IFNγ induces epigenetic programming of human T-bethi B cells and promotes TLR7/8 and IL-21 induced differentiation

Although B cells expressing the IFNγR or the IFNγ-inducible transcription factor T-bet promote autoimmunity in Systemic Lupus Erythematosus (SLE)-prone mouse models, the role for IFNγ signaling in human antibody responses is unknown. We show that elevated levels of IFNγ in SLE patients correlate with expansion of the T-bet expressing IgDnegCD27negCD11c+CXCR5neg (DN2) pre-antibody secreting cell (pre-ASC) subset. We demonstrate that naïve B cells form T-bethi pre-ASCs following stimulation with either Th1 cells or with IFNγ, IL-2, anti-Ig and TLR7/8 ligand and that IL-21 dependent ASC formation is significantly enhanced by IFNγ or IFNγ-producing T cells. IFNγ promotes ASC development by synergizing with IL-2 and TLR7/8 ligands to induce genome-wide epigenetic reprogramming of B cells, which results in increased chromatin accessibility surrounding IRF4 and BLIMP1 binding motifs and epigenetic remodeling of IL21R and PRDM1 loci. Finally, we show that IFNγ signals poise B cells to differentiate by increasing their responsiveness to IL-21.


Introduction
Systemic Lupus Erythematosus (SLE) is characterized by progressive dysregulation of the innate and adaptive arms of the immune system, which ultimately leads to loss of immune tolerance in B and T lymphocytes and the production of autoantibodies (Abs) by Ab-secreting B cells (ASCs) . The hallmark SLE autoAbs recognize nuclear proteins and nucleic acids (Gatto et al., 2016), which are also ligands for TLR7 and TLR9 that are expressed by innate immune cells and B cells (Avalos et al., 2010). SLE autoAbs bound to their autoAgs form immune complexes, which are responsible for many of the clinical manifestations of SLE, particularly those associated with organ damage (Gatto et al., 2016). Consistent with the important role for B cells and ASCs in SLE pathogenesis (Sanz, 2014), the only new drug approved to treat SLE in decades, Belimumab, targets B cells.
Inflammatory cytokines and chemokines also contribute to SLE pathogenesis (Apostolidis et al., 2011). SLE patient PBMCs often exhibit a type I interferon (IFN) transcriptional signature and systemic IFNa is elevated in many patients (Obermoser and Pascual, 2010). It is less well appreciated that IFNg is also increased in some SLE patients (Csiszár et al., 2000;Harigai et al., 2008;Pollard et al., 2013) and that a distinct IFNg transcription signature can be detected in PBMCs from a portion of SLE patients (Chiche et al., 2014;Welcher et al., 2015). Interestingly, elevated serum IFNg can be observed years before IFNa or autoAbs are detected in SLE patients and much earlier than clinical disease Lu et al., 2016). Consistent with these observations, B cells from SLE patients can exhibit signs of prior IFNg exposure. For example, two IFNg-inducible proteins, CXCR3 and T-bet, are more highly expressed by circulating B cells from SLE patients compared to healthy controls (Harigai et al., 2008;Nicholas et al., 2008;Lit et al., 2007;Wang et al., 2018;Jenks et al., 2018). Moreover, data from mouse SLE models show that clinical disease is dependent on B cell-specific expression of the IFNgR and the IFNg-induced transcription factors (TF) STAT1 (Domeier et al., 2016;Jackson et al., 2016;Thibault et al., 2008) and T-bet in some Liu et al., 2017) but not all (Jackson et al., 2016;Du et al., 2019) models. Taken together, these data suggest that IFNg-driven inflammation may contribute to SLE B celldriven pathophysiology.
Two interrelated populations of circulating B cells present in SLE patients, namely the CD11c hi B cells, which are also called age associated B cells (ABCs) (Karnell et al., 2017;, and the IgD neg CD27 neg B double negative (B DN ) B cells, which are often referred to as 'atypical' memory B cells (Wei et al., 2007;Portugal et al., 2017), are reported to express the IFNginducible TF T-bet. These B cells, which may have been exposed to IFNg at some step in their developmental process, are present in low numbers in the blood or tonsils of healthy individuals (Ehrhardt et al., 2005) and are reported to be expanded in chronically infected, aging and autoimmune individuals (reviewed in Naradikian et al., 2016), including patients with SLE Jenks et al., 2018). The CD11c hi population found in SLE patients is heterogeneous and contains CD11c-expressing IgD neg CD27 + switched memory (B SW ) cells, IgD neg CD27 neg naïve (B N ) cells and B DN cells . The B DN population is also heterogeneous and can be subdivided using CD11c and CXCR5 into the DN2 (CD11c hi CXCR5 neg ) subset, which express T-bet, and the DN1 (CD11c lo CXCR5 + ) subset, which does not express T-bet (Jenks et al., 2018).
Despite extensive data showing that these overlapping populations of CD11c hi B cells and B DN cells are expanded in a number of human diseases (Naradikian et al., 2016), our understanding regarding their origin and function is incomplete. Although initial studies examining B DN cells from malaria or HIV-infected individuals described these B cells as anergic (reviewed in Portugal et al., 2017), more recent studies reported that the CD11c-expressing IgD neg CD27 + CD21 lo activated B SW cells from influenza vaccinated humans (Lau et al., 2017) and HIV infected patients (Knox et al., 2017), as well as the CD11c hi cells from SLE patients  and the CD11c hi DN2 cells from SLE patients (Jenks et al., 2018) possess phenotypic and molecular characteristics of pre-ASCs. Both the CD11c hi B cells and the more narrowly defined DN2 subset from SLE patients differentiate into ASCs following stimulation Jenks et al., 2018). Moreover, the T-bet hi DN2 subset from SLE patients can produce autoAbs (Jenks et al., 2018), suggesting that these cells can potentially contribute to disease.
Since T-bet hi DN2 pre-ASCs produce autoAbs and correlate with disease severity in SLE patients (Jenks et al., 2018), we set out to identify the signals that control the development and differentiation of this population into ASCs. Here we show that expansion of the DN2 cells in SLE patients correlates with systemic concentrations of IFNg and IFNg-induced cytokines. We demonstrate that activation of naïve B (B N ) cells from healthy donors or SLE patients with IFNg-producing T cells or IFNg+TLR7/8 and BCR ligands induces formation of a T-bet hi pre-ASC population that is phenotypically, transcriptionally and functionally similar to the SLE T-bet hi DN2 subset. We show that IFNg signals significantly augment ASC differentiation by sensitizing B N cells to respond to BCR, IL-2 and TLR7/8 signals and by promoting global epigenetic changes in the activated B cells that lead to significantly increased chromatin accessibility surrounding binding sites for the key ASC commitment TFs, BLIMP1 and IRF4. We find IFNg-dependent differentially accessible regions (DARs) within the IL21R and PRDM1 (BLIMP1) loci and show that early IFNg signaling promotes increased IL-21R expression and responsiveness. Finally, we observe that the key IFNg-regulated epigenetic changes in the in vitro generated T-bet hi B DN pre-ASC subset and the molecular signals required to induce ASC development are conserved in the SLE patient DN2 cells. Collectively, these data suggest that IFNg signals can augment ASC development and may regulate the formation of pathogenic autoreactive pre-ASCs in some SLE patients.

Expansion of T-bet hi DN2 cells correlates with systemic IFNg levels in SLE patients
Recent studies from our group (Stone et al., 2019) revealed that differentiation of mouse B cells activated in the presence of IFNg-producing T cells was dependent on B cell intrinsic expression of the IFNgR and the IFNg-induced transcription factor (TF), T-bet. This result fit well with data from our group (Figure 1-figure supplement 1) and others Liu et al., 2017) showing that B cell intrinsic expression of T-bet is required for the development of autoAb-mediated disease in SLE mouse models and suggested that IFNg signaling in B cells might also regulate development of ASCs from autoreactive B cells. Consistent with this possibility, we and others Jenks et al., 2018) identified a population of circulating T-bet-expressing B cells in SLE patients, referred to as DN2 cells (Jenks et al., 2018), that express high levels of T-bet and exhibit phenotypic and functional properties of pre-ASCs. Based on these data, we postulated that the T-bet hi DN2 pre-ASC population that is expanded in a subset of SLE patients likely arises in response to IFNg-dependent signals. To test this hypothesis, we first assessed whether expansion of the T-bet hi DN2 pre-ASC subset in SLE patients correlated with IFNg levels in these patients. Consistent with our prior studies using a different cohort of SLE patients (Jenks et al., 2018), we observed that a subset of our SLE patients presented with an expanded population of circulating IgD neg CD27neg (double negative, B DN cells) (Figure 1a-b) that could be subdivided into CD11c + CXCR5 neg DN2 cells and CD11c neg CXCR5 + DN1 cells (Figure 1c). The DN2 cells, but not the DN1 cells, uniformly expressed high levels of T-bet ( Figure 1d) and also expressed high levels of CD19 and FcRL5  Figure 1g). Next, we measured cytokines in plasma from the SLE patients (Figure 1h-k). Consistent with our hypothesis, we observed a significant positive correlation between IFNg, as well as the IFNg-induced cytokines CXCL10, IL-6 and TNFa, and the frequency of T-bet hi DN2 cells in these individuals ( Figure 1h). These data therefore indicated that the circulating T-bet hi B DN cells present in our SLE patient cohort were phenotypically identical to the previously described (Jenks et al., 2018) DN2 pre-ASC subset and that this pre-ASC population is most expanded in SLE patients with elevated amounts of autoAbs, IFNg and IFNg-driven inflammatory cytokines.

