A negative feedback loop mediated by the Bcl6–cullin 3 complex limits Tfh cell differentiation

Cul3cKO mice show spontaneous enlargement of secondary lymphoid tissues because of the selective accumulation of Tfh cells and the resulting expansion of GC B cells, whereas other T helper responses are unaffected (Mathew et al., 2012). We found that the Tfh phenotype was not observed in thymic CD4⁺ single-positive (SP) cells or in recent thymic emigrants (RTEs) but developed progressively in the spleen and in mesenteric, cervical, and pancreatic lymph nodes, whereas axillary/inguinal lymph nodes and Peyer’s patches were much less affected (Fig. 1, a and b). Interestingly, although the Tfh expansion was mostly made of conventional CD4⁺ T cells, it also included Foxp3⁺ T follicular regulatory (Tfr) cells (Fig. 1 d). In lethally irradiated mice reconstituted with CD45 congenic mixtures of WT and KO bone marrows, the T cell defect was clearly cell intrinsic, but a modest Tfh conversion of WT cells could also be observed at high KO/WT ratio (Fig. 1 c), suggesting some bystander effect, perhaps caused by increased expression of the Tfh-promoting factors cMaf and Il21 (Fig. 1 e).

The Tfh and GC B cell expansions developed progressively over the normal lifespan of Cul3cKO mice, with only slightly increased antinuclear antibodies detected in old age, but without overt disease or notable anomalies in blood cell counts or in function and histology of lung, liver, and kidney (Fig. 1, f and g; and not depicted). Altogether, the general picture was suggestive of exaggerated Tfh responses to environmental antigens at mucosal surfaces, without overt immunopathological consequences in the B6 background.

To test whether the Tfh expansion was dependent on antigen stimulation, we crossed the Cul3cKO mice to OTII TCR transgenic (Tg) mice. As shown in Fig. 2 a, a Tfh expansion still occurred in TCR Tg mice, but it was strictly restricted to the cells that expressed endogenous TCRs (Vα2^loVβ5^lo), whereas the cells expressing high levels of Tg Vα2 and Vβ5 were not involved. Thus, Tfh expansion depended on TCR specificity.

Tfh cell differentiation occurs in two steps, including a pre-Tfh stage, characterized by intermediate levels of Tfh markers, and a GC-Tfh stage, which depends on SAP-mediated interactions with B cells. Upon deletion of SAP in Cul3-deficient T cells, the expansion of GC-Tfh and GC B cells was markedly decreased (Fig. 2, b and c). However, the pre-Tfh population was still expanded. Thus, Cul3-deficient Tfh cells differentiated along the classical SAP-independent pre-Tfh and SAP-dependent GC-Tfh stages, and their expansion was initiated before GC entry.

We transferred 0.5 × 10⁶ WT or Cul3cKO CD4⁺ enriched SP OTII thymocytes i.v. into unirradiated CD45 congenic recipients to study their response after immunization with cognate antigen OVA in alum. First, we followed unimmunized recipients. Surprisingly, we noted that, although both CD4⁺ SP thymocytes and 24-h RTEs in Cul3cKO mice failed to express CXCR5 (chemokine [C-X-C motif] receptor 5) and other surface Tfh markers (Fig. 1 a), the transferred cells expressed conspicuously higher amounts of basal Batf (67%) and Bcl6 (41%) on average than their WT counterparts 3 d after transfer, as measured by intracellular flow cytometry (Fig. 3, a [left column] and b). These increased basal levels were maintained without further increase at day 7 after transfer (Fig. 3 b). However, these changes were not sufficient to induce the acquisition of the stereotypical Tfh phenotype because levels of CXCR5, PD1 (programmed cell death 1), and ICOS were not detectably increased in the absence of immunization. Thus, Cul3 ablation at the DP stage resulted in delayed but stable increases of basal Batf and Bcl6 proteins, without overt Tfh transformation, suggesting that some degree of Tfh dysregulation was already present before antigen exposure.

After immunization, PD1^hiCXCR5^hi Cul3cKO Tfh cells appeared earlier, expanded faster, and were maintained longer than WT Tfh cells (Fig. 3 a, top two rows). As soon as day 3 after immunization, 62 ± 2% of Cul3cKO cells expressed very high levels of PD1 and CXCR5 compared with only 33 ± 3% of WT. A similar increase was noted for ICOS, GL7, BTLA, and OX40 (not depicted). At day 7, when WT cells were already losing their Tfh phenotype, Cul3cKO cells kept expressing even higher levels of the Tfh markers, reaching 72 ± 5% compared with 13 ± 6% for controls. These differences were confirmed in multiple independent experiments and were highly significant (P < 0.001). This exaggerated Tfh response was associated with a rise of Batf and Bcl6 detected by intracellular flow cytometry, which occurred earlier and reached higher and more sustained levels than WT cells (Fig. 3 a, bottom four rows). Importantly, after immunization with OVA-NP₁₆ mixed with alum, the Cul3cKO OTII cells drove correspondingly higher titers of anti-NP antibodies detected by ELISA with BSA-NP₄₁, as well as high-affinity antibodies detected with BSA-NP₄, indicating normal Tfh functions (Fig. 3 c).

The exaggerated Tfh response was associated with increased expression of Ki67, suggesting markedly increased proliferation of Cul3cKO Tfh cells (Fig. 4 a, second row). It did not appear to occur at the expenses of other T helper programs because Cul3cKO mice immunized with OVA mixed with alum or with OVA mixed with CFA exhibited the same bias toward Gata3⁺ Th2 cells or Tbet⁺ Th1 cells, respectively (Fig. 4, a and b). Altogether, these results indicated that although Cul3cKO CD4⁺ T cells exhibited baseline increases in Batf and Bcl6 after exiting the thymus, they acquired their exaggerated Tfh cell differentiation after immunization in the periphery.

The stable baseline increase of Batf and Bcl6 proteins observed shortly after thymic maturation of Cul3cKO CD4⁺ T cells suggested that fresh thymocytes might already express detectable changes at the transcriptional level. Using quantitative real-time PCR (qRT-PCR), we measured the Batf transcript levels in large and small DP thymocytes, as well as in CD4⁺ SP thymocytes. Fig. 5 a shows that although Batf transcripts were not quantitatively altered at the early, large-DP stage in Cul3cKO mice, they were already significantly increased at the small-DP stage compared with littermate controls and further elevated at the SP stage. This kinetics was consistent with the genetic ablation of Cul3 at the early DP stage by the Cd4-Cre transgene. Thus, although increased Batf protein was detected soon after thymic emigration at the single cell level, transcriptional changes were already evident at the DP stage immediately after Cul3 ablation. A similar picture was noted for Bcl6, which was expressed at a higher level in Cul3cKO than in WT CD4⁺ SP thymocytes (Fig. 5 b). Importantly, these alterations were cell intrinsic, as the analysis included several mixed (1:1) bone marrow chimeras in which Cul3cKO and WT CD4⁺ SP thymocytes could be compared side by side.

We performed a comparative whole genome microarray analysis of CD4⁺ SP thymocytes in WT and Cu3-deficient littermates and examined a larger Tfh set of 14 genes, defined based on previously published work (Vinuesa et al., 2005; Nurieva et al., 2008; Yusuf et al., 2010), which included Slamf6, Il21, Cxcr5, Maf, Icos, Irf4, Batf, Bcl6, Pdcd1, Sh2d1a, Tiam1, Stat4, Tnfsf11, and Cd44. By examining gene resolution fold changes of Cul3cKO over WT with biological replicates plotted as x and y axes, we found that Tfh genes were particularly up-regulated relative to other genome-wide changes in expression (Fig. 5 c). The Tfh-specific up-regulation was highly significant, as shown by comparing the Tfh gene set with randomly generated sets of genes using Monte Carlo–generated simulations (P = 2 × 10⁻⁶). In contrast, a similar analysis of the Th1 and Th2 gene sets (detailed in the Materials and methods section) failed to show significant enrichment in Cul3cKO compared with WT CD4⁺ SP thymocytes, confirming the Tfh-specific nature of the thymic changes.

Recent studies suggested a potential correlation between strength of TCR signaling and Tfh responses (Fazilleau et al., 2009; Tubo et al., 2013). When stimulated in vitro with APC and peptide, Cul3cKO CD4⁺ SP OTII thymocytes exhibited a massive increase of Batf, compared with WT cells, over a large range of peptide antigen concentrations, but, importantly, other proteins induced by TCR signaling such as CD69 were at similar levels, indicating the specific nature of Batf dysregulation and the absence of global alteration of TCR signaling (Fig. 5 d). We also asked whether the Tfh bias might be dependent on TCR signaling events during thymic development. We transplanted bone marrow from OTII Tg mice with a WT or Cul3cKO background into lethally irradiated MHC II–deficient hosts. Although, as expected, the TCR Tg thymocytes did not progress beyond the DP stage in these chimeras because of lack of positive selection, Fig. 5 e shows that Cul3cKO DP thymocytes did up-regulate Batf transcripts (compared with WT DP) in the MHC II–deficient background as well, indicating that the up-regulation was independent of TCR engagement. A similar conclusion was derived from the analysis of unselected DP thymocytes obtained from MHC I/II double KO irradiated recipients reconstituted with WT or Cul3cKO bone marrow (Fig. 5 f). Thus, the basal up-regulation of Batf transcript observed in thymocytes lacking Cul3 did not depend on TCR signaling.

Human cortical thymocytes were previously shown to express the Bcl6 protein (Hyjek et al., 2001), and their mouse counterparts transiently express Bcl6 mRNA at the DP stage according to a public database (ImmGen). Flow cytometric analysis of WT mouse thymocytes revealed high amounts of Bcl6 in DP thymocytes and rapid down-regulation in SP cells (Fig. 6 a). Notably, the thymic expression levels were comparable or superior to those elicited after immunization in peripheral Tfh cells.

Immunoprecipitation of Cul3 followed by Western blot analysis of fresh DP thymocytes demonstrated that >20% of Bcl6 was bound to Cul3 (Fig. 6 b). Other BTB-ZF transcription factors such as Bach2 and c-Krox, which are expressed later at the SP stage, were not coprecipitated by anti-Cul3 antibodies in SP cells. The binding of Bcl6 to Cul3 was further confirmed by transfection in HeLa cells (Fig. 6 c). Although single transfection experiments in HeLa cells showed that Bcl6 located to the nucleus and Cul3 was mainly found in the cytosol, cotransfection resulted in the transport of Cul3 to the nucleus. Furthermore, nuclear transport depended on the Cul3 residues L52 and E55, which were previously shown to be specifically involved in binding BTB proteins (Fig. 6 c, bottom row; Wimuttisuk and Singer, 2007). Altogether, these results indicated that Bcl6 protein was naturally and transiently expressed in mouse DP thymocytes where it was prominently associated with Cul3.

As Batf was recently shown to be a key inducer of several Tfh genes, including Bcl6, cMaf, and Il21 (Ise et al., 2011), the simplest hypothesis explaining the above observations was that Bcl6–Cul3 complexes exerted a negative feedback on the Tfh program by repressing Batf and Bcl6. Recent chromatin immunoprecipitation (ChIP)–Seq studies of Bcl6-expressing macrophages identified Bcl6 binding to the first intron of the Batf gene, which contains a putative Bcl6-binding sequence motif (Barish et al., 2010, 2012). Using ChIP-qPCR, we demonstrated direct binding of Bcl6 to this intronic region of Batf in fresh thymocytes (Fig. 7 a). We also detected binding of Bcl6 to the Bcl6 proximal promoter, which includes a binding site previously identified in B cell lymphoma cells (Wang et al., 2002; Pasqualucci et al., 2003), as well as to Prdm1 and Bcl2 (Parekh et al., 2007). Thus, these findings demonstrated that, despite its transient expression, Bcl6 protein did bind key regulatory elements of multiple genes in fresh DP thymocytes.

Notably, Batf transcripts were significantly up-regulated in Bcl6cKO CD4⁺ SP thymocytes, supporting the conclusion that Bcl6 and Cul3 together regulate Batf during thymic development (Fig. 7 b). Furthermore, we detected a significant increase in the mRNA segment encompassing the exon 2–3 junction of Bcl6 in exon 7–9–deleted Bcl6cKO compared with WT (Fig. 7 b). These findings are consistent with an inhibitory function of Bcl6 on the transcription of Batf and Bcl6.

To further dissect how Bcl6 could regulate the transcription of Batf, we cotransfected Bcl6 along with a luciferase-based Batf reporter in HeLa cells (Fig. 7 c). The pBATF-luc plasmid alone was strongly repressed in cells cotransfected with Bcl6. As there were at least two potential binding sites for Bcl6 in the first intron of Batf based on sequence analysis, we examined the effects of individual and double mutations at these sites. Only the double mutation significantly relieved the repression mediated by Bcl6, suggesting that the two binding sites were used in vivo. Altogether, these results indicated that Bcl6 could directly bind regulatory elements repressing Batf transcription in thymocytes.

Because the up-regulation of Batf mRNA begins in Cul3cKO DP thymocytes and persists after emigration to the periphery, at a time when there is little residual Bcl6, we examined whether Cul3 might have helped establish lasting epigenetic marks in the Batf gene. A significant increase in H3K4Me3 and H3Ac marks were detected in Cul3cKO CD4⁺ SP thymocytes, mostly around the first intron region harboring the Bcl6-binding sites (Fig. 7 d). Thus, deletion of Cul3 at the DP stage resulted in altered epigenetic marks at the Batf gene in emerging CD4⁺ T cells.

In Cul3cKO mice, the early dysregulation of Tfh-inducing genes in thymocytes preempted the assessment of an additional role of Cul3 during antigen stimulation of mature CD4⁺ T cells. To test whether a similar autoregulatory feedback existed in mature CD4⁺ T cells undergoing Tfh cell differentiation, we deleted Cul3 by retroviral expression of Cre recombinase in Cul3^fl/fl OTII Tg CD4⁺ splenocytes. After transduction, cells were reinjected into naive hosts, which were subsequently immunized with OVA, following the same protocol as in Fig. 3. Fig. 8 shows that the GFP-expressing cells derived after transduction with MIGR1-Cre, but not control MIGR1, also exhibited a highly significant exaggeration of Tfh cell differentiation at various times after immunization (P < 0.001). Thus, we conclude that the Cul3-mediated negative feedback of Tfh cell differentiation also operates in mature CD4⁺ T cells undergoing antigen stimulation.

This study provides several new insights into the regulation of Tfh responses. First, it reveals the exquisite role of Cul3, the E3 ligase which is specifically recruited by Bcl6, in regulating the Tfh transcriptional program of CD4⁺ T cells. Second, it establishes the existence of a potent negative feedback loop whereby Bcl6–Cul3 complexes can directly repress the key Tfh-inducing genes Bcl6 and Batf. Third, and unexpectedly, it shows that the feedback loop also operates before antigen encounter by epigenetic marking during thymic development.

The BTB domain of the transcription factor Bcl6 homodimerizes to provide a platform that recruits a complex set of proteins with transcriptional regulatory activity. These include the corepressor proteins NCOR and SMRT, which bind in a mutually exclusive manner in a groove formed at the dimer interface, the MTA3 subunit of the nucleosome remodeling and deacetylase complex, the DNA methyltransferase DNMT1, the histone deacetylase HDAC1, and Polycomb group proteins (Basso and Dalla-Favera, 2012). Cul3 is known to directly bind BTB proteins and use them as adaptors for substrate recognition and ubiquitination (Duda et al., 2011; Errington et al., 2012). In a previous study of PLZF, a BTB-ZF transcription factor which directs NKT lineage cell differentiation, we demonstrated that the targets of ubiquitination included histone proteins and the genome organizers SATB1 and LaminB1, which colocalized with PLZF in the nuclear environment, but not the BTB protein itself (Mathew et al., 2012). Thus, Bcl6 is associated with a complex set of partner proteins that can mediate various epigenetic modifications at different target sites, and selective disruptions of these interactions have the potential to reveal novel functions of Bcl6 (Huang et al., 2013).

The conditional ablation of either Cul3 or Bcl6 in thymocytes resulted in significant changes of expression of Batf and of Bcl6 itself, revealing a negative feedback loop exerted by Bcl6–Cul3 complexes in thymocytes. This regulation involved the direct binding to, and repression of, Batf, a global initiator of the Tfh pathway, which is essential for induction of Bcl6, cMaf, and Il21 (Betz et al., 2010; Ise et al., 2011). In addition, as previously reported for B cell lymphomas (Pasqualucci et al., 2003), our studies suggested that, in thymocytes, Bcl6 could also bind and repress its own gene in an autoregulatory loop. Thus, our findings identify a potent negative feedback loop built into the Tfh program, an unusual feature among other T helper effector programs, which generally show self-reinforcement rather than self-limitation. Notably, previous studies demonstrated that Bcl6 could induce several genes associated with the Tfh program, including the endogenous Bcl6 itself (Nurieva et al., 2009), after overexpression in vitro. Although this study of Bcl6 autoinduction is in apparent conflict with our findings, we note that our results were obtained in vivo and that negative autoregulation of Bcl6 has also been demonstrated in B cells (Pasqualucci et al., 2003). Thus, it is possible that Bcl6 might contribute to both its induction and feedback inhibition, perhaps depending on the stage of the Tfh response, the level of Bcl6 expression, or the changing environment of other transcription factors associated with the Bcl6 network. Although our findings are consistent with the possibility that Cul3 regulated Bcl6 function, for example through the ubiquitination of critical components of the Bcl6 transcriptional complex, as suggested for PLZF (Mathew et al., 2012), the precise nature of these changes and their causal relationship remain elusive. It is also possible that Cul3 directly ubiquitinates Bcl6, thereby altering its function or its lifespan, although we have failed to document such changes so far. In any case, our findings clearly document the key role of Cul3 in regulating the Bcl6-mediated Tfh program, providing an explanation for the exaggerated Tfh responses in the absence of Cul3 in vivo.

It is noteworthy that conditional ablation of Cul3 in B cells resulted in the loss of GC formation (Mathew et al., 2012), whereas ablation in T cells increased Tfh cell differentiation and GC formation. It has been shown that by modulating the DNA damage response and controlling apoptosis, Bcl6 was critical for B cells to tolerate the DNA breaks associated with somatic hypermutation (Parekh et al., 2007). This critical antiapoptotic function of Bcl6 in GC B cells might explain why the absence of Cul3 has such drastic consequences. Interestingly, mice engineered to express a mutant form of Bcl6 lacking the ability to bind the corepressor NCOR also exhibited markedly diminished GC formation, whereas Tfh generation appeared intact (Huang et al., 2013). These studies emphasize the exquisite functional impact of individual components of Bcl6 transcriptional complexes at different genes or in different lineages.

The derepression of Bcl6 and Batf occurred in the thymus and, in the case of Batf, was clearly associated with an increase in active epigenetic marks. Although a broader analysis of epigenetic changes should reveal the biochemical details and the extent of the Bcl6–Cul3-mediated repression program in thymocytes, the current results provide a simple explanation for the increased basal overexpression of these key Tfh-inducing genes detected first at the mRNA level in thymocytes, and then at the protein level around the time of emigration. Flow cytometric analysis of Batf and Bcl6 demonstrated that these changes were widespread at the single-cell level as shown by the global shifts of FACS profiles and therefore seemed sufficient to explain the long-range effect in mature peripheral CD4⁺ T cells and their exaggerated induction of Batf and Bcl6 upon antigen stimulation. However, because ablation of Cul3 directly in mature post-thymic T cells also resulted in sustained Tfh responses, indicating that the negative autoregulatory feedback operated also during the peripheral activation of mature CD4⁺ T cells, it remains unclear whether the thymic-stage alterations alone are sufficient to deregulate Tfh responses.

There is increasing understanding that Tfh responses are exquisitely regulated at multiple levels and through separate mechanisms. For example, previous studies have highlighted the relevance of TCR signal strength during antigen stimulation (Fazilleau et al., 2009; Deenick et al., 2010; Tubo et al., 2013). Another level of regulation of Tfh responses is mediated by the E3 ligases roquin 1 and 2, which act independently of Bcl6 and directly bind Icos mRNA to regulate its degradation (Vinuesa et al., 2005; Glasmacher et al., 2010; Pratama et al., 2013; Vogel et al., 2013). In contrast with these reported mechanisms, our study showed that the regulation of Tfh-inducing genes by Cul3 was independent of TCR signal strength and occurred directly at the level of Batf and Bcl6 transcription.

