|Home | About | Journals | Submit | Contact Us | Français|
Conceived and designed the experiments: SMO QPH LH RSH. Performed the experiments: SMO QPH. Analyzed the data: SMO QPH LH RSH. Contributed reagents/materials/analysis tools: SMO QPH LH RSH. Wrote the paper: SMO RSH. Edited the paper: QPH LH.
Despite considerable effort, the genetic factors responsible for >90% of the antibody deficiency syndromes IgAD and CVID remain elusive. To produce a functionally diverse antibody repertoire B lymphocytes undergo class switch recombination. This process is initiated by AID-catalyzed deamination of cytidine to uridine in switch region DNA. Subsequently, these residues are recognized by the uracil excision enzyme UNG2 or the mismatch repair proteins MutSα (MSH2/MSH6) and MutLα (PMS2/MLH1). Further processing by ubiquitous DNA repair factors is thought to introduce DNA breaks, ultimately leading to class switch recombination and expression of a different antibody isotype.
Defects in AID and UNG2 have been shown to result in the primary immunodeficiency hyper-IgM syndrome, leading us to hypothesize that additional, potentially more subtle, DNA repair gene variations may underlie the clinically related antibody deficiencies syndromes IgAD and CVID. In a survey of twenty-seven candidate DNA metabolism genes, markers in MSH2, RAD50, and RAD52 were associated with IgAD/CVID, prompting further investigation into these pathways. Resequencing identified four rare, non-synonymous alleles associated with IgAD/CVID, two in MLH1, one in RAD50, and one in NBS1. One IgAD patient carried heterozygous non-synonymous mutations in MLH1, MSH2, and NBS1. Functional studies revealed that one of the identified mutations, a premature RAD50 stop codon (Q372X), confers increased sensitivity to ionizing radiation.
Our results are consistent with a class switch recombination model in which AID-catalyzed uridines are processed by multiple DNA repair pathways. Genetic defects in these DNA repair pathways may contribute to IgAD and CVID.
Upon antigen stimulation, the constant region exons of the expressed antibody heavy chain gene can be replaced with downstream ones that encode an alternative antibody isotype (IgM -> IgG, IgE or IgA). At the molecular level, antibody class switch recombination (CSR) occurs between C/G-rich switch regions upstream of each set of constant region exons (Figure 1). In recent years a consensus model has emerged whereby CSR is initiated by activation-induced deaminase (AID)-catalyzed cytidine deamination to uridine within switch region DNA –. These DNA uridines are subsequently recognized and removed by the base excision repair enzyme uracil DNA glycosylase 2 (UNG2) or the mismatch repair proteins MutSα and MutLα (heterodimers of MSH2/MSH6 and MLH1/PMS2, respectively) , –. Additional base excision repair, mismatch repair, and recombination repair factors are then proposed to help convert these DNA repair intermediates to double-strand breaks and ultimately to CSR products –.
A subset of primary antibody deficiency syndromes has been explained by defects in these DNA repair pathways. For instance, hyper-IgM syndrome type 2 (HIGM2) is caused by mutations in AID and this disease is characterized by high levels of IgM at the expense of the other antibody isotypes (; reviewed in ). Mutations in UNG2 lead to the less severe HIGM5 , and defects in PMS2 have been associated with decreased antibody production . Varying degrees of antibody deficiency have also been noted in chromosomal instability syndromes such as ataxia-telangiectasia (A-T, ATM mutations), Nijmegen breakage syndrome (NBS, NBS1 mutations), and ataxia-telangiectasia-like disorder (ATLD, MRE11 mutations) –.
Prior studies have shown that missense mutations that impair MSH5 binding to its obligate heterodimerization partner MSH4 associate with immunoglobulin A deficiency (IgAD) and common variable immunodeficiency (CVID) . IgAD and CVID often occur in different individuals of the same family, suggesting a common genetic components in at least a subset of patients . Mutations in the B cell surface receptor genes TACI – and CD19 , , and the T cell receptor gene ICOS  are also responsible for a subset of CVID cases. Nevertheless, despite this considerable progress, the genetic causes of >90% of IgAD and CVID cases have yet to be identified (reviewed in –).
Given that defects in DNA metabolism proteins are responsible for a fraction of primary antibody deficiencies, we hypothesized that variations in DNA repair genes could also underlie additional IgAD and CVID cases. To test this hypothesis, we performed a genetic association study of 27 candidate DNA metabolism genes in IgAD/CVID. Based on those results, we sequenced selected genes in a panel of 96 IgAD/CVID patients and subsequently genotyped the non-synonymous alleles we identified in a large case-control association study. Novel coding and non-coding variations were detected in several genes, including seven rare variants found only in IgAD and two specific to CVID. Four of these mutations, MLH1-S247A, MLH1-Q409P, NBS1-P401R, and RAD50-Q372X showed significant association with IgAD/CVID. One IgAD patient carried three previously unreported mutations, MSH2-A727S, MLH1-G638R, and NBS1-P401R. A second IgAD patient was compound heterozygous for RAD50-P165H and the previously reported RAD50-R327H variation. Two novel single nucleotide polymorphisms (SNPs) in the 3′ untranslated region of MRE11 were also associated with CVID. Finally, both patient-derived and purposefully engineered cells harboring the RAD50-Q372X mutation exhibited increased sensitivity to ionizing radiation.
To screen for candidate genes in IgAD and CVID, we genotyped 140 IgAD patients, 48 CVID patients, and 92 healthy controls for SNPs selected from 26 known DNA repair genes and AID (collectively these genes define several DNA metabolism pathways). To test if a given SNP was associated with IgAD and/or CVID, allele frequencies were compared to the healthy control cohort. Significant single marker associations with IgAD and/or CVID (p<0.01) were noted for SNPs in the mismatch repair complexes, MutSα, MutSβ, and MutSγ, the MRN complex, the extended RAD52 epistasis group, and AID. A summary of single marker associations with p-values <0.01 is presented in Table 1. Full association data can be found in Table S1.
To further refine the list of candidates, we performed a relaxed correction for multiple testing by permutation testing markers grouped by DNA repair complex or pathway. After applying this statistical method, 13 SNPs remained significantly associated with IgAD/CVID (Table 1). Markers in RAD54B (rs3019279), AID (rs2580874 and rs714629) showed significant association with the combined IgAD/CVID cohort, and SNPs in MSH2 (rs3771276 and rs6729015), RAD50 (rs2237060), and RAD52 (rs10849605) were associated with IgAD. In agreement with our prior studies, six markers in the MHC class III region gene MSH5 were associated with IgAD .
Next, we constructed multi-marker haplotypes and tested for association with IgAD and/or CVID. The markers comprising each haplotype block are listed in Table S2 and association data for haplotypes with a frequency greater than 0.1% are reported in Table S3. Significantly associated haplotypes with an uncorrected p-value less than 0.01 are summarized in Table 2. After pathway-based permutation testing, haplotypes of MLH1, MSH5, and MUS81 were associated with IgAD and the combined IgAD/CVID group. Association with CVID was noted for haplotypes of MSH2 and RAD50.
Since notable IgAD/CVID associations were observed for genes encoding DNA repair proteins that could convert AID-catalyzed uracils into DNA breaks, we resequenced select DNA repair genes in IgAD and CVID patient samples. Specifically, we sequenced the MutSα genes MSH2 and MSH6, the MutLα gene MLH1, and the MRN complex genes MRE11, RAD50, and NBS1. The genes encoding AID, APEX1, ERCC1, and RAD52 were also sequenced.
In all, 242 genetic variants were detected, of which, 93 did not have records in dbSNP or SwissProt and were therefore considered novel (Table S4). 24 alleles encoded amino acid changes, 13 of which were previously unreported (all detected in a heterozygous state). Of the remaining 80 novel SNPs, 5 were synonymous coding alleles, 31 were located in mRNA untranslated regions (UTRs), 40 were intronic, and 4 were in flanking non-transcribed regions.
We next determined the allele frequencies of identified non-synonymous alleles and selected UTR, synonymous, and intronic SNPs in a large cohort of healthy controls, IgAD patients, and CVID patients. The most significant data are summarized in Table 3, and full datasets can be found in Table S5. In the mismatch repair pathway, two novel MSH2 variations were discovered in IgAD samples: MSH2-T292S and MSH2-A727S. The latter variation was unique to a single IgAD patient. Four novel amino acid mutations were detected in MLH1. MLH1-S247A and MLH1-G638R were specific to IgAD and MLH1-Q409P was specific to CVID (Table 3). MLH1-H727L and the previously reported mutations MLH1-K618A, MLH1-V716M, and MLH1-I219V occurred at similar frequencies in IgAD/CVID patients and controls.
The most striking variations were observed in RAD50. In all, 16 novel SNPs were found: 3 encoded amino acid changes, 2 were located in the 5′ UTR, 1 was synonymous, and 10 were intronic (Table S4). Most notably a single CVID sample contained the heterozygous transition mutation (1114C>T) that creates a premature stop codon at amino acid 372 (RAD50-Q372X, Table 3). One IgAD patient carried RAD50-R1077Q and another had both RAD50-P165H and RAD50-R327H. RAD50-R327H was also detected in our control samples, but never linked with RAD50-P165H. Pedigree analysis of the RAD50-P165H/R327H patient revealed that the two alleles were present in a compound heterozygous state.
Novel variations were also detected in the other two MRN complex genes, MRE11 and NBS1. In MRE11, 18 novel SNPs were detected (Table S4). Of these, two resulted in amino acid changes (MRE11-A492D and MRE11-E494K), eight were in the 3′ UTR, and eight were intronic. Both MRE11-A492D and MRE11-E494K were found at similar frequencies in controls and IgAD/CVID cases (Table 3). Two 3′ UTR SNPs, ss206257789 and ss206257804, were, however, significantly enriched in the CVID population. In NBS1, 11 unreported SNPs were detected (Table S4). Three were non-synonymous, and the remaining 8 were a mixture of synonymous, 5′ UTR, 3′ UTR, and intronic. Two novel alleles, NBS1-P401R and NBS1-D527Y, were specific to IgAD (Table 3). NBS1-V210F and NBS1-L421S were detected in IgAD/CVID and control samples, as was the previously reported common polymorphism, rs1805794 (NBS1-E185Q).
A number of previously unreported SNPs were also discovered in other genes. In AID, two novel 3′ UTR SNPs and three novel synonymous SNPs were detected (Table S4). Two novel SNPs were found in APEX1, one synonymous and one in the 5′ UTR. The previously reported 5′ UTR SNP rs2307490 showed modest association with CVID (Table 3). In ERCC1, one 3′ flanking SNP and five novel intronic SNPs were detected (Table S4). 18 new SNPs were discovered in RAD52, 14 in the 3′ UTR, and four in introns. No amino acid substitution mutations variations were detected in these four genes.
While no patients were homozygous for any single SNP, some patients were found to carry mutant alleles in more than one DNA repair gene. An IgAD patient heterozygous for the previously reported APEX1-Q51H allele also carried three novel mutations, MLH1-G638R, MSH2-A727S, and NBS1-P401R. One IgAD patient carried both RAD50-R1077Q and MLH1-K618A. Additionally, as described earlier, an IgAD patient was found to be biallelic for RAD50-P165H and RAD50-R327H. These observations hint that the combinatorial effects of multiple mutations may underlie a subset of IgAD/CVID cases.
Since CSR is likely to involve DNA breakage and end processing by exonucleases, we next decided to test the effect of one of the identified mutations (RAD50-Q372X) on cellular DNA repair capacity. To this end, immortalized B lymphoblast lines were generated from the CVID patient carrying RAD50-Q372X. Immunoblotting experiments indicated that RAD50 protein levels were reduced approximately 2-fold compared to three similarly derived cell lines from individuals with no detected RAD50 mutation (Figure 2a). MRE11 levels were also reduced approximately 2-fold, consistent with prior reports indicating that all three components of the MRN complex are subject to coordinated post-translational regulation . Compared to a control EBV immortalized cell line, the Q372X cells were modestly, but significantly, sensitive to higher doses of ionizing radiation, but not to the extent of cells derived from an ataxia telangiectasia patient (Figure 2b).