IFNg-producing Th1 cells promote development of T-bet hi B DN cells and ASCs
Since IFNg can induce T-bet expression in B cells (Stone et al., 2019) and the T-bet hi DN2 pre-ASCs are expanded in SLE patients with higher systemic levels of IFNg, we predicted that the IFNg might regulate the formation of T-bet hi pre-ASCs. To test this, we developed an in vitro B cell/T cell mixed lymphocyte reaction (MLR) paired co-culture system ( Figure 2a) containing B N cells (purified as described in Figure 2-figure supplement 1a) purified from the peripheral blood or tonsil of one HD and highly polarized human Th1 and Th2 effectors (Zhu et al., 2010), which were generated in vitro using purified naïve peripheral blood T cells isolated from a second unrelated HD. The Th1 cells expressed T-bet and produced IFNg and IL-8 following restimulation while Th2 cells expressed GATA-3 and produced elevated levels of IL-4, IL-5, and IL-13 (  (a-f) Characterization of T-bet hi B cells in peripheral blood B cell subsets from healthy donor (HD) and SLE patients. Gating strategy to identify CD38 hi CD27 + ASCs, B cells (non-ASCs) (a, left) and double negative IgD neg CD27 neg (B DN ) cells (a, right) from the peripheral blood of HD and SLE patients. Frequency of B DN cells (b) within the total B cells. Subdivision of the SLE B DN population into CXCR5 + CD11c lo DN1 and T-bet hi CXCR5 neg CD11c hi DN2 populations (c) with T-bet expression levels (d) in each subset shown as a histogram. Expression of BLIMP1 (e) and IRF4 (f) by ASCs, T-bet hi B DN cells and T-bet lo B cells from SLE patients. Representative flow plots and mean fluorescence intensity (MFI) expression of BLIMP1 and IRF4 in each population are shown. (g) Correlation analysis between frequency of circulating T-bet hi B DN cells and anti-Smith autoAb titers in SLE patients. (h-k) Correlation (h) between plasma cytokine levels and frequency of T-bet hi B DN cells in SLE patient peripheral blood. Plasma concentration of IFNg (i), CXCL10 (j) and TNFa (k) in HD (blue symbols) and SLE patients (red symbols). See Horizontal black lines represent the median (b,i-k) within the group. Data shown from n = 20 HD and 40 SLE patients (b), representative flow plots from 16 SLE patients (c-d), 3 SLE patients (e-f), [16][17][18] or 5 HD and 26 SLE patients (i-k). Statistical analyses were performed using a non-parametric Mann-Whitney test (b,ik), a one-way paired T test (e-f) or Spearman Correlation test (g-h). Correlation P and r values listed in the figure. P values * 0.05, **<0.01, ***<0.001. DOI: https://doi.org/10.7554/eLife.41641.002 The following figure supplements are available for figure 1: Since neither the Th1 nor Th2 cells produced IL-21 following restimulation (Figure 2-figure supplement 1d), we added IL-21 to the co-cultures to ensure optimal B N activation (Ettinger et al., 2008;Tangye, 2015) and included IL-2 to enhance the survival of the T effectors (Rochman et al., 2009). After 6 days in culture, approximately 50% of the HD B cells activated in the presence of IFNg-producing Th1 cells (Be1 cells) expressed T-bet while very few (<3%) of the HD B cells activated with IL-4 producing Th2 cells (Be2 cells) upregulated T-bet ( Figure 2b). Approximately half of the T-bet hi B cells present in the Be1 cultures downregulated IgD and these cells were CD19 hi CD27 neg CD11c + -FcRL5 + CD23 neg (Figure 2c). Therefore, activation of B N cells with Th1 cells and IL-21 +IL-2 resulted in the formation of a T-bet hi IgD neg CD27 neg B DN population that was phenotypically similar to the SLE patient-derived T-bet hi DN2 cells.
In addition to observing the T-bet hi B DN pre-ASC like population in the Be1 co-cultures, we also identified CD38 hi CD27 + ASCs in both the Be1 and Be2 co-cultures ( Figure 2d). However, we always found more ASCs in the Be1 co-cultures, even across multiple experiments using B N and T effectors from different HD pairs ( Figure 2e). To address whether the increased ASC formation observed in the Be1 co-cultures was limited to isotype switched or unswitched B cells, we measured the frequency of IgM and IgG-producing (gated as in Figure 2-figure supplement 2) ASCs across multiple paired Be1 and Be2 co-cultures. Again, we found that ASCs, regardless of isotype, were greatly enriched in the Be1 co-cultures (Figure 2f-g). This increase in ASCs in the Be1 co-cultures was not due to intrinsic differences in the proliferative rates of the cells in each culture but rather that a higher proportion of the Be1 cells at each cell division committed to the ASC lineage ( Figure 2-figure supplement 3). These data indicated that Be1 co-cultures efficiently promoted the formation of T-bet hi B DN pre-ASC-like cells and ASCs.

T-bet hi B DN cells induced with Th1 cells and IL-21 are pre-ASCs
Given the phenotypic similarities between the in vitro generated T-bet hi B DN cells and SLE patient T-bet hi DN2 cells and the fact that the in vitro cultures containing T-bet hi B DN cells also efficiently formed ASCs, we predicted that the in vitro generated Tbet hi B DN cells were likely to be pre-ASCs. To test this, we first asked whether the in vitro generated T-bet hi B DN cells were transcriptionally related to SLE patient-derived T-bet hi DN2 pre-ASCs or to ASCs from HD. We therefore sort-purified IgD neg CD27 neg B DN cells ( Figure 3a) from 3 independent paired day 6 Be1 and Be2 co-cultures and performed RNA-seq analysis (Supplementary file 1). We identified 427 differentially expressed genes (DEGs) between the B DN cells from the Be1 and Be2 co-cultures ( Figure 3b). Consistent with our data showing that T-bet was selectively upregulated in the B cells from Be1 co-cultures, we observed significantly higher levels of TBX21 mRNA in the in vitro induced B DN Be1 cells compared to B DN Be2 cells ( Figure 3c). Next, we used Gene Set Enrichment Analysis (GSEA) to compare the transcriptomes of the in vitro generated Be1 and Be2 B DN cells to the T-bet hi DN2 population isolated from SLE patients (Jenks et al., 2018, Supplementary file 2) and to curated ASC transcriptome datasets (Abbas et al., 2005;Tarte et al., 2003). Consistent with our phenotyping data, the transcriptome of the T-bet expressing B DN Be1 cell subset was highly enriched relative to the B DN Be2 cells for genes that are specifically upregulated in the SLE-derived T-bet hi DN2 subset ( Figure 3d). Moreover, the transcriptome of the in vitro-induced B DN Be1 population was significantly enriched in expression of genes that are upregulated in ASCs compared to B N cells (Figure 3e), mature B cells ( Figure 3f) and switched memory B cells (Figure 3g). In addition, genes that are direct targets of IRF4 and upregulated in ASCs (Shaffer et al., 2008)      To determine whether the Be1 T-bet hi B DN cells were functional pre-ASCs, we sort-purified the IgD neg CD27 neg B DN cells from day 6 Be1 and Be2 co-cultures, labeled the sorted B DN cells with Cell Trace Violet (CTV), incubated the cells for 18 hr in conditioned media and enumerated CD38 hi CD27 + ASCs in the cultures. As expected, the sorted Be1 and Be2 B DN cells were activated, with 47-65% of the cells undergoing one cell division within 18 hr (Figure 3k). CD38 hi CD27 + ASCs were only detected in proliferating cells (Figure 3k), indicating that the sorted B DN cells include pre-ASCs that are poised to differentiate within one round of replication. Although ASCs were detected in the cultures containing either Be1 or Be2 B DN cells, significantly more ASCs were found in cultures containing the sorted T-bet expressing Be1 B DN cells ( Figure 3l). Thus, activation of B N cells with Th1 cells and IL-21 +IL-2 gave rise to a population of T-bet hi B DN cells that were phenotypically, transcriptionally and functionally similar to the T-bet hi DN2 pre-ASCs that are expanded in SLE patients (Jenks et al., 2018).

