A significant causal association between C-reactive protein levels and schizophrenia.

Many observational studies have shown elevated blood CRP levels in schizophrenia compared with controls, and one population-based prospective study has reported that elevated plasma CRP levels were associated with late- and very-late-onset schizophrenia. Furthermore, several clinical studies have reported the efficacy of anti-inflammatory drugs on the symptoms in patients with schizophrenia. However, whether elevated CRP levels are causally related to schizophrenia is not still established because of confounding factors and reverse causality. In the present study, we demonstrated that serum CRP levels were significantly higher in patients with schizophrenia than in the controls by conducting a case-control study and a meta-analysis of case-control studies between schizophrenia and serum CRP levels. Furthermore, we provided evidence for a causal association between elevated CRP levels and increased schizophrenia risk by conducting a Mendelian randomization analysis. Our findings suggest that elevated CRP itself may be a causal risk factor for schizophrenia.

Schizophrenia is a devastating psychiatric disorder with a median lifetime prevalence rate of 0.7–0.8%1. It is also a complex disorder that results from genetic and environmental etiological influences2. Previous studies suggest that inflammation and immune system dysfunctions may play an important role in the pathogenesis of schizophrenia345.

C-reactive protein (CRP) is an acute-phase reactant whose levels increase in response to proinflammatory cytokines and other endogenous signals of innate immunity or tissue damage6. Many studies have been conducted to evaluate the association between CRP levels in the blood789101112131415161718192021222324252627, including plasma and serum CRP, and recent two meta-analyses have demonstrated elevated CRP levels in schizophrenia compared with controls2829. However, these previous association studies have been conducted without consideration of the effects of genetic variants, even though recent meta-analyses of genome-wide association studies of serum hs-CRP levels have identified several genetic variants3031. Furthermore, the results from cross-sectional studies still do not clarify whether elevated blood CRP levels are causally related to schizophrenia because of known and unknown confounding factors and reverse causality.

Mendelian randomization analysis, which uses genetic variants as instrumental variables for exposures of interest, is a useful method for assessing causal relationships between risk factor and outcome32333435363738. Mendelian randomization refers to the random allocation of alleles at the time of gamete formation. A specific genotype carried by a person results from 2 such randomized transmissions, one from the paternally inherited allele and the other from the maternally inherited allele. As a consequence of these randomizations, genotypes are not expected to be associated with known or unknown confounders for any outcome of interest, except those lying on the causal pathway between the genotype and the outcome. By combining the results of the genotype-risk factor association (in this case, genotype-CRP levels) and the genotype-outcome association (in this case, genotype-schizophrenia), one can produce an estimate of the risk factor-outcome association (in this case, CRP levels-schizophrenia) that is free from confounding factors34.

In the present study, we tested the hypothesis that CRP itself may play a causal role in the development of schizophrenia. For this purpose, we first investigated whether serum CRP levels were higher in patients with schizophrenia than in non-psychiatric controls by conducting an analysis of covariance (ANCOVA) on a large Japanese cohort (N = 1,337). Second, we conducted a meta-analysis of case-control studies between serum CRP levels and schizophrenia (N = 4,826). Finally, we investigated the causal relationship between CRP levels and schizophrenia with a Mendelian randomization approach in the Japanese population and in the world-wide population using 2 single-nucleotide polymorphisms (SNPs) (rs2794520 and rs1183910) as instrumental variables.

Materials and Methods

Association study between serum CRP levels and schizophrenia

Subjects

Four hundred and eighteen patients with schizophrenia (241 men, mean age: 62.5 ± 11.5 y; 177 women, mean age: 62.6 ± 11.5 y) were recruited from Tokushima University Hospital in Japan. The diagnosis of schizophrenia was made according to DSM-IV criteria by at least 2 expert psychiatrists on the basis of extensive clinical interviews and a review of medical records. These patients were free of cardiovascular diseases, cancers, and past history of inflammatory diseases according to a review of medical records. All patients were being treated with various antipsychotic drugs. The mean chlorpromazine equivalent dose was 616.8 ± 551.4 mg/day. One thousand, three hundred sixty-five control subjects (422 men, mean age: 41.6 ± 12 y; 943 women, mean age: 41.6 ± 12 y) were selected from volunteers who were recruited from hospital staff, students, and company employees, and were documented to be free from psychiatric problems and past histories of mental illness. All subjects who participated in this study were of Japanese origin. All subjects signed written informed consent forms. This study was conducted in accordance with the World Medical Association’s Declaration of Helsinki and approved by the institutional ethics committee of the University of Tokushima Graduate School.

Serum CRP measurements

Blood was collected at the time of the study’s visit. CRP measurements were performed by a Nephelometric Immuometric assay with a measurement sensitivity >0.02 mg/dl. The same CRP assays were used for all subjects.

SNP selection and genotyping

Among the 7 serum CRP-associated SNPs from a meta-analysis of genome-wide association studies (GWAS) comprising 7,514 individuals in Asian populations30 and the 15 serum CRP-associated SNPs from a meta-analysis of GWAS comprising 82,725 individuals of European ancestry31, which reached genome-wide significance (p < 5.0 × 10−8), 2 common SNPs between the 2 meta-analyses (rs2794520 in the CRP gene and rs1183910 in the HNF1 homeobox A (HNF1A) gene) were selected. Genotyping was performed using commercially available TaqMan probes with the Applied Biosystems 7500 Fast Real Time PCR System, according to the protocol recommended by the manufacturer (Applied Biosystems, California, U.S.A.).

Statistical methods

An ANCOVA was performed to examine the presence of the differences in the natural log-transformed serum CRP levels between the 2 groups (schizophrenia versus control) separately by genotypes (rs2794520 and rs1183910) after adjusting for age and gender.

Meta-analysis of case-control studies between serum CRP levels and schizophrenia

Study selection

Eligible studies were identified using the PubMed search engine with the terms “CRP”, “C-reactive protein”, and “schizophrenia.” We also conducted an additional manual search of reference lists and review articles. Studies meeting the following criteria were included for further meta-analysis: (1) included laboratory assessment of serum CRP levels, (2) performed a case-control study (schizophrenia versus control), and (3) published in the English language. The 2 reviewers (Inoshita M and Umehara H) selected the articles independently according to the above inclusion criteria, and then discussed the articles until they reached a consensus on every study used for the meta-analysis. The scheme of the study selection using a preferred reporting items for systematic review and meta-analysis (PRISMA) checklist39 for this meta-analysis is shown in Supplementary Fig. 1.

A meta-analysis of case-control studies between serum CRP levels and schizophrenia was performed on the standardized mean differences (SMD) as we did in our previous study40. Heterogeneity was assessed using the I2 statistic. If heterogeneity across studies was found, then a random-effects model was applied; otherwise, a fixed-effects model was applied. Publication bias was assessed using funnel plots and a regression test41. Odds ratio (OR) and 95% confidence intervals (CI) were calculated by ‘metafor’, an R package.

Mendelian randomization analysis

To evaluate the causal role of CRP levels in schizophrenia, we conducted a Mendelian randomization analysis by using the estimate for gene-serum CRP association (a beta coefficient value) and the estimate for gene-schizophrenia association (OR). Graphical representation of our Mendelian randomization approach is shown in Supplementary Fig. 2.

The estimate for gene-serum CRP association

We examined the association between natural log-transformed serum CRP levels and each SNP (rs2794520 and rs1183910) in the non-psychiatric Japanese control subjects (N = 932) using a multiple linear regression analysis with adjustments for age and sex, and then calculated a beta coefficient value for each SNP, which represents the number of standard deviation (SD) differences in the natural log-transformed serum CRP levels per the C allele of rs2794520 or the G allele of rs1183910, as done in previous studies3031. The estimates of their effects on CRP levels were also obtained from a meta-analysis of GWAS comprising 82,725 individuals31.

