Article Text
Abstract
Objective To investigate whether a child’s genotype affects a mother’s risk of rheumatoid arthritis (RA) beyond the risk associated with her genotype and to test whether exposure to fetal alleles inherited from the father increases risk of RA among mothers without risk alleles.
Methods A case–control study was conducted among 1165 mothers (170 cases/995 controls) and their respective 1482 children. We tested the association between having any child with alleles encoding amino acids (AAs) associated with RA including the ‘shared epitope’ (SE) and DERAA AA sequences at positions 70–74; AA valine, lysine and alanine at positions 11, 71 and 74 of HLA-DRB1; aspartic acid at position 9 of HLA-B and phenylalanine at position 9 of DPB1. We used logistic regression models to estimate OR and 95% CI for each group of alleles, adjusting for maternal genotype and number of live births.
Results We found increased risk of RA among mothers who had any child with SE (OR 3.0; 95% CI 2.0 to 4.6); DERAA (OR 1.7; 95% CI 1.1 to 2.6); or valine (OR 2.3; 95% CI 1.6 to 3.5), lysine (OR 2.3; 95% CI 1.5 to 3.4) and alanine (OR 2.8; 95% CI 1.2 to 6.4) at DRB1 positions 11, 71 and 74, respectively. Among non-carrier mothers, increased risk of RA was associated with having children who carried DERAA (OR 1.7; 95% CI 1.0 to 2.7) and alleles encoding lysine at DRB1 position 71 (OR 2.3; 95% CI 1.5 to 4.8).
Conclusion Findings support the hypothesis that a child’s genotype can contribute independently to risk of RA among mothers.
- Rheumatoid arthritis
- epidemiology
- autoimmune disease
Statistics from Altmetric.com
Introduction
Rheumatoid arthritis (RA) affects two to three times more women than men1 2 for reasons not entirely understood.3 Various aspects of pregnancy have been investigated due to the observed overall sex dimorphism and findings indicating this time period is relevant to disease risk. Among disease-free women, an increased risk of RA onset takes place during the postpartum period.4 5 Furthermore, improvement of symptoms during pregnancy among cases has been described.6 Paternal antigens from the fetus are hypothesised to influence risk of developing RA. During pregnancy, mothers are exposed to fetal material shed from the placenta as well as fetal cells.7 Fetal cells may persist in small quantities for up to decades after delivery.8 It is possible that fetal antigens could contribute to the development of RA among some women.
The strongest genetic risk factors for RA are variants of the HLA-DRB1 gene. The ‘shared epitope’ (SE) alleles encoding the QKRRA and QRRRA amino acid (AA) sequences at positions 70–74 explain much of the genetic predisposition to RA.9 10 At the same positions, the sequence DERAA has been associated with reduced risk of RA.11 More recently, a large study demonstrated that the association between the major histocompatibility complex and RA is best explained by five AAs: three in HLA-DRB1 and one each in HLA-B and HLA-DPB1, of which two are the same as those of the shared epitope.12 Of the 16 resulting DRB1 haplotypes, valine/lysine/alanine at positions 11/71/74 is the most strongly associated with RA and it corresponds to the DRB1*04:01 allele,12 a well-documented association.13 Clinically, the SE is associated with anticitrullinated peptide antibody-positive (ACPA) RA and antibody titres,14 disease severity15 and mortality.16 The SE alleles have been shown to be more predictive of ACPA presence than of RA.17 Exposure to SE alleles through non-host genetics during pregnancy is potentially relevant to RA aetiology. Two studies found that among SE-negative mothers, SE-positive microchimerism is more frequently found among cases than controls.18 19 For this reason, we would expect to find that, among SE-negative women, cases with RA have more SE-positive children compared with controls. Exposure to DERAA through non-inherited maternal antigens is associated with decreased risk of RA.20 By analogy, it is possible that having children positive for DERAA reduces the risk of RA for mothers. The objective of this study was to investigate whether a child’s genotype affects a mother’s risk of disease beyond the risk associated with her own genotype and to test whether exposure to risk alleles of the fetus inherited from the father is associated with risk of RA among mothers who do not carry the risk alleles.
Methods
Study population
We conducted a case–control study of 1165 mothers and 1482 children using data from the University of California San Francisco (UCSF) Mother–Child Immunogenetic Study (MCIS) and research studies conducted at the Inova Translational Medicine Institute (ITMI), Inova Health System, Falls Church, Virginia. White women of European ancestry with at least one living child were eligible to participate. Cases were identified from patients enrolled in genetic studies of autoimmunity at UCSF between 1997 and 2010. All cases with RA met the 1987 revised criteria of the American College of Rheumatology (ACR)21 and had at least one live birth prior to diagnosis. Only children born prior to diagnosis were included in this study. Control mothers had no prior history of autoimmune disease and had at least one live born child. Controls were recruited from various sources including blood donors at the Blood Centers of the Pacific and the Institute for Transfusion Medicine in Pittsburgh, Pennsylvania, and from families who enrolled in studies at the Inova Women’s Hospital, Inova Fairfax Medical Center, Falls Church, Virginia. Only participants with genotype data for both mother and at least one child were included in this study. All participants provided written informed consent. The study protocol is in accordance with the Declaration of Helsinki and was approved by the UCSF and University of California Berkeley Institutional Review Boards (IRBs). The Western IRB and the Inova Health System IRB approved the ITMI studies.
Clinical and questionnaire data
For cases, we obtained the date of diagnosis and clinical characteristics from medical records. The MCIS collected data from case and control mothers on reproductive history and potential confounders through a self-administered questionnaire. For ITMI control mothers, reproductive history and mother’s and child’s date of birth were obtained from electronic medical records. Seropositive RA was defined as rheumatoid factor (RF) and/or cyclic citrullinated peptide (CCP) antibody test-positive. A combination of RF and anti-CCP was used to define seropositive RA due to the lack of available anti-CCP data for some patients; patients recruited at earlier dates primarily had RF status available documenting their diagnosis.
Human leucocyte antigen allele imputation
We used SNP2HLA22 to impute human leucocyte antigen (HLA) alleles using post-QA/QC genotype and whole-genome sequencing data. In order to minimise confounding by ancestry, we selected participants of European ancestry for inclusion in this study. Using ancestry informative markers for Northern and Southern Europeans,23 we adjusted for ancestry proportions estimated using STRUCTURE (V.2.3.4).24 A detailed description of genotyping, QA/QC steps and imputation methods can be found in online supplementary materials.
Supplemental material
Statistical analyses
We classified mothers and children as carriers (yes/no) of RA-associated alleles. We included SE alleles and alleles containing the five AAs associated with increased risk of RA as well as alleles containing the AA sequence DERAA. Figure 1 includes the alleles in each group as well as the overlap in allele groups. We created a binary variable for each mother to indicate whether she had any children positive for any of the risk-associated alleles prior to diagnosis.
We used logistic regression models to estimate ORs and 95% CIs to investigate three questions. First, we tested the independent association between allele groups and RA among mothers in a multivariable model including all the risk allele groups, without inclusion of children’s genotype. Second, we tested the association between mother’s RA case status and having any children who carried one or two alleles from each allele group. We modelled exposure to a child’s genotype using two approaches. One model considered the additional risk associated with having any allele-positive children by including maternal carrier status of the same allele group. The third model specifically addressed whether the allele inherited from the father was associated with increased risk of RA among allele-negative mothers. We used directed acyclic graphs to identify our sufficient adjustment set of variables that met the definition of a confounding variable. Maternal genetic ancestry was considered in all models, but it was not included in final models since it did not affect our estimates. The number of live births and maternal carrier status of risk alleles was included in models for the second and third questions.
We tested the robustness of our estimates through a number of sensitivity analyses. In models testing the independent effect of a child’s genotype on maternal risk of RA, we adjusted for maternal genotype by including a categorical variable for DRB1 haplotype carrier status instead of individual AA positions to evaluate if the method of defining maternal genotype impacted model estimates. The DRB1 haplotype is based on AAs at positions 11, 71 and 74 as defined in previous studies.12 25 Haplotype frequencies can be found in online supplementary table 2. Since our exposure of interest pertained to children born prior to diagnosis, we created a variable for the number of live births prior to RA diagnosis and compared the effect on our estimates for models 2 and 3. We repeated our analyses in the seropositive subgroup and same number of controls. Analyses were conducted using Stata V.13 and R.26
Results
Cases and controls differed in age at study enrolment and number of live births. (table 1). The average number of live births before diagnosis in cases was similar to the average number of births among controls (2.0±0.9 vs 1.9±1.0, p=0.09) and slightly higher at the time of interview. Clinical characteristics for cases are presented in table 1. The average age at diagnosis was 40 years of age and 69% were seropositive, per medical records. Approximately one-third had evidence of rheumatoid nodules and 55% had evidence of radiographic changes. Frequencies of additional ACR criteria in cases are in online supplementary table 1.
Maternal risk alleles and maternal RA
We evaluated the association between known risk alleles and case–control status for mothers (table 2). As expected, cases with RA were more likely to have one or two SE alleles compared with controls, 82% and 39%, respectively. Associations with DRB1 risk allele groups were in the expected direction. The SE alleles and lysine-encoding alleles at position 71 were independently associated with risk of RA. Results were very similar when seropositive cases were compared with controls (data not shown).
Children with risk alleles and maternal RA
Next, we evaluated whether having at least one child positive for any of the risk alleles of each classification (SE, DERAA and five AAs) was associated with risk of RA in the mother (table 3). Final models were adjusted for number of live births and maternal genotype (SE, DERAA and five AAs). We found a threefold increased risk of RA for mothers who had any child positive for the SE, independent of maternal genotype (OR 3.0; 95% CI 2.0 to 4.6). A twofold increase in risk for maternal RA was also present for other HLA-DRB1 alleles encoding valine at amino acid position 11, lysine at position 71 and alanine at position 74. Having a child with alleles encoding the reduced risk sequence DERAA was associated with increased risk of RA (OR 1.7; 95% CI 1.1 to 2.6). Although among mothers, carrying alleles coding for phenylalanine at position 9 of DPB1 was not associated with RA, having children with one or more alleles was associated with a fourfold increase in risk of maternal RA (OR 4.0; 95% CI 1.2 to 13.4). The effect size for DPB1 alleles was stronger in the seropositive group (OR 8.5; 95% CI 1.1 to 63.5) than among seronegative cases (OR 1.9; 95% CI 0.4 to 8.2). No other striking differences were found by serostatus (data not shown). Estimates adjusted for maternal genotype defined by presence or absence of the highest risk AA at each of the five positions (table 3) did not differ from estimates adjusted for DRB1 haplotype and HLA-B and DPB1 AAs (see online supplementary table 3).
We repeated our analyses among mothers who did not carry any of the alleles for each of the groups investigated (table 4). The association between having children who carried DERAA alleles and those for AA position 71 and RA was also observed in allele-negative mothers. However, association between children who carried SE alleles or from the other allele groups and RA was not observed. The association between children who carried one or two SE alleles among SE-negative mothers and RA was attenuated to OR 1.7 from the threefold increase in the previous model that included all mothers (table 3). Among allele-positive mothers, positive associations were present between having children positive for the SE; AA positions 11, 71 and 74 of DRB1; and AA position 9 of DPB1 and maternal RA. The association between having children with one or two DERAA alleles and RA was not statistically significant among DERAA-positive mothers (table 5).
At the individual allele level, DRB1*04:01 is strongly associated with risk of RA. DRB1*04:01 encodes valine, lysine and alanine at positions 11, 71 and 74, respectively, and it is the AA combination most strongly associated with risk of RA.12 In our study, the association between *04:01 and RA among mothers was OR 2.9 (95% CI 1.5 to 5.6), adjusting for the shared epitope, DERAA and five AAs. Having any *04:01 allele-positive children and adjusting for number of live births and mother’s genotype resulted in an OR of 2.3 (95% CI 1.5 to 3.5). Among allele-negative mothers, the association between *04:01 allele-positive children and RA was similar (OR 2.2; 95% CI 1.1 to 4.3).
Discussion
To our knowledge, this is the first report to investigate the association between a child’s genotype and maternal RA. We found increased risk of RA among women with children who carried one or two alleles encoding AAs or AA sequences of DRB1 and DPB1 molecules associated with RA, after adjusting for maternal genotype and number of live births. Thus, a child’s genotype is independently associated with maternal RA possibly through exposure to risk-associated HLA alleles. The additive effects of a child’s genotype may in part explain why women are more likely to develop RA compared with men. Female cases are less likely to carry SE alleles compared with male cases,27 28 possibly implicating non-host genetic factors in RA pathogenesis.
An increase in risk of RA was also associated with having children who carried DERAA-encoding alleles among mothers who did not carry DERAA alleles. Our findings are in contrast to what we might expect given that the sequence is associated with reduced risk of RA.11 The alleles that encode DERAA also encode alanine at position 74 of the DRB1 molecule (figure 1), which is associated with increased risk of RA. It is possible that the association is not due to DERAA but only to alanine at position 74 and other AAs at different positions. However, our estimates are adjusted for maternal genotype at these other AA positions. Another possible explanation is that the observed association is due to one of the DERAA alleles. We excluded one allele at a time, and all estimates were within the 95% CI for the reported DERAA ORs. Therefore, the observed association is not due to a single allele. Increased risk associated with exposure during pregnancy is consistent with a mechanism mediated by DRB1-derived epitopes. T-cell cross-reactivity with the DERAA sequence of microbial as well as of self-origin has been identified in patients with RA.29 Likewise, it is possible that exposure to fetal DERAA during pregnancy affects maternal risk of RA through molecular mimicry. Fetal antigens could contribute to the process of epitope spreading prior to disease onset.30 More work is needed to understand the biological mechanisms underlying the association between DRB1 alleles and risk of RA in general.
In our study, the SE alleles were strongly associated with RA among mothers. The observed SE frequency among cases (82%) and controls (39%) was similar to the range (69%–80% and 42%–45%) reported in previous studies.14 31–33 The estimate for maternal SE status and RA was attenuated once we took into account having any SE-positive children (OR 3.1; 95% CI 2.0 to 4.7). These findings support the hypothesis that a child’s genotype can independently contribute to risk of maternal RA.
SE-positive women are more likely to have children who carry at least one SE allele. This is evident in both cases and controls. Since more cases with RA carry SE alleles than controls, cases with RA are more likely to have SE-positive children than controls (88% vs 62%, respectively). This translates to an increase in exposure through non-host genetics of 6% among cases with RA compared with 23% among controls. Results from our logistic regression models suggest that the increase in risk associated with having SE-positive children (and other alleles associated with risk) is not entirely due to the difference in maternal genotype since it is accounted for in our models. Our results support previous work that demonstrated a dose effect of SE alleles.14 31 RA is a complex disease likely caused by a combination of genetic and environmental factors. Exposure to children’s SE alleles, regardless of their origin, could serve as one of many environmental ‘hits’ contributing to RA pathogenesis.
Among mothers who did not carry the alleles, we observed an association between RA and having children who carried DERAA, lysine at position 71 and DRB1*04:01. The small number of allele-negative cases for various groups including the SE, led to a lack in precision for some estimates. We found an 11% difference in the frequency of SE-positive children that may help explain the excess of SE-positive microchimerism among cases compared with controls previously reported.18 19 One limitation of our study is that we did not have measures of microchimerism to test whether maternal and/or fetal genotype combinations influence its presence and quantity.
In a previous study of patients with systemic lupus erythematosus (SLE), an increase in risk of maternal SLE associated with children who carried DRB1*04:01 was observed.34 DRB1*04:01 does not have a strong association with SLE but it does share sequence similarities with the Epstein–Barr virus (EBV)35; EBV is a risk factor for SLE.36 Studies in RA have demonstrated that ACPAs react with EBV viral sequences and may contribute to disease-associated antibody formation.37 It is possible that EBV-DRB1*04:01 molecular mimicry may trigger or contribute to autoimmunity.
Our mother–child study had many strengths. Cases with RA and controls were clinically well characterised; comprehensive reproductive histories were obtained for each participant, genetic data were collected and HLA genotypes were derived for classical loci using established computational methods. We performed QA/QC measures that increase confidence in our findings. We took into account potential confounding variables in our analyses. Similar results were obtained when adjusting for the number of live births before diagnosis for cases rather than their total live births reported at the time of interview.
Despite the large overall number of mother–child pairs, we had limited sample sizes for some of the allele groups we tested. We did not correct p values for multiple comparisons due to the lack of independence between allele groups (figure 1). Among SE-negative mothers, we had only 30% power to detect an association of the observed magnitude or greater. ITMI controls were younger at the time of study enrolment and had fewer births. However, comparing the number of children born prior to diagnosis, cases with RA and controls did not differ, and inclusion of any version of number of live births in our models did not affect our results. Potential misclassification due to younger age of controls would bias results towards the null; RA is a relatively rare disease, and population rates would be expected to apply to our control group. Results by serostatus are limited by the lack of complete anti-CCP data, and future studies are needed to confirm and extend these findings. Our study was conducted among women who have given birth to a child and therefore apply to a subset of cases who became pregnant before RA diagnosis.
In conclusion, exposure to a child’s genotype during pregnancy may contribute to risk of RA among mothers. Non-host genetic exposure may be relevant to consider in understanding RA pathology. Functional studies are needed to characterise the biological pathways that can explain our observations.
Acknowledgments
We thank Ann Guiltinan and Ed Murphy from Blood Centers of the Pacific; and Ram Kakaiya, MD, and Pam D’Andrea, RN, from the Institute for Transfusion Medicine, Pittsburgh, Pennsylvania.
References
Footnotes
Contributors Study design, drafting and interpretation: GIC, LAC, LFB. Data analysis: GIC, XS, NAP. Data collection, revising manuscript and approval to submit: HQ, KAH, KS, JAN, MPB, DJT, WSWW, BDS, JEN.
Funding Funding provided by the National Institute of Allergy and Infectious Diseases (NIAID) grants R01AI059829, R21AI117879, R01AI065841, F31AI116064; the Robert Wood Johnson Foundation Health & Society Scholars Program and the Rheumatology Research Foundation’s Health Professional Research Preceptorship award.
Competing interests None declared.
Ethics approval UC San Francisco, UC Berkeley, Inova Health Systems and the Western Institutional Review Board.
Provenance and peer review Not commissioned; externally peer reviewed.