Ontario, Canada, Faculty of Medicine, University of Ottawa, Ottawa, Ontario, Canada

    ORCID http://orcid.org/0000-0001-7269-9429

  • Ahmad Vaez ,

    Contributed equally to this work with: Ali Abbasi, Anson Wong, Ahmad Vaez

    Affiliations Department of Epidemiology, University of Groningen, University Medical Center Groningen, Groningen, The Netherlands, School of Medicine, Isfahan University of Medical Sciences, Isfahan, Iran

  • Ilja Nolte,

    Affiliation Department of Epidemiology, University of Groningen, University Medical Center Groningen, Groningen, The Netherlands

  • Nora Franceschini,

    Affiliation Department of Epidemiology, University of North Carolina, Chapel Hill, North Carolina, United States of America

  • Philip E. Stuart,

    Affiliation Department of Dermatology, Veterans Affairs Hospital, University of Michigan, Ann Arbor, Michigan, United States of America

    ORCID http://orcid.org/0000-0002-9546-6529

  • Javier Guterriez Achury,

    Affiliation Department of Genetics, University of Groningen, University Medical Centre Groningen, Groningen, the Netherlands

  • Vanisha Mistry,

    Affiliations Metabolic Research Laboratories, Wellcome Trust–MRC Institute of Metabolic Science, Addenbrooke’s Hospital, University of Cambridge, Cambridge, United Kingdom, Blizard Institute, Barts and The London School of Medicine and Dentistry, Queen Mary University of London, London, United Kingdom

  • Jonathan P. Bradfield,

    Affiliation Center for Applied Genomics, Division of Human Genetics, Children’s Hospital of Philadelphia, Philadelphia, United States of America

  • Ana M. Valdes,

    Affiliation Department of Academic Rheumatology, University of Nottingham, Nottingham, United Kingdom

  • Jose Bras,

    Affiliation Department of Molecular Neuroscience, Institute of Neurology, London, United Kingdom

  • Aleksey Shatunov,

    Affiliation Department of Basic and Clinical Neuroscience, Institute of Psychiatry, Psychology and Neuroscience, King’s College London, London, United Kingdom

  • PAGE Consortium,

    Affiliation Department of Dermatology, Veterans Affairs Hospital, University of Michigan, Ann Arbor, Michigan, United States of America

  • International Stroke Genetics Consortium,

    Affiliation Neurology Unit, Department of Clinical Neurosciences, University of Cambridge, Cambridge, United Kingdom

  • Systemic Sclerosis consortium,

    Affiliations Division of Rheumatology and Clinical Immunogenetics, University of Texas Health Science Center at Houston, Houston, Texas, United States of America, Instituto de Parasitologia y Biomedicina Lopez-Neyra, Consejo Superior de Investigaciones Científicas, Granada, Spain

  • Treat OA consortium,

    Affiliation Department of Hygiene and Epidemiology, University of Ioannina Medical School, Ioannina, Greece

  • DIAGRAM Consortium,

    Affiliation Wellcome Trust Centre for Human Genetics, University of Oxford, Oxford, United Kingdom

  • CARDIoGRAMplusC4D Consortium,

    Affiliation Institut für Integrative und Experimentelle Genomik, Universität zu Lübeck, Lübeck, Germany

  • ALS consortium,

    Affiliation Department of Basic and Clinical Neuroscience, Institute of Psychiatry, Psychology and Neuroscience, King’s College London, London, United Kingdom

  • International Parkinson’s Disease Genomics Consortium,

    Affiliation Department of Molecular Neuroscience, UCL Institute of Neurology, London, United Kingdom

  • Autism Spectrum Disorder Working Group of the Psychiatric Genomics Consortium,

    Affiliation McKusick-Nathans Institute of Genetic Medicine and Department of Medicine, Division of Cardiology, Johns Hopkins University School of Medicine, Baltimore, Maryland, United States of America

  • CKDGen consortium,

    Affiliation NHLBI’s Framingham Heart Study, Center for Population Studies and Harvard Medical School, Framingham, Massachusetts, United States of America

  • GERAD1 Consortium,

    Affiliation Institute of Psychological Medicine and Clinical Neurosciences, School of Medicine, Cardiff University, Cardiff, United Kingdom

  • International Consortium for Blood Pressure,

    Affiliation NIHR Barts Cardiovascular Biomedical Research Unit, William Harvey Research Institute, Queen Mary University of London, London, United Kingdom

  • Schizophrenia Working Group of the Psychiatric Genomics Consortium,

    Affiliation MRC Centre for Neuropsychiatric Genetics and Genomics, Institute of Psychological Medicine and Clinical Neurosciences, School of Medicine, Cardiff University, United Kingdom

  • Inflammation Working Group of the CHARGE Consortium,

    Affiliation Department of Epidemiology, University of Groningen, University Medical Center Groningen, Groningen, The Netherlands

  • Chen Lu,

    Affiliation Department of Biostatistics, Boston University School of Public Health, Boston, Massachusetts, United States of America

  • Buhm Han,

    Affiliation Program in Medical and Population Genetics, Broad Institute of Harvard and MIT, Cambridge, Massachusetts, United States of America

  • Soumya Raychaudhuri,

    Affiliations Program in Medical and Population Genetics, Broad Institute of Harvard and MIT, Cambridge, Massachusetts, United States of America, Division of Genetics, Brigham and Women’s Hospital, Harvard Medical School, Boston, Massachusetts, United States of America, Division of Rheumatology, Brigham and Women’s Hospital, Harvard Medical School, Boston, Massachusetts, United States of America, Partners HealthCare Center for Personalized Genetic Medicine, Boston, Massachusetts, United States of America, Faculty of Medical and Human Sciences, University of Manchester, Manchester, United Kingdom

  • Steve Bevan,

    Affiliation Neurology Unit, Department of Clinical Neurosciences, University of Cambridge, Cambridge, United Kingdom

  • Maureen D. Mayes,

    Affiliation Division of Rheumatology and Clinical Immunogenetics, University of Texas Health Science Center at Houston, Houston, Texas, United States of America

  • Lam C. Tsoi,

    Affiliations Department of Dermatology, Veterans Affairs Hospital, University of Michigan, Ann Arbor, Michigan, United States of America, Department of Biostatistics, University of Michigan, Ann Arbor, Michigan, United States of America

  • Evangelos Evangelou,

    Affiliations Department of Hygiene and Epidemiology, University of Ioannina Medical School, Ioannina, Greece, Department of Epidemiology and Biostatistics, School of Public Health, Imperial College London, London, United Kingdom

  • Rajan P. Nair,

    Affiliation Department of Dermatology, Veterans Affairs Hospital, University of Michigan, Ann Arbor, Michigan, United States of America

  • Struan F. A. Grant,

    Affiliations Center for Applied Genomics, Division of Human Genetics, Children’s Hospital of Philadelphia, Philadelphia, United States of America, Department of Pediatrics, Perelman School of Medicine, University of Pennsylvania, Philadelphia, United States of America

  • Constantin Polychronakos,

    Affiliation Endocrine Genetics Research Institute, McGill University Health Center, Montreal, Quebec, Canada

  • Timothy R. D. Radstake,

    Affiliation Department of Rheumatology & Clinical Immunology and Laboratory of Translational Immunology, University Medical Center Utrecht, Utrecht, the Netherlands

  • David A. van Heel,

    Affiliation Blizard Institute, Barts and The London School of Medicine and Dentistry, Queen Mary University of London, London, United Kingdom

  • Melanie L. Dunstan,

    Affiliation Institute of Psychological Medicine and Clinical Neurosciences, School of Medicine, Cardiff University, Cardiff, United Kingdom

    ORCID http://orcid.org/0000-0003-3817-2142

  • Nicholas W. Wood,

    Affiliation Department of Molecular Neuroscience, UCL Institute of Neurology, London, United Kingdom

  • Ammar Al-Chalabi,

    Affiliations Department of Basic and Clinical Neuroscience, Institute of Psychiatry, Psychology and Neuroscience, King’s College London, London, United Kingdom, Complex Disease Genetics, Cold Spring Harbor Laboratory, Cold Spring Harbor, New York, United States of America

  • Abbas Dehghan,

    Affiliation Department of Epidemiology, Erasmus University Rotterdam, University Medical Centre Rotterdam, Rotterdam, the Netherlands

  • Hakon Hakonarson,

    Affiliations Center for Applied Genomics, Division of Human Genetics, Children’s Hospital of Philadelphia, Philadelphia, United States of America, Department of Pediatrics, Perelman School of Medicine, University of Pennsylvania, Philadelphia, United States of America

  • Hugh S. Markus,

    Affiliation Neurology Unit, Department of Clinical Neurosciences, University of Cambridge, Cambridge, United Kingdom

  • James T. Elder,

    Affiliation Department of Dermatology, Veterans Affairs Hospital, University of Michigan, Ann Arbor, Michigan, United States of America

  • Jo Knight,

    Affiliations Institute of Medical Sciences, University of Toronto, Toronto, Ontario, Canada, Campbell Family Mental Health Research Institute, Centre for Addiction and Mental Health, Toronto, Ontario, Canada, Department of Psychiatry, University of Toronto, Toronto, Ontario, Canada, Biostatistics Division, Dalla Lana School of Public Health, University of Toronto, Toronto, Ontario, Canada

  • Dan E. Arking,

    Affiliation McKusick-Nathans Institute of Genetic Medicine and Department of Medicine, Division of Cardiology, Johns Hopkins University School of Medicine, Baltimore, Maryland, United States of America

  • Timothy D. Spector,

    Affiliation Department of Twin Research and Genetic Epidemiology, King’s College London, London, United Kingdom

  • Bobby P. C. Koeleman,

    Affiliation Complex Genetic Section, Department of Medical Genetics, University Medical Center Utrecht, Utrecht, the Netherlands

  • Cornelia M. van Duijn,

    Affiliation Department of Epidemiology, Erasmus University Rotterdam, University Medical Centre Rotterdam, Rotterdam, the Netherlands

  • Javier Martin,

    Affiliation Instituto de Parasitologia y Biomedicina Lopez-Neyra, Consejo Superior de Investigaciones Científicas, Granada, Spain

  • Andrew P. Morris,

    Affiliations Wellcome Trust Centre for Human Genetics, University of Oxford, Oxford, United Kingdom, Department of Biostatistics, University of Liverpool, Liverpool, United Kingdom

  • Rinse K. Weersma,

    Affiliation Department of Gastroenterology and Hepatology, University of Groningen, University Medical Center Groningen, Groningen, the Netherlands

  • Cisca Wijmenga,

    Affiliation Department of Genetics, University of Groningen, University Medical Centre Groningen, Groningen, the Netherlands

  • Patricia B. Munroe,

    Affiliations NIHR Barts Cardiovascular Biomedical Research Unit, William Harvey Research Institute, Queen Mary University of London, London, United Kingdom, Clinical Pharmacology, William Harvey Research Institute, Barts and the London School of Medicine, Queen Mary University of London, London, United Kingdom

  • John R. B. Perry,

    Affiliation MRC Epidemiology Unit, University of Cambridge School of Clinical Medicine, Institute of Metabolic Science, Addenbrooke’s Hospital, Cambridge, United Kingdom

  • Jennie G. Pouget,

    Affiliations Institute of Medical Sciences, University of Toronto, Toronto, Ontario, Canada, Campbell Family Mental Health Research Institute, Centre for Addiction and Mental Health, Toronto, Ontario, Canada, Department of Psychiatry, University of Toronto, Toronto, Ontario, Canada

  • Yalda Jamshidi,

    Affiliation Cardiogenetics Lab, Cardiovascular and Cell Sciences Institute, St George’s Hospital Medical School, London, United Kingdom

  • Harold Snieder,

    Affiliation Department of Epidemiology, University of Groningen, University Medical Center Groningen, Groningen, The Netherlands

  •  [ ... ],
  • Behrooz Z. Alizadeh

    bram.prins@sanger.ac.uk (BPP); b.z.alizadeh@umcg.nl (BZA)

    Affiliations Department of Epidemiology, University of Groningen, University Medical Center Groningen, Groningen, The Netherlands, Department of Gastroenterology and Hepatology, University of Groningen, University Medical Center Groningen, Groningen, the Netherlands

  • [ view all ]
  • [ view less ]

Investigating the Causal Relationship of C-Reactive Protein with 32 Complex Somatic and Psychiatric Outcomes: A Large-Scale Cross-Consortium Mendelian Randomization Study

  • Bram. P. Prins, 
  • Ali Abbasi, 
  • Anson Wong, 
  • Ahmad Vaez, 
  • Ilja Nolte, 
  • Nora Franceschini, 
  • Philip E. Stuart, 
  • Javier Guterriez Achury, 
  • Vanisha Mistry, 
  • Jonathan P. Bradfield
PLOS
x

Abstract

Background

C-reactive protein (CRP) is associated with immune, cardiometabolic, and psychiatric traits and diseases. Yet it is inconclusive whether these associations are causal.