The Tfh expansion of Cul3cKO mice was conspicuously localized to mucosal draining lymph nodes and spleen and was not associated with the florid signs of systemic autoimmunity reported in san/san mice. This difference might suggest the presence of additional defects in san/san and roquin-deficient T cells (Pratama et al., 2013). Alternatively, associated B cell defects might be required to produce anti-DNA antibodies, as suggested by the absence of systemic autoimmunity in mice lacking roquin 1 and 2 selectively in their T cell department (Vogel et al., 2013). Rather than autoimmunity, a key role for environmental antigens is suggested by the mucosal distribution of the Tfh expansion in Cul3cKO mice and the requirement for antigen stimulation implied by the TCR Tg experiments. Future studies will address the role that microbiota and dietary antigens might play in these immune responses.

Bcl6-deficient mice were initially reported to exhibit poor health and eosinophilic infiltration of heart and lung, with Th2 skewing of their responses to environmental and experimental antigens (Dent et al., 1997; Ye et al., 1997). This dysregulation likely results from the combination of cell-intrinsic defects in several of the cell types in which Bcl6 is expressed, including macrophages, T cells, and B cells (Kusam et al., 2003; Mondal et al., 2010; Barish et al., 2012; Hollister et al., 2013). Unlike Cul3cKO mice, Bcl6cKO mice did not exhibit Tfh expansion because this process requires Bcl6 itself, but their CD4⁺ SP thymocytes did exhibit elevated basal levels of Batf, which, in conjunction with Irf4, can bind and activate many T helper effector genes (Ciofani et al., 2012; Glasmacher et al., 2012; Li et al., 2012; Tussiwand et al., 2012). In the absence of Bcl6, these derepressed factors might promote aberrant T helper effector responses that cause inflammatory pathology.

In conclusion, our findings reveal the role of Cul3 in the transcriptional repression of key Tfh-inducing genes by Bcl6, identifying a negative autoregulatory loop that is critical for limiting Tfh responses. This negative regulation operates when mature CD4⁺ T cells develop Tfh responses in response to antigen in the periphery, but also during thymic development where the transient expression of Bcl6 appears to imprint multiple Tfh genes before thymocyte maturation and export.

C57BL/6, B6.SJL-Ptprca Pep3b/BoyJ (CD45.1), and Cd4-Cre (B6 Tg(cd4-cre)1Cwi) mice were obtained from the Jackson Laboratory. MHC II^−/− (I-Aβ^b−/−), MHC I^−/− (β2microglobulin^−/−), and MHC I/II^−/− mice in B6 background were purchased from Taconic. Sh2d1a (SAP)^−/− (Wu et al., 2001) and Cul3^fl/fl mice (McEvoy et al., 2007) in a B6 background were bred in our colony. We used Cre⁻Cul3^fl/fl mice as controls for Cul3cKO and OTII Cre⁻Cul3^fl/fl as controls for OTII Cul3cKO. Bcl6^fl/fl mice have been described previously (Hollister et al., 2013). All mice were raised in a specific pathogen–free environment at the University of Chicago, and experiments were performed in accordance with the guidelines of the Institutional Animal Care and Use Committee.

HeLa cells were maintained in DMEM (Gibco) plus 10% FBS, 1% penicillin/streptomycin. All of the cell lines were initially purchased from the ATCC.

Human pcDNA3-myc-Cul3 and pcDNA3-myc-Cul3^mut (L52AE55A) were previously described (Mathew et al., 2012). pcDNA3-Bcl6 was provided by B. Baron (University of Chicago, Chicago, IL). Anti-Myc (71D10 from Cell Signaling Technology or 9E10 from EMD Millipore) and anti-Bcl6 (Santa Cruz Biotechnology, Inc.) were used to stain confocal microscopy slides as described previously (Mathew et al., 2012).

20 × 10⁶ thymic DP or CD4⁺ SP from B6 mice were lysed and immunoprecipitated with anti-Cul3 antibody (C-0871; Sigma-Aldrich) and Western blotted with anti-Bcl6 (BD), anti–c-Krox (Abcam), anti-Bach2 (Novus Biologicals), and anti-Cul3 (Singer et al., 1999).

Fluorochrome-labeled monoclonal antibodies against CD4 (GK1.5), CD8a (53-6.7), TCRβ (H57-597), CD44 (IM7), CD25 (PC61), B220 (RA3-6B2), CXCR5 (2G8), PD1 (29F.1A12, J43), ICOS (C398.4A), Fas (JO2), GL7 (GL7), CD3e (17A2), CD45.1 (A20), CD45.2 (104), CD19 (ID3), IgD (11-26c.2a), CD69 (H1.2F3), γδ-TCR (GL3), Vα2 (B20.1), Vβ5 (MR9-4), CXCR5 (L138D7), BTLA (HMBT-6B2), and OX40 (OX86) were purchased from e-Bioscience, BD, or BioLegend. We used either fluorophore-conjugated CXCR5 antibody (L138D7) or purified CXCR5 (2G8) antibody. For CXCR5 staining using purified antibody, cells were first incubated with 5 µg/ml of purified anti-CXCR5 (2G8; BD), followed by biotinylated AffiniPure anti–rat IgG (Jackson ImmunoResearch Laboratories, Inc.) and APC-streptavidin (Invitrogen). After addition of 100 µg/ml of purified rat IgG2a isotype, cells were incubated with other extracellular antibodies. Bcl6 (clone K12-91; BD), Batf (clone MBM7C7; eBioscience), Foxp3 (clone FJK-16s; eBioscience), Tbet (clone 4B10; BioLegend), Ki-67 (clone SolA15; eBioscience), and Gata3 (TWAJ; eBioscience) were stained after fixation and permeabilization with the Foxp3 staining buffer set from eBioscience. Samples were analyzed on an LSRII (BD) or sorted on a FACSAria (BD) or MoFlo (Dako). Data were analyzed using FlowJo software (Tree Star).

6–8-wk-old B6 (CD45.1 or CD45.1/CD45.2) mice were irradiated (100 Gy) using a gamma cell 40 irradiator with a cesium source and i.v. injected with 2–5 × 10⁶ bone marrow cells from donor mice after T cell depletion using anti-CD3 and AutoMACS (Miltenyi Biotec). The recipient mice were analyzed 8–10 wk after irradiation.

Serum dilutions were incubated with Hep-2 cells (AN-1012; MBL International), followed by FITC-conjugated anti–mouse IgG (eBioscience). Mean nuclear fluorescence intensities of 20 cells per sample were recorded using ImageJ software (National Institutes of Health). Positive control included a serum from MRL^lpr/lpr mice.

For in vitro OTII cell stimulation, thymocytes were depleted of CD69⁺ cells using PE-conjugated anti-CD69 and anti-PE magnetic microbeads followed by AutoMACS. 5 × 10⁴ CD69-depleted thymocytes were incubated with 5 × 10⁵ CD3-depleted CD45 congenic splenocytes and cultured with a range of OVAp (323–339) concentrations (GenScript) in duplicate in 96-microwell plates for 20 h at 37°C before FACS analysis. For in vivo cell transfer experiments, 5 × 10⁵ CD8-depleted thymocytes from OTII or OTII Cul3cKO mice were i.v. transferred into CD45 congenic recipient mice 1 d before i.p. injection of 50 µg OVA (Worthington Biochemical Corporation) mixed 1:1 with alum (Thermo Fisher Scientific) or s.c. injection in the left and right hock of a total of 50 µg OVA mixed 1:1 with CFA (Sigma-Aldrich). Each mouse was injected s.c. in the left and right hind hock with 50 µl of the CFA emulsion containing 25 µg of OVA protein. OTII cells in recipient spleens were characterized by FACS after identification based on anti-CD45 allele-specific antibodies, anti-Vβ5 and anti-Vα2 antibodies.

To measure the T helper–dependent antibody response to the hapten NP, 5 × 10⁵ CD8-depleted thymocytes from OTII or OTII Cul3cKO mice were i.v. transferred into CD45.1/CD45.2 congenic recipient mice 1 d before i.p. immunization with 50 µg OVA-NP₁₆ (Biosearch Technologies) mixed 1:1 with alum. Serum was collected 0 and 21 d later to titer anti-NP antibodies by ELISA against BSA-NP₄₁ or BSA-NP₄ (Biosearch Technologies) coated on 96-well plates, using anti-IgG1 isotype-specific antibodies (BD). Values between different experiments were normalized to the same serum standard.

6–8-wk-old mice were injected with 5 µl of a 1 mg/ml solution of FITC (Sigma-Aldrich) into each thymic lobe under anesthesia and killed 20 h later for analysis of FITC⁺ spleen cells.

MIGR1 or MIGR1-Cre (Pear et al., 1998) was transfected into Plat-E cells, and viral supernatants were collected 48 and 72 h later. Splenic CD4⁺ cells were enriched by MACS depletion with a mixture of anti-B220, -CD19, -Ter119, and -CD8a and stimulated with plate-bound anti-CD3 (5 µg/ml) and anti-CD28 (2 µg/ml) for 24 h before infection with viral supernatant (1 ml for 10⁶ cells) in the presence of 5 µg/ml polybrene by spin infection for 90 min at 800 g. After transduction, 10⁶ cells were injected i.v. into CD45 congenic mice, followed by i.p. immunization 24 h later with 50 µg OVA in alum (1:1).

HeLa cells were transfected using Lipofectamine (Invitrogen) as per the manufacturer’s instruction with 0.5 µg Bcl6 and 2 µg Cul3 plasmids. Cells were harvested 24 h after transfection for confocal microscopy.

For qRT–PCR, total RNA was prepared from 10⁶ sorted thymic CD4⁺ SP (CD4⁺CD8⁻TCRγδ⁻CD25⁻CD1d–αGalCer tetramer⁻), large DP (CD71⁺FSC^hi) or small DP (CD71⁻FSC^lo) cells from Cul3cKO or Bcl6cKO, and their littermate controls by using TRIzol (Invitrogen), the RNeasy kit (QIAGEN), and reverse transcription with the Superscript III first strand synthesis super mix cDNA kit (Invitrogen). cDNA expression was determined with the Mx3005P using Brilliant SYBR green master mix (Agilent Technologies). Fluorescence signals were measured over 40 PCR cycles, and the cycle (Ct) at which signals crossed a threshold set within the logarithmic phase was recorded. The Ct for the target gene was subtracted from the Ct for 18S (ΔCt), and ΔΔCt was calculated by subtracting ΔCt for KO from ΔCt for littermate (ΔΔCt). The fold change of mRNA in KO over littermate was calculated as 2^−ΔΔCt.

ChIP experiments were performed as previously described (Mathew et al., 2012). In brief, cells were fixed and isolated nuclei were resuspended in 130 µl nuclei lysis buffer (50 mM Tris-HCl, pH 8.1, 10 mM EDTA, 0.5% SDS, and 1× protease inhibitor) sonicated to obtain 100–500-bp fragments (Covaris-S220). Chromatin from 20 million cells was used for immunoprecipitation using anti-Bcl6 (sc-N3), anti-H3K4me3 (07-473; EMD Millipore), or anti-H3Ac (06-599; EMD Millipore). Specific DNA sequences were assessed by qRT-PCR using primers as listed.

The following qRT-PCR primers were used: batf_f, 5′-ACAGACACAGAAAGCCGACAC-3′; batf_r, 5′-CTCGGTGAGCTGTTTGATCTC-3′; bcl6_f, 5′-ACTGGGCAAACACAACATGG-3′; bcl6_r, 5′-GGAGGCGATTAAGGTTGAGAAG-3′; bcl6_exon 2–3_f, 5′-ACTGGGCAAACACAACATGG-3′; bcl6_exon 2–3_r, 5′-GGAGGCGATTAAGGTTGAGAAG-3′; il21_f, 5′-GCCAAACTCAAGCCATCAAACC-3′; il21_r, 5′-TTCTCATACGAATCACAGGAAGGG-3′; c-maf_f, 5′-AGCAGTTGGTGACCATGTCG-3′; c-maf_r, 5′-TGGAGATCTCCTGCTTGAGG-3′; 18s_f, 5′-TTGACTCAACACGGGAAACC-3′; 18s_r, 5′-ACCCACGGAATCGAGAAAGA-3′; gapdh_f, 5′-CTGGAGAAACCTGCCAAGTATG-3′; and gapdh_r, 5′-TCCTCAGTGTAGCCCAAGATG-3′.

The following Blc6 ChIP-qPCR primers were used: bcl6-f, 5′-GGGTTCTTAGAAGTGGTGATGC-3′; bcl6-r, 5′-CAGCAACAGCAATAATCACCTG-3′; batf-f, 5′-CGTCTCTGCCACTTAGGTTCC-3′; batf_r, 5′-TGGAACATTCCTCTGCTTATGG-3′; bcl2-f, 5′-TCTGCCCTGGAGGTCTGAAG-3′; bcl2-r, 5′-CTGTGATTCTCCCTTCTTCTCG-3′; prdm1-f, 5′-TGCAGAACTTCTAGGTTCTTTCTTG-3′; and prdm1-r, 5′-GCTTTGGAAGGTTCTGAATGG-3′.

The following histone ChIP-qPCR primers were used: batf(−100)-f, 5′-TCGCTCTGCCTTTCAAACTG-3′; batf(−100)-r, 5′-AAGTCACATGGGCGGAAAC-3′; batf(+580bp)-f, 5′-CGTCTCTGCCACTTAGGTTCC-3′; batf(+580bp)-r, 5′-TGGAACATTCCTCTGCTTATGG-3′; batf(+700bp)-f, 5′-TGTTCCAGTAGATCGCTGCTTC-3′; batf(+700bp)-r, 5′-TCTGTCTTGCCACTTGCTTG-3′; batf(+1.5kb)-f, 5′-TTGCCCTCTCTAACCCAATG-3′; batf(+1.5kb)-r, 5′-CTTGGGTCAGGTTCTGATTCC-3′; batf(+2k)-f, 5′-AACGCGCTTGTACAGAAGTTG-3′; batf(+2k)-r, 5′-TGCGCACATGCTAAACCTTC-3′; batf(+2.5k)-f, 5′-CCTCCTGAGGCTTCAATTTCTC-3′; batf(+2.5k)-r, 5′-TCTTCAGCATTCCCTGCTACC-3′; batf(+3k)-f, 5′-CAGGTTGTGGAAAGGGCTTC-3′; batf(+3k)-r, 5′-AGAAAGGTTCAGTGTCCAGTGC-3′; batf(+3.5k)-f, 5′-TCATCGCATCCTTCTTCCTG-3′; batf(+3.5k)-r, 5′-CTTATCTGGGTTTGGCACCTC-3′; batf(+4k)-f, 5′-TCACACACGCATGTGCATC-3′; batf(+4k)-r, 5′-TATGTGCGTGTGGTGTTTCC-3′; batf(+9.5k)-f, 5′-ATGTCTGCGAGGCTGAATTG-3′; batf(+9.5k)-r, 5′-CCACCATTGGTAATTTGGAGAC-3′; batf(+18.5k)-f, 5′-TGATGGGCCTAACCACATTC-3′; batf(+18.5k)-r, 5′-TAGGCAGTGTGGAGATCAAGC-3′; batf(+18.8k)-f, 5′-TCCGAGCTTACAGACATTCCTC-3′; batf(+18.8k)-r, 5′-TGACTAACCACTGCCTGCTG-3′; batf(+19.4k)-f, 5′-AGCATTCAAAGTGTCGCTGAG-3′; batf(+19.4k)-r, 5′-TTAAGTTGTGGAGGAGCAGAGG-3′; batf(+20k)-f, 5′-CAAGCTTCTGGCCTGAAATG-3′; and batf(+20k)-r, 5′-CCCGGACCTGAATAAACAGAAC-3′.

The proximal promoter, first exon, and first intron containing the Bcl6-binding motif of mouse Batf (−572 to 924) were cloned into the SmaI and XhoI site of pGL3-basic vector (Promega). To disrupt the Bcl6-binding site, a site-directed mutagenesis kit (Agilent Technologies) was used to change the sequences from 5′-TTTCTAGGAA-3′ (576/585: Bcl6_1) and 5′-TTTCCAGGAA-3′ (696/705: Bcl6_2) to 5′-TTGTCGACTA-3′ (m1) and 5′-TTGATATCGAC-3′ (m2), respectively. These plasmids were transfected along with a cmv-Renilla plasmid into 293T cells (ATCC) using Lipofectamine (Invitrogen). Reporter assays were performed using the dual luciferase detection system (Promega), and firefly luciferase levels were normalized to Renilla luciferase values.

RNA was isolated from FACS-sorted thymic CD4⁺ SP cells (CD4⁺CD8⁻TCRγδ⁻CD25⁻CD1d–αGalCer tetramer⁻) from 4.5-wk-old Cul3cKO and littermate controls using TRIzol and processed for 430A.v2 arrays (Affymetrix) according to the manufacturer’s instructions. To test whether Tfh-related genes were non-randomly up-regulated in Cul3cKO thymocytes, we compared the SSMDs of Tfh-, Th1-, and Th2-related genes with the SSMDs of randomly generated subsets of genes. First, we selected Tfh-related genes (Slamf6, Il21, Cxcr5, Maf, Icos, Irf4, Batf, Bcl6, Pdcd1, Sh2d1a, Tiam1, Stat4, Tnfsf11, and CD44), Th1 genes (Eomes, Tbet, Fasl, IL12rb2, Ifng, IL18rap, Stat1, Stat4, IL27ra, IL18r1, Cxcr3, Zbtb32, and Ccr5), and Th2 genes (pparg, Il5, Ccr1, Il1rl1, Areg, Il13, Il4, Gata3, Ccr3, Il13ra1, Il24, Ccr8, Ccr4, Stat6, IL4ra, Ccl5, Il9, and Ccl11) based on previously published work (Vinuesa et al., 2005; Nurieva et al., 2008; Yusuf et al., 2010). Then, drawing from all genes represented on the microarray, we generated 10 million randomly selected, equally sized subsets of genes using the Monte Carlo method, producing an ensemble of gene sets to serve as a null distribution of possible observed effects. For each set of genes, we compared probe set intensities between Cul3 KO mice and their WT littermates, quantifying the overall effect size using a paired-sample SSMD (Zhang, 2011). The SSMD for paired observations is defined as the mean of the log fold changes (FC) divided by the standard deviation of the log fold changes:

FC_n denotes the log fold change for the n-th observation with respect to a negative reference, μ is the arithmetic mean, and σ is the standard deviation. The SSMD of the Tfh-related probe sets estimates the effect size of Cul3 deficiency on Tfh-specific genes, whereas the SSMDs for the randomly generated sets approximate the distribution of possible effects observed by chance alone. To estimate the statistical significance of Tfh-related gene up-regulation, we compared the SSMD of Tfh genes (SSMD_Tfh = 0.788) with the Monte Carlo–generated distribution of random SSMDs. In our sample of 10 million sets of genes, SSMDs greater than or equal to SSMD_Tfh occurred 24 times, corresponding to a p-value of 2 × 10⁻⁶. A similar analysis conducted for the Th1 and Th2 gene sets showed nonsignificant p-values of 0.3 in each case.

Unpaired Student’s t test was performed with Prism (GraphPad Software). *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Gene expression data are available at the Gene Expression Omnibus (GEO) database under accession no. GSE56514.

This work was supported by the Cancer Research Institute (R. Mathew), National Institutes of Health (NIH) grants RO1 GM106173 and AI038339 and the Howard Hughes Medical Institute (to A. Bendelac), the University of Chicago Digestive Diseases Research Core Center grant P30 DK42086 and NIH grant R21 AI099825 (to A.L. Dent), and NIH grant RO1 GM082940 (to J.D. Singer). A.H. Chiang and A.R. Dinner’s effort is in part sponsored by the Defense Advanced Research Projects Agency (D12AP00023), and the content of the information does not necessarily reflect the position or the policy of the Government, and no official endorsement should be inferred.

The authors declare no competing financial interests.