BAFF Produced by Neutrophils and Dendritic Cells Is Regulated Differently and Has Distinct Roles in Antibody Responses and Protective Immunity against West Nile Virus

This article is featured in In This Issue, p.1419

B cell activating factor (BAFF) (also known as BLyS or Tnfsf13b) is a member of the TNF family of ligands that binds to three different TNFR family members: BAFFR, TACI, and BCMA (1–3). Baff^−/− and Baffr^−/− mice develop few or no mature B cells, displaying a block in B cell development from the transitional 1 (T1) to transitional 2 (T2) stage onward (4–6). BAFF signaling, like signaling through the BCR, functions as a key survival factor for maintaining mature B cell homeostasis (5, 6). BAFF is overexpressed in patients with autoimmune diseases such as Sjögren syndrome, systemic lupus erythematosus, type I diabetes, B cell lymphoma, and chronic lymphocytic leukemia (3, 7–9). In addition to BAFFs contributing to the development of autoimmune diseases (10), we and others have shown that BAFF plays an important role in infectious diseases, including HIV and West Nile virus (WNV) (11–15). BAFF may also contribute to the pathology of HIV (16). Moreover, BAFF has been associated with inflammatory conditions, such as inflammatory bowel disease (17) and allergic airway inflammatory diseases (18).

Although BAFF dysregulation has been implicated in a number of human diseases, little is known about the regulation of BAFF expression and function in vivo. Most studies have focused on the effects of blocking BAFF receptors and the development of BAFF-blocking drugs (17, 19, 20). However, which BAFF-producing cells contribute to specific B cell responses and B cell–associated diseases is not entirely clear. Studies primarily done in vitro have established that BAFF can be expressed by different immune cell types, including monocytes (MOs), macrophages, dendritic cells (DCs), neutrophils (Nphs), and follicular DCs as well as epithelial cells and stromal cells (1, 11, 21–23). In particular, BAFF produced by DCs, Nphs, epithelial cells, follicular DCs, and stromal cells helps to regulate humoral immune responses (24). Several cytokines, including IFN-γ, IL-10, G-CSF, and GM-CSF, upregulate BAFF expression in different cell types (1, 21, 25–28).

Type I IFN is a key regulator of BAFF. IFN-α induces DCs to produce BAFF, which can play a role in class-switch recombination and drive B cells to become Ab-producing cells (29–33). In particular, type I IFN–dependent BAFF produced by DCs is likely to play a key role in T-independent type 2 (TI-2) Ab responses (29, 32–35). Both ssRNA and dsRNA viruses induce type I IFN via RNA-sensing pathways, including TLR3, TLR7, and the RIG-I/MAVS pathway (36). Polyinosinic:polycytidylic acid (Poly[I:C]) promotes BAFF production in tonsillar mononuclear cells, which in turn triggers IgA class-switch recombination and aggravates IgA nephropathy (37). However, which specific myeloid cell subset is responsible for this effect is not known. Infection with either ssRNA or dsRNA viruses also induces BAFF production in vitro in human salivary gland epithelial cells, DCs, and MOs (11).

To define the BAFF-producing cells implicated in specific immune responses, we developed two novel transgenic mouse lines, BAFF reporter (BAFF-RFP) mice and Baff floxed (Baff^fl/fl) mice. Using BAFF-RFP mice to monitor BAFF levels, we detected changes in BAFF expression in specific myeloid populations after in vivo activation of RNA-sensing pathways. Changes in BAFF production by MOs, conventional DCs (cDCs), and Nphs were mostly dependent on type I IFN receptor (IFNAR) but differentially affected by the absence of MAVS. Furthermore, we identified a new BAFF-expressing CD11b^hi NK subset induced by either Poly(I:C) or WNV infection. Selective removal of Baff from either cDCs or Nphs revealed that BAFF produced from both DCs and Nphs is required for optimal TI-2 Ag-specific Ab responses. Mice lacking BAFF expression on cDCs were more susceptible to infection with WNV and had reduced WNV-specific IgG and neutralizing Abs as compared with controls. Thus, BAFF produced by cDCs is required to protect against a lethal viral infection. These findings underscore that BAFF produced by Nphs and cDCs is regulated differently and has distinct roles in Ab responses and protective immunity.

C57BL/6J (B6) mice were purchased from The Jackson Laboratory (Bar Harbor, ME). BAFF-RFP and Baff^fl/fl mice (on the B6 background) were developed at the University of California Davis Mouse Biology Program (University of California Davis, Davis, CA) (see Supplemental Fig. 1A). Mavs^−/− (B6 × 129Sv/Ev) mice were kindly provided by Dr. S. Akira (Osaka University, Osaka, Japan) and backcrossed onto the B6 background (38). Ifnar^−/− B6 mice were a kind gift from Dr. Michael Gale (University of Washington, Seattle, WA). Mrp8^Cre and zDC^Cre B6 mice were purchased from The Jackson Laboratory. All mice were age and sex matched for experiments and used at 8–11 wk of age. Mice were housed in a specific pathogen-free environment; all procedures were approved by the University of Washington Institutional Animal Care and Use Committee.