Methods and Findings

We performed Mendelian randomization (MR) analyses using two genetic risk scores (GRSs) as instrumental variables (IVs). The first GRS consisted of four single nucleotide polymorphisms (SNPs) in the CRP gene (GRSCRP), and the second consisted of 18 SNPs that were significantly associated with CRP levels in the largest genome-wide association study (GWAS) to date (GRSGWAS). To optimize power, we used summary statistics from GWAS consortia and tested the association of these two GRSs with 32 complex somatic and psychiatric outcomes, with up to 123,865 participants per outcome from populations of European ancestry. We performed heterogeneity tests to disentangle the pleiotropic effect of IVs. A Bonferroni-corrected significance level of less than 0.0016 was considered statistically significant. An observed p-value equal to or less than 0.05 was considered nominally significant evidence for a potential causal association, yet to be confirmed.

The strengths (F-statistics) of the IVs were 31.92–3,761.29 and 82.32–9,403.21 for GRSCRP and GRSGWAS, respectively. CRP GRSGWAS showed a statistically significant protective relationship of a 10% genetically elevated CRP level with the risk of schizophrenia (odds ratio [OR] 0.86 [95% CI 0.79–0.94]; p < 0.001). We validated this finding with individual-level genotype data from the schizophrenia GWAS (OR 0.96 [95% CI 0.94–0.98]; p < 1.72 × 10−6). Further, we found that a standardized CRP polygenic risk score (CRPPRS) at p-value thresholds of 1 × 10−4, 0.001, 0.01, 0.05, and 0.1 using individual-level data also showed a protective effect (OR < 1.00) against schizophrenia; the first CRPPRS (built of SNPs with p < 1 × 10−4) showed a statistically significant (p < 2.45 × 10−4) protective effect with an OR of 0.97 (95% CI 0.95–0.99). The CRP GRSGWAS showed that a 10% increase in genetically determined CRP level was significantly associated with coronary artery disease (OR 0.88 [95% CI 0.84–0.94]; p < 2.4 × 10−5) and was nominally associated with the risk of inflammatory bowel disease (OR 0.85 [95% CI 0.74–0.98]; p < 0.03), Crohn disease (OR 0.81 [95% CI 0.70–0.94]; p < 0.005), psoriatic arthritis (OR 1.36 [95% CI 1.00–1.84]; p < 0.049), knee osteoarthritis (OR 1.17 [95% CI 1.01–1.36]; p < 0.04), and bipolar disorder (OR 1.21 [95% CI 1.05–1.40]; p < 0.007) and with an increase of 0.72 (95% CI 0.11–1.34; p < 0.02) mm Hg in systolic blood pressure, 0.45 (95% CI 0.06–0.84; p < 0.02) mm Hg in diastolic blood pressure, 0.01 ml/min/1.73 m2 (95% CI 0.003–0.02; p < 0.005) in estimated glomerular filtration rate from serum creatinine, 0.01 g/dl (95% CI 0.0004–0.02; p < 0.04) in serum albumin level, and 0.03 g/dl (95% CI 0.008–0.05; p < 0.009) in serum protein level. However, after adjustment for heterogeneity, neither GRS showed a significant effect of CRP level (at p < 0.0016) on any of these outcomes, including coronary artery disease, nor on the other 20 complex outcomes studied. Our study has two potential limitations: the limited variance explained by our genetic instruments modeling CRP levels in blood and the unobserved bias introduced by the use of summary statistics in our MR analyses.

Conclusions

Genetically elevated CRP levels showed a significant potentially protective causal relationship with risk of schizophrenia. We observed nominal evidence at an observed p < 0.05 using either GRSCRP or GRSGWAS—with persistence after correction for heterogeneity—for a causal relationship of elevated CRP levels with psoriatic osteoarthritis, rheumatoid arthritis, knee osteoarthritis, systolic blood pressure, diastolic blood pressure, serum albumin, and bipolar disorder. These associations remain yet to be confirmed. We cannot verify any causal effect of CRP level on any of the other common somatic and neuropsychiatric outcomes investigated in the present study. This implies that interventions that lower CRP level are unlikely to result in decreased risk for the majority of common complex outcomes.

Editors' Summary

Background

Inflammation is an important part of the human immune response, the network of cells and molecules that protects the body from attack by pathogens (infectious organisms) and from harmful substances and foreign particles (for example, splinters). When human cells are attacked by pathogens or injured by trauma or chemicals, molecules called inflammatory mediators induce fluid leakage from the blood vessels into the damaged tissue and attract “phagocytes” (a type of immune cell) to the site of infection or injury to “eat” the germs and dead or damaged cells. The end result is inflammation, which is characterized by swelling, redness, heat, and pain. The inflammatory response, although unpleasant, limits the damage caused by foreign invaders or chemicals by preventing further contact with body tissues. Sometimes, however, inflammation can be harmful. Persistent dysregulation of the inflammatory response is implicated in numerous somatic disorders (diseases that affect the body, such as cardiovascular disease) and neuropsychiatric disorders (mental disorders attributable to diseases of the nervous system, such as schizophrenia).

Why Was This Study Done?

Observational studies suggest that increased blood levels of C-reactive protein (CRP, an inflammatory protein) are associated with certain somatic and neuropsychiatric disorders. But observational studies cannot prove that changes in CRP levels actually cause any of these disorders. It could be that the individuals who develop a specific disease and who have a high CRP level also share another unknown characteristic that is actually responsible for disease development (confounding). Alternatively, it could be that the disease itself increases CRP levels (reverse causation). It is important to know whether CRP is causally involved in the development of specific diseases because it might then be possible to prevent or treat these diseases using drugs that control CRP levels. Here, the researchers undertake a Mendelian randomization study to determine whether CRP has a causal relationship with 32 common complex somatic and neuropsychiatric outcomes. Because gene variants are inherited randomly, they are not prone to confounding and are free from reverse causation. So, if CRP levels actually cause a specific somatic or neuropsychiatric disease, genetic variants that affect CRP levels should be associated with an altered risk for that disease.

What Did the Researchers Do and Find?

The researchers used data collected by several consortia involved in large genome-wide association studies (studies that ask whether specific genetic changes across the whole human genome, or blueprint, are associated with specific diseases) to look for associations between 32 somatic and neuropsychiatric outcomes and two genetic risk scores (GRSs) for CRP level. GRSCRP consisted of four single nucleotide polymorphisms (SNPs; a type of genetic variant) in the gene encoding CRP; GRSGWAS consisted of 18 SNPs that were associated with CRP level in a genome-wide association study. The researchers report that a genetically increased CRP level was significantly associated with a reduced risk of schizophrenia (a significant association is one unlikely to have arisen by chance). In addition, they found a nominally significant association (an association that needs to be confirmed) between genetically increased CRP levels and an increased risk of knee osteoarthritis, raised diastolic and systolic blood pressure, and bipolar disorder. Notably, there was no evidence for an effect of genetically increased CRP levels on any of the other 27 outcomes studied.

What Do These Findings Mean?

These findings suggest that genetically raised levels of CRP are causally associated with protection against schizophrenia, an unexpected finding given other recent studies that suggest that raised CRP levels and brain inflammation predispose individuals to schizophrenia. The findings also provide preliminary evidence that genetically raised levels of CRP may be causally associated with an increased risk of raised blood pressure, knee arthritis, and bipolar disorder. The lack of any association between genetically raised levels of CRP and the other outcomes studied suggests, however, that many previously identified disease-associated rises in CRP levels might be a response to the disease process rather than a cause of these diseases. Like all Mendelian randomization studies, the reliability of these findings depends on the validity of several assumptions made by the researchers and on the ability of the GRSs used in the study to explain variations in CRP level. Importantly, however, these findings suggest that interventions designed to lower CRP level are unlikely to decrease the risk of people developing the majority of common complex somatic and neuropsychiatric outcomes.

Additional Information

This list of resources contains links that can be accessed when viewing the PDF on a device or via the online version of the article at http://dx.doi.org/10.1371/journal.pmed.1001976.

Introduction

Emerging evidence suggests that persistent dysregulation of the inflammatory response is linked to a plethora of complex somatic and neuropsychiatric disorders [118]. Epidemiological studies have shown that C-reactive protein (CRP), a well-studied biomarker of inflammation, is associated with and exhibits reliable predictive value for cardiovascular disease [19,20], type 2 diabetes [21], immunity-related disorders such as inflammatory bowel disease (IBD) [22], rheumatoid arthritis [23], and all-cause mortality [20,24]. Nevertheless, the evidence for a causal involvement of CRP in these outcomes from traditional experimental or observational studies remains controversial [25,26], fueling the debate surrounding whether CRP contributes to the chain of causality in disease mechanisms [27]. The use of genetically informed instrumental variables (IVs), termed Mendelian randomization (MR), is a complementary approach to epidemiological observations and allows investigation of whether the effect of an exposure (i.e., CRP level) on observed outcome phenotypes is likely to be causal [28].

Recent large-scale MR studies, focusing mainly on cardiovascular disease and metabolic traits, failed to show a causal association between CRP level and these outcomes (S1 Table). This has led to the notion that elevated CRP levels do not causally contribute to these traits and disorders. However, these studies used a single CRP-associated single nucleoid polymorphism (SNP) or a very limited set of CRP-associated SNPs (S1 Table). Common SNPs serving as proxies for CRP level represent only a small effect on CRP level per se and thus require a large enough sample size to detect causal effects on the outcome. Moreover, most studies have generally included a limited range of common complex diseases, often not more than two or three outcomes, or they have been performed in a single or small population, yielding inadequate study power (S1 Table). In other words, the evidence for a causal relationship between CRP and a broad range of common traits or diseases remains inconclusive. This is mostly due to the lack of well-powered MR studies that use optimally informative genetic IVs for CRP. Here, we sought to comprehensively examine the hypothesis that genetically determined CRP level directly contributes to common somatic and psychiatric outcomes. To optimize IV power, we applied a MR approach using summary statistics from large-scale genome-wide association study (GWAS) consortia of 32 somatic and psychiatric phenotypes for the four CRP variants representing 98% of the common variation in the CRP gene and for the largest known set of independent SNPs known to be associated with CRP. We further aimed to confirm the identified association between CRP and schizophrenia using a CRP polygenic risk score (CRPPRS) from individual-level genotype data from the largest consortium of schizophrenia to date. We performed an in silico pathway analysis (see Discussion) to provide insights into the possible mechanism underlying the observed association of CRP level with schizophrenia.

