Human T-bet governs the generation of a distinct subset of CD11chighCD21low B cells

High level expression of the transcription factor T-bet characterizes a phenotypically distinct murine B-cell population known as ‘age-associated B cells’ (ABCs). T-bet-deficient mice have reduced ABCs and impaired humoral immunity. We describe a patient with inherited T-bet deficiency and largely normal humoral immunity including intact somatic hypermutation, affinity maturation and memory B-cell formation in vivo, and B-cell differentiation into Ig-producing plasmablasts in vitro. Nevertheless, the patient exhibited skewed class switching to IgG1, IgG4 and IgE, along with reduced IgG2, both in vivo and in vitro. Moreover, T-bet was required for the in vivo and in vitro development of a distinct subset of human B cells characterized by reduced expression of CD21, and the concomitantly high expression of CD19, CD20, CD11c, FCRL5, and T-bet, a phenotype which shares many features with murine ABCs. Mechanistically, human T-bet governed CD21loCD11chi B cell differentiation by controlling chromatin accessibility of lineage-defining genes in these cells: FAS, IL21R, SEC61B, DUSP4, DAPP1, SOX5, CD79B and CXCR4. Thus, human T-bet is largely redundant for long-lived protective humoral immunity but is essential for the development of a distinct subset of human CD11chi CD21lo B cells.

T-bet has also been implicated in fate decisions for murine B-cell subsets. ABC, as a subset of T-bet + B cells, has been the focus of growing interest over the last decade (4,6,9,23). In mice, numbers of these cells increase in lymphoid tissues with age (4,23,24), as well as following infection with certain viruses (6,25), bacteria (26,27), parasites (28), and in the context of autoimmunity (29)(30)(31). Depletion of ABCs in mouse models of systemic lupus erythematosus (SLE) decreases anti-chromatin IgG levels (32) and disease severity (30), but the non-redundant function of ABCs remains much less well understood than the conditions in which this B-cell subset expands in mice. Moreover, T-bet expression in murine B cell subsets has yet to be clearly defined, as it is detected in only ~50% of all ABCs (8,9). It is unclear whether this heterogeneity results from the presence of distinct subsets of ABCs, or simply the existence of numerous differentiation stages of the same ABC lineage. Furthermore, the requirement for T-bet in the development of murine ABCs in vivo is a matter of debate, as some studies indicated T-bet is indispensable for ABC formation (6,30), whereas others showed ABCs are generated normally from T-bet deficient B cells (33,34). It is thus unclear if T-bet is essential or redundant for generating ABCs or a subset of ABCs.
In humans, numerous subsets of B cells resembling murine ABCs have been identified. These include CD21 lo T-bet + B cells that were first discovered to be increased in some individuals with common variable immunodeficiency (CVID) (35). Subsequent studies identified similar populations that have been termed "activated naïve-like B cells", "atypical memory", or "IgD − CD27 − double-negative" B cells that are significantly expanded in as chronic viral (HIV, HCV) and malaria infections (36)(37)(38)(39)(40)(41)(42), and autoimmune conditions including SLE and rheumatoid arthritis (9,31,(43)(44)(45)(46)(47)(48)(49). CD21 lo T-bet + B cells are also detected at increased frequencies following influenza vaccination (50,51) and SARS-CoV-2 infection causing severe COVID19 (52,53). Thus, depending on the condition, human CD21 lo T-bet + B cells might be pathogenic, protective, or both. In contrast to mice, CD21 lo Tbet + B cells only mildly increase until 30 years of age but do not increase afterwards (54). Moreover, the role of T-bet in the development, maintenance, and function of human B cells, including CD21 lo T-bet + B cells, and humoral immunity in response to pathogen infections, is completely unknown.
We recently reported the first patient with autosomal recessive (AR) complete T-bet deficiency (55,56). This patient carries a homozygous in-del mutation in TBX21 that abolishes DNA-binding activity without abrogating protein expression (55). By disrupting the development of IFN-γ-producing innate or innate-like adaptive lymphocytes, human T-bet deficiency causes Mendelian susceptibility to mycobacterial disease (55). The patient also developed upper airway inflammation and peripheral eosinophilia due to T H 2-skewing of T-bet-deficient CD4 + T cells (56). However, the development and function of B cells in this patient have not been studied. Human T-bet deficiency therefore provides a unique opportunity to determine the physiological requirement for T-bet in the induction and maintenance of humoral immunity, and the development and homeostasis of T-betexpressing B cells. By thoroughly assessing B cells in inherited T-bet deficiency, we delineated the essential and redundant roles of human T-bet in humoral immunity and serological memory.

Altered class switch recombination and production of polyclonal antibodies in human T-bet deficiency
To investigate the impact of T-bet deficiency on human B cells, we first performed flow cytometric immunophenotyping of the patient (P or M/M -referring to genotype Mutant/ Mutant)'s peripheral blood mononuclear cells (PBMCs). Frequencies of total B cells in P was greater than in healthy donors (aged 16 -65 years), but similar to age-matched healthy donors (n=5-8, range 1-7 yrs, mean age 3.7 yrs) (Fig. 1A). This is consistent with the decline in B cell frequency in peripheral blood during the first decade of life (57,58). Proportions of transitional, naïve, and memory B-cell subsets, as well as of IgG + and IgA + memory B cells, were also similar between P, healthy donors, and age-matched controls ( Fig. 1B and C). We investigated whether T-bet regulated production of different classes of Ig by human B cells in vivo, as observed for mice (15). Plasma IgA and IgM levels in P were normal, while IgE (Fig. 1D). Consistent with these serological findings, which reflect constitutive production of Ig isotypes by plasma cells, memory B cells in the T-bet-deficient patient contained markedly increased proportions of IgG1 + and fewer IgG2 + cells than healthy donors (Fig. S1A). We also investigated antigen-specific antibody (Ab) responses of the T-bet deficient patient by determining levels of IgG specific for tetanus toxoid, diphtheria toxoid, and Haemophilus influenzae b, which P had been vaccinated against. P had IgG levels against these three vaccines in the normal range of healthy donors (Table S1). In addition, his antibody titer against pneumococcal antigen was also normal (Table S1). Thus, while human T-bet deficiency does not impact the generation of antigen-specific Ab or differentiation of human naïve B cells into memory or plasma cells per se, it does skew Ig class switching towards IgG1, IgG4 and IgE, and away from IgG2.