A two-construct approach was used to generate BAFF-RFP mice expressing a Tag–red fluorescent protein–T (RFP) (39) under the Tnfsf13b/Baff promoter (BAFF-RFP) and Baff^fl/fl mice to conditionally knockout the Tnfsf13b/Baff gene. First, the endogenous reporter BAFF-RFP mice were generated by replacing a Tnfsf13b allele with a targeting construct expressing internal ribosome entry site (IRES)–RFP between exon 2 and exon 3 of Baff gene and containing loxP sites between exon 3 and exon 7. In the BAFF-RFP allele, Baff is functionally knocked out where the endogenous reporter expresses BAFF-IRES-RFP (BAFF-RFP), the RFP under the control of the BAFF promoter but with an IRES. In a second step, Baff^fl/fl mice were generated by excising the RFP sequence and leaving wild-type (WT) Baff gene with floxed exons 3–7 (for details about the construct and the generation of the transgenic mice, see Supplemental Fig. 1A, 1B). The Baff^fl/fl mice express WT Baff until Cre-mediated deletion of exons 3–7 render the Baff gene inactive. Baff^fl/fl mice were crossed with Mrp8^Cre mice (40, 41) and zDC^Cre mice (42, 43), to generate a conditional knockout (cKO) where Baff is selectively deleted in Nphs (Baff^fl/fl Mrp8^Cre) or cDCs (Baff^fl/fl zDC^Cre).

For in vivo experiments, we used heterozygous BAFF-IRES-RFP^+/− (BAFF-RFP^+/−) mice, which still have mature B cells and, therefore, are suitable for functional studies. Unless otherwise specified, “BAFF-RFP” indicates heterozygous BAFF-RFP^+/− mice. BAFF-RFP^+/− mice were crossed with Ifnar^−/− mice or Mavs^−/− mice to generate BAFF-RFP Ifnar^−/− mice and BAFF-RFP Mavs^−/− mice, respectively.

Mice were injected i.p. with 200 μg/mouse Poly(I:C), 200 μg/mouse R848, or 20 nmol/mouse CpG; 6 h after injection of the TLR ligand, brefeldin A (eBioscience, San Diego, CA) was administered i.v. (44, 45). Brefeldin A was injected in vivo to optimize the retention of the BAFF-RFP signal inside the cells after TLR agonist administration; 24 h after TLR agonist injection, spleens and bone marrows (BMs) were harvested for cell subset analysis. Spleens from naive or TLR-injected mice were harvested and dissociated into single-cell suspensions as previously described, with some modification (33). Briefly, spleens were removed and dissociated by enzymatic digestion at 37°C with Liberase TL and Dnase I (Roche, Indianapolis, IN), followed by mincing the tissue between the ends of two frosted microscope glass slides to obtain a single-cell suspension. BM cells were isolated by cutting one end of the femur and flushing BM cells out of the bone by centrifugation. For PBMC harvesting, blood was collected by retro-orbital eye bleeds using heparinized capillary tubes. After erythrocytes were lysed, PBMC, splenocytes, and BM cell suspensions were processed for staining for flow cytometry. In initial experiments, BAFF-RFP mice were injected with brefeldin A only for 18 h as controls, and no significant differences in the BAFF-RFP signal/subset by flow cytometry were observed between naive BAFF-RFP mice and BAFF-RFP mice only injected with brefeldin A.

The pathogenic lineage 1 WNV-TX infectious clone, derived from the Texas 2002-HC strain, was generously provided by Dr. Michael Gale (University of Washington, Seattle, WA) and prepared as described (13). For infections for inguinal lymph node (iLN) analysis, mice were inoculated under anesthesia with 1000 PFU of WNV-TX s.c. into both footpads (f.p.) in a total volume of 40 μl; 20 μl (500 PFU) were injected in each f.p. (13). Twenty-four or forty-eight hours after WNV infection, iLNs were removed, minced into small fragments, and digested for 40 min at 37°C with Liberase TL and Dnase I (Roche), as described (46). Single-cell suspensions were processed for staining for flow cytometry. For survival studies, infected mice were inoculated under anesthesia with 100 PFU of WNV-TX s.c. into one f.p. in a total volume of 20 μl. Mice were monitored at least once daily for survival. Serum was isolated from blood, collected via the retro-orbital route at day 4 and day 7, and stored at −80°C until use.

Viral RNA was extracted from sera by using QiAMP viral RNA extraction kit (Qiagen, Valencia, CA). Sequences for primers and TaqMan probes used were as described (13). WNV envelope protein (WNVE)–specific IgG was quantitated by ELISA using a recombinant WNVE (MyBioSource, San Diego, CA) as described (13). Titers of WNV neutralizing Abs were quantitated by a focus forming reduction neutralization test using Vero cells, as described (47). The 50% focus reduction neutralization test (FRNT₅₀) titers were calculated as the lowest dilution of serum with <50% of the average number of spots in the virus-only wells.

Mice were i.p. administered 20 μg/mouse 4-hydroxy-3-nitrophenylacetic (NP) hapten conjugated to AminoEthylCarboxyMethyl-FICOLL, NP-Ficoll (Biosearch Technologies, Petaluma, CA). At the indicated time points, mice were bled and serum was isolated. Serum NP Ab titers were measured as described (33). Plates were developed with tetramethylbenzidine substrate and read at 450 nm absorbance. Values were compared with known dilutions of IgM or IgG to calculate Ab concentrations.

Splenocytes, PBMCs, BM cells, and iLNs were processed into single-cell suspensions as above and stained for flow cytometry analysis, as described (13). Cells were incubated with an Aqua Live-Dead fixable viability dye (Molecular Probes, Life Technologies, Waltham, MA) in the absence of FBS, to discriminate dead cells. Cells were then blocked using an anti-Fc receptor Ab (anti-CD16/CD32) (2.4G2) (BioLegend, San Diego, CA) and stained for surface markers and then fixed in 1–2% paraformaldehyde. Cells were stained with mAbs conjugated to FITC, allophycocyanin, eFluor450, allophycocyanin-eFluor780, PerCPCy5.5, PE-Cy7, AlexaFluor647, BUV395, BV605, BV421, BV711, BV650, and BUV395. For analysis of splenic or BM cell subsets, 11- to 12-color flow cytometry was performed using combinations of mAbs against the following: CD19 (1D3), CD11b (M1/70), and CD11c (N418) from eBioscience; B220 (RA3-6B2), CD93 (AA4.1), and Ly6C (AL-21) from BD Horizon/Biosciences (San Jose, CA); and CD19 (1D3), B220 (RA3-6B2), CD3 (17A2), NK1.1 (PK136), CD8α (53-6.7), Ly6G (1A8), SiglecH (440c), Ly49c (14B11), CD49b (DX5), NKp46/CD335 (29A1.4), CD127 (SB/199), CD21/35 (7E9), CD23 (B3B4), and CD24 (M1/69) from BioLegend. The BAFF-RFP signal was detected in the PE channel. Cells were processed with an LSRII flow cytometer (Becton Dickinson, Franklin Lakes, NJ) using FACSDiva software, and data were analyzed using FlowJo (v.10; Tree Star). See Supplemental Fig. 2A, 2B for gating strategies for spleens (and iLNs) and BM, respectively. Splenic myeloid cell subsets were defined as described previously (13), as in Supplemental Fig. 2A. For BM, we identified five populations of myeloid cells and precursors in the CD11b⁺B220⁻CD19⁻Ly6G⁻SSC⁻ gate and defined them based on the literature (48–50) (see Supplemental Fig. 2B, Supplemental Table I). A similar gating strategy as for spleen cells was used for PBMC. PBMC subsets were defined as follows: Nphs, CD11b^hiLy6G^hiLy6C^intSSC^int−NK1.1⁻; Ly6C^hi MOs, CD11b^hiLy6C^hiCD11c⁻SSC⁻Ly6G⁻NK1.1⁻SiglecH⁻; DCs, CD11b⁺Ly6C⁻CD11c⁺SSC⁻Ly6G⁻NK1.1⁻SiglecH⁻; NK cells, NK1.1^hiCD11b^int; T cells, CD3⁺CD19⁻; and B cells, CD3⁻CD19⁺.

Splenic B cell subsets were defined as described previously (13, 51). After gating out debris, doublets, and dead cells, B cell subsets in the CD19⁺ B220⁺ gate were defined as follows: follicular B cells, CD24^midCD21/35^midCD93⁻CD23⁻; marginal zone (MZ) B cells, CD24^hiCD21/35^hiCD93⁻CD23⁻; MZ B cell precursors, CD24^hiCD21/35^hiCD93^loCD23⁺; T2 B cells, CD24^hiCD21/35^int/hiCD93⁺CD23⁺; and T1 B cells, CD24^hiCD21/35^loCD93⁺CD23⁻. For BM B cell precursors and long-lived plasma cells (PCs) (52), we gated out NK cells, p-DCs, and CD11b⁺ cells. In the NK1.1⁻SiglecH⁻CD11b⁻ gate, we defined PreProB cells as B220⁻CD43⁺CD19⁻, ProB cells as B220^loCD43⁺CD19⁺, PreB cells as B220⁺CD43⁻CD19⁺IgM⁻IgD⁻, newly formed B cells as B220⁺CD43⁻CD19⁺IgM⁺IgD⁻, mature B cells as B220⁺CD43⁻CD19⁺IgM⁺IgD⁺, and PreBNFMatB (B220⁺CD43⁻). Splenic PCs and long-lived PCs in the BM were defined as B220^loCD138⁺. Splenic T cell subsets are defined as follows: CD8 T cells as CD3⁺ CD8⁺CD19⁻ and CD4 T cells as CD3⁺ CD8⁻CD19⁻.

BM cells from WT and BAFF-RFP (^+/−) mice were harvested as described above, and cell subsets were sorted with a FACS Aria cell sorter (Becton Dickinson). Nphs were sorted from WT and BAFF-RFP mice as CD11b⁺Ly6G^hiLy6C^intSSC^int−. B cells were sorted from WT mice as Ly6G⁻B220⁺ cells and from BAFF-RFP mice as Ly6G⁻B220⁺RFP⁺ B cells and Ly6G⁻B220⁺RFP⁻ B cells.

Splenocytes were harvested from Baff^fl/fl, Baff^fl/fl MRP8^Cre, or Baff^fl/fl zDC^Cre mice as described above and enriched with CD11b⁺ CD11c⁺ magnetic bead positive selection (Milteny, CA). Cell populations were sorted using a FACS Aria cell sorter (Becton Dickinson). For gating strategies, see Supplemental Fig. 2A. Sorted cells were lysed and RNA was extracted for quantitative RT-PCR analysis as described (33). The primer sequences used were as follows: mBaff-forward 5′-AGGCTGGAAGAAGGAGATGAG-3′ and mBaff-reverse 5′-CAGAGAAGACGAGGGAAGGG-3′. See Supplemental Fig. 1C, 1D for verification of BAFF Nph cKO (Baff^fl/fl MRP8^Cre) mice or BAFF cDC cKO (Baff ^fl/fl zDC^Cre) mice. As additional controls, sorted cells from WT mice, MRP8^Cre mice, and zDC^Cre mice all showed similar levels of Baff mRNA as Baff^fl/fl mice (data not shown).

Survival data were analyzed by Mantel–Cox log-rank test. Timeline data were analyzed using a two-way ANOVA with Tukey multiple comparison test. Analysis between more than two groups was performed using one-way ANOVA with Holm–Sidak method for multiple comparisons. Analyses between two groups were performed using unpaired Student t test. GraphPad Prism 7 was used for statistical analyses. Differences of p < 0.05 were considered significant.

We developed mouse models to define how BAFF expression by cell subsets is regulated and what BAFF-producing cells are required for B cell responses. We used a two-construct approach to generate BAFF IRES-TagRFP-T (BAFF-RFP) reporter mice concurrently with Baff^fl/fl mice (see Supplemental Fig. 1A for details). In BM from naive BAFF-RFP (^+/−) mice, most Nphs were BAFF-RFP⁺, and a small population of BAFF-RFP⁺ B220⁺ B cells was detectable (Fig. 1A). To determine whether RFP expression in BAFF-RFP mice correlated with Baff mRNA expression, we sorted BM Nphs, BAFF-RFP⁺ B220⁺ cells, and BAFF-RFP⁻ B220⁺ B cells and quantified Baff mRNA levels (Fig. 1B, for sorting strategy see 2Materials and Methods). As expected, Baff mRNA levels in Nphs from heterozygous BAFF-RFP mice were about half the levels detected in WT mice. Baff mRNA expression was lower in B cells than in Nphs in both WT and BAFF-RFP^+/− mice. Importantly, Baff mRNA levels were higher in BAFF-RFP⁺ B cells compared with BAFF-RFP⁻ B cells or unsorted WT B cells (Fig. 1B). Therefore, BAFF-RFP detected by flow cytometry correlates with Baff mRNA expression.

Analysis of cell populations from spleens or PBMCs (see 2Materials and Methods and Supplemental Fig. 2A for gating strategy) from naive BAFF-RFP^+/− mice revealed that BAFF is constitutively expressed at high levels in Nphs, expressed at lower levels in DCs and MOs, and had very low expression or was undetectable in other leukocyte subsets (Fig. 1C, 1D, Supplemental Fig. 3A, 3B). These data are consistent with previous reports (1, 11, 21). Although BAFF-RFP expression overall was very low in B cells, BAFF was detectable particularly in MZ B cells, MZ B cell precursors, T2 B cells, and in small numbers of follicular and T1 B cells (Supplemental Fig. 3A), in agreement with Baff mRNA expression in B cell subsets (53).

Interestingly, in the BM of naive BAFF-RFP^+/− mice, in addition to Nphs, other myeloid cell subsets and myeloid precursors constitutively expressed significant levels of BAFF-RFP (Fig. 2). Using mice with either one Rfp allele (RFP^+/−) or two Rfp alleles (RFP^+/+), we examined the CD11b⁺ BM population in detail and designated subsets based on their cell surface phenotypes (see 2Materials and Methods and Supplemental Fig. 2B, Supplemental Table I). In particular, a high percentage of Nph precursors, CD11b⁺ myeloid precursors, and pre-DCs expressed BAFF-RFP compared with MOs (Fig. 2A). BAFF-RFP expression levels were also greater in Nphs and other myeloid precursors compared with MOs, as shown by their higher levels of expression (Fig. 2B). The BAFF-RFP signals in both BAFF-RFP^+/− and BAFF-RFP^+/+ were significantly different from RFP background detection in WT mice (Fig. 2A–D). The higher levels of BAFF-RFP in BAFF-RFP^+/+ mice versus BAFF-RFP^+/− mice in all BM populations confirmed that BAFF-RFP detection by flow cytometry correlates with the amount of Baff gene expression (Fig. 2A–D).

We also analyzed BM B cell subsets and precursors. BAFF-RFP was expressed at low levels in B220^lo CD138⁺ long-lived PCs and in a subset of CD11b⁺ B220⁺ CD19⁺ B cells but not in ProB cells or more mature B cell precursors (Fig. 2C, 2D). As expected, BAFF-RFP^+/− mice had fewer mature B cell precursors in the BM than WT mice, whereas BAFF-RFP^+/+ mice had almost no mature B cells and normal numbers of earlier B cell precursors (Supplemental Fig. 2C, 2D). Although the BAFF-RFP^+/− mice had lower expression of BAFF, they were still capable of producing mature B cells and thus could be used to measure B cell functions and Ab responses in vivo. For subsequent studies, we used BAFF-RFP^+/− mice and refer to them simply as BAFF-RFP mice.

To address how the activation of RNA-sensing pathways regulate BAFF expression in specific cell subsets in vivo, we injected BAFF-RFP mice with TLR3 agonist Poly(I:C) or the TLR7/8 agonist R848 and analyzed BAFF-RFP expression in splenic cell populations 24 h later (Fig. 3A–D). Poly(I:C) administration led to significant increases in BAFF-RFP mean fluorescence intensity (MFI) (Fig. 3A, 3B) and the percentage of BAFF-RFP⁺ cells (data not shown) in inflammatory Ly6C^hi MOs but not in cDCs, Nphs, or B cells. In contrast, R848 administration had no effect on BAFF-RFP expression in splenic Ly6C^hi MOs but slightly upregulated BAFF-RFP in Nphs (Fig. 3C, 3D). Furthermore, the TLR9 agonist CpG did not affect BAFF-RFP expression in either Ly6C^hi MOs or Nphs (Fig. 3D). Thus, changes in BAFF-RFP expression in myeloid subsets varied depending on the TLR ligand used.

Next we challenged BAFF-RFP mice with WNV-TX (WNV), a pathogenic ssRNA virus, and examined BAFF-RFP expression in myeloid cells within the draining iLNs 24 and 48 h postinfection. Similar to the effect of the TLR3 agonist in the spleen, WNV infection slightly increased BAFF-RFP levels in Nphs from iLNs (Fig. 3E), and in addition, significantly upregulated BAFF-RFP expression in CD8⁻ cDCs 24 h postinfection (Fig. 3E). BAFF-RFP levels did not change in iLN CD8⁺ cDC, MO subsets, B cells, or T cells after WNV infection (Fig. 3E, data not shown).

Interestingly, after either Poly(I:C) administration or WNV infection, we detected increased numbers of an activated NK1.1 subset, which expressed high levels of both BAFF-RFP and CD11b (CD11b^hi NK). In contrast, CD11b^int− NK cells were BAFF-RFP⁻ (Fig. 3F–H left panel and I). This BAFF-RFP⁺ CD11b^hi NK subset was not upregulated after in vivo administration of R848 or CpG (Fig. 3H, right panel). A phenotypic analysis revealed that the CD11b^hi NK cells, like other NK cell subsets, express CD49b, NKp46, and Ly-49C and in addition express Ly6G, a marker found on Nphs (Supplemental Fig. 3C). Furthermore, CD11b^hi NK cells did not express CD127, suggesting they were not innate lymphoid cell 1 (ILC1) (54). In summary, by using BAFF-RFP mice, we detected cell- and tissue-specific upregulation of BAFF after in vivo challenge with different RNA-sensing TLR agonists or WNV infection.

Both dsRNA and ssRNA viruses trigger three RNA-sensing pathways: TLR3 (dsRNA), TLR7/TLR8 (ssRNA), and RIG-I/MDA5/MAVS (ssRNA, dsRNA), which induce type I IFN production in plasmacytoid DCs or other cells (36, 55). Type I IFN is a major inducer of BAFF. However, direct regulation of BAFF by TLR engagement also has been described (22). Thus, to test whether type 1 IFN was required for BAFF upregulation, we crossed BAFF-RFP mice with Ifnar^−/− mice and examined BAFF-RFP expression after either Poly(I:C) administration or WNV infection (Fig. 4). Unlike in BAFF-RFP WT mice, BAFF-RFP Ifnar^−/− mice did not upregulate BAFF-RFP in Ly6C^hi MOs and Ly6C^hi DCs in response to Poly(I:C), whereas there was no change in the constitutive BAFF-RFP expression in Nphs (Fig. 4A, 4B, data not shown). Furthermore, BAFF-RFP Ifnar^−/− mice did not upregulate BAFF in either iLN Nphs or CD8⁻ cDC postinfection with WNV (Fig. 4C, not shown). Thus, the increases in BAFF expression in splenic inflammatory MO subsets in response to Poly(I:C) immunization and in iLN Nphs and CD8⁻ cDCs after WNV infection all require type I IFN signaling.

Because both dsRNA and ssRNA can induce type I IFN via the MAVS pathway, we next tested whether Poly(I:C)- or WNV-induced changes in BAFF-RFP expression required MAVS. Using BAFF-RFP Mavs^−/− mice, we found that myeloid subsets differed in their requirement for MAVS to regulate BAFF expression. Poly(I:C)-induced increases in BAFF-RFP MFI in Ly6C^hi DCs required MAVS, whereas the BAFF-RFP MFI increase in Ly6C^hi MOs and Ly6C^lo MOs was only partially inhibited in BAFF-RFP Mavs^−/− mice (Fig. 5A, 5B). As expected, there was no change in the BAFF-RFP expression in Nphs. These data suggest that for MO subsets, type I IFN production that upregulates BAFF occurs through either the TLR3 pathway or the MAVS pathway. In contrast, the BAFF increase in inflammatory Ly6C^hi DCs is dependent solely on type I IFN produced via the MAVS pathway.

We also analyzed BM cells from BAFF-RFP Ifnar^−/− and BAFF-RFP Mavs^−/− mice after Poly(I:C) immunization. Both the IFNAR and MAVS pathways were required to induce increases in the percentage of inflammatory BAFF-RFP⁺ Ly6C^hi MOs and BAFF-RFP⁺ Ly6C^hi CD11c⁺ cells as well as in other myeloid precursors and BAFF⁺ CD11b⁺ B cells (Supplemental Fig. 3D, 3E). Thus, BAFF is differentially regulated in both a cell-specific manner and in a tissue-specific fashion. For example, BAFF regulation in Ly6C^hi MOs is only partially dependent on MAVS in the spleen but requires MAVS in the BM.

BAFF MFI was not significantly upregulated in CD8⁻ cDCs in BAFF-RFP Mavs^−/− mice after WNV infection; however, BAFF upregulation in Nphs postinfection was independent of MAVS (Fig. 5C). Therefore, after WNV infection, the BAFF increase in CD8⁻ cDCs requires the activation of the MAVS pathway, whereas TLR7/8 pathway may be required for increasing BAFF expression in Nphs. In summary, the changes in BAFF expression by myeloid subsets induced after Poly(I:C) immunization or WNV infection require type I IFN signaling but only in some cases are dependent on MAVS.

BAFF-RFP mice allowed us not only to detect BAFF expression per cell but also to analyze changes in populations of BAFF-RFP–producing cells. Several BAFF-RFP⁺ myeloid cell subsets, including Nphs, inflammatory MOs, and cDCs, were expanded in the iLNs 24 h after WNV infection (Fig. 6A). A number of factors, including chemokines and chemokine receptors, could regulate migration and the expansion of myeloid cells upon infection. Therefore, we tested whether MAVS/IFNAR signaling pathways were required for the changes in cell numbers. Interestingly, MAVS contributed to BAFF-RFP⁺ inflammatory Ly6C^hi MO (Fig. 6A) and Ly6C^hi DC (data not shown) expansion but not to the increased numbers of BAFF⁺ CD8⁻ cDCs and BAFF⁺ Nphs (Fig. 6A). Thus, although BAFF-RFP expression per cell in CD8⁻ cDC required MAVS, the expansion of CD8⁻ cDCs was MAVS independent. The requirements for IFNAR in the expansion of myeloid subsets were similar to the MAVS requirements (data not shown). Taken together, these results suggest that the RIG-I/MAVS–type I IFN signaling axis plays a role in the expansion of inflammatory Ly6C^hi MOs and Ly6C^hi DCs, whereas other factors are responsible for the increased numbers of Nphs and cDCs upon WNV infection.

The expansion of the BAFF-RFP⁺ NK CD11b^hi cells in the iLN after WNV infection required the IFNAR signaling pathway but was independent of MAVS (Fig. 6B). In contrast, administration of Poly(I:C) to either BAFF-RFP Ifnar^−/− or BAFF-RFP Mavs^−/− mice failed to expand BAFF⁺ CD11b^hi NK cells (Fig. 6C). Taken together, these data suggest that type I IFN–dependent expansion of activated BAFF⁺ CD11b^hi NK cells upon WNV infection is mostly dependent on TLR7/8 signaling. In contrast, after dsRNA challenge, the type I IFN–dependent expansion of this BAFF⁺ NK subset occurs via engagement of the MAVS pathway rather than TLR3 activation.

BAFF expression increases in cDCs and Nphs, and BAFF⁺ cDCs and BAFF⁺ Nphs expand after WNV infection (Figs. 3E, 6A). To examine the role of these sources of BAFF during infection, we generated Baff cKO (BAFF cKO) mice, in which we specifically deleted Baff in cDCs or Nphs; Baff^fl/fl mice (see Supplemental Fig. 1A and 2Materials and Methods) crossed with either zDC^Cre mice or Mrp8^Cre mice to generate mice deficient in BAFF expression on cDCs (Baff^fl/fl zDC^Cre); and mice deficient in BAFF expression on Nphs (Baff^fl/fl Mrp8^Cre). Analysis of the expression of Baff mRNA in purified cDCs or Nphs confirmed that BAFF expression was knocked out appropriately in the respective BAFF cKO mice (Supplemental Fig. 1C, 1D). Baff mRNA reduction was restricted to cDCs in Baff^fl/fl zDC^Cre mice. BAFF mRNA in Baff^fl/fl Mrp8^Cre mice was dramatically reduced in Nphs, but some reduction was evident in MOs. In agreement with previous findings suggesting that radiation-resistant but not hematopoietic cells were responsible for B cell homeostasis (56), we found that depleting BAFF from either Nphs or DCs did not alter the splenic B cell compartment in naive mice (data not shown).

We infected Baff^fl/fl zDC^Cre mice and Baff^fl/fl Mrp8^Cre mice and monitored their survival and Ab responses. Baff^fl/fl zDC^Cre mice had significantly increased mortality compared with Baff^fl/fl and zDC^Cre controls (Fig. 7A). In contrast, Baff^fl/fl Mrp8^Cre mice had similar survival rate as Baff^fl/fl mice (Fig. 7A). Furthermore, Baff^fl/fl zDC^Cre mice had increased virus titers (Fig. 7B) and decreased WNV-specific IgG and WNV neutralizing Abs (Fig. 7C, 7D) compared with controls. Thus, BAFF produced by DCs is required for optimal protective immune responses to WNV infection. Apparently, BAFF from cDCs helps to sustain or promote B cell humoral responses to WNV because WNV-specific Ab responses are decreased in mice lacking BAFF expression on cDC.

