Objective Complement-mediated vasculopathy of muscle and skin are clinical features of juvenile dermatomyositis (JDM). We assess gene copy-number variations (CNVs) for complement C4 and its isotypes, C4A and C4B, in genetic risks and pathogenesis of JDM.
Methods The study population included 105 patients with JDM and 500 healthy European Americans. Gene copy-numbers (GCNs) for total C4, C4A, C4B and HLA-DRB1 genotypes were determined by Southern blots and qPCRs. Processed activation product C4d bound to erythrocytes (E-C4d) was measured by flow cytometry. Global gene-expression microarrays were performed in 19 patients with JDM and seven controls using PAXgene-blood RNA. Differential expression levels for selected genes were validated by qPCR.
Results Significantly lower GCNs and differences in distribution of GCN groups for total C4 and C4A were observed in JDM versus controls. Lower GCN of C4A in JDM remained among HLA DR3-positive subjects (p=0.015). Homozygous or heterozygous C4A-deficiency was present in 40.0% of patients with JDM compared with 18.2% of controls (OR=3.00 (1.87 to 4.79), p=8.2×10−6). Patients with JDM had higher levels of E-C4d than controls (p=0.004). In JDM, C4A-deficient subjects had higher levels of E-C4d (p=0.0003) and higher frequency of elevated levels of multiple serum muscle enzymes at diagnosis (p=0.0025). Microarray profiling of blood RNA revealed upregulation of type I interferon-stimulated genes and lower abundance of transcripts for T-cell and chemokine function genes in JDM, but this was less prominent among C4A-deficient or DR3-positive patients.
Conclusions Complement C4A deficiency appears to be an important factor for the genetic risk and pathogenesis of JDM, particularly in patients with a DR3-positive background.
- Autoimmune Diseases
- Gene Polymorphism
- Disease Activity
Statistics from Altmetric.com
Juvenile dermatomyositis (JDM) is a rare, autoimmune, multisystem inflammatory disease affecting primarily muscle and skin in children. Characteristic clinical features and diagnostic criteria include proximal muscle weakness and inflammation, increased levels of serum muscle enzymes, distinct skin rashes such as Gottron's papules or heliotrope rash and pathological changes on muscle biopsy or MRI.1–6
The human leucocyte antigen (HLA) class II gene DRB1 allele *0301 (also known as DR3) has been identified as a major immunogenetic risk factor for JDM and was reaffirmed as the predominant risk locus of juvenile and adult dermatomyositis in a genome-wide association study.3 ,4 ,7 HLA class I and class II genes are engaged in antigen presentation and processing. The class III genes are heterogeneous in structure and function, and include genes encoding for components of the complement system C4, C2 and factor B, and for cytokines such as tumour necrosis factor-α (TNF-α), and α and β lymphotoxins (figure 1).8 Previous studies revealed that class II genes DRB1 allele *0301 and DQA1 allele *0501, class III gene TNFA−308A allele, and class I gene variants B*08 and A*01 are in strong linkage disequilibrium among human subjects of European descent and this is described as the ancestral haplotype AH8.1.9–11 Also present in AH8.1 is a single C4B gene but the absence of a C4A gene.12 ,13 Remarkably, there are extensive interindividual gene copy-number variations (CNVs) and gene-size dichotomy for complement C4. Two to eight copies of C4 genes can be present in a diploid genome.14 ,15 Segmental duplications for complement C4 genes occur as RCCX modules, which always include the RP (STK19) gene upstream of C4, and the downstream genes CYP21 and TNX (RCCX). Each C4 gene can be a long gene of 20.6 kb or a short gene of 14.2 kb.16 ,17 Each C4 gene either encodes for an acidic C4A or a basic C4B protein, with only four amino acid changes (PCPVLD 1101–1106 for C4A and LSPVIH for C4B), but these result in substantial differences in chemical reactivity for peptide and carbohydrate antigens.18–21
Complement-mediated destruction of perivascular endothelium and perifascicular ischaemia of muscle fibres in biopsies from patients with dermatomyositis have been demonstrated by multiple investigators.22–27 Circulating immune complexes, immunoglobulins IgG and IgM, complement C3 and C5b-9 membrane attack complex were shown in dermatomyositis muscle and skin biopsies. However, the initiation for complement activation and the potential role of complement in the breakdown of immune tolerance in JDM remain unclear. Continuous CNVs with one to four copies of C4 genes on a haplotype with different combinations of C4A and C4B genes in human populations have only been established since 1999.15 ,28 ,29 Low gene copy number (GCN) for total C4 or C4A deficiency has been observed as a risk factor of human systemic lupus erythematosus.30–33 Many earlier epidemiological studies of complement C4A and C4B in rheumatic diseases, including JDM, were based on an incomplete or inaccurate model with two-locus (C4A–C4B) on a haplotype for data interpretation, and thus the conclusions drawn became uncertain.34–37 Here we perform an investigation of C4 genetic diversity and examine the effects on the risk and pathogenesis of JDM, with further considerations to the presence and absence of HLA-DRB1 risk and protective alleles.
Patients and methods
IRB approval was obtained from Nationwide Children's Hospital (NCH) and the National Institutes of Health (National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK)/National Institute of Arthritis and Musculoskeletal and Skin Diseases (NIAMS), NIH). One hundred and five patients with JDM were enrolled, of whom 60 were from NIH and 45 were from the NCH. Subjects with JDM were required to meet the original or revised Bohan and Peter criteria.1 ,2 Typical characteristic MRI abnormalities of muscle were applied in place of biopsy as a modification of the Bohan and Peter criteria for the NCH cohort.38 The mean age (±SD) at recruitment was 10.8±7.6 years, and at disease diagnosis was 7.4±4.2 years. The self-reported racial distribution was 90.5% Caucasian, 6.7% African-American and 2.9% Hispanic. Complete demographics and disease characteristics are displayed in table 1. Ten non-Caucasian patients with JDM were not included in the genetic analysis. Race-matched healthy control subjects included 500 European Americans residing in Midwest America.
Determination of total C4, C4A and C4B genotypes and phenotypes
Previously, we described protocols for genotyping and phenotyping of complement C4 by Southern blot analyses and immunofixation, respectively.39–41 In the cases of limited DNA quantities or ambiguous results, quantitative real-time PCR experiments for GCN of total C4, C4A and C4B were performed as described.14 All C4-CNV calls were validated rigorously by independent technology, or multiple amplicons in qPCR, and matched genotype and phenotype interpretations (see online supplementary figures S1–S3 and tables S1 and S2).
Flow cytometric detection of erythrocyte-bound complement activation fragments
Erythrocytes from whole blood were used for antibody staining and flow cytometry. Mouse monoclonal antibodies specific for human C4d, for human C3d, or the isotype-matched control MOPC21 (Quidel, San Diego, California, USA) were used.42 ,43 Phycoerythrin-conjugated goat anti-mouse IgG F(ab′)2 (Jackson ImmunoResearch, West Grove, Pennsylvania, USA) was used as a secondary antibody. FlowJo software (Tree Star, V.7.6) was used to electronically gate erythrocytes based on forward and sideward scatter properties. Among the gated cells, E-C4d or E-C3d was reported as median fluorescence intensity (MFI), which was calculated using C4d-specific (or C3d-specific) MFI minus the MOPC-isotype control MFI.
Genotyping of HLA-DRB1 alleles for JDM and all control samples were determined using genomic DNA for sequence-specific primer PCR.44 HLA-DRB1 frequency was calculated by the number of allele-positive subjects divided by the total number of subjects.
Gene expression profiling
RNA was extracted from whole blood using PAXgene collection tubes (PreAnalytiX, Becton, Dickinson and Company). Microarray analysis was performed by the Biomedical Genomics Core facility at the NCH. RNA samples passing quality control were labelled with Agilent's One-Color microarray-based gene expression analysis labelling protocol and hybridised to the SurePrint G3 Human V.2 GE 8×60K Microarray (AMADID 039494). Images were analysed with Feature Extraction 10.9 (Agilent Technologies). Median foreground intensities were obtained for each spot and imported into the mathematical software package R. After preprocessing, the data were quantile normalised using the linear models for microarray data (LIMMA) package.45 Statistical analysis was performed via Significance Analysis of Microarray (SAM) implemented using the Bioconductor Siggenes package to identify differentially expressed genes between JDM and control groups.46 Changes in expression ≥1.5-fold and a 15% false discovery rate as estimated by SAM were considered significantly different. Selected genes were validated by SYBR-Green qPCR using PAXgene RNA. The housekeeping gene GAPDH was used as a normalisation standard.47
Statistical analyses were performed using Prism6 (GraphPad Software, San Diego, California, USA) and JMP Genomics V.6.0 (SAS Institute, Cary, North Carolina, USA) software. Descriptive statistics are displayed as mean±SD for normally distributed data, and simple comparisons were made using Student's t test for continuous data, or by χ2 analysis for categorical data. For non-normally distributed data, median with IQR is reported, and Mann–Whitney test was used for comparisons. For all analyses, p≤0.05 was considered to be significant.
Total C4 genes
In healthy controls (N=500), the variation of C4 GCN ranged from two to seven total copies. In Caucasian patients with JDM (N=95), total C4 genes ranged from two to five copies, with a shift of distribution to the lower copy number compared with controls (χ2=20.7; degree of freedom=4; p=0.0004, χ2 analysis; panel B, figure 1). Patients with JDM had a lower mean GCN than did controls (JDM: 3.49±0.71; controls: 3.83±0.69; p=1.8×10−5, t test; table 2A). Low copy number of total C4 (C4T, GCN≤3) was present in 50.5% of patients with JDM and 29.2% of controls. The OR and 95% CI of total C4 with GCN≤3 among patients with JDM was 2.48 (1.59 to 3.87) and the p value was 7.5×10−5.
Copy numbers of C4A and C4B genes
The reduction in copy number of total C4 in JDM can be the result of a decrease in C4A, C4B or both. We observed a significant shift to lower GCN of C4A in patients with JDM (p=0.0004; panels C and E, figure 1). The presence of homozygous or heterozygous deficiency of C4A genes (GCN=0 or 1) had a frequency of 40.0% in patients with JDM, compared with 18.2% in controls (OR=3.00 (1.87 to 4.79); p=8.2×10−6). The overall mean GCN of C4A observed in patients with JDM was 1.79±0.86 compared with 2.09±0.75 in controls (p=0.0004, t test).
As for C4B, no significant differences in the distribution of C4B GCN or C4B deficiency were observed between patients with JDM and controls (panel D, figure 1). Therefore, the basis for decreased GCN of total C4 in Caucasian patients with JDM was attributable to lower GCN of C4A.
HLA-DRB1 alleles, C4A GCN and C4A deficiency on disease risks of JDM
The frequency of HLA-DRB1*0301 alleles (DR3) was 46.3% in Caucasian patients with JDM (N=95), compared with 25.8% in a race-matched healthy population (N=500). HLA-DR3 was associated with JDM with an OR=2.48 (1.58 to 3.89); p value=9.5×10−5. The concurrence of C4A deficiency and DR3 in a subject was present in 36.8% of patients with JDM and 15.4% of controls (OR=3.20 (1.98 to 5.19); p value=4.8×10−6). By contrast, the frequency of HLA-DRB1*1501 (DR2) was 11.6% in Caucasian patients with JDM compared with 27.8% in healthy controls. DRB1*1501 was a protective factor against JDM with OR=0.34 (0.18 to 0.66) and p=0.0004 (figure 1E and table 2A).
Multiple logistic regression analyses were performed to investigate if C4A deficiency, the presence of DR3 and the presence of DRB1*1501 could serve as independent risk factors for JDM, conditional upon presence of other factor(s) in five different combinations of regression models (table 2B). In models when DR3+ and C4A deficiency were put together as individual factors (model a or b), the relevance of DR3+ as an independent parameter became insignificant. The presence of DRB1*1501 plus C4A deficiency, or DR3+, or C4A deficiency with DR3+ all yielded statistical significance to account for JDM genetic risks. The last model yielded the best fit: C4A deficiency plus DR3+ was a strong risk factor with OR of 2.96 (1.84 to 4.80) and DRB1*1501 was a protective factor with OR of 0.34 (0.21 to 0.55).
To further evaluate the relative roles of C4A deficiency and DRB1*0301 on disease risk of JDM, we performed subgroup analyses (table 2C). Among the DRB1*0301-positive (DR3+) subjects, patients with JDM had a significantly lower mean GCN of C4A than controls (JDM: 1.18±0.54; controls: 1.47±0.72; p=0.015). Similarly, among the DR3+ subjects, C4A deficiency had a greater prevalence in patients with JDM (79.6%) than controls (59.7%) (OR=2.63 (1.17 to 5.92), p=0.014). Among DR3-negative subjects, however, there were no apparent differences in mean GCN of C4A or the prevalence of C4A deficiency between patients with JDM and controls, suggesting the heightened risk of lower C4A GCN or C4A deficiency on JDM required a DR3+ background.
Reciprocal subgroup analyses to compare the prevalence of DR3+ between patients with JDM and controls among C4A-deficient subjects (GCN≤1), or among C4A-proficient subjects (GCN≥2) revealed slight but insignificant increases in the frequency of DR3 in JDM (table 2C).
Levels of erythrocyte-bound C4d (E-C4d) or C3d (E-C3d) in patients with JDM and controls
Cell-bound complement levels were determined in 40 Columbus patients with JDM and 206 healthy subjects of European ancestry by flow cytometry. Comparing patients with JDM and healthy controls, significant elevation of E-C4d levels (p=0.004, Mann–Whitney test, figure 2A) but not E-C3d levels (figure 2B) was observed in JDM. The MFI for E-C3d levels were substantially lower than those of E-C4d levels, which is consistent with the presence of complement regulation mechanisms on self surfaces.
We investigated if there was a correlation of E-C4d levels with C4A or C4B gene dosages in JDM. The C4A-deficient group (GCN≤1; N=15) had a median MFI of 1426 (IQR: 601–1744), which was significantly higher than that of the C4A-proficient group (GCN≥2; N=25; median MFI=454 (234–718); p=0.0003, Mann–Whitney test). On the other hand, the C4B-deficient group (GCN≤1; N=11) had a median E-C4d MFI of 308 (226–505), which was significantly lower than that of the C4B-proficient group (GCN≥2, N=29; median MFI=775 (495–1458); p=0.003). Thus, C4A and C4B appeared to play opposite roles on the deposition of cell-bound E-C4d: high GCN of C4A dampened activation, while high GCN of C4B amplified activation.
Patients with JDM with elevated levels of multiple serum muscle enzymes had low GCN of C4A
Patients with JDM exhibited elevated levels of a variety of serum muscle enzymes. We performed intragroup comparisons to investigate if C4A deficiency was related to elevated muscle enzyme levels at the time of disease diagnosis. Indeed, patients with C4A deficiency had higher prevalence of abnormal serum muscle enzymes such as creatine kinase (C4A deficient: 86.1%, C4A proficient: 58.2%; p=0.0034) and aldolase (C4A deficient: 94.1%, C4A proficient: 78.4%; p=0.038) and elevations of multiple (≥2) serum muscle enzymes (C4A deficient: 97.1%; C4A proficient: 74.5%; p=0.0025) than C4A-proficient patients. The prevalence of elevated levels of serum aspartate aminotransferase and lactate dehydrogenase was not associated with C4A deficiency (table 3).
Differential gene expression profiling of patients with JDM and controls
Global gene expression profiling was performed using PAXgene blood RNA from 19 consecutive patients with JDM with defined disease status and treatment, plus seven controls (see figure 3A and online supplementary tables S3–S4). Expression profiles revealed differential expression of transcripts in JDM from 56 genes that were significantly different using SAM criteria. Differentially expressed genes included 24 upregulated and 32 downregulated genes. Of the upregulated genes, the most remarkable are nine type I interferon (IFN-I) response genes and three genes related to B-cell functions. Of the downregulated genes, the most distinct were genes related to T-cell functions, chemokines, and chemokine receptors. Six patients with JDM exhibited the most polarised upregulation of IFN-I genes and downregulation of genes for chemokine/chemokine receptor and T-cell functions (figure 3A). Of interest, five of these six patients with JDM were C4A proficient (GCN≥2), and did not carry the HLA-DR3 allele.
To validate gene expression changes from microarray, we performed SYBR-Green qPCR using cDNA from PAXgene-blood RNA for IFI44, IFI17, CXCR6 and CCR5 transcripts. Results revealed in JDM upregulation of IFI44 (2.7-fold; p=0.028) and IFI17 (3.0-fold; p=0.0054), and downregulation of CXCR6 (1.9 fold; p=0.031) and CCR5 (1.5 fold; p=0.048) (figure 3B). The housekeeping gene GAPDH was used as a normalisation standard.47
A great challenge for studying complex diseases associated with the HLA region, including JDM, is to determine which gene(s) or polymorphic variants contribute to disease development under the background of strong allelic associations or linkage disequilibrium. This is the first study to decipher the gene CNVs for complement C4 and its isotypes C4A and C4B in JDM with definitive techniques, and to dissect the relative roles of HLA-DRB1*0301 and C4A deficiency on JDM disease risk in subjects of European ancestry.
The carriage of DRB1*0301 was present in 46.3% of our patients with JDM compared with 25.8% of race-matched healthy controls. DRB1 is a medium effect-size risk factor for JDM (OR=2.48). Homozygous and heterozygous deficiency of complement C4A had a frequency of 40.0% in Caucasian patients with JDM and 18.2% in healthy controls (OR=3.00). The coexistence of HLA-DR3 and C4A deficiency confers higher risk than either individual risk factors, and such concurrence in a diploid genome was present in 36.8% of patients with JDM and 15.4% of controls with an OR=3.20. The roles of DRB1 variants and C4A deficiency in JDM were further examined by multiple logistic regression analyses. Moreover, among DR3-positive subjects, lower mean GCN of C4A or higher prevalence of C4A deficiency persisted in JDM versus controls. An interpretation to this phenomenon is that DR3-positivity contributes to a permissive background and C4A-deficiency significantly elevated the vulnerability to an autoimmune disease including JDM. While HLA-DRB1 and complement C4 each is engaged in specific immunological functions such as antigen presentation to effector T cells, and complement-mediated cytolysis and immunoclearance, they are both involved in the recognition of self and non-self, and are key players for the process of archiving memory and tolerance in the immune system.
Destruction of perifascicular capillaries by complement and subsequent ischaemia of muscle fibres in dermatomyositis have been demonstrated by multiple investigators over the past three decades.22–27 Activation of complement can be initiated via the classical pathway that is triggered by immune complexes formed between myositis-associated or myositis-specific autoantibodies and self-antigens abundant in muscles and skin. Physiologically, low GCN or low production of C4A protein systemically may dampen immune complex clearance and therefore promote autoimmunity. Compared with controls, we observed a moderate but significant increase in the deposition of C4d on red blood cells among patients with JDM, which reiterates involvement of complement activation in the pathogenesis of JDM. Almost all patients with JDM with two or more elevated muscle enzymes at disease diagnosis had a C4A deficiency but normal mean GCN of C4B. In other words, immune-mediated tissue injuries in JDM might have resulted through the activation of C4B. Consistent with this phenomenon, we observed increased deposition of processed complement activation product E-C4d in patients with JDM than in controls. Interestingly, the levels of E-C4d were directly proportional to the GCN of C4B, and inversely proportional to the GCN of C4A. Physiologically, activated C4B protein is highly reactive and overactivation could lead to complement-mediated injuries. In addition to its role in immunoclearance and protection against autoimmunity, the presence of activated C4A may attenuate the activity of C4B and minimise its potential deleterious effect.
Among the patients with JDM, 63.2% were not associated with C4A deficiency on a DR3+ background, and the underlying genetic risk factors in this group of patients (C4A-proficient and DR3-negative) are yet to be identified. An emerging feature in JDM is the upregulation of IFN-I–stimulated gene expression in many patients.48 ,49 Our microarray studies of peripheral blood samples revealed marked increase in transcripts in JDM for many IFN-I stimulated genes and B-cell-specific genes, but diminished transcript levels of many genes related to chemokines and T-cell functions. Such differential levels of transcripts reflect both different gene expression levels and also compositions of leucocytes in the peripheral blood samples (see online supplementary figures S4–S5). The polarised upregulation of IFN-stimulated genes and B-cell function genes, and downregulation of chemokine receptor and T-cell function genes were more marked among C4A-proficient or DR3-negative patients, implying the presence of an alternative or additional mechanism(s) leading to the pathogenesis of JDM.
In conclusion, we report the novel finding of low GCN of complement C4 and C4A deficiency associated with JDM. Patients with JDM with C4A deficiency were more likely to have elevated levels of multiple serum muscle enzymes at diagnosis and high levels of E-C4d. Further in-depth studies through HLA-DRB1 and C4A genotypes, cell-bound C4d levels and differentially expressed genes including those engaged in muscle cell functions and signalling, and characterisation of clinical phenotypes50 may help understanding the pathogenic mechanisms, enable patient stratification and facilitate genotype- and gene expression-guided therapies of JDM.
We are indebted to blood donors and patients with juvenile dermatomyositis for their invaluable specimens and support, and to the following referring physicians for patient recruitment: Stacy Ardoin, Sharon Bout-Tabaku, Gloria Higgins, Bita Arabshahi, Ruy Carrasco, Victoria Cartwright, Anne Eberhard, Barbara Edelheit, Harry Gewanter, Donald Goldsmith, Beth Gottlieb, Thomas Griffin, Melissa Hawkings-Holt, Roger Hollister, Yukiko Kimura, Patrick Knibbe, Lauren Pachman, Maria Perez, Abigail Smukler, Sangeeta Sule, Carol Wallace and Lawrence Zemel. We are grateful to Ms Jeanie Shaw for help in recruiting healthy subjects, and Drs Joe Ahearn and CC Liu for assistance in interpreting erythrocyte C4d data. We thank Drs Elaine Remmers and Adriana Almeida de Jesus for critical review of the manuscript.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
- Data supplement 1 - Online supplement
Handling editor Tore K Kvien
KEL and AP have contributed equally.
Contributors CHS, AP, KEL, LGR, LP and CYY conceived and designed the study. AP, RA-A, LGR, FWM and CHS recruited patients with JDM and investigated patients’ clinical presentations. KBJ, YLW, KEL and CYY recruited healthy controls. KEL, YLW, BZ, EL, AA, DN, TPO'H and CYY performed experiments. KEL, RA-A, CHS, PW, LP, LGR and CYY performed data interpretation and analyses. KEL, CHS, LGR and CYY wrote the manuscript. All authors (except DN who is deceased) revised and approved the manuscript as written.
Funding This work was supported in part by National Institute of Arthritis, Musculoskeletal and Skin Diseases, the National Institutes of Health grant 5R01 AR054459, the CureJM Foundation, an institutional grant from the Research Institute of the Nationwide Children's Hospital, and the intramural research programme of the National Institute of Environmental Health Sciences, National Institutes of Health.
Competing interests None declared.
Patient consent Obtained.
Ethics approval IRBs from the Research Institute of the Nationwide Children's Hospital (Columbus, Ohio), and from NIAMS/NIDDK, National Institutes of Health (Bethesda, Maryland) in the USA.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement Additional data are included in the online supplementary data.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.