To overcome the natural variability associated with patient-derived cell lines, we asked if we could recapitulate the sensitivity to ionizing radiation by introducing the RAD50 c.1114C>T mutation encoding for RAD50-Q372X into the endogenous locus of an immortalized human colon cancer cell line HCT116 (Figure 2c). HCT116 is a near-diploid line that has been used extensively in gene targeting experiments, particularly for studies of DNA metabolism proteins . Targeting events were identified by insertion-specific PCR and confirmed by Southern blot using a flanking probe. Specific incorporation of the c.1114C>T mutation was confirmed by sequencing insertion-specific PCR products. Cells containing the heterozygous c.1114C>T mutation displayed an increased sensitivity to ionizing radiation compared to the control parental line, non-targeted sister clones, or targeted sister clones in which the mutation was not integrated (Figure 2d and data not shown). In comparison, a DNA-PKcs-deficient HCT116 cell line showed higher levels of ionizing radiation sensitivity . These results demonstrate that the RAD50-Q372X mutation impairs the ability of the MRN complex to repair ionizing radiation-induced DNA double strand breaks.
We have shown that variations in multiple DNA repair pathways show association with the primary immunodeficiency syndromes CVID and IgAD. To our knowledge this is the first report associating immunodeficiency with markers in many of these genes, notably MSH2, MLH1, and RAD50. We also identified new non-synonymous alleles of MSH2, MLH1, RAD50, and NBS1, and UTR SNPs in RAD50 and MRE11 that were unique to IgAD and/or CVID. Cells carrying one of these mutations, RAD50-Q372X, were incapable of wildtype levels of recombination repair, likely due to haploinsufficiency. Overall, these data suggest that variations in the mismatch and recombination repair pathways may underlie some IgAD and CVID cases.
Mismatch repair proteins directly recognize and excise single base DNA mismatches and short patches of mismatched DNA (reviewed in ). Genetically, the involvement of mismatch repair in CSR has been well established in mice, with the most prominent defects observed in animals that also lack the uracil excision repair protein UNG2 . Humans with heterozygous inactivating mutations in the mismatch repair proteins MSH2, MSH6, MLH1, or PMS2 are at increased risk for developing hereditary non-polyposis colorectal cancer (HNPCC) , . IgA deficiency has been noted in a handful of HNPCC patients, one with a homozygous MSH2 mutation  and three with biallelic MSH6 mutations , .
Our results are consistent with a role for mismatch repair in CSR. Most of the IgAD/CVID-specific alleles that we detected in the mismatch repair pathway are located in functionally important domains. MSH2-A727S maps to a conserved ABC domain (amino acids 718-731) that is predicted to be involved in ATP hydrolysis  (Figure 3). Two of the three IgAD/CVID specific mutations identified in MLH1 map to domains of known importance. MLH1-S247A is located in a region that has both nucleotide hydrolysis and signal transduction functions . Mutation of this residue to proline (S247P) has been reported in 14 HNPCC2 patients and is associated with decreased MLH1 protein expression and impaired mismatch repair activity using an in-vitro assay . MLH1-G638R introduces a positively charged amino acid into a neutral area of the region known to be important for interaction with PMS2, MLH1, and PMS1 , . Interestingly, two additional rare alleles that we detected in both IgAD/CVID cases and control samples were previously identified and characterized in HNPCC studies. MLH1-K618A was suggested to bind less efficiently to PMS2 , however follow-up studies with this allele found PMS2 binding and overall mismatch repair function to be normal . The MLH1-V716M variant has been suggested to be a protective allele, as it was found enriched in controls compared to HNPCC cases . Taken together, these results indicate that cancer predisposition alleles that are DNA damage sensitizing may also perturb the ability of the DNA repair pathways to elicit CSR.