Methods

Study Design and Rationale

The present MR study consists of two key components. First, we used established gene variants associated with CRP level and combined them to build two genetic risk scores (GRSs) for CRP. The first GRS consisted of only four SNPs in the CRP gene (GRSCRP) selected from the largest recent MR study of CRP [29], and the second consisted of 18 SNPs that were associated with CRP level at a genome-wide significance level in the largest GWAS for CRP to date (GRSGWAS) [30]. Second, we obtained summary association statistics from GWAS consortia for a panel of 32 common somatic and psychiatric outcomes (Table 1). BPP and BZA selected the studies, and contacted each consortium with a standardized request for study data, including the name of the study or consortium, the number of cases and controls, the number of available CRP SNPs for GRSCRP and GRSGWAS, and the estimated effect for each SNP (or its proxy) on outcome, i.e., the per allele regression coefficient with standard error or the odds ratio (OR) and corresponding 95% confidence interval. Data were available for 32 different outcomes in five broad disease/trait classes (autoimmune/inflammatory, cardiovascular, metabolic, neurodegenerative, and psychiatric), including at least 1,566, and up to 184,305, participants per outcome from populations of European ancestry (Table 1). These outcomes were selected based on the following two inclusion criteria: (i) the outcome having been associated with CRP level in epidemiological studies and (ii) availability of large meta-GWAS analyses for the outcome (Table 1).

Genetic Instruments

Weak IVs yielding insufficient statistical power may have hampered estimation of causal effects of CRP on the outcomes in previous analyses (S1 Table). Our MR approach, by using GWAS data and combining multiple independent SNPs into a GRS (i.e., IV), has the potential to greatly increase power. The selected SNPs have been described elsewhere [30,55,56] and are further detailed in S2S4 Tables. These IVs were used to test the combined effect of the associations of CRP-level-influencing alleles with the outcomes. Our approach was implemented in such a way that the effects of both independent SNPs in the CRP gene (GRSCRP) [55,56] (S1 Methods) and independent SNPs known to be genome-wide significantly associated with CRP levels (GRSGWAS) [30], as well as pleiotropic effects of SNPs, could be discriminated [57]. Pleiotropy exists if CRP SNPs influence exposures (risk factors) other than CRP level and therefore violate one of the key MR assumptions.

Statistical Analysis

All analyses were done using the GRS function implemented in the grs.summary module of the R package Genetics ToolboX (version 2.15.1 for Windows). The grs.summary module approximates the regression of an outcome onto an additive GRS, using only single SNP association summary statistics extracted from GWAS results. The method is described in more detail elsewhere [58]. In brief, we performed MR analyses using GRS IVs in two steps. First, we used four individual CRP gene SNPs (i.e., IVs) associated with CRP level [56,59] (S2 and S3 Tables) to create a weighted GRS, named GRSCRP, corresponding to the joint effect of the four SNPs within the CRP gene [55]. We extracted ω (the estimated coefficient, or weight) for individual SNPs from association results reported by the CRP Coronary Heart Disease Genetics Collaboration (CCGC) [29,55]; ω represents a one-unit (in mg/l) increase of the natural log of CRP level (lnCRP) per dose of the coded allele. The four tagging SNPs represent 98% of the common variation in the CRP gene, assuming a minor allele frequency of ≥0.05 and an r2 threshold of ≥0.8, and aggregately explain ~2% of the total variation (i.e., phenotypic variance) in serum CRP level in populations of European descent [55,59]. Second, we constructed a multilocus GRS, named GRSGWAS, that combined 18 SNPs associated with serum CRP level at a genome-wide significance level (p < 5×10−8; S2 and S3 Tables), derived from a large meta-GWAS analysis of CRP conducted by the CHARGE (Cohorts for Heart and Aging Research in Genomic Epidemiology) Consortium [30]. This multilocus GRS explains approximately ~5% of the total variation in serum CRP level [30].

We integrated ω for each CRP SNP from the reference data of CCGC [55] or meta-analysis of GWASs [30] for CRP level with the summary association statistics extracted from the GWAS consortium data for each outcome (S1 Data; S2 Methods). This MR approach using meta-GWAS summary statistics data is equivalent to an inverse-variance-weighted meta-analysis and has previously been validated in comparison to individual-level data [57,60]. To estimate the causal effect of CRP level on an outcome, we obtained the β values (estimated effects from regression analysis) for the effects of CRP SNPs on the outcome, with standard errors, seβ, from the corresponding GWAS results. Where no summary statistics for a CRP SNP in the GRS IVs were available in the look-up dataset, we chose the proxy SNP that had the highest linkage disequilibrium with the initial SNP (r2 > 0.9 in HapMap release 22; S3 Table). If several proxy SNPs had the exact same r2 value, we chose the proxy nearest to the original SNP in the instrument. Separate regressions of outcomes on GRSs were performed to calculate αIV estimators (i.e., causal IV estimators) for each outcome. Correspondingly, the value of a GRS is the sum of the ω values multiplied by the allele dosage (i.e., 0, 1, or 2) for each CRP SNP in the CCGC or in the CHARGE Consortium data [30,55]. For uncorrelated SNPs, when maximizing the likelihood function, the αIV value and its standard error, seα, can be approximated with the formula α ≅ (Σω × β × seβ−2) /(Σω2 × seβ−2), with seα ≅ √1/ Σω2 × seβ−2. lnCRP was used as the outcome in reference studies [30,55], so in obtaining the ω values (i.e., effect sizes) for each of the CRP SNPs, a unit increase in lnCRP equals a 10 symmetric percentage (s%) increase in CRP level, which corresponds to a unit change in the level of a continuous outcome or logit of risk estimate (i.e., beta coefficient) for a dichotomous outcome [61]. The αIV value (i.e., causal estimate) for each CRP SNP is, therefore, presented for each outcome as corresponding to a 10-s% increase in actual CRP level. During the course of this study, an updated, larger GWAS dataset for coronary artery disease (CAD) became publicly available (CARDIoGRAMplusC4D Consortium, release 2015 [41]); we therefore redid the analysis for CAD using the release 2015 data.

To assess which SNPs might have violated the key MR assumption regarding pleiotropy, we performed goodness-of-fit tests to correct both GRSs for the heterogeneity of their corresponding SNPs’ effects on each outcome. Heterogeneity, which indicates the potential presence of pleiotropy, was measured using the Q statistic and was considered statistically significant at a conservative uncorrected p-value of <0.05. Although heterogeneity could be an indicator of pleiotropy, there are other factors that could introduce heterogeneity in the analyses. Even though the adjustments for heterogeneity that we have made could be overconservative, we have used this method in order to minimize false positives. After stepwise removal of SNPs with potential pleiotropic effects, we repeated the analyses until significant heterogeneity was no longer observed.

To further ensure the strength of these two GRSs as IVs, we generated an F-statistic for each outcome. We used variance in lnCRP explained by each set of CRP SNPs (2% and 5%, respectively, for GRSCRP and GRSGWAS) to calculate the F-statistic using the formula F-statistic = [R2 × (n − 1 − K)]/[(1 − R2) × K], where R2 represents the proportion of variability in CRP level that is explained by the GRS, n represents sample size, and K represents the number of IVs included in model (i.e., for this study K = 1) [62]. As a rule of thumb, an F-value above ten indicates that a causal estimate is unlikely to be biased due to weak instruments [57].

Multiple Testing

The present study included 32 independent sample sets. For each sample set, we did one statistical test, for which a global nominal significance level of ≤0.05 was considered as satisfactory to derive conclusions. The need for correction for multiple testing is debatable. Nevertheless, to ensure the validity of our conclusions, we took a conservative approach and applied a Bonferroni-corrected significance threshold calculated as 0.05 divided by 32 (i.e., 0.0016). We considered a statistical test with an observed p-value more than 0.05 as a definitely nonsignificant result, i.e., no association; an observed p-value equal to or less than 0.05 as nominally significant evidence for a potential, but yet to be confirmed, causal association; and an observed p-value equal to or less than 0.0016 as statistically significant evidence for a causal association.

CRP Polygenic Risk Score and Schizophrenia Using Individual-Level Data

In an ancillary follow-up study, inspired by comments by the editors and the reviewers, we aimed further to determine whether GRSGWAS was causally associated with schizophrenia using individual-level data retrieved from the Psychiatric Genomics Consortium (PGC) schizophrenia dataset (S3 Methods) [54]. This dataset consisted of 36 independent cohorts with a combined 25,629 cases and 30,976 controls for which we had ethics approval (S4 Methods). Three family-based samples of European ancestry (1,235 parent–affected offspring trios) were excluded from our analysis. To evaluate whether the observed protective causal association between GRSGWAS and schizophrenia was persistent, we investigated whether the CRPPRS was also protectively associated with schizophrenia. Briefly, CRPPRS values were calculated for each individual by summing the total effect of the SNP dosages by their effect size. In addition to the 18 genome-wide significant CRP SNPs, we grouped subthreshold CRP-associated SNPs at the following p-value thresholds: 1 × 10−4, 0.001, 0.01, 0.05, and 0.1. Standardized CRPPRS values were tested for association with schizophrenia case status in each cohort with adjustment for ten principal components (PCs). A fixed effects inverse-variance-weighted meta-analysis was performed across all 36 cohorts to obtain the overall effect size estimate as explained in S4 Methods and elsewhere [63]. The variance in schizophrenia case status explained by CRPPRS was estimated using the deviation in Nagelkerke’s pseudo-R2 between a null model (which included ten PCs) and the full model (which included GRS in addition to the ten PCs), calculated in R using the Functions for Medical Statistics Book with Some Demographic Data (fmsb) R package (S3 Methods). Similar to previous studies, the statistical significance of CRPPRS values was estimated based on their logistic regression coefficient [64], and reported CRPPRS ORs correspond to a 1-SD increase in CRPPRS [65].

Results

Using GRSCRP, we first tested whether a CRP-gene-determined increase in lnCRP was associated with each outcome. In Table 2, the causal effects of lnCRP estimated for each outcome are summarized. We found no heterogeneity in the IV analyses (pheterogeneity ≥ 0.11 for all outcomes), and GRSCRP was a strong instrument (F ≥ 31). IV analyses provided nominal evidence for potential causal relationships of lnCRP with risk of Crohn disease (OR 0.78 [95% CI 0.65–0.94]; p < 0.009), psoriatic arthritis (1.45 [1.04–2.04]; p < 0.03), and schizophrenia (0.90 [0.82–0.99]; p < 0.03), and with an increase in systolic blood pressure (SBP) (mean increase 1.23 mm Hg per 10-s% increase in CRP level [95% CI 0.45–2.01]; p < 0.002) and diastolic blood pressure (DBP) (0.70 [0.20–1.19]; p < 0.006). GRSCRP showed no significant effect on any of the other outcomes (Table 2; S1 Fig).

thumbnail
Table 2. The effect of the CRP genetic risk score instrument of four SNPs in CRP (GRSCRP) with somatic and neuropsychiatric outcomes.

https://doi.org/10.1371/journal.pmed.1001976.t002

GRSGWAS showed a statistically significant protective effect of lnCRP on the risk of schizophrenia (per 10-s% increase in CRP level, OR 0.86 [95% CI 0.79–0.94]; p < 0.0010) (Figs 1 and S1; Table 3). In a follow-up analysis using the individual-level PGC data, we found that a GRS incorporating the same 18 CRP SNPs used to construct the GRSGWAS was again significantly associated with a lower risk of schizophrenia (OR 0.96 [95% CI 0.94–0.98]; p < 1.72 × 10−6). This signal persisted when we included all SNPs meeting a less stringent p-value threshold of 1 × 10−4 (OR 0.97 [95% CI 0.95–0.99]; p < 2.45 × 10−4). At less stringent p-value thresholds, less variance was explained by the logistic model, and the protective effect of CRP risk scores became less significant, but across all p-value thresholds, the direction of the effect was consistently protective (Figs 2 and 3). To ensure that the association between risk alleles for CRP and schizophrenia was not driven by a small number of genome-wide significant SNPs, we performed a leave-one-out sensitivity analysis of the 18 genome-wide SNPs. In the 18 sets of 17 SNPs, the variance explained (Nagelkerke’s pseudo-R2) ranged from 0.012% to 0.034%, with p-values ranging from 9.3 × 10−5 to 1.6 × 10−2, suggesting that the protective effect observed between risk alleles for CRP and schizophrenia was not driven by a small number of SNPs with large effects.

thumbnail
Fig 1. Genetic risk score GRSGWAS for schizophrenia.