Altered B-cell receptor repertoire in inherited human T-bet deficiency
We further investigated the consequences of inherited human T-bet deficiency on humoral immunity by analyzing the B cell receptor (BCR) repertoire in P and comparing it to healthy donors. This confirmed the flow cytometric analysis of memory B cells (Fig. 1C), and serum Ig levels (Fig. 1D), which suggested CSR was generally intact in P, however with some differences (Fig. 1E). Specifically, we detected fewer clones expressing IgG2 and more expressing IgG4 in P (Fig. 1E). This difference was particularly marked in comparison with age-matched healthy children and is consistent with the lower levels of serum IgG2 and higher serum levels of IgG4 in P relative to healthy donors. Somatic hypermutation (SHM) was also intact (Fig. 1F). The targeting and nature of SHM within complementaritydetermining regions of the Ig H and L chains of human T-bet-deficient total B cells were consistent with B cells from healthy donors. The repertoire of IGHM-expressing B cells from P was slightly less diverse than that of healthy donors (Fig. S1B), probably due to the presence of more expanded clones within the IgM repertoire (Fig. S1C). Diversity was also lower for IGK and IGL in P (Fig. S1B). This may also be due to the presence of some larger clones, based on D20 metrics (Fig. S1D). By contrast, the diversity of IgG and IgA expressed by T-bet-deficient B cells was similar to healthy donors (Fig. S1D). Further analysis indicated that IGH, IGK and IGL gene usage in B cells of P was not different to B cells from healthy donors. However, within IgM-expressing T-bet-deficient B cells, usage of IGHV3-15, IGHV3-43 and IGHV7-4-1 genes was higher than in healthy donors (Fig. S1E). Interestingly, IgG + memory B cells of P also displayed higher usage of IGHV4-34 (20.2% of IgG1 clones) than those from healthy donors (8.5% of clones) (Fig. S1F). Although IGHV4-34 genes are expressed by autoreactive B cells, the significance of an enriched population of T-bet-deficient IgG + B cells expressing putative self-reactive BCRs is unknown, as there was no clinical or serological evidence of autoantibodies in the patient. Overall, analysis of the BCR repertoire revealed that inherited human T-bet deficiency increased and decreased CSR to IgG4 and IgG2 respectively, but otherwise had no major impact on the generation of diverse polyclonal antibodies or affinity maturation.

T-bet-deficient B cells differentiate normally into Ig-secreting plasmablasts in vitro
We then investigated intrinsic consequences of T-bet deficiency on the differentiation of human B cells into Ig-secreting plasmablasts in vitro. Naïve and memory B cells were cytokines (59). CD40L/CpG induced marked IgM secretion by naïve B cells, at similar levels for P and healthy donors (Fig. 1G). Stimulation with CD40L/IL-21 resulted in similarly high levels of IgM, and induction of the class-switched isotypes IgG and IgA, in cultures of naïve B cells from healthy donors and from P ( Fig. 1H -J). T-bet-deficiency also had no effect on the ability of memory B cells to differentiate into plasmablasts secreting IgM or IgG, regardless of the nature of the stimulus: CD40L/CpG (Fig. 1K), CD40L/IL-10 (60, 61) (Fig. 1L), or CD40L/IL-21 (Fig. 1M). T-bet deficiency slightly increased IgA production by memory B cells stimulated with CD40L/IL-10 (Fig. 1L). Interestingly, T-betdeficient memory B cells tended to produce larger amounts of IgG than memory B cells from most healthy donors (Fig. 1M). Quantification of IgG subclasses revealed increases in secretion of IgG1 and IgG4 (Fig 1N and O). IgE secretion by memory B cells in response to CD40L/IL-4/IL-21 (62) was also unaffected by inherited T-bet deficiency (Fig. 1P). Thus, human T-bet deficiency does not affect CSR and B-cell differentiation in vitro in response to stimulation with cytokines (IL-4, IL-10, IL21), CD40L or TLR agonists. However, similar to findings for serum IgG subclasses, these in vitro functional analyses revealed that T-bet deficiency intrinsically alters the capacity of human B cells to differentiate into IgG subclass-specific plasma cells.

CD21 lo B cells are reduced in human T-bet deficiency
We then investigated whether CD19 hi CD21 lo B cells, corresponding to murine ABCs, were affected by inherited human T-bet deficiency. Conventional flow cytometry established that these cells represent ~2.5% of peripheral blood B cells in healthy aged-matched and adult donors but only ~0.5% in T-bet-deficient patient (Fig. S2A). Phenotypic analysis of CD19 hi CD21 lo B cells in healthy donors revealed downregulation of CCR7, CXCR4, CXCR5, and increased expression of CD11c, CXCR3, FCRL5, CD86, CD95, and T-bet relative to CD21 + B cells (Fig. S2B). Notably, expression of CXCR3, FCRL5 and T-bet by the residual CD19 hi CD21 lo B cells detected in the T-bet-deficient patient were not upregulated, and CCR7 was not down-regulated, relative to expression levels on CD21 + B cells (Fig. S2B).

T-bet deficiency disrupts generation of the CD11c + subset of CD21 lo B cells
We extended these findings by performing in-depth analysis of B-cell subsets with a 29color spectral flow cytometry panel. We focused on expression of CD21 and CD11c on circulating CD3 − CD56 − CD19 + CD20 + B cells as CD21 lo CD11c hi is a common phenotype of murine ABCs and human CD21 lo and atypical memory B cells (9,(63)(64)(65). CD21 lo B cells in healthy donors were heterogeneous, expressing various levels of CD11c ( Fig. 2A) (9,43,66). CD21 lo CD11c + B cells from healthy donors expressed higher levels of CD19 and T-bet than corresponding CD21 lo CD11c − B cells, which mostly lacked T-bet ( Fig.  2A -C). CD21 lo CD11c + B cells were markedly lower in P than in healthy donors ( Fig.  2A and B). The residual CD21 lo CD11c + B cells in peripheral blood of P displayed no detectable expression of T-bet and did not upregulate CD19 ( Fig. 2A). CD11c + T-bet + B-cells are reduced in IFN-γ deficient mice (34). By contrast, frequencies of CD21 lo CD11c + B cells in patients with inherited IFN-γR1 deficiency were similar to healthy donors ( Fig.  2A and B). However, proportions of CD21 lo CD11c + B cells with the highest expression of T-bet and CD19 were reduced in patients with IFN-γR1 complete deficiency compared to healthy donors ( Fig. 2A and C). This suggests that while IFN-γ signaling, which can induce T-bet in human and murine B cells (6,7), is not required for the in vivo development of CD21 lo CD11c + B cells in humans, IFN-γ does modulate differentiation of these cells, evidenced by lower expression of CD19 and T-bet on CD21 lo B cells from IFNGR1-deficient individuals (67).

