The Alternative NF-κB Pathway in Regulatory T Cell Homeostasis and Suppressive Function

This article is featured in In This Issue, p.2227

Maintenance of immune homeostasis and tolerance is achieved through multiple feedback mechanisms. CD4⁺Foxp3⁺ regulatory T cells (Tregs) represent 5–15% of the CD4⁺ T cell population and develop both in the thymus (natural Treg [nTreg]) and the periphery (peripheral Treg [pTreg]). They play a pivotal role in the control of innate and adaptive immune responses (1), and are generally characterized by the constitutive expression of CD25 and the forkhead-box transcription factor Foxp3, which is required for their suppressive activity (2). Perturbations in the homeostasis and/or function of these cells are associated with the development of autoimmune diseases such as type 1 diabetes and rheumatoid arthritis (3). Depletion of Tregs in mice leads to a lethal multifocal inflammation; moreover, mutations in the foxp3 locus are responsible for the Scurfy phenotype in mice and the IPEX (immunodysregulation, polyendocrinopathy, and enteropathy, X-linked) syndrome in humans (4). Thus, Treg development, maintenance, and function must be tightly regulated through various molecular mechanisms. In recent years, several studies have demonstrated a crucial role for the NF-κB transcription factor in the development of Tregs and the expression of Foxp3 (5–7).

NF-κB signaling can be separated in two main pathways. The canonical pathway leads to the phosphorylation and degradation of IκBα/β and consequent nuclear translocation of the NF-κB1 p50 protein bound to either c-Rel or p65, whereas the alternative pathway leads to activation of NF-κB–inducing kinase (NIK), which promotes processing of NF-κB2 from the full-length precursor protein p100 to the p52 form, which results in formation of transcriptionally active p52:RelB complexes (8, 9). The biological processes regulated by these two NF-κB pathways are distinct. The canonical p65 and c-Rel subunits are well-defined inducers of inflammation and have clear roles in the activation of B and T cells, whereas the alternative NF-κB RelB:p52 heterodimer has primarily been implicated in lymphoid tissue organogenesis and cell migration (9, 10). NF-κB2 and RelB are important regulators of immune tolerance, because NIK^−/−, relb^−/−, and nfkb2^−/− mice develop spontaneous autoimmunity (reviewed in Ref. 11). Importantly, the two NF-κB pathways are triggered by different receptors and signaling pathways. Upon engagement of the TCR/CD28 Ag receptor, the canonical pathway is activated and NF-κB complexes containing c-Rel can regulate de novo expression of Foxp3 in developing thymocytes, and therefore c-Rel–deficient animals exhibit fewer nTregs (5–7, 12–15). Recently, we and others have demonstrated a central role for the NF-κB p65 subunit in the maintenance of mature Treg identity and in the prevention of autoimmunity (16, 17). In contrast, c-Rel, but not p65, is required for the homeostasis of Tregs during antitumor responses (18). These observations highlight canonical NF-κB signaling as a master regulator of Treg development and function, and demonstrate the discrete functions of individual NF-κB subunits in Treg-dependent immune tolerance.

The alternative NF-κB signaling pathway, by contrast, is largely inactive during normal T cell homeostasis. In lymphocytes, NF-κB2 resides predominantly in the cytoplasm in its unprocessed p100 form, where it is believed to function as an inhibitor of other NF-κB proteins (9). However, activation of a subset of TNFR family members expressed in lymphocytes, including OX40, CD40, or LT-βR, leads to activation of the alternative pathway. Specifically, these receptors direct the recruitment and inhibition of the TRAF2/3/cIAP1/2 ubiquitin ligase complex, which allows the stabilization and activation of NIK. The alternative pathway also appears to be indirectly involved in Treg development. For example, NIK^−/− mice and NIK mutant aly/aly mice have reduced numbers of Tregs (19, 20). The same observation was made more recently in mice when IKKα was conditionally deleted in T cells (21). Finally, constitutive expression of NIK in T cells induced the peripheral expansion of poorly functional Tregs (22). However, enforced expression of NIK can activate both alternative and canonical pathways (23), and therefore the specific role for the alternative NF-κB pathway remained unclear. Likewise, germline deletions of relb or nfkb2 were associated with developmental and autoimmune defects associated with changes with lymphoid organogenesis, obscuring analysis of their functional role in Tregs (9). To overcome these difficulties, we have now used conditional deletion of nfkb2 and/or relb in both total T cells and Tregs, and demonstrated a critical function for NF-κB2 in maintaining Treg homeostasis. Surprisingly, rather than functioning as a crucial component of alternative pathway-dependent transcriptional complexes, we found that the key function of p100 in Tregs was as a negative regulator of p52-independent, RelB-containing complexes. Hence our studies suggest that p100 and RelB have critical and unexpected functions in Treg homeostasis and suppressive function.

Nfkb2-floxed and relb-floxed mice were generated by U. Klein (Columbia University, New York, NY) (24). CD4^cre [Tg(CD4-cre)^1Cw1], Foxp3^CRE-YFP [Foxp3^{tm4(YFP/cre)Ayr}], CD45.1 (Ptprc^a Pepc^b/BoyJ), and RAG1^−/−, all on a C57BL/6J background, were originally purchased by the Jackson Laboratory and maintained in our animal facility. C57BL/6J mice were purchased from the Jackson Laboratory. All mice were kept in specific pathogen-free conditions in the animal care facility at Columbia University (New York, NY). All mouse experiments were approved by the Institutional Animal Care and Use Committee of Columbia University.

Cells were isolated from thymus, spleen, and lymph nodes (LNs) by mechanical desegregation in PBS+FBS 3%. Colon lamina propria lymphocytes were isolated upon digestion and Percoll for intracellular cytokines analyses, and cell suspensions were incubated 3 h with PMA (50 ng/ml; Sigma) and ionomycin (1 μg/ml; Sigma) in the presence of Golgi Plug (BD Biosciences). Cells were then stained with mAbs purchased from eBioscience or Tonbo Biosciences. Foxp3 and cytokine staining were performed using the eBioscience kit and protocol. Cells were acquired on an LSR II (BD Biosciences) and analyzed with FlowJo (Tree Star) software. Cell sorting was achieved on a FACSAria II (BD Biosciences).

CD4⁺CD44^lowCD25⁻ naive T cells were FACS-sorted from splenocyte suspensions. A total of 10⁵ T cells were cultured in complete RPMI 1640 (Life Technologies) with 10⁵ T cell–depleted, mitomycin C–treated wild-type (WT) splenocytes and 2.5 μg/ml anti-mCD3 (Bio X Cell), in the presence of 10 ng/ml mIL-2 (PeproTech) and grading doses of human TGF-β1 (PeproTech), for 4 d at 37°C. Cells were then stained for flow cytometry analysis.

For in vitro assays, LN-derived CD45.1⁺ naive conventional CD4⁺ T (Tconv) cells were magnetically isolated (Miltenyi) and labeled with CellTrace Violet Proliferation Tracker (CTV; Life Technologies). They were cultured with T cell–depleted, mitomycin C–treated WT splenocytes and 2.5 μg/ml anti-mCD3, in the presence or not of FACS-sorted WT or nfkb2^−/− CD4⁺YFP⁺ Tregs. Proliferation of CD4⁺CD45.1⁺ T cells was assessed by FACS at day 4. The percentage of suppression was calculated as described. For in vivo assays, 4 × 10⁵ naive CD45.1⁺ Tconv cells were isolated as described earlier and transferred with or without 1 × 10⁵ Tregs to the retro-orbital sinus of 6- to 9-wk-old RAG1^−/− mice. Recipients were then weighed every week and euthanized when weight loss was >30%.

Total lysates were extracted using radioimmunoprecipitation assay buffer and protease inhibitors with SDS. Cytosolic and nuclear fractions were fractionated using the classical hypo- and hypertonic lysis buffers protocol. Twenty micrograms of protein extracts was run in polyacrylamide gels and transferred onto polyvinylidene difluoride membranes. Membranes were incubated with polyclonal anti–NF-κB2 (K-27), RelB (C-19), p65 (C-20), c-Rel (C, sc-71), HDAC-1 (H51) (Santa Cruz), and monoclonal anti-GAPDH (10R-G109a; Fitzgerald) Abs, followed by HRP-coupled secondary Abs from Jackson Immunoresearch.

Total RNA was extracted using a Qiagen RNeasy Mini Kit with DNase treatment. For quantitative PCR, RNA was reverse transcribed by Superscript III (Invitrogen). cDNAs were used for PCR with SYBR Green reagents (Quanta Biosciences, Gaithersburg, MD) on a C1000 Touch thermal cycler (Bio-Rad, Hercules, CA). The data were normalized to GAPDH expression. Primers sequences can be sent under request. For RNA sequencing, RNA was extracted from Tregs isolated from three to four pooled mice per genotype. Libraries were prepared using an Illumina TruSeq Library Kit and sequenced by an Illumina 2500 instrument. Upon sequencing, raw FASTQ files were aligned on the mm10 genome using STAR aligner with default parameters (25). Aligned fragments were then counted and annotated using Rsamtools v3.2 and the TxDb.Mmusculus.UCSC.mm10.knownGene version 3.1.2 transcript database, respectively. Normalized fragments per kilobase per million mapped reads were obtained using the robust fragments per kilobase per million mapped reads estimate function of DeSeq2 v1.10.1. Differentially expressed genes were obtained using the DESeqResults function of the same package. For gene set enrichment analysis, we acknowledge our use of the Gene Set Enrichment Analysis (GSEA) software and Molecular Signature Database (MSigDB) including the ImmuneSigDB (C7 collection) (26, 27). We also used the Panther Database for further function analysis (28). The Gene Expression Omnibus accession number is GSE108532 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE108532).

Colons were washed and fixed in 4% formalin acetate. Five-micrometer sections were then prepared from the middle section of the colon; three to four sections, each separated by 300 μm, were prepared for each sample. Tissues were then stained with H&E. Quantification of abscesses was performed in a double-blind fashion. The mean number of abscesses among the sections is shown.

Experimental groups were compared statistically using the nonparametric Mann–Whitney U test or the unpaired, two-tailed Student t test.

To investigate the T cell–intrinsic role of NF-κB2 p100/p52 in the differentiation and function of CD4⁺ T cell subsets, we first ablated nfkb2 in total T cells using the CD4^cre deleter strain. This led to the efficient depletion of p100 in splenic CD4⁺ T cells. As expected, p52 expression was not observed at steady-state in WT cells (Supplemental Fig. 1A and data not shown). Deletion of nfkb2 was not associated with significant changes in the proportion and numbers of CD4 and CD8 T cells in both primary and secondary lymphoid organs (Fig. 1A–C). Next, we examined the differentiation of Foxp3⁺ Tregs. The proportion and number of immediate Foxp3⁻ Treg precursors and mature Foxp3⁺ Tregs were similar in the thymus of WT and knockout (KO) animals (Fig. 1E, Supplemental Fig. 1B). However, we observed a significant increase in the percentage of Tregs in the spleen and peripheral LNs (pLNs) upon nfkb2 deletion (Fig. 1D, 1E). This increase in the fraction of Tregs corresponded to an increase in the absolute number of Tregs in these tissues (Fig. 1E), consistent with there being an expansion of Treg numbers rather than a decrease in the Tconv populations. Moreover, the activation state of nfkb2-deficient Tregs was altered, as demonstrated by significantly increased expression of CD44 and Ki67 relative to WT Tregs (Supplemental Fig. 1C). Interestingly, we also observed a trend toward decreased expression of Foxp3 in nfkb2^−/− Tregs (Supplemental Fig. 1D). Thus, ablation of nfkb2 in T cells led to a selective expansion of the peripheral Treg population without apparent effects on the thymic differentiation of Tregs.

Although we did not observe any changes in CD4 or CD8 T cell phenotype, we wished to confirm that these changes in Treg homeostasis were due to intrinsic changes in p100/p52 function in Tregs. We therefore selectively deleted nfkb2 in Tregs using the Foxp3^CRE deleter mice. In 6- to 8-wk-old animals, there were no obvious signs of systemic inflammation or autoimmunity. Consistent with the lack of gross phenotypic changes in these mice, there were no changes in thymic Treg differentiation or in the distribution or effector function of conventional CD4 (Tconv) and CD8 T cells in the mutant mice (data not shown). However, consistent with the results obtained with Cd4^CRE mice, the Treg population was significantly expanded in the spleen and LN of Foxp3^CREnfkb2^F/F animals when compared with littermate controls (Fig. 2A, 2B). Again, nkfb2-deficient Tregs displayed significantly impaired expression of Foxp3, whereas Treg activation markers were upregulated in nfkb2-deleted Tregs (Fig. 2C and data not shown). Taken together, these results demonstrate an important role for p100, and/or p52, in the regulation of Treg homeostasis.

Although these results demonstrate a role for p100/p52 in Tregs, we wondered whether nfkb2 controlled Treg homeostasis in a cell-intrinsic manner or indirectly, for example, by regulating production of a factor that was driving Treg expansion. Therefore, to formally demonstrate a Treg-autonomous role for nfkb2, we performed mixed bone marrow (BM) chimera experiments. In such an experimental setting, any cell-extrinsic defect in the KO BM would either be compensated for by the presence of WT BM or be revealed by an effect on WT Tregs. Lethally irradiated WT recipients were transplanted with a 1:1 mixture of congenic WT BM and either Foxp3^CRE or Foxp3^CREnfkb2^F/F BM (Fig. 2D). Eight weeks after BM reconstitution, the proportion of Tregs in KO spleen and LN compartments was dramatically increased when compared with their WT counterparts (Fig. 2E). Consistent with the increase in deleted Tregs, upregulation of the Ki67 proliferation marker was also selectively observed in Tregs in which nfkb2 was deleted. Thus, nfkb2 has a cell-intrinsic function in restricting Treg proliferation in the periphery. Next, we assessed whether these expanded Tregs were of thymic origin or derived from a peripheral conversion of Tconv cells. The percent expression of two genes associated with nTregs, the membrane receptor Nrp-1 and the transcription factor Helios, were not affected by loss of nfkb2. This suggested that the enhanced nfkb2^−/− Treg expansion was occurring in nTregs of thymic origin (Fig. 2G). We also measured the effect of ablating nfkb2 on the in vitro conversion of Tconv into Foxp3⁺ Tregs, which mimic pTregs. After 4 d of culture, the proportion of Foxp3⁺ cells was slightly reduced in the absence of nfkb2 (Fig. 2H). Therefore, NF-κB2 does not have effects on either nTreg or pTreg development that explain the expansion of the peripheral Treg population. Instead, the data suggest that nfkb2 is a key, cell-autonomous regulator of nTreg homeostasis.

As described earlier, young Foxp3^CREnfkb2^F/F mice did not display abnormal inflammation or autoimmunity. However, beginning around 12 mo of age, we began to observe variable weight loss and hunched posturing in the Foxp3^CREnfkb2^F/F animals. Histological examination revealed increased immune infiltrate in the colons of Foxp3^CREnfkb2^F/F mice, whereas other tissues (lungs, liver, and kidney) were unaffected (Fig. 3A and data not shown). Of note, these localized colonic infiltrates were not associated with significant histological changes at the mucosa (Fig. 3A). FACS analysis revealed increased numbers of IFN-γ–producing CD4 and CD8 T cells in the colon lamina propria of nfkb2-deficient mice compared with WT controls, whereas the IL-17A⁺ cell population was unchanged (Supplemental Fig. 2A, 2B). Next, the lymphoid tissue composition was analyzed. The cellularity of both spleen and pLNs was significantly increased in Foxp3^CREnfkb2^F/F animals (Fig. 3B). Interestingly, the CD4⁺ T cell compartment was significantly expanded, leading to an increased CD4/CD8 ratio (Supplemental Fig. 2C). Moreover, we observed enhanced activation of both CD4⁺ Tconv and CD8⁺ T cells, as shown by the increased expression of CD44 and Ki67 (Fig. 3C–E). Strikingly, we observed a nearly 2-fold increase in the proportion of CD4 and CD8 T cells expressing IFN-γ (Fig. 3F–G). However, the expression of IL-17A, TNF, and IL-2 was not significantly different between WT and KO mice (Fig. 3H, Supplemental Fig. 2D). Finally, we could not detect any change in the levels of serum TNF and total IgG (data not shown). Thus, the expression of NF-κB2 by Tregs prevents the late development of a mild autoimmune syndrome, likely mediated by Th1 CD4 and Tc1 CD8 T cells.

Given the expansion of Tregs in young Foxp3^CREnfkb2^F/F animals, we hypothesized that the observed autoimmune phenotype in aged animals was likely the consequence of an impaired Treg suppressive function. Indeed, FACS analysis revealed that the inflammatory syndrome observed in aged Foxp3^CREnfkb2^F/F mice was associated with continued expansion of Tregs (Fig. 4A). Therefore, we next examined the suppressive capacity of nfkb2-deficient Tregs in vitro and in vivo. For the in vitro assay, the proliferative capacity of responder CD4 T cells was measured in the absence or presence of varying ratios of WT or KO Tregs. Nfkb2-deficient Tregs exhibited reduced inhibitory activity at each ratio relative to WT Tregs (Fig. 4B). We next assessed the activity of WT and nfkb2^−/− Tregs in vivo using a T cell transfer colitis model (29). Although WT Tregs effectively prevented weight loss and reduced IFN-γ expression by splenic Tconv cells transferred into RAG1^−/− mice, nfkb2^−/− Tregs were considerably less effective (Fig. 4C, 4D). This was not due to a loss of the Treg population per se, because the proportion of nfkb2^−/− Tregs 5 wk after transfer was three times higher than that of WT Tregs (Fig. 4E). To better understand this loss of Treg suppressive function, we measured the mRNA expression of several well-characterized genes involved in Treg function. Whereas the expression of Tgfb1, Lag3, and CD73 was unaffected in Tregs lacking NF-κB2, Ctla4 mRNA was strongly downregulated (Supplemental Fig. 2E). This was associated with a decreased expression of intracellular CTLA-4 in nfkb2^−/− Tregs, both in unmanipulated Foxp3^CREnfkb2^F/F mice and in the T cell transfer colitis model (Supplemental Fig. 2F). Taken together, these results demonstrate that the expression of NF-κB2 is required to restrict Treg proliferation and also to maintain optimal suppressive capacity.

We next explored the molecular mechanism driving dysregulated Treg homeostasis in the absence of NF-κB2. NF-κB2 can act as both a negative regulator of other Rel-proteins, by sequestering them through interaction via its ankyrin domains, and as a positive regulator by generating the p52 subunit, which can associate with other NF-κB subunits to form transcriptionally active NF-κB heterodimers. Therefore, we first sought to examine the expression and activation state of the transcriptionally active NF-κB subunits, RelB, p65, and c-Rel. Both the expression and stimulation of induced nuclear localization of the canonical NF-κB subunits RelA and c-Rel were unaffected in nfkb2^−/− Tregs (Fig. 5A). In contrast, there was a dramatic increase in the nuclear accumulation of RelB in TCR/CD28-stimulated nfkb2^−/− Tregs (Fig. 5A). Therefore, the data suggested that NF-κB2 predominantly functions as an inhibitor of RelB activation in Tregs. To test the hypothesis that the primary function of NF-κB2 in Tregs is to regulate RelB activation, we deleted relb in both total T cells (using CD4^cre) and Tregs (using Foxp3^cre). RelB protein was undetectable in splenic T cells of CD4^crerelb^F/F mice (Supplemental Fig. 3A). Surprisingly, Relb deletion did not have a significant effect on either the overall distribution of CD4 and CD8 T cells in lymphoid organs or the corresponding number of Tregs (Supplemental Fig. 3B, 3C). Strikingly, ablation of relb in the CD4^crenfkb2^F/F background restored a normal proportion and number of peripheral Tregs (Fig. 5B). Hence loss of NF-κB2 resulted in increased RelB activation, and the Treg expansion observed in nfkb2-deficient animals was dependent on RelB.

We next sought to determine whether the altered in vivo suppressive function observed in Foxp3^CREnfkb2^F/F mice was also dependent on RelB. Therefore, we next measured inflammation/autoimmunity symptoms in Treg-specific, Foxp3^CREnfkb2^F/Frelb^F/F, double-KO mice. Consistent with the normalization of Treg homeostasis upon RelB loss, no increase in colonic abscesses was evident in 10- to 12-mo-old Foxp3^CREnfkb2^F/Frelb^F/F mice (Fig. 5C, 5D). Moreover, the global cellularity of spleen and LN was also normal (Fig. 5E). As in the CD4^cre deleter strain, Foxp3^CREnfkb2^F/Frelb^F/F mice also displayed normal numbers of peripheral Tregs (Fig. 5F). Finally, we examined the activation state of conventional CD4 and CD8 T cells. Again, the level of CD44 and IFN-γ expression remained unchanged in KO mice when compared with WT littermates (Fig. 5G–I, Supplemental Fig. 3D–F). Taken together, these data indicate that p100 is an essential negative regulator of RelB in Tregs and that the uncontrolled expansion of Tregs, and the subsequent autoimmunity that develops after deletion of nfkb2, is secondary to enhanced RelB activation.

To further explore how NF-κB2 controls Treg homeostasis, we compared the transcriptomes of WT, nfkb2^−/−, relb^−/−, and nfkb2^−/−relb^−/− Tregs by RNA sequencing. To ensure an optimal activation of the alternative NF-κB pathway, we stimulated Tregs with anti-CD3, anti-CD28, anti–OX40, and IL-2. We initially analyzed global changes in gene expression between the genotypes (Fig. 6A, 6B). Surprisingly, we found that loss of NF-κB2 resulted in a far broader effect on the Treg transcriptome than did loss of RelB. Specifically, the analysis of the nfkb2^−/− Tregs transcriptome was notable for the large number of upregulated transcripts (175 upregulated versus 84 downregulated), reinforcing the idea the NF-κB2 predominantly affects Treg function through its IκB-like repressor activity. Interestingly, the dysregulated gene set in nfkb2^−/− Tregs showed only minor overlap with other genotypes. Furthermore, relb deletion did not impair gene expression, which was at odds with its role in B cells (24) but was consistent with the lack of a detectable phenotype in Foxp3^CRErelb^F/F mice. Remarkably, however, dysregulated expression of the vast majority of genes in nfkb2-deficient Tregs was rescued upon relb ablation (Fig. 6B). Therefore, these data further support the idea that NF-κB2 primarily acts as an inhibitor of RelB-mediated transcription. Interestingly, neither partial nor complete ablation of the alternative NF-κB pathway drove a change in the expression of other NF-κB subunits genes (Supplemental Fig. 4A). GSEA and other gene ontology analyses revealed significant changes in the expression of genes related to proliferation, survival, and migration, as well as cytokine expression (Fig. 6B, Supplemental Fig. 4B). We observed a dramatic increase in the expression of genes involved in cell proliferation and antiapoptotic functions in nfkb2-only deleted cells (Fig. 6C). The expression of the proapoptotic gene Bcl2l11, which encodes Bim, was downregulated in nfkb2^−/− Tregs, whereas the kinase Kit was upregulated. These genes were restored to the WT expression levels upon additional ablation of relb. This perturbed expression profile may explain the specific accumulation of Tregs in the tissues of nfkb2-deficient mice.

It has previously been suggested that the alternative pathway may control the expression of migration and homing molecules in stromal cells and innate immune cells. In Tregs, the expression of numerous chemokines and chemokine receptors were modified by the absence of NF-κB2, but not RelB (Fig. 6C). However, most of these genes were upregulated and not downregulated, revealing a previously unknown facet of the alternative NF-κB pathway as an inhibitor of expression of homing molecules in Tregs.

Finally, we observed changes in the expression of many genes related to inflammation and Th cell differentiation. Nfkb2^−/− Tregs displayed a significant overexpression of inflammatory molecules, in particular cytokines and transcription factors that are hallmarks of Th17 effector cells, such as Il-17a, Il-21, Ccl20, and Rorc (Fig. 6C, Supplemental Fig. 4B). This was further confirmed by quantitative RT-PCR and flow cytometry analyses (Fig. 6D, 6E). Moreover, the aberrant expression of most of these genes was “rescued” by knocking out relb, underscoring a deleterious role for uncontrolled RelB activation on the Treg-associated molecular profile. These results are consistent with a predominantly inhibitory function of NF-κB2 in the expression of cell cycle and proinflammatory genes. As we recently demonstrated, the canonical NF-κB subunits p65 and c-Rel are critical for maintaining the expression of Treg hallmark genes (16). However, there was not a significant positive or negative correlation in gene expression between Foxp3^CRERela^F/FRel^F/F and Foxp3^CREnfkb2^F/F Tregs (Supplemental Fig. 4C), suggesting that perturbations in gene expression in these two genotypes were distinct, and further supporting the idea that the inhibitory effect of NF-κB2 is on RelB-containing complexes. Therefore, taken together, our observations demonstrate a novel and unique function of NF-κB2 in maintaining Treg homeostasis and function.

Recently, significant progress has been made in understanding the intrinsic role of the canonical NF-κB pathway in the biology of Tregs (16–18). Indeed, we have shown that conditional deletion of p65 in Tregs drove a systemic autoimmune syndrome, which was further aggravated by deletion of c-Rel. c-Rel itself exhibited a critical function in maintaining Treg homeostasis specifically during antitumor immunity (18). We now demonstrate that the alternative NF-κB pathway, acting through its NF-κB2 subunit, is also crucial for the maintenance of Treg homeostasis and function. Forced activation of the alternative pathway by deleting TRAF3 (Traf3^−/− mice), or overexpressing NIK (NIK-transgenic [Tg] mice), leads to increased numbers of Tregs (22, 30). Based on these findings it would have been predicted that loss of nfkb2 would result in impaired Treg expansion. Instead, surprisingly, we find that deletion of nfkb2 drives a cell-autonomous increase in Treg numbers in secondary lymphoid organs. This is likely due to the expansion of nTregs in the periphery because in vitro–differentiated Treg generation was only slightly impaired.

Interestingly, Treg numbers returned to baseline upon additional deletion of RelB (i.e., nfkb2^−/−relb^−/− double KO). Of note, in both Foxp3^CRErelb^F/F and Foxp3^CREnfkb2^F/Frelb^F/F mice, the relative proportion of Tregs was not decreased compared with WT mice. This suggested that NF-κB2 functions to prevent aberrant RelB activation, which otherwise has no role in Treg expansion at steady-state. Previous studies had indirectly addressed the role of the alternative NF-κB pathway in Treg expansion. For instance, several reports showed a cell-intrinsic requirement for NIK in the peripheral homeostasis of Tregs (20, 31). Our data suggest that this result is likely due to the effect of NIK on the function of canonical NF-κB. A similar decrease of Tregs was observed in conditional IKKα-KO mice (21); but once again, because IKKα can also activate the canonical NF-κB pathway (32), this suggests that the effects of NIK or IKKα relied more on impaired p65 or c-Rel activation, rather than on a diminished activation of the alternative NF-κB pathway.

Nfkb2 encodes for both p100 and p52. In this study, we show that deletion of nfkb2 can drive uncontrolled Treg activation; however, it is not clear whether this effect results from the ablation of p100, p52, or both. Previous studies have suggested that p100 can act as a repressor of TCR-triggered signaling events; for instance, overexpression of p100 in cell lines decreased initial NF-κB activation and IL-2 transcription upon TCR triggering (33). Similarly, constitutive activation of NIK in c-IAP^H570A mutant mice, or in NIK-Tg animals, led to overactivation of CD4⁺ T cells (22, 34). Moreover, T cells isolated from mice carrying a germline deletion of nfkb2 were also more proliferative in vitro (34). Finally, NIK-deficient T cells were slightly hyporesponsive in vitro and in vivo (19, 31). In summary, these studies suggest that Treg expansion in nfkb2-conditional KO animals was likely due to the absence of the inhibitor p100, rather than the absence of transcriptionally active p52-containing complexes. Mechanistically, several studies have demonstrated the role of p100 as a potent inhibitor of both canonical and alternative NF-κB signaling. For example, the p100 C-terminal ankyrin domains have been shown to physically interact and inhibit the translocation of RelB, p65, and c-Rel (19, 33, 35). Most compelling, however, are studies highlighting RelB as the critical target of p100 in the cytosol of mouse embryonic fibroblasts, Hela cells, dendritic cells, and T cells (36–39). Consistent with this, our studies in Tregs revealed that the nuclear translocation of RelB, but not p65 or c-Rel, was significantly increased in the absence of nfkb2. This suggested that an aberrant RelB activation was responsible for the accumulation of Tregs in vivo. This hypothesis was confirmed by abolishment of nfkb2-deficient Treg expansion upon the additional ablation of relb.

Analysis of the transcriptional landscape associated with our KO lines further confirmed the specific role of NF-κB2 in restraining the activity of the other NF-κB subunits. Indeed, many genes whose expression was increased in nfkb2^−/− Tregs were normalized in the nfkb2^−/−relb^−/− cells. Because we did not observe a correlation of the nfkb2^−/− Treg gene signature with the Rela^−/−Rel^−/− Treg gene signature, it supported our hypothesis that nfkb2 deletion did not affect the function of the canonical NF-κB transcription factors. To our surprise, only a small number of genes were affected by the loss of RelB, or both NF-κB2 and RelB, suggesting that transcription by alternative NF-κB subunits was mainly expendable for the maintenance of Treg identity in the steady-state. However, it remains possible that p52 and/or RelB could display a more prominent function in Treg biology during acute or chronic inflammation, which is known to drive the engagement of a number of TNFR family members and, therefore, leads to alternative NF-κB activation and p100 processing (11, 40).

Previous studies have shown a role for the alternative pathway in the expression of chemokines and adhesion molecules by stromal cells upon LT-βR triggering (9). This pattern was only partially confirmed in BAFF and CD40-stimulated nfkb2^−/− B cells (24). In our experiments, we actually observed an increase in the expression of migration/homing genes, upon CD3/CD28/OX40 stimulation of nfkb2^−/− Tregs, but not in double-KO (nfkb2^−/−relb^−/−) Tregs. Another striking difference between nfkb2^−/− and double-KO Tregs is the significant enrichment of a Th17 cell signature in the absence of p100 alone. The increased expression of Rorc or Il17a might explain why these p100-deficient Tregs partially lose their inhibitory function, even when the expression of suppressive cytokines or costimulation markers remain intact. IL-17–producing Foxp3⁺ Tregs have been implicated in the development of certain autoimmune and chronic inflammatory conditions including psoriasis (41, 42). Interestingly, high numbers of IL-17⁺Foxp3⁺ cells were found in the lamina propria and blood of patients with inflammatory bowel disease and mouse models of colitis (43, 44). Therefore, it is possible that development of colonic inflammation in aging Foxp3^CREnfkb2^F/F animals relies on the aberrant expression of inflammatory cytokines by Tregs.

Even though aging Foxp3^CREnfkb2^F/F mice had an expanded Treg population in their pLNs, the conventional CD4 and CD8 T cells exhibited increased activation and secretion of inflammatory cytokines, leading to undesired infiltration in nonlymphoid tissues. This suggested that despite a large increase in numbers, these Tregs had impaired suppressive activity. This functional deficit was confirmed both in vitro and in vivo. Mice with conditional deletion of Ctla4 in Tregs also develop a severe autoimmune syndrome despite a significant increase in their number of Tregs (45), and we observed reduced CTLA-4 expression by nfkb2^−/− Tregs. This may partially explain both Treg expansion and their lack of suppressive function in Foxp3^CREnfkb2^F/F mice. Moreover, uncontrolled activation of RelB might be deleterious for Tregs. This possibility has been previously addressed indirectly. For instance, the triggering of the GITR receptor, which induces activation of the alternative pathway (46), led to increased T cell survival and proliferation (47). This drove a significant Treg accumulation in GITRL-Tg animals (48). Meanwhile, Treg suppressive function was impaired by GITR signaling (49). These data also fit our model. Therefore, all of these observations may help unmask a specific role for the alternative pathway in Treg homeostasis.

Based on the previous analyses of TNFR superfamily members in Tregs and the data presented in this article, we propose a model in which p100 is a crucial regulator of RelB and Treg homeostasis and function. Following TCR stimulation, induced synthesis of p100 results in the inhibition of RelB-containing complexes. However, upon activation of the alternative NF-κB pathway, for example, through OX40 or GITR, increased processing of p100 results in the activation of RelB-containing complexes leading to Treg expansion coupled with loss of Treg suppressive function. This model is consistent with recent reports analyzing OX40 and GITR function and provides a molecular explanation for the ability of these alternative NF-κB pathway stimuli to inhibit Treg function and tumor tolerance. Likewise, these data provide further insight into how constitutive activation of the canonical and alternative NF-κB pathways might drive pathology in patients with chronic inflammation and autoimmune disease (11, 40, 50). Therefore, taken together, our data demonstrate that the adequate balance of NF-κB activity, regulated by NF-κB2 p100, is essential for maintaining optimal Treg function and immune tolerance.

We thank Dev Bhatt, Alexis Desrichard, Alice Lepelley, Gaelle H. Martin, and Thomas S. Postler for technical help and constructive discussions on the project.

The authors have no financial conflicts of interest.