A rt i c l e s

Lineage-specific functions of Bcl-6 in immunity and inflammation are mediated by distinct biochemical mechanisms

Chuanxin Huang, Katerina Hatzi & Ari Melnick

The transcription factor Bcl-6 orchestrates germinal center (GC) reactions through its actions in B cells and T cells and regulates inflammatory signaling in . Here we found that genetic replacement with mutated Bcl6 encoding Bcl-6 that cannot bind corepressors to its BTB domain resulted in disruption of the formation of GCs and affinity maturation of immunoglobulins due to a defect in the proliferation and survival of B cells. In contrast, loss of function of the BTB domain had no effect on the differentiation and function of follicular helper T cells or that of other helper subsets. Bcl6-null mice had a lethal inflammatory phenotype, whereas mice with a mutant BTB domain had normal healthy lives with no inflammation. The repression of inflammatory responses by Bcl-6 in macrophages was accordingly independent of the repressor function of the BTB domain. Bcl-6 thus mediates its actions through lineage-specific biochemical functions.

Bcl-6 is a transcriptional repressor originally identified as being Bcl-6 is a member of the BTB–zinc finger family of proteins. Its BTB encoded by a locus frequently translocated in diffuse large B cell lym- domain forms an obligate homodimer and it contains C2H2 zinc fin- phomas (DLBCLs)1. During normal development, Bcl-6 has critical gers that bind to DNA. The interface between monomers of the Bcl-6 functions in various cell types of the adaptive and innate compart- BTB domain creates two symmetrical extended lateral grooves that ments of the immune system. Bcl-6 expression undergoes substantial form docking sites for the corepressors SMRT, NCOR and BCOR20–22. upregulation in B cells after challenge with T cell–dependent antigens2, Those three corepressors bind to Bcl-6 via an unstructured 18–amino and Bcl-6 is required for the formation of germinal centers (GCs) acid Bcl-6-binding domain23. The Bcl-6-binding domains of NCOR in which B cells undergo affinity maturation of immunoglobulins. and SMRT are identical, whereas that of BCOR is completely differ- © 2013 Nature America, Inc. All rights reserved. America, Inc. © 2013 Nature Bcl-6-deficient (Bcl6–/–) mice do not form GCs and thus are unable ent, yet all three bind to the lateral groove of the Bcl-6 BTB domain to generate high-affinity antibodies3–5. The proposed biological func- in perfectly overlapping configurations23,24. Substitution of lysine tion of Bcl-6 in B cells in the GC is to facilitate the simultaneous rapid for asparagine at position 21 (N21K) and of alanine for histamine at npg proliferation and tolerance to the genomic damage that occurs during position 116 (H116A) in the Bcl-6 BTB domain completely abrogates clonal expansion and somatic hypermutation by directly repressing the binding of Bcl-6 to NCOR, SMRT and BCOR without impairing genes encoding molecules involved in DNA-damage sensing and folding or dimerization23. Mutant Bcl-6 with those N21K and H116A checkpoints, such as ATR6, CHEK1 (ref. 7), EP300 (ref. 8), TP53 (ref. 9) substitutions in its BTB domain is completely inactive, which indicates 10 and CDKN1A . Follicular helper T cells (TFH cells), specifically that the lateral groove–Bcl-6-binding domain interface is responsible those in GCs, are specialized CD4+ helper T cells that provide help to for the repressor activity of the Bcl-6 BTB domain23. However, Bcl-6 B cells during the GC reaction11,12. Bcl-6 expression is upregulated also has a middle autonomous repression region often called ‘repres- –/– 25 during the differentiation of TFH cells, and Bcl6 T cells do not dif- sion domain 2’ (RD2) , which may recruit other corepressors, such as 13–15 26,27 ferentiate into TFH cells in vivo . Constitutive expression of Bcl-6 NuRD and CTBP . The sequence of the reported Bcl-6 consensus 14–16 enhances the differentiation of TFH cells . The requirement for binding site (TTCCT(A/C)GAA, where A/C indicates either adeno- Bcl-6 in both B cells and TFH cells in GCs is cell autonomous, and loss sine or cytosine) overlaps that of STAT transcription factor–binding of Bcl-6 in either cell type leads to abrogation of the GC reaction17. sites25,28,29, and several lines of evidence indicate that Bcl-6 may Bcl-6 also has an important role in macrophages, in which it mediates antagonize STAT signaling, with potential relevance for inflamma- a dampening effect on inflammatory signaling through repression tory and innate immunological functions3,30,31. of expression and target genes of the transcription fac- Collectively, the data suggest complex biological roles for Bcl-6 in tor NF-κB18,19. Bcl6–/– mice develop a lethal inflammatory disease the immune system. However, the link between the transcriptional caused by the interaction and crosstalk between macrophages and mechanisms of action of Bcl-6 and its biological actions in the immune helper T cells. system remains unknown. Here we generated a knock-in mouse model

Division of Hematology and Oncology, Weill Cornell Medical College, New York, New York, USA. Correspondence should be addressed to A.M. ([email protected]).

Received 10 September 2012; accepted 10 January 2013; published online 3 March 2013; doi:10.1038/ni.2543

380 VOLUME 14 NUMBER 4 APRIL 2013 nature immunology A rt i c l e s

a PNA B220 + Bcl-6 b 15 Day 7 Day 10 Day 14 d e +/+ BTBMUT –/– +/+ BTBMUT –/–

) Bcl6 Bcl6 Bcl6 Bcl6 Bcl6 Bcl6 4 5 5 15.10 8.12 8.20 2.25 1.10 0.28 ** ** ** 10 10 10 4 4 Bcl6+/+ 10 10 103 103 5 102 5.77 0.42 0.20 102 0 0 CXCR5 CD38

GC size ( µ m × 10 2 3 4 5 0 102 103 104 105 0 10 10 10 10 0 Fas PD-1 +/+ +/+ +/+ Bcl6BTBMUT Bcl6 BTBMUTBcl6 BTBMUTBcl6 BTBMUT Day 0 Day 10 Day 0 Day 10 Day 0 Day 10 8 15 20 Bcl6 Bcl6 Bcl6 ** ** * Bcl6+/+ 6 cells (%) 15

4 cells (%) c BTBMUT cells (%) 10 lo–hi Bcl6 hi )

2 4 10 3 ** lo–neg PD-1 –/– ** PD-1

Bcl6 * hi 5 5

2 lo–hi 2 CD38 +

0 CXCR5 0 0 1 Fas GCs (per mm +/+ –/– +/+ –/– +/+ –/– +/+ –/– CXCR5 +/+ –/– +/+ –/–

0 Bcl6BTBMUTBcl6Bcl6 BTBMUTBcl6 Bcl6 BTBMUTBcl6Bcl6 BTBMUTBcl6 Bcl6 BTBMUTBcl6Bcl6 BTBMUTBcl6 Day 7 Day 10 Day 14 Bcl6 Bcl6 Bcl6 Bcl6 Bcl6 Bcl6

Figure 1 Impaired GC formation in Bcl6BTBMUT mice. (a) Immunohistochemistry of paraffin-embedded serial spleen sections 10 d after immunization of mice with SRBCs. Arrows (middle) indicate GCs; insets present enlargement of those GCs. PNA, peanut agglutinin. Scale bars, 200 µm (main images); original magnification (insets), ×20. (b,c) Size (b) and number (c) of GCs in spleen sections 7, 10 and 14 d after immunization of mice with SRBCs. (d) Flow cytometry (top) of GC B cells, gated on live B220+ splenic lymphocytes obtained from mice 10 d after immunization with SRBCs. Numbers in outlined areas indicate Fas+CD38lo–neg cells (GC B cells). Below, quantification of GC B cells among live B220+ cells, before and 10 d – + after immunization. (e) Flow cytometry of GC and total TFH cells, gated on live B220 CD4 splenic lymphocytes from mice 10 d after immunization hi hi lo–hi lo–hi with SRBCs. Numbers above outlined areas indicate CXCR5 PD-1 cells (right; GC TFH cells) and CXCR5 PD-1 cells (left; total TFH cells). – + Below, quantification of GC TFH and TFH cells among live B220 CD4 cells. Each symbol (b,d,e) represents an individual GC (b) or mouse (d,e); small horizontal lines indicate the mean (± s.e.m.). *P < 0.01 and **P < 0.001 (two-tailed t-test). Data are representative of five (a) or three (b–e) independent experiments (mean and s.e.m. of two to five mice).