BAFF produced by DCs, Nphs, or MOs has been reported to contribute to TI-2 Ab responses (25, 32, 33, 57). To examine further whether BAFF from Nphs or cDCs play a role in TI-2 Ab responses, we immunized Baff^fl/fl Mrp8^Cre mice and Baff^fl/fl zDC^Cre mice with the TI-2 Ag, NP-Ficoll (Fig. 8). Both Baff^fl/fl Mrp8^Cre mice and Baff^fl/fl zDC^Cre mice had reduced NP-specific IgM and IgG3 production (Fig. 8A, 8B, respectively). Thus, both Nph and cDC BAFF sources are required for normal Ag-specific Ab responses to TI-2 Ags. Because BAFF depletion in Nph in Baff^fl/fl Mrp8^Cre mice was selective but not restricted to Nphs, we cannot formally exclude that BAFF from MOs may also contribute to the reduced Ab response to NP-Ficoll. However, Baff^fl/fl Cx3cr1^Cre mice, which have reduced BAFF expression in MOs, did not have reduced NP-specific Ab responses compared with control mice (Supplemental Fig. 1E).

BAFF is essential for the host response to pathogens and provides a critical link between innate and adaptive immune responses (1, 58). Although BAFF dysregulation has been implicated in several diseases, studies investigating BAFF-producing cells have been limited either to in vitro experiments or used models in which BAFF or its receptors were completely blocked or ablated (13, 29, 32, 33, 57, 59). In this study, we developed and used BAFF-RFP mice and Baff^fl/fl mice to better understand BAFF sources and their functions in B cell responses. We found that BAFF is expressed in several blood cell subsets as well as lymphoid tissues, including the spleen and LNs. The major cells expressing BAFF are Nphs, DCs, and MO subsets. Furthermore, BAFF expression can undergo cell- and tissue-specific upregulation upon in vivo activation of different RNA-sensing pathways. In addition, selective deletion of BAFF expression in cDCs or Nphs revealed that BAFF from both cDCs and Nphs contribute to TI-2 Ab responses, whereas cDC BAFF, unlike Nph BAFF, is required for optimal protective immunity against WNV.

Nphs express the highest BAFF levels in blood and in every tissue we tested, including the spleen, iLNs, and BM. These data are consistent with previous reports describing Nphs as a major source of BAFF (21, 60, 61). BAFF production by Nphs has been reported in human autoimmune diseases and mouse models (62–64). However, BAFF production by Nphs during viral infections has not been previously noted. BAFF expression by Nphs was slightly upregulated after in vivo challenge with two ssRNA stimuli: R848 and WNV. WNV-induced BAFF increase in Nphs was IFNAR dependent but MAVS independent, suggesting that type 1 IFN production that regulates BAFF in Nphs might be induced through the TLR7/8 pathway. Although Nph BAFF production increased after viral infection, it is not required for protective immunity to WNV. In contrast, BAFF produced by Nphs is required for protective immunity against Salmonella typhimurium infection (R. Kuley, K.E. Draves, K.D. Smith, N.V. Giltiay, E.A. Clark, and D. Giordano, manuscript in preparation). Further investigation is required to elucidate the possible roles of BAFF produced by Nphs during viral and bacterial infections. Puga et al. (56) previously reported that BAFF produced by human Nphs may contribute to TI Ab responses (57). In support of this study, we demonstrated that selective depletion of BAFF from Nphs reduces Ag-specific Ab responses to a TI-2 Ag. Nphs strategically located in the peri-MZ of the spleen may help B cells to rapidly respond to blood-borne TI Ags (57, 65, 66).

Earlier studies also implicated DCs in TI Ab responses, and more recently, several cDC subsets were reported to be localized in the MZ (32, 67–69). In agreement with these findings, we found that BAFF from cDCs is required for optimal humoral responses to a TI-2 Ag. The reduction but not full ablation of Ab responses to NP-Ficoll in either BAFF Nph cKO or BAFF cDC cKO mice suggests that both Nphs and cDCs contribute to TI-2 Ab responses. It is also possible that some BAFF-producing cells compensate for the absence of BAFF production by another source. Further studies are needed to identify specific niches in lymphoid tissues where BAFF-producing cDCs and Nphs provide help to B cell responses (70, 71).

BAFF plays an important protective role during viral infections (13, 72). In particular, BAFF signaling is an important regulator of humoral responses and is required for survival against lethal WNV infection (73). In this study, we found that cDC BAFF is essential for protection from WNV infection. In contrast to BAFF Nph cKO mice, BAFF cDC cKO mice after WNV infection had increased mortality and increased virus titers. The reduced WNV-specific IgG and WNV neutralizing Abs in BAFF cDC cKO mice suggest that BAFF from cDCs is essential for the development of protective humoral responses to WNV. How BAFF-producing cells contribute to control WNV replication, whether it is through the regulation of B cell responses or other mechanism, needs further investigation. Consistent with the protective role of cDC-derived BAFF, WNV infection upregulated BAFF expression in CD8⁻ cDCs. WNV-induced upregulation of BAFF in CD8⁻ cDCs was dependent on MAVS, suggesting that this occurred through WNV activation of the RIG-I pathway (36, 55). Because both p-DCs and cDCs express the RIG-I/MAVS pathway and can produce type I IFN upon viral challenge (47, 55, 74), it remains to be determined if BAFF upregulation in cDCs after WNV infection is dependent on type I IFN produced by cDCs or by p-DCs.

The IFNAR-dependent and MAVS-dependent BAFF regulation in inflammatory MO subsets by Poly(I:C) most likely occurs via MDA5 triggered by long dsRNA (36, 55, 75). Our finding that Poly(I:C) increased BAFF levels in splenic inflammatory Ly6C^hi MO and Ly6C^hi DC subsets is consistent with previous reports. Poly(I:C) and virus-induced type I IFN production by p-DCs upregulates BAFF levels in MOs and myeloid DCs (12, 31, 76). Inflammatory MO/DCs also express TLR3 and produce IFN-α in response to Poly(I:C) and dsRNA (74, 77–80). Furthermore, the MAVS pathway has been implicated in type I IFN production by inflammatory MOs and MO-derived DCs (74, 79). Further studies are needed to establish whether the MAVS/IFNAR requirement for Poly(I:C)-induced BAFF upregulation in inflammatory MO subsets occurs via type I production by p-DCs or in an autocrine fashion.

Using BAFF-RFP mice, we identified a murine NK cell subset expressing BAFF induced by the activation of RNA-sensing pathways. The increased expression of CD11b, NKp46, and Ly49C in BAFF⁺ CD11b^hi NK cells, compared with CD11b^int/− NK cells, suggests that the BAFF⁺ CD11b^hi NK cells are activated (54, 81). The fact that the expansion of BAFF⁺ CD11b^hi NK cells is dependent on IFNAR suggests type I IFN plays a major role not only in the upregulation of BAFF expression by NK cells but also in the induction of this activated NK subset. These findings are in agreement with previous reports implicating type I IFN in the activation of NK cells during viral infections (82–85). NK cells play an important role in antiviral immunity not only through the NK-mediated killing of virus-infected cells but also by regulating adaptive immune responses (84, 86, 87). NK cells can enhance Ag presentation by B cells and can promote TI Ab responses against bacteria or in autoimmune settings (88–90). Also, a human NK subset has been shown to produce BAFF (91). BAFF expression by human NK cells inhibits sensitivity to treatment of chronic lymphoid leukemia and has been correlated to a single nucleotide polymorphism associated with susceptibility to autoimmune disease (92, 93). It would be interesting to investigate whether the activated BAFF-expressing NK cells we have characterized play a role in regulating B cell responses during infections or during the development of autoimmune diseases.

Remarkably, in the BM, not only Nphs but also some CD11b⁺ myeloid precursors express high levels of BAFF. Interestingly, Poly(I:C)-dependent BAFF regulation in BM MO subsets was similar to their counterparts in the spleen and required the MAVS/IFNAR pathway. These BAFF⁺ myeloid precursors are most likely late precursors because they express CD11b. Among the myeloid precursors expressing higher BAFF levels, the population we define as CD11b⁺ myeloid precursor lacking Ly6C may be an earlier progenitor of a precursor not committed to a specific granulocyte/MO lineage, as described by Yáñez et al. (48, 50). Consistent with reports of BM-derived DCs being a source of BAFF (33), we found that BM pre-DCs expressed substantial BAFF levels. Despite the high BAFF expression in Nph and DC precursors in the BM and spleens, these cell sources were not required to maintain the general mature B cell compartment. Our results are in agreement with earlier findings suggesting that BAFF produced by radiation-resistant cells, possibly stromal cells, are responsible for normal B cell homeostasis (56),

Interestingly, a subset of CD19⁺B220⁺CD11b⁺ B cells in the BM expressed BAFF; these cells were CD43^+loCD5⁻CD11c⁻, and some were IgM⁺ and IgD⁺. Further analysis is required to define this subset better. They may belong to the B1 B cell lineage because they express CD11b and BAFF (53, 94) or be a BM B cell subset providing BAFF to BM long-lived PCs as a survival factor (1, 5, 95). We also found that BAFF is expressed in some peripheral B cell subsets; further studies are required to define the functions of this source of BAFF.

Our new BAFF-RFP and Baff^fl/fl mouse lines may be useful for identifying BAFF sources and defining their in vivo functions. Localization of BAFF-RFP positive cells in the spleen and lymph nodes would also be very helpful to identify specific niches of BAFF-producing cells during immune responses. Given the major role of BAFF in the host response to pathogens (58) and in autoimmune diseases, more insights into the localization, timing, and function of BAFF-producing cells regulating B cell responses may help in the development of more targeted vaccines and therapies.

We thank Dr. Keith Elkon for helpful insights and discussion. We thank Michel Black and the Cell and Analysis Flow Cytometry and Imaging Core in the Department of Immunology at the University of Washington for their support.

The authors have no financial conflicts of interest.