Coexistent alterations of BAFF and B cell phenotypes in complicated CVID course

Abstract

Background:

Common variable immunodeficiency (CVID) is a primary immunodeficiency marked by impaired antibody production and infection susceptibility, with about half developing non-infectious complications. BAFF elevation and genetic variants in a BAFF receptor, transmembrane activator and CAML interactor (TACI), frequently occur in CVID. While these findings associate with autoimmune and lymphoproliferative complications, pathogenic mechanisms remain incompletely defined.

Objective:

This work explored how coexistent changes in BAFF, its receptors, and B cell subsets may shape CVID.

Methods:

Plasma protein measurement, spectral flow cytometry, and single-cell RNA sequencing was applied.

Results:

CVID with autoimmune cytopenias and lymphoid hyperplasia had elevated plasma BAFF:TACI ratio and increased transitional and activated naïve B cells. Activated naïve B cells from CVID patients had increased mRNA of genes downstream of BAFF receptor (BAFF-R) that promote B cell survival. Also on these expanded subsets, BAFF-R surface protein decreased, consistent with negative feedback, while autoreactive B cell receptor (BCR) VH4–34 clonality increased.

Conclusions:

Novel application of spectral flow cytometry, paired BCR sequencing RNA-seq, and measurement of BAFF and related proteins in plasma found CVID with autoimmune and lymphoproliferative complications to be marked by coexistent BAFF and B cell subset dysregulation. This included increased plasma BAFF, BAFF:TACI ratio, and transitional and activated naïve B cells with reduced BAFF-R expression and BCR repertoire diversity. The expanded activated naïve B cell subset in CVID had increased expression of BAFF-R-driven genes that subvert B cell tolerance as well as increased autoreactive VH4–34 clonality. Convergence of BAFF, its receptors, and B cell subset dysregulation, should be further explored in CVID.

CAPSULE SUMMARY

Imbalance of BAFF neutralization by soluble TACI may subvert B cell tolerance mechanisms through heightened BAFF-R signaling in CVID patients with more complicated disease course.

Graphical Abstract

INTRODUCTION

Common variable immunodeficiency (CVID) is a heterogeneous primary immunodeficiency characterized by impaired antibody production and susceptibility to infections.^{1, 2} Numerous monogenetic etiologies have been identified, but only in a minority of CVID cases.³ About half of CVID patients develop non-infectious complications, including autoimmune cytopenias (AIC) and interstitial lung disease (ILD), for unclear reasons.⁴ As they are the greatest source of morbidity and mortality, understanding mechanisms driving non-infectious complications is the most significant knowledge gap in CVID.^{5, 6}

B cell activating factor (BAFF) has a central function in B cell homeostasis and promotes autoimmune and lymphoproliferative disease when in excess.^7–10 BAFF has three receptors, BAFF receptor (BAFF-R), transmembrane activator and CAML interactor (TACI), and B cell maturation antigen (BCMA). While BAFF-R is broadly expressed across B cell subsets, highest expression of BCMA and TACI occurs on antibody-producing cells and activated/memory subsets, respectively.¹¹ BAFF-R is distinguished from BCMA and TACI by its ability to promote survival of B cells via the non-canonical NF-κB pathway.⁷ Accordingly, excessive BAFF-R stimulation can subvert peripheral B cell tolerance, driving autoimmune and lymphoproliferative disease.^12–14

Variants of TNFRSF13B, which encodes TACI, are among the most frequent genetic findings in CVID and increase non-infectious complications, including AIC, ILD, and splenomegaly.^15–19 As variants also occur in asymptomatic individuals, pathogenicity of heterozygous TNFRSF13B variants may require a modifying factor, like BAFF elevation. If TACI counterbalance is deficient, increased BAFF-R signaling may promote autoimmunity or lymphoproliferation via its unique ability among BAFF receptors to signal via non-canonical NF-κB.^{20, 21} Yet, interplay between BAFF and its receptors in CVID remains incompletely understood.

We aimed to study BAFF and its signaling within the context of B cell subset alteration and CVID manifestations. Applying a combination of plasma protein analysis, spectral flow cytometry, and single cell RNA sequencing (scRNA-seq) to a characterized cohort, we measured levels of BAFF and its receptors together with that of B cell subsets and gene expression in CVID. Our findings indicate that unchecked BAFF signaling may promote pathogenic B cell responses in CVID and should be explored further.

METHODS

Participants

Diagnosis of CVID was defined as a reduction of ≥2 standard deviations below the mean for serum IgG and IgA and/or IgM, poor response to vaccination, and exclusion of other causes of hypogammaglobulinemia.¹ All CVID subjects with non-infectious complications, and 17 without, had genetic analysis by whole exome sequencing or gene panel (Invitae). Those with monogenic CVID etiologies were excluded. Three subjects had heterozygous variants in TNFRSF13B, two with C104R and the third with A181E. IgG, IgA, IgM, neutrophil, monocyte, CD4+ and CD8+ T cell, B cells, and NK cell values were from the clinical laboratory. This study was approved by the institutional review board of the Boston University Chobanian & Avedisian School of Medicine.

Isolation of PBMCs and Plasma

Peripheral blood mononuclear cells (PBMCs) and plasma were isolated using Ficoll-Paque density gradient centrifugation. Blood samples were collected while subjects were on immunoglobulin replacement therapy and had no acute infection within the preceding 3 months. Plasma was stored at −80°C and PBMCs were stored at −135°C.

Clinical Phenotype Clustering

Cohort data was loaded into ClustVis to visualize clustering of multivariate data using principal component analysis (PCA) and heatmap.²² No row scaling was applied to categorical data and singular value decomposition (SVD) with imputation was used for PCA, with clustering by Euclidean clustering distance using the Ward clustering method.

Enzyme-Linked Immunosorbent Assay (ELISA)

Plasma BAFF, APRIL, TACI, and BCMA (R&D Systems) and BAFF-R (Novus Biologicals) were measured in 96-well polystyrene plates following manufacturer’s protocols.

Flow Cytometry

PBMCs underwent live/dead staining (Thermo Fisher) and Fc receptor block (BioLegend) before application of surface marker antibodies (Supplemental Table I). Cells were washed before fixation and permeabilization buffer (Invitrogen) and antibodies against intracellular markers. Data collected by Cytek Aurora cytometer was spectrally unmixed using SpectroFlo v3.2 (Cytek Biosciences) and computationally processed using OMIQ cloud analysis platform (OMIQ, Dotmatics). Briefly, data were scaled, gated for singlet live B cells and relevant dataset features (CD10, CD11c, CD38, IgM, IgD, CD21, IgA, T-bet, CD24, CD27, IgG) were clustered via the Phenograph algorithm (kNN = 15, distance metric = Euclidean).²³ Twenty-eight raw Phenograph clusters were combined into ten phenotypically distinct metaclusters based on Eucledian distance hierarchy. The dataset features were projected into 2D mapping space using opt-SNE dimensionality reduction with PCA-guided map initialization (perplexity = 60) to aid cluster visualization.²⁴

scRNA-seq

scRNA-seq was performed using Chromium 5’ Gene Expression and V(D)J Enrichment Kit (10X Genomics). CD3-CD27-IgD+CD19+ B cells were collected from PBMCs by BD FACSAria II (BD Biosciences). Single-cell suspensions were loaded onto the 10X Chromium Controller, where individual cells were partitioned into gel beads-in-emulsion for reverse transcription and cDNA amplification. Following library preparation, samples were sequenced on the NextSeq 2000 (Illumina) to generate paired end reads. Raw sequencing data were processed using the Cell Ranger pipeline (10X Genomics).²⁵ Briefly, reads were aligned to the GRCh38 human reference genome and gene expression matrices were generated. Cell barcodes and unique molecular identifiers distinguished individual cells. B cell receptor (BCR) sequences were reconstructed using Cell Ranger VDJ. Quality control of processed data was performed using the Single Cell ToolKit (SCTK) pipeline to detect and remove doublets and cells contaminated with ambient RNA.²⁶ Cells were removed from analysis according to the following parameters: DecontX contamination score >0.3, mitochondrial gene expression >20%, total UMIs <100 or >14,000, total genes detected <4,000, top 200 gene expression proportion > 85%, identified as doublet by two or more doublet detection algorithms. Mitochondrial and immunoglobulin genes were filtered out and gene expression scaled by subject prior to clustering. PCA was performed used highly variable genes (n=2,000) selected via the variance stabilizing transformation method in Seurat (version 5.2.1). Cell clustering was performed using the Louvain algorithm implemented in FindClusters() function in Seurat (dims=1:10, resolution = 0.5) to identify six clusters and generate UMAP embeddings. Visual representations of the data were generated using Seurat and dittoSeq (version 1.10.0).^{27, 28} Single cell BCR sequencing data was processed and combined with the general gene expression dataset for paired analysis using the scRepertoire (version 2.2.1) package.²⁹ Gene set enrichment analysis was conducted using the fgsea (version 1.24.0) and msigdbr (version 7.5.1) packages in R.^{30, 31}

Statistical Analysis

Data were analyzed using GraphPad Prism and R Studio (version 2023.09.01). Fisher Exact Test was used for categorical variables and Kruskal-Wallis test for continuous values, with adjustment for multiple comparisons.

RESULTS

CVID clustering based on history of ILD, ITP, and splenomegaly/splenectomy.

Our CVID cohort had 108 patients, 61% female, with a median age of 48 years (Table I). History of clinical complications included 34% ILD and 32% immune thrombocytopenia (ITP). Notably, 23% of subjects had a history of both ILD and ITP. Interquartile ranges for neutrophils, monocytes, NK cells, CD4+ T cells, CD8+ T cells, and B cells were within reference range at time of CVID diagnosis, but IgG, IgA, and IgM were not.

Table I.

Clinical and laboratory features of CVID subjects.

Clinical or Laboratory Parameter	Value
Number of CVID subjects	108
Median age in years (interquartile range)	48 (38 – 55)
Number of female subjects (%)	66 (61)
Infectious complications	Number of subjects (% of total subjects)
Pneumonia	60 (56%)
Sinusitis	72 (67%)
Non-infectious complications	Number of subjects (% of total subjects)
Immune thrombocytopenia (ITP)	35 (32%)
Autoimmune hemolytic anemia	13 (12%)
Total autoimmune cytopenia (AIC)	36 (33%)
Interstitial lung disease (ILD)	37 (34%)
Both AIC and ILD	25 (23%)
Splenomegaly or splenectomy	23 (21%)
Nodular regenerative hyperplasia	14 (13%)
Inflammatory bowel disease (IBD)	7 (6%)
Enteropathy (non-IBD)	30 (28%)
Bronchiectasis	21 (19%)
Laboratory findings (reference range)	Median (interquartile range)
IgG, mg/dL (700 – 1600)	248 (107 – 431)
IgA, mg/dL (70 – 400)	4 (0 – 26)
IgM, mg/dL (46 – 304)	20 (9 – 38)
Neutrophils, cells/μL (1800 – 7000)	3250 (2500 – 4430)
Monocytes, cells/μL (200 – 900)	400 (300 – 600)
NK cells, cells/μL (48 – 450)	90 (57 – 168)
B cells, cells/μL (25 – 285)	135 (39 – 232)

Undetectable IgA (<7mg/dL) was considered 0 for calculation of interquartile range. IgG, IgA, and IgM levels were at time of diagnosis and prior to immunoglobulin replacement therapy. Data indicates medical history of clinical complications listed.

We analyzed patterns of clinical complications to identify distinct phenotypic groupings, a strategy previously utilized.^{5, 32} Three CVID clusters were identified by Ward clustering method with Euclidean clustering distance (Figure 1A). The CVID-1 cluster had 28 subjects (26%) and higher prevalence of ITP (86%), ILD (93%), concurrence of ITP and ILD (79%), and those with splenomegaly or splenectomy (82%) compared to CVID-2 and CVID-3 (Figure 1B). CVID-2 was an intermediate cluster, with 14% having history of ITP and ILD but none with splenectomy or splenomegaly. CVID-3 comprised 59 subjects (55%) with the fewest complications, including none with ITP. Principal component 1 (PC1) provided the cluster separation (Figure 1C), driven most prominently by history of ILD, ITP, and splenomegaly/splenectomy (Figure 1D).

Figure 1.

CVID clustering based on history of ILD, ITP, and splenomegaly/splenectomy. A. Heatmap of CVID complications grouped by phenotypic cluster. Column clustering by Euclidean clustering distance using the Ward clustering method. B. Complication prevalence in each CVID cluster shown as percentage of subjects. Statistics via Fisher’s Exact Test. **** = P < 0.0001; * = P < 0.05. C. PCA plot of CVID subjects according to prevalence of complications, generated using singular value decomposition with imputation. D. PCA loadings for PC1, indicating the relative contribution of each disease variable. PCA, principal componenet analysis.

CVID-1 subjects had significantly fewer total lymphocytes, CD4+ T cells, CD8+ T cells, B cells, and NK cells compared to CVID-3 (Supplemental Figure 1). CVID-1 also had significantly fewer CD8+ T cells compared to CVID-2, and CVID-2 had significantly reduced total B cells and NK cells compared to CVID-3. There were no differences between CVID clusters for age (data not shown), IgG, IgA, IgM, neutrophils, or monocytes.

Elevated plasma BAFF to TACI ratio distinguishes complicated CVID.

BAFF is elevated in CVID, including higher levels in those with lymphoproliferative disease including progressive ILD,^33–36 and is a potential driver of complications derived from B cell dysregulation, including ILD, ITP, and splenomegaly, the clinical features that largely determined our clustering.^{37, 38} Thus, we measured plasma BAFF, soluble forms of its receptors (BAFF-R, BCMA, and TACI) which may regulate BAFF, and a proliferation-inducing ligand (APRIL), which also binds BCMA and TACI.^39–41 BAFF was elevated in all three CVID clusters compared to healthy donors (HD), and was higher in CVID-1, the most complicated cluster, compared to CVID-3, the least complicated (Figure 2A). While APRIL was elevated in CVID-1 and CVID-3 compared to HD, this cytokine was not elevated in CVID-2, nor did it distinguish between CVID-1 and CVID-3 like BAFF (Figure 2B). Consistent with prior study, BCMA was similarly reduced across CVID clusters compared to HD (Figure 2C).⁴² BAFF-R was increased in plasma of all CVID clusters compared to HD (Figure 2D), while TACI was elevated in CVID-1 and CVID-3, but not CVID-2, relative to HD (Figure 2E).

Figure 2.

Elevated plasma BAFF to TACI ratio distinguishes complicated CVID. Plasma A. BAFF, B. APRIL, C. BCMA, D. BAFF-R, E. TACI measured by ELISA in CVID by cluster and HD. F. Ratio of plasma BAFF:TACI for CVID clusters and HD. G. Percentage of subjects from each CVID cluster above or below the median for BAFF:TACI. Bar graphs represent median values. Statistical comparisons performed by Kruskal-Wallis test with Dunn’s multiple comparisons test. * = P < 0.05, ** = P < 0.01, *** = P < 0.001, **** = P < 0.0001.

Given association of TACI variants with CVID, inconsistency of soluble TACI levels across CVID clusters, and postulated role of TACI as a regulator of BAFF, we assessed significance of soluble BAFF:TACI ratio. BAFF:TACI was increased in CVID-1 compared to HD and CVID-3 (Figure 2F). While BAFF:BAFF-R ratio was also increased, it was to a lesser extent than BAFF:TACI (data not shown). Elevation of plasma BAFF:TACI ratio corresponded with more severe clinical course, as it was greater than the median ratio of our CVID cohort in 78% of CVID-1 patients, compared to 57% and 38% of CVID-2 and CVID-3 patients (Figure 2G).

Elevated BAFF and BAFF:TACI together marks worsened CVID course.

To test whether relative deficiency of soluble TACI when BAFF is elevated is linked to more complicated disease, we stratified CVID subjects into two groups based on plasma BAFF (median = 236 pg/mL) and BAFF:TACI ratio (median = 0.54). Thirty CVID subjects had both plasma BAFF and BAFF:TACI above the median (Figure 3A). These high BAFF and BAFF:TACI subjects had greater occurrence of ILD, ITP, AIHA, nodular regenerative hyperplasia of the liver (NRH), and splenomegaly or splenectomy (Figure 3B). In contrast, no differences were observed for pneumonia, sinusitis, or enteropathy. Grouping by plasma BAFF alone was not significant for AIHA, and was less significant for NRH and splenomegaly (Supplemental Table II). Thus, BAFF and BAFF:TACI together distinguished CVID with complicated course better than BAFF alone.

Figure 3.

Elevated BAFF and BAFF:TACI together marks worsened CVID course. A. CVID subjects with plasma BAFF and BAFF:TACI above the median as red dots, other subjects as blue dots. B. Percentage of CVID subjects with (red) or without (blue) BAFF and BAFF:TACI above the median for each listed clinical complication. * = P < 0.05, ** = P < 0.01, **** = P < 0.0001. ILD, interstitial lung disease; ITP, immune thrombocytopenia; AIHA, autoimmune hemolytic anemia; nodular regenerative hyperplasia, NRH.

Transitional and activated naive B cells with reduced BAFF-R expression are increased in CVID with elevated BAFF and BAFF:TACI ratio and B cell-related complications.

We next explored how plasma BAFF, BAFF:TACI ratio, and CVID clusters related to peripheral B cells. Performing spectral flow cytometry on PBMCs from 40 CVID and 18 HD subjects, high-dimensional clustering using the Phenograph algorithm identified distinct B cell subsets (Figure 4A & B). Identified subsets included transitional 1 (CD27-IgD+CD38++CD24++IgM++CD10++), transitional 2/marginal zone precursor (MZP) (CD27-IgD+CD38++CD24++IgM+CD10+CD21++), transitional 3 (CD27-IgD+CD38++CD24+CD10+), resting naïve (CD27-IgD+CD38+CD24+CD21+IgM+), activated naïve (CD27-IgD+CD38-CD21-IgM+CD11c+T-bet+CXCR5-), unswitched memory/marginal zone (CD27+IgD^lowCD38^+/lowCD24+CD21+IgM+/++), IgG memory (CD27+IgD-IgG+), IgA memory (CD27+IgD-IgA+), other switched memory (CD27+IgD-IgG-IgA-), and plasmablasts (CD27++IgD-CD38+++CD24-Ki67++) (Figure 4C, Supplemental Figure 2).⁴³ Subjects with high BAFF and BAFF:TACI had reduced memory B cell subsets relative to other CVID and HD, with a reciprocal expansion of transitional and activated naïve B cells (Figure 4D). CVID-1 had similar memory B cell reduction and transitional B cell expansion (Figure 4E).

Figure 4.

B cell subsets in CVID and HD characterized through spectral flow cytometry. B cells from all subjects were concatenated and projected into opt-SNE space (A, B) with metaclusters representing biological subsets overlaid and color-coded (B). Averaged mean fluorescence intensities of metaclustered (B) subsets from all subjects are presented as clustered heatmaps (C) and abundancies of each subset are presented as heatmaps across high BAFF and BAFF:TACI, all other CVID, and HD (D) and CVID cluster and HD (E).

Proportion of transitional 1 B cells was increased in CVID-1 and those with high BAFF and BAFF:TACI compared to HD, as well as CVID-1 versus CVID-3 (Figure 5A). Additionally, transitional 2/MZP B cells were elevated in CVID-1 compared to CVID-3. Activated naïve B cells may develop in a T cell-independent manner, with expansion in autoimmune disorders and infections.⁴⁴ While the proportion of activated naïve B cells did not differ across CVID clusters, elevation of this subset in those with high BAFF and BAFF:TACI approached significance (P = 0.057) and was elevated in CVID with AIC compared to HD (Figure 5A, Supplemental Figure 3A). CD27+IgM+IgD+ IgM unswitched memory/marginal zone B cell proportion was increased in CVID-3 relative to HD and CVID-1. IgG memory B cells were decreased in all CVID clusters relative to HD, and reduction approached significance in those with high BAFF and BAFF:TACI ratio compared to other CVID subjects (P = 0.0742) (Supplemental Figure 3B). Levels of IgA memory B cells were also decreased in all CVID clusters relative to HD. Other memory B cells (IgD-IgG-IgA-) were reduced in all CVID clusters relative to HD and were significantly reduced in those with high BAFF and BAFF:TACI ratio compared to other CVID subjects.

Figure 5.

Transitional and activated naive B cells with reduced BAFF-R expression are increased in CVID with elevated BAFF and BAFF:TACI ratio and B cell-related complications. A. Percentage of transitional 1, transitional 2, activated naïve, and unswitched memory/marginal zone B cells in high BAFF and BAFF:TACI, all other CVID, and HD and CVID clusters and HD. B. BAFF-R MFI on transitional 1, transitional 2, activated naïve, and unswitched memory/marginal zone B cells in high BAFF and BAFF:TACI, all other CVID, and HD and CVID clusters and HD. Bar graphs represent median values. Statistical comparisons performed by Kruskal-Wallis test with Dunn’s multiple comparisons test. * = P < 0.05, ** = P < 0.01, *** = P < 0.001, **** = P < 0.0001.

As reduced BAFF-R expression is indicative of negative feedback, we next examined surface protein expression of BAFF-R and TACI.⁴⁵ Surface BAFF-R was reduced on transitional and activated naïve B cells from CVID subjects compared to HD (Figure 5B). Notably, BAFF-R on transitional 2/MZP and activated naïve B cells was significantly lower for CVID-1 compared to CVID-3. In contrast, there were no differences in TACI surface expression (Supplemental Figure 4).

Increased mRNA expression of BAFF-R related pathways in transitional and activated naïve B cells in complicated CVID.

We next investigated transcriptional changes of transitional and activated naïve B cells, the subsets elevated in clinically complicated and high BAFF/BAFF:TACI CVID. PBMC-derived CD27-IgD+ B cells from 3 CVID patients, all with a history of ITP and 2 also with ILD, and 3 HD underwent scRNA-seq with paired BCR sequencing. We included CD38 in our flow cytometry sort to ensure adequate collection of transitional (CD38^high), resting naïve (CD38^mid), and activated naïve (CD38^low) subsets.⁴³

Clustering patterns revealed distinct transcriptional signature among B cell subsets, and between CVID and HD (Figure 6A). Using expression levels of CD38, CD24, MME (CD10), and TCL1A, transitional, resting naïve, and activated naïve/CD38low B cell subsets were identified for both CVID and HD subjects (Figure 6B). Activated naïve B cells from CVID subjects and a corresponding cluster from HD exhibited low or absent expression of CD38, CR2, CD24, and CXCR5. ⁵¹ While activated naïve B cells from CVID subjects highly expressed ITGAX (CD11c) and TBX21 (T-bet), the corresponding cluster in HD (CD38low-HD) did not. Single cell composition analysis found each B cell subset (transitional, resting, and activated naïve/CD38low) had nearly pure separation into CVID and HD comparators (Figure 6C).

Figure 6.

Increased mRNA expression of pathways downstream of BAFF-R in transitional and activated naïve B cells of complicated CVID subjects. A. Seurat clusters of B cells on UMAP. B. Expression overlay of canonical markers defining B cell subsets. C. Number of cells by subject group in each B cell cluster. D. Numbers of genes up- and downregulated in CVID relative to HD for each B cell subset. E. Gene set enrichment analysis compared for B cell subset between CVID and HD. F. Heatmap showing mRNA expression of Bcl-2 family members. G. mRNA expression of TNFRSF13C and TNFRSF13B in transitional, resting naïve, and activated naïve B cells of CVID and HD.

Comparative differential gene expression analysis between CVID and HD for transitional, naïve, and activated naïve/CD38^low-naïve B cell clusters revealed extensive transcriptional differences, with the most pronounced when comparing aNaive-CVID and CD38low-HD B cells, with 5692 genes downregulated and 2435 upregulated (Figure 6D). Analysis of gene expression in CVID relative to HD found downregulation of 4229 genes and upregulation of 1596 for resting naïve B cells, and downregulation of 2912 genes and upregulation of 912 for transitional B cells. Gene set enrichment analysis uncovered distinct changes between CVID and HD subjects for each of the B cell subsets analyzed (Figure 6E). For activated naïve/CD38low B cells, expression of B cell activation, survival, and PI3K/AKT signaling gene sets were increased in CVID at false discover rate (FDR) < 0.25 and P adjusted < 0.25. For resting naïve B cells, increased expression of gated channel activity, G protein coupled receptor signaling, cell maturation, and regulation of cytosolic calcium and decreased expression of FOXO4 target genes, PI3K/AKT, and non-canonical NF-κB signaling was found (FDR < 0.25, P adj < 0.25). For transitional B cells, decreased expression of gene sets was found for B cell activation, survival, and PI3K/AKT (FDR < 0.25, P adj < 0.25).

Expression heatmap of Bcl-2 gene family members, which are downstream of the non-canonical NF-κB pathway and impact cell survival, showed several differences between CVID and HD comparators for each of the B cell subsets examined (Figure 6F).⁵⁶ The greatest differences occurred in CVID activated naïve B cells relative to their HD comparators. This included increased expression of the majority of Bcl-2 family members examined (FDR < 0.05) in CVID relative to HD, including genes like BCL2A1 and MCL1 that are strongly tied to cell survival. Notably, Bcl-2 family members for which expression increased in HD relative to CVID, PMAIP1 and BNIP3, have been shown to be pro-apoptotic.^{46, 47} Results for resting naïve and transitional B cells were mixed, with expression of similar numbers of Bcl-2 family members increased or decreased between CVID and HD.

As we observed reduction of BAFF-R surface protein on CVID B cells, we assessed mRNA expression of TNFRSF13C (BAFF-R) and TNFRSF13B (TACI). TNFRSF13C transcription was lower in activated and resting naïve B cells from CVID subjects compared to HD comparators (Figure 6F). TNFRSF13B expression was also lower in CVID across B cell subsets.

Reduced BCR repertoire in CVID and increased autoreactive VH4–34+ transitional and activated naïve B cells.

BCR analysis revealed no significant differences in constant heavy chain gene usage between CVID and HD B cells (Figure 7A). However, CVID B cells exhibited significantly increased usage of IGLC2 and IGLC3 constant light chains, and reduced usage of IGKC and IGLC1. Additionally, CVID displayed reduced diversity in variable heavy chain (VH) gene usage compared to HD, indicative of a restricted repertoire (Figure 7B). Increased usage of the specific VH genes VH4–34, VH4–39, and VH4–59 occurred in CVID transitional and activated naïve B cells, but not resting naïve B cells (Figure 7C). VH4–34-expressing B cells have been linked to several forms of autoimmunity, including AIC, as well as complicated CVID.^48–50 VH4–34+ B cells were significantly increased in transitional and activated naïve B cell subsets of CVID relative to HD comparators (Figure 7D). The light chain repertoire typically lacks diversity, a feature believed to limit autoimmunity.⁵¹ Consistent with autoreactive potential, we found VH4–34+ B cells from CVID had increased diversity of variable light chain (VL) gene usage compared to HD (Figure 7E). We also found VH4–34 B cells from CVID had elevated expression of the pro-survival Bcl-2 family member MCL1 compared to CVID B cells expressing other VH genes as well as VH4–34+ B cells from HD, with the result most profound in activated naïve B cells (Figure 7F).

Figure 7.

Reduced BCR repertoire of CVID B cells and increased autoreactive VH4–34+ transitional and activated naïve subsets. A. Constant heavy (CH) and light (CL) chain gene usage was determined by quantifying B cells expressing the specific constant gene as a percentage of the total B cells. Statistics by Fisher’s Exact Test. **** = P < 0.0001. B. Diversity of Variable heavy chain (VH) gene usage was calculated using the cloncalDiversity() function in the scRepertoire package with diversity index calculated by the methods shown. C. The vizGenes function of the scRepertoire package was used to generate a heatmap to visualize the distribution of IGHV genes by assigned cluster. Percentage of cells in each cluster (column) expressing distinct IGHV genes (row). D. Prevalence of VH4–34-expressing B cells was determined using paired HD & CVID clusters from 6 individuals and quantifying the frequency of B cells expressing IGHV4–34 as a percentage of the total B cells in each cluster (x-axis). Statistics by Fisher’s Exact Test. E. The vizGenes function of the scRepertoire package generated a heatmap to visualize the distribution of IGLV-IGLJ pairings as a percentage of total B cells. F. Expression of MCL1 visualized using the dittoRidgePlot() function of the DittoSeq package. MCL1 expression levels are illustrated on the x axis, with the vertical axis of each plot representing the cell count. Cells were grouped by B cell cluster and split by VH4–34 assignment.

Discussion

Our study links B cell-related complications of CVID with elevated BAFF and greater BAFF:TACI ratio, together with BAFF-R-related pro-survival gene expression and autoreactive BCR VH4–34 clonality within expanded activated naïve B cells. BAFF elevation, in concert with relative deficiency of soluble TACI and increased BAFF-R signaling, may subvert tolerance mechanisms that restrain pathogenic B cell responses underlying CVID complications like AIC, ILD, and splenomegaly. While additional study is necessary, our work provides fresh insight into the long-appreciated association of BAFF elevation and TACI variants with CVID.

Our CVID cohort clustered into 3 groups shaped by ITP, ILD, and splenomegaly, and may be most applicable to cohorts with similar prevalence of these complications.^45,46 While BAFF was elevated across CVID clusters consistent with previous studies,^31–32 BAFF:TACI ratio was significantly increased in CVID-1 compared to CVID-3 and HD. Combining elevated BAFF with increased BAFF:TACI ratio amplified association with ILD, AIC, and splenomegaly/splenectomy. These data implicate a deficiency of soluble TACI neutralization of BAFF in CVID complications linked with B cell dysregulation. As deficiency of soluble BCMA is a feature of CVID, deficiency of soluble TACI may have greater significance in these patients, particularly when BAFF is elevated.

Spectral flow cytometry revealed expansion of transitional B cells in CVID-1 as well as elevated BAFF/BAFF:TACI patients. Transitional B cells typically circulate at low levels, but can be elevated in autoimmunity, and are increased in CVID with autoimmunity and lymphoproliferation.^{33, 52, 53} Activated naïve B cells may develop into plasmablasts in autoimmunity and infection, but have had limited study in CVID.⁴⁴ Activated naïve B cells are contained within the CD21^low subset, reported elevated in complicated CVID.⁵⁴ We found activated naïve B cells elevated in CVID with AIC and approaching significance in high BAFF and BAFF:TACI CVID (P = 0.057). Surface BAFF-R, and its corresponding mRNA expression in activated naïve B cells, was reduced in complicated CVID and those with elevated BAFF/BAFF:TACI ratio, which may reflect negative feedback. Indeed, we observed elevated expression of genes downstream of BAFF-R that promote B cell survival and may promote autoimmune and lymphoproliferative disease, including non-canonical NF-κB pathway pro-survival genes BCL2A1 and MCL1. Despite reduced BCR diversity overall, autoreactive VH4–34 heavy chains with greater diversity of light chain usage was enriched in the expanded transitional and activated naïve subsets of complicated CVID.⁵¹ VH4–34 CVID B cells had increased expression of MCL1, a pro-survival Bcl-2 family member downstream of BAFF-R. Thus, expansion of pathogenic activated naïve B cells in more complicated as well as high BAFF/BAFF:TACI CVID patients could result from BAFF-R-driven subversion of immune tolerance, leading to B cell-related complications. Given the modest P value and FDR of our findings due to the small sample size, additional studies are needed.

While we identified altered transcription and shifts in CVID B cell subsets, further studies are needed to determine functional consequences. Signaling overlap of BAFF-R and TACI, as well as other receptor-induced pathways in B cells, complicates ability to delineate effects of BAFF. Moreover, while we found associations between elevated BAFF:TACI, B cell subset and transcriptional changes, and disease manifestations in CVID, providing definitive connections of these factors may require investigation of BAFF neutralization. It is also important to note that follicular helper CD4+ T cells have been reported elevated in CVID patients with non-infectious complications, and BAFF may promote survival of this lymphocyte subset.^55–57 Additionally, we do not know the impact of acute infections upon this cytokine or related CVID complications as we did not collect samples if the subject had recent infection. By measuring BAFF and its receptors in the same plasma samples we hoped to gain a sense of how much free BAFF may potentially be present – as those with higher levels of BAFF and lower levels of its receptors (high BAFF to TACI ratio) could plausibly have more unbound BAFF. However, our assay may not differentiate between free and receptor-bound levels of BAFF.

This study links alterations in BAFF regulation and B cell subset phenotypes to B cell-related complications in CVID. Using approaches like spectral flow cytometry and paired BCR and gene expression RNA-seq to measurements of BAFF and its receptors, we found evidence that excess BAFF-R signaling and increased autoreactivity in activated B cells expanded in CVID with elevated BAFF and BAFF:TACI ratio. Thus, deficient regulation by soluble TACI may worsen CVID course. Continued study of BAFF and its receptors within the B cell dysregulation and complications of CVID is needed.

Key Messages.

• Application of spectral flow cytometry, paired BCR sequencing RNA-seq, and measurement of BAFF and related proteins in plasma found CVID with autoimmune and lymphoproliferative complications to be marked by coexistent BAFF and B cell subset dysregulation.

• Immune dysregulated observed included increased BAFF, BAFF:TACI ratio, and transitional and activated naïve B cells in the blood, with reduced BAFF-R and BCR repertoire diversity.

• The expanded activated naïve B cell subset in CVID had increased expression of BAFF-R-driven genes that subvert B cell tolerance as well as increased autoreactive VH4–34 clonality.

Abbreviations:

AIC

autoimmune cytopenia

AIHA

autoimmune hemolytic anemia

APRIL

a proliferation-inducing ligand

BAFF

B cell activating factor

BAFF-R

BAFF receptor

BCMA

B cell maturation antigen

BCR

B cell receptor

CVID

common variable immunodeficiency

ELISA

enzyme-linked immunosorbent assay

HD

healthy donor

ITP

immune thrombocytopenia

MZP

marginal zone progenitor

NF-κB

nuclear factor kappa-light-chain enhancer of activated B cells

PBMC

peripheral blood mononuclear cell

PCA

principal component analysis

scRNA-seq

single cell ribonucleic acid sequencing

SVD

singular value decomposition

TACI

transmembrane activator and CAML interactor

Data Availability

The data is available under accession number: GSE333202.