IFN-γ, STAT1 and T-bet cooperate to induce the generation of human CD11c hi CD21 lo B cells in vivo
We investigated this aspect further by studying B cells from patients with AR complete STAT1 deficiency, which abolishes signaling via IFN-γ and other STAT1-dependent cytokines (68). Proportions of CD21 lo CD11c + B cells in AR STAT1-deficient patients were also similar to healthy donors ( Fig. 2D and E). Interestingly, STAT1-deficient CD21 lo CD11c + B cells expressed high levels of CD19, but only intermediate levels of CD11c (CD11c int ), whereas most of the CD21 lo CD11c + B cells of healthy donors were CD11c hi (Fig. 2D and F). Similarly, IFN-γR1-deficient CD21 lo CD11c + B cells expressed lower levels of CD11c compared to CD21 lo CD11c + B cells from most age-matched controls (Fig. S2C). Interestingly, frequencies of CD21 lo CD11c + B cells expressing the highest levels of CD11c and T-bet (i.e. CD11c hi T-bet hi ) in AR STAT1 or IFN-γR1 deficiencies were lower than in age-matched and most adult controls ( Fig. S2D and E). Strikingly, CD11c hi T-bet hi B cells were completely absent in T-bet deficiency (Fig. S2E). Thus, CD21 lo cells can be generated in the absence of T-bet, STAT1 or IFN-γR. However, T-bet is strictly required to induce the canonical CD21 lo CD11c + T-bet hi phenotype of this B-cell subset, while STAT1 or IFN-γR are only required to generate the CD11c hi T-bet hi subset of human CD21 lo CD11c + B cells.

Comprehensive characterization of CD21 lo CD11c + B cells
Based on spectral flow cytometry, CD21 lo CD19 hi CD11c + B cells constitute only a subset of human CD21 lo B cells ( Fig. 2A). We therefore further explored the nature of human B cell subsets defined by T-bet, CD21, and CD11c expression. In healthy donors, CD21 lo CD11c + B cells expressed T-bet more strongly than CD21 + B cells, but this expression was also heterogeneous, as most CD21 lo CD11c + B cells lacked T-bet (Fig. S2F). We overlaid CD21 lo CD11c + T-bet hi B cells with their CD21 lo CD11c + T-bet lo counterparts and CD21 hi CD11c − B cells (Fig. S2F). CD21 lo CD11c + T-bet hi B cells had the highest levels of CD19 and CD20 (Fig. 2G, H; Fig. S2F). We further defined the phenotype of CD21 lo CD11c + T-bet hi B cells relative to CD21 hi CD11c − , CD21 lo CD11c − , and CD21 lo CD11c + T-bet lo B cells. In addition to CD19 and CD20, HLA-DR and FCRL5 were expressed at the highest levels by CD21 lo CD11c + T-bet hi B cells, and at intermediate levels on CD21 lo CD11c + T-bet lo B cells of adult and age-matched healthy donors (Fig. 2I, Fig.  S2G). By contrast, CD23, CD24, CD38 and CD40 levels were lowest on CD21 lo CD11c + Tbet hi B cells, intermediate on CD21 lo CD11c + T-bet lo B cells, and highest on CD21 hi CD11c − B cells (Fig. S2H -K). CD95 was strongly expressed on all CD21 lo CD11c + B cells, with expression on CD21 lo CD11c + T-bet lo B cells being slightly higher than on CD21 lo CD11c + Tbet hi cells (Fig. S2L). CXCR3 and CD86 were strongly expressed, whereas CXCR4 was only weakly expressed, on CD21 lo CD11c + cells, but the level of expression of these molecules on CD21 lo CD11c + T-bet hi cells was similar or lower than on CD21 lo CD11c + T-bet lo B cells (Fig. S2M -O). CD21 lo CD11c + T-bet lo B cells also had unusually high levels CD11b, FCRL4, CD10, CD269, CD5, and CD80 generally not observed on their T-bet hi counterparts ( Fig. S2P -T). Thus, CD21 lo CD11c + CD19 hi CD20 hi T-bet + B cells constitute a distinct subset of human B cells, with expression of most signature markers being consistent across different age groups, and being T-bet dependent. In contrast, CD21 lo CD11c + T-bet lo B cells appear to be an intermediate precursor of CD21 lo CD11c + T-bet hi B cells or represent a distinct effector B-cell subset.

Unsupervised analysis confirms the depletion of CD21 lo CD11c + B cells in T-bet deficiency
To prevent bias introduced by manual gating, and identify other B-cell perturbations in inherited T-bet deficiency, we performed an unsupervised analysis of CD3 − CD56 − CD19 + CD20 + B cells with FlowSOM (69,70). We excluded T-bet from the initial clustering because T-bet-deficient B cells had lower basal levels of T-bet. When B-cell data for all individuals -including healthy donors, IFN-γR1-, STAT1-, and T-bet-deficient patients -were combined, 30 self-organizing clusters were identified ( . Notably, five B-cell clusters were enriched or diminished in the T-bet-deficient patient relative to healthy donors. The percentage of B cells corresponding to cluster 25 (CD21 hi CD24 hi CD40 hi CD23 − CD11c − CD38 int ), probably representing a subset of immature B cells, was higher in P than in adult and agematched healthy donors but lower in a patient with AR complete STAT1 deficiency ( Fig. 3A and 3B). By contrast, four clusters were depleted in P: clusters 13 and 14 (CD19 hi CD20 hi CD21 lo CD23 − CD24 − CD11c int CD27 − CD38 − CD40 lo ) and clusters 9 and 10 (CD19 hi CD20 hi CD21 lo CD23 − CD24 − CD11c hi CD27 +/− CD38 − CD40 lo ) (Fig. 3A, C and D). CD21 lo CD11c + B cells, including both CD11c int and CD11c hi , were therefore the only B-cell subset strictly dependent on T-bet (Fig. 3E). These four clusters (Clusters 9, 10, 13, 14) all expressed high levels of intracellular T-bet (Fig. S3), and the frequencies of each was lowest in the T-bet deficient patient ( Fig. 3C -E). Remarkably, these subsets had different developmental requirements. Depletion of clusters 9 and 10, but not clusters 13 or 14, was observed in patients with inherited AR STAT1 deficiency (Fig. 3A, C and D). However, IFN-γR is redundant for the development of cluster 9 (CD19 hi CD20 hi CD21 lo CD23 − CD11c hi CD95 hi FCRL4 hi CXCR3 hi ), whereas cluster 10 (CD19 hi CD20 hi CD21 lo CD23 − CD11c hi CD95 int CD27 lo ) seemingly required both STAT1 and IFN-γR for their proper development (Fig. 3C). Although signaling via T-bet was indispensable, intact signaling via IFN-γR or STAT1 was not required for the generation of CD21 lo CD11c int B cells corresponding to clusters 13 and 14 (Fig. 3D). This is consistent with our earlier observation that most CD21 lo CD11c + B cells in STAT1-deficient individuals expressed intermediate levels of CD11c (Fig. 2F).