The x-axis shows the effect size for the 15 SNPs for which data were available in the PGC schizophrenia dataset comprising the GRSGWAS influencing levels of CRP, with corresponding standard error bars. The y-axis shows the log OR of the GRSGWAS SNPs for schizophrenia (SCZ) with corresponding standard error bars. The effect estimate of CRP level on disease risk is represented by the red solid line, with gradient α. The 95% CI of this α estimate is represented by the grey dashed lines. The included SNPs are shown by Arabic numbering: #1, rs2847281 (gene: PTPN2; chromosome: 18; basepair position: 12811593); #2, rs340029 (RORA; 15; 58682257); #3, rs6901250 (GPRC6A; 6; 117220718); #4, rs10745954 (ASCL1; 12; 102007224); #5, rs4705952 (IRF1; 5; 131867517); #6, rs12037222 (PABPC4; 1; 39837548); #7, rs12239046 (NLRP3; 1; 245668218); #8, rs6734238 (IL1F10; 2; 113557501); #9, rs13233571 (BCL7B; 7; 72609167); #11, rs1260326 (GCKR; 2; 27584444); #12, rs4129267 (IL6R; 1; 152692888); #13, rs1800961 (HNF4A; 20; 42475778); #14, rs4420065 (LEPR; 1; 5934049); #15, rs10521222 (SALL1; 16; 49716211); 12; 119905190); #17, rs2794520 (CRP; 1; 157945440). The three SNPs of #10, rs9987289 (PPP1R3B; 8; 9220768); #16, rs1183910 (HNF1A; and #18, rs4420638 (APOC1; 19; 50114786) were not present in the data of the PGC.

https://doi.org/10.1371/journal.pmed.1001976.g001

thumbnail
Fig 2. Polygenic risk scores for elevated CRP level and protective effect on schizophrenia, using individual-level genetic data.

https://doi.org/10.1371/journal.pmed.1001976.g002

thumbnail
Fig 3. Polygenic risk scores for elevated CRP level and explained variance of schizophrenia using individual-level genetic data.

https://doi.org/10.1371/journal.pmed.1001976.g003

thumbnail
Table 3. The effect of the CRP genetic risk score instrument of 18 SNPs associated with CRP (GRSGWAS) on somatic and neuropsychiatric outcomes.

https://doi.org/10.1371/journal.pmed.1001976.t003

GRSGWAS also showed moderate but nominally significant effects of lnCRP on the risk of IBD (OR 0.85 [95% CI 0.74–0.98]; p < 0.03), Crohn disease (0.81 [0.70–0.94]; p < 0.005), psoriatic arthritis (1.36 [1.00–1.84]; p < 0.049), knee osteoarthritis (1.17 [1.01–1.36]; p < 0.04), and bipolar disorder (1.21 [1.05–1.40]; p < 0.007), while its effect was statistically significant for CAD (0.88 [0.84–0.94]; p < 2.4 × 10−5) (Table 3; Figs 4 and S1). GRSGWAS revealed a nominally significant effect of lnCRP on blood pressure: an increase of 0.72 (95% CI 0.11–1.34; p < 0.02) and 0.45 (0.06–0.84; p < 0.02) mm Hg in SBP and DBP, respectively (Table 3; S1 Fig). Likewise, a genetically determined 10-s% increase in CRP level was nominally associated with a 0.01 ml/min/1.73 m2 (95% CI 0.003–0.02; p < 0.005) higher estimated glomerular filtration rate from serum creatinine (eGFRcr), a 0.01 g/dl (0.0004–0.02; p < 0.04) higher serum albumin level, and a 0.03 g/dl (0.008–0.05; p < 0.009) higher serum protein level. The remaining outcomes tested for causal associations using GRSGWAS did not reach statistical significance, though the corresponding GRSGWAS proved to be a strong IV, with F-values ≥ 82 (Table 3; S1 Fig).

thumbnail
Fig 4. Genetic risk score GRSGWAS for bipolar disorder.

The x-axis shows the effect size for the 18 SNPs comprising the GRSGWAS influencing levels of CRP, with corresponding standard error bars. The y-axis shows the log OR of the GRSGWAS SNPs for bipolar disorder (BIP) with corresponding standard error bars. The effect estimate of CRP level on disease risk is represented by the red solid line, with gradient α. The 95% CI of this α estimate is represented by the grey dashed lines. The included SNPs are shown by Arabic numbering: #1, rs2847281 (gene: PTPN2; chromosome: 18; basepair position: 12811593); #2, rs340029 (RORA; 15; 58682257); #3, rs6901250 (GPRC6A; 6; 117220718); #4, rs10745954 (ASCL1; 12; 102007224); #5, rs4705952(IRF1; 5; 131867517); #6, rs12037222 (PABPC4; 1; 39837548); #7, rs12239046 (NLRP3; 1; 245668218); #8, rs6734238 (IL1F10; 2; 113557501); #9, rs13233571 (BCL7B; 7; 72609167); #10, rs9987289 (PPP1R3B; 8; 9220768); #11, rs1260326 (GCKR; 2; 27584444); #12, rs4129267 (IL6R; 1; 152692888); #13, rs1800961 (HNF4A; 20; 42475778); #14, rs4420065 (LEPR; 1; 5934049); #15, rs10521222 (SALL1; 16; 49716211); #16, rs1183910 (HNF1A; 12; 119905190); #17, rs2794520 (CRP; 1; 157945440); #18, rs4420638 (APOC1; 19; 50114786).

https://doi.org/10.1371/journal.pmed.1001976.g004

Using GRSGWAS, there was no significant evidence of heterogeneity of the effect size for knee osteoarthritis, bipolar disorder, schizophrenia, or SBP, while the heterogeneity test was statistically significant for psoriatic arthritis, IBD, Crohn disease, CAD, DBP, eGFRcr, serum albumin, and serum protein. These heterogeneities in the effects of GRSGWAS may be attributable to pleiotropic effects of the SNPs used to build the GRSGWAS. We subsequently performed a stepwise removal of SNPs from GRSGWAS until no significant heterogeneity remained (Table 4). This adjustment in the GRSGWAS resulted in the removal of three SNPs from the GRSGWAS for IBD (in GCKR, IRF1, and PTPN2), five SNPs from the GRSGWAS for Crohn disease (in GCKR, IL6R, IRF1, PABPC4, and PTPN2), one SNP from the GRSGWAS for psoriatic arthritis (in IRF1), three SNPs for CAD (in APOC1, HNF1A, and IL6R), one SNP from the GRSGWAS for DBP (in PABPC4), two SNPs from the GRSGWAS for eGFRcr (in LEPR and GCKR), six SNPs from the GRSGWAS for serum albumin level (in APOC1, BCL7B, GCKR, PPP1R3B, PTPN2, and IRF1), and one SNP from the GRSGWAS for serum protein level (in GCKR). After removal of these variants from the GRSGWAS, we found no statistically significant (at p < 0.0016) association between genetically increased lnCRP level and any of these outcomes (Table 4). However, the effect estimate of CRP on DBP, serum albumin, and psoriatic arthritis showed nominal association at p < 0.05. For example, for DBP, 17 SNPs remained in the GRSGWAS and yielded a slightly lower causal estimate (compared to the values before adjustment) of a 0.39 (95% CI −0.01 to 0.78) mm Hg increase in DBP per 10-s% increase in lnCRP level, with a nominal significance of p < 0.05.

thumbnail
Table 4. The effect of the CRP genetic risk score instrument of 18 SNPs associated with CRP (GRSGWAS) on somatic and neuropsychiatric outcomes after correcting for heterogeneity.

https://doi.org/10.1371/journal.pmed.1001976.t004

Likewise, we hypothesized that the fact that GRSGWAS showed a nonsignificant effect of CRP on celiac disease, ulcerative colitis, rheumatoid arthritis, type 1 diabetes, and type 2 diabetes can be to some extent explained by the significant heterogeneity observed for these outcomes (Table 3). The stepwise adjustment in the GRSGWAS resulted in the removal of two SNPs from the GRSGWAS for celiac disease (in PABPC4 and PTPN2), one SNP from the GRSGWAS for ulcerative colitis (in GCKR), five SNPS from the GRSGWAS for rheumatoid arthritis (in HNF4A, IL6R, SALL1, NLRP3, and PTPN2), one SNP from the GRSGWAS for type 1 diabetes (in PTPN2), and one SNP from the GRSGWAS for type 2 diabetes (in APOC1). After adjusting for heterogeneity, the association of GRSGWAS with these outcomes remained statistically nonsignificant (Table 4).

Discussion

In this large-scale cross-consortium MR study of 32 complex outcomes, we found evidence for a potential protective causal relationship between elevated CRP level and schizophrenia in both genetic IVs (i.e., GRSCRP and GRSGWAS) and confirmed this protective relationship in follow-up analyses using individual-level genotype data from the schizophrenia GWAS. We also found a statistically significant association of CRP level with CAD, and nominally significant evidence for a predisposing causal association of CRP level with IBD, Crohn disease, psoriatic arthritis, knee osteoarthritis, SBP, DBP, eGFRcr, serum albumin level, serum protein level, and bipolar disorder, using GRSGWAS as an IV. However, after adjustment for heterogeneity, neither GRS showed a significant effect (at p < 0.0016) of CRP level on any of these outcomes, including CAD, nor on the 20 other common somatic and psychiatric outcomes we investigated, including celiac disease, ulcerative colitis, psoriasis (all types), rheumatoid arthritis, systemic lupus erythematous, systemic sclerosis, type 1 and 2 diabetes, stroke (all types), body mass index, chronic kidney disease, amyotrophic lateral sclerosis, Alzheimer disease, Parkinson disease, autism, and major depressive disorder.

CRP Protection against Schizophrenia

Strikingly, as opposed to the current literature and previous inconclusive small-scale studies [6668], our findings suggest that genetically elevated levels of CRP are not predisposing but in fact protective for schizophrenia. The significant causal protective role of CRP for schizophrenia was consistent in both IVs using summary statistics, i.e., GRSCRP and GRSGWAS. When incorporating 18 genome-wide CRP-associated SNPs using individual-level data, we confirmed a modest, but significant, protective effect of CRP level for schizophrenia. This signal persisted when we included all SNPs meeting a less stringent p-value threshold of 1×10−4. Notably, the leave-one-out sensitivity analysis revealed that the genetic overlap between CRP level and schizophrenia we observed at genome-wide and 1×10−4 significance thresholds was not driven by a few major SNPs. In contrast, others have previously shown that CRP levels are significantly elevated in patients with schizophrenia [69,70], with a recent meta-analysis concluding that the association between elevated CRP and schizophrenia is indeed robust [71]. Given that clinical studies report elevated CRP levels in schizophrenia, one would expect to find that alleles for elevated CRP would confer an increased risk for schizophrenia. The fact that we found the completely opposite effect—in a cohort of over 25,000 cases and 30,000 controls—should give one pause when deriving clinical meaning from these results. Our observation that a genetically determined marginal increase in the level of CRP is likely to be protective for schizophrenia may fuel the debate about whether the observed CRP elevation in schizophrenia is a by-product of the pathogenesis of schizophrenia or directly contributing to clinical manifestations of the disorder [6]. Our finding may also point out potential biases in previous studies regarding the causes of elevated CRP levels in patients with schizophrenia, such as reverse causality and/or pleiotropic effects within chosen instruments.

The exact mechanism for how elevated CRP levels are linked to schizophrenia requires a well-defined experimental analysis. In addition to CRP variants, other recent studies have identified several inflammatory genetic variants associated with schizophrenia and bipolar disorder, which include variants in the major histocompatibility complex (MHC) region on Chromosome 6p21 [72]—harboring many cytokine genes [54,7376]—and in the IL10 promoter [77], TNF promoter [78], IL1B [79], and C4 [80].

Biological Annotation

Following comments made by the reviewers, we explored the possible underlying pathways that may explain the potential protective causal association between CRP and schizophrenia. We performed a follow-up in silico functional pathway analysis using a previously reported approach [81] as summarized in S5 Methods and S4S13 Tables. In brief, our results show that pathways associated with the interferon response are significantly enriched amongst genes harbored by CRP loci and their associated expression quantitative trait loci (eQTLs) and that there are differentially expressed genes between schizophrenia cases and controls. Previous studies showed that the induction of T cell IFN cytokine release stimulates microglia and astrocytes to facilitate glutamate clearance in neuronal cells without evoking inflammatory mediators [82,83]. One could speculate that CRP-interferon pathways may induce neuroprotection by contributing to glutamate clearance, leading to the protection of neurons against the oxidative stress associated with an excess of glutamate [84,85], and thereby offering a protective effect against schizophrenia.

CRP GRSGWAS Association with Bipolar Disorder

As for bipolar disorder, we found a nominal effect of a 1.21-fold increase in risk for bipolar disorder with a 10-s% increase in CRP level. Though this nominal predisposing effect needs to be confirmed, our finding corroborates epidemiological observations suggesting that elevated CRP is associated with the disease and supports a potential causal influence of general inflammation in bipolar disorder [86]. We note that, though it may be biologically sensible, this result failed to pass multiple testing correction. Confirmation by replication in independent cohorts, functional follow-up analyses, or the use of a stronger CRP GRSGWAS in upcoming studies is required to draw a definitive conclusion.

CRP GRSGWAS Association with Blood Pressure and Hypertension

We found nominally significant evidence for an up to ~0.70-mm Hg increase in blood pressure with a 10-s% increase in CRP level and no evidence of heterogeneity for SBP. Additionally, there was nominally borderline significance for a causal association between CRP and DBP after adjustment for heterogeneity. These nominally significant findings, on the one hand, are in line with numerous epidemiological studies that have highlighted an association between elevated CRP and an increased risk of hypertension. For instance, one study found an association between CRP loci and hypertension in Asian individuals [87]. An additional line of support for a possible causal association of CRP and blood pressure comes from an experimental study in which an increase in CRP gene expression in mice, and subsequently CRP protein levels, led to a rise in SBP particularly [88]. Moreover, an ex vivo study by Zhou et al. showed that combining IL6 treatment and mechanical strain leads to a consistent increase in CRP expression at the protein and mRNA levels in smooth muscle cells [89]. Both inflammatory factors and local mechanical strains are abundant in blood vessels and are well-known risk factors for high blood pressure. Our finding did not reach a statistically significant level after correction for multiple testing; thus, it may echo previous MR studies that have failed to find a causal relationship between CRP level and blood pressure or hypertension in Europeans [90,91]. However, our systematic literature review showed that previous studies had some limitations (S1 Table). For instance, no study used a refined GWAS set of 18 CRP-associated SNPs; instead, they tested single or a limited set of CRP SNPs. Using such instruments might have led to biased estimates as their corresponding effects on CRP levels have been found to be small [30,57]. A combination of weak instruments and small sample sizes might have led to type II error [28,57] and hence to a conclusion of no causal association between CRP and blood pressure traits in previous studies. When all of the evidence is taken together, a direct link between CRP and blood pressure remains to be elucidated, though our nominal associations between GRSCRP and GRSGWAS and blood pressure do add to a line of findings from experimental studies suggesting a potential causal relationship between CRP and blood pressure.

CRP GRSGWAS Association with Osteoarthritis

Our nominally significant finding that CRP might be a potential causal factor for knee osteoarthritis (using GRSGWAS) should be interpreted with caution. In line with our findings, we have previously shown that levels of CRP were higher in women with early radiological knee osteoarthritis (i.e., Kellgren-Lawrence grade 2+) and in women whose disease progressed [92]. Additionally, another study showed that genetically elevated CRP levels contribute to osteoarthritis severity [93]. However, other studies have found contrasting results [71,72,94]. One systematic review provided evidence that the relationship between CRP and osteoarthritis does exist but is dependent on body mass index [95]. It remains to be further investigated whether weight gain over the lifetime mediates the potential causal association between genetically elevated CRP and knee osteoarthritis.

CRP GRSGWAS Shows No Association with Other Remaining Outcomes

The present study was able to calculate nominal causal estimates for IBD, Crohn disease, psoriatic arthritis, CAD, eGFRcr, serum albumin level, and serum protein level using CRP GRSGWAS, but the estimates were altered by removal of SNPs from GRSGWAS based on heterogeneity tests, resulting in nominal or nonsignificant associations. These outcomes appeared therefore to have heterogeneity in the causal estimates, suggesting that these observed estimates were biased, likely due to pleiotropic effects of CRP loci. These results corroborate negative findings of previous studies (S1 Table), suggesting that a causal role of CRP in these traits and diseases is unlikely.

Methodological Concerns and Advantages

Pleiotropic biases in Mendelian randomization analyses using CRP GRSGWAS.

A detailed evaluation of pleiotropic SNPs in our study showed that the method applied to identify heterogeneity sources was able to indicate and exclude several already known pleiotropic loci from the GRSGWAS IV. For instance, the use of a SNP in IL6R (rs4129267), amongst others, resulted in heterogeneity of effects on CAD risk. The same variant contributed to heterogeneity of effects for Crohn disease in our study, and it has been shown that this SNP is associated with levels of biomarkers other than CRP [56]. Further, a MR study found that IL6R SNPs, specifically the nonsynonymous SNP rs8192284, are associated with CAD risk and CRP levels [96]. Our selected IL6R SNPs, namely rs4537545 and rs4129267, are in extremely high linkage disequilibrium with rs8192284 (r2 ≥ 0.96 for both SNPs in HapMap data, CEU population). Carriers of the risk allele of rs8192284 have higher CRP, IL6, and fibrinogen levels [96]. Fibrinogen is also a well-known risk factor for CAD. Therefore, it is unclear so far which biomarker(s) mediates the effect of IL6R SNPs on CAD. Besides the IL6 locus, APOC1 and PABPC4 have been indicated as pleiotropic in three out of 32 our investigated outcomes, and PTPN2 and GCKR in six. With this information taken together, we were able to disentangle at least part of the pleiotropy regarding the causal estimates of CRP for outcomes. Again, we found no significant association of CRP GRSGWAS with IBD, Crohn disease, psoriatic arthritis, CAD, eGFRcr, serum albumin level, and serum protein level after adjustment for heterogeneity.

Using summary statistics of large-scale consortia.

It is of utmost interest whether the observed effect of CRP as a risk predictor for human disease is causal, and thus whether reduction of CRP levels will lower the risk of disease. Here, we investigated the causality of CRP in 32 phenotypes by leveraging very large sample sizes collected by GWAS consortia, an approach that was much better powered than most previous MR studies. We found that genetically elevated CRP levels approximated by powerful instruments did not appear to contribute directly to most of the studied somatic and psychiatric outcomes. Our findings are consistent with previous MR studies reporting null associations of genetically elevated CRP levels with inflammation-related outcomes including CAD [56,59,97], type 2 diabetes [98], high body mass index [99], Alzheimer disease, and depression [100]. All previous MR studies were substantially limited to a single or a few outcomes, used only SNPs in the CRP gene, or had sample sizes much smaller than that of the present study (S1 Table). In addition to these studies, the current GWAS data do not corroborate epidemiological observations suggesting that elevated CRP levels are associated with amyotrophic lateral sclerosis [101], Alzheimer disease [102], Parkinson disease [103], and major depressive disorder [104]. Furthermore, patients with immunity-related disorders frequently have a very high CRP level (as high as 100 mg/l) due to their disease status. Our findings may therefore more favorably indicate reverse causality. Taken together, these results show that CRP is highly unlikely to contribute causally to most of the major common somatic and neuropsychiatric outcomes that were investigated in the present study, with the possible exception of schizophrenia.

Strength of instrumental variables.

The results presented in Table 2 show that our GRSCRP is not a weak instrument, as indicated by its high F-values owing to the large sample sizes of available outcomes from GWASs for the phenotypes under study. The strength of our instrument increased considerably in all disease classes when we used variants of multiple loci associated with CRP in GWASs. However, the variants comprising the CRP GRSGWAS explain on average only a moderate ~5% of the total variance in baseline CRP levels [30]. Moreover, the possibility of effect modification by nongenetic CRP-related factors on the outcomes remains to be investigated. We may be able to create even stronger instruments based on ongoing efforts to identify additional variation influencing CRP levels. Even if larger sample sizes and stronger instruments can be realized, the overwhelming lack of causal effects observed for most outcomes in our study implies that therapies targeted at lowering CRP will not directly result in decreased risk of the investigated outcomes, or in better symptom management [105,106].

Using summary statistics instead of individual-level data.

Here we used summary association statistics obtained from previously conducted meta-GWASs in order to maximize our study power. One may argue this may induce bias compared to when one uses individual-level data. Nevertheless, previous studies showed high agreement in results from MR methods using GWAS summary data and individual-level data [60,107]; Furthermore, our analyses of individual-level data for schizophrenia led to the same conclusion as our analyses using summary statistics data, confirming the robustness of our methodological approach.

Other potential sources of bias.

An important rationale for MR is that the gene variants do not change over time and are inherited randomly. Thus, the genetic variants are considered free from confounding and reverse causation [108]. However, one cannot completely control for the possibility of confounding of genotype–intermediate phenotype–disease associations. For instance, there could be a confounding effect by ethnic/racial group (i.e., population stratification), but this is unlikely to be a major problem in most situations [108]. In the present study, we included summary statistics data from highly credible results of meta-GWASs. All the original meta-GWASs corrected for population stratification in cohort-level analyses and at meta-GWAS level.

Another caveat of MR is that developmental compensation might occur, through a genotype being expressed during fetal development that in turn buffers the effects of either environmental or genetic factors, a process called canalization [108,109]. Therefore, buffering mechanisms could hamper the associations between genetic variants and the outcome of interest. As opposed to this, a lifetime exposure to a risk factor may enhance its effects on the disease [109]. However, it is not clear to what extent genetically determined small changes in any given exposure would be sufficient to induce compensation [108].

All 32 of the meta-GWASs from which instrument summary estimates were taken were performed in individuals of European descent in Europe and the US and included thousands of samples for each outcome (S1 Table), which was also the case for our previous CRP meta-GWAS from which we chose the CRP-associated SNPs to calculate GRSGWAS. Therefore, the results of this MR study are applicable to individuals of European descent and are not necessarily generalizable to other ethnic groups.

Conclusion

We showed that elevated CRP levels driven by genetic factors are causally associated with protection against schizophrenia, suggesting that CRP may be one important puzzle piece that leads to an improved understanding of the pathogenesis of schizophrenia. We observed nominal evidence that genetically elevated CRP is causally associated with SBP, DBP, knee osteoarthritis, and bipolar disorder. Based on current GWAS data, we cannot verify any causal effect of CRP on the other 27 common somatic and neuropsychiatric outcomes investigated in the present study. Therefore, disease-associated rise in CRP levels may be a response to the disease process rather than a cause for these 27 outcomes. This implies that interventions to lower CRP levels are unlikely to result in decreased risk for the majority of common complex outcomes.

Supporting Information

S1 Consortia. Consortia coauthors and collaborators.

https://doi.org/10.1371/journal.pmed.1001976.s001

(DOCX)

S1 Data. Individual association summary statistics of CRP lead SNPs and/or proxies with traits and diseases.

https://doi.org/10.1371/journal.pmed.1001976.s002

(XLSX)

S1 Fig. GRSCRP and GRSGWAS for each studied outcome.

https://doi.org/10.1371/journal.pmed.1001976.s003

(DOCX)

S1 Methods. Linkage disequilibrium of the four GRSCRP SNPs.

https://doi.org/10.1371/journal.pmed.1001976.s005

(DOCX)

S2 Methods. CRP GRSGWAS for Alzheimer disease and body mass index.

https://doi.org/10.1371/journal.pmed.1001976.s006

(DOCX)

S4 Methods. CRP polygenic risk score (CRPPRS) for schizophrenia.

https://doi.org/10.1371/journal.pmed.1001976.s008

(DOCX)

S5 Methods. In silico (gene) pathway analyses highlight the role of interferon in the causal pathway between CRP and schizophrenia.

https://doi.org/10.1371/journal.pmed.1001976.s009

(DOCX)

S1 Table. Previous Mendelian randomization analyses using CRP variants as instruments.

https://doi.org/10.1371/journal.pmed.1001976.s010

(XLSX)

S2 Table. CRP lead variants used in the genetic risk scores as instrumental variables.

https://doi.org/10.1371/journal.pmed.1001976.s011

(XLSX)

S3 Table. Proxy SNPs of CRP lead variants used in the genetic risk scores as instrumental variables.

https://doi.org/10.1371/journal.pmed.1001976.s012

(XLSX)

S4 Table. Biologically prioritized candidate gene set associated with CRP used as the input query to the gene set enrichment analysis.

https://doi.org/10.1371/journal.pmed.1001976.s013

(XLSX)

S5 Table. Pathway enrichment results for the biologically prioritized candidate gene set associated with CRP used as the input query to the gene set enrichment analysis.

https://doi.org/10.1371/journal.pmed.1001976.s014

(XLSX)

S6 Table. Genes (n = 144) that were significantly differentially expressed between schizophrenia and unaffected controls in the hippocampus.

https://doi.org/10.1371/journal.pmed.1001976.s015

(XLSX)

S7 Table. Pathway enrichment results for 144 genes that were significantly differentially expressed between schizophrenia and unaffected controls in the hippocampus.

https://doi.org/10.1371/journal.pmed.1001976.s016

(XLSX)

S8 Table. Biologically prioritized candidate gene set of CRP from Vaez et al. [81] (from S5 Table, “CRP genes,” in blue) and differentially expressed genes in schizophrenia cases versus controls from Hwang et al. [110] (from S7 Table, “SCZ expr genes,” in red) used as the input query to the pathway analysis.

https://doi.org/10.1371/journal.pmed.1001976.s017

(XLSX)

S9 Table. Pathway enrichment results for the combined set of the biologically prioritized candidate gene set of CRP from Vaez et al. [81] and differentially expressed genes in schizophrenia cases versus controls from Hwang et al. 2013.

https://doi.org/10.1371/journal.pmed.1001976.s018

(XLSX)

S10 Table. List of genes at 108 genome-wide significant loci associated with schizophrenia.

https://doi.org/10.1371/journal.pmed.1001976.s019

(XLSX)

S11 Table. Brain and blood eQTL for credible sets of SNPs of the 108 schizophrenia loci.

https://doi.org/10.1371/journal.pmed.1001976.s020

(XLSX)

S12 Table. List of genes at 108 genome-wide significant loci associated with schizophrenia (yellow), brain eQTL (red), and blood eQTL (blue).

https://doi.org/10.1371/journal.pmed.1001976.s021

(XLSX)

S13 Table. Pathway enrichment results for the list of genes at 108 genome-wide significant loci associated with schizophrenia (yellow) and those associated with brain eQTL (red) and blood eQTL (blue).

https://doi.org/10.1371/journal.pmed.1001976.s022

(XLSX)

Acknowledgments

We thank for Dan Arking for providing the genome-wide summary statistics data on CRP polymorphisms for the autism GWAS. We would also like to acknowledge the rheumatoid arthritis and systemic lupus erythematosus consortia, the Genetic Investigation of Anthropometric Traits (GIANT) consortium, and the PGC for making publicly available the data from which we extracted summary association statistics for CRP SNPs. The PAGE Consortium is grateful to Alex Tsoi for adding the consortium information and membership list.

The Type 1 Diabetes Genetics Consortium would like to thank all participants and their families, and also thank Smari Kristinsson, Larus Arni Hermannsson, and Asbjörn Krisbjörnsson of Raförninn for their extensive software design and contribution. We would like to thank the people with motor neuron disease and their families for their participation in this project. We acknowledge sample management undertaken by Biobanking Solutions funded by the Medical Research Council at the Centre for Integrated Genomic Medical Research, University of Manchester.

Consortia Coauthors

Please refer to S1 Consortia for the full names and affiliations of coauthors from the following five consortia: the Schizophrenia Working Group of the Psychiatric Genomics Consortium, the Autism Spectrum Disorder Working Group of the Psychiatric Genomics Consortium, the GERAD1 Consortium, the International Consortium for Blood Pressure Genome-Wide Association Studies, and the PAGE Consortium.

Author Contributions

Conceived and designed the experiments: BPP AA HS BZA CVD. Performed the experiments: NF JGA VM PES JPB AMV JB AS. Analyzed the data: BPP AA AV AM IN NF JGA VM PES JPB AMV JB AS AD BH SR SB MDM LT EE RPN SG CP TR DVH MD JN NW DA AAC BK HH HM JE TS CVD JM APM RW JP CW PM JGP YJ HS BZA JK. Contributed reagents/materials/analysis tools: CL AAC AD JK CVD. Wrote the first draft of the manuscript: BPP AA BZA. Contributed to the writing of the manuscript: BPP AA AV AW JGA VM PES AMV DEA BK TS JM RW JP JGP YJ HS BZA JK. Agree with the manuscript’s results and conclusions: BZA BPP AA AW AV IN NF PES JGA VM JPB AMV JB AS CL BH SR SB MDM LT EE RPN SG CP TR DVH MD NW AAC AD HH HM JE JK DA TS BK CVD JM APM RW CW PM JP JGP YJ HS. All authors have read, and confirm that they meet, ICMJE criteria for authorship.