IFNg is required for in vitro development of T-bet hi B DN pre-ASCs and ASCs from B N cells
Since the in vitro generated Th1-induced T-bet hi B DN subset and the SLE patient derived T-bet hi DN2 pre-ASC population (Jenks et al., 2018) were quite similar, we asked whether we could use our in vitro co-culture system to define the minimal signals required to generate this potentially pathogenic population of T-bet hi pre-ASCs. Using Ingenuity Pathway Analysis (IPA) to interrogate the Be1 and Be2 B DN cell RNA-seq data-sets, we identified predicted upstream regulators of the T-bet hi B DN pre-ASC transcriptional network. These included antigen receptor signaling molecules, like Btk, cytokines, like IFNa, IFNg, IL-2 and IL-21, and cytokine-induced TFs, like STAT1 and STAT3 ( Figure 4a). In addition, both TLR7 and TLR9 were predicted as upstream regulators of the T-bet hi B DN Be1 cells (Figure 4a). This was unexpected, given that we did not add exogenous TLR ligands to the co-cultures, however, endogenous TLR7 and TLR9 ligands are known to be released by dying cells in vitro (Sindhava et al., 2017).
Next, we addressed whether stimulation of B N cells with the IPA-predicted activators of the T-bet hi B DN transcriptional network was sufficient to induce the formation of the T-bet hi B DN pre-ASC population. We therefore stimulated HD B N cells with anti-Ig, cytokines (IFNg, IL-2, IL-21 and BAFF) and the TLR7/8 ligand, R848 ( Figure 4b) and evaluated the B cells on day 6. We found that >95% of the B N cells activated with these defined stimuli resembled SLE patient T-bet hi DN2 cells (Jenks et al., 2018) as the in vitro activated cells were IgD neg CD27 neg T-bet hi IRF4 int , expressed the DN2 markers, CD11c and FcRL5, and were losing expression of CD21 and CXCR5 (Figure 4c-d).
To address which signals were critical for the in vitro development of T-bet hi B DN cells we set up 'all minus one cultures' by activating B N cells for 3 days with or without individual stimuli ( Figure 4e). As expected, when HD B N cells were activated for 3 days in the presence of anti-Ig and all cytokines + R848 (ALL condition), essentially all of the cells upregulated T-bet and IRF4 (Figure 4fg). Similar results were observed when the B N cells were activated for 3 days without anti-Ig ( Figure 4f) or without R848, IL-21, BAFF or IL-2 ( Figure 4g). By contrast, when the cells were activated without IFNg, more than 80% of the cells were T-bet neg/lo (Figure 4g). While this wasn't particularly surprising, given that T-bet is IFNg-inducible (Stone et al., 2019), the cells also failed to upregulate IRF4 (Figure 4g), indicating that IFNg signals are obligate for the in vitro generation of the T-bet hi IRF4 int B DN pre-ASC like population.
Although HD B N cells activated with anti-Ig, cytokines and R848 developed in an IFNg-dependent fashion into T-bet hi IRF4 int B DN pre-ASC like cells (Figure 4g), CD38 hi CD27 + ASCs did not accumulate in the cultures containing all stimuli ( Figure 4h). This suggested that our defined cultures lacked a factor that was necessary for the differentiation of the T-bet hi pre-ASC-like cells into ASCs. Alternatively, it was possible that one or more of the stimuli present in the cultures either blocked differentiation or needed to be provided transiently during a discrete temporal window. Since anti-Ig was not added in our original in vitro Th1 and B N co-cultures, we first examined whether BCR signaling was  blocking ASC development. We therefore stimulated B N cells for 6 days with the complete activation cocktail (+,+) or removed the anti-Ig from the activation cocktail for the first three days (-,+), last three days (+,-), or throughout the entire culture period (-,-) ( Figure 4i). Consistent with our prior results, few ASCs were recovered when anti-Ig was included throughout the culture period ( Figure 4j). Similarly, excluding anti-Ig from the culture for all 6 days or for the first 3 days also resulted in poor ASC recovery ( Figure 4j). However, when anti-Ig was present only during the first 3 days of culture ASCs accumulated in the cultures ( Figure 4j). These data therefore argued that early but transient BCR signals were important for the development and recovery of ASCs from cytokine and R848 stimulated B N cells.
Next, we asked whether IFNg signals were required for the development of ASCs in the culture. We therefore activated B N cells with the cytokine cocktail and R848 for 6 days, including 3 days in the presence of anti-Ig and 3 days without anti-Ig. In individual cultures we excluded specific cytokines or R848 for all 6 days ( Figure 4k). In agreement with our earlier experiment, ASCs were recovered ( Figure 4l) when B cells were transiently activated with anti-Ig in the continuous presence of R848 and the complete cytokine cocktail. Although elimination of BAFF or IL-2 from the cultures decreased the number of ASCs recovered from the cultures (Figure 4l), neither cytokine was obligate for ASC development. By contrast, and consistent with prior reports showing that ASC development from B N cells requires IL-21 (Ettinger et al., 2008;Tangye, 2015), no ASCs were detected in the cultures lacking IL-21 ( Figure 4l). Likewise, ASC recovery in cultures lacking R848 or IFNg was also at background levels ( Figure 4l). Collectively, the data indicated that formation of the T-bet hi IR-F4 int pre-ASC like population required IFNg signals while the development and recovery of ASCs were dependent on transient BCR signals, IFNg, R848 and IL-21.
Temporal control of ASC development from T-bet hi IRF4 int pre-ASCs by IFNg, R848 and IL-21 Although the number of ASCs recovered from cultures lacking IL-21, R848 or IFNg was equally low (Figure 4l), the frequencies of ASCs and number of total cells recovered from each culture differed dramatically (Figure 4-figure supplement 1). These data suggested that the different stimuli were likely to play distinct roles in the development and recovery of ASCs. Since IFNg, but not R848 or IL-21, was required for the formation of the pre-ASC population, we postulated that IFNg signals would be required during the initial activation (Days 0-3, priming phase) while TLR7/8 and IL-21 signals subsets. Data are reported as Enrichment Score (ES) plotted against the ranked B DN Be1 and Be2 gene list (n = 11598). DEG lists used for GSEA include: DEGs that are upregulated in sort-purified SLE patient-derived T-bet hi DN2 cells (CD19 hi IgD neg CD27 neg CXCR5 neg IgG + ) compared to other SLE patient-derived mature B cell subsets (d, Jenks et al., 2018); DEGs that are upregulated in human plasma cells (ASCs) relative to: B N cells (e, Abbas et al., 2005), total B cells (f, Tarte et al., 2003) or switched memory (B SW ) B cells (g, Abbas et al., 2005); and IRF4-dependent upregulated target genes in ASCs (h, Shaffer et al., 2008). (i-j) IgD neg CD27 neg T-bet hi B DN cells express intermediate levels of IRF4. Gating strategy (i) to identify CD38 hi CD27 + ASCs, IgD + CD27 neg B cells and IgD neg CD27 neg B DN cells in day 6 Be1 co-cultures generated from HD B N cells. Expression of T-bet and IRF4 (j) by ASCs (blue), IgD + CD27 neg B cells (green) and IgD neg CD27 neg B DN cells (red) from day 6 Be1 co-cultures. (k-l) B DN Be1 cells rapidly differentiate into ASCs. B DN cells from day 6 HD Be1 and Be2 cultures were sort-purified, Cell-Trace Violet (CTV) labeled and incubated 18 hr in conditioned medium. Enumeration of ASCs (CD19 lo CD38 hi CD27 + ) in the undivided cells (D0, ) and the cells that divided one time (D1, ). Representative flow plots (k) showing the frequency of cells in D0 or D1 in each culture and the frequency of CD19 lo CD38 hi CD27 + ASCs present in the D0 or D1 fraction. Panel (  showing T-bet and IRF4 expression (f-g) by day 3 B cells in each culture. Enumeration of CD38 hi CD27 + ASCs (h) in day 6 'ALL' cultures. (i-j) Transient BCR activation is required for ASC development. Cartoon (i) depicting activation of HD B N cells for 3 days with R848, cytokines (IFNg, IL-2, IL-21, BAFF) ±anti-Ig (Step 1). Cells were then washed and recultured for an additional 3 days with the same stimuli ± anti-Ig (Step 2). Enumeration of CD38 hi CD27 + ASCs (j) on day 6 in cultures that were not exposed to anti-Ig during Steps 1 and 2 (-,-); were exposed to anti-Ig throughout Steps 1 and 2 (+,+); were exposed to anti-Ig only in Step 1 (+,-); or were exposed to anti-Ig only in Step 2 (-,+). (k-l) IFNg, R848 and IL-21 are required for ASC development. Cartoon (k) showing HD B N cells activated with anti-Ig + cytokine cocktail (IFNg, IL-2, IL-21, BAFF) and R848 for 3 days (Step 1) and then cultured for an additional 3 days (Step 2) with cytokine cocktail and R848. Alternatively, individual stimuli (as indicated) were excluded from the cultures for all 6 days. Enumeration of day 6 CD38 hi CD27 + ASCs (l). See Figure 4-figure supplement 1 for % ASCs and number of total cells recovered in cultures lacking individual stimuli. RNA-seq IPA analysis was performed on n = 3 samples/subset derived from 3 independent paired co-culture experiments. Data in (c-l) are representative of !3 Figure 4 continued on next page would be more critical later in the culture period (Days 4-6, differentiation phase). To test this hypothesis, we activated CTV-labeled B N cells for 3 days in the presence of anti-Ig and 3 days without anti-Ig -while adding the various stimuli minus one during the priming phase (+,-), during the differentiation phase (-,+) or throughout (+,+) the culture period ( Figure 5a). We then measured proliferation, cell recovery and the frequency and number of ASCs present in cultures on day 6 (see Next, we analyzed when TLR7/8 signals were necessary for ASC development. When R848 was only added during the first 3 days, ASCs could not be detected in the cultures, whether measured as the frequency (Figure 5f) or number (Figure 5g) of ASCs. This was due, at least in part, to the fact that proliferation was severely stunted (Figure 5h), resulting in greatly reduced cell recovery (Figure 5i) in the day 6 cultures. When R848 was only added to the cultures between days 3-6, we observed no impact on pre-ASC formation ( Figure 5-figure supplement 1d) or the frequency of ASCs in the day 6 cultures (Figure 5f). However, the number of cells recovered on day 6 was significantly reduced (Figure 5i), which affected the number of ASCs recovered in the cultures ( Figure 5g). Despite the poor recovery of cells in the cultures that received TLR7/8 stimulation only between days 3-6, proliferation of the cells was not impacted (Figure 5h). These data therefore indicated that R848 played both early and late roles in the development of ASCs, with early TLR7/8 signals appearing to promote B cell survival and late TLR7/8 signals promoting proliferation.
Finally, we assessed when IL-21 signals were required for ASC development. When IL-21 was only included for the first 3 days of the culture, pre-ASCs formed normally (  (Figure 5k) of ASCs recovered from the cultures that were exposed to IL-21 between days 3-6 only were very similar to cells that were stimulated all 6 days in the presence of IL-21. Therefore, late IL-21 signals were sufficient to drive ASC formation. Thus, while inclusion of IFNg, TLR7/8 ligand and IL-21 throughout the entire culture period promoted optimal ASC recovery, IFNg and BCR signals were required during the priming phase, IL-21 was necessary during the later expansion and differentiation phase and R848 was important throughout the culture period (Figure 5n). . Temporally distinct regulation of T-bet hi IRF4 int pre-ASC and ASC development by IFNg, R848 and IL-21. Cartoon (a) depicting stimulation of CTV-labeled HD B N cells for 3 days with anti-Ig, R848, IL-21 and IFNg (Step 1). Cells were washed and re-cultured for 3 days with R848, IFNg, and IL-21 (Step 2, +,+ condition) or individual stimuli were included in Step 1 only (+,-condition) or in Step 2 only (-,+ condition). Cells from day 6 cultures containing IFNg (b-e), R848 (f-i) or IL-21 (j-m) in Step 1, Step 2 or both steps were analyzed to determine ASC frequencies (b, f, j), ASC recovery (c, g, k), cell division (d, h, l) and total cell recovery (e, i, m). Summary of data (n) showing that ASC development and recovery from T-bet hi IRF4 int B DN pre-ASCs requires early IFNg, R848 and BCR 'priming' signals and late R848 and IL-21 proliferation and differentiation signals. See Figure 5-figure supplement 1 for representative flow cytometry plots from each culture showing T-bet hi IRF4 int B DN cells on day 3, CD38 hi CD27 + ASCs on day 6 and CTV dilution on day 6. Data are representative of !3 experiments. The percentage of cells in each division, the frequency of ASCs and cell recovery (total and ASCs) are shown as the mean ±SD of cultures containing purified B N cells from 3 independent healthy donors. All statistical analyses were performed using one-way ANOVA with Tukey's multiple comparison test. P values *<0.05, **<0.01, ***<0.001, ****<0.0001. DOI: https://doi.org/10.7554/eLife.41641.013 The following figure supplement is available for figure 5:  IFNg synergizes with R848 and IL-2 to promote proliferation, IL-21 responsiveness and ASC recovery Our data indicated that IFNg played a non-redundant and critical role in the formation of the Tbet hi IRF4 int B DN cells in vitro, and was necessary for development and recovery of ASCs, even when IL-21 and R848 were present. These data led us to hypothesize that IFNg signaling might sensitize B cells to respond to other stimuli, like IL-21, IL-2 and TLR ligands, that promote B cell proliferation and differentiation. To test whether IFNg signals promoted B cell responsiveness to R848 we activated CTV-labeled HD B N cells with anti-Ig, IL-2 and increasing concentrations of R848 in the presence and absence of IFNg for 3 days, washed the cells and then re-cultured them for an additional 3 days with IL-21 and the same concentration of R848 that the cells were exposed to during the priming phase. On day 6 we measured cell division and ASC formation. Consistent with our earlier experiments ( Figure 5), the B cells remained largely undivided when R848 was completely excluded from the cultures (Figure 6a). By contrast, when high dose R848 was included in the cultures, the cells proliferated regardless of whether IFNg was included in the cultures for the first 3 days ( Figure 6b). However, when we activated B N cells with a 100-fold lower dose of TLR7/8 ligand, proliferation was only seen in the cultures that contained IFNg (Figure 6c). Moreover, we observed that the frequency of ASCs in the cultures that were activated with low dose TLR ligand in the presence of IFNg was approximately 10-fold higher than that observed for the cultures that lacked IFNg (Figure 6d). Similar results were seen when we cross-titrated the IFNg and R848 in the cultures (Figure 6-figure supplement 1). Thus, exposure of B N cells to IFNg during the initial priming phase allowed these cells to differentiate even in the face of sub-optimal stimulation with R848.
Next, we asked whether the IFNg priming signals enhanced the early response of B cells to cytokines. We first assessed cooperation between IFNg and IL-2 as IL-2, while not obligate for ASC development, did significantly enhance ASC recovery in our in vitro cultures (Figure 4). We activated HD B N cells for 3 days with anti-Ig +R848 (Be.0 conditions), anti-Ig +R848+IL-2 (Be.IL2 conditions), anti-Ig +R848+IFNg (Be.IFNg conditions) or with anti-Ig +R848+IL-2+IFNg (Be.g2 conditions). We then washed and stimulated the cells for an additional 3 days with R848 +IL-21 ( Figure 6e) and evaluated cell recovery and ASC formation (see Figure 6-figure supplement 2 for representative flow cytometry plots). As expected, we recovered very few viable cells (Figure 6f (Figure 6h-i). However, when B cells were exposed to both IL-2 and IFNg during the early priming phase, the number of ASCs recovered on day 6 ( Figure 6i) was significantly more than seen in the Be.IFNg or Be.IL2 cultures. This was due to an increase in the number of cells recovered (Figure 6g) and to an increase in the frequency of ASCs (Figure 6h) in the cultures. Thus, early IFNg and IL-2 signals cooperate to induce formation and recovery of ASCs.
Finally, since IL-21 signaling was obligate for ASC differentiation in our in vitro cultures, we hypothesized that early IFNg signals might program the B cells to respond to IL-21. To test this hypothesis, we measured phosphorylation of the IL-21R associated TF, STAT3, before and after IL-21 stimulation in day 3 Be.0, Be.IL2, Be.IFNg and Be.g2 cells. Day 3 basal levels of phospho-STAT3 were similar and low in the Be.0, Be.IL2 and Be.IFNg cells and modestly higher in the Be.g2 cells (Figure 6j, see Figure 6-figure supplement 3 for flow cytometry plots). However, following a 20 min exposure to IL-21, phospho-STAT3 levels were increased significantly in the B cells that were exposed to IFNg during the priming phase (Figure 6k), indicating that early IFNg stimulation enhanced IL-21R signaling. Collectively, these data show that early IFNg signals sensitize human B N cells to respond more robustly to stimuli, like TLR7/8 ligands, IL-2 and IL-21, that promote B cell activation, proliferation and differentiation. Step 1) with anti-Ig, IL-2, and increasing concentrations of R848 (as indicated) in the presence or absence of IFNg (10 ng/ml). Cells were washed and re-cultured for 3 additional days ( Step 2) with IL-21 and the same concentration of R848 that was used in Step 1. B cell division was measured on day 6 in cultures that were activated with IFNg (green circles) or without IFNg (orange circles) in the presence of no R848 (0 mg/ml, (a), high dose R848 (10 mg/ml, (b) or low dose R848 (0.1 mg/ml, (c). The frequency of CD38 hi CD27 + ASCs (d) on day 6 is shown. (e-i) IFNg cooperates with IL-2 to promote ASC development and recovery. Cartoon (e) depicting CTV-labeled HD B N cells activated for 3 days (Step 1) with anti-Ig and R848 alone (Be.0); with anti-Ig +R848+IFNg (Be.IFNg); with anti-Ig +R848+IL-2 (Be.IL2); or with anti-Ig +R848+IFNg+IL-2 (Be.g2). Cells were then washed and recultured for an additional 3 days (Step 2) with R848 and IL-21. The percentage of cells that have undergone cell division (f), the total cell recovery (g), the ASC frequencies (h) and total ASCs recovered (i) from each day 6 culture are shown. (j-k) Early IFNg signals regulate IL-21R signaling. Phospho-STAT3 (pSTAT3) expression levels (reported as Mean Fluorescence Intensity (MFI)) in day 3 HD Be.0, Be.IFNg, Be.IL2 and Be.g2 cells under basal conditions (j) or following 20 min IL-21 stimulation (k). See   Early IFNg signals cooperate with IL-2 and R848 to initiate ASC epigenetic programming and IL-21R expression Our data showed that early IFNg signals cooperated with both IL-2 and R848 to promote IL-21 dependent ASC formation and recovery. Given the importance of IFNg in driving the development of the T-bet hi pre-ASC like population, we hypothesized that IFNg might induce molecular and epigenetic changes that would initiate early commitment to the ASC lineage and/or regulate IL-21R expression and responsiveness. To test this possibility, we used ATAC-seq analysis (Supplementary file 3) to identify differentially accessible regions (DAR) in the genome of Be.0, Be. IL2, Be.IFNg and Be.g2 cells on day 3 -a time in which cell recovery was similar in the cultures (Figure 7a-b, see Figure 7-figure supplement 1 for representative flow plots) and the T-bet hi pre-ASC like population was easily detected in the IFNg-containing cultures. As expected, distinct sets of DAR were found in all 4 groups of activated B cells (Figure 7c), however the largest number of chromatin accessible regions was seen in the day 3 Be.g2 cells (Figure 7c). Moreover, the chromatin accessibility pattern in the Be.g2 cells appeared to reflect cooperation or synergy between the IFNg and IL-2 signals (Figure 7c). Examination of chromatin accessibility within 100 bp surrounding consensus TF binding motifs revealed significant (see Supplementary file 4 for statistical analyses) enrichment in accessibility near T-bet binding sites in the B cells that were exposed to IFNg (Figure 7d). Similarly, accessibility around STAT5 binding motifs was enriched in IL-2 exposed B cells ( Figure 7e). However, the Be.g2 cells exhibited the greatest enrichment in chromatin accessibility surrounding both T-bet and STAT5 binding sites (Figure 7d-e), suggesting that IFNg and IL-2 cooperate to remodel the epigenome. Consistent with this, binding motifs for NF-kB p65 and REL, TFs activated by anti-Ig and TLR7/8 stimulation (Kaileh and Sen, 2012), were most accessible in the Be.g2 cells compared to all other groups (Figure 7f-g). Moreover, chromatin accessibility surrounding the HOMER-defined IRF4 and BLIMP1 binding motifs (Heinz et al., 2010) was also highly enriched in the Be.g2 cells (Figure 7h-i). These data therefore suggested that these key ASC initiating TFs were already exerting epigenetic changes to the genome of the Be.g2 cells, even before these cells were exposed to IL-21. Consistent with this finding, when we examined the PRDM1 (BLIMP1) locus, we identified 4 DAR that were each more accessible in the Be.g2 cells relative to the other cells (Figure 7j). Although none of these DAR contained a T-bet binding motif, each DAR directly aligned with peaks previously identified in a published T-bet ChIP-seq analysis of GM12878 cells (ENCODE Project Consortium, 2012), suggesting that T-bet could be associated with TF complexes that bind to these regulatory regions. Moreover, 3 of the 4 PRDM1-associated DAR were also seen in T-bet hi DN2 cells purified from SLE patients (Figure 7j), indicating that these DAR were present in the pre-ASC population found in SLE patients.
Finally, given our data showing that IFNg and IL-2 potentiated signaling through the IL-21R, we examined the 2 DAR assigned to the IL21R locus of the day 3 cells (Figure 7k). One of the DAR contained two putative T-bet binding motifs and was directly aligned with a T-bet ChIP-seq peak from GM12878 cells (ENCODE Project Consortium, 2012) (Figure 7k). This DAR was only observed in the cells that were exposed to IFNg and was most enriched in the Be.g2 population. Interestingly, we identified the same DAR in the SLE patient T-bet hi DN2 cells (Figure 7k), which are reported to be highly responsive to IL-21 (Jenks et al., 2018). To address whether these early IFNg-dependent epigenetic changes in the IL21R were associated with altered expression of IL-21R, we measured IL-21R expression in the day 3 and day 6 stimulated cells. Although day 3 B cells from Be.IFNg and Be.g2 cultures expressed slightly higher levels of IL-21R compared to B cells from Be.0 and Be.IL2 cultures (Figure 7l), IL-21R expression were comparable between all groups at this timepoint. By day 6 however, IL-21R expression levels were 5.5-6-fold higher in the B cells that were cultured in the presence of IFNg during the first 3 days (Figure 7m). Taken together, the data suggested that early IFNg signals synergize with BCR, TLR and IL-2 signals to induce global changes in chromatin accessibility and promote increased TF binding at T-bet, NF-kB, STAT5, BLIMP1 and IRF4 binding sites as well as chromatin remodeling at the PRDM1 and IL21R loci.

SLE patient T-bet hi DN2 cells differentiate into ASCs without a further requirement for BCR stimulation
Previous data from our group (Jenks et al., 2018) showed that the T-bet hi DN2 cells from SLE patients were transcriptionally distinct from conventional memory cells and, like B N cells (Tangye, 2015), require IL-21 signals to differentiate. Since our in vitro culture system accurately predicted that the T-bet hi DN2 cell differentiation would be IL-21 dependent, we hypothesized that the in vitro culture data could be used to make additional testable predictions about the molecular properties of the T-bet hi DN2 cells found in SLE patients. To evaluate this possibility, we first tested the prediction that IFNg-dependent ASC formation from the B N cells isolated from SLE patients would require transient BCR stimulation. We therefore purified T-bet lo B N cells (see Figure 8-figure supplement 1 for purification strategy) from the peripheral blood of SLE patients and stimulated the cells with the complete cytokine cocktail (IFNg, IL-2, IL-21 and BAFF) plus R848 for 6 days in the continuous presence of anti-Ig (+,+), in the complete absence of anti-Ig (-,-) or in the presence of anti-Ig for the first 3 days (+,-) ( Figure 8a). Consistent with our prediction, SLE patient B N cells did acquire phenotypic characteristics of the T-bet hi DN2 subset following in vitro activation with R848, IL-2, IFNg and IL-21 ( Figure 8b). Moreover, the recovery of ASCs in the cultures started with SLE patient B N cells was highly dependent on transient but early stimulation with anti-Ig as continuous stimulation with anti-Ig or no stimulation with anti-Ig reduced both the frequency and number of ASCs recovered in the cultures (Figure 8c-d). Thus, these data indicated that transient BCR stimulation was required for ASC development from SLE patient-derived B N cells activated with R848, IFNg, IL-2 and IL-21.
Based on these data and our in vitro experiments, we made two additional testable predictions. First, we postulated that T-bet hi DN2 cells isolated from SLE patients should differentiate without a requirement for BCR stimulation. Second, we predicted that SLE patient T-bet hi DN2 cells should differentiate more rapidly than B N cells. To test these predictions, we sort-purified (see  Jenks et al., 2018) and IgD neg CD27 + memory (conventional B mem ) subsets. We stimulated the cells for 2.5 days with R848, IFNg, IL-21 and IL-2 and then enumerated IgG-producing ASCs. As expected, the conventional B mem and DN1 memory cells efficiently formed ASCs in this short timeframe (Figure 8e), while B N cells failed to differentiate (Figure 8e). Consistent with our predictions, ASCs were easily identified in the day 2.5 cultures containing T-bet hi DN2 cells (Figure 8e). Indeed, ASC recovery was at least 50-fold higher in T-bet hi DN2 cell cultures compared to the B N cultures and only 2-3 times less than that seen with the memory B cell populations (Figure 8e). These data therefore suggested that the expanded population of T-bet hi DN2 cells present in some SLE patients likely represent a population of IFNg, TLR ligand and antigen programmed primary effectors that can rapidly differentiate Chromatin accessibility plots and histograms for T-bet (d), STAT5 (e), NF-kB p65 (f), NF-kB REL (g), IRF4 (h) and BLIMP1 (i). Plots report reads per million (rpm) in the 100 bp surrounding the transcription factor binding motifs and histograms show accessibility at the indicated motif and for the indicated surrounding sequence. Genome plots showing chromatin accessibility for the PRMD1 (j) and IL21R (k) loci. DAR are shown and consensus T-bet, IRF4 and STAT5 binding motifs within DAR are indicated. DAR are aligned with previously reported T-bet binding sites in GM12878 cells (assessed by ChIP, ENCODE Project Consortium, 2012) and with ATAC-seq data derived from B cell subsets purified from SLE patients (Jenks et al., 2018). Data reported in rpm. (l-m) Early IFNg signals control IL-21R expression levels. Representative flow plots showing IL-21R expression in day 3 (l) and day 6 (m) Be.0, Be.IFNg, Be.IL2 and Be.g2 cells. See Step 1) and then washed and recultured for an additional 3 days with the same stimuli ± anti-Ig (Step 2). Cells were analyzed by flow cytometry on day 6 (b-d). Phenotypic characterization (b) of IgD neg CD27 neg B DN cells in cultures containing anti-Ig for all 6 days showing expression of T-bet, CD11c, FcRL5, CD21 and CXCR5 by the T-bet hi B DN subset. The frequency (c) and number (d) of CD38 hi CD27 + ASCs in cultures lacking anti-Ig (-,-), containing anti-Ig for all 6 days (+,+) or exposed to anti-Ig for the first 3 days only (+,-). (e) SLE patient T-bet hi B DN cells rapidly differentiate in ASCs. Purified SLE B cell subsets (T-bet lo B N , T-bet lo CD11c neg CXCR5 + CD27 neg IgD neg DN1 memory cells, T-bet lo CD27 + memory B cells (B mem ) and T-bet hi CD11c + CXCR5 neg DN2 cells) were stimulated with cytokines (IFNg, IL-21, IL-2, BAFF) and R848 for 2.5 days then counted and transferred to anti-IgG ELISPOT plates for 6 hr. The frequency of IgG ASCs derived from each B cell subset is shown. See  in a BCR-signaling independent manner into ASCs following IL-21 exposure. The importance of IFNg in driving human ASC commitment and differentiation in the context of autoimmune disease is discussed.

Discussion
Here we show that IFNg promotes the in vitro formation of a T-bet hi IRF4 int IgD neg CD27 neg (B DN ) population that is similar to the T-bet expressing CD11c hi CXCR5 neg B DN (referred to as DN2 cells) subset found in SLE patients (Jenks et al., 2018) and the CD11c hi Age-Associated B cells (ABCs) that accumulate in aged and autoimmune mice and humans . Both the in vitro generated T-bet hi IRF4 int B DN cells and SLE patient-derived DN2 cells (Jenks et al., 2018) exhibit transcriptional and functional properties of pre-ASCs, suggesting that B cell intrinsic IFNgR signals could regulate human ASC responses. While this hypothesis is supported by mouse experiments showing that IFNg signals enhance autoAb responses (Domeier et al., 2016;Jackson et al., 2016;Lee et al., 2012) and recent studies from our group demonstrating a role for B cell expression of the IFNgR and the IFNg-inducible transcription factor T-bet in ASC development (Stone et al., 2019), the role for IFNg and STAT1 signaling in human B cell differentiation is less clear. In fact, prior data showing that IFNg has only very modest effects on activation and differentiation of human B cells (Nakagawa et al., 1985;Splawski et al., 1989;Rousset et al., 1991) and that patients deficient in the IFN-activated transcription factor STAT1 produce Abs in response to some vaccines (Chapgier et al., 2009;Chapgier et al., 2006) argue that IFNg signaling is not obligate for the formation of human ASCs. Our in vitro studies do not contradict this conclusion as we also find that human B cells can differentiate in the absence of IFNg-induced signals. However, we show that B cell intrinsic IFNg signals significantly enhance ASC differentiation induced in response to stimulation with anti-Ig, TLR7/8 ligand, IL-2 and IL-21. Indeed, we routinely recover 5-to 10 fold more ASCs in the B N cultures that contain IFNg or IFNg-producing T cells compared to cultures that lack IFNg. Thus, we argue that IFNg signaling has the potential to augment ASC development in settings, like autoimmunity and viral infection, where IFNg and TLR ligands are present.
Our data show that IFNg signals, when delivered in conjunction with IL-2 and BCR +TLR7/8 ligand during the initial activation of B N cells, greatly increase ASC recovery in vitro. The co-activation of B N cells with IFNg and IL-2 +BCR +TLR7/8 ligand results in IFNg-dependent chromatin remodeling and the formation of the T-bet hi IRF4 int pre-ASC subset. These early IFNg signals are required for subsequent proliferation and differentiation following stimulation with IL-21 and TLR7/8 ligand. IFNg is not, in and of itself, a B cell mitogen and is reported to induce apoptosis of human B cells (Bernabei et al., 2001;Sammicheli et al., 2011). However, we find that IFNg synergizes with TLR7/8 ligand to promote multiple rounds of B cell proliferation -an important prerequisite of human ASC differentiation (Tangye et al., 2003). Although in vitro experiments using human B cells show that IFNg can synergize with TLR7 and CD40 signals to promote upregulation of Bcl6 and the acquisition of a germinal center-like phenotype (Jackson et al., 2016), our data extend these studies to show that IFNg and TLR7/8 signals also cooperate to promote human B cell differentiation. These results are analogous to studies showing that IFNa-directed signals can enhance TLR7-mediated human B cell differentiation (Jego et al., 2003;Bekeredjian-Ding et al., 2005). Given the considerable overlap between genes regulated by IFNa and IFNg (Pollard et al., 2013), it is possible that IFNa and IFNg may augment TLR7 signaling in human B cells by similar mechanisms.
Our in vitro data suggest multiple ways in which early IFNg priming signals promote subsequent ASC differentiation and recovery. First, we show that IFNg cooperates with IL-2, BCR, TLR7/8 ligand to globally alter the epigenetic landscape of the activated B cells and to specifically increase chromatin accessibility surrounding NF-kB, STAT5 and T-bet binding sites. While it is not particularly surprising that IFNg signaling induces increased T-bet expression and alterations in chromatin accessibility around T-bet binding sites (see for example Iwata et al., 2017), the finding that chromatin accessibility surrounding NF-kB and STAT5 binding motifs is also regulated by IFNg suggests that IFNg must augment TLR7/8 and IL-2-dependent signaling. This is consistent with our in vitro data showing synergistic effects on ASC recovery when IFNg is combined with IL-2 or R848. Second, we show that IFNg promotes commitment to the ASC lineage by inducing expression of IRF4 and modifying chromatin accessibility surrounding IRF4-binding sites within regulatory regions in the genome of the activated B N cells. Third, we find that IFNg signals promote chromatin accessibility within the PRDM1 (BLIMP1) locus and initiate chromatin remodeling around BLIMP1-binding sites within the genome of the activated B N cells. Finally, we demonstrate that IFNg signals alter chromatin accessibility within the IL21R locus of the activated B N cells and that this change in accessibility is associated with IFNg-dependent, increased expression of the IL-21R by the activated B N cells and with increased responsiveness of these cells to IL-21, as measured by phosphorylation of STAT3. Collectively, these data suggest that IFNg signals, when combined with BCR, TLR and IL-2R signals, poise human B N cells to differentiate in response to IL-21.
One key finding from this study is that IFNg-augmented ASC formation and recovery is highly reliant on TLR7/8 activation by its RNA and RNA/protein ligands, which are derived from viral pathogens and dead and dying cells (Avalos et al., 2010). Signaling through TLR7 is known to be important in SLE as prior studies reveal that SNPs in the human TLR7 locus (Lee et al., 2016) and overexpression of TLR7 in mice (Pisitkun et al., 2006) are associated with increased SLE susceptibility while deletion of TLR7 protects mice from the development of SLE (Christensen et al., 2006). Our data show that deletion of the IFNg-inducible transcription factor T-bet in B lineage cells prevents autoAb responses in a mouse model (Pisitkun et al., 2006) of TLR7-dependent SLE. Moreover, our data demonstrates that B cell intrinsic IFNg signaling induces a TLR7/8 hyperresponsive state in human B cells. This finding does not appear to be due to IFNg-dependent changes in the expression of TLR7 by the B cells (data not shown). Rather, we find that IFNg-exposed B N cells can respond and differentiate into ASCs when exposed to 100-fold lower concentration of TLR7/8 ligands than normally used to activate B cells. Given that we observed that even low levels of IFNg are sufficient to synergize with suboptimal concentrations of TLR7/8 ligands, we predict that B cells from autoimmune patients with detectable systemic levels IFNg will be highly sensitive to the presence of endogenous and exogenously derived TLR7 ligands.
Our data predict that TLR7-driven ASC responses are likely to be further enhanced in individuals who have increased levels of circulating IFNg. Consistent with this, we show that SLE patients who have higher systemic levels of IFNg also have more T-bet hi DN2 cells and higher autoAb titers. We and others Jenks et al., 2018) also report that the size of the T-bet hi DN2 population correlates with disease activity, particularly in African-American SLE patients. However, it is important to note that T-bet hi DN2 cells are unlikely to represent a purely 'pathogenic' population as we also find an inducible population of vaccine-specific T-bet hi CD27 neg DN2 cells in healthy individuals who were immunized with inactivated influenza virus (data not shown). Similarly, others report (Lau et al., 2017;Knox et al., 2017) a T-bet expressing CD27 + CD21 lo switched memory subset with pre-ASC attributes, which is induced following vaccination or infection. Thus, we speculate that the T-bet hi B cells, which are found in HD and autoimmune patients in the settings of acute and chronic inflammation driven by vaccination, infection, autoimmunity and aging, are formed in an IFNg-dependent manner and likely represent a pool of primary and secondary pre-ASCs as well as effector memory B cells that are epigenetically poised to differentiate.
The IFNg-induced T-bet hi IRF4 int pre-ASC population that we characterized in our in vitro studies is similar to the T-bet hi DN2 subset that is expanded in SLE patients. Since the expansion of the T-bet hi DN2 cells in SLE patients correlates with systemic levels of IFNg and IFNg-induced cytokines, we postulate that the DN2 cells likely arise in an IFNg-dependent fashion in these patients. In support of this possibility, we demonstrate that the IFNg-directed changes in chromatin accessibility within the IL21R and PRDM1 loci seen in the in vitro generated T-bet hi IRF4 int pre-ASCs are also found in T-bet hi DN2 cells isolated from SLE patients. Moreover, we show that the molecular properties of the SLE DN2 subset and the in vitro generated IFNg-dependent T-bet hi B DN cells are similar and unique when compared to conventional memory B cells or B N cells. For example, as discussed above, IFNg primes the T-bet hi B DN cells to respond to subthreshold concentrations of R848. Similarly, SLE DN2 cells make augmented responses to TLR7/8-stimulation compared to other B cell subsets (Jenks et al., 2018). We also show that SLE patient DN2 pre-ASCs and the in vitro generated T-bet hi B DN subset can differentiate without a need for additional BCR stimulation. This is similar to memory B cells but unlike what we find for B N cells. However, like B N cells (Deenick et al., 2013), both the SLE DN2 subset Jenks et al., 2018) and the in vitro generated T-bet hi B DN subset require IL-21 to differentiate into ASCs. Thus, given the many shared phenotypic, molecular and functional properties of the in vitro generated T-bet hi B DN subset and the T-bet hi DN2 cells found in SLE patients, we think that the in vitro pre-ASC cultures described here could be used to better understand the development, maintenance and functional attributes of the T-bet hi DN2 cells that are expanded and associated with more severe disease in SLE patients.
In summary, we demonstrate that IFNg is critical for the in vitro formation of a T-bet hi IRF4 int pre-ASC population that is remarkably similar to the T-bet hi DN2 cells that accumulate in SLE patients who present with high autoAb titers, elevated disease activity and increased systemic levels of IFNg. We show that IFNg signals, particularly when combined with IL-2 and TLR7/8 + BCR ligands, initiate epigenetic reprogramming of human B cells -changes which poise the activated B N cells to respond to IL-21 and fully commit to the ASC lineage. Based on these results, we argue that blocking IFNg signaling in SLE patients should curtail development of T-bet hi DN2 pre-ASCs from primary B N cells, which would result in decreased autoAb production and reduced disease activity. However, results from a phase I trial examining IFNg blockade in SLE patients did not reveal a therapeutic benefit (Boedigheimer et al., 2017). Interestingly, no African Americans SLE patients with nephritis were included in the study (Boedigheimer et al., 2017). Given the data showing that T-bet hi DN2 cells are most expanded in African American patients with severe disease Jenks et al., 2018) and our data presented here showing that the DN2 cells likely develop in response to IFNg, we propose that future studies evaluating the efficacy of IFNg blockade in SLE patients should focus specifically on the subset of patients who present with elevated IFNg levels and significant expansion of the IFNg-inducible T-bet expressing DN2 pre-ASC population. Continued on next page À80˚C. Human PBMCs and tonsil mononuclear cells were either used immediately or were cryopreserved at À150˚C.

Human lymphocyte purification
Naïve CD4 + T cells and CD19 + B cells were isolated from human PBMCs or tonsils using EasySep enrichment kits (StemCell). B N cells were then positively selected using anti-IgD microbeads (Miltenyi). B cell subsets were sort-purified from PBMCs and tonsils as described in text.

B cell activation with defined stimuli
Purified B cell subsets isolated from the tonsil or blood of HD or SLE patients were cultured (1 Â 10 6 cells/ml) for 3 days with 5 mg/ml anti-Ig (Jackson ImmunoResearch), 5 mg/ml R848 (InvivoGen), 50 U/ ml IL-2, 10 ng/ml BAFF, 10 ng/ml IL-21 (Peprotech) and 20 ng/ml IFNg (R and D) (Step 1). Cells were either directly analyzed or washed and recultured (2 Â 10 5 cells/ml) for an additional 3 days with the same stimuli ( Step 2). The number of ASCs and total cells recovered in cultures on day 6 were determined and then normalized based on cell input. In some experiments, anti-Ig, R848, IL-21, IL-2, IFNg or BAFF were omitted from the cultures during Step 1, or Step 2 or both steps. In other experiments, the concentration of R848 in Step 1 and Step 2 and/or the concentration of IFNg in Step1 was varied, as indicated in the text. In some experiments, B cell subsets isolated from blood of SLE patients and HD were stimulated for 2.5-6 days with R848 and IL-21, IL-2, BAFF and IFNg.

STAT3 phosphorylation assays
HD B N cells were cultured with 5 mg/ml anti-Ig and 5 mg/ml R848 alone (Be.0) or in combination with IFNg (Be.IFNg), IL2 (Be.IL2), or IL2 plus IFNg (Be.g2). On day 3 cells were washed and restimulated with medium alone or with IL-21 (10 ng/ml) for 20 min at 37˚C. The cells were fixed and permeabilized with BD Transcription Factor Phospho Buffer Set and intracellular staining with anti phospho-STAT3 was performed.

In vitro B cell proliferation
Purified B cell subsets (1À5 Â 10 6 cells/ml) were stained for 10 min at 37˚C with PBS diluted Cell-Trace Violet (Molecular Probes, Thermofisher). The cells were washed and either used in T effector co-culture experiments or were cultured in the presence of defined stimuli.
In vitro ASC differentiation B N cells were co-cultured with in vitro generated Th1 or Th2 cells plus IL-2 and IL-21. On day 6 of the co-culture B DN cells from both cultures were sort-purified and then cultured in 0.22mM-filtered conditioned media (media collected from the original T/B co-cultures). ASCs were enumerated after 18 hr by flow cytometry.

Cytokine measurements
Th1 and Th2 cells were restimulated with platebound anti-CD3 and anti-CD28 (both 5 mg/ml). Cytokine levels in restimulated T cell cultures and SLE patient plasma samples was measured using Milliplex MAG Human Cytokine/Chemokine Immunoassays (Millipore).

Elispot
Serial diluted B cells were transferred directly to anti-IgG (Jackson ImmunoResearch) coated ELI-SPOT plates (Millipore) for 6 hr. Bound Ab was detected with alkaline phosphatase-conjugated antihuman IgG (Jackson ImmunoResearch) followed by development with alkaline phosphatase substrate (Moss, Inc). ELISPOTs were visualized using a CTL ELISPOT reader. The number of spots detected per well (following correction for non-specific background) was calculated.

Anti-SMITH ELISAs
Anti-Smith IgG autoantibodies in plasma from SLE patients and healthy donors were detected using the enzymatic immunoassay kit (Alpha Diagnostic) according to the manufacturer protocol.

Flow cytometry
Single cell suspensions were blocked with 10 mg/ml FcR blocking mAb 2.4G2 (mouse cells) or with 2% human serum or human FcR blocking reagent (Miltenyi) (human cells) and then stained with fluorochrome-conjugated Abs. 7AAD or LIVE/DEAD Fixable Dead Cell Stain Kits (Molecular Probes/ ThermoFisher) were used to identify live cells. For intracellular staining, cells were stained with Abs specific for cell surface markers, fixed with formalin solution (neutral buffered, 10%; Sigma) and permeabilized with 0.1% IGEPAL (Sigma) in the presence of Abs. Alternatively, the transcription factor and phospho-transcription factor staining buffers (eBioscience) were used. Stained cells were analyzed using a FACSCanto II (BD Bioscience). Cells were sort-purified with a FACSAria (BD Biosciences) located in the UAB Comprehensive Flow Cytometry Core. Analysis was performed using FlowJo v9.9.3 and FlowJo v10.2.

RNA-seq library preparation and analysis
RNA samples were isolated from TRIzol (FisherThermo) treated sort-purified day 6 Be1 and Be2 IgD neg CD27 neg B cells. 300 ng of total RNA from 3 biological replicates per B cell subset was used as input for the KAPA stranded mRNA-seq Kit with mRNA capture beads (KAPA Biosystems). Libraries were assessed for quality on a bioanalyzer, pooled, and sequenced using 50 bp paired-end chemistry on a HiSeq2500. Sequencing reads were mapped to the hg19 version of the human genome using TopHat with the default settings and the hg19 UCSC KnownGene table as a reference transcriptome. For each gene, the overlap of reads in exons was summarized using the Genomi-cRanges package in R/Bioconductor. Genes that contained two or more reads in at least 3 samples were deemed expressed (11598 of 23056) and used as input for edgeR to identify differentially expressed genes (DEGs). P-values were false-discovery rate (FDR) corrected using the Benjamini-Hochberg method and genes with a FDR of <0.05 were considered significant. Expression data was normalized to reads per kilobase per million mapped reads (FPKM). Data processing and visualization scripts are available (Scharer, 2019a;Scharer, 2019b;Scharer, 2019c; copies archived at https://github.com/elifesciences-publications/genomePlots, https://github.com/elifesciences-publications/heatmap, and https://github.com/elifesciences-publications/plotScaledBEDfeatures respectively). All RNA-seq data is available from the GEO database under the accession GSE95282. See also Supplementary file 1.

ATAC-seq preparation and analysis
ATAC-seq data generated from the SLE B cell subsets was previously reported (Jenks et al., 2018). ATAC-seq analysis on in vitro generated B cell was performed on 10,000 Be.0, Be.IFNg, Be.IL2 or Be.g2 cells as previously described (Scharer et al., 2016). Sorted cells were resuspended in 25 ml tagmentation reaction buffer (2.5 ml Tn5, 1x Tagment DNA Buffer, 0.2% Digitonin) and incubated for 1 hr at 37˚C. Cells were lysed with 25 ml 2x Lysis Buffer (300 mM NaCl, 100 mM EDTA, 0.6% SDS, 1.6 mg Proteinase-K) for 30 min at 40˚C, low molecular weight DNA was purified by size-selection with SPRI-beads (Agencourt), and then PCR amplified using Nextera primers with 2x HiFi Polymerase Master Mix (KAPA Biosystems). Amplified, low molecular weight DNA was isolated using a second SPRI-bead size selection. Libraries were sequenced using a 50 bp paired-end run at the NYU Genome Technology Center. Raw sequencing reads were mapped to the hg19 version of the human genome using Bowtie (Langmead et al., 2009) with the default settings. Duplicate reads were marked using the Picard Tools MarkDuplicates function (http://broadinstitute.github.io/picard/) and eliminated from downstream analyses. Enriched accessible peaks were identified using MACS2 (Zhang et al., 2008) with the default settings. Differentially accessible regions were identified using edgeR v3.18.1 (Robinson et al., 2010) and a generalized linear model. Read counts for all peaks were annotated for each sample from the bam file using the Genomic Ranges (Lawrence et al., 2013) R/Bioconductor package and normalized to reads per million (rpm) as previously described (Scharer et al., 2016). Peaks with a greater than 2-fold change and FDR < 0.05 between comparisons were termed significant. Genomic and motif annotations were computed for ATAC-seq peaks using the HOMER (Heinz et al., 2010) annotatePeaks.pl script. The findMotifsGenome.pl function of HOMER v4.8.2 (42) was used to identify motifs enriched in DAR and the 'de novo' output was used for downstream analysis. To generate motif footprints, the motifs occurring in peaks were annotated with the HOMER v4.8.2 annotatePeaks.pl function (Heinz et al., 2010) using the options '-size given'. The read depth at the motif and surrounding sequence was computed using the Genomi-cRanges v1.22.4 (66) package and custom scripts in R/Bioconductor. All other analyses and data display were performed using R/Bioconductor with custom scripts (Scharer, 2019a;Scharer, 2019b;Scharer, 2019c). ATAC-seq data has been deposited in the NCBI GEO database under accession number GSE119726. See also Supplementary files 3-5 for complete list of DAR and for analysis of TF motif enrichment in the ATAC-seq dataset.

GSEA
For gene set enrichment analysis samples were submitted to the GSEA program (http://software. broadinstitute.org/gsea/index.jsp). For the comparison of interest (i.e., B DN Be1 and B DN Be2 cells), all detected genes were ranked by multiplying the -log 10 of the P-value from edgeR by the sign of the fold change and used as input for the GSEA Preranked analysis. The custom gene set defining genes upregulated in SLE T-bet hi B DN relative to other B cell subsets were derived from Jenks et al. (2018) and are listed in Supplementary file 2.

Ingenuity Pathway Analysis (IPA)
IPA upstream regulator analysis (Krämer et al., 2014, Qiagen, Redwood City CA) was performed using the log 2 fold-change in gene expression between genes that were significantly differentially expressed (FDR < 0.05) in B DN Be1 and B DN Be2 cells. Upstream regulators with an activation z-score of !2 or À2 were considered to be activated or inhibited in B DN Be1 cells. Overlap P-value (between the regulator's downstream target list and the DEG list was based on Fisher's exact test.

Statistical analysis
Comparisons between two groups were performed with the Student's t test for normally distributed variables and the Mann-Whitney test for non-normally distributed variables. The one-way ANOVA test was used to compare mean values of 3 or more groups and the Kruskal-Wallis nonparametric test was used to compare medians. Strength and direction of association between two variables measures was performed using the D'Agostino-Pearson normality test followed by Pearson's or Spearman's correlation test. Data were considered significant when p 0.05. Analysis of the data was done using the GradhPad Prism version 7.0a software (GraphPad). See Supplementary file 5 for all statistical comparisons.

Mouse ANA detection and imaging
Antinuclear antibodies (ANA) were detected by an indirect immunofluorescence assay using HEp-2 cells. Fixed HEp-2-coated microscope slides (Kallestad, BioRad) were blocked, incubated with serum diluted 1:100 and stained with anti-IgG-FITC (Southern Biotech) (10 mg/ml). Slides were mounted with SlowFade Gold Antifade Mountant with DAPI (ThermoFisher) and imaged. Anti-nuclear staining was quantitated as the mean flourescence intensity (MFI) of IgG-FITC over DAPI-staining areas (nuclei) using NIS-Elements AR software (Nikon). Data are presented as log nuclear IgG MFI normalized by subtracting the MFI of negative control serum from B6 mice. ANA images were collected using a Nikon Eclipse Ti inverted microscope and recorded with a Clara interline CCD camera (Andor). The images were taken with a 20X (immunofluorescence) objective for 200-400X final magnification. Images were collected using NIS Elements software, scale bars were added and images were saved as high-resolution JPEGs. JPEG images were imported into Canvas Ver 12 software and were resized, cropped with the identical settings applied to all immunofluorescence images from the same experiment. Final images presented at 600-650 dpi (ANA).

Urinary Albumin to Creatinine Ratio (UACR)
Albumin concentrations in urine samples, collected from live or euthanized mice, were measured using the Mouse Albumin ELISA Quantitation Set (Bethyl Labs) according to manufacturer's protocol using a mouse reference serum as an albumin standard. To normalize for urine concentration, urinary creatinine was measured by liquid chromatography-mass spectrometry in the UAB/UCSD O'Brien Core Center for Acute Kidney Injury Research. The UACR was calculated as mg/ml albumin divided by mg/ml creatinine and is reported as mg albumin/mg creatinine.  (Jenks et al., 2018) on sort-purified T-bet hi -expressing IgD neg CD27 neg IgG + CXCR5 neg B cells from HD and SLE patients (DN2 cells). The DN2 Up DEG list is defined as genes that are significantly upregulated in SLE and HD DN2 cells relative to at least one other B cell subset (B N , switched memory or CXCR5-expressing (T-bet lo ) DN1 memory = cells). DOI: https://doi.org/10.7554/eLife.41641.024 . Supplementary file 3. ATAC-seq data set from day 3 Be.0, Be.IFNg, Be.IL2 and Be.g2 B cell subsets. HD B N cells were activated for 3 days with anti-Ig and R848 alone (Be.0) or in combination with: IFNg (Be.IFNg), IL-2 (Be.IL2) or both IFNg+IL-2 (Be.g2). ATAC-seq analysis was performed on DNA isolated from each B cell subset. Data availability Sequencing data have been deposited in GEO under accession codes GSE95282 and GSE118984. All data generated or analyzed during this study are included in the manuscript and supporting files. Source data files for sequencing analysis are included as Supplementary Files 1 and 2 (excel files).
The following datasets were generated: