|Home | About | Journals | Submit | Contact Us | Français|
Serum galactose-deficient IgA1 (Gd-IgA1) is an inherited risk factor for adult IgA nephropathy (IgAN). The goal of this study is to determine the heritability of serum Gd-IgA1 levels in children with IgAN and Henoch-Schönlein nephritis (HSPN). For this purpose, we obtained serum from 34 Caucasian families (20 pediatric cases of HSPN, 14 pediatric cases of IgAN, and 54 of their first-degree relatives), as well as from 51 age- and ethnicity-matched pediatric controls and 141 healthy adult controls. Serum Gd-IgA1 levels were quantified using an HAA-lectin-based ELISA. Children with either IgAN or HSPN had significantly higher Gd-IgA1 levels compared with pediatric controls (p = 1.7 × 10−7 and p = 6.5 × 10−9, respectively). Serum levels of Gd-IgA1 were also elevated in a large fraction of the first-degree relatives of pediatric IgAN and HSPN patients compared with unrelated adult controls (p = 3.2 × 10−6 and p = 5.1 × 10−4, respectively). The unilineal transmission of the trait was observed for 75% of families, bilineal transmission in 5%, and sporadic occurrence in 20%. The age-, gender-, and household-adjusted heritability of serum Gd-IgA1 level was estimated at 76% (p = 0.021) in pediatric IgAN patients and at 64% (p = 0.018) in HSPN patients. Our data demonstrate that serum Gd-IgA1 levels are highly inherited in pediatric IgAN and HSPN, providing support for yet another shared pathogenic link between these disorders.
IgA nephropathy (IgAN) was first described in 19681 and is considered to be the most common type of primary glomerulonephritis in the world.2 Henoch-Schönlein purpura (HSP) is the most frequent form of vasculitis in children, with renal involvement in up to 40% of cases.3 About 3% of children with HSP nephritis (HSPN) progress to end-stage renal disease.4 The hypothesis that IgAN and HSPN represent clinical phenotypes that share a common pathogenic mechanism is strongly supported by evidence from the indistinguishable renal immunohistopathology5–7 and clinical observations.7–10 Genetic factors are known to play an important role in susceptibility to IgAN, and multiple extended pedigrees with familial forms of the disorder have been reported worldwide.11–13 Interestingly, patients with HSPN have been documented in several pedigrees of related patients with IgAN.8, 14 This observation suggests that the same genetic factors that are involved in IgAN may also operate in the pathogenesis of HSPN.
The pathogenetic entity shared by IgAN and HSPN is aberrant glycosylation of O-linked glycans in the hinge region of a fraction of IgA1 molecules.15–17 Rather than terminating with galactose, the aberrant galactose-deficient O-glycans end with N-acetylgalactosamine (GalNAc) or sialylated GalNAc. The terminal GalNAc moiety on the aberrantly glycosylated IgA1, in turn, may be recognized by anti-glycan antibodies,18 leading to formation of nephritogenic circulating immune complexes that subsequently deposit in the glomerular mesangium to induce kidney injury.19 Serum galactose-deficient IgA1 (Gd-IgA1) level is elevated in patients with IgAN regardless of ethnicity or age,20–24 and is also elevated in children with HSPN.21, 25 Serum Gd-IgA1 levels appear to be heritable in a dominant pattern in IgAN, although most relatives with high levels have no clinical manifestation of renal injury.26 Inheritance of high levels of circulating Gd-IgA1 has recently been demonstrated in cohorts of Caucasian, Asian, and African-American families with IgAN.23, 24, 26 The purpose of this study is to determine whether the serum levels of Gd-IgA1 in children with HSPN and IgAN are similarly heritable, providing further evidence for a shared mechanism in the pathogenesis of the two diseases.
For the purpose of this study, we recruited 34 pediatric patients with HSPN or IgAN, and their first-degree relatives (20 pediatric cases of HSPN, 14 pediatric cases of IgAN, and 54 of their relatives), as well as 51 healthy pediatric controls and 141 healthy adult controls (Table 1). All participants were of Caucasian ethnicity and were recruited from the same geographic region. The clinical characteristics of the enrolled cases are summarized in Table 2. All 14 cases of IgAN and 11 of 20 cases of HSPN were diagnosed by renal biopsy. The remaining 9 cases of HSPN were diagnosed based on clinical presentation (Supplementary Table 1). The group of pediatric patients with HSPN was slightly younger compared with children with IgAN (mean age 10.1 vs. 14.3 years, p=0.002). However, there were no statistically significant differences between the two case groups in terms of gender, body mass index, blood pressure, estimated glomerular filtration rate (eGFR), microscopic hematuria, or degree of proteinuria.
