A TNF‐p100 pathway subverts noncanonical NF‐κB signaling in inflamed secondary lymphoid organs

Abstract

Lymphotoxin‐beta receptor (LTβR) present on stromal cells engages the noncanonical NF‐κB pathway to mediate RelB‐dependent expressions of homeostatic chemokines, which direct steady‐state ingress of naïve lymphocytes to secondary lymphoid organs (SLOs). In this pathway, NIK promotes partial proteolysis of p100 into p52 that induces nuclear translocation of the RelB NF‐κB heterodimers. Microbial infections often deplete homeostatic chemokines; it is thought that infection‐inflicted destruction of stromal cells results in the downregulation of these chemokines. Whether inflammation per se also regulates these processes remains unclear. We show that TNF accumulated upon non‐infectious immunization of mice similarly downregulates the expressions of these chemokines and consequently diminishes the ingress of naïve lymphocytes in inflamed SLOs. Mechanistically, TNF inactivated NIK in LTβR‐stimulated cells and induced the synthesis of Nfkb2 mRNA encoding p100; these together potently accumulated unprocessed p100, which attenuated the RelB activity as inhibitory IκBδ. Finally, a lack of p100 alleviated these TNF‐mediated inhibitions in inflamed SLOs of immunized Nfkb2 ^−/− mice. In sum, we reveal that an inhibitory TNF‐p100 pathway modulates the adaptive compartment during immune responses.

Introduction

Homing of naïve lymphocytes into SLOs, such as lymph nodes (LN) and the spleen, is important for continuous monitoring of the tissues that these SLOs drain, and for mounting antigen‐specific adaptive immune response. Homeostatic chemokines expressed by lymphoid stromal cells ensure the steady‐state ingress of naïve lymphocytes and their segregation into specific compartments in SLOs. CCL21 and CCL19 produced by fibroblastic reticular cells (FRCs) congregate T lymphocytes in T‐cell zones (Cyster, 2005). CXCL13 and CXCL12 derived from follicular dendritic cells (FDCs) and marginal reticular cells (MRCs) guide B lymphocytes to B‐cell follicles. Lymphotoxin‐α₁β₂ (LTα₁β₂) present on the surface of lymphocytes chronically activates LTβR expressed by FRCs, FDCs, and MRCs that sustains the transcriptions of genes encoding these chemokines (Boulianne et al, 2012). Disruption of lymphotoxin signaling in adult mice abrogates the expressions of homeostatic chemokines and obstructs the homing of naïve lymphocytes in SLOs (Ngo et al, 1999; Browning et al, 2005).

LTβR induces the expressions of homeostatic chemokines through the noncanonical NF‐κB pathway. In unstimulated cells, Nfkb2‐encoded p100 retains RelB and other NF‐κB subunits in the cytoplasm in a multimeric IκBδ complex (Sun, 2012; Tao et al, 2014). LTβR‐induced noncanonical signaling activates NIK (Sanjo et al, 2010), which in association with inhibitor of NF‐κB kinase 1 (IKK1 or IKKα) phosphorylates p100 (Dejardin et al, 2002; Yilmaz et al, 2003). Subsequent proteasomal processing removes the C‐terminal inhibitory domain from p100 that produces the mature p52 subunit, and concomitantly releases the RelB:p52 NF‐κB dimer and a minor RelB:p50 NF‐κB dimer from the IκBδ‐inhibited complex into the nucleus (Derudder et al, 2003; Basak et al, 2008). It was shown that NIK, IKK1, and RelB are required for the expression of CCL21, CXCL13, and CXCL12 in LTβR‐stimulated cells and in SLOs (Weih et al, 2001; Dejardin et al, 2002; Bonizzi et al, 2004; Basak et al, 2008). Interestingly, RelB:p52 and RelB:p50 dimers possess overlapping functions; an elevated basal activity of RelB:p50, which accumulates in the nucleus in the absence of inhibitory p100‐IκBδ, compensates for the lack of the LTβR‐induced RelB:p52 dimer in Nfkb2 ^−/− cells and in lymphoid tissues of Nfkb2 ^−/− mice, and mediates the expression of homeostatic chemokines (Lo et al, 2006; Basak et al, 2008).

TNF engages a separate canonical pathway to activate the RelA:p50 NF‐κB dimer, which induces the expression of pro‐inflammatory cytokines and chemokines. In this pathway, phosphorylation of NF‐κB inhibitor α (IκBα) bound to the RelA:p50 dimer by a complex composed of NF‐κB essential modulator (NEMO) and inhibitor of NF‐κB kinase 2 (IKK2) triggers the proteasomal degradation of IκBα and nuclear activation of RelA:p50. Of note, TNF‐activated RelA:p50 dimer induces the synthesis of its own inhibitor IκBα as well as the noncanonical signal transducer p100 (Mitchell et al, 2016).

Systemic as well as localized infection of mice often leads to attenuated expressions of homeostatic chemokines in reactive SLOs (Chang & Turley, 2015). Cytotoxic T cells generated against lymphocytic choriomeningitis virus (LCMV) destroy infected FRCs, which produce these chemokines (Scandella et al, 2008). Infection of stromal cells by mouse cytomegalovirus downregulates the expression of CCL21 (Benedict et al, 2006). Salmonella typhimurium invades draining LNs and obstructs CCL21 and CXCL13 expressions by stromal cells (St John & Abraham, 2009). Importantly, microbial downregulation of these chemokines restricts the ingress of naïve lymphocytes in reactive SLOs (Benedict et al, 2006; Mueller et al, 2007; St John & Abraham, 2009). It is thought that inhibition of homeostatic chemokines contributes to the subversion of adaptive immunity by pathogens. In addition, microbial infections trigger the accumulation of pro‐inflammatory cytokines in reactive SLOs, and interferon‐γ (IFNγ) have been implicated in the suppression of homeostatic chemokines during LCMV infection (Mueller et al, 2007). However, it remains unclear if infection‐inflicted disruptions of lymphoid stromal cells are required, or inflammation per se is sufficient for depleting these chemokines. More so, cell‐intrinsic mechanisms that may limit the transcription of these noncanonical NF‐κB target genes in reactive SLOs have not been examined.

Here we report that TNF accumulated in inflamed SLOs upon non‐infectious immunization of mice with ovalbumin (OVA) in complete Freund's adjuvant (CFA) restricts the expressions of RelB‐target homeostatic chemokines and thereby limits the trafficking of lymphocytes. Our mechanistic study revealed that TNF inactivated NIK and induced the expression of Nfkb2 mRNA; these together potently accumulated the unprocessed p100, which abrogated the pre‐existing RelB activity as IκBδ in LTβR‐stimulated cells. Finally, a lack of p100 alleviated these TNF‐mediated inhibitions in inflamed SLOs of immunized Nfkb2 ^−/− mice. Our study suggests that an inhibitory mechanism involving p100‐IκBδ links dynamical signaling induced by pro‐inflammatory TNF and immune homeostatic processes directed by LTβR during immune responses.

Results

TNF suppresses the expression of homeostatic chemokines in draining lymph nodes of OVA–CFA‐immunized mice

Microbial infections often cause a downregulation of homeostatic chemokines in SLOs and thereby prevent the homing of naïve lymphocytes. However, less is known about the impact that inflammation by itself produces on these homeostatic processes. To address this question, we immunized mice by footpad injection with OVA–CFA, which elicits inflammatory Th1 responses in reactive SLOs (Ke et al, 1995), and examined draining as well as control, contralateral popliteal lymph nodes (pLNs). We transferred naïve CD45.1⁺ splenocytes into the immunized C57BL/6 CD45.2⁺ recipient mice and scored the presence of CD45.1⁺ cells in pLNs by FACS. We identified a twofold to threefold decrease in the frequency of transferred B and T cells in draining pLNs at day 2 post‐immunization (Fig 1A). Our ELISA analysis revealed a close to threefold reduction in the abundances of CCL21, CXCL13, and CXCL12 in draining pLNs as compared to control pLNs (Fig 1B). Furthermore, we measured the abundances of mRNAs encoding these chemokines by quantitative reverse transcription polymerase chain reaction (qRT–PCR). We observed that the expressions of these mRNAs were downregulated in draining pLNs within 1 day of immunization, and were slowly restored to that detected before immunization by second week except for CXCL13 (Fig 1C). The abundance of CXCL13 mRNA returned to normal in 5 days.

Figure 1. Homing of naïve lymphocytes and the homeostatic chemokine expressions in draining lymph nodes of OVA–CFA‐immunized mice.

1. Naïve splenocytes were transferred from B6.SJL CD45.1⁺ congenic donors via retro‐orbital route to WT C57BL/6 CD45.2⁺ recipient mice (n = 8), which were administered with OVA–CFA through subcutaneous, footpad injections 2 days prior to the transfer. At 6 h post‐transfer, pLNs were harvested and the frequencies of the transferred B (B220⁺CD45.1⁺) and T (CD90⁺CD45.1⁺) cells in draining, ipsilateral LNs were determined by FACS. Data were normalized against the respective control pLNs obtained after injecting saline in the contralateral footpad.
2. WT mice (n = 8) were immunized as in (A), and the levels of CCL21, CXCL13, and CXCL12 in draining and control pLNs were measured at day 2 by ELISA.
3. The expressions of homeostatic chemokine mRNAs were measured by qRT–PCR in draining pLNs of WT and Ifng ^−/− mice (n = 6 for each data point) at the indicated days post‐immunization. Data were normalized against the respective control pLNs.
4. IFNγ (top) and TNF (bottom) levels were measured by ELISA in draining pLNs of immunized WT mice (n = 5 per group).
5. qRT–PCR demonstrating relative abundances of the indicated mRNAs at day 2 post‐immunization in draining pLNs of WT mice (n = 5 per group), which were intraperitoneally administered with PBS or etanercept. Etanercept was injected 6 h prior to immunization with OVA–CFA.
6. FACS revealing relative abundances of the indicated adoptively transferred lymphocytes at day 2 post‐immunization in draining pLNs of WT mice (n = 6 per group), which were subjected to etanercept treatment prior immunization. Arrows indicate corresponding frequencies of these transferred lymphocytes in the WT mice subjected to OVA–CFA immunization in the absence of etanercept, as presented in (A).

Data information: In all panels, data are means ± SEM. ***P < 0.001; **P < 0.01; *P < 0.05 (paired two‐tailed Student's t‐test).

Local antigen challenge leads to the accumulation of pro‐inflammatory cytokines in draining LNs (Cyster, 2005). We also noted that OVA–CFA immunization potently accumulated TNF and moderately increased the level of IFNγ in draining LNs of WT mice within 24 h (Fig 1D). An earlier investigation implicated IFNγ in the reduced expressions of homeostatic chemokines upon LCMV infection (Mueller et al, 2007). We instead noticed rapid depletion of these chemokine mRNAs and proteins in draining pLNs of IFNγ‐deficient mice upon OVA–CFA immunization (Figs 1C and EV1A). In comparison to WT mice, however, restoration of the levels of these mRNAs was subtly accelerated in IFNγ‐null mice. Inhibition of TNF by administering etanercept in WT mice preserved fully the abundances of CXCL13 and CXCL12 mRNAs, and partially restored the expression of CCL21 mRNA, at day 2 post‐immunization with OVA–CFA (Fig 1E). Similar results were obtained upon OVA–CFA immunization of Tnfr1 ^−/− mice (Fig EV1B). On the other hand, our analysis revealed that OVA–CFA‐immunized IFNγ‐deficient mice accumulated TNF (Fig EV1C), and that inhibition of TNF restored the expressions of homeostatic chemokines in these knockout mice (Fig EV1D). Finally, etanercept treatment preceding immunization prevented completely the reduction in the frequency of adoptively transferred B cells and partially restored the frequency of transferred T lymphocytes in the reactive pLNs of WT mice (Fig 1F). Our in vivo studies broadly suggest that TNF accumulated upon non‐infectious OVA–CFA immunization suppresses the expressions of homeostatic chemokines and thereby diminishes the ingress of naïve lymphocytes in inflamed SLOs.

Figure EV1. Analysis of reactive pLNs derived from IFNγ‐deficient and Tnfr1 ^−/− mice.