IgA-and IgG-expressing B cells are enriched in CD21 lo CD11c + B cells, which correlates strongly with high T-bet expression
To evaluate expression of Ig isotypes by CD21 lo CD11c + and other B cell subsets, we integrated the surface expression of BCRs into a 30-color spectral flow phenotyping panel. Frequencies of CD21 lo CD11c + B cells in inherited human T-bet deficiency were significantly lower than in adult and age-matched controls (Fig. S4A). Frequencies of unswitched IgM + IgD + B cells were significantly reduced in CD21 lo CD11c + B cells from adult, age-matched healthy donors and T-bet deficiency compared to CD21 hi CD11c − and CD21 lo CD11c − B cell subsets ( Fig. 3F -H). By contrast, frequencies of IgG + or IgA + switched B cells were significantly increased in CD21 lo CD11c + B cells among healthy donors, but not T-bet deficient P, relative to these other B-cell subsets (Fig. 3F, 3G, 3I and 3J). These IgG + or IgA + switched CD21 lo CD11c + B cells did not express the surface memory marker CD27 (Fig. S4B). This was in striking contrast to IgG + or IgA + CD21 hi CD11c − B cells that were mostly CD27 hi (Fig. S4B). Most IgG + or IgA + CD21 lo CD11c + B cells from healthy donors displayed the highest expression of T-bet amongst all B-cell subsets examined (Fig. S4C -E and Fig. 3K). CD71 can delineate early activated CD20 hi B cells from resting naïve B cells (71). CD71 + CD20 hi B cells share some phenotypic similarity with CD21 hi CD11c − cells and CD21 lo CD11c + B cells such as high CD80 expression (Fig. S4F). Interestingly, frequencies of CD71 hi CD80 hi cells were slightly enriched in CD21 lo CD11c + B cells and drastically increased in the remaining few CD21 lo CD11c + B cells in T-bet deficiency ( Fig. S4F and G). Therefore, IgG + or IgA + B cells are enriched in CD21 lo CD11c + B cells in healthy donors but not T-bet deficiency. Expression of IgA and IgG on these CD21 lo CD11c + B cells correlate strongly with high T-bet expression, collectively suggesting a critical role of T-bet in the development of this distinct subset of B cells.

Single-cell proteotranscriptomics of CD21 lo B cells reveals the complete depletion of a distinct subset of human CD21 lo B cells in inherited T-bet deficiency
To test if the apparent absence of the CD21 lo CD11c + CD23 lo CD24 lo CD38 lo B-cell subset in T-bet deficiency may result from the lack of surface markers potentially regulated by T-bet, we performed single-cell (sc) proteotranscriptomic profiling of CD21 lo B cells. PBMCs from age-matched healthy donors, IFN-γR1-deficient and T-bet-deficient patients were labeled individually with oligonucleotide (OGN)-barcoded Hashtag Abs and OGNconjugated TotalSeq Abs against CD21, CD11c, CD95, CXCR3, and FCRL5. To avoid epitope competition, different clones of anti-CD21 and CD11c Abs that recognize unique epitopes, were used for flow and TotalSeq purposes ( Fig. S5A and B). Cells were pooled, FACS-sorted as live CD20 + CD21 lo B cells, followed by CITE-seq (cellular indexing of transcriptomes and epitopes by sequencing) and sc-VDJ sequencing ( Fig. 4A and B, Fig.  S5C). We analyzed 328 and 273 CD21 lo cells from two age-matched controls, 913 CD21 lo cells from the IFN-γR1-deficient patient, and 937 CD21 lo cells from P. Based on unbiased automated clustering, CD21 lo B cells from these four individuals formed five distinct clusters -0, 2, 3, and 4 ( Fig. S5D, Fig. 4C and Data file S1) -each with a unique transcriptomic signature (Fig. S5D). Cluster 0 corresponded to memory B cells, with high levels of CD27, CD99, LTB, and CD53; cluster 1 corresponded to transitional B cells with high levels of IGHM, IGHD, ISG20, IL4R and CCR7; cluster 2 had the highest levels of CXCR5 and SOCS3; cluster 3 corresponded to CD21 lo CD11c hi T-bet + B cells, defined by spectral flow cytometry, as shown by high expression of CD19, MS4A1 (CD20), ITGAX (CD11c), FCRL2, FCRL3, and FCRL5; cluster 4 corresponded to cells with high levels of AP-1 subunits as well as CD69, CD9, CD38 and CD55 (Fig. S5D). Strikingly, the T-betdeficient patient was completely devoid of cluster 3, which was present in age-matched controls ( Fig. 4D and E). Consistent with cytometric data, cluster 3 B cells were reducedbut still detectable -in IFN-γR1-deficient patients (Fig. 4E). Therefore, CD21 lo CD11c + B cells form a distinct T-bet-dependent B-cell subset with a distinguishable pattern of signature gene expression that is consistent with the unique pattern of surface marker expression.

Normal CSR and SHM in CD21 lo CD11c + B cells
We then analyzed sc-VDJ sequencing data. A few B cells expressed >1 Ig H or L chain gene, consistent with recent reports (74,75). The frequencies of cells with 3-4 consensus Ig chains were similar in P and age-matched controls (Fig. 5A). The frequencies of cells expressing >1 Ig H or L chain gene were similar for CD21 lo CD11c + B cells and other cells (Fig. 5B). We then assessed the clonality of each sample. Cells with identical sequences for the junctional region, the CDR3 region, or both IgH or IgL chains were considered to be of the same clonotype. Most B cells were unique. However, 1-10% of clonotypes were common to >1 CD21 lo B cell. The frequency of expanded clonotypes among CD21 lo CD11c + B cells did not differ between the T-bet-deficient patient and healthy donors (Fig. 5C). Frequencies of expanded clonotypes were similar among CD21 lo CD11c + B cells from healthy donors and IFN-γR1-deficient patients (Fig. 5D). Consistent with enriched IgG-and IgA-expressing B cells in CD21 lo CD11c + B cells (Fig. 4), single-cell studies showed that the frequencies of IgG-and IgA1/2-switched B cells among CD21 lo CD11c + B cells of most healthy donors were higher than those among other CD21 lo B-cell counterparts (Fig. 5E). We also assessed SHM frequency in a 280-nucleotide (nt) region upstream from the CDR3 site at single-cell level, by comparing the assembled sequences with their predicted germline sequences (76,77) (Fig. 5F). SHM frequencies ranged from 0 to 10%. CD21 lo B cells from T-bet-deficient or IFN-γR1-deficient patients displayed similar levels of SHM to CD21 lo B cells of age-matched controls (Fig. 5G). CD21 lo CD11c + B cells from healthy donors and IFN-γR1-deficient patients had SHM rates similar to those in other CD21 lo B-cell counterparts (Fig. 5H). Thus, CD21 lo CD11c + B cells had levels of clonal expansion, CSR and SHM similar to those of other CD21 lo B cells. However, IgG-and IgA-enriched B cells appear accumulate in CD21 lo CD11c + B cells.

Signaling via TLR, BCR and IFN-γ or IL-27 induce T-bet hi CD19 hi CXCR3 + B cells in vitro
The lack of CD21 lo CD11c + B cells in P demonstrates an indispensable role for T-bet in generating and/or maintaining these cells in vivo, but it remains unknown whether the requirement for T-bet in this process is B cell intrinsic or extrinsic. We addressed this by investigating the ability of naïve B cells from healthy donors to differentiate into T-betexpressing B cells in vitro. CpG stimulation induced expression of T-bet in, and increased CD19 on, human naïve B cells (Fig. 6A) (6,7,78). The proportion of T-bet hi CD19 hi B cells was further enhanced by costimulation with anti-Ig (αIg). Addition of IFN-γ or IL-27 modestly increased expression of T-bet and/or CD19 by CpG/αIg-primed naïve B cells (Fig.  6A-C). Extended phenotypic analysis of in vitro-derived T-bet hi CD19 hi B cells revealed that CpG/αIg-stimulation induced stronger FCRL5, CD95, and CD19 expression than CpG alone (Fig. 6B -D). Notably, addition of IFN-γ or IL-27 to CpG/αIg-stimulated naïve B cells led to further increases in expression of FCRL5, CD95, and CD19 (Fig. 6A, D, E). Furthermore, CXCR3 was induced on 30 to 65% of T-bet hi CD19 hi B cells following culture with CpG/αIg and either IFN-γ or IL-27, but on fewer than <10% of cells in response to CpG/αIg alone in vitro (Fig. 6F, G). Thus, these in vitro culture conditions provide a model to determine the intrinsic molecular requirements for generating T-bet hi CD19 hi B cells from naïve B cell precursors.

T-bet functions in a B cell-intrinsic manner to induce the generation of CD21 lo CD11c + B cells
We then subjected naïve B cells from P, and patients with autosomal dominant (AD) partial IFN-γR1 deficiency (79)(80)(81), dominant negative (DN) STAT3 deficiency (82), AD or AR STAT1 deficiency (82)(83)(84), AR complete IL-27R deficiency (unpublished), partial recessive JAK1 deficiency (unpublished), or AR IRAK4 deficiency (67,85) to these culture conditions. As expected, IRAK4 deficiency completely abolished induction of T-bet in B cells stimulated with CpG alone or together with other stimuli, establishing a requirement for TLR signaling in generating T-bet-expressing B cells in vitro (Fig. 6H). In contrast, T-bet was induced in T-bet-, AD STAT1, AR STAT1, AR IL27R-and AD IFN-γR1-deficient naïve B cells stimulated with CpG/αIg, CpG/αIg/IFN-γ or CpG/αIg/IL-27 (Fig. 6H). However, T-bet-deficient B cells had lower levels of T-bet than naïve B cells from most healthy donors (Fig. 6H), suggesting T-bet promotes its own expression. Neither IFN-γ nor IL-27 induced CXCR3 expression on naïve B cells from patients with AD IFN-γR1 deficiency or AR complete IL-27R-deficiency, respectively (Fig. 6H-K). The ability of these cytokines to induce CXCR3 on CpG/αIg-stimulated naïve B cells was also reduced by AD STAT1 deficiency and completely abolished by AR STAT1 deficiency but was unaffected by DN mutations in STAT3 (Fig. 6H -K). Similar results were obtained for upregulation of CD19 and FCRL5 expression mediated by IFN-γ, inasmuch that this was abolished by bi-allelic mutations in TBX21, STAT1, and impaired by DN mutations in IFNGR1 or STAT1 (Fig. 6H -K). The role for STAT1 in this process was also revealed by the ability of JAK inhibitors to prevent the generation of T-bet hi CXCR3 + B cells from naïve B cells from healthy donors in vitro (67). These results establish that signals mediated by TLRs and the BCR in the presence of cytokine inputs initiate the differentiation of naïve B cells into T-bet + CD19 hi B cells independently of T-bet. However, T-bet is strictly required for the generation of T-bet + CD19 hi CXCR3 + FCRL5 hi B cells in vitro.