in which the endogenous Bcl6 locus encodes a mutant form of Bcl-6 chromatin-immunoprecipitation (ChIP) assays showed that Bcl-6 protein with the N21K and H116A point substitutions in the BTB protein with the mutant BTB domain retained the same ability to domain (Bcl6BTBMUT). The facts that SMRT, NCOR and BCOR are bind target genes in cultured macrophages as that of wild-type Bcl-6 coexpressed with Bcl-6 in the relevant cell types and that the mecha- protein (Supplementary Fig. 2d). The abundance of SMRT at Bcl-6 nism of the BTB domain is the only well-characterized biochemical target genes in Bcl6BTBMUT mice was identical to that in Bcl6–/– mice function of Bcl-6 favor the proposal that the biological ‘readout’ of such (Supplementary Fig. 2d), which confirmed that these substitutions in a knock-in model would be most rigorously interpretable. Notably, our the lateral groove abrogated interaction with lateral groove–binding data suggest that the transcriptional mechanisms of action of Bcl-6 are corepressors. The mutated Bcl6 alleles thus encoded a viable tran- specific to the lineage and biological function, with notable implica- scription factor able to bind to its targets but unable to recruit core- tions for the general understanding of how Bcl-6 and other transcrip- pressors through its BTB domain. © 2013 Nature America, Inc. All rights reserved. America, Inc. © 2013 Nature tion factors work, as well as for the ‘translation’ of such results into the clinical use of inhibitors of Bcl-6. Impaired GC formation in Bcl6BTBMUT mice Phenotypic analysis showed normal early development of B cells in the npg RESULTS and spleens of Bcl6BTBMUT mice, as well as normal distri- Viability of Bcl6BTBMUT mice bution of peripheral T cells (Supplementary Fig. 3a–c). Bcl6BTBMUT To address the biological function of interactions between the Bcl- mice also formed normal splenic primary lymphoid follicles (data not 6 BTB domain and its corepressors in vivo, we introduced point shown). Analysis of the spontaneous formation of GCs in spleens of mutations in exon 3 and exon 4 of Bcl6 that resulted in the desired unimmunized mice by staining with peanut agglutinin showed a few N21K and H116A substitutions in the BTB domain (Bcl6BTBMUT; small GCs scattered in Bcl6+/+ mice, whereas none were present in Supplementary Fig. 1a,b). We crossed Bcl6+/BTBMUT mice hetero- Bcl6BTBMUT mice (data not shown). After immunization with sheep zygous for the neomycin-resistance cassette with mice with transgenic red blood cells (SRBCs), a T cell–dependent antigen, Bcl6BTBMUT mice expression of Cre recombinase from the adenovirus EIIa promoter developed only a few, scattered and small peanut agglutinin–positive (which targets Cre expression to the early mouse embryo) to gener- and Bcl-6+ cell clusters (Fig. 1a). Serial examination of GCs by immu- ate heterozygous Bcl6+/BTBMUT mice, which we then intercrossed to nohistochemistry at days 7, 10 and 14 after immunization with SRBCs generate homozygous Bcl6BTBMUT/BTBMUT mice (called ‘Bcl6BTBMUT showed that Bcl6BTBMUT mice had considerably smaller and fewer mice’ here; Supplementary Fig. 2a). Bcl6BTBMUT mice were born GCs than Bcl6+/+ mice had at all three time points (Fig. 1b,c). Lymph at the expected Mendelian ratios, were viable and were develop- nodes in Peyer’s patches of in Bcl6BTBMUT mice showed a similar loss mentally indistinguishable from their Bcl6+/+ littermates (data not of GCs and the presence of residual cell clusters (Supplementary shown), in contrast to Bcl6–/– mice, which are runted and sickly and Fig. 4a), which emphasized the general requirement for Bcl-6 BTB are born at lower than the expected ratios. Quantitative RT-PCR and domain–mediated repression in the formation of GCs. immunoblot analysis showed normal expression of mutated Bcl6 To better understand the mechanism of the GC impairment in transcripts and mutant Bcl-6 protein in splenic B220+ cells from Bcl6BTBMUT mice, we first compared the frequency of GC B cells Bcl6BTBMUT mice (Supplementary Fig. 2b). Sequencing of genomic (CD38lo–negFas+B220+) among splenic B cells from Bcl6+/+, Bcl6–/– DNA from tails of Bcl6BTBMUT mice confirmed the presence of the and Bcl6BTBMUT mice before and at day 10 after immunization with introduced point mutations (Supplementary Fig. 2c). Quantitative SRBCs. The frequency of GC B cells in unimmunized Bcl6+/+ mice

nature immunology VOLUME 14 NUMBER 4 APRIL 2013 381 A rt i c l e s

Bcl6+/+ Bcl6+/+ a BTBMUT b BTBMUT c Bcl6 Bcl6 +/+ +/+ BTBMUT ** 100 Bcl6 Bcl6 Bcl6 * 5 5 105 105 105 10 105 5 NS NS 10 0.76 1.10±0.15 1.20±0.20 80 NS 104 titer (RU)

titer (RU) 3 titer (RU) 26 titer (RU) titer (RU) 4 60 10 4 4 26 4 4

26 10 26

26 10 10 10 10 102 40 0 CD138 2 3 4 5 3 3 3 3 3 010 10 10 10 10 10 10 10 10 20

NP-specific spots /10 B220 IgG2b anti-NP IgM anti-NP IgG1 anti-NP IgG3 anti-NP IgG2a anti-NP 0 8 d 21 d 8 d 21 d 8 d 21 d 8 d 21 d 8 d 21 d IgM IgG 80 60 0.13 8.90±1.40 9.50±1.90 +/+ +/+ NS Figure 2 Intact extrafollicular responses but impaired GC responses Bcl6 Bcl6 40 d BTBMUT e BTBMUT BTBMUT Bcl6 Bcl6 20

in Bcl6 mice. (a) Titers of NP-specific immunoglobulins, Cells 1.5 0 5 *** measured with NP26-BSA in serum from mice 8 d and 21 d 10 ** 0 102 103 104 105 (horizontal axes) after immunization with NP-CGG and presented ** NP 1.0 4 ratio 105 0.02 0.25±0.04 0.05±0.01 in relative units (RU) as serial dilution of serum relative to 10 26 titer (RU)

4 * 104 antibody end-point titers. (b) Enzyme-linked immunospot assay /NP 4 0.5 103 of the frequency of NP-specific IgM- or IgG-secreting cells 103 NP Anti-NP 102 among splenocytes from mice 7 d after immunization with 0 NP 200 0 2 3 4 5 NP-CGG. (c) Flow cytometry analysis of CD138 and B220 on IgG1 IgG2a IgG1 IgG2a 010 10 10 10 live CD11c–CD4–CD8– spleen cells (top) from mice left GL-7 unimmunized (UI) or at 7 d after immunization with NP-CGG UI NP-CGG (7 d) (NP-CGG (7 d)), with further gating of the NP+ compartment above on total plasma cells (middle) and of NP-specific GC B cells on live splenic B220+ cells (bottom). Numbers above outlined areas indicate percent CD138+B220lo–neg cells (total plasma cells; top row) or NP+GL-7+ cells (NP-specific GC B cells; bottom row); numbers above bracketed lines (middle) indicate percent NP+ plasma cells. P values (NS or *), Bcl6BTBMUT versus Bcl6+/+. (d) Titers of NP-specific IgG1 and IgG2a, measured with NP4-BSA in serum from mice 21 d after immunization with NP-CGG. (e) Ratio of the titer of IgG1 or IgG2a detected with NP4-BSA to that detected with NP26-BSA. Each symbol (a,b,d,e) represents an individual mouse; small horizontal lines indicate the mean. NS, not significant; *P < 0.05, **P < 0.01 and ***P < 0.001 (two-tailed t-test). Data are from three independent experiments with four to six mice (a,b,d,e) or are representative of two independent experiments (c, infected Bcl6+/+ or Bcl6BTBMUT mice; mean ± s.e.m. of three mice) or one experiment (c, uninfected Bcl6+/+ mice).

was 0.38% ± 0.14%, and this increased to 5.07% ± 0.85% at day 10 NP26-BSA (NP conjugated to bovine serum albumin (BSA) at a after immunization (Fig. 1d). As reported before3–5, GC B cells were molecular ratio of 26:1; Fig. 2a). In contrast, at day 21 after immuni- essentially undetectable in Bcl6–/– mice before and after immunization zation with NP-CGG, the titers of NP-specific immunoglobulin G1 (<0.2%). Likewise, Bcl6BTBMUT mice had an almost complete loss of (IgG1) and IgG2a were significantly lower in Bcl6BTBMUT mice than in GC B cells, with a frequency of 0.23% ± 0.09% before immunization Bcl6+/+ mice, with a trend toward lower titers of other immunoglobu- and 0.57% ± 0.068% at day 10 after immunization (Fig. 1d). lins (Fig. 2a). Early after immunization (at day 7 d), Bcl6BTBMUT and lo–hi lo–hi +/+ We next examined the development of TFH cells (CXCR5 PD-1 ) Bcl6 mice formed a similar number of antigen-specific IgM- or hi hi + + – and TFH cells in the GC (CXCR5 PD-1 ). The frequency of GC IgG-secreting cells (Fig. 2b) and plasma cells (NP CD138 CD11c © 2013 Nature America, Inc. All rights reserved. America, Inc. © 2013 Nature + – – neg–lo TFH cells among splenic CD4 cells was much lower in immunized CD4 CD8 B220 cells; Fig. 2c). However, at that time point, Bcl6BTBMUT mice (2% ± 0.3%) than in Bcl6+/+ mice (8.05% ± 1.9%; the abundance of antigen-specific GC B cells (NP+GL7+B220+) in lo lo BTBMUT +/+ Fig. 1e). The proportion of CXCR5 PD-1 TFH cells, a population that Bcl6 mice was less than 20% of that in immunized Bcl6