References

  1. 1. Stockinger B. Immunology: cause of death matters. Nature. 2009;458:44–45. pmid:19262664
  2. 2. Baumgart DC, Carding SR. Inflammatory bowel disease: cause and immunobiology. Lancet. 2007;369:1627–1640. pmid:17499605
  3. 3. Bruunsgaard H, Pedersen M, Pedersen BK. Aging and proinflammatory cytokines. Curr Opin Hematol. 2001;8:131–136. pmid:11303144
  4. 4. Cesari M, Penninx BWJH, Newman AB, Kritchevsky SB, Nicklas BJ, Sutton-Tyrrell K, et al. Inflammatory markers and cardiovascular disease (The Health, Aging and Body Composition [Health ABC] Study). Am J Cardiol. 2003;92:522–528. pmid:12943870
  5. 5. Couzin-Frankel J. Inflammation bares a dark side. Science. 2010;330:1621. pmid:21163993
  6. 6. Fan X, Goff DC, Henderson DC. Inflammation and schizophrenia. Expert Rev Neurother. 2007;7:789–796. pmid:17610386
  7. 7. Grimaldi MP, Vasto S, Balistreri CR, di Carlo D, Caruso M, Incalcaterra E, et al. Genetics of inflammation in age-related atherosclerosis: its relevance to pharmacogenomics. Ann N Y Acad Sci. 2007;1100:123–131. pmid:17460170
  8. 8. Hanson DR, Gottesman II. Theories of schizophrenia: a genetic-inflammatory-vascular synthesis. BMC Med Genet. 2005;6:7. pmid:15707482
  9. 9. Johnson TE. Recent results: biomarkers of aging. Exp Gerontol. 2006;41:1243–1246. pmid:17071038
  10. 10. Kiecolt-Glaser JK, Glaser R. Depression and immune function: central pathways to morbidity and mortality. J Psychosom Res. 2002;53:873–876. pmid:12377296
  11. 11. Kiecolt-Glaser JK, McGuire L, Robles TF, Glaser R. Emotions, morbidity, and mortality: new perspectives from psychoneuroimmunology. Annu Rev Psychol. 2002;53:83–107. pmid:11752480
  12. 12. Lynch MA, Mills KHG. Immunology meets neuroscience—opportunities for immune intervention in neurodegenerative diseases. Brain Behav Immun. 2012;26:1–10. pmid:21664452
  13. 13. Meyer U, Schwarz MJ, Müller N. Inflammatory processes in schizophrenia: a promising neuroimmunological target for the treatment of negative/cognitive symptoms and beyond. Pharmacol Ther. 2011;132:96–110. pmid:21704074
  14. 14. Pardo CA, Vargas DL, Zimmerman AW. Immunity, neuroglia and neuroinflammation in autism. Int Rev Psychiatry. 2005;17:485–495. pmid:16401547
  15. 15. Sansoni P, Vescovini R, Fagnoni F, Biasini C, Zanni F, Zanlari L, et al. The immune system in extreme longevity. Exp Gerontol. 2008;43:61–65. pmid:17870272
  16. 16. Hotamisligil GS. Inflammation and metabolic disorders. Nature. 2006;444:860–867. pmid:17167474
  17. 17. Mantovani A, Allavena P, Sica A, Balkwill F. Cancer-related inflammation. Nature. 2008;454:436–444. pmid:18650914
  18. 18. Dantzer R, O’Connor JC, Freund GG, Johnson RW, Kelley KW. From inflammation to sickness and depression: when the immune system subjugates the brain. Nat Rev Neurosci. 2008;9:46–56. pmid:18073775
  19. 19. Emerging Risk Factors Collaboration, Kaptoge S, Di Angelantonio E, Lowe G, Pepys MB, Thompson SG, et al. C-reactive protein concentration and risk of coronary heart disease, stroke, and mortality: an individual participant meta-analysis. Lancet. 2010;375:132–140. pmid:20031199
  20. 20. Emerging Risk Factors Collaboration, Kaptoge S, Di Angelantonio E, Pennells L, Wood AM, White IR, et al. C-reactive protein, fibrinogen, and cardiovascular disease prediction. N Engl J Med. 2012;367:1310–1320. pmid:23034020
  21. 21. Wang X, Bao W, Liu J, Ouyang Y-Y, Wang D, Rong S, et al. Inflammatory markers and risk of type 2 diabetes: a systematic review and meta-analysis. Diabetes Care. 2013;36:166–175. pmid:23264288
  22. 22. Henriksen M, Jahnsen J, Lygren I, Stray N, Sauar J, Vatn MH, et al. C-reactive protein: a predictive factor and marker of inflammation in inflammatory bowel disease. Results from a prospective population-based study. Gut. 2008;57:1518–1523. pmid:18566104
  23. 23. Rhodes B, Merriman ME, Harrison A, Nissen MJ, Smith M, Stamp L, et al. A genetic association study of serum acute-phase C-reactive protein levels in rheumatoid arthritis: implications for clinical interpretation. PLoS Med. 2010;7:e1000341. pmid:20877716
  24. 24. Ridker PM. High-sensitivity C-reactive protein as a predictor of all-cause mortality: implications for research and patient care. Clin Chem. 2008;54:234–237. pmid:18223130
  25. 25. Dehghan A, Kardys I, de Maat MPM, Uitterlinden AG, Sijbrands EJG, Bootsma AH, et al. Genetic variation, C-reactive protein levels, and incidence of diabetes. Diabetes. 2007;56:872–878. pmid:17327459
  26. 26. Lee CC, Adler AI, Sandhu MS, Sharp SJ, Forouhi NG, Erqou S, et al. Association of C-reactive protein with type 2 diabetes: prospective analysis and meta-analysis. Diabetologia. 2009;52:1040–1047. pmid:19326095
  27. 27. Danesh J, Pepys MB. C-reactive protein and coronary disease: is there a causal link? Circulation. 2009;120:2036–2039. pmid:19901186
  28. 28. Lawlor DA, Harbord RM, Sterne JAC, Timpson N, Davey Smith G. Mendelian randomization: using genes as instruments for making causal inferences in epidemiology. Stat Med. 2008;27:1133–1163. pmid:17886233
  29. 29. C Reactive Protein Coronary Heart Disease Genetics Collaboration (CCGC), Wensley F, Gao P, Burgess S, Kaptoge S, Di Angelantonio E, et al. Association between C reactive protein and coronary heart disease: Mendelian randomisation analysis based on individual participant data. BMJ. 2011;342:d548. pmid:21325005
  30. 30. Dehghan A, Dupuis J, Barbalic M, Bis JC, Eiriksdottir G, Lu C, et al. Meta-analysis of genome-wide association studies in >80 000 subjects identifies multiple loci for C-reactive protein levels. Circulation. 2011;123:731–738. pmid:21300955
  31. 31. Dubois PCA, Trynka G, Franke L, Hunt KA, Romanos J, Curtotti A, et al. Multiple common variants for celiac disease influencing immune gene expression. Nat Genet. 2010;42:295–302. pmid:20190752
  32. 32. Franke A, McGovern DPB, Barrett JC, Wang K, Radford-Smith GL, Ahmad T, et al. Genome-wide meta-analysis increases to 71 the number of confirmed Crohn’s disease susceptibility loci. Nat Genet. 2010;42:1118–1125. pmid:21102463
  33. 33. Anderson CA, Boucher G, Lees CW, Franke A, D’Amato M, Taylor KD, et al. Meta-analysis identifies 29 additional ulcerative colitis risk loci, increasing the number of confirmed associations to 47. Nat Genet. 2011;43:246–252. pmid:21297633
  34. 34. Nair RP, Duffin KC, Helms C, Ding J, Stuart PE, Goldgar D, et al. Genome-wide scan reveals association of psoriasis with IL-23 and NF-kappaB pathways. Nat Genet. 2009;41:199–204. pmid:19169254
  35. 35. Ellinghaus E, Ellinghaus D, Stuart PE, Nair RP, Debrus S, Raelson JV, et al. Genome-wide association study identifies a psoriasis susceptibility locus at TRAF3IP2. Nat Genet. 2010;42:991–995. pmid:20953188
  36. 36. Stahl EA, Raychaudhuri S, Remmers EF, Xie G, Eyre S, Thomson BP, et al. Genome-wide association study meta-analysis identifies seven new rheumatoid arthritis risk loci. Nat Genet. 2010;42:508–514. pmid:20453842
  37. 37. Hom G, Graham RR, Modrek B, Taylor KE, Ortmann W, Garnier S, et al. Association of systemic lupus erythematosus with C8orf13-BLK and ITGAM-ITGAX. N Engl J Med. 2008;358:900–909. pmid:18204098
  38. 38. Radstake TRDJ, Gorlova O, Rueda B, Martin J-E, Alizadeh BZ, Palomino-Morales R, et al. Genome-wide association study of systemic sclerosis identifies CD247 as a new susceptibility locus. Nat Genet. 2010;42:426–429. pmid:20383147
  39. 39. Bradfield JP, Qu H-Q, Wang K, Zhang H, Sleiman PM, Kim CE, et al. A genome-wide meta-analysis of six type 1 diabetes cohorts identifies multiple associated loci. PLoS Genet. 2011;7:e1002293. pmid:21980299
  40. 40. Ramos YFM, Metrustry S, Arden N, Bay-Jensen AC, Beekman M, de Craen AJM, et al. Meta-analysis identifies loci affecting levels of the potential osteoarthritis biomarkers sCOMP and uCTX-II with genome wide significance. J Med Genet. 2014;51:596–604. pmid:25057126
  41. 41. Nikpey M, Goel A, Won H, Hall LM, Willenborg C, Kanoni S, et al. A comprehensive 1,000 Genomes–based genome-wide association meta-analysis of coronary artery disease. Nat Genet. 2015;47:1121–1130. pmid:26343387
  42. 42. International Consortium for Blood Pressure Genome-Wide Association Studies, Ehret GB, Munroe PB, Rice KM, Bochud M, Johnson AD, et al. Genetic variants in novel pathways influence blood pressure and cardiovascular disease risk. Nature. 2011;478:103–109. pmid:21909115
  43. 43. International Stroke Genetics Consortium (ISGC), Wellcome Trust Case Control Consortium 2 (WTCCC2), Bellenguez C, Bevan S, Gschwendtner A, Spencer CCA, et al. Genome-wide association study identifies a variant in HDAC9 associated with large vessel ischemic stroke. Nat Genet. 2012;44:328–333. pmid:22306652
  44. 44. Speliotes EK, Willer CJ, Berndt SI, Monda KL, Thorleifsson G, Jackson AU, et al. Association analyses of 249,796 individuals reveal 18 new loci associated with body mass index. Nat Genet. 2010;42:937–948. pmid:20935630
  45. 45. Voight BF, Scott LJ, Steinthorsdottir V, Morris AP, Dina C, Welch RP, et al. Twelve type 2 diabetes susceptibility loci identified through large-scale association analysis. Nat Genet. 2010;42:579–589. pmid:20581827
  46. 46. Köttgen A, Pattaro C, Böger CA, Fuchsberger C, Olden M, Glazer NL, et al. New loci associated with kidney function and chronic kidney disease. Nat Genet. 2010;42:376–384. pmid:20383146
  47. 47. Franceschini N, van Rooij FJA, Prins BP, Feitosa MF, Karakas M, Eckfeldt JH, et al. Discovery and fine mapping of serum protein loci through transethnic meta-analysis. Am J Hum Genet. 2012;91:744–753. pmid:23022100
  48. 48. Shatunov A, Mok K, Newhouse S, Weale ME, Smith B, Vance C, et al. Chromosome 9p21 in sporadic amyotrophic lateral sclerosis in the UK and seven other countries: a genome-wide association study. Lancet Neurol. 2010;9:986–994. pmid:20801717
  49. 49. Hollingworth P, Harold D, Sims R, Gerrish A, Lambert J-C, Carrasquillo MM, et al. Common variants at ABCA7, MS4A6A/MS4A4E, EPHA1, CD33 and CD2AP are associated with Alzheimer’s disease. Nat Genet. 2011;43:429–435. pmid:21460840
  50. 50. International Parkinson Disease Genomics Consortium, Nalls MA, Plagnol V, Hernandez DG, Sharma M, Sheerin U-M, et al. Imputation of sequence variants for identification of genetic risks for Parkinson’s disease: a meta-analysis of genome-wide association studies. Lancet. 2011;377:641–649. pmid:21292315
  51. 51. Weiss LA, Arking DE, Gene Discovery Project of Johns Hopkins & the Autism Consortium, Daly MJ, Chakravarti A. A genome-wide linkage and association scan reveals novel loci for autism. Nature. 2009;461:802–808. pmid:19812673
  52. 52. Psychiatric GWAS Consortium Bipolar Disorder Working Group. Large-scale genome-wide association analysis of bipolar disorder identifies a new susceptibility locus near ODZ4. Nat Genet. 2011;43:977–983. pmid:21926972
  53. 53. Major Depressive Disorder Working Group of the Psychiatric GWAS Consortium, Ripke S, Wray NR, Lewis CM, Hamilton SP, Weissman MM, et al. A mega-analysis of genome-wide association studies for major depressive disorder. Mol Psychiatry. 2013;18:497–511. pmid:22472876
  54. 54. Schizophrenia Working Group of the Psychiatric Genomics Consortium. Biological insights from 108 schizophrenia-associated genetic loci. Nature. 2014;511:421–427. pmid:25056061
  55. 55. C Reactive Protein Coronary Heart Disease Genetics Collaboration (CCGC), Danesh J, Hingorani A, Wensley F, Casas JP, Smeeth L, Samani NJ, et. al. Collaborative pooled analysis of data on C-reactive protein gene variants and coronary disease: judging causality by Mendelian randomisation. Eur J Epidemiol. 2008;23:531–540. pmid:18425592
  56. 56. Elliott P, Chambers JC, Zhang W, Clarke R, Hopewell JC, Peden JF, et al. Genetic loci associated with C-reactive protein levels and risk of coronary heart disease. JAMA. 2009;302:37–48. pmid:19567438
