IgA defects in CVID lead to bacterial translocation, increased serum γ-interferon, and BAFF

Common variable immunodeficiency (CVID), the most prevalent symptomatic primary immunodeficiency, is characterized by low levels of serum immunoglobulin (Ig) G, A, and/or M and a lack of specific antibodies after vaccination or disease exposure (1, 2, 3). Due to loss of protective antibodies, subjects with CVID are susceptible to recurrent, severe infections, commonly of the respiratory tract. However, noted for years is that up to 30–50% of CVID subjects may develop a variety of additional inflammatory or autoimmune conditions (4, 5), commonly termed “noninfectious complications.” These conditions include autoimmunity (especially thrombocytopenia and hemolytic anemia), interstitial lung disease, lymphoid hyperplasia, enteropathy, nodular regenerative hyperplasia of the liver, systemic granulomatous disease, and in some, lymphoid malignancy (4, 5, 6, 7). These complications pose difficulties for treatment, as standard immune globulin therapy provides neither a treatment nor a prevention of these conditions, and immune suppression in immune deficiency is generally to be avoided. Taken together, these inflammatory conditions lead to an 11-fold increased morbidity and mortality in CVID (5, 8).

One of the most useful methods of subdividing CVID subjects is by considering the number of isotype-switched memory (SM) B cells in circulation. Subjects with the lowest numbers of these cells have been shown to be patients with the greatest risk of inflammatory complications (9, 10), but the reasons for this have not been clear. While recent investigations have demonstrated that 25–30% of subjects have one of a number of genetic defects, this work has not supplied a better understanding nor suggested a therapeutic option for the majority of subjects (11, 12, 13).

On investigating inflammatory pathways that might distinguish CVID subjects, we previously demonstrated a marked upregulation of IFN-γ pathways by mRNA transcriptional profiling (14, 15). The IFN signature was also associated with an expanded population of IFN-γ–, IL-17A–, and IL-22–positive innate lymphoid cells (ILCs) in peripheral blood and also in gastrointestinal mucosal and lung tissues of CVID subjects (15). As ILCs play an important role in host-commensal homeostasis (16), we suggested that their excessive activity and/or proliferation could contribute to systemic, mucosal, and organ-specific inflammation in CVID. However, the stimulus underlying these immune responses has remained unknown. We have recently shown that circulating bacterial 16S ribosomal DNA (rDNA), primarily from gut commensals, is significantly increased in CVID serum (P < 0.0001), but especially in those subjects with low numbers of SM B cells and increased numbers of inflammatory conditions (P = 0.0007) (17). These bacterial ribosomal DNA (rDNA) levels were also significantly associated with increased serum IFN-γ, TNF-α, IL-6, soluble CD14, lipopolysaccharide-binding protein, and the mucosal markers of gastrointestinal mucosal leakage, zonulin and intestinal fatty acid binding protein (intestinal fatty acid binding protein (I-FABP)) (17, 18).

In mice and humans, secretory IgA and IgM have been shown to limit bacterial translocation from mucosal compartments (19, 20, 21, 22, 23). Here we show that the levels of bacterial ribosomal DNA (rDNA) in CVID serum are inversely correlated with the numbers of isotype SM B cells, the numbers of circulating IgA⁺ SM B cells, and residual serum IgA concentrations. We find that the levels of bacterial DNA are closely associated with loss of serum IgA and also with the levels of serum IFN-γ and B cell–activating factor (BAFF). In previous work, we showed that the serum of subjects with CVID contained high BAFF (24, 25), and separately, that increased serum BAFF was correlated with increasing B cell–rich lymphocytic infiltrates in the lung in CVID (26). Based on the current observations, we suggest that the loss of IgA, along with mucosal defects, permits gastrointestinal bacterial transcytosis, which then leads to the production of both of these inflammatory signals.

55 of the 114 subjects studied here (23 females and 32 males) (48%) had experienced one or more “noninfectious” complications; 31 had had one or more episodes of  immune thrombocytopenia (ITP), in some cases also hemolytic anemia, at the same or different intervals; 15 of these had interstitial lung disease, 16 had biopsy-identified granulomatous disease, 10 had marked lymphoid hyperplasia, 14 have significant enteropathy, 6 had severe liver disease, 3 had lymphoma, and 7 have had a splenectomy. Of the 114 subjects, 32 had one or more genetic mutations previously associated with CVID or considered to be pathogenic (28% of the group) (Table 1). In most cases, these were heterozygous, but one was homozygous. One subject had two pathogenic heterozygous mutations associated with CVID. Of the 32 subjects with known gene mutations, 19 had experienced one or more of the inflammatory complications; however, 14 other subjects (with no identified gene mutations) also had one or more of these complications.

A useful method of subdividing CVID subjects has been by considering the number of SM B cells in circulation, as subjects with the lowest numbers of these cells are known to be patients with the greatest risk of inflammatory complications (9, 10). This is illustrated for this cohort, examining the 55 CVID subjects with inflammatory complications (Group 2) as compared to 59 other CVID subjects without these medical issues (Group 1) or 21 normal controls, showing highly significant differences between the patient groups (P = 0.0001.) Fig. 1 A. We previously showed that the subjects with inflammatory complications also have substantially more bacterial ribosomal DNA (rDNA) in their serum than subjects with higher numbers of these SM B cells or normal controls (17). This is again illustrated for this cohort, examining here 55 CVID subjects with inflammatory conditions as compared to the 59 CVID subjects without these conditions (or controls), also showing highly significant differences between these same groups (P = 0.0008) (Fig. 1 B). The numbers of isotype SM B cells are also significantly and inversely related to the amount of bacterial DNA in the serum of the same patients, as shown for the total group (P = 0.038) (Fig. 2).

As increased numbers of isotype SM B cell phenotype suggest that the B cells of these subjects may retain some capacity for normal function and may produce some serum IgA, we then tested this association. First, serum IgA concentrations for the entire group were overall correlated with the numbers of SM B cells in the same CVID subjects (P = 0.01). Second, suggesting a protective role of IgA in bacterial translocation, there was an overall inverse correlation between baseline serum IgA concentrations and the level of bacterial ribosomal DNA (rDNA) in the blood of 111 patients (P = 0.006). In Fig. 3 A, dividing CVID subjects into the two groups shows that Group 1 subjects (“infections only” phenotype, without inflammatory complications) had higher serum IgA levels and less bacterial DNA than Group 2 CVID subjects with inflammatory conditions. In contrast, there were no correlations between the serum concentrations of either serum IgG or baseline IgM and the amounts of circulating bacterial DNA of these same samples (P = 0.36; P = 0.33). We then examined if the numbers of circulating IgA+ SM B cells in CVID subjects might provide another useful biomarker for transcytosis of gut commensal bacteria. First, as one would expect, the numbers of IgA+ SM B cells were closely correlated with the numbers of total isotype class SM B cells in 94 CVID subjects, again shown in these groups (P = 0.0001) (Fig. 3 B), as well as with the baseline level of serum IgA concentrations of the same patients, (P = 0.002) (Fig. 3 C). The overall numbers of IgA⁺ SM B cells in blood were also inversely correlated with bacterial 16S ribosomal DNA (rDNA) levels in the same patient’s serum (P = 0.02). Fig. 3 D shows these data and also indicates the differences between these two cohorts, (Group 1) without as opposed to those with inflammatory complications (Group 2). We noted also that the CVID subjects with lower numbers of SM B cells had significantly fewer numbers of IgA+ SM B cells, as compared to those subjects with the infection only phenotype (mean 2.1% ± 2.8 for 46 Group 2 subjects vs. 3.6% ± 3.9% of B cells for 43 Group 1 subjects; P = 0.03) (data not shown).

We previously showed that circulating bacterial 16S rDNA in the serum of subjects with CVID was associated with markedly increased serum IFN-γ levels (17). In addition, when CVID mononuclear cells of patients with inflammatory complications were exposed to 16S ribosomal DNA (rDNA), they produced large amounts of this cytokine in vitro (17, 18). Increased serum IFN-γ was also the case for the subjects studied here, particularly for those subjects with fewer isotype SM B cells. While CVID subjects of both groups had more serum IFN-γ than controls (P = 0.0001), subjects with <2% of SM B cells had more serum IFN-γ than those with more than this number (P = 0.05) (using the previously suggested cutoff of 2%) (9) (Fig. 4 A). We also noted that as a measure of overall increased serum IFN-γ levels in CVID, the serum chemokine CXCL9 was also significantly increased in CVID subjects as compared to control sera (P < 0.0001) (Fig. 4 B). As one would expect, the level of CXCL9 was closely related to the level of IFN-γ in the same CVID samples (P = 0.0001). This is shown for the patients, again divided into those with (Group 2) or without (Group 1) inflammatory complications (Fig. 4 C). (Subjects with inflammatory complications had increased CXCL9 as compared to those with the infection only phenotype, P = 0.01.)

In other publications, IFN-γ has been shown to stimulate monocyte, myeloid, and T cell production of BAFF (27, 28). As we had shown previously that the serum of subjects with CVID commonly contains increased amounts of BAFF (24, 25), we also examined this cytokine in the sera of this CVID cohort. First, we show here that overall, serum BAFF concentrations were also significantly correlated with serum IFN-γ concentrations in the same CVID samples (P = 0.002) (Fig. 5 A). Second, for subjects with SM B cells of <2%, the median serum BAFF was 11,486 pg/ml, while for those with SM B cells >2%, the median BAFF was 5,549 pg/ml (P = <0.0001) (Fig. 5 B), possibly not surprising as these sera also contained increased IFN-γ concentrations. In fact, BAFF levels overall in CVID sera were closely related to the numbers of isotype SM B cells; P = 0.0013 (data not shown). Again suggesting the potential positive role of IgA, there was also an inverse correlation between serum BAFF concentrations and baseline serum IgA levels (P = 0.02) (Fig. 5 C).

One of the central questions regarding CVID has been why a significant proportion of patients have at diagnosis, or later develop, severe inflammatory and/or autoimmune complications, in spite of receiving the same treatment with substantial doses of immune globulin and antibiotics as needed. While genetics has supplied some of the answers, about 70% of subjects still do not have a known genetic basis (12, 13, 29). For this reason, we have been pursuing additional means to unravel the inflammatory pathways that underlie what is often termed the “noninfectious” complications of CVID. In previous work, seeking an unbiased approach, we demonstrated that a hallmark of the subjects with inflammatory/autoimmune complications was an IFN-γ mRNA signature, with a number of genes in this pathway being upregulated; in addition, we noted expanded numbers of IFN-γ secreting ILCs in blood and mucosal tissues of these subjects (14, 15, 30). To seek to explain these observations, and the cause of the IFN-γ elevation, we later identified that these subjects also had increased mucosal bacterial transcytosis, with high serum levels of 16S bacterial ribosomal DNA (rDNA), from species identified of gastrointestinal origin (17). This phenomenon was also associated with elevation of corresponding markers of systemic inflammation (lipopolysaccharide-binding protein, sCD14) as well as specific markers of gut mucosal barrier defects (zonulin and intestinal fatty acid binding protein (I-FABP)). We also showed that these biomarkers were a particular characteristic of subjects with very low numbers of circulating isotype SM B cells (17), previously also found to be characteristic of CVID subjects with greater risk of inflammatory/autoimmune complications (9, 10, 18). Bacterial transcytosis has previously been suggested to occur in CVID, as endotoxin in plasma and increased lipopolysaccharide have been identified, along with functional impairments of CD4 T cell proliferation, with increased programmed death 1 expression, as well as serum and plasma markers of systemic immune activation, sCD14 and sCD25 (31, 32, 33). Several studies have also noted dysbiosis with reduced stool microbial diversity and increases of selected microbial species in CVID, associated with elevated markers of immune activation in subjects with these inflammatory conditions. The acquisition of selected bacterial strains, potentially in the absence of IgA, could also promote the observed inflammatory phenotype (33, 34, 35).

As the loss of isotype-switched B cells in peripheral blood suggested to us that the greater losses of serum or mucosal IgA may be connected to the mucosal bacterial transcytosis, resulting in immune activation and the IFN-γ RNA signature, we then examined this connection. Although deficient serum IgA is a general characteristic of the CVID immune defect, many subjects still have at least some detectable levels in peripheral blood. While we are not able to quantitate levels of mucosal IgA, we show here that the amounts of residual baseline serum IgA is significantly associated with the level of microbial 16S ribosomal DNA (rDNA) in blood and that the baseline level of serum IgA, and even the number of circulating IgA+ SM B cells, can serve as biomarkers for bacterial transcytosis. This also agrees with previous work, which showed that the lowest levels of IgA were noted in the duodenal tissues of subjects with enteropathy (35) and in another study, lower levels of plasma IgA in CVID subjects were also associated with reduced stool microbial alpha diversity (33) potentially contributing to the inflammatory phenotype. These observations are also compatible with our previous work, in which retained serum IgA and increased numbers of IgA+ SM B cells in CVID subjects with autoimmune cytopenias were associated with lower serum endotoxin levels (36). In fact, in the CVID subjects with autoimmune cytopenias in this study, CD27⁺IgA⁺ SM B cells were virtually absent in peripheral blood (36). In addition, CVID patients with very rare circulating IgA⁺ SM B cells were also characterized by an increase in VH4-34⁺IgG⁺ B cells that were reported to recognize commensal bacteria, further illustrating the impaired gut microbiota containment in these patients (36, 37). As a consequence, increased commensal bacteria transcytosis in CVID patients with very low or absent serum IgA may trigger inflammatory immune responses, potentially also related to the hyperplastic but inefficient germinal center responses in these patients, associated with enhanced T follicular helper numbers and decreased regulatory T (Treg) cell numbers and function (36). Since Treg cells play an important role in preventing the accumulation of autoreactive naïve B cells in the periphery (38, 39, 40), it may not be surprising that elevated frequencies of autoreactive clones were previously detected in the mature naïve B cell compartment of CVID patients in which Treg cell functions were compromised (25). Together, these observations may also help to explain why CVID subjects with the lowest numbers of isotype SM B cells, with greater losses of IgA, may be more likely to have inflammatory/autoimmune complications. While some subjects with CVID have some retained IgM, and if secreted, this Ig could play a protective role in gastrointestinal mucosal tissues, we found no relationship between serum IgM and bacterial DNA levels in blood. In addition, one study suggested that the IgM produced in CVID may possess an autoimmune potential, with the ability to bind to cellular antigens (41).

Our previous work also showed that CVID subjects with increased microbial ribosomal DNA (rDNA) in blood also had increased serum IFN-γ and that adding microbial ribosomal DNA (rDNA) in vitro to mononuclear cells of subjects with known inflammatory conditions led to the production of more IFN-γ in these cultures (17). Here we show that, as one might expect, subjects with the lowest numbers of isotype SM B cells, those known to be more likely to have these complications (9, 10), also have significantly more IFN-γ in their serum. IFN-γ was previously shown in other studies to stimulate the production of BAFF by monocytes and myeloid cells (27, 28), an observation of potential importance in lymphoid hypertrophy and autoimmunity (42). Perhaps, in line with this, serum BAFF levels in CVID were also directly correlated with serum IFN-γ levels in paired serum samples. We previously showed that the serum of subjects with CVID contains large amounts of BAFF (24, 25), a factor also associated with pulmonary B cell infiltrates with germinal center formation in the lungs of our CVID patients (43). Here we show that serum BAFF concentrations were also inversely correlated with serum IgA concentrations, suggesting that the mucosal defects may be connected to the overproduction of these inflammatory signals. Decreased numbers of B cells, and especially memory B cells that express several receptors for BAFF (i.e., BAFF receptor and TACI), which consume this growth factor, could also result in an increase in serum BAFF concentrations (44). In agreement with this scenario, we found that CVID subjects with the lowest numbers of isotype SM B cells are also those subjects with significantly more serum BAFF.

In summary, the data from this large cohort of CVID subjects provide an overall framework to better understand some of the poorly understood aspects of this complex immune defect. Why do those subjects with the lowest numbers of isotype SM B cells have the greatest risk of inflammatory and autoimmune complications? And what are the mediators of this inflammation? Here we show that bacterial transcytosis due to mucosal barrier defects and loss of B cell isotype class-switched B cells, and therefore IgA production, promotes an IFN-γ signature, leading to inflammation, which in turn activates also numerous mediators of inflammation, one of which is excess BAFF, which may promote survival and proliferation of immature B cells. While IgA from normal donors might not address the dysbiotic microbial species found in subjects with CVID (34), polyclonal IgA supplementation, in parallel with IgG replacement, (for subjects who have not developed anti-IgA antibodies) could represent a novel therapeutic strategy to restore proper gut mucosal barriers and commensal bacteria containment, which may then thwart inflammation and autoimmunity in these patients.

Patients with CVID (59 females and 55 males, median age of 55, range 24–90) were enrolled with informed consent in a Mount Sinai Institutional Review Board–approved protocol designed to investigate human B cell defects. Patients were diagnosed by standard criteria, including reduced serum IgG, IgA, and/or IgM with loss of antibody to vaccine protein and polysaccharide antigens (1). Baseline patient characteristics showed a mean of IgG = 200 mg/dl (range 9–593 mg/dl), IgA = 16.5 mg/dl (0–179 mg/dl range), and IgM = 46.1 mg/dl (range 0–81 mg/dl). The age range was 22–72 years of age. The inflammatory medical complications experienced by each subject were categorized as previously outlined: infections only or subjects with autoimmunity, interstitial lung disease, granulomatous disease, lymphoid hyperplasia/splenomegaly, gastrointestinal enteropathy, liver disease, and/or lymphoma (4, 5). Control subjects were healthy adults, 13 females and 8 males, ages 22–64.

B cell populations from patients were examined as previously described (9, 10), identifying total B cells, CD27⁺ memory B cells, and isotype SM cells (CD27+, IgD−IgM−). Separately, we also identified IgA+CD27+ IgD−IgM− B cells from patients and controls. Subjects were stratified into two groups, as those with 2% or more SM B cells and those with fewer than this number, as previously published (9).

Genetic evaluation was done by whole-exome sequencing as previously described (11, 12, 13, 45, 46). In some cases, patient exomes were also examined for mutations in a panel of 429 genes associated with primary immune deficiency disease (Invitae Diagnostics).

Serum from patients were drawn into 6-ml gold top rubber-sealed sterile Vacutainer SST II tubes (BD Diagnostics) as described (17). The tubes maintained in the upright position until the serum was aseptically removed in a sterile hood. Bacterial DNA analysis was performed as previously described, using a DNA standard prepared from Escherichia coli competent cells (Qiagen); concentrations were measured by NanoDrop. Using the E. coli genome length (4,700,000), the number of E. coli DNA copies was calculated as described (17), and serial dilutions of this standard were used to define DNA copy numbers in a standard curve.

Serum IFN-γ was determined using a human IFN-γ ELISA (R&D Systems) using dilutions of (1:2,500, 1:1,000, and 1:10,000) (R&D Systems). Serum CXCL9 was also examined in the same sera by ELISA (R&D Systems); both are expressed in pg/ml. Serum BAFF was quantitated in serum, diluted 1:2 by ELISA using a commercial kit (Axxora LLC), and expressed in pg/ml.

The nonparametric Mann–Whitney test was used to compare bacterial DNA and class SM B cells between subjects with inflammatory complications from those without these medical problems. This test was also used to compare serum BAFF, IFN-γ, and CXCL9 levels in patient groups and controls. For determining the correlations between the paired variables, bacterial DNA, IFN-γ, CXCL9, BAFF, serum IgA, and isotype SM B cells, the nonparametric Spearman rank correlation test was used. In all tests, P values <0.05 were considered significant. All calculations were performed using Prism GraphPad software.

This study was approved by the institutional review board of the Icahn School of Medicine at Mount Sinai and was carried out in accordance with the Code of Ethics of the World Medical Association (Declaration of Helsinki). Written informed consent was received from participants prior to inclusion in the study.

The data underlying these results and methods of analysis are available on request.

We thank the patients and their families for their participation in this research as well as the nursing and support staff for sharing in the care of these patients.

This work was supported by the National Institutes of Health, AI-061093, AI-086037, and AI-48693, and the David S. Gottesman Immunology Chair.

Author contributions: Hsi-en Ho: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, supervision, validation, and writing—original draft, review, and editing. Lin Radigan: data curation, investigation, methodology, project administration, resources, supervision, and validation. Eric Meffre: conceptualization and writing—original draft, review, and editing. Charlotte Cunningham-Rundles: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, software, supervision, validation, visualization, and writing—original draft.