npg BTBMUT BTBMUT includes pre-GC TFH cells, was also lower in Bcl6 mice than in mice (Fig. 2c). Immunized Bcl6 mice also had much lower +/+ BTBMUT Bcl6 mice (Fig. 1e). Accordingly, Bcl6 mice had significantly titers of high-affinity IgG1 and IgG2a able to bind NP4-BSA (NP lo lo fewer total TFH cells, which included GC TFH cells and CXCR5 PD-1 conjugated to BSA at z molecular ratio of 4:1), which recognizes +/+ +/+ TFH cells, than did Bcl6 mice (Fig. 1e). We observed similar defects high-affinity immunoglobulin, than did their Bcl6 counterparts in Bcl6BTBMUT mice challenged with a second T cell–dependent anti- (Fig. 2d), indicative of impaired affinity maturation. That defect gen, NP-CGG (4-hydroxy-3-nitrophenylacetyl conjugated to chicken was also evident by calculation of the ratio of IgG1 or IgG2a titers γ-globulin; Supplementary Fig. 4b). Collectively, these results dem- detected with NP4-BSA to those detected with NP26-BSA (Fig. 2e). onstrated that BTB domain–mediated transcriptional repression was Therefore, recruitment of corepressors by the BTB domain was indis- absolutely required for the formation of GCs. pensable for Bcl-6 to drive the T cell–dependent formation of high- affinity immunoglobulins. Impaired immunoglobulin affinity maturation in Bcl6BTBMUT mice Bcl-6 is not required for the response to T cell–independent anti- B cell–intrinsic GC impairment in Bcl6BTBMUT mice gens, and accordingly, Bcl6BTBMUT mice had normal low-affinity anti- Bcl6–/– mice have cell-autonomous defects in both GC B cells and GC 4,5,13–15 BTBMUT body responses to the T cell–independent antigen of NP conjugated TFH cells . To determine whether Bcl6 mice had similar to the hydrophilic polysaccharide Ficoll (data not shown). The defects, we generated mixed–bone marrow chimeras by transplant- T cell–dependent B cell immune response triggers both an extra- ing a mixture of congenic CD45.1+ Bcl6+/+ bone marrow cells (50%) follicular response, which generates short-lived plasma cells and an together with CD45.2+ bone marrow cells from Bcl6+/+, Bcl6–/– or early wave of low-affinity antibody production, and a GC response, Bcl6BTBMUT mice (50%) into sublethally irradiated mice deficient in which gives rise to long-lived plasma or memory cells and a later wave recombination-activating gene 1 (Rag1–/– mice), which have a con- of high-affinity antibodies. Immunization with NP-CGG induced genital deficiency in mature B cells and T cells (Fig. 3a). All CD45.2+ a normal extrafollicular response in Bcl6BTBMUT mice, with the donor cells showed a normal developmental pattern for cells of the expected production of low-affinity immunoglobulin specific to T or B lineage before the GC stage (date not shown). As expected,

382 VOLUME 14 NUMBER 4 APRIL 2013 nature immunology A rt i c l e s

CD45.2+ Bcl6+/+ Bcl6BTBMUT Bcl6–/– + + + + a b + +/+ +/+ +/+ CD45.1 Bcl6 Bcl6 Bcl6 + 8 CD45.2 + ***

) CD45.1 +/+ BTBMUT 50% Bcl6+/+ BM 50% Bcl6 , Bcl6 or + + –/– + + 6 *** (CD45.1 ) Bcl6 BM (CD45.2 ) CD45.2 B cells (CD45.1+ + CD45.2+) gated T cells (CD45.1+ + CD45.2+) 4.63 0.60 0.24 4 Injection 650 rads 5 + 10 2 CD45.1 Formation of

4 GC B cells (% B220 8 weeks 10 d GC B cells 10 CD45.1+ SRBCs + and T cells 3 0 CD45.2 FH gated 10 CD45.2+ Bcl6+/+ Bcl6BTBMUT Bcl6–/– 2 Rag1–/– 10 4.07 4.26 6.53 + + + + 0 + +/+ +/+ +/+

CD38 CD45.1 Bcl6 Bcl6 Bcl6 0102 103 104 105 Fas

+/+ BTBMUT –/– c CD45.2+ Bcl6 Bcl6 Bcl6 d e + + + + + CD45.1+ Bcl6+/+ Bcl6+/+ Bcl6+/+ CD45.2 Bcl6+/+ CD45.1+ Bcl6BTBMUT –/– 6.19 6.34 0.60 12 Bcl6 1.0 + 5 * CD45.2 ) + NS 10 gated 10 ** * 0.8 8 NS * 0.6 104 6 105 NS

cells (% CD4 0.4 4 4 10 7.00 8.33 7.75 FH + CD45.1 3 NP4/NP26 ratio 3 10 0.2 10 2

gated GC T IgG1 titer (relative units) 2 10 0 200 0 0 + +/+ BTBMUT –/– +/+ BTBMUT CXCR5 CD45.2 Bcl6 Bcl6 Bcl6 NP26 NP4 Bcl6 Bcl6 0102 103 104 105 + + + + + +/+ +/+ +/+ PD-1 CD45.1 Bcl6 Bcl6 Bcl6

Figure 3 Bcl6BTBMUT mice show impaired GC responses in a B cell–autonomous manner. (a) Strategy for generating chimeras by transplantation of mixed bone marrow (BM). (b) Frequency of GC B cells among live B220+ cells from CD45.1+ and CD45.2+ donors, quantified by flow cytometry 7 d – + + + after immunization of chimeras with SRBCs (left). (c) Frequency of GC TFH cells among live B220 CD4 T cells from CD45.1 and CD45.2 donors, quantified by flow cytometry as in b (left). Numbers in outlined areas (left; b,c) indicate percent Fas+CD38lo–neg cells (GC B cells; b) or CXCR5hiPD-1hi +/+ BTBMUT cells (GC TFH cells; c). Right (b,c), summary of results at left. (d) IgG1 titers in serum from chimeras given µMT bone marrow plus Bcl6 , Bcl6 –/– or Bcl6 bone marrow (Supplementary Fig. 5a), measured with NP26-BSA or NP4-BSA 21 d after immunization of chimeras with NP-CGG. (e) Ratio of the titer of IgG1 detected with NP4-BSA to that detected with NP26-BSA (as in d). Each symbol (b–e) represents an individual mouse; small horizontal lines indicate the mean. *P < 0.05, **P < 0.01 and ***P < 0.001 (two-tailed (b,c) or one-tailed (d,e) t-test). Data are from three independent experiments with four mice per experimental condition.

+ in the chimeras, GC B cells and TFH cells derived from CD45.2 in the affinity maturation of immunoglobulins and the development Bcl6–/– donor cells were effectively absent after immunization with of functional GC B cells. SRBCs (Fig. 3b,c). In contrast, whereas CD45.2+ donor Bcl6BTBMUT © 2013 Nature America, Inc. All rights reserved. America, Inc. © 2013 Nature cells showed profound deficiency in formation of GC B cells, they Lower proliferation and survival of Bcl6BTBMUT GC B cells formed normal numbers of TFH cells (Fig. 3b,c). The lower abundance Bcl-6 suppresses checkpoints that control cellular proliferation and BTBMUT of TFH cells noted in the Bcl6 setting was thus secondary to survival, which may explain how GC B cells proliferate so rapidly npg the lack of GC B cells. and tolerate somatic hypermutation6,7,9,10. We reasoned that the ina- To determine whether impaired affinity maturation of immu- bility to achieve that proliferative and prosurvival phenotype might noglobulins in Bcl6BTBMUT mice was also intrinsic to B cells, we explain the failure of Bcl6BTBMUT mice to form functional GCs. We constructed chimeras by transferring bone marrow from µMT mice determined the rate of cell proliferation in vivo by measuring incor- (which lack B cells; 50%) along with bone marrow from Bcl6+/+, poration of the thymidine analog BrdU. Less than 1% of non-GC Bcl6BTBMUT or Bcl6–/– mice (50%) into sublethally irradiated Rag1–/– B cells incorporated BrdU in either Bcl6+/+ mice or Bcl6BTBMUT mice mice (Supplementary Fig. 5a). The µMT bone marrow provides (Supplementary Fig. 6a). Notably, whereas 32% ± 4.3% of GC B cells a source of normal T cells but no B cells; thus, all B cells in these in Bcl6+/+ mice incorporated BrdU, only 16.8% ± 3.1% of GC B cells chimeras originate from Bcl6+/+, Bcl6BTBMUT or Bcl6–/– donor bone in Bcl6BTBMUT mice were BrdU+ (Fig. 4a). Cell-cycle analysis showed marrow cells. Chimeras given a mixture containing Bcl6BTBMUT bone a greater fraction of GC B cells arrested in the G1 phase with corre- marrow formed few GC B cells and GCs after SRBC immunization, spondingly fewer in S phase in Bcl6BTBMUT mice than in Bcl6+/+ mice similar to chimeras given a mixture containing Bcl6–/– bone mar- (Fig. 4a). We next assessed the abundance of apoptotic cells by evalu- row (Supplementary Fig. 5b–d). The titers of NP-specific IgG1 ating active caspase and staining with the membrane-impermeable antibodies in chimeras given a mixture containing Bcl6BTBMUT bone DNA-intercalating dye 7-AAD. Whereas 14.7% ± 1.1% of GC B cells in marrow were 16% or 6% of those in chimeras given a mixture con- Bcl6+/+ mice were positive for activation of total caspase, this fraction taining Bcl6+/+ bone marrow after immunization with NP-CGG was higher (21% ± 1.75%) in Bcl6BTBMUT mice (Fig. 4b). Bcl6BTBMUT +/+ when NP26-BSA or NP4-BSA was used as capture antigen, respectively GC B cells also had more caspase-3 activity than did Bcl6 GC – (Fig. 3d). The ratio of titers detected with NP4-BSA to those detected B cells (7-AAD cells, 6% ± 1.1% (mutant) versus 3% ± 0.8% (wild-type); + with NP26-BSA was significantly lower in chimeras given a mixture 7-AAD cells, 10% ± 0.6% (mutant) versus 5% ± 1.2% (wild-type); containing Bcl6BTBMUT bone marrow than in those given a mixture Fig. 4b). Notably, after 3 h of incubation, Bcl6BTBMUT GC B cells still containing Bcl6+/+ bone marrow (Fig. 3e). These results demonstrated included a greater proportion of positive for active caspase-3 than did a B cell–intrinsic requirement for BTB domain–mediated repression Bcl6+/+ GC B cells (7-AAD– cells, 10% ± 1.5% (mutant) versus 6.7 ± 2.3%

nature immunology VOLUME 14 NUMBER 4 APRIL 2013 383 A rt i c l e s

Figure 4 The lateral groove of the Bcl-6 BTB a b 0 h 0 h 3 h domain is required for the proliferation and survival 600 1,000 G1: 64 600 S: 21 of GC B cells. (a) BrdU incorporation (left) and 800 * ** +/+ 400 G2-M: 7 400 * 5 ± 1.2 18 ± 2.5 Bcl6 * +/+ 14.7 ± 1.1 DNA content (right) in GC B cells from Bcl6BTBMUT 32 ± 4.3 600 Bcl6 400 +/+ 200 200 * or Bcl6 mice immunized with SRBCs, assessed by 200 3 ± 0.8 6.7 ±* 2.3 0 flow cytometry. Numbers above bracketed lines (left) 0 0 150 5 indicate percent BrdU+ (proliferating) cells; numbers 120 G1: 71 10 120 S: 10 BTBMUT 120 4 90 Bcl6 10 at top (right) indicate percent cells in phases of the 90 G2-M: 8 90 21 ± 1.8 10 ± 0.6 38 ± 3.1 BTBMUT 16.8 3.1 3 Bcl6 ± 10 60 60 60 cell cycle. (b) Active total (pan-) caspase (left) or 2 30 30 10 30 0 6 ± 1.1 10 ± 1.5 caspase-3 (right) in fresh isolated splenocytes 7-AAD Cells 0 Cells 0 0 2 3 4 5 2 3 4 5 (0 h; left and middle) or splenocytes cultured for 0102 103 104 105 50 k 100 k 010 10 10 10 010 10 10 10 3 h ex vivo (3 h; right), from mice as in a, assessed BrdU DNA content Active pan-caspase Active caspase-3

by incubation with fluorescein isothiocyanate– Chr 6 Chr 17 c 10 kb 10 kb conjugated stain for caspase activity (VAD-fmk) 5.6 3 or activated caspase-3 (DEVD-fmk), respectively, Bcl-6 Bcl-6 0 0 and flow cytometry with gating on splenic 5.1 5.4 BCOR Fas+CD38lo–negB220+ cells (GC B cells). Numbers BCOR 0 0 above bracketed lines (left) indicate percent cells 3.3 3 SMRT with active total caspase; numbers in outlined areas SMRT 0 0 (right) indicate percent live cells (7-AAD–; bottom) or CDKN1A TP53 dead cells (7-AAD+; top) with active caspase-3. Chr 6 50 kb (c) ChIP-seq analysis of the presence of Bcl-6, SMRT 2.8 Bcl-6 or BCOR on CDKN1A, TP53 and ATR in human 0 tonsil GC B cells. Below plots, coding sequences. 3.7 BCOR Chr, chromosome. *P < 0.05 and **P < 0.01 0 (two-tailed t-test). Data are from two independent 1.4 SMRT experiments with three to four mice per genotype 0 (a,b; mean ± s.d.) or are representative of three ATR independent experiments (c).

(wild-type); 7-AAD+ cells, 38 ± 3.1% (mutant) versus 18 ± 2.5% in chimeras that received a mixture containing Bcl6+/+ bone marrow (wild-type); Fig. 4b). As expected, less than 4% of freshly isolated or when either NP26-BSA or NP4-BSA was used as the capture antigen +/+ BTBMUT + ex vivo–cultured Bcl6 or Bcl6 non-GC B cells were 7-AAD (Fig. 5c). Their ratio of titers detected with NP4-BSA to those detected − or 7-AAD cells positive for active caspase-3 (Supplementary Fig. 6b). with NP26-BSA was also similar to that in chimeras that received a These data indicated that the lateral groove of the Bcl-6 BTB domain mixture containing Bcl6+/+ bone marrow (Fig. 5d). Thus, Bcl6BTBMUT was needed to facilitate the rapid proliferation and survival of GC mice formed fully functionally competent TFH cells. B cells. Along those lines, we assessed the binding profiles of Bcl-6, BCOR and SMRT in primary human GC B cells by ChIP followed by Normal differentiation of helper T cell subsets in Bcl6BTBMUT mice deep sequencing (ChIP-seq) and found that both corepressors were Bcl6–/– mice show universal skewing of T cells toward the T helper present together with Bcl-6 at the promoters of key checkpoint target type 2 (TH2) and 17 (IL-17)-producing helper T cell © 2013 Nature America, Inc. All rights reserved. America, Inc. © 2013 Nature genes, including CDKN1A, TP53 and ATR (Fig. 4c). The presence (TH17) lineages, which contributes to aberrant inflammatory sign- of such complexes was consistent with published data showing that aling and the eventual death of these mice3,4,33. That defect can be expression of those genes is induced by exposure to peptides that triggered and aggravated by immunization34. To better define the npg block the lateral groove of the BTB domain6,28,32. contribution of the lateral groove of the BTB domain to the differen- tiation of those effector helper T cell subtypes, we isolated splenocytes BTBMUT BTBMUT +/+ Bcl6 TFH cells support normal GC responses from age-matched Bcl6 and Bcl6 mice, then measured the The results presented above indicated that the Bcl-6 BTB domain was abundance of T helper type 1 (TH1), TH2 and TH17 cells. At base- BTBMUT dispensable for the formation of TFH cells (Fig. 3c). To determine line, Bcl6 mice had an abundance of TH1, TH2 and TH17 cells BTBMUT +/+ whether Bcl6 TFH cells were able to support GC functions, similar to that of their Bcl6 littermates (data not shown). We next including the affinity maturation of immunoglobulins, we generated analyzed helper T cell subtypes in mice immunized with SRBCs. As –/– –/– 3,4,33 –/– chimeras by transferring bone marrow from Tcrb Tcrd mice (which reported before , Bcl6 mice had significantly more TH2 cells and +/+ –/– lack the genes encoding the T cell antigen receptor β- and δ-chains; TH17 cells than did Bcl6 mice (TH2 cells, 11.5% ± 0.4% (Bcl6 ) +/+ BTBMUT –/– +/+ –/– 80%) along with bone marrow from Bcl6 , Bcl6 or Bcl6 versus 1% ± 0.2% (Bcl6 ); TH17 cells, 4.8% ± 0.4% (Bcl6 ) versus mice (20%) into sublethally irradiated Rag1–/– mice (Supplementary 0.6% ± 0.2% (Bcl6+/+); Fig. 6a,b). In contrast, Bcl6BTBMUT mice had –/– –/– +/+ Fig. 7a). Tcrb Tcrd bone marrow cells provide a source of normal an abundance of TH2 and TH17 cells similar to that of Bcl6 mice B cells but no T cells; thus, the chimeras would be able to form functional (0.98 ± 0.3% (Bcl6BTBMUT) versus 1 ± 0.2% (Bcl6+/+); 0.62 ± 0.3% GCs only if the other donor bone marrow provided a source of normal (Bcl6BTBMUT) versus 0.6 ± 0.2% (Bcl6+/+); Fig. 6a,b). Thus, the lateral TFH cells. After immunization with SRBCs, chimeras that received a groove of the BTB domain was not involved in Bcl-6-mediated sup- BTBMUT mixture containing Bcl6 bone marrow formed an abundance of pression of TH2 or TH17 differentiation in vivo. GC B cells and normal GCs in spleens equal to that formed by chimeras that received a mixture containing Bcl6+/+ bone marrow, but those that No Bcl6–/– inflammatory phenotype in Bcl6BTBMUT mice received a mixture containing Bcl6–/– bone marrow did not (Fig. 5a,b Bcl6–/– mice are born at lower than expected frequencies, are runted and Supplementary Fig. 7b). Moreover, titers of NP-specific IgG1 and develop a severe TH2-type inflammatory syndrome that involves antibodies in NP-CGG-immunized chimeras that received a mixture many organs and frequently includes the lungs and spleen3,4. However, containing Bcl6BTBMUT bone marrow were indistinguishable from those we observed that Bcl6BTBMUT mice were born at the expected

384 VOLUME 14 NUMBER 4 APRIL 2013 nature immunology A rt i c l e s

+/+ Bcl6+/+ Bcl6BTBMUT Bcl6–/– Bcl6 a b +/+ BTBMUT –/– c BTBMUT Bcl6 Bcl6 Bcl6 Bcl6 5 NS 10 5 Bcl6–/– 10 NS 104

103 4 102 10 * 0 3.20 2.92 0.34 CD38 0102 103 104 105 * Fas 103 lgG1 titer (relative units) 500 NP26 NP4

d 1.0 Figure 5 Normal GC responses in chimeras given Tcrb–/–Tcrd–/– and Bcl6BTBMUT bone marrow. (a) Flow cytometry of GC NS –/– –/– +/+ BTBMUT –/– B cells from chimeras given Tcrb Tcrd bone marrow plus Bcl6 , Bcl6 or Bcl6 bone marrow (Supplementary 0.8 Fig. 6a), assessed 10 d after immunization of chimeras with SRBCs, gated on live splenic B220+ lymphocytes. Numbers in outlined areas indicate percent Fas+CD38lo–neg cells (GC B cells). (b) Staining of spleen sections from mice in a with 0.6 peanut agglutinin. Scale bars, 200 µm. (c) IgG1 titers in serum from chimeras as in a, 21 d after immunization with 0.4 NP-CGG, measured with NP26-BSA and NP4-BSA. (d) Ratio of the titer of IgG1 or IgG2a detected with NP4-BSA to NP4/NP26 ratio 0.2 that detected with NP26-BSA. Each symbol represents an individual mouse; small horizontal lines indicate the mean. * < 0.05 (one-tailed -test). Data are representative of three independent experiments (a,b) or two independent P t 0 experiments with four mice (c,d). Bcl6+/+ Bcl6BTBMUT

Mendelian frequencies, were developmentally indistinguishable Bcl6BTBMUT mice prompted us to measure chemokine expression and from their Bcl6+/+ littermates and had normal body weights (Fig. 7a). the proliferation of macrophages isolated from these mice. We first By 6 weeks of age, more than half of the Bcl6–/– mice had died, measured the abundance of mRNA from genes encoding molecules whereas all Bcl6BTBMUT mice remained healthy even after 1 year of involved in the inflammatory response (Ccl2, Ccl3, Ccl6, Il1a, Ccl7 monitoring. Bcl6–/– mice had the expected infiltrates of inflammatory and Ccnd2) in resting and lipopolysaccharide (LPS)-stimulated mac- cells in their lungs, as well as multinodular lesions characterized by rophages from Bcl6+/+, Bcl6BTBMUT and Bcl6–/– mice, by quantitative the infiltration of eosinophils into the spleen, whereas histological PCR. Most of these genes had 20-fold higher expression in resting studies of Bcl6BTBMUT organs (including lungs, spleen, heart, liver, Bcl6–/–macrophages than in resting Bcl6+/+ cells and were further thymus, kidney and intestine) showed no such infiltrates (Fig. 7b induced by treatment with LPS (Fig. 7c). In contrast, Bcl6BTBMUT and Supplementary Fig. 8a). We confirmed that the mice had that macrophages showed no more than 1.4- to 4-fold higher basal expres- phenotype in two independent founder lines with different genetic sion of these genes relative to their expression in Bcl6+/+ cells, and that backgrounds (data not shown). limited higher expression was not further enhanced by LPS treatment (Fig. 7c). As expected, Bcl6–/– macrophages divided faster than Bcl6+/+ Minimal effect on in Bcl6BTBMUT macrophages macrophages did, whereas Bcl6BTBMUT macrophages had a prolifera- Bcl6–/– macrophages have substantial upregulation of the expression tion rate very similar to that of Bcl6+/+ macrophages (Fig. 7d). of inflammatory chemokines such as CCL2, CCL3, CCL6, CCL7 and To determine whether other Bcl-6 domains contribute to the IL-1a, which are believed to be critical mediators of the lethal inflam- repression of genes encoding inflammatory molecules to a greater © 2013 Nature America, Inc. All rights reserved. America, Inc. © 2013 Nature matory response in Bcl6–/– mice18,19. Bcl-6 also suppresses extent than the BTB domain does, we transduced Bcl6–/– macro- proliferation through repression of CCND2, a critical driver of the phages with retrovirus expressing wild-type Bcl6 or Bcl6 with vari- cell cycle35. The unexpected lack of an inflammatory phenotype in ous mutations. Bcl6–/– macrophages transduced to express wild-type npg Bcl6, Bcl6 with point mutations that result in the substitutions BN21K and H116A (as in +/+ BTBMUT –/– a Bcl6 Bcl6 Bcl6 Bcl6BTBMUT mice), Bcl6 with mutations that 5 10 result in inactivation of the RD2 repression 4 10 5.62 4.60 3.67 36 +/+ domain or Bcl6 with both sets of mutations 3 Bcl6 10 b BTBMUT Bcl6 retained the ability to suppress the expression 102 0 Bcl6–/– IFN- γ 15 0 102 103 104 105 * Figure 6 Normal TH2 and TH17 differentiation CD4 12 BTBMUT T cells) in mice. (a) Flow cytometry of + Bcl6 NS 9 splenic CD4+ T cells stimulated for 5 h 5 10 * with the phorbol ester PMA (20 ng/ml) and 104 6 1.51 1.97 12.50 ionomycin (1 µg/ml) in the presence of the 3 10 3 NS NS protein-transport inhibitor GolgiPlug (1 µg/ml), 102 followed by staining for CD4 and IFN-γ, IL-4

0 Frequency (% CD4 0 IL-4 or IL-17. Numbers adjacent to outlined areas 2 3 4 5 IFN-γ IL-4 IL-17 010 10 10 10 indicate percent cells in each. (b) Frequency CD4 + + of TH1 (IFN-γ ) cells, TH2 (IL-4 ) cells and + + 105 TH17 (IL-17 ) cells among CD4 T cells, 104 0.49 0.59 4.10 based on flow cytometry as in a. Each 103 symbol represents an individual mouse; small horizontal lines indicate the mean (± s.d. of 102 0 two to three mice). *P < 0.001 (two-tailed IL-17 0102 103 104 105 t test). Data are representative of two CD4 independent experiments.

nature immunology VOLUME 14 NUMBER 4 APRIL 2013 385 A rt i c l e s

BTBMUT Figure 7 Bcl6 mice do not develop TH2 a b Bcl6+/+ Bcl6BTBMUT Bcl6–/– inflammatory disease and have nearly normal inflammation-related gene expression in 30 ** NS macrophages. (a) Body weight of 8-week-old Lung Bcl6+/+, Bcl6BTBMUT and Bcl6–/– mice. 20 (b) Hematoxylin and eosin staining of lung and spleen sections from mice as in a. Dotted

Weight (g) 10 circle (bottom right) outlines multinodular lesion enlarged at right; arrow (inset, far right) Spleen indicates infiltration by eosinophil. Scale bars, 0 +/+ –/– 200 µm (main images); original magnification BTBMUT (inset), ×40. (c) Expression of mRNA (genes, Bcl6 Bcl6 +/+ BTBMUT Bcl6 horizontal axes) in resting Bcl6 , Bcl6 Resting LPS and –/– bone marrow–derived macrophages c d Bcl6 +/+ 26,200 +/+ Bcl6+/+ Bcl6 2,718 Bcl6 (BMDMs; left) or those treated with LPS 400 256 3,000 BTBMUT 20 BTBMUT Bcl6BTBMUT Bcl6 Bcl6 –/– –/– (5 µg/ml; right); results are normalized to 300 Bcl6–/– 2,000 Bcl6 Bcl6

) 15 those of the housekeeping gene Hprt and are 200 131 1,043 5 1,000 604 presented relative to those of resting Bcl6+/+ 100 37 31 21 108 10 cells. Numbers above bars indicate exact values 15 80 NS 15 80 +/+ BTBMUT mRNA (fold) mRNA (fold) for each. (d) Growth of Bcl6 , Bcl6 and 10 60 Cells (1 × 10 5 –/– BMDMs. (e) Expression of mRNA (genes, 4.6 40 Bcl6 5 20 11 horizontal axes) in sorted green fluorescent 0 0 0 + –/– 0 63 protein–positive (GFP ) populations from Bcl6 Ccl2 Ccl3 Ccl6 Il1a Ccl7 Ccl2 Ccl3 Ccl6 Il1a Ccl7 BMDMs infected with a bicistronic retrovirus Ccnd2 Ccnd2 Time (d) expressing GFP alone or GFP and wild-type Bcl6 GFP Bcl-6(QQYQ) e 14 f 1.25 (Bcl-6), mutated Bcl6 encoding Bcl-6 with the Bcl-6 Bcl-6(N21K,H116A,QQYQ) * ** Mock 12 substitutions BN21K and H116A (Bcl-6(N21K, Bcl-6(N21K,H116A) Bcl-6(ZNFMUT) 1.00 ** Bcl-6 H116A) or an inactivated RD2 (Bcl-6(QQYQ))36, 10 * Bcl-6(N21K, 0.75 H116A,QQYQ) with both sets of mutations (Bcl-6(N21K, 8 Bcl6 Bcl-6(ZNFMUT) 6 H116A,QQYQ)) or mutated Bcl6 encoding a STAT5 0.50 mutant zinc finger (Bcl-6(ZNFMUT)); results 4 enrichment (%) 0.25 are normalized to Hprt and are presented 2 mRNA repression (fold) relative to those of GFP. (f) Binding of STAT5 0 0 to Ccl2 and Ccnd2 in puromycin-resistant Il1a –/– Ccl2 Ccl3 Ccl6 Ccl7 Ccl2 Bcl6 BMDMs mock infected (Mock) or Ccnd2 Ccnd2 Gapdh infected with retroviruses as described in e (key); results are presented relative to input. *P < 0.05 and **P < 0.01 (two-tailed t test). Data are representative of two experiments with six mice (a) or three mice (c), at least three independent experiments (d–f; mean and s.d. in a,c,d–f) or three independent experiments (b).

of genes encoding chemokines and inflammatory molecules (Fig. 7e). DISCUSSION In contrast, those transduced to express Bcl6 with a mutation that Given that Bcl-6 and corepressors of its BTB domain are expressed © 2013 Nature America, Inc. All rights reserved. America, Inc. © 2013 Nature affects the third C2H2 zinc finger of Bcl-6 that abolishes DNA binding together in most Bcl-6-expressing tissues, the biochemical function of without affecting localization to the nucleus37 were completely unable Bcl-6 would be expected to be similar among cell types, and its tissue- to repress those target genes (Fig. 7e). Hence, whereas DNA binding specific function would be expected to be perhaps mediated by binding npg was required for the repression of chemokine expression, the lateral to distinct sets of target genes. Indeed, genome-localization studies of groove of the BTB domain was dispensable for that function, consist- Bcl-6 in various cell types have suggested that its transcriptional targets ent with the lack of inflammatory phenotype of Bcl6BTBMUT mice. are partially cell type specific19,28,29. However, evidence of a more pro- ChIP-seq analysis of Bcl-6 in macrophages has shown that Bcl-6- found, biochemical level of functional diversification has been hinted binding sites show considerable enrichment for STAT-binding motifs, at by studies of cell-penetrating peptides designed to dissociate the including those in the chemokine-encoding genes we assessed here19. BTB domain of Bcl-6 from its corepressors39,40. The administration of Bcl-6 has been proposed to potentially compete with STAT proteins such peptides to mice bearing human lymphoma xenografts induces for binding to certain target genes30,31,38. Loss of the binding of Bcl-6 growth arrest and apoptosis in lymphoma cells but does not induce the to DNA might thus enhance the recruitment of STAT proteins to those Bcl6–/– inflammatory phenotype39,40. Unfortunately, the interpretation inflammation-related loci. Here we assessed STAT5 as an example. of such peptide studies has been limited by the unknown kinetics and Bcl6+/+ and Bcl6–/– primary mouse macrophages have a similar abun- degree of inhibition of Bcl-6 in various cell types, as well as a relatively dance of active STAT5 (ref. 35). By quantitative ChIP assay, we observed time-limited exposure. Our introduction of point mutations in the reciprocal changes in the binding of STAT5 and Bcl-6 at the promoters native Bcl6 locus in mice afforded constitutive loss of the repressor of Ccl2 and Ccnd2 in Bcl6+/+ versus Bcl6–/– primary mouse macro­ function of the BTB domain in all tissues while preserving proper tim- phages, with substantially enhanced enrichment for STAT5 in the ing and amount of expression, which allowed us to gain critical insights absence of Bcl-6 (Supplementary Fig. 8b,c). More notably, expression into the function of this unique biochemical mechanism of Bcl-6. of wild-type Bcl6 plus Bcl6 with the mutations resulting in the BN21K Bcl6BTBMUT mice had a normal early extrafollicular response but an and H116A substitutions and inactivation of the RD2 repression impaired late GC response after immunization with a T cell–dependent domain (as described above) in Bcl6–/– primary mouse macrophages antigen. Notably, loss of the repression function of the BTB domain impaired enrichment for STAT5 at both of those chemokine-encoding manifested as failure of GC B cells to proliferate and survive, even in loci, but expression of wild-type Bcl6 plus Bcl6 with the mutation affect- the presence of wild-type TFH cells and a wild-type microenviron- ing its third C2H2 zinc finger (described above) did not (Fig. 7f). ment. Hence, the function of the BTB domain of Bcl-6 was specifically

386 VOLUME 14 NUMBER 4 APRIL 2013 nature immunology A rt i c l e s

to enable the proliferative and DNA damage–tolerant phenotype of Our data have shown that such side effects would be unlikely to GC B cells, possibly by repressing ATR, TP53 and CDKN1A. Blockade occur, consistent with reported toxicity studies of such inhibitors in of the BTB domain with peptide inhibitors in primary GC B cells mice39. In summary, by constructing a knock-in mutation that results and DLBCL cells results in derepression of those genes6,32. One in constitutive disruption of the binding of corepressors to the BTB implication of these results is that the same biochemical function domain, we were able to demonstrate that Bcl-6 mediated immuno- through which Bcl-6 specifically mediates the GC B cell phenotype logical lineage-specific effects through various protein interactions also drives the survival and proliferation of DLBCL cells. Targeting and thus provide a new paradigm for the functional diversification the BTB domain of Bcl-6 with peptides or small-molecule inhibitors of transcription factors. kills DLBCL cell lines in vitro and in vivo, as well as primary human DLBCL cells ex vivo39–41. Thus, DLBCLs are essentially driven by a Methods normal transcriptional mechanism derived from their cell of origin. Methods and any associated references are available in the online Similar to Bcl6–/– mice, Bcl6BTBMUT mice had fewer CXCR5+PD-1+ version of the paper. TFH cells and GC TFH cells. However, in contrast to the T cell devel- opment in Bcl6–/– mice, Bcl6BTBMUT mice generated perfectly func- Accession codes. GEO: microarray data, ChIP-seq data, GSE43350. tional TFH cells after immunization in the presence of normal cognate B cells. As cognate B cells are absolutely required for the differentia- Note: Supplementary information is available in the online version of the paper. 42–44 tion and maintenance of GC TFH cells , the defect in GC TFH cells Acknowledgments in Bcl6BTBMUT mice was probably secondary to the defect in the B cell We thank H. Ye (Albert Einstein College of Medicine) for Bcl6–/– mice; compartment. The finding that the deficit in CXCR5+PD-1+ T cells W. Pear (University of Pennsylvania) for the MIGR1 expression vector; and D. FH BTBMUT was less pronounced was probably related to the preservation of the Wen and S. Rafii for assistance in generating Bcl6 mice. Supported by the US National Cancer Institute (R01 104348 to A.M.), the Burroughs Wellcome early, extrafollicular phase of the immune response that we observed Foundation and Chemotherapy Foundation (A.M.) and the March of Dimes in Bcl6BTBMUT mice. The mechanisms underlying the deficiencies in (A.M.), and facilitated by the Sackler Center for Biomedical and Physical Sciences BTBMUT –/– the TFH cell phenotype in Bcl6 and Bcl6 mice are completely at Weill Cornell Medical College. different. Further research is thus needed to explain the biochemical AUTHOR CONTRIBUTIONS mechanism through which Bcl-6 mediates the TFH cell phenotype. C.H. designed and did most of the experiments; K.H. did and analyzed ChIP-seq Perhaps different sets of corepressors that bind to different sites on experiments; and A.M. conceived of the project and wrote the manuscript. Bcl-6 are involved. Another salient finding of our phenotypic analysis of Bcl6BTBMUT COMPETING FINANCIAL INTERESTS The authors declare no competing financial interests. mice was the absence of the lethal inflammatory disease of Bcl6–/– mice, which indicated that the lateral groove of the BTB domain was Reprints and permissions information is available online at http://www.nature.com/ dispensable for the functions of Bcl-6 in the innate immune system. reprints/index.html. In macrophages, Bcl-6 attenuates inflammatory responses through the 18,19,35 repression of chemokines and NF-κB target genes . Enhanced 1. Ye, B.H. et al. Alterations of a zinc finger-encoding gene, BCL-6, in diffuse large- inflammatory signaling in Bcl6–/– macrophages also deregulates cell lymphoma. Science 262, 747–750 (1993). T cell development, which results in skewing toward T 2 and T 17 2. Cattoretti, G. et al. BCL-6 protein is expressed in germinal-center B cells. Blood H H 86, 45–53 (1995). 33 differentiation and eventually forms a vicious circle that leads to 3. Dent, A.L., Shaffer, A.L., Yu, X., Allman, D. & Staudt, L.M. Control of inflammation, © 2013 Nature America, Inc. All rights reserved. America, Inc. © 2013 Nature destructive infiltration of various tissues by T cells. Loss of Bcl6 is expression, and germinal center formation by BCL-6. Science 276, also reported to confer an atherogenic and xanthomatous tendinitis 589–592 (1997). 4. Ye, B.H. et al. The BCL-6 proto-oncogene controls germinal-centre formation and 45 phenotype reminiscent of human familial hypercholesterolemia. Th2-type inflammation. Nat. Genet. 16, 161–170 (1997). npg We did not observe any of those phenotypes in Bcl6BTBMUT mice. 5. Fukuda, T. et al. Disruption of the Bcl6 gene results in an impaired germinal center formation. J. Exp. Med. 186, 439–448 (1997). The expression of inflammatory was at best partially 6. Ranuncolo, S.M. et al. Bcl-6 mediates the germinal center B cell phenotype and deregulated and proliferation was normal in Bcl6BTBMUTmacrophages lymphomagenesis through transcriptional repression of the DNA-damage sensor ATR. relative to that in Bcl6–/– macrophages. It was the DNA-binding Nat. Immunol. 8, 705–714 (2007). 7. Ranuncolo, S.M., Polo, J.M. & Melnick, A. BCL6 represses CHEK1 and suppresses domain and not the BTB domain of Bcl-6 that was most critical in DNA damage pathways in normal and malignant B-cells. Blood Cells Mol. Dis. 41, the repression of chemokine expression in Bcl6–/– macrophages. The 95–99 (2008). explanation for that finding may be found in how transcriptional 8. Cerchietti, L.C. et al. BCL6 repression of EP300 in human diffuse large B cell lymphoma cells provides a basis for rational combinatorial therapy. J. Clin. Invest. 120, repressors and activators can compete for promoter occupancy by 4569–4582 (2010). mutually exclusive binding to overlapping DNA elements46. Indeed 9. Phan, R.T. & Dalla-Favera, R. The BCL6 proto-oncogene suppresses p53 expression in germinal-centre B cells. Nature 432, 635–639 (2004). Bcl-6 and STAT proteins bind to overlapping consensus binding 10. Phan, R.T., Saito, M., Basso, K., Niu, H. & Dalla-Favera, R. BCL6 interacts with sites19,25,28,29. Our results presented here showed reciprocal binding the transcription factor Miz-1 to suppress the cyclin-dependent kinase inhibitor of Bcl-6 and STAT5 to key macrophage genes encoding inflamma- p21 and cell cycle arrest in germinal center B cells. Nat. Immunol. 6, 1054–1060 (2005). tory molecules. This, passive DNA-binding competition between 11. Crotty, S. Follicular helper CD4 T cells (TFH). Annu. Rev. Immunol. 29, 621–663 Bcl-6 and STAT proteins, rather than the interaction of Bcl-6 with (2011). its corepressors, may be the dominant biochemical function of Bcl-6 12. McHeyzer-Williams, L.J., Pelletier, N., Mark, L., Fazilleau, N. & McHeyzer-Williams, M.G. Follicular helper T cells as cognate regulators of B cell immunity. Curr. Opin. in innate immunity. Immunol. 21, 266–273 (2009). Finally, our study has implications for the clinical use of inhibi- 13. Yu, D. et al. The transcriptional repressor Bcl-6 directs T follicular helper cell lineage commitment. Immunity 31, 457–468 (2009). tors of Bcl-6 designed to disrupt the binding of corepressors to its 14. Nurieva, R.I. et al. Bcl6 mediates the development of T follicular helper cells. BTB domain39,40, which are now being ‘translated’ for use in the Science 325, 1001–1005 (2009). treatment of patients with Bcl-6-dependent tumors. The potential 15. Johnston, R.J. et al. Bcl6 and Blimp-1 are reciprocal and antagonistic regulators of T follicular helper cell differentiation. Science 325, 1006–1010 (2009). for such drugs to cause systemic inflammation and atherosclerosis 16. Kroenke, M.A. et al. Bcl6 and Maf cooperate to instruct human follicular helper is a potential concern for humans treated with such compounds. CD4 T cell differentiation. J. Immunol. 188, 3734–3744 (2012).

nature immunology VOLUME 14 NUMBER 4 APRIL 2013 387 A rt i c l e s

17. Crotty, S., Johnston, R.J. & Schoenberger, S.P. Effectors and memories: Bcl-6 and 32. Cerchietti, L.C. et al. Sequential transcription factor targeting for diffuse large B-cell Blimp-1 in T and B lymphocyte differentiation. Nat. Immunol. 11, 114–120 lymphomas. Cancer Res. 68, 3361–3369 (2008). (2010). 33. Mondal, A., Sawant, D. & Dent, A.L. Transcriptional repressor BCL6 controls Th17 18. Toney, L.M. et al. BCL-6 regulates chemokine gene transcription in macrophages. responses by controlling gene expression in both T cells and macrophages. Nat. Immunol. 1, 214–220 (2000). J. Immunol. 184, 4123–4132 (2010). 19. Barish, G.D. et al. Bcl-6 and NF-κB cistromes mediate opposing regulation of the 34. Dent, A.L., Hu-Li, J., Paul, W.E. & Staudt, L.M. T helper type 2 inflammatory innate immune response. Genes Dev. 24, 2760–2765 (2010). disease in the absence of and transcription factor STAT6. Proc. Natl. 20. Dhordain, P. et al. Corepressor SMRT binds the BTB/POZ repressing domain of Acad. Sci. USA 95, 13823–13828 (1998). the LAZ3/BCL6 oncoprotein. Proc. Natl. Acad. Sci. USA 94, 10762–10767 35. Yu, R.Y. et al. BCL-6 negatively regulates macrophage proliferation by suppressing (1997). autocrine IL-6 production. Blood 105, 1777–1784 (2005). 21. Huynh, K.D. & Bardwell, V.J. The BCL-6 POZ domain and other POZ domains interact 36. Bereshchenko, O.R., Gu, W. & Dalla-Favera, R. Acetylation inactivates the with the co-repressors N-CoR and SMRT. Oncogene 17, 2473–2484 (1998). transcriptional repressor BCL6. Nat. Genet. 32, 606–613 (2002). 22. Huynh, K.D., Fischle, W., Verdin, E. & Bardwell, V.J. BCoR, a novel corepressor 37. Mascle, X., Albagli, O. & Lemercier, C. Point mutations in BCL6 DNA-binding involved in BCL-6 repression. Genes Dev. 14, 1810–1823 (2000). domain reveal distinct roles for the six zinc fingers. Biochem. Biophys. Res. 23. Ahmad, K.F. et al. Mechanism of SMRT corepressor recruitment by the BCL6 BTB Commun. 300, 391–396 (2003). domain. Mol. Cell 12, 1551–1564 (2003). 38. Fernández de Mattos, S. et al. FoxO3a and BCR-ABL regulate cyclin D2 transcription 24. Ghetu, A.F. et al. Structure of a BCOR corepressor peptide in complex with the through a STAT5/BCL6-dependent mechanism. Mol. Cell Biol. 24, 10058–10071 BCL6 BTB domain dimer. Mol. Cell 29, 384–391 (2008). (2004). 25. Chang, C.C., Ye, B.H., Chaganti, R.S. & Dalla-Favera, R. BCL-6, a POZ/zinc-finger 39. Cerchietti, L.C. et al. A peptomimetic inhibitor of BCL6 with potent antilymphoma protein, is a sequence-specific transcriptional repressor. Proc. Natl. Acad. Sci. USA effects in vitro and in vivo. Blood 113, 3397–3405 (2009). 93, 6947–6952 (1996). 40. Polo, J.M. et al. Specific peptide interference reveals BCL6 transcriptional and oncogenic 26. Fujita, N. et al. MTA3 and the Mi-2/NuRD complex regulate cell fate during mechanisms in B-cell lymphoma cells. Nat. Med. 10, 1329–1335 (2004). B lymphocyte differentiation. Cell 119, 75–86 (2004). 41. Cerchietti, L.C. et al. A small-molecule inhibitor of BCL6 kills DLBCL cells in vitro 27. Mendez, L.M. et al. CtBP is an essential corepressor for BCL6 autoregulation. and in vivo. Cancer Cell 17, 400–411 (2010). Mol. Cell Biol. 28, 2175–2186 (2008). 42. Kerfoot, S.M. et al. Germinal center B cell and T follicular helper cell development 28. Ci, W. et al. The BCL6 transcriptional program features repression of multiple oncogenes initiates in the interfollicular zone. Immunity 34, 947–960 (2011). in primary B cells and is deregulated in DLBCL. Blood 113, 5536–5548 (2009). 43. Goteri, G. et al. Comparison of germinal center markers CD10, BCL6 and human 29. Basso, K. et al. Integrated biochemical and computational approach identifies BCL6 germinal center-associated lymphoma (HGAL) in follicular lymphomas. Diagn. direct target genes controlling multiple pathways in normal germinal center B cells. Pathol. 6, 97 (2011). Blood 115, 975–984 (2010). 44. Choi, Y.S. et al. ICOS receptor instructs T follicular helper cell versus effector cell 30. Reljic, R., Wagner, S.D., Peakman, L.J. & Fearon, D.T. Suppression of signal differentiation via induction of the transcriptional repressor Bcl6. Immunity 34, transducer and activator of transcription 3-dependent B lymphocyte terminal 932–946 (2011). differentiation by BCL-6. J. Exp. Med. 192, 1841–1848 (2000). 45. Barish, G.D. et al. The Bcl6-SMRT/NCoR cistrome represses inflammation to 31. Harris, M.B. et al. Transcriptional repression of Stat6-dependent interleukin- attenuate atherosclerosis. Cell Metab. 15, 554–562 (2012). 4-induced genes by BCL-6: specific regulation of iepsilon transcription and 46. Hermsen, R., Tans, S. & ten Wolde, P.R. Transcriptional regulation by competing immunoglobulin E switching. Mol. Cell Biol. 19, 7264–7275 (1999). transcription factor modules. PLoS Comput. Biol. 2, e164 (2006). © 2013 Nature America, Inc. All rights reserved. America, Inc. © 2013 Nature npg

388 VOLUME 14 NUMBER 4 APRIL 2013 nature immunology ONLINE METHODS Fix/Perm buffer and intracellular staining with fluorescein isothiocyanate–con- Generation of Bcl6BTBMUT mice. The mutations encoding N21K and jugated anti-IFN-γ, phycoerythrin-conjugated anti-IL-4 or Alexa Fluor 488–con- H116A were introduced into exon 3 and exon 4 of a Bcl6 bacterial artificial jugated antibody to mouse IL-17A in Perm/Wash buffer (BD Pharmingen). ­chromosome (RP24-371N16; Children’s Hospital Oakland Research Center) by a GalK positive-counterselection strategy. A neomycin-resistance (neor) BrdU detection, and cell-cycle and apoptosis assays. For BrdU labeling, mice cassette flanked by two loxP sites was inserted into the Bcl6 intron 3,800 base were given intravenously injection of 2 mg BrdU (Sigma-Aldrich) 2 h before pairs upstream to the residue encoding His116. Finally, 2.0-kb diphtheria toxin being killed. Splenocytes were prepared and stained with phycoerythrin- α-chain cassette replaced the 1.0-kilobase genomic fragment 2.0 kilobases conjugated antibody to mouse B220, phycoerythrin-indotricarbocyanine– downstream of the 3′ loxP site (Supplementary Fig. 1a). The targeting conjugated anti-Fas and allophycocyanin-conjugated antibody to mouse CD38. vectors were made linear and were transfected by electroporation into 129 × Then, BrdU+ cells were detected with a BrdU Flow kit according to the manu- C57Bl/6 embryonic stem cells. Embryonic stem cell clones were screened facturer’s protocol (BD Pharmingen). For cell-cycle analysis, splenocytes were for integration by PCR (primers F1 and R1, Supplementary Table 1). Two stained with phycoerythrin-conjugated antibody to mouse B220, phycoeryth- clones confirmed to contain the homologous-targeted mutation were injected rin-indotricarbocyanine–conjugated anti-Fas and allophycocyanin-conjugated into C57Bl/6 blastocysts, and those blastocytes were implanted intp pseudo- antibody to mouse CD38 and were fixed for 45 min in fixation/permeable pregnant female mice. Germline transmission resulted in the generation of buffer (FoxP3 staining set; eBioscience), followed by staining with DAPI Bcl6+/BTBMUT mice with the neor cassette in the knocked-in allele. Those mice (4,6-diamidino-2-phenylindole). Phase distribution was analyzed automati- were mated with C57Bl/6 EIIa-Cre mice (with transgenic expression of Cre cally with the Dean-Jett-Fox model (Flowjo). For detection of apoptosis in situ, recombinase controlled by the EIIa promoter, which targets Cre expression fresh isolated splenocytes or splenocytes incubated for 3 h were maintained to the early mouse embryo, to remove the loxP-flanked neor cassette); this for 1 h at 37 °C in FITC-VAD-fmk (fluorescein isothiocyanate–Val-Ala- generated Bcl6+/BTBMUT mice. Those mice were further bred to C57Bl/6 mice Asp–fluoromethylketone) or FITC-DEVD-fmk (fluorescein isothiocyanate– for at least five generations and were intercrossed to obtain homozygous mice Asp-Glu-Val-Asp–fluoromethylketone; BioVision) in RPMI medium. Cells for most experiments. The genotyping primers are shown in Supplementary were washed according to the manufacturer’s protocol and then were labeled Table 1. For study of the development of inflammatory diseases in Bcl6BTBMUT with phycoerythrin-conjugated antibody to mouse B220, phycoerythrin- mice, Bcl6+/BTBMUT mice were bred with Sv129 mice (Jackson Laboratory) for indotricarbocyanine–conjugated anti-Fas and allophycocyanin-conjugated three generations and were intercrossed to obtain homozygous mice. antibody to mouse CD38 for surface phenotyping.

Mice and mixed–bone marrow chimera studies. Bcl6–/– mice were provided ChIP with quantitative PCR. Cells were crosslinked with 1% formaldehyde by H. Ye. For the generation of mixed bone marrow chimera, 4 × 106 cells and neutralized with 0.125 M glycine. Cell lysates were sonicated to a length of from a mixture of B6.SJL bone marrow (CD45.1+; Jackson Laboratory) plus 300–500 base pairs, and proteins were immunoprecipitated with antibody to Bcl- Bcl6+/+, Bcl6–/– or Bcl6BTBMUT bone marrow, at a ratio of 1:1, or a mixture of 6 (anti-Bcl-6; N3; Santa Cruz), anti-SMRT (Millipore), anti-STAT5 (sc-835; Santa µMT bone marrow (Jackson laboratory) plus Bcl6+/+, Bcl6–/– or Bcl6BTBMUT Cruz) or IgG (ab-37415; abcam) as a control. After complete washing, immuno- bone marrow, at a ratio of 1:1, or a mixture of Tcrb–/–Tcrd–/– bone marrow precipitated DNA was eluted in elution buffer and reverse-crosslinked overnight (Jackson laboratory) plus Bcl6+/+, Bcl6–/– or Bcl6BTBMUT bone marrow, a ratio at 65 °C. DNA was purified and quantified by real-time PCR (primer sequences, of 4:1, were transferred intravenously into sublethally irradiated Rag1–/– mice Supplementary Table 2). Enrichment was calculated relative to input. (Jackson Laboratory). Then, 8 weeks later, the recipient mice were immunized and killed for further analysis of GC formation and antibody production. Mice Immunization, enzyme-linked immunosorbent assay and enzyme-linked were housed in the specific pathogen–free animal facility at Weill Cornell immunospot assay. For analysis of the formation of GCs, mice were immu- Medical College and animal experiments were done with protocols approved nized intraperitoneally for the appropriate number of days with SRBCs by the Institutional Animal Care and Use Committee. (1 × 108 cells per mouse) or NP21-CGG (Biosearch Technologies) in Imject

© 2013 Nature America, Inc. All rights reserved. America, Inc. © 2013 Nature alum (Pierce). For analysis of T cell–independent antibody production, Antibodies and flow cytometry. Phycoerythrin-, fluorescein isothiocyanate– mice were immunized intraperitoneally with 100 µg NP26-Ficoll (Biosearch or allophycocyanin-conjugated antibody to mouse B220 (RA3-6B2), allophyco- Technologies) and analyzed for 8 d. For analysis of T cell–dependent antibody cyanin-conjugated antibody to mouse IgM (II/41), phycoerythrin-conjugated production, mice were immunized intraperitoneally with 100 µg NP21-CGG. npg antibody to mouse IgD (11-26c.2a), phycoerythrin-indotricarbocyanine– On days 8 and 21 after immunization, serum was collected and titers of iso- conjugated antibody to Fas (anti-Fas; jo2), phycoerythrin-indotricarbocyanine– type-specific antibodies to NP were measured in plates coated with NP26-BSA conjugated antibody to mouse CXCR5 (2G8), phycoerythrin-conjugated anti- or NP4-BSA with the SBA Clonotyping System, according to the manufacturer’s IL-4 (11b11), fluorescein isothiocyanate–conjugated anti-GL-7 (GL-7), Pacific protocol (Southern Biotech). Titers are presented as the greatest serum dilu- blue–conjugated anti-CD138- (181-2), allophycocyanin-indotricarbocyanine– tion that provided an average absorbance exceeding 1.5-fold above the aver- conjugated anti-CD11c (N418) and fluorescein isothiocyanate–conjugated age background absorbance at 405 nm. For enzyme-linked immunospot anti-IFN-γ (XMG1.2) were from BD Bioscience. Fluorescein isothiocyanate– assay, spleen cells were incubated for 20 h at 37 °C on NP26-BSA-coated conjugated anti-CD45.2 FITC (104), Phycoerythrin- or fluorescein isothiocyanate– 96-well MultiScreen-HA filter plates (Millipore). Spots were visualized with conjugated antibody to mouse PD-1 (J43), Alexa Fluor 488–conjugated antibody goat antibody to mouse IgG (1034-05) or IgM (102105) conjugated to horse- to mouse CD1d (1B1), fluorescein isothiocyanate–conjugated antibody to mouse radish peroxidase (Southern Biotechnology), and color was visualized by the CD21 (8D9), allophycocyanin-conjugated antibody to mouse CD4 (GK1.5), addition of 3,3′,5,5′-tetramethylbenzidine (Southern Biotechnology). phycoerythrin-conjugated antibody to mouse CD23 (B3B4), fluorescein iso- thiocyanate– or allophycocyanin-indotricarbocyanine–conjugated anti-mouse Immunoblot analysis. B220+ cells were isolated from spleens with Mouse CD43 (eBioR2/60) or to mouse CD8α (53.67), allophycocyanin-conjugated anti- B220 MicroBeads (Miltenyi Biotec) and were analyzed by immunoblot with body to mouse CD38 (90), and Alexa Fluor 488–conjugated antibody to mouse anti-Bcl6 (D8) and anti-actin (C-11; both from Santa Cruz). IL-17A (ebioTC11-18H10.1) were from eBiosciences. NP-phycoerythrin was from Biosearch Technologies. All flow cytometry data were acquired on a BD Quantitative RT-PCR. Total RNA was prepared with TriZol regent (Invitrogen) LSR II and were analyzed with FlowJo software package (Tri-Star). or an RNeasy Mini kit (Qiagen), then cDNA were synthesized with Superscript reverse transcriptase and random primers (Invitrogen). Quantitative PCR was Intracellular cytokine staining. For analysis of cytokine secretion by flow done with Power SYBR Green PCR master mix (Applied Biosystems) and the cytometry, spleen cells were stimulated for 5 h with PMA (phorbol 12-myristate appropriate primers (sequences, Supplementary Table 3). 13-acetate; 20 ng/ml; Invitrogen) and ionomycin (1 µg/ml; Invitrogen) in the presence of GolgiPlug (1 µg/ml, Invitrogen) and were stained with allophyco- BMDM culture, retrovirus production and transduction. BMDMs were cul- cyanin-conjugated antibody to mouse CD4, followed by permeabilization in tured as described18. After 8 d of culture, mature macrophages were scraped

doi:10.1038/ni.2543 nature immunology off the dishes and washed with PBS, then were replated in complete DMEM. through a descending alcohol series to distilled water. Slides were boiled for Cells were allowed to adhere overnight, and then fresh medium, with or with- 20 mi in citrate antigen retrieval buffer, followed by washes under running out 5 µg/ml LPS (from Eshecrichia coli strain 055:B5; Sigma-Aldrich), was water. Endogenous peroxidase activity was blocked by treatment of the sec- added for 6 h before collection of cells. Expression constructs for Bcl6+/+ and tions for 20 min with 3% hydrogen peroxide in methanol. Tissue sections were Bcl6 with point mutations that result in the substitutions BN21K and H116A then incubated overnight at 4 °C with biotin-conjugated peanut agglutinin have been described24. Expression constructs for the latter and for Bcl6 with (Vector Laboratories). After a further wash in TBS, streptavidin–horseradish mutations that result in inactivation of RD2 or Bcl6 with mutation affecting peroxidase was added, followed by incubation for 30 min. Horseradish per- its third C2H2 zinc finger were generated with QuikChange II Site-Directed oxidas activity was detected with a DAB kit (Vector Laboratories). Finally, Mutagenesis kits (Agilent Technologies). The cDNA fragments for Bcl6+/+ and sections were counterstained with hematoxylin if necessary. For double stain- those mutated forms of Bcl6 were subcloned into MIGR1-GFP or MIGR1- ing, sections were incubated overnight at 4 °C with anti-Bcl-6 (N3; Santa puromycin retroviral expression vector. Viral supernatants were prepared with Cruz), followed incubation for 1 h with biotin-conjugated secondary antibody Plat-E cells according to a standard protocol. For retroviral infection, bone (sc-2030; Santa Cruz). Streptavidin–alkaline phosphatase was added after a marrow cells were maintained in complete DMEM for 4 d and infected with further wash in TBS followed by incubation for 30 min. Alkaline phosphatase viral supernatants in the presence of 8 µg/ml polybrene (Sigma). For MIGR1- activity was detected with a Alkaline Phosphatase Substrate Kit III (Vector GFP-infected cells, GFP+ cells were sorted 7 d after infection for analysis of Laboratories). Sections were boiled for 10 min followed by incubation with gene expression by quantitative RT-PCR. For MIGR1-puromycin–infected biotin-conjugated anti-B220 (RA3-6B2; Caltag Metsystems), then incubation cells, puromycin-resistant cells were selected by the addition of 2 µg/ml puro- with streptavidin–horseradish peroxidase. Horseradish peroxidase activity was mycin (Invitrogen) and used for quantitative ChIP assays. detected with a DAB kit (Vector Laboratories).

Immunohistology. Spleens and Peyer’s patches of the terminal ileum were Statistical analysis. Student’s t-test was used for statistical analysis with fixed in 4% paraformaldehyde and embedded in paraffin. Sections of each the software GraphPad Prism 5. P values above 0.05 were considered not sample 6 µm in thickness were prepared, cleared in xylene and hydrated significant. © 2013 Nature America, Inc. All rights reserved. America, Inc. © 2013 Nature npg

nature immunology doi:10.1038/ni.2543