  57. 57. Burgess S, Butterworth A, Thompson SG. Mendelian randomization analysis with multiple genetic variants using summarized data. Genet Epidemiol. 2013;37:658–665. pmid:24114802
  58. 58. Dastani Z, Hivert M-F, Timpson N, Perry JRB, Yuan X, Scott RA, et al. Novel loci for adiponectin levels and their influence on type 2 diabetes and metabolic traits: a multi-ethnic meta-analysis of 45,891 individuals. PLoS Genet. 2012;8:e1002607. pmid:22479202
  59. 59. Zacho J, Tybjaerg-Hansen A, Jensen JS, Grande P, Sillesen H, Nordestgaard BG. Genetically elevated C-reactive protein and ischemic vascular disease. N Engl J Med. 2008;359:1897–1908. pmid:18971492
  60. 60. Voight BF, Peloso GM, Orho-Melander M, Frikke-Schmidt R, Barbalic M, Jensen MK, et al. Plasma HDL cholesterol and risk of myocardial infarction: a Mendelian randomisation study. Lancet. 2012;380:572–580. pmid:22607825
  61. 61. Cole TJ. Sympercents: symmetric percentage differences on the 100 log(e) scale simplify the presentation of log transformed data. Stat Med. 2000;19:3109–3125. pmid:11113946
  62. 62. Rice JA. Mathematical statistics and data analysis. 2nd ed. Belmont (California): Duxbury Press; 1995.
  63. 63. Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MA, Bender D, et al. PLINK: a tool set for whole-genome association and population-based linkage analyses. Am J Hum Genet. 2007;81:559–575. pmid:17701901
  64. 64. Cross-Disorder Group of the Psychiatric Genomics Consortium. Identification of risk loci with shared effects on five major psychiatric disorders: a genome-wide analysis. Lancet. 2013;381:1371–1379. pmid:23453885
  65. 65. Power RA, Steinberg S, Bjornsdottir G, Rietveld CA, Abdellaoui A, Nivard MM, et al. Polygenic risk scores for schizophrenia and bipolar disorder predict creativity. Nat Neurosci. 2015;18:953–955. pmid:26053403
  66. 66. Zakharyan R, Chavushyan A, Khoyetsyan A, Stahelova A, Arakelyan A, Boyajyan A, et al. Genetic variants of the inflammatory C-reactive protein and schizophrenia in Armenian population: a pilot study. Int J Immunogenet. 2010;37:407–410. pmid:21182750
  67. 67. Singh B, Chaudhuri TK. Role of C-reactive protein in schizophrenia: an overview. Psychiatry Res. 2014;216:277–285. pmid:24565000
  68. 68. Wium-Andersen MK, Ørsted DD, Nordestgaard BG. Elevated C-reactive protein associated with late- and very-late-onset schizophrenia in the general population: a prospective study. Schizophr Bull. 2014;40:1117–1127. pmid:23996346
  69. 69. Miller BJ, Culpepper N, Rapaport MH. C-reactive protein levels in schizophrenia: a review and meta-analysis. Clin Schizophr Relat Psychoses. 2014;7:223–230. pmid:23428789
  70. 70. Dickerson F, Stallings C, Origoni A, Vaughan C, Khushalani S, Yang S, et al. C-reactive protein is elevated in schizophrenia. Schizophr Res. 2013;143:198–202. pmid:23218564
  71. 71. Fernandes BS, Steiner J, Bernstein H-G, Dodd S, Pasco JA, Dean OM, et al. C-reactive protein is increased in schizophrenia but is not altered by antipsychotics: meta-analysis and implications. Mol Psychiatry. 2016;21:554–564. pmid:26169974
  72. 72. Lewis CM, Levinson DF, Wise LH, DeLisi LE, Straub RE, Hovatta I, et al. Genome scan meta-analysis of schizophrenia and bipolar disorder, part II: schizophrenia. Am J Hum Genet. 2003;73:34–48. pmid:12802786
  73. 73. International Schizophrenia Consortium, Purcell SM, Wray NR, Stone JL, Visscher PM, O’Donovan MC, et al. Common polygenic variation contributes to risk of schizophrenia and bipolar disorder. Nature. 2009;460:748–752. pmid:19571811
  74. 74. Shi J, Levinson DF, Duan J, Sanders AR, Zheng Y, Pe’er I, et al. Common variants on chromosome 6p22.1 are associated with schizophrenia. Nature. 2009;460:753–757. pmid:19571809
  75. 75. Stefansson H, Ophoff RA, Steinberg S, Andreassen OA, Cichon S, Rujescu D, et al. Common variants conferring risk of schizophrenia. Nature. 2009;460:744–747. pmid:19571808
  76. 76. Schizophrenia Psychiatric Genome-Wide Association Study (GWAS) Consortium. Genome-wide association study identifies five new schizophrenia loci. Nat Genet. 2011;43:969–976. pmid:21926974
  77. 77. Al-Asmari SM, Kadasah S, Arfin M, Tariq M, Al-Asmari A. Genetic variants of interleukin-10 gene promoter are associated with schizophrenia in Saudi patients: A case-control study. N Am J Med Sci. 2014;6:558–565. pmid:25535603
  78. 78. Saviouk V, Chow EWC, Bassett AS, Brzustowicz LM. Tumor necrosis factor promoter haplotype associated with schizophrenia reveals a linked locus on 1q44. Mol Psychiatry. 2005;10:375–383. pmid:15340354
  79. 79. Hänninen K, Katila H, Saarela M, Rontu R, Mattila KM, Fan M, et al. Interleukin-1 beta gene polymorphism and its interactions with neuregulin-1 gene polymorphism are associated with schizophrenia. Eur Arch Psychiatry Clin Neurosci. 2008;258:10–15. pmid:17901998
  80. 80. Sekar A. A natural allelic series of complex structural variants and its influence on the risk of lupus and schizophrenia. 2014 [cited 19 May 2016]. Digital Access to Scholarship at Harvard. Available: http://dash.harvard.edu/handle/1/13070061.
  81. 81. Vaez A, Jansen R, Prins BP, Hottenga J-J, de Geus EJC, Boomsma DI, et al. In silico post genome-wide association studies analysis of C-reactive protein loci suggests an important role for interferons. Circ Cardiovasc Genet. 2015;8:487–497. pmid:25752597
  82. 82. Shaked I, Tchoresh D, Gersner R, Meiri G, Mordechai S, Xiao X, et al. Protective autoimmunity: interferon-gamma enables microglia to remove glutamate without evoking inflammatory mediators. J Neurochem. 2005;92:997–1009. pmid:15715651
  83. 83. Garg SK, Banerjee R, Kipnis J. Neuroprotective immunity: T cell-derived glutamate endows astrocytes with a neuroprotective phenotype. J Immunol. 2008;180:3866–3873. pmid:18322194
  84. 84. Javitt DC. Glutamatergic theories of schizophrenia. Isr J Psychiatry Relat Sci. 2010;47:4–16. pmid:20686195
  85. 85. Marsman A, van den Heuvel MP, Klomp DWJ, Kahn RS, Luijten PR, Hulshoff Pol HE. Glutamate in schizophrenia: a focused review and meta-analysis of 1H-MRS studies. Schizophr Bull. 2013;39:120–129. pmid:21746807
  86. 86. Chung K-H, Huang S-H, Wu J-Y, Chen P-H, Hsu J-L, Tsai S-Y. The link between high-sensitivity C-reactive protein and orbitofrontal cortex in euthymic bipolar disorder. Neuropsychobiology. 2013;68:168–173. pmid:24051690
  87. 87. Hong EP, Kim DH, Suh JG, Park JW. Genetic risk assessment for cardiovascular disease with seven genes associated with plasma C-reactive protein concentrations in Asian populations. Hypertens Res. 2014;37:692–698. pmid:24671014
  88. 88. Vongpatanasin W, Thomas GD, Schwartz R, Cassis LA, Osborne-Lawrence S, Hahner L, et al. C-reactive protein causes downregulation of vascular angiotensin subtype 2 receptors and systolic hypertension in mice. Circulation. 2007;115:1020–1028. pmid:17283257
  89. 89. Zhou H, Li Y, Huang G, Gu X, Zeng J, Li Y, et al. Interleukin 6 augments mechanical strain-induced C-reactive protein synthesis via the stretch-activated channel-nuclear factor κ B signal pathway. Heart. 2013;99:570–576. pmid:23257175
  90. 90. Timpson NJ, Lawlor DA, Harbord RM, Gaunt TR, Day INM, Palmer LJ, et al. C-reactive protein and its role in metabolic syndrome: Mendelian randomisation study. Lancet. 2005;366:1954–1959. pmid:16325697
  91. 91. Davey Smith G, Lawlor DA, Harbord R, Timpson N, Rumley A, Lowe GDO, et al. Association of C-reactive protein with blood pressure and hypertension: life course confounding and Mendelian randomization tests of causality. Arterioscler Thromb Vasc Biol. 2005;25:1051–1056. pmid:15731495
  92. 92. Spector TD, Hart DJ, Nandra D, Doyle DV, Mackillop N, Gallimore JR, et al. Low-level increases in serum C-reactive protein are present in early osteoarthritis of the knee and predict progressive disease. Arthritis Rheum. 1997;40:723–727. pmid:9125256
  93. 93. Bos SD, Suchiman HED, Kloppenburg M, Houwing-Duistermaat JJ, le Graverand MPH, Seymour AB, et al. Allelic variation at the C-reactive protein gene associates to both hand osteoarthritis severity and serum high sensitive C-reactive protein levels in the GARP study. Ann Rheum Dis. 2008;67:877–879. pmid:18055473
  94. 94. Vlad SC, Neogi T, Aliabadi P, Fontes JDT, Felson DT. No association between markers of inflammation and osteoarthritis of the hands and knees. J Rheumatol. 2011;38:1665–1670. pmid:21572158
  95. 95. Kerkhof HJM, Bierma-Zeinstra SMA, Castano-Betancourt MC, de Maat MP, Hofman A, Pols HAP, et al. Serum C reactive protein levels and genetic variation in the CRP gene are not associated with the prevalence, incidence or progression of osteoarthritis independent of body mass index. Ann Rheum Dis. 2010;69:1976–1982. pmid:20511616
  96. 96. Interleukin-6 Receptor Mendelian Randomisation Analysis (IL6R MR) Consortium. The interleukin-6 receptor as a target for prevention of coronary heart disease: a Mendelian randomisation analysis. Lancet. 2012;379:1214–1224. pmid:22421340
  97. 97. Anand SS, Yusuf S. C-reactive protein is a bystander of cardiovascular disease. Eur Heart J. 2010;31:2092–2096. pmid:20675658
  98. 98. Brunner EJ, Kivimäki M, Witte DR, Lawlor DA, Davey Smith G, Cooper JA, et al. Inflammation, insulin resistance, and diabetes—Mendelian randomization using CRP haplotypes points upstream. PLoS Med. 2008;5:e155. pmid:18700811
  99. 99. Timpson NJ, Nordestgaard BG, Harbord RM, Zacho J, Frayling TM, Tybjærg-Hansen A, et al. C-reactive protein levels and body mass index: elucidating direction of causation through reciprocal Mendelian randomization. Int J Obes 2005. 2011;35:300–308.
  100. 100. Wium-Andersen MK, Orsted DD, Nordestgaard BG. Elevated C-reactive protein, depression, somatic diseases, and all-cause mortality: a Mendelian randomization study. Biol Psychiatry. 2014;76:249–257. pmid:24246360
  101. 101. Ryberg H, An J, Darko S, Lustgarten JL, Jaffa M, Gopalakrishnan V, et al. Discovery and verification of amyotrophic lateral sclerosis biomarkers by proteomics. Muscle Nerve. 2010;42:104–111. pmid:20583124
  102. 102. Kok EH, Alanne-Kinnunen M, Isotalo K, Luoto T, Haikonen S, Goebeler S, et al. CRP gene variation affects early development of Alzheimer’s disease-related plaques. J Neuroinflammation. 2011;8:96. pmid:21831326
  103. 103. Song I-U, Chung S-W, Kim J-S, Lee K-S. Association between high-sensitivity C-reactive protein and risk of early idiopathic Parkinson’s disease. Neurol Sci. 2011;32:31–34. pmid:20532580
  104. 104. Pasco JA, Nicholson GC, Williams LJ, Jacka FN, Henry MJ, Kotowicz MA, et al. Association of high-sensitivity C-reactive protein with de novo major depression. Br J Psychiatry. 2010;197:372–377. pmid:21037214
  105. 105. Prasad K. C-reactive protein (CRP)-lowering agents. Cardiovasc Drug Rev. 2006;24:33–50. pmid:16939632
  106. 106. Wray NR, Yang J, Hayes BJ, Price AL, Goddard ME, Visscher PM. Pitfalls of predicting complex traits from SNPs. Nat Rev Genet. 2013;14:507–515. pmid:23774735
  107. 107. Vimaleswaran KS, Berry DJ, Lu C, Tikkanen E, Pilz S, Hiraki LT, et al. Causal relationship between obesity and vitamin D status: bi-directional Mendelian randomization analysis of multiple cohorts. PLoS Med. 2013;10:e1001383. pmid:23393431
  108. 108. Smith GD, Ebrahim S. Mendelian randomization: prospects, potentials, and limitations. Int J Epidemiol. 2004;33:30–42. pmid:15075143
  109. 109. Jansen H, Samani NJ, Schunkert H. Mendelian randomization studies in coronary artery disease. Eur Heart J. 2014;35:1917–1924. pmid:24917639
  110. 110. Hwang Y, Kim J, Shin JY, Kim JI, Seo JS, Webster MJ, et al. Gene expression profiling by mRNA sequencing reveals increased expression of immune/inflammation-related genes in the hippocampus of individuals with schizophrenia. Transl Psychiatry. 2013;3:e321. pmid:24169640