The estimate for gene-schizophrenia association

The risk estimate for the gene-schizophrenia association of the C allele of rs2794520 and the G allele of rs1183910 was evaluated by conducting a meta-analysis of 4 Japanese case-control studies (2,593 cases and 4,247 controls). Genotyping was performed using commercially available TaqMan probes with the Applied Biosystems 7500 Fast Real Time PCR System, according to the protocol recommended by the manufacturer (Applied Biosystems, California, U.S.A.). The genotyping data in the Fujita Heath University set was obtained from a previous Japanese GWAS conducted by the Affymetrix Genome-Wide Human SNP Array 5.042. All subjects who participated in this study were of Japanese origin. All subjects signed written informed consent forms. This study was conducted in accordance with the World Medical Association’s Declaration of Helsinki and approved by the institutional ethics committees of the University of Tokushima Graduate School, Osaka University, Osaka Medical College, and Fujita Health University. Heterogeneity was assessed using the I2 statistic. If heterogeneity across studies was found, then a random-effects model was applied; otherwise, a fixed-effects model was applied. Publication bias was assessed using funnel plots and a regression test. OR and 95% confidence intervals (CI) were calculated by ‘metafor’, an R package. The risk estimate for the gene-schizophrenia association of each SNP was also obtained from a meta-analysis of GWAS comprising 36,989 cases and 113,075 controls43.

The estimate of the effect of CRP on the risk of schizophrenia

We calculated a Mendelian randomization estimate of the effect of the CRP levels on the risk of schizophrenia using rs2794520 and rs1183910 as the instrumental variables, representing an ORscz/crp. As done in previous studies4044, the estimate was calculated for each of two instruments separately. Then these two estimates were combined using an inverse-variance-weight-fixed meta-analysis. The equation we used is log ORscz/crp = (log ORscz/per allele)/betacrp/per allele, where log ORscz/crp is the (log) increase of schizophrenia risk by 1-SD in the natural log-transformed CRP (CRP-schizophrenia association); log ORscz/per allele is the (log) increase in schizophrenia risk per allele (gene-schizophrenia association); and betacrp/per allele is the number of SD differences in the natural log-transformed CRP levels per allele (SD/allele) (gene-CRP association).

Results

Differences in the serum CRP levels between patients with schizophrenia and controls

Among 418 patients, 10 patients who had a CRP concentration either below 0.02 mg/dl (N = 9) or above 10 mg/dl (N = 1) were excluded in our analysis, because 0.02 mg/dl is the lower limit of detection, and CRP levels higher than 10 mg/dl are often suggestive of an underlying an infection, an inflammatory disease, or cancer45. Among the remaining 408 patients, we obtained genomic DNA for 405 patients. Among the 1,365 controls, 118 subjects who had a CRP concentration either below 0.02 mg/dl (N = 118) or above 10 mg/dl (N = 0) were also excluded in our analysis. Among the remaining 1,247 controls, we obtained genomic DNA for 932 controls. The mean serum CRP levels in the 408 patients with schizophrenia and the 1,247 control subjects were 0.36 mg/dl (SD = 0.81) and 0.03 mg/dl (SD = 0.01), respectively. An ANCOVA was performed to examine the presence of differences between the 2 groups in the natural log-transformed serum CRP levels separately by the three genotypes of each SNP (rs2794520 and rs1183910). When we examined the differences between 2 groups separately by the 3 genotypes (TT, TC and CC) of rs2794520, a significant effect on diagnosis (higher in patients than in the controls) was observed in all of the 3 strata (p = 6.7 × 10−13, p = 5.5 × 10−6, and p = 8.8 × 10−3, respectively) after adjusting for age and gender (Fig. 1). When we examined the differences between the 2 groups separately by the 3 genotypes (AA, AG and GG) of rs1183910, a significant effect on diagnosis (higher in schizophrenia than in the control) was observed in all of the 3 strata (p = 4.5 × 10−6, p = 7.0 × 10−11, and p = 7.3 × 10−6, respectively) after adjusting for age and gender (Fig. 1). Among these six strata, the difference between 2 groups in the stratum of rs2794520 CC genotype did not reach statistical significance after the Bonferroni adjustment (p > 8.3 × 10−3).

Figure 1

Differences in the serum CRP levels between patients with schizophrenia and controls separately by genotypes (rs2794520 and rs1183910).

The ANCOVA demonstrated that the serum CRP levels in patients with schizophrenia were significantly higher than in controls in most of the strata after the Bonferroni adjustment (*age and gender-adjusted p < 0.05).

A meta-analysis of case-control studies between serum CRP levels and schizophrenia

We performed a meta-analysis of previous case-control studies between serum CRP levels and schizophrenia. The studies included in this meta-analysis are shown in Supplementary Table 1. We used 14 case-control studies78910111213142122242526, including our data, for a total of 1,664 patients with schizophrenia and 3,070 control subjects. As shown in Fig. 2, the random-effects model revealed that the serum CRP levels were significantly higher in patients with schizophrenia than in the controls (SMD = 0.62; 95% CI, 0.24–0.99; p = 1.4 × 10−3) with significant heterogeneity among studies (I2 = 96.4%; p < 0.05). The funnel plot analysis indicated no evidence of publication bias in these 14 association studies (p = 0.15). In the subgroup analysis based on serum studies evaluated by high-sensitive CRP assays, the serum high-sensitivity CRP levels was also significantly higher in patients with schizophrenia than in the controls (SMD = 0.54; 95% CI, 0.20–0.87; p = 1.7 × 10−3) with significant heterogeneity among studies (I2 = 93.2%; p < 0.05).

Figure 2

Meta-analyses of case-control studies between serum CRP levels and schizophrenia.

The results of the meta-analysis of fifteen case-control studies which measured serum CRP levels (N = 4,734). The serum CRP levels were significantly higher in patients with schizophrenia than in the controls (SMD = 0.62; 95% CI, 0.24–0.99; p = 1.4 × 10−3 in the random-effects model).

Mendelian randomization results

Gene-CRP association

For the gene-CRP association, beta coefficient values of rs2794520 and rs1183910, which are expressed as the number of SD differences in the natural log-transformed serum CRP levels per allele, were calculated in the Japanese non-psychiatric control subjects (N = 932). Each SNP had a significant effect on the natural log-transformed serum CRP levels with a betacrp/per alleleof 0.20 (SE, 0.05; p = 7.3 × 10−5) for the C allele of rs2794520 and 0.18 (SE, 0.05; p = 6.6 × 10−5) for the G allele of rs1183910. The beta coefficient value of each SNP from a meta-analysis of GWAS31 was 0.16 (SE, 0.006; p = 2.0 × 10−186) and 0.15 (SE, 0.006; p = 2.1 × 10−124), respectively.

Gene-schizophrenia association

We first examined whether the 2 serum CRP-related SNPs (rs2794520 and rs1183910) were associated with schizophrenia in the 4 Japanese case-control sample sets (N = 6,840, 2,593 cases and 4,247 controls). The genotypic distributions of these 2 SNPs did not deviate significantly from the Hardy-Weinberg equilibrium (HWE) in the control groups of all sample sets (p > 0.05). No significant difference between any SNP and schizophrenia was observed in any of the sample sets (P > 0.05). Next, we performed a meta-analysis using these 4 Japanese genetic association studies. No significant heterogeneity was detected at either SNP among the 4 studies (p > 0.05), and the funnel plot analysis indicated no evidence of publication bias (p > 0.05). No significant difference between each of the SNPs and schizophrenia was observed in the meta-analysis, with an ORscz/per allele of 0.99 (95% CI, 0.90–1.08; p = 0.77) for the C allele of rs2794520 and 0.96 (95% CI, 0.89–1.03; p = 0.23) for the G allele of rs1183910 (Supplementary Fig.3). The OR of each SNP from a meta-analysis of GWAS43 was 1.02 (95%CI, 1.00–1.04; p = 0.046) and 1.03 (95% CI, 1.00–1.05; p = 0.013), respectively. None of these two SNPs exhibited genome-wide levels of significant (p < 5 × 10−8).

CRP-schizophrenia association

By combining the 2 pooled estimates, which include the ORscz/per allele from the meta-analysis of 4 Japanese case-control studies and the betacrp/per allele from the Japanese control subjects, we calculated the effect of the CRP levels on the risk of schizophrenia (ORscz/crp). No significant effect of the CRP levels on schizophrenia risk was observed in the Japanese population, with an ORscz/crp of 0.99 (95% CI, 0.69–1.42; p = 0.94) using rs2794526 and 0.79 (95% CI, 0.53–1.18; p = 0.25) using rs1183910. On the other hand, when we calculated the effect of the CRP levels on the risk of schizophrenia in the world-wide population by combining 2 pooled estimates, OR of schizophrenia from a meta-analysis of GWAS43 and beta of CRP from a meta-analysis of GWAS31, we found a significant effect of CRP levels on schizophrenia, representing an ORscz/cro of 1.15 (95% CI, 1.00–1.31; p = 0.046) using rs2794526 and 1.21 (95% CI, 1.04–1.40; p = 0.014) using rs1183910, as well as an ORscz/cro of 1.17 (95% CI, 1.06–1.30; p = 0.0017) across these 2 SNPs (Fig. 3). Furthermore, the pooled Mendelian randomization estimate across 15 CRP-associated SNPs, which reached genome-wide significance in a meta-analysis of GWAS31, showed the similar result (ORscz/cro 1.10, 95% CI, 1.02–1.19; p = 0.015) (Supplementary Table 2).

Figure 3

Mendelian randomization analysis.

Both a Mendelian randomization estimate based on each SNP (rs2794520 and rs1183910) and a pooled Mendelian randomization estimate across these 2 SNPs showed significant effects of CRP levels on schizophrenia risk (p = 0.046, p = 0.014, and p = 0.0017, respectively).

Discussion

In this study, we revealed the presence of higher serum CRP levels in patients with schizophrenia compared to controls in the Japanese population by conducting an ANCOVA with separate genotypes of the 2 SNPs (rs2794520 and rs1183910) identified in the meta-analyses of genome-wide association studies of CRP although the difference between 2 groups in the stratum of rs2794520 CC genotype did not reach statistical significance after the Bonferroni adjustment. This is the first study to examine the association between serum CRP levels and schizophrenia with consideration of the effects of genetic variants associated with CRP levels. Our findings are consistent with the results of previous association studies7891011121314151617181920212223.

We provided further evidence that serum CRP levels are elevated in patients with schizophrenia by conducting a meta-analysis of 14 case-control studies between serum CRP levels and schizophrenia, including our data. Our result is consistent with the results from two previous meta-analyses of association studies between blood CRP levels, including plasma and serum CRP, and schizophrenia2829. However, the considerable heterogeneity was observed in our meta-analysis using only serum studies and in a recent meta-analysis using serum and plasma studies (I2 = 96.4% and 91.1%, respectively).

We also demonstrated the causal relationship between CRP levels and schizophrenia in the world-wide population by conducting a Mendelian randomization analysis based on the selected SNPs. Investigating whether elevated CRP levels in schizophrenia is a cause or consequence is important for the understanding of the pathology and treatment of schizophrenia. Our finding in the present Mendelian randomization study is consistent with the results of Wium-Andersen MK et al.20. They have demonstrated that an elevated plasma CRP was associated with late- and very-late-onset schizophrenia by conducting a prospective analysis as well as Mendelian randomization analysis based on 78,810 individuals from the Copenhagen General Population Study and the Copenhagen City Heart Study combined. Our finding suggests that CRP levels are causally associated with not only late- and very-late-onset schizophrenia but also general schizophrenia.

The underlying mechanisms of elevated CRP levels in the pathogenesis of schizophrenia are still unclear. CRP has been reported to have proinflammatory effects46, and an increased paracellular permeability of the blood-brain barrier at elevated CRP levels that enables CRP entry into the central nervous system has been reported47. So chronic elevated CRP levels during development may alter the blood-brain barrier functions and cause neuroinflammation in the central nervous system. It has also been proposed that elevated CRP levels may affect the micro-vascular system in the brain and cerebral flow, neurotransmitter synthesis, and neurotransmission484950515253. Furthermore, elevated CRP levels have been associated with the severity of clinical symptoms, cognitive impairments, and sensory impairments in schizophrenia54555657.

Our finding of the Mendelian randomization analysis leads to the hypothesis that medications that reduce CRP levels can be used in schizophrenia. In fact, several clinical studies have demonstrated the efficacy of anti-inflammatory drugs, such as aspirin and the cyclooxygenase-2 (COX-2) inhibitor celecoxib, on the symptoms in patients with schizophrenia5859.

There are some limitations to the present study. First, we excluded from our association study any subjects who had a CRP concentration below the assay’s limit of 0.02 mg/dl (a total of 127 subjects; patients N = 9, controls N = 118). However, when these subjects were assigned a value of 0.02 mg/dl and included in our ANCOVA with separate genotypes of the 2 SNPs, significantly elevated serum CRP levels in patients were also observed in all of the six strata after the Bonferroni adjustment. Second, all patients were receiving various antipsychotic drugs in our samples, and these medications might influence our results. However, when we examined the relationship between an equivalent dose of antipsychotics and the serum CRP levels in patients by using a univariate linear regression model, we did not observe a significant correlation (p = 0.53). In addition, an elevated level of serum CRP in patients with antipsychotic-free schizophrenia has been reported9 and the initiation of antipsychotics has not been associated with an increase in CRP levels29. Third, we considered only a few determinants in our case-control study because of lack of information. Several determinants that may affect CRP concentrations, such as smoking, BMI, obesity, HDL cholesterol, triglycerides, diabetes mellitus, hypertension, genetic variants, and medications like statins have been reported30313859606162. And other unknown confounding factors of non-diseases related conditions that impact cytokine and CRP levels might be distributed in our cohort. Fourth, there was significant heterogeneity among the studies that we used in our meta-analysis of association studies between serum CRP and schizophrenia. This heterogeneity might be caused by the clinical heterogeneity of the patients and some of the determinants that may affect CRP concentrations. Fifth, the Mendelian randomization estimate showed a significant effect of CRP levels on schizophrenia only in the world-wide population, not in the Japanese population. The discrepancy of these results might be mainly caused by sample size (36,989 cases and 113,075 controls vs. 2,593 cases and 4,247 controls)63. Finally, an elevated level of CRP in the blood has been observed not only in schizophrenia but also in depression and bipolar disorder6465. Wium-Andersen MK et al. have demonstrated with a Mendelian randomization approach that CRP was not a causal risk factor for depression, but a risk factor for bipolar disorder6466. These results suggest that some common CRP-related inflammatory biological mechanisms may contribute to the development of schizophrenia and bipolar disorder. Further studies will be necessary to reveal how elevated CRP levels are involved in the pathophysiology of each disease.

In conclusion, we demonstrated higher serum CRP levels in patients with schizophrenia compared to controls in the Japanese population. The meta-analysis of case-control studies which measured serum CRP levels supported our findings. We provided evidence for a causal association between elevated CRP levels and schizophrenia risk when we conducted a Mendelian randomization analysis. Our findings suggest that elevated CRP itself may be a causal risk factor for schizophrenia and lead to the hypothesis that medications that reduce CRP levels can be used in schizophrenia. This hypothesis should be tested in future studies.

Additional Information

How to cite this article: Inoshita, M. et al. A significant causal association between C-reactive protein levels and schizophrenia. Sci. Rep. 6, 26105; doi: 10.1038/srep26105 (2016).