Circulating IgA levels have been reported to increase with age, a pattern that may be attributable to the cumulative number of mucosal antigenic exposures and/or normal immune aging process.27 In addition, prior studies noted that serum Gd-IgA1 levels are lower in healthy children compared with healthy adults.21, 24, 25 As serum total IgA and Gd-IgA1 levels were highly correlated in this study, we first explored the relationship with age for both measurements in our cohorts (Figure 1). Total IgA and Gd-IgA1 levels were significantly higher in sera of adult controls compared with pediatric controls (p < 0.0001 for both traits). The levels of total IgA and Gd-IgA1 also positively correlated with age within the pediatric healthy-control group (r = 0.42, p = 0.002 for total IgA; r = 0.29, p = 0.04 for Gd-IgA1). Among cases, however, older age was not associated with higher levels, suggesting that the effect of age may be masked by other factors with substantially larger effects in the affected children. Considering a significant effect of age among healthy subjects, we utilized a separate control group for adult and pediatric subjects.
Children with either IgAN or HSPN had significantly higher serum Gd-IgA1 levels compared with pediatric healthy controls (p = 1.7 × 10−7 and p = 6.5 × 10−9, respectively). These differences were robust to additional adjustment for age and gender (Figures 2c and d). There were no significant differences in serum Gd-IgA1 levels between children with IgAN and children with HSPN (p = 0.50). In addition, these differences were less pronounced for serum total IgA levels, providing support for serum Gd-IgA1 being the better disease biomarker. The sensitivity and specificity analysis of serum Gd-IgA1 level confirmed its superior predictive characteristics compared with serum total IgA level (Figures 2a and b). In pediatric IgAN patients, the area under the receiver operating characteristics (ROC) curve was estimated at 0.96 (95% CI: 0.91–1.00) for Gd-IgA1 levels and 0.85 (95% CI: 0.73–0.98) for total IgA levels. In HSPN, the area for Gd-IgA1 was 0.94 (95% CI: 0.90–0.99), and for total IgA, 0.87 (95% CI: 0.78–0.96). Additional adjustments of serum measurements for age and gender did not substantially alter the shape of the ROC curves.
Serum levels of Gd-IgA1 were elevated in the first-degree relatives of pediatric IgAN and HSPN patients compared with adult healthy controls (p = 3.2 × 10−6 and p = 5.1 × 10−4, respectively; Figure 3a). In total, 11 of 25 (44%) relatives of IgAN patients and 9 of 29 (31%) relatives of HSPN patients had levels above the 95th percentile for healthy adult controls (Table 1). In addition, the Gd-IgA1 levels did not differ between the cases and their relatives (p = 0.8 and p = 0.5, respectively; Figures 3b and 3c). We also observed a strong positive correlation between the serum Gd-IgA1 levels for index cases and parents (r = 0.42, p = 1.4 × 10−3; Figure 3d). When we stratified the relatives based on the Gd-IgA1 level in the index cases, the relatives of cases with high levels (above 95th percentile for pediatric controls) had higher levels as compared with those of the relatives of index cases with normal Gd-IgA1 levels (p = 0.042, Figure 3e). Next, we analyzed the transmission of high serum Gd-IgA1 level in the 23 trios (both parents and the patient) available for analysis. The unilineal transmission of the trait was observed in 17 of 23 (~75%) families, bilineal transmission in 1 (~5%) family, and sporadic occurrence in 5 (~20%) families. This pattern is consistent with an autosomal dominant inheritance with incomplete penetrance, but high sporadic rate is also apparent. We observed no appreciable differences in the clinical characteristics of cases with high versus low/normal Gd-IgA1 levels (Supplementary Table 2).
The heritability was modeled using variance component methods.28 The distribution of Gd-IgA1 levels in families was positively skewed, and required normalization by logarithmic transformation prior to modeling (Supplementary Figure 1). The significant covariates in the polygenic models of Gd-IgA1 included: age (p = 0.016), sex (p = 0.038), and household membership (p = 0.015). In the combined cohorts of HSPN and IgAN, age and sex alone explained approximately 8% of the variability, while household effects accounted for approximately 5% of the overall phenotypic variance in Gd-IgA1 levels. Collectively, these covariates explained 14.4% of the trait’s variance in the fully adjusted polygenic model. None of the pairwise interaction terms between age, sex, or household membership tested significant in the polygenic models. In the combined cohorts, the disease type (cohort membership) or BMI (obesity) were not statistically significant. These covariates and their interaction terms were subsequently dropped from the final models.
The estimated heritability of the serum Gd-IgA1 level was substantial and statistically significant in both cohorts (Table 3). Age-, sex-, and household-adjusted heritability was estimated at 63.7% (p = 0.018) in the HSPN cohort, 76.0% (p = 0.021) in the IgAN cohort, and 70.5% (p = 0.0017) in both cohorts combined. In contrast to the Gd-IgA1 levels, the unadjusted heritability of total IgA levels was low and failed to reach statistical significance, even after combining the cohorts (h2 = 0.215, p = 0.17). Additional adjustments for the disease type, age, gender, BMI, and household effects did not improve this estimate (h2 = 0.132, p=0.29). Moreover, addition of serum total IgA as a covariate in the fully adjusted polygenic model of Gd-IgA1 did not substantially alter its heritability (h2 = 0.685, p=0.0017); thus, high heritability of serum Gd-IgA1 level appears to be independent of any genetic factor that may regulate serum total IgA level. Estimation of heritability for age- and sex-adjusted BMI in our cohort (h2 = 0.50, p = 0.015) demonstrated values consistent with published data in other cohorts29–32, demonstrating lack of bias in our calculations.
This study is the first one to demonstrate that circulating levels of Gd-IgA1 are highly heritable in children with IgAN and HSPN. High heritability of Gd-IgA1 in our cohort is consistent with three other family-based studies of adult IgAN involving Caucasians26, Asians23, and, most recently, African Americans24. Our data support a shared inherited defect in the glycosylation of circulating IgA1 as a central unifying link in the pathogenesis of IgAN and HSPN. Similarly as for adult IgAN patients, we postulate that the development of high levels of Gd-IgA1 is likely antecedent to the disease process in children with IgAN and may represent an inherited risk factor for nephritis due to mesangial deposition of IgA1.13, 33 However, we also note that high Gd-IgA1 levels were observed in many asymptomatic first-degree relatives of the affected children. Therefore, a high serum Gd-IgA1 level is clearly not sufficient for the development of clinical symptoms, and it is likely that a “second hit” (i.e., other environmental or inherited risk factor) is required to produce overt disease.33 Recent studies suggest that such a “second hit” may be related to the development of anti-glycan antibodies that recognize the galactose-deficient hinge-region O-glycans of IgA1 to form nephritogenic immune complexes in patients with IgAN18 as well as HSPN34. Additionally, we observed a smaller subgroup of affected children with normal Gd-IgA1 levels (Supplementary Table 2). This subgroup may represent a disease category with distinct pathogenesis that is independent of IgA1 glycosylation defects. Larger studies will be required to better characterize this subgroup of patients.
In our study, serum Gd-IgA1 levels were similarly elevated in children with IgAN or HSPN. In both disorders, the ROC curve for the Gd-IgA1 level was superior compared with the ROC curve for serum total IgA. These observations are consistent with prior reports of detectable glycosylation defects in patients with Henoch-Schönlein purpura with renal involvement.21, 25 Furthermore, the area under the ROC curve for serum Gd-IgA1 appeared to be greater for pediatric patients (AUC ~0.94–0.96) when compared with adults with IgAN (AUC ~0.90).20 Although the age adjustment did not significantly alter the ROC curves, the effect of age on Gd-IgA1 levels in healthy pediatric controls was statistically significant. Increasing levels of total IgA with age in healthy children have been well described in the literature; total IgA concentration is lowest at birth and mean adult levels are typically not reached until after puberty.35–37 The effect of age on adult IgA levels is much less pronounced.27 Our data indicate that serum Gd-IgA1 levels appear to follow a similar pattern in healthy children and adult controls. Because of the correlation of Gd-IgA1 levels with age, standardized age-specific normal ranges will likely need to be developed for increased diagnostic utility of the serum test in children. This study is not sufficiently powered to develop such ranges, and larger studies of healthy pediatric populations will be required to develop age-normative data.
We also recognize several other limitations of our study. Most importantly, the studied cohorts are relatively small and limited to participants of Caucasian ancestry from a small region of the United States. Despite small sample size, however, we achieved statistical significance of heritability estimates, largely because of a substantial genetic component to the variability in serum Gd-IgA1. Additional limitations stem from the method of heritability estimation used in this study. Here, we examine the narrow sense heritability of Gd-IgA1, defined as the proportion of phenotypic variance explained by the additive genetic component.38 This model ignores the effects of dominance, epistasis, or gene-environment interactions, although these effects are likely to cause only a downward bias in heritability estimates.39 A more precise dissection of heritability could be achieved in twin studies and more complete sets of larger pedigrees, but such cohorts are not readily available for pediatric IgAN or HSPN. Approximately 20% of total variance in the serum Gd-IgA1 level could not be explained by either the additive genetic component or modeled confounders. This finding may be due to assay variability, non-additive genetic effects, random fluctuations of Gd-IgA1 levels, or other environmental/epigenetic factors. Lastly, we recognize that our heritability estimates are strictly relative and population-specific. Studies in cohorts of diverse ethnic backgrounds and patients residing in different geographic regions will be needed to establish the generalizability of our findings.
We conclude that a serum Gd-IgA1 level is, in part, genetically determined and may constitute a useful tool for screening and stratification of pediatric patients at risk for HSPN or IgAN. Age may represent an important confounder in the studies of pediatric populations, and should be taken into account in the interpretation of serum Gd-IgA1 levels. Our findings of high heritability of Gd-IgA1 in pediatric patients are consistent with prior studies involving adult cases of IgAN.26 In aggregate, our observations highlight potential clinical utility of Gd-IgA1 testing to identify individuals at genetic risk of nephropathy. Additional studies will be needed to determine if a high serum Gd-IgA1 level correlates with any specific clinico-pathologic feature, or differential response to treatment. For this reason, we strongly advocate for inclusion of serum Gd-IgA1 levels in the evaluation and follow-up of patients in randomized controlled trials for treatment of IgAN or HSPN. The acquisition of prospective longitudinal data will help to define the prognostic utility of this test. Lastly, our data lay a basis for future quantitative genetic mapping studies of Gd-IgA1 aiming at identification of specific gene(s) responsible for this phenotype. Identification of genes and pathways responsible for aberrant glycosylation of IgA1 may ultimately lead to the development of novel therapeutic and prophylactic approaches for these common childhood disorders.
Our study included families recruited at the University of Tennessee Health Sciences Center in Memphis. In total, we analyzed 33 Caucasian families (14 trios with pediatric IgAN, 18 trios with pediatric HSPN, and 1 nuclear family with two siblings affected by HSPN), as well as 51 age- and ethnicity-matched pediatric controls and 141 healthy adult controls (Table 1). A single blood collection from patients, their relatives, and healthy controls was performed for the purpose of measurement of serum total IgA and Gd-IgA1. Screening urine dipstick test was performed in all study participants at the time of recruitment. All healthy controls included in this study had a negative urine test for blood and protein. Spot urine protein and creatinine concentrations were measured in all cases and proteinuria was defined as a urinary protein/creatinine ratio > 0.15 g/g. Estimated GFR was calculated with the 4-variable Modification of Diet in Renal Disease (MDRD) formula for adults40 and the Schwartz formula for pediatric patients.41 We considered chronic kidney disease (CKD) to be present in individuals with proteinuria or eGFR < 60 mL/min/1.73m2. Microscopic hematuria was defined as ≥ 1+ blood on urine dipstick test or > 5 RBC per high-power field on microscopic examination of the spun urinary sediment. Macroscopic hematuria was defined by clinical history. Body mass index (BMI) was calculated as weight (kg) / height (m)2; obesity was defined by BMI ≥ 30 kg/m2 in adults, and above 95th percentile for age- and sex-matched healthy children of Caucasian ethnicity. Adults with BMI 25–30 kg/m2 were considered overweight. Similarly, children with BMI between 85th and 95th percentile for age- and sex-matched healthy children were classified as overweight and at risk of obesity, according to the guidelines of the American Academy of Pediatrics (AAP).42 Consequently, we coded BMI as a categorical variable with 3 levels (normal/overweight/obese) for the purpose of polygenic modeling across family members of varying age. All kidney-biopsy diagnoses were confirmed by review of the original pathology reports. The clinical (non-biopsy) diagnosis of HSPN was defined by the presence of characteristic purpuric skin lesions of HSP in combination with heme-positive dipstick test (≥1+) and confirmed microscopic hematuria (> 5 RBCs / hpf). The clinical characteristics of the enrolled patients are summarized in Table 2 and Supplementary Table 1. All patients or their parents/guardians provided written informed consent, and participants aged 8 to 18 years provided written assent. The IRB committees at the participating institutions approved the study protocol.
Serum levels of IgA were determined by capture ELISA.20, 43 ELISA plates were coated with F(ab’)2 fragment of goat IgG specific for human IgA (Jackson ImmunoResearch Inc.) at 1 ug/ml. The bound human IgA was then detected with F(ab’)2 fragment of goat IgG specific for human IgA (Biosource). The IgA concentration was expressed in mg/ml, based on standard curves generated with calibrated human IgA standard.
The measurements of serum Gd-IgA1 were performed using a lectin-based ELISA, as previously described in detail.19 Briefly, the IgA from samples and standard captured on ELISA plates were treated with 1 mU per well neuraminidase for 3 hr at 37°C. Samples were then incubated with biotinylated GalNAc-specific lectin from Helix aspersa (HAA; Sigma-Aldrich). The bound IgA1 was detected with avidin-horseradish peroxidase conjugate and the reaction was developed with the peroxidase chromogenic substrate o-phenylenediamine-H2O2 (OPD-H2O2) (Sigma-Aldrich). The concentration of Gd-IgA1 was calculated by interpolating the optical densities on calibration curves constructed using neuraminidase-treated standard Gd-IgA1 myeloma protein. The results were expressed in units/ml, where 1 unit of Gd-IgA1 was defined as 1 ug of the standard Gd-IgA1 protein.
Summary statistics were calculated for demographic, clinical, and biochemical characteristics and were expressed as proportions/percentages for categorical data and means (+/− standard deviations) or medians (ranges) for quantitative data. Pearson’s correlation coefficients were used to explore relationships between quantitative variables. Partial correlation coefficients were used to adjust for selected covariates. Pairwise group comparisons were performed using 2-sided Student’s t-test (or Mann Whitney U test for non-normally distributed data) and Chi-sq test for categorical data. Multiple group comparisons in quantitative characteristics were performed using ANOVA, or Wilcoxon’s rank-sum test for non-normally distributed data. Two-tailed P < 0.05 was considered statistically significant. Multiple regression and ROC analyses were performed in SPSS Statistics 17.0.
Before polygenic modeling, all traits were tested for normality by Shapiro-Wilk test and visual examination of histograms and quantile-quantile (qq) plots. All non-normal trait distributions were normalized by logarithmic transformation (Supplementary Figure 1). Summary statistics (mean, SD, kurtosis) were derived for all transformed traits before and after covariate adjustment. The covariates were screened by stepwise selection in polygenic models of quantitative traits. The step criteria included entry and removal p-value of 0.05 (likelihood ratio test). Age, sex, and BMI, as well as all of their first order interaction terms, were tested for significance. We also accounted for shared family environment by modeling household effects. The polygenic analyses were performed using variance components methods, as implemented in SOLAR version 220.127.116.11, 44 Heritability estimates were derived assuming additive genetic variance component. Summary statistics, normality testing, transformations, and linear regression analyses were performed with SPSS Statistics 17.0.
We thank our patients and their families for participation in this study. Krzysztof Kiryluk was supported by the Daland Fellowship from the American Philosophical Society and Grant Number KL2 RR024157 from the National Center for Research Resources (NCRR), a component of the National Institutes of Health (NIH), and NIH Roadmap for Medical Research. Robert J. Wyatt, Jan Novak, Bruce A. Julian, Zina Moldoveanu, and Ali G. Gharavi are supported by Grant Number DK082753 from the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK). John T. Sanders was supported by a fellowship training award from the National Kidney Foundation. The authors also acknowledge other grants from NIDDK supporting their research of IgAN: DK078244, DK080301, DK075868, DK083663, DK071802, and DK077279. The contents of this publication are solely the responsibility of the authors and do not necessarily represent the official view of NCRR, NIDDK, or NIH.
Conflicts of Interest: