Common variable immunodeficiency-associated endotoxemia promotes early commitment to the T follicular lineage

Abstract

Background:

Although chiefly a B-lymphocyte disorder, several research groups have identified common variable immunodeficiency (CVID) subjects with numerical and/or functional T helper cell alterations. The causes, interrelationships and consequences of CVID-associated CD4⁺ T-cell derangements to hypogammaglobulinemia and/or autoantibody production remain unclear.

Objective:

To determine how circulating CD4⁺ T-cells are altered in CVID subjects with autoimmune cytopenias (CVID+AIC) and the causes of these derrangements.

Methods:

Using hypothesis-generating, high-dimensional single-cell analyses, we created comprehensive phenotypic maps of circulating CD4⁺ T cells. Differences between subject groups were confirmed in a large, genetically diverse CVID subject cohort (n=69) using flow cytometry, transcriptional profiling, multiplex cytokine/chemokine detection and a suite of in vitro functional assays measuring naive T cell differentiation, B cell/T cell co-cultures and Treg suppression.

Results:

Whereas CD4⁺ T helper cell profiles from healthy donors and CVID subjects without autoimmune cytopenias were virtually indistinguishable, CVID+AIC T cells exhibited follicular features as early as thymic egress. Follicular skewing correlated with IgA deficiency-associated endotoxemia and endotoxin-induced expression of activin A and ICOSL. The resulting enlarged CVID+AIC circulating Tfh (cTfh) cell population provided efficient help to receptive HD B cells but not unresponsive CVID B cells. Despite this, CVID+AIC cTfh exhibited aberrant transcriptional profiles and altered chemokine/cytokine receptor expression patterns that interfered with Treg suppression assays and were associated with autoantibody production.

Conclusions:

Endotoxemia is associated with early commitment to the follicular T cell lineage in IgA-deficient CVID subjects, particularly those with AICs.

Graphical Abstract

Capsule Summary:

Endotoxin and plasma from IgA-deficient common variable immunodeficiency subjects with autoimmune cytopenias promote enlarged and altered T follicular helper cell populations.

INTRODUCTION

Common variable immunodeficiency (CVID) subjects are antibody deficient and approximately 20% experience autoantibody-mediated autoimmune cytopenias (AICs)¹. Several lines of evidence suggest CVID is chiefly due to B-cell dysfunction. CVID B cells display an array of impairments including defective tolerance^2,3, diminished toll-like receptor seven/nine signaling^4,5, absent thymic-independent antibody responses⁶, curtailed class-switch recombination (CSR)⁷, somatic hypermutation (SHM)^8–10, blocked plasma cell differentiation¹¹ and IgG/IgA deficiency-associated endotoxemia^8,12. Additionally, most monogenic forms of CVID are linked to genes critical to B-cell centric pathways including early B-cell development¹³, the BAFF/APRIL axis^3,14–16, B-cell receptor signal transduction^17–19, B cell co-stimulation^20,21, and terminal B-cell differentiation²².

Despite B lymphocytes’ central role in CVID pathogenesis, disease-associated alterations in several CD4⁺ T-cell subsets have been reported, including diminished numbers of recent thymic emigrants (RTEs)^23–25, poorly suppressive T regulatory cells (Tregs)^3,8,26, and increased frequencies of circulating T follicular helper (cTfh) cells^3,8,27. The causes, interrelationships and consequences of CVID-associated CD4⁺ T-cell derangements to hypogammaglobulinemia and/or autoantibody production remain unclear.

To more wholly capture CVID T helper cell landscapes, we compared high-dimensional single-cell analyses of CVID subject and healthy donor peripheral blood samples. Whereas the CD4⁺ T-cell profiles of CVID subjects without AICs (CVID-AIC) were indistinguishable from healthy donor controls, we report here that CVID+AIC T helper profiles possess a pervasive follicular signature. We further demonstrate the direct and indirect effects of systemic endotoxemia favor commitment to the T follicular helper cell (Tfh) lineage as early as thymic egress. This results in enlarged circulating Tfh pools that express aberrant chemokine and cytokine receptors. In total, our data provide a unifying inflammatory framework connecting seemly disparate CVID-associated T-cell subset alterations to IgA deficiency-associated endotoxemia.

METHODS

Human samples.

Peripheral blood samples from 69 CVID subjects with (17 subjects) and without AICs (52 subjects) (Table E1) were obtained by standard phlebotomy. Each CVID subject met 1999 PAGID CVID diagnostic criteria²⁸ at the time of enrollment and all were receiving antibody replacement therapy. None were receiving chronic immune modulatory therapy at enrollment. We defined immune thrombocytopenia (ITP) using established clinical criteria^29–31. Autoimmune hemolytic anemia (AIHA) was defined by a hemoglobin less than 11g/dL, evidence of hemolysis and a positive direct antibody test (DAT). DAT-negative subjects with hemolysis were also considered to have AIHA if other acquired and hereditary forms of hemolytic anemia were excluded^32,33. Evan Syndrome (ES) was defined as those meeting both AIHA and ITP diagnostic criteria simultaneously or sequentially within at-most a ten-year period. The average age of enrolled CVID+AIC subjects was 23.9 years old (range 4–57 years); 58.8% were male. CVID subject genotyping was performed at genes listed by the International Union of Immunologic Societies to be associated with primary immune deficiencies³⁴. The average age of enrolled CVID-AIC subjects was 18.8 years old (range 3–60 years); 50.0% were male. We also analyzed peripheral blood samples from 33 related and unrelated HD controls with a mean age of 25.2 years (range 4–59); 42.2% were male. As an additional comparator we analyzed peripheral blood samples from a cohort of 14 immunocompetent AIC subjects comprised of 12 ITP subjects and 2 ES subjects. Their mean age was 14.1 years (range 3–18); 36% were male Study protocols were approved by the institutional review boards of Children’s Hospital of Philadelphia, the University of Pennsylvania and Baylor College of Medicine.

Mass cytometry.

Mass cytometry reagents were obtained or generated by custom conjugation to isotope-loaded polymers using MAXPAR kits (Fluidigm). Mass cytometry antibodies were chosen to target 38 molecules relevant to CD4⁺ T cell biology including chemokine receptors, transcription factors, cell-cycle antigens and activating/inhibitory immunoreceptors (Table E2). Cell staining was performed as described³⁵. Briefly, circulating CD4⁺ T cells from three CVID-AIC subjects, three CVID+AIC subjects and three sex/age matched HDs were purified using the MojoSort™ Human CD4 T Cell Isolation Kit (Biolegend). 2×10⁶ circulating CD4⁺ T cells were then stained with 20µ M L/D-139 for 10 minutes at room temperature (RT) for live/dead cell discrimination. Cells were washed in PBS 1% FBS. After incubation with a surface antibody cocktail for 30 min at room temperature, cells were fixed and permeabilized with the FOXP3/Transcription Factor Staining Buffer Set (eBioscience) before intracellular staining. Cells were further fixed in 1.6% PFA (Electron Microscopy Sciences) solution containing 125nM Iridium overnight at 4°C and characterized by mass cytometry using a CyTOF Helios instrument (Fluidigm). Bead-based normalization of CyTOF data was performed using the Nolan Lab Normalizer (https://github.com/nolanlab/bead-normalization/releases). Analyses of individual subject samples from the same clinical group were pooled and then reduced into two-dimensional t-Distributed Stochastic Neighbor Embedding (t-SNE) plots³⁶. Cell clustering was performed with self-organizing map software (FlowSOM; https://github.com/SofieVG/FlowSOM)³⁷. HD and CVID+AIC clusters containing four-fold or greater cell number differences were chosen for downstream analyses.

Cell preparation, flow cytometry and cell sorting.

Mononuclear cells (PBMCs) were isolated from peripheral blood samples using Ficoll-Paque PLUS density gradient centrifugation (GE Healthcare Life Sciences). Cells were surface antibody stained (Table E3) and when appropriate fixed and permeabilized for intracellular staining. Cells were analyzed using a LSRFortessa (BD Bioscience) and/or sorted using a MoFlo Astrios EQ (Beckman Coulter). FACS data was visualized with FlowJo (TreeStar). For cytokine staining, mononuclear cells were rested overnight at 37° Celsius followed by 6 hour stimulation with phorbol 12-myristate 13-acetate (PMA, 25ng/ml, Sigma) and ionomycin (1µ g/ml, Sigma) in the presence of brefeldin A (5 µ g/ml, BD Bioscience). After activation, cells were incubated with a surface antibody cocktail for 30 minutes at RT, then fixed, permeabilized, and intracellularly stained. All antibodies used for FACS analyses are listed in Table E3.

In vitro naive CD4⁺ T cell activation.

Batch-sorted CD4⁺CD45RO⁻ HD naive CD4⁺ T cells were plated at 100,000 cells/well with anti-CD2/CD3/CD28 coated beads (Miltenyi) in the presence of subject plasma (20%) or FBS (20%). After 24 hours, cells, beads, and plasma/FBS containing media were transferred to a recombinant ICOSL-coated (5 µ g/ml, R&D Systems) plate. On culture day five, cells were analyzed for CXCR5 and PD1 expression by FACS. In some cases, 50 µ g/ml of polymyxin B (InvivoGen) was added to CVID+AIC plasma samples or FBS was spiked with lipopolysaccharide (LPS; 100ng/ml, sigma) on culture day zero.

T-cell/B-cell co-cultures.

2.5×10⁴ CD19⁺CD21⁺CD27⁻IgM⁺ sorted naive B cells or CD19⁺CD21⁺CD27⁺ sorted memory B cells were co-cultured with an equal number of CD4⁺CD45RO⁺CD25⁻CD127⁺CXCR5⁺PD1⁺ sorted cTfh cells. Co-cultures were activated by addition of anti-CD2/CD3/CD28 coated beads at a ratio of 1 bead per T cell. On culture day seven, the frequency of surface IgG, IgA and CD38 expressing B cells was measured by FACS. Culture supernatant IgM, IgG, and IgA concentrations were determined by ELISA. In some co-cultures, naive B cells were carboxyfluorescein diacetate succinimidyl ester (CFSE; ThermoFisher) stained to determine proliferative responses.

In vitro T regulatory (Treg) suppression assay.

5×10³ CD4⁺CD45RO⁻CD127⁺CD25⁻ sorted naive T responder (Tresp) cells were CFSE-labeled and co-cultured with an equal number of either CD4⁺CD25^hiCD127^lo/- “all”, CD4⁺CD25^hiCD127^lo/-CXCR5^- “Tregs”, CD4⁺CD25^hiCD127^lo/-CXCR5⁺PD1^hi “CD25^hi “cTfh”. Cultures were activated with anti-CD2/CD3/CD28 coated beads. Co-cultures were stained for viability with the LIVE/DEAD kit (Thermo Fisher), and the proliferation of viable Tresp cells was determined by CFSE dilution at culture day 3.5–4.5 by FACS.

Cytokine, chemokine and immunoglobulin quantification.

Cytokine and chemokine concentrations were measured in thawed HD, CVID and CVID+AIC plasma using Milliplex (T-cell panel; MilliporeSigma) and LEGENDplex (human proinflammatory chemokine panel, BioLegend). Activin A plasma concentrations and IgG, IgA, IgM supernatant concentrations were measured via ELISA (R&D and Jackson Immuno Research respectively). Serum immunoglobulin isotype concentrations were determined by the clinical immunology laboratories at the Children’s Hospital of Philadelphia and the University of Pennsylvania.

Gene expression microarrays.

CD4⁺CD25⁻CD127⁺CXCR5⁺PD1^hiCXCR3⁺ cells and CD4⁺CD25^-CD127⁺CXCR5⁺PD1^hiCXCR3⁻ cells from three HD and CD4⁺CD25⁻ CD127⁺CXCR5⁺PD1^hi cells from four CVID+AIC subjects were FACS sorted. RNA was isolated using the Direct-zol™ RNA MicroPrep (Zymo Research) and transcriptional profiles were determined using the Clarion™ S Pico Array (Affymetrix) in accordance with the manufacturer’s instructions.

Differentially expressed transcripts were subjected to hierarchical clustering (Affymetrix console) and gene set enrichment analysis with the Molecular Signatures Database version 6.2 (http://software.broadinstitute.org/gsea/index.jsp)^38,39.

Statistics.

Data were analyzed with GraphPad Prism using either Mann-Whitney U tests, or paired Student’s t-tests, or Pearson correlation coefficient tests.

RESULTS

Activated and follicular features predominate in CVID+AIC T helper cell landscapes

From a large, genetically diverse CVID cohort (n=69; Table E1) we chose three representative subjects with AICs, three without AICs and three age/gender matched HD controls for high dimensional T helper cell analyses. All three CVID+AIC subjects lacked detectable serum IgA; they included a 57-year-old female with immune thrombocytopenia carrying a heterozygous W783X NFKB1 variant, a 30-year-old male with both Evans Syndrome and granulomatous lymphocytic interstitial lung disease carrying a heterozygous C172Y TACI variant, and a 15-year-old male with Evans Syndrome but no identifiable mutations in CVID-associated genes. All three CVID-AIC subjects possessed detectable, albeit decreased, serum IgA concentrations (mean 9 mg/dL). All HD controls were IgA replete. To catalogue the phenotypic differences between these subjects’ circulating CD4⁺ T lymphocytes, we stained them with a custom panel of 38 heavy metal-conjugated antibodies targeting activation markers, chemokine receptors, inhibitory proteins, transcription factors, and cell cycle proteins (Table E2). Stained cells were analyzed using mass cytometry or cytometry by time-of-flight (CyTOF) to produce high-dimensional datasets that we subsequently reduced into two-dimensional t-distributed Stochastic Neighbor Embedding (t-SNE) plots³⁶. Some markers were expressed more ubiquitously whereas others were expressed only by distinct clusters of T helper cells (Fig E1). As plots within subject groups displayed similarities (Fig E2), we pooled them and scrutinized each groups’ plot with self-organizing map software which ignores a priori assumptions about T-cell biology (Fig E3)³⁷. Using this hypothesis generating strategy, we found HD and CVID-AIC subject pooled CD4⁺ T cell t-SNE plots were virtually indistinguishable, whereas CVID+AIC subject plots displayed unique T-cell distribution patterns (Fig 1A, left) that were concentrated in three spatially discrete populations (P1, P2 and P3, Fig 1, A right). In CVID+AIC samples the frequency of cells belonging to P1 (2.1%) was clearly diminished compared to CVID-AIC (29.6%) and HD P1 frequencies (27.3%), with redistribution into P2 (27.8%) and P3 (5.2%). Only 1.3–2.2% of CVID-AIC and HD CD4⁺ cells belonged to P2 and only 0.5–1.1% belonged to P3.

Figure 1. High-dimensional analyses of CVID+AIC subject circulating CD4⁺ T cells reveal distinct distribution patterns and pervasive follicular signatures.

(A) 38-dimension mass cytometry analyses of HD (n=3), CVID-AIC subject (n=3) and CVID+AIC subject (n=3) peripheral blood CD4⁺ T cells are pooled and reduced into 2-dimensional t-SNE plots (left). Between plots cell distribution differences are concentrated into three spatially-distinct populations: P1, P2 and P3 ⁶⁹. Density of cells co-expressing surface markers indicative of (B) naive cells (C) Tfh cells and (D) “Tregs” in healthy donors are overlaid on ungated t-SNE plots. Heatmaps ⁶⁹ compare the frequency of cells from different populations expressing the indicated markers. Compared populations include, (B) naive cells belonging to P1, P2 or a third CD45RA⁺CD45RO⁻ population (P1’) (C) cTfh cells from each subject by group and (D) cells in the “Treg” gate belonging to P3 and not.

To determine identifying features of P1, P2 and P3 we first mapped conventional markers of naive T cells (CD45RA⁺CD45RO^-) onto our t-SNE plots (green; Fig 1, B left). In HD and CVID-AIC samples 41.6–51% of naive T cells belonged to P1 with the remaining cells belonging to a phenotypically related population (P1’). P1’ cells were distinct from P1 cells by expression of activation markers associated with follicular T cell development including CCR7, T-cell factor one (TCF1), CD38 and inducible T-cell costimulator (ICOS) expression (Fig 1B, right; Fig E1)^40,41. Instead of belonging to P1, most naive CVID+AIC T helper cells mapped to P1’ (49.7%) and to a lesser extent P2 (10.3%). Like P1’, naive P2 cells expressed CCR7, TCF1, CD38, and ICOS but additionally expressed CXCR5 and programmed cell death protein one (PD1), suggesting greater follicular commitment (Fig 1B, right; Fig E1)^42–44.

Given the observed follicular skewing of CVID+AIC naive T helper cells, we next mapped canonical Tfh surface markers CXCR5 and PD1 onto t-SNE plots (Fig 1, C left). 85.9% of CVID+AIC T helper cells expressing these markers belonged to P2. As mentioned above, some P2 cells were naive but most (>90%) expressed CD45RO. Although scarce in HD/CVID-AIC samples and abundant in CVID+AIC samples, CXCR5⁺PD1⁺ cells from all three subject groups appeared phenotypically similar except CVID+AIC cells expressed more ICOS, CXCR3, and T-cell immunoreceptor with Ig and ITIM domains (TIGIT) but less CCR6 (Fig 1, C right). Although absent in HD and CVID-AIC subjects, 12.8% of CXCR5⁺PD1⁺ cells from CVID+AIC subjects expressed the interleukin-2 (IL-2) receptor alpha chain (CD25; Fig E4), a surface marker categorically absent in bona fide Tfh cells from secondary lymphoid tissues⁴⁵. In CVID+AIC blood samples these CD25 expressing follicular cells primarily localized to P3.

CD25 expression by healthy donor T helper cells denotes an activated or regulatory phenotype while IL-7 receptor (CD127) co-expression excludes FOXP3⁺ Tregs⁴⁶. As FOXP3 staining requires fixation and permeabilization, the CD25^hiCD127^lo/- “Treg” sorting gate has historically been used to isolate viable human Tregs, including CVID Tregs^3,8,26, for in vitro suppression assays. To compare phenotypic diversity within the traditional “Treg” gate of CVID subjects we mapped CD25 expression and CD127 low/non-expression onto our t-SNE plots (Fig. 1, D, left). We found a substantially increased proportion of CD25^hiCD127^lo/- cells belonged to P3 in CVID+AIC subject samples (21.7%) whereas this distribution pattern was rare in HD (3.9%) and CVID-AIC subject samples (3.3%). Compared to CD25^hiCD127^lo/- CVID+AIC cells lying outside P3 (“Treg”-P3), those belonging to P3 expressed lower amounts of regulatory transcription factors (FOXP3, Helios) but higher amounts of classic Tfh (ICOS, CXCR5, PD1), and activation/cell cycle markers (CD95, Ki67; Fig. 1, D right; Fig E1). Hence, the CVID+AIC T helper cell landscape demonstrates unique features including naive T cell activation, prominent follicular skewing, and the presence of a uniquely expanded population possessing some features that are characteristic of Tregs (CD25^hiCD127^lo/^-) but others that are not (FOXP3^lo/- CXCR5⁺PD1^hiKi67⁺).

Endotoxemia favors early-stage follicular differentiation

To confirm and further explore the T-cell alterations observed in our CyTOF analysis, we analyzed peripheral blood samples from our full subject cohort (Table E1) and an age/gender-matched cohorts of immunocompent AIC subjects and HD controls using standard flow cytometry (FACS). We first turned our attention to naive (CD45RA⁺CD45RO^-) T helper cells which comprised on average 56.6% of HD, 54.4% of AIC and 58.1% CVID-AIC CD4⁺ cells but only 23.1% of CVID+AIC cells (p<0.0001 for each comparison; Fig E5). As suggested by our CyTOF analyses, a small percentage of CVID+AIC naive T helper cells, but not cells from other subjects, co-expressed CXCR5 and PD1 (mean 4.7%; Fig E6). Diminishment of the CVID+AIC naive T helper cell pool was primarily due to a twofold relative decrease in CD45RA⁺CD31⁺ RTEs (Fig 2, A)⁴⁷, a reduction not observed in the CD8 compartment (Fig E7). Although scarce, a significantly greater fraction of CVID+AIC CD4⁺ RTEs expressed activation (CD95), cell-cycle (Ki67), and follicular molecules (CXCR5, PD1) than counterparts from CVID-AIC subjects and HD controls (Fig 2, B–D). Thus as early as thymic egress, CVID+AIC T helper cells display signs of activation, proliferation, and Tfh differentiation.

Figure 2. CVID+AIC CD4⁺ recent thymic emigrants (RTEs) express activation and follicular markers.

(A) The frequencies of RTEs among total CD4⁺ T cells are displayed for representative subjects (left) and all available subjects ⁶⁹. Comparison of (B) CD95 expression intensities and (C) Ki67 (D) CXCR5, and PD1 expression frequencies on CD4⁺ RTEs from representative (left) and all available ⁶⁹ HD, AIC, CVID-AIC and CVID+AIC subjects. *P<.05, **P<.01, and ****P<.0001, Mann-Whitney U tests.

Tfh differentiation requires naive T helper cells and antigen presenting cells (APCs) to interact via soluble factors (IL-12, IL-6, activin A)^48–50 and cell-cell contacts (TCR-MHCII, ICOS-ICOSL)⁵¹. To determine if serum soluble factors contribute to the observed follicular skewing of CVID+AIC T helper cells, we cultured HD naive CD4⁺ T cells with anti-CD2/CD3/CD28 beads, recombinant ICOS ligand (rICOSL), and either HD, or CVID-AIC, or CVID+AIC subject plasma. After five days, an average of 6% of cells exposed to HD plasma and 5.7% exposed to CVID-AIC plasma co-expressed CXCR5 and PD1. When the same HD cells were cultured with CVID+AIC plasma, co-expression frequency increased significantly to 11% (p<0.01 and p<0.001 respectively, Fig 3, A). In all cases, CXCR5⁺PD1⁺ cells expressed more BCL6, the master Tfh transcriptional regulator, and more OX40, an important amplifier of Tfh development⁵², than CXCR5⁻PD1⁻ cells from the same cultures (Fig E8).

Figure 3. Endotoxemia promotes T follicular helper cell (Tfh) differentiation.

(A) Day five CXCR5/PD1 coexpression on HD CD45RO⁻ naive cells cultured with anti-CD3/CD28 beads, recombinant ICOSL and 20% HD, CVID-AIC, or CVID+AIC plasma with and without polymyxin B (PMB). (B) Day two CCR7, ICOS and CD95 expression and (C) day five CXCR5 and PD1 expression on activated HD CD45RO^- cells similarly cultured but in 20% fetal bovine serum spiked with LPS (100ng/ml) or not. (D) Activin A plasma concentrations between groups and against paired cTfh cell frequencies are displayed. *P<.05, **P<.01, ***P<.001, and ****P<.0001, Mann-Whitney U tests or paired student’s t-tests or a Pearson correlation coefficient.

Recently we reported that most CVID+AIC subjects, including many analyzed here, possess undetectable serum IgA concentrations and varying degrees of endotoxemia⁸. To determine if CVID+AIC plasma endotoxin skews T-cell development, we neutralized it with polymyxin B and repeated in vitro Tfh differentiation experiments. At day five, CXCR5 PD1 co-expression frequencies dropped precipitously in polymyxin B containing cultures (4.6% vs. 11%, p<0.01; Fig 3, A). To confirm this effect was due to LPS neutralization, we again Tfh differentiated HD naive T helper cells in culture but replaced subject plasma with LPS-spiked FBS (0.5µ g/ml). After two days of LPS exposure, cells began expressing characteristic P1’ cell surface markers CCR7, CD95 and ICOS (Fig 3, B). After five days of LPS exposure, HD cells co-expressed CXCR5 and PD1 at a frequency similar to cells cultured in CVID+AIC plasma (Fig 3, C). Thus, the endotoxin contained within CVID+AIC plasma directly promotes follicular molecule expression by CD4⁺ T cells in vitro.

The inflammatory consequences of endotoxemia are diverse but include myeloid lineage cell release of inflammatory cytokines, including a powerful inducer of Tfh differentiation, activin A^50,53. To determine if chronic CVID-related endotoxemia was associated with greater activin A release, we measured its concentration in subject plasma. Indeed, the average activin A concentrations were much higher in CVID+AIC samples (1722 pg/ml) than in HD, AIC and CVID-AIC samples (526, 791 and 904 pg/ml, P<0.0001, P<0.001, P<0.05 respectively; Fig 3, D), whereas concentrations of other Tfh promoting cytokines, like IL-12 and IL-6, were not different (Fig E9)^48,49,54. Further, we found that activin A concentrations positively correlated with cTfh frequencies in all analyzed subjects, regardless of their disease status (r= 0.43, P<0.0001, Fig 3, D), and that serum IgA concentrations negatively correlated with both cTfh frequencies (r= −0.54, p<0.001) and plasma activin A concentrations (r= −0.37, p<0.01; Fig E10).

In addition to promoting release of soluble mediators like activin A, endotoxin also induces ICOSL expression by APCs, including human monocytoid lines and monocyte-derived DCs⁵⁵. To determine if CVID+AIC T helper cells are more likely to encounter monocyte expressed ICOSL in vivo we surveyed different circulating monocyte subsets for ICOSL expression (Fig 4, A). In all subjects we found ICOSL to be primarily expressed on CD14⁺⁺CD16⁺ intermediate monocytes (iMo, Fig 4, C), a population enlarged nearly three-fold in CVID+AIC blood samples versus CVID-AIC, AIC and HD samples (15.6% vs. 5.7%, 4.1% and 6.1%, respectively; P<0.0001 for each comparison, Fig 4, B, Fig E11, Fig E12). The absolute number of ICOSL expressing iMo cells was also highest in CVID+AIC subjects. Hence, either through a direct effect or via activin A/ICOSL induction, endotoxin promotes early helper T cell commitment to the follicular lineage.

Figure 4. CVID+AIC monocytes express increased ICOSL.

(A) Representative dot plot (upper) of classical monocyte (cMo, CD14⁺CD16⁻), intermediate monocyte (iMo, CD14⁺CD16⁺) and non-classical monocyte (ncMo, CD14^intCD16⁺). (B) HD, AIC, CVID-AIC and CVID+AIC iMo frequencies. (C) ICOSL mean fluorescence intensities on monocyte subsets from representative donors (left) and all subjects ⁶⁹. **P<.01, ****P<.0001, Mann-Whitney U tests.

CVID+AIC cTfh cells provide efficient B-cell help

Unlike more phenotypically homogenous tonsillar Tfh cells, cTfh cells exhibit more heterogeneous chemokine receptor expression, cytokine secretion, and functional profiles^44,56. For instance, healthy donor cTfh cells can be divided into at least two distinct functional subsets: 1) a CXCR3 expressing, interferon gamma (IFNγ) secreting subset that does not provide efficient B-cell help, and 2) a CXCR3 non-expressing, IL-21, IL-4 and IL-17 secreting subset that does (Fig E13 and E14)⁵⁶. As CVID+AIC subjects are hypogammaglobulinemic, yet also possess enlarged cTfh pools (Fig 1, C and Fig E15)⁸, we sought to compare CXCR3 expression, cytokine secretion, and functionality of their cTfh cells with counterparts from HDs, AIC only subjects and CVID-AIC subject whose cTfh frequencies were not elevated even in the context of non-cytopenic autoimmune diseases. Unlike cells from other subjects which were equally divided between CXCR3 expressing and non-expressing cells, CVID+AIC cTfh cells uniformly and overwhelmingly (>90%) express CXCR3⁺ (p<0.0001 against all other subject groups; Fig 5, A). Corresponding with increased CXCR3⁺ cTfh frequencies, CVID+AIC subject samples also contained significantly higher plasma concentrations of CXCR3 ligands CXCL9, CXCL10 and CXCL11 (Fig E16, A). CCR4 and CCR6 ligands, CCL17 and CCL20, were not significantly different between subject groups (Fig E16, B).

Figure 5. Most CVID+AIC cTfh cells express CXCR3⁺ but are functionally distinct from HD CXCR3⁺ cTfh cells.

(A) The frequencies of cTfh cells expressing CXCR3 are displayed for representative subjects (left) and all available subjects ⁶⁹ including immunocompetent autoimmune cytopenia ²⁰ subjects. (B) Differentially expressed transcripts between CVID+AIC cTfh and HD CXCR3⁺ cTfh are displayed in a volcano plot with enriched gene sets listed (C) The frequency of activated cTfh cells staining positive for intracellular cytokines are shown for representative (left) and all available subjects ⁶⁹. *P<.05, **P<.01, ***P<.001 and ****P<.0001, Mann-Whitney U tests.

Despite near universal CXCR3 expression, CVID+AIC cTfh cell transcriptomes were significantly different from HD CXCR3⁺ cTfh cells and assorted separately when subjected to a hierarchical cluster analysis algorithm (Fig E17). Gene set enrichment analyses revealed CVID cTfh cells displayed prominent anaerobic respiration and glycolysis signatures consistent with activation (data not shown). CVID+AIC cTfh cells also lacked the inflammatory IFNγ response transcriptional signature which defined HD CXCR3⁺ cTfh cells (Fig 5, B), a finding consistent with non-elevated CVID+AIC plasma IFNγ concentrations (Fig E18). Indeed, we found that CVID+AIC cTfh cells did not exhibit the IFNγ-dominant cytokine secretion profile characteristic of HD CXCR3⁺ cTfh cells (Fig 5, C). Instead they expressed IL-21 at a frequency twice that of HD and CVID-AIC cTfh cells (p<0.01 for both comparisons). Despite more IL-21, CVID+AIC cTfh cell cytokine secretion profiles still differed from characteristic CXCR3^- cTfh cells. For instance, IL-4 secreting cTfh cell frequencies were not different and IL-17 secreting cTfh cell frequencies were significantly lower in CVID+AIC cTfh cells than in HD and CVID-AIC cTfh cells.

Arguably it is function, not surface markers nor cytokine profiles, that should arbitrate helper T-cell identity. To determine if CVID+AIC cTfh cells are capable of providing B-cell help, we co-cultured them with either heterologous HD or homologous CVID+AIC naive B cells (CD19⁺CD21⁺IgM⁺CD27). At day seven we found CVID+AIC cTfh were just as effective as HD cTfh at inducing HD B-cell CSR and IgA/IgG secretion (Fig 6, A and B). In contrast, CVID+AIC naive B cells were generally unable to secrete class-switched antibodies no matter the cTfh donor. Thus, despite near unanimous CXCR3⁺expression, cTfh cells from CVID+AIC subjects are primarily IL-21 secreting cells capable of providing efficient help to responsive B cells.

Figure 6. CVID cTfh efficiently help receptive B-cells.

(A) Day seven IgG and IgA cell-surface expression frequencies on naive B cells (nB) from representative indicated subjects in homologous or heterologous co-cultures with cTfh cells from indicated subjects (left) and mean frequencies tallied from four separate experiments ⁶⁹. (B) Mean immunoglobulin concentrations in supernatants from the described co-cultures are displayed. Error bars indicate mean ± SEM. * P<.05, **P<.01. Mann-Whitney U tests.

Circulating CD25^hiCD127^loCXCR5⁺PD1⁺ helper cells are not regulatory T cells

Our high-dimensional CyTOF analysis of three pooled CVID+AIC subject samples identified a unique population of CD25^hiCD127^lo cells within the P3 gate that also expressed Tfh makers ICOS, CXCR5, and PD1 (Fig 1, D). To confirm this finding in our larger cohort and to further investigate this novel cell population, we analyzed follicular subsets contained within the CD25^hiCD127^lo/- “Treg” gate by FACS. We found HD, CVID-AIC, and immunocompetent AIC subjects possessed two distinct sub-populations (Fig 7, A and Fig E19). Both expressed the regulatory transcription factor FOXP3 (Fig 7, B) and suppressed HD T-effector cell proliferation in vitro (Fig E20). Only one expressed CXCR5 consistent with circulating T follicular regulatory cells (cTfr shown in green)⁵⁷. The CXCR5^- population, which comprised approximately 75% of the CD25^hiCD127^lo gate, corresponded to traditional Tregs (grey).

Figure 7. The CVID+AIC CD25^hiCD127^lo “Treg gate” is contaminated by a poorly suppressive cTfh-like subset.

(A) Peripheral CD4⁺ cells in the CD25^hiCD127^lo/- “Treg” gate include conventional Tregs (grey), CXCR5⁺PD1^lo circulating T follicular regulatory cells (Tfr, green) and a third CXCR5⁺PD1^hi population, CD25^hi Tfh-like cells (red). (B) FOXP3, Ki67, (C) IL-21 and IL-10 expression and expression frequencies on CVID+AIC and HD T cell subsets (D) Representative histograms (left) of CFSE-labeled HD naive T effector cells (Teff) stimulated (solid line) or not (dashed) in cocultures with indicated CVID+AIC T helper cell subsets. Columns ⁶⁹ represent mean percent inhibition relative to unstimulated Teff cells. Error bars indicate mean ± SEM. * P<.05, **P<.01, ***P<.001, and ****P<.0001, Mann-Whitney U tests or paired t-tests.

CVID+AIC subjects possessed a third CD25^hiCD127^lo subpopulation expressing CXCR5, PD1, Ki67, ICOS, IL-21, and IL-10 (red, Fig 7, B, C and E21, A). The subpopulation comprised an average 12.4% of the “Treg” gate, a share gained primarily at the expense of cTfrs (Fig 7, A). Although highly expressing CD25, IL-10 and the inhibitory co-receptor molecule cytotoxic T-Lymphocyte associated protein four (CTLA4; Fig 7, C and Fig E21, B), these cells did not stain FOXP3 positive (Fig 7, B) and could not suppress T-effector cell proliferation in vitro (Fig 7, D). Furthermore, depletion of PD1⁺CXCR5⁺ cells significantly improved the suppressive function of remaining CVID+AIC FOXP3⁺ cells (Fig 7, D). Hence, although they highly express CD25 and do not express CD127, circulating CD25^hiCD127^loCXCR5⁺PD1⁺ T helper cells are not regulatory cells.

DISCUSSION

Here, using high-dimensional analysis, we systematically catalogued the phenotypic distribution of circulating CVID CD4⁺ T cells. Our data establishes CVID subjects in our cohort to display one of two immune phenotypes: either pure antibody deficiency without significant T helper cell impacts (CVID-AIC), or antibody deficiency, including undetectable serum IgA, complicated by the T cell-altering effects of systemic endotoxemia (CVID+AIC).

Endotoxin’s adjuvantic properties were first described by Condie and Good in 1955⁵⁸ and were later attributed to direct TLR4-mediated mitogenic effects on murine B cells^59,60. As human B cells do not express TLR4, we think it unlikely LPS exerts direct effects on CVID B cells⁶¹. Instead, our own in vitro data demonstrate plasma endotoxin promotes a follicular program in naive T cells, a population known to express TLR4⁶². We propose the observed Tfh skewing effect of endotoxin may maintain antibody/microbial balance at mucosal surfaces in immunocompetent hosts but may promote pathologic autoantibody production in antibody-deficient CVID subjects. Although others have demonstrated a Th1, not a Tfh bias in LPS-exposed human T cells, our results are likely influenced by the additional in vitro effects of rICOSL. Induced by LPS⁵⁵ and expressed by an increased number of CVID+AIC monocytes, it is likely endogenous ICOSL similarly promotes CVID+AIC subject Tfh differentiation in vivo. Another protein induced by endotoxin⁵³ and highly concentrated in CVID+AIC plasma, is the inflammatory protein activin A. Activin A was previously shown to be an elite Tfh lineage promoter in a high-throughput in vitro screen of more than 2000 human inflammatory proteins⁵⁰. The strong correlations between cTfh cell frequencies, plasma activin A concentrations, and serum IgA that we report here provides the first in vivo proof of this concept in a human disease.

Several CVID-related T helper cell phenomena found by our analyses were previously observed in more narrowly focused studies. Our intentionally broader approach provides us the opportunity to both corroborate earlier findings and, in some cases, re-interpret them aided by additional contextual clues. For instance, reduced naive T cell frequencies and T cell receptor excision circle concentrations in CVID subjects were previously thought to signal poor thymic output^23–25, yet our data indicate CVID+AIC RTEs are likely scarce because they leave the thymus in a pre-activated state favoring proliferation and accelerated differentiation. Similarly, it was recently proposed, but not demonstrated in co-cultures, that increased CXCR3 expression by CVID cTFh cells was the cause of traditional CVID B-cell defects in CSR, SHM and memory B-cell formation²⁷. Although we too find increased frequencies of CXCR3 expressing cTfh cells in CVID+AIC subjects, in our hands these cells are primarily IL-21, not IFNγ producers that are as effective at promoting in vitro class-switched antibody production as HD cTfh cells are. Further, instead of predicting poor helper function, we find CXCR3 expression on CVID+AIC Tfh cells to instead correlate with stark IgA deficiency and endotoxin-induced release of CXCR3 ligands, CXCL9, CXCL10, and CXCL11 from innate cells^63–67. Thus, other than a few notable exceptions^22,68,69, CVID-related hypogammaglobulinemia appears primarily attributable to intrinsically unresponsive B-cells and not to Tfh dysfunction. Although it is possible that IFNγ producing cTfh cells may predominate in other CVID cohorts more enriched with granulomatous diseases than ours^27,70, our data clearly demonstrate that cTfh function cannot not be intuited from chemokine receptor expression alone, especially in immunodeficient subjects.

Finally, previous reports, including our own, demonstrate CVID CD25^hiCD127^- CVID “Tregs” perform poorly in traditional in vitro suppression assays^3,8,26. Here we demonstrate this phenomenon is unlikely due to Treg intrinsic defects but is instead secondary to contamination of the traditional human “Treg” sorting gate with CD25^hi cTfh-like cells. A notable exception may be truly dysfunctional CTLA4-haploinsufficient or LRBA-deficient Tregs that poorly suppress^71,72 even when depleted of CD25^hi cTfh cells⁷³. In all cases, and especially in CVID, CXCR5⁺PD1⁺ cells, including CD25^hi cTfh-like cells, should be excluded from in vitro Treg suppression assays to avoid confounding effects. The biologic role of the IL-10 producing CD25^hi cTfh-like cells described here is currently unclear, but phenotypicaly similar cells with regulatory function were recently described in human tonsil⁷⁴. The remarkable expansion of the circulating CD25^hi cTfh-like cells we describe here suggests they are pathologicly linked to autoantibody production. Identification, localization, and functional investigation of tissue resident counterparts to circulating CD25^hi cTfh-like cells in the secondary lymphoid tissues of HDs and CVID subjects will be critical to understanding if and how they influence disease pathogenesis.

Key Messages.

• Endotoxin promotes Tfh cell differentiation in vitro.

• CVID Tfh cells can help receptive healthy donor B cells secrete class-switched immunoglobulins.

• Tregs from many CVID subjects are suppressive if CD25⁺ Tfh-like cells are excluded from in vitro assays.

Abbreviations used:

AIC

autoimmune cytopenia

APC

antigen presenting cells

CFSE

Carboxyfluorescein diacetate succinimidyl ester

CSR

class-switch recombination

CyTOF

cytometry by time of flight

CVID

common variable immunodeficiency

FACS

Flow cytometry

ICOS

inducible T-cell costimulatory

PD1

programmed cell death protein one

RTE

recent thymic emigrants

SHM

somatic hypermutation

TCF1

T cell factor one

Tfh

T follicular helper

t-SNE

t-distributed stochastic neighbor embedding

Treg

regulatory T cell