Our results are the first to implicate MRN defects in IgAD/CVID. The premature stop codon encoded by the RAD50-Q372X allele likely leads to nonsense-mediated decay of the mRNA and haploinsufficiency (Figure 4). Other identified mutations may slightly perturb the structure and/or function of the MRN complex, and could act as weak hypomorphic alleles. The RAD50-P165H mutation replaces a neutral, non-polar proline with a positively charged, polar histidine at a position that is evolutionarily conserved through vertebrates. This residue is just outside of the N-terminal NTPase-containing globular domain and at the edge of the region that is thought to be important for interaction with MRE11. At the edge of the N-terminal MRE11-interating domain, RAD50-R1077Q replaces a conserved, positively charged arginine with a neutral glutamine. RAD50-R327H, which by itself did not associate with IgAD/CVID, exchanges one neutral, polar residue for another in the N-terminal coiled-coil domain. Neither of the additional IgAD/CVID-specific alleles detected in the MRN complex, NBS1-P401R and NBS1-D527Y, disrupt known or predicted functionally important protein domains. Future studies on the detected MRN alleles will provide greater insight into how they functionally impact recombination repair and CSR.
The association noted for two markers in AID suggests that a subclass of IgAD/CVID syndromes may be due to variations at this locus. These markers may be linked to undiscovered alleles that impact the expression of functional AID. This could be due to variations that create splice variants, UTR mutations that impact mRNA stability and protein expression, or mutations that impact transcription of AID, including alterations to transcription factor binding sites or sites important for epigenetic modulation of expression. In these studies, we identified two novel SNPs in the AID UTR and three novel synonymous mutations (Table S4). The further studies of these alleles may yield that some IgAD/CVID patients represent a new subclass of Hyper-IgM type 2 patients.
Overall, these studies suggest that, like many cancer predisposition syndromes such as HNPCC, IgAD and CVID may be a heterogeneous class of diseases due to defects in multiple pathways. Rather than straightforward causal alleles giving rise to disease, IgAD and CVID alleles likely represent risk factors that predispose an individual to developing immune deficiency. Additional studies to determine what effects these unique variations have on DNA repair and CSR are ongoing and long-term.
The Karolinska Institutet human subjects research review board provided written approval for the informed collection of DNA samples from CVID/IgAD patients for sequence analysis. The sample identities were unknown to the University of Minnesota team and therefore this arm of the project was exempt from institutional review.
The initial association screen was performed in DNA from 140 IgAD patients, 48 CVID patients, and 92 healthy controls. To characterize all coding variation for the selected candidate genes in the CVID population, genomic resequencing was performed in 58 CVID DNAs, 48 from the original population and 10 that were subsequently recruited. 38 IgAD patients, each carrying at least one allele/haplotype associating with disease, were also included. 334 IgAD patients, 108 CVID patients, and 991 healthy controls were used for the subsequent large genetic association study of detected non-synonymous SNPs. Where adequate DNA was available, samples that had been used for earlier studies were utilized. All DNA samples used in this study were collected from unrelated individuals of self-reported Swedish descent and isolated at the Karolinska Institutet (Stockholm, Sweden). Informed consent was obtained from all subjects, and these studies were approved by the Karolinska Institutet human subjects research institutional review board.
For the initial association screen, tag SNPs were selected to cover the genomic locus of each gene plus an additional 10 kb upstream and 5 kb downstream using Tagger . As phase II hapmap data was not yet available at the time, phase I data was used. Additional non-tag SNPs were included to fill out the assay. Genotypes were determined using the Illumina GoldenGate platform at the Broad Institute. Of the 384 markers genotyped, 81 failed to pass quality control tests and were not used in subsequent studies (Table S6). Two markers rs2020908 and rs689754 deviated significantly from Hardy-Weinberg equilibrium in the control DNA set (p<0.001), and 79 markers had genotype call rates that were below the 90% confidence threshold. Therefore, a total of 303 SNPs passed quality control criteria and were used in the subsequent studies.
For the large genetic association study, genotyping was performed by the BioMedical Genomics Center at the University of Minnesota. Two Sequenom iPLEX gold genotyping assays were designed using Spectrodesigner, part of the MassARRAY software package, to include non-synonymous coding SNPs detected by resequencing. Assays were filled out by including synonymous coding SNPs, UTR SNPs, and intronic SNPs located within 10 bases of an exon. Primers were synthesized by IDT and sequences are available upon request.