Chromatin accessibility of B cells is altered in inherited T-bet deficiency
To explore mechanisms by which T-bet controls the lineage determination of T-betexpressing B cells in humans, naive B cells from healthy donors and P were stimulated with CpG/αIg in the presence of IFN-γ or IL-27. Cells were subjected to Omni-ATACseq for the genome-wide investigation of chromatin accessibility (86). In the absence of stimuli, chromatin accessibility differed between B cells from healthy donors and P for only 33 loci (Fig. S6A). Only five of these loci also presented differences in chromatin accessibility between naïve B cells from healthy donors and P in response to stimulation with CpG/αIg/IFN-γ or CpG/αIg/IL-27 (Fig. 7A). Three of these were encompassed by the CCL3L1 locus, and their chromatin was in the closed configuration in T-bet deficiency, with another proximal locus within CCL4L1 following the same trend ( Fig. 7A and Fig.   7B). This suggests that T-bet-deficient B cells are less poised to secrete chemokines required for T-cell recruitment and are therefore less likely to receive sufficient T-cell help (87). Following in vitro stimulation with CpG/αIg/IFN-γ, chromatin accessibility differed significantly between B cells from healthy donors and P for 2391 loci. For 139 loci, chromatin accessibility differed significantly between B cells from healthy donors and those of P following CpG/αIg/IL-27 stimulation, and 50 of these loci overlapped with those displaying differential chromatin accessibility after CpG/αIg/IFN-γ stimulation (Fig. S6A). This finding suggests that IFN-γ and IL-27 stimulate B cells through a common mechanism, probably involving T-bet, but that IFN-γ is the more potent stimulus, consistent with larger proportions of T-bet + CXCR3 + FCRL5 + cells induced by IFN-γ from CpG/αIg-stimulated naïve B cells ( Figure 6E and F). Thus, inherited human T-bet deficiency leads to changes in chromatin accessibilities of targets common to stimuli known to induce T-bet in B cells.

Changes in the epigenetic landscape determined by T-bet program B cell differentiation in vitro
We analyzed epigenetic changes governed by T-bet in B cells by first studying changes in chromatin accessibility in activated B cells from healthy donors. Chromatin accessibility was upregulated at 2017 loci and downregulated at 461 loci in response to CpG/αIg/ IFN-γ in B cells from healthy donors relative to unstimulated B cells. 89% (2208) of these 2478 differentially regulated loci remained unaltered in CpG/αIg/IFN-γ-stimulated T-bet-deficient B cells (Fig. S6B). Chromatin accessibilities of 1184 loci were upregulated, and those of 352 loci were downregulated in B cells from healthy donors in response to CpG/αIg/IL-27. As for CpG/αIg/IFN-γ stimulation, most (89%) remained unaltered by CpG/αIg/IL-27 in T-bet-deficient B cells (Fig. S6B). Thus, the majority of epigenetic changes caused by stimuli that induce T-bet in human B cells were T-bet-dependent. These 2208 (CpG/αIg/IFN-γ) and 1363 (CpG/αIg/IL-27) loci therefore represent the landscape of a T-bet-dependent chromatin signature driving lineage determination in T-betexpressing B cells (Fig. S6C). Notably, 902 loci (66% of T-bet-dependent targets induced by CpG/αIg/IL-27) overlapped with those induced by CpG/αIg/IFN-γ stimulation (Fig.  7C, Fig. S6C, and Data file S2). DNA binding motifs for IRF1, JUNB, and RUNX1 were most significantly enriched in these 902 shared loci (Fig. 7D and Data file S2). Notably, enrichment in DNA binding motifs of IRF1, a crucial transcription factor downstream of IFN-γ-dependent response in humans (88,89), suggests T-bet provides permissive environment for binding of IRF1 to IFN-γ-and IL-27-dependent targets in human B cells. Chromatin at the FAS, IL21R, SEC61B, DUSP4, DAPP1, and SOX5 loci, which are all were strongly expressed by CD21 lo CD11c + B cells, was in an open configuration, whereas that at the CD79B and CXCR4 loci, which are weakly expressed in CD21 lo CD11c + B cells, was closed by both stimuli in a T-bet-dependent manner (Fig. 7E -G, Fig. S6D). These investigations revealed many new T-bet-dependent epigenetic targets. For example, chromatin was in an open configuration at three loci within IRF4 and three within GFI1 in B cells from healthy donors, but not in those of P; chromatin accessibility at these loci was increased by CpG/αIg/IFN-γ and CpG/αIg/IL-27, whereas it was decreased at the SEMA4B, CCR6, and CD37 loci, by both stimuli, in a T-bet-dependent manner ( Fig. 7G and Fig. S6E). These findings suggest that T-bet poises the cells for differentiation into T-bet-expressing B cells by creating a permissive chromatin environment facilitating the efficient differentiation of human CD21 lo CD11c + B cells (90).

Discussion
We report that while T-bet is largely redundant for in vivo functions of human B cells and humoral immunity, it has a nuanced role in regulating Ig CSR, evidenced by increased serum levels of IgG1, IgG4 and IgE, reduced serum IgG2 levels, and increased proportions of IgG1 + and IgG4 + memory B cells, in a patient with complete T-bet deficiency. The alteration to IgG subclasses and serology is unlikely to reflect infectious history of the patient as he has been in remission and free of mycobacterial infection for several years. These perturbations to Ig levels are likely results from B-cell intrinsic and/or extrinsic mechanisms. On one hand, T-bet may directly regulate IgG subclasses in human B cells. On the other hand, increased serum IgG1, IgG4 and IgE in T-bet deficiency are consistent with skewing of T-bet deficient CD4 + T cells to a T H 2-type effector function, evidenced by increased production of IL-4, IL-5 and IL-13 (56), and the well-established role of these cytokines in inducing human B-cell class switching to IgG1, IgG4 and IgE (59). Thus, dysregulated T H 2 cytokine production by T-bet deficient CD4 + T cells may contribute to altered levels of some serum Ig classes in the patient.
These findings from human T-bet deficiency are similar to those from mice which established that T-bet is required for CSR to IgG2a/c in vitro and in vivo (10,11,15,16).
Interestingly, B-cell intrinsic T-bet-dependent IgG2a/c production appears to be important in mice in vivo for long-lived humoral immunity following immunization (16) or during viral and parasitic infections (17)(18)(19)(20), and in the pathogenesis of autoimmune disease models (30,45,91). Despite altered serum Ig levels in the human T-bet deficient patient, he has not presented any clinical disease due to infections with, for example, S. pneumoniae, to which he has been exposed and can be life-threatening in patients with B-cell immunodeficiency disorders (92,93). This is also consistent with our findings of intact SHM, affinity maturation, memory B-cell formation in the patient, as well as intact differentiation of his naïve and memory B cells into Ig-secreting cells in response to polyclonal stimulation in vitro. Thus, these clinical and immunological features suggest T-bet constrains CSR to IgG1, IgG2, and IgG4, but is largely redundant for clinically meaningful B cell-mediated humoral immunity against most common infections in humans, at least for the functions tested to date.
Whilst humoral immunity was essentially unaffected by T-bet deficiency, a major discovery from our study was that T-bet is essential for the generation of the CD11c hi CXCR3 + subset of human CD21 lo CD19 hi B cells in vivo and in vitro. We also identified pathways upstream of T-bet fundamental for generating human CD11c hi CXCR3 + CD21 lo CD19 hi B cells. In vitro co-stimulation of human naïve B cells with TLR9, BCR, and either IFN-γ or IL-27 induced high level co-expression of T-bet, CXCR3, FCRL5 and CD19. Despite this, neither IFN-γ nor IL-27 were uniquely required to generate human T-bet + B cells in vivo, as frequencies of these cells were intact in patients with IFN-γR or IL-27Rdeficiencies. However, CD21 lo CD11c hi CD19 hi CD20 hi CXCX3 + B cells were reduced in peripheral blood of patients with complete AR STAT1 deficiency, which impairs IFN-γ and IL-27 signaling, or IFN-γR1 deficiency, which abolishes IFN-γ signaling. These findings indicate that T-bet and STAT1, downstream of IFN-γ or IL-27, co-operate to induce the transcriptomic and epigenetic imprinting necessary to generate CD21 lo CD11c + B cells in vivo. Furthermore, these cytokines compensate for each another in individuals with defective signaling due to loss-of-function mutations in IFNGR1 or IL27R. As CD21 lo T-bet + B cells are overrepresented in several human immune dysregulatory diseases, these findings indicate that a selective JAK1/STAT1 inhibitor, or directly targeting T-bet, may yield beneficial clinical outcomes by preventing or controlling expansion of pathogenic CD21 lo T-bet + B cells. Indeed, JAK inhibitors can suppress the in vitro generation of T-bet + B cells from naïve B cells from healthy donors (67).
The physiological or pathogenic roles of CD21 lo CD11c + B cells remain enigmatic. First, although frequencies of CD21 lo CD11c + B cells increases in peripheral blood following vaccination, chronic infections and in autoimmune disorders (9,31,47,(50)(51)(52)(39)(40)(41)(42)(43)(44)(45)(46), the predominance of these cells in these conditions is largely correlative. Furthermore, it remains unclear how they contribute to immunopathology. Similarly, it is unclear whether the expansion of these cells is a cause or consequence of the immune stimulatory environment of infection or autoimmunity. Second, despite lacking CD21 lo CD11c + B cells, the T-bet deficient patient has largely normal humoral immunity in vivo and B-cell function in vitro, despite skewing of IgG subclasses. It is thus possible that CD21 lo CD11c + B cells play roles in processes other than humoral immunity. Indeed, spectral flow cytometry and CITE-seq revealed an enrichment in expression of genes encoding proteins involved in antigen presentation, cognate T-B cell interactions, and peripheral tolerance in CD21 lo CD11c + B cells. As the T-bet-deficient patient is young, long-term follow up may reveal whether he is protected from or prone to certain conditions. The identification of additional T-bet-deficient patients is required to draw firm conclusions. Overall, findings from our study establish a framework to investigate CD21 lo CD11c + B cells in human health and disease, particularly other patients with known or newly discovered genetic defects. These future studies will shed more light on the molecular requirements for the development and function of this intriguing B-cell subset.

Study design
We investigated the B cell and antibody phenotypes in a patient with autosomal recessive T-bet deficiency. We also enrolled his relatives and healthy controls in the study as controls. We performed ex vivo and in vitro experiments using peripheral blood mononuclear cells derived from the patient and controls. We also obtained DNA, plasma, and other biospecimens from the patient and controls to analyze their in vivo phenotypes. Both biological and technical replicates were used to validate the findings. Experiments were performed at least twice with appropriate replications. Conclusions were drawn from analyzing the results from aforementioned approaches collectively.

Human subjects
The T-bet-deficient patient and the relatives studied here were living in and followed up in Morocco. The case report has already been published (55,56). The study was approved by and performed in accordance with the requirements of the institutional ethics committees of Necker Hospital for Sick Children, Paris, France, and the Rockefeller University, New York, USA. Informed consent was obtained from the patient, his relatives, and the healthy control volunteers enrolled in the study. This study was also approved by the Sydney Local Health District RPAH Zone Human Research Ethics Committee and Research Governance Office, Royal Prince Alfred Hospital, Camperdown, New South Wales, Australia (protocol X16-0210/LNR/16/RPAH/257). Experiments using samples from human subjects were conducted in the United States, France, Australia and Sweden, in accordance with local regulations and with the approval of the IRBs of corresponding institutions.

Bulk sequencing and analysis of immunoglobulin transcripts from PBMCs
Immunoglobulin heavy chain (IGH), kappa chain (IGK) and lambda chain (IGL) repertoires were sequenced from PBMCs from the T-bet-deficient patient (4 years of age at the time of sampling) and five healthy donors aged 11 months, 24 months, 16 years, 45 years and 66 years. Independent amplifications for each isotype (IgM, IgG, IgA and IgE) were performed for the heavy chain, and IGK and IGL were amplified in separate reactions, as previously described, with the addition of IgA and IgE reverse primers (94); IgA: GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGCAGGTCACACTGAGTGGCTCC, IgE: GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGCCAGGCAGCCCAGAGTCACGG . Samples were indexed and pooled for sequencing on an Illumina NextSeq with 2x300 PE.
Sample datasets were demultiplexed during FASTQ generation on the basis of sample indices. Paired-end reads were merged with FLASH (95) and the merged sequences were quality-filtered with FilterSeq from the presto (v0.5.13 2019.08.29) package (96), with a minimum quality of 20. Forward and reverse primers were trimmed, and constant regions were tagged with MaskPrimers (presto package), with a requirement for exact matches and the discarding of reads not meeting this requirement. Datasets were deduplicated (only unique sequences retained) with CollapseSeq (presto package) and the deduplicated datasets were input into stand-alone IgBLAST (v1.14) (97) for alignment against the IMGT human germline reference directories (downloaded 16 Jan 2020). IgBLAST results were filtered to remove truncated transcripts and transcripts lacking an identifiable CDR3. B-cell clones were inferred for each subject for IGH (combining transcripts from all isotypes), IGK and IGL. Clones were generated by first subsetting the VDJs from each donor on the basis of V gene, J gene and CDR3 length and then clustering CDR3 nucleotide sequences, with a 90% threshold, with cd-hit (98). Each cluster was inferred to be a clone of related VDJs stemming from a lineage that shared the same progenitor B cell. Median somatic hypermutation (SHM) for each clone was calculated per isotype for the V-REGION (percentage of V-REGION nucleotides mutated, based on IgBLAST alignment), and clone size, as both total and unique read numbers were also calculated.

B-cell differentiation
Naïve (CD20 + CD10 − CD27 − IgG − ) and memory (CD20 + CD10 − CD27 + IgG + ) B cells were purified by sorting from the PBMCs of healthy donors or P with a FACSAria III. Purity was typically >97%. We assessed the in vitro induction of T-bet + B cells by culturing naïve B cells in media alone (RPMI1640/10% FCS), or with presence of F(ab') 2 goat anti-human Ig (0.8 μg/mL) and CpG (0.35 μg/mL) with or without IFN-γ (333 U/mL) or IL-27 (50 ng/mL). After 3.5 days, the B cells were harvested, and stained for the surface expression of CD19, FCRL5 and CXCR3, fixed and permeabilized and then stained for intracellular expression of T-bet. Proportions of T-bet + B cells, and expression of CD19, FCRL5 and CXCR3 on T-bet + and T-bet − B cells present in the cultures, were then determined. B-cell viability was determined with the Zombie Aqua Viability dye (BioLegend). We investigated in vitro differentiation into Ig-secreting cells, by culturing naïve and memory B cells with CD40L (200 ng/mL) cross-linked with anti-HA mAb (50 ng/mL, R&D Systems) alone or together with IL-21 (50 ng/mL, PeproTech), IL-10 (100 U/mL; provided by R. de Waal Malefyt -DNAX Research Institute, Palo Alto, CA), IL-21 plus IL-4 (100 U/mL; provided by R. de Waal Malefyt), or CpG 2006 (1 μg/mL, Sigma-Aldrich). Culture supernatants were harvested after 7 days and the amount of IgM, IgG and IgA secreted into the supernatant was determined in Ig heavy chain-specific ELISAs (61,99). Secretion of IgG1, IgG2, IgG3, and IgG4 was determined using an IgG subclass ELISA kit (Invitrogen, catalogue # 99-1000) as per manufacturers' instructions.

Immunophenotyping of age-associated B cells with spectral flow cytometry
Experiments were performed in two batches. In the first batch, PBMCs were obtained from 20 healthy adult donors, four age-matched controls (2, 6, 7, 8 years of age), P (4 years of age at the time of sampling), P's healthy brother (8 years of age at the time of sampling), who is wild-type for the TBX21 locus, and P's healthy mother, who is heterozygous for the mutation (55,(100)(101)(102). In the second batch, PBMCs were obtained from 10 healthy adult donors and a patient with complete STAT1 deficiency (68). We stained 1 x 10 6 to 2 x 10 6 PBMCs from each individual with Zombie-NIR live-dead exclusion dye (BioLegend). Cells were then labeled with FcBlock It should be noted that IgG-PE/Cy5 signal was found to be masked by a humanized FcBlock antibody which was mistakenly used in the study. Due to this technical reason, IgG signal was almost undetectable from this immunophenotyping. However, IgG-PE/Cy5 did not affect the balance of other 28 markers in this spectral flow experiment.

Cellular indexing of transcriptomes and epitopes by sequencing (CITE-seq)
PBMCs from four age-matched controls (sample codes: #906, #4156, #1093, and #14881 with ages ranging from 2 to 8 years), three IFN-γR1-deficient patients (sample codes: Single-cell 5' expression and BCR libraries were generated with the Chromium Single Cell 5' Library & Gel Bead kit Version 2 (10x Genomics cat # PN1000265). BCR cDNA was amplified from the total cDNA pool with the 10X Chromium Single-Cell Human BCR Amplification Kit (PN-1000253) before library construction. Hashtag libraries were generated with the Chromium Single Cell 5' Feature Barcode Kit (PN 1000256). Standard protocols from 10X Genomics were followed for library generation. The quality of the libraries was assessed on an Agilent TapeStation and the three libraries were pooled in the following ratio: expression 10:BCR 1:HTO 1. The pooled libraries were sequenced on an Illumina NovaSeq 6000 sequencer with a 100-cycle SP flow cell. In total, 800 million paired reads were generated (read 1 = 26 bp, read 2 = 90 bp).

Additional materials and methods located in Supplemental Methods
Some detailed materials and methods are provided in a Supplemental Methods section. This material includes the methods used for analysis of B cells with conventional flow cytometry, immunophenotyping of surface B cell receptors with spectral flow cytometry and unsupervised analysis of data from spectral flow cytometry. The Supplemental Methods section also describes the methods used for real-time quantitative ENC1 PCR, analysis of CITE-seq data, single-cell VDJ sequencing analysis, naïve B-cell differentiation for Omni-ATAC-seq, and analysis of Omni-ATAC-seq.

Statistical analysis
Student's t-test, Mann-Whitney test, one-way ANOVA, and two-way ANOVA were used in their corresponding datasets to investigate statistical difference. Bar graphs throughout the figures represent either the mean and the standard deviation or the mean and the standard error of the mean. Dots present individual samples or technical replicates. P values of 0.05 and below are considered to be statistically significant. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001, and ns = not significant (or not marked). Details of the statistical methods used in individual experiments are provided in the corresponding figure captions. All raw data are provided in Data file S4.

Supplementary Material
Refer to Web version on PubMed Central for supplementary material.  light chains for each given cell was divided by the total number of nucleotides counted, to calculate the mutation frequency. (G) Mutation frequency of each CD21 lo B cell from the indicated individuals. The frequencies of cells with mutation rates greater than 1% are highlighted. (H) Mutation frequency of each CD11c + CD21 lo B cells (+) or CD11c − CD21 lo (−) B cells from the indicated individuals. The frequencies of cells with mutation rates greater than 1% are highlighted. In Fig. 5A -E, bars represent values of each individual sample. In Fig. 5G and H, dots represent values for individual cells. showing a selection of 51 loci from the list, as in (C), the chromatin accessibilities of which were differentially regulated in control cells, but not in T-bet-deficient cells, in response to both aIg+CpG+IFN-γ and aIg+CpG+IL-27.