Association between C4, C4A, and C4B copy number variations and susceptibility to autoimmune diseases: a meta-analysis

Although several studies have investigated the association between C4, C4A, and C4B gene copy number variations (CNVs) and susceptibility to autoimmune diseases, the results remain inconsistency for those diseases. Thus, in this study, a comprehensive meta-analysis was conducted to assess the role of C4, C4A, and C4B CNVs in autoimmune diseases in different ethnic groups. A total of 16 case-control studies described in 12 articles (8663 cases and 11099 controls) were included in this study. The pooled analyses showed that a low C4 gene copy number (GCN) (<4) was treated as a significant risk factor (odds ratio [OR] = 1.46, 95% confidence interval [CI] = 1.19–1.78) for autoimmune diseases compared with a higher GCN (>4). The pooled statistical results revealed that low C4 (<4) and low C4A (<2) GCNs could be risk factors for systemic lupus erythematosus (SLE) in Caucasian populations. Additionally, the correlation between C4B CNVs and all type of autoimmune diseases could not be confirmed by the current meta-analysis (OR = 1.07, 95% CI = 0.93–1.24). These data suggest that deficiency or absence of C4 and C4A CNVs may cause susceptibility to SLE.

The complement system, which is involved in both innate and adaptive immunity, is characterized by triggered-enzyme cascades activated by alternative, lectin or classical pathways 1 . As a necessary and important component of the complement system, complement component C4 plays a pivotal role in the activation of immune defenses and the clearance of immune complexes or apoptotic debris in vitro and in vivo 2,3 . If a structural variation such as fragmentary deficiency or deletion of a complement C4 gene occurs, it could cause an aberrant immune response and an autoimmune or inflammatory disorder 4 . Genetically, the complement component C4 gene is mapped in the class III region of the major histocompatibility complex (MHC) on chromosome 6p21. 3 and there are two isotypes of C4 gene products, which are acidic C4A and basis C4B 5 . The C4A and C4B isotypes can be distinguished by four specific amino acids at positions 1101-1106 (or 1120-1125 if the protein numbering starts at the initiation codon for Methionine), which were the results of five nucleotide polymorphisms 6,7 . C4A prefers to bind to amino groups of immune complexes, while C4B binds to hydroxyl groups when activated 8,9 . Cumulative research has confirmed that gene expression as well as C4, C4A, and C4B CNVs could affect the strength of an individual's immunity and susceptibility to autoimmune diseases 8,[10][11][12] . The link between complement component C4 complete protein deficiency and autoimmune diseases was first observed among patients with systemic lupus erythematosus (SLE) in 1974 13 .
CNVs in the human genome belong to a significant sequence variation, indicating that large stretches of DNA may exist in a number of different forms in different individuals 14 . CNVs in genes performing functions related to immunity can increase immunological diversity and genetic predisposition to autoimmune diseases 15 . It has been reported that partial C4 deficiency is one of the most frequent immunoprotein deficiencies in Homo sapiens 10 . Additionally, both C4A and C4B have various GCNs according to the Database of Genomic Variants (http://projects.tcag.ca/variation). The C4 gene encodes for either C4A or C4B, so the GCN of total C4 is equal to the sum of the GCNs of C4A and C4B 16 . CNV patterns in individuals belonging to different racial groups vary in terms of the number and size of C4 genes. C4 exists as two isoforms encoded by different genes, C4A and C4B. 1 Meta-analysis results. The relationship between C4, C4A, and C4B CNVs and autoimmune diseases was investigated, and the detailed results of pooled odds ratios (ORs) and a stratification analysis are presented in Table 3. The pooled analyses suggested significant between-study heterogeneity among the three genes (C4: I 2 = 85.3%, P H < 0.001; C4A: I 2 = 89.5%, P H < 0.001; C4B: I 2 = 72.8%, P H < 0.001). Hence, the meta-analysis detected a random-effect model. As presented in Fig. 2, there was a significant association between C4 GCN and autoimmune diseases. Individuals with low C4 copy numbers (< 4) were more likely to develop an autoimmune disease (pooled OR = 1.46, 95% CI: 1.19-1.78). In the subgroup analysis involving disease type, the risk of SLE was clearly increased in individuals with low C4 GCNs (pooled OR = 1.80, 95% CI: 1.51-2.13). Moreover, the subgroup analysis involving ethnicity revealed that low C4 GCNs were significantly associated with the risk of autoimmune diseases among Caucasian individuals (pooled OR = 1.91, 95% CI:1.42-2.56), while such an association was not confirmed in an Asian population (pooled OR = 1.13, 95% CI: 0.88-1.45). Furthermore, the results suggested that low C4A GCNs (< 2) are more associated with apparent risk for autoimmune diseases compared with higher GCNs (≥ 2) (pooled OR = 1.46, 95% CI: 1.10-1.94) (Fig. 3). The subgroup analysis involving both disease type and ethnicity indicated that low C4A GCNs were significantly associated with susceptibility to SLE and a Caucasian population (pooled OR = 2.13, 95% CI: 1.71-2.64; pooled OR = 2.05, 95% CI: 1.49-2.83, respectively). No significant association between low C4B GCNs (< 2) and autoimmune diseases was found in this meta-analysis (OR = 1.07, 95% CI: 0.93-1.24) (Fig. 4). The fixed effect model was applied to these data in accordance with Chen et al.'s report 30 . In the subgroup analysis involving disease type, we found a significantly increased risk for SLE among carriers with low C4 or C4A GCNs (pooled OR = 1.68, 95% CI: 1.51-1.86; pooled OR = 1.99, 95% CI: 1.77-2.24, respectively). Finally, the studies were stratified by ethnicity, which found an apparently increased risk of autoimmune diseases associated with low C4 or C4A GCNs in Caucasians (pooled OR = 1.57, 95% CI: 1.41-1.75; pooled OR = 1.83, 95% CI: 1.62-2.06, respectively). These results suggest that, regardless of the random effect model or fixed effect model used for meta-analysis, the pooled data was consistent, believable, and stable.
Heterogeneity test and sensitivity analysis. As suggested in Table 3, there was significant heterogeneity between studies in terms of GCNs (p < 0.05). The results of our subgroup analysis confirmed that disease type and ethnicity were the main sources of heterogeneity. Additionally, a sensitivity analysis was conducted to evaluate the effect of individual studies on the pooled ORs by sequentially omitting each study. The pooled ORs were not affected by excluding any study (data not shown).

Refs
Year  Publication bias. Begg's funnel plots and Egger's regression tests were applied to determine the potential publication bias for C4, C4A, and C4B CNVs. As presented in Table 4, there was no obvious publication bias among these detection. In order to validate whether there was potential publication bias regarding C4, we performed a funnel plot using the trim and fill method. LnOR and 95% CI were 0.376 (0.173, 0.579) and 0.246 (0.042, 0.450), respectively, before and after applying the trim and fill method, which indicated that publication bias was present in the meta-analysis (Fig. 5).

Discussion
In our meta-analysis of 8663 cases and 11099 controls in 16 studies from 12 articles, we drew a general conclusion that C4 and C4A CNVs are tightly associated with autoimmune diseases, especially with SLE. Individuals with low GCNs of C4 (< 4) or C4A (< 2) are predisposed to autoimmune disorders in the presence of environmental triggers. Consistent with previous studies, both C4 and C4A CNVs were regarded as pivotal genetic factors in the pathogenesis of SLE. Furthermore, our meta-analysis demonstrated that low GCNs of C4 (< 4) or C4A (< 2) could lead to increased risk of autoimmune diseases among Caucasian populations. In other words, differences in the association between different ethnicities may result from other factors, such as geography, socioeconomic development, or race. Complement C4 plays an essential role in innate and adaptive immune responses, which are involved in the classical and mannose-binding lectin complement activation pathways and help to direct against external attacks such as microbial infection, clearance of immune complexes, and removal of apoptotic cells 31,32 . Human C4 protein is encoded by two polymorphic genes, C4A and C4B, which are located in the MHC on chromosome 6. C4 deficiency and dysfunction are linked to the pathogenesis of many autoimmune and inflammatory diseases 33 . Several studies investigating different ethnic groups have shown that C4 deficiency may be one of the most penetrant genetic risk factors of SLE 4,21,23 . The results of the current meta-analysis were consistent with earlier observations.      C4A and C4B are two isotypes of C4. Although they share > 99% of their amino acid sequences, their chemical reactions to substrates are remarkably different. C4A tends to combine amino group-containing antigens or immune complexes, while C4B combines hydroxyl group-containing antigens. Because C4A more efficiently handles immune complexes, deficiency of C4A while C4B affects the development of SLE 34,35 .
It is generally accepted that CNVs are a genetic determinant of phenotypic variation 9 . A CNV is a type of structural variation in which large segments of DNA are altered due to duplication, deletion, insertion, inversion, or complex combination or rearrangement 36 . Loci involved in immunity are prone to CNV, leading to differences in the intrinsic strength of the immune system and variations in susceptibility to immune disorders 8,37 . Haraksingh et al. 38 found that CNVs may affect gene dosage due to various copies of a certain gene that is present in the genome. Yang et al. 8 has reported a strong, positive correlation between C4 GCN and serum C4 protein concentration as well as C4A and C4B gene dosage and serum C4A and C4B concentration. In their study, a low GCN group (for C4, n < 4; for C4A, n < 2) had a significantly lower serum protein concentration than that of a medium GCN group (for C4, n = 4; for C4A, n = 2) and a high GCN group (for C4, n > 4; for C4A, n > 2) 8 . Since C4 plays a pivotal role in innate and adaptive immune responses, individuals with low GCNs of C4 (< 4) or C4A (< 2) could show reduced serum C4 or C4A concentrations, resulting in dysfunction in the ability to resist microbial infection, clear immune complexes, and remove apoptotic cells. Thus, low C4 CNV is a risk factor for SLE. As we know, SLE, RA, BD, and AS are immune-mediated autoimmune diseases. In this meta-analysis, we planned to summarize the current knowledge related to the relationship between C4 CNVs and SLE and other systemic autoimmune diseases and to determine whether C4 is a genetic master key to autoimmunity. Liu et al. 22 revealed that individuals with less than 4 copies of C4 and those with less than 2 copies of C4A and C4B tended to be at less risk for GD. Liu et al. 22 also found that less than 2 copies of C4A may be associated with high risk for vitiligo in patients with GD. Hou et al. 19 found that patients with Vogt-Koyanagi-Harada (VKH) syndrome have decreased frequency of high C4 GCNs (> 4) and decreased expression of serum C4 in patients. Hou et al. 18 indicated that high GCNs (> 2) of C4A lead to risk for BD but not acute anterior uveitis by modulating the expression of C4A and enhancing IL-6 production. In another study, Lintner et al. 28 discovered that there were significant differences in the distribution of lower C4A and C4 GCNs (< 2 and < 4, respectively) among JDM patients and controls, suggesting that complement C4A deficiency appeared to be an important risk factor for JDM. Furthermore, Rigby et al. 26 demonstrated that C4B deficiency resulted in increased risk for RA and had broad implications for the pathogenesis of RA. However, data from pooled analyses of C4 and non-SLE autoimmune diseases are relatively insufficient. Additionally, data from several published articles were only done by single method without validation using the other methods and therefore further studies with different methods and larger samples including individuals from different ethnic backgrounds would help reveal the role of C4 in GD, uveitis, JDM, and RA.
Some limitations of the current meta-analysis of the potential relationship between C4, C4A, and C4B CNVs and autoimmune diseases need to be addressed. First, heterogeneity among ethnic groups or disease types was discovered when the association between C4, C4A, and C4B CNVs and autoimmune diseases was investigated. However, based on the results of the sensitivity analysis, it is clear that the overall results were not affected by heterogeneity. Second, an obvious publication bias was detected in the comparison between C4 CNV and autoimmune diseases. Third, the size of the patient and control groups was relatively small in each study; therefore, future studies should employ a much larger sample size and include individuals from different ethnic populations. Fourth, the effects of common confounding factors, including sex, age, and medical condition, were not assessed in the present study because of insufficient data. Fifth, autoimmune diseases at various sites may vary widely in terms of the etiology of pathology, including genetic factors, immunologic factors, and environmental factors. Finally, the electronic databases from which we selected eligible studies were listed in English and Chinese; thus, the meta-analysis may have a language bias.
In conclusion, the present meta-analysis provides evidence-based pooled data revealing a significant association between low C4 or C4A CNVs and susceptibility to autoimmune diseases, especially for Caucasian individuals with SLE.

Materials and Methods
Literature search. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) criteria were using as a guide for this meta-analysis 39  Inclusion and exclusion criteria. Studies selected from electronic databases were included when they met the following criteria: (1) focus on the CNVs of total C4, C4A, or C4B in relation to autoimmune diseases; (2) a case-control study design; (3) presence of enough data to calculate an OR and corresponding 95% CI; and (4) inclusion of patients diagnosed according to international standards. The following items were regarded as the primary exclusion criteria: (1) focus on families or twins rather than random general populations; (2) inclusion of individuals with overlapping investigated diseases; (3) a case study design; (4) insufficient data applied after emailing the relevant corresponding author. When several articles with the same types of patients were identified, the most detailed article with the largest sample size and most comprehensive analysis was included in the meta-analysis.
Data extraction. Data from the retrieved studies were extracted independently by two authors (N.L. and J.Z.). The following information was collected from each eligible study: first author, year of publication, country of origin, ethnicity of subjects, type of disease, sample size, genotyping methods, and information about GCN distribution in cases and controls. Two authors carefully screened the collected data and reached agreement in all respects. When the authors disagreed, a third reviewer (D.L.) weighed the arguments and then helped reach a consensus.
Quality assessment. Evaluation of the quality of extracted studies was also performed by two authors (N.L. and J.Z.) based on the HuGENet Handbook 40 . Six bias assessment items referring to the association between genes and disease were derived from this handbook, including bias in selection of cases, bias in selection of controls, bias in genotyping cases, bias in genotyping controls, bias in population stratification, confounding bias, multiple tests, and selective outcome reports. The quality of every item was labeled as "Yes" or "No, " while "Unclear" was used if there was not enough information to make a determination. A correction and review were performed independently by the investigator (D.L.) if the two coauthors dissented. Consensus regarding all labels was achieved after discussion.

Statistical analysis.
ORs and the corresponding 95% CIs were estimated to evaluate the amount of correlation between C4, C4A, and C4B CNVs and susceptibility to autoimmune diseases. The diversity of genotypes (for total C4, the number of subjects with < 4 GCNs compared to those with ≥ 4 or > 4, and for C4A and C4B, the number of subjects with < 2 GCNs compared to those with ≥ 2 or > 2) among cases and healthy controls was compared. The heterogeneity between studies was estimated and measured by Cochran's Q statistic as well as the I 2 statistic. When the p value of chi-square statistic was less than 0.05 or the I 2 value was more than 50%, the random-effect model was adopted for meta-analysis, indicating that heterogeneity existed across studies 41 . In addition, a goodness of fit test proposed by Chen et al. 30 was used to explore the adequacy of the model for our systematic meta-analysis. Sensitivity analyses were performed to assess the effect of individual studies on pooled ORs by omitting each study in turn. Publication bias was estimated by inspecting Begg's funnel plots 42 and Egger's regression test 43 . Statistical data was analyzed using STATA 12.0 software (StataCorp LP, College Station, Texas, USA). A significant difference was estimated at p < 0.05 (a two-tailed p value). The final conclusions of the study were independently validated by two authors (J. Z. and S. H.).