Primer design, sequencing, and polymorphism detection were performed by the Broad Institute Center for Genotyping and Analysis. Briefly, primers were designed for each target gene to sequence the known exons, the 5′ UTR, and the 3′UTR, with at least 100 nucleotides of flanking sequence on each side. Genetic variants were identified by a combination of the PolyPhred and PolyDhan programs. Primer sequences for each assay are available upon request. To increase the power to detect informative alleles, the sample set was enriched for DNAs that carried disease-associated single markers and haplotypes as determined in the initial genotyping screen. In all, 3,086,784 nucleotides were examined, with average 1× coverage of 92.0% and 2× coverage of 88.4% (Tables S7). All novel non-synonymous SNPs were confirmed by fluorescence-based Sanger di-deoxy sequencing at either the University of Minnesota Sequencing and Analysis Facility or the Macrogen Company (South Korea).
For both genotyping studies, allele calls were made using the appropriate built-in automatic allele calling specific to each platform. Call clustering was inspected manually and adjusted when obvious calling errors were made by the software. For genetic association studies, we required that each SNP have a missing genotype rate <10% and, in controls, have a Hardy-Weinberg p-value >0.001. Samples were excluded if genotyping failed for >50% of markers. Haploview was used to determine haplotype structure from control genotypes and to perform association testing for both single markers and haplotypes .
To correct for multiple testing bias, association results were subjected to 100,000 permutation tests by pathway using Haploview. For this analysis, genes were grouped based on functional pathways. Each group that contained at least one marker or haplotype with an uncorrected p-value <0.01 was permuted. These pathways were MutSα (MSH2 and MSH6), MutSβ (MSH2 and MSH3), MutLα (MLH1 and PMS2), MRN (MRE11, RAD50, NBS1), MutSγ (MSH4 and MSH5), and RAD51-mediated homologous recombination (RAD51, RAD52, RAD54B, DMC1, and BLM). AID and MUS81 were permuted on an individual gene basis. Permutation testing was conducted for SNPs and haplotypes independently.
EBV-immortalized lymphoblast lines were generated at the Karolinska Institutet (Stockholm, Sweden). The HCT116-RAD50-c.1114C>T (Q372X) knock-in cell line and a wildtype knockin control line were generated using recombinant adeno-associated virus (rAAV)-assisted gene targeting, as previously described . Briefly, targeting arms were amplified from genomic DNA by PCR and cloned into pJET1.2 (CloneJet PCR kit, Fermentas). Site directed mutagenesis was used to introduce the RAD50 c.1114C>T stop mutation and targeting arms were assembled into pNeDaKO . rAAV was produced in HEK-293T cells and used to transduce HCT116 cells. The resulting G418-resistant clones were screened for targeted insertion by PCR and confirmed by Southern blotting. The introduction of the mutation was confirmed by sequencing.
Cell extracts were prepared using RIPA buffer, fractionated by SDS PAGE, and transferred to PVDF membrane for immunoblotting. The membranes were probed with antibodies to RAD50 (Novus), MRE11 (Genetex), or tubulin (Bethyl) followed by appropriate HRP-conjugated secondary antibodies (BioRad). Radiation sensitivity assays were conducted exposing cultures in a cesium irradiator for the indicated doses. Cell viability was assayed by Propidium Iodide staining and flow cytometry (lymphoblasts) or crystal violet colony staining of plated dilutions (HCT116 cells) .
(0.12 MB PDF)
(0.04 MB PDF)
(0.09 MB PDF)
(0.07 MB PDF)
(0.05 MB PDF)
(0.11 MB PDF)
(0.04 MB PDF)
We thank T.W. Behrens, R. Ferreira, and R. Graham for initiating aspects of these studies, A. Kotnis for confirming some of the rare alleles, the Broad Institute for Illumina genotyping and DNA sequencing, the University of Minnesota Biostatistics and Informatics Shared Resource for statistical support, the University of Minnesota Supercomputing Institute for computational support, and the University of Minnesota BioMedical Genomics Center for Sequenom genotyping.
Competing Interests: The authors have declared that no competing interests exist.
Funding: This work was supported by the National Institutes of Health (U19 AI067152 to L.H., R21 AI079743 and R01 GM80437 to R.H.) and by Swedish Cancerfound (to Q.P.). S.M.O. was also supported in part by a University of Minnesota Graduate School Doctoral Dissertation Fellowship and by the Curtis L. Carlson University of Minnesota/Karolinska Institutet Medical Research and Education Program Award. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.