1. IFNγ‐deficient mice (n = 6) were immunized with OFA‐CFA through footpad injections, and the levels of CCL21 and CXCL12 in draining pLNs and control, contralateral pLNs were measured by ELISA after 2 days.
2. qRT–PCR demonstrating the relative levels of the indicated chemokine mRNAs in draining pLNs at day 2 post‐immunization of Tnfr1 ^−/− mice (n = 14); relative levels of these mRNAs in immunized WT mice have been indicated by arrows.
3. TNF levels were measured by ELISA in draining pLNs of immunized Ifng ^−/− mice (n = 5 per group).
4. qRT–PCR of indicated mRNAs in draining pLNs at day 2 post‐immunization of IFNγ‐deficient mice (n = 5 for each set), which were also administered with PBS or etanercept.

Data information: In all panels, data are means ± SEM; ***P < 0.001; *P < 0.05; ns, not significant (paired two‐tailed Student's t‐test).

TNF subverts noncanonical RelB NF‐κB signaling in LTβR‐stimulated cells

In resting SLOs, chronic LTβR signal sustains the transcription of genes encoding homeostatic chemokines in stromal cells. We asked whether TNF‐dependent depletion of these chemokines in immunized mice involved cell‐intrinsic signaling mechanisms. We stimulated cultured cells either with an agonistic anti‐LTβR antibody (αLTβR) or with TNF or subjected them to LTβR stimulation for 36 h and then treated them with TNF in the continuing presence of αLTβR (Fig 2A). We determined signal‐induced gene expressions in the lymphoid stromal‐derived BLS4 and BLS12 cell lines (Katakai et al, 2004). Our qRT–PCR analysis demonstrated that LTβR stimulation induced, whereas TNF alone did not discernibly alter, the expressions of CXCL13 and CXCL12 mRNAs (Fig 2B and C). Interestingly, TNF treatment for 12 h attenuated the expressions of these mRNAs in LTβR‐stimulated cells (Fig 2B and C). The RelB heterodimers activated by the noncanonical NF‐κB pathway mediate the expressions of homeostatic chemokines (Fig 2D) (Bonizzi et al, 2004; Basak et al, 2008). TNF activates the RelA:p50 dimer, which induces the expression of pro‐inflammatory cytokines. Our analyses in the genetically tractable mouse embryonic fibroblasts (MEFs) confirmed that RelB was required for LTβR‐stimulated, sustained expressions of CXCL13 and CXCL12 mRNAs, whose abundances at 36 and 48 h post‐stimulation were nearly equivalent (Fig EV2A). When MEFs stimulated for 36 h with αLTβR were subsequently treated with TNF for another 12 h in the continuing presence of αLTβR, the expressions of these RelB‐target genes were reduced (Fig 2E). TNF‐induced expression of RANTES mRNA, encoded by a RelA‐target gene, was not inhibited in the combinatorial treatment regime.

Figure 2. TNF inhibits the nuclear RelB activity and represses the expressions of RelB‐target chemokine genes in LTβR‐stimulated cells.

• A
  Schematic presentation of treatment regimes. Cells were stimulated with an agonistic anti‐LTβR antibody (αLTβR) for 36 h and then either were additionally treated with TNF in the presence of αLTβR for the indicated times (L‐T, representing 12 h of costimulation unless mentioned otherwise), or αLTβR stimulation was continued for another 12 h (L, representing 48 h of LTβR stimulation altogether). Alternatively, cells were left untreated for 36 h and then treated with TNF alone (T, representing 12 h of TNF treatment).
• B, C
  qRT–PCR analysis of the indicated homeostatic chemokine mRNAs in lymphoid stromal‐derived BLS4 (B) and BLS12 (C) cells subjected to the indicated treatments. U denotes untreated cells. Data represent four biological replicates.
• D
  Graphical depiction of the TNF‐induced canonical pathway, which activates the RelA:p50 dimer, and the LTβR‐stimulated noncanonical pathway, which activates the RelB:p52 as well as the RelB:p50 heterodimers. The RelB heterodimers mediate the expression of homeostatic chemokines, whereas RelA:p50 induces the expression of pro‐inflammatory mediators.
• E
  qRT–PCR analysis of the indicated homeostatic chemokine mRNAs in WT MEFs subjected to the indicated treatments. U denotes untreated cells. Data represent four biological replicates.
• F, G
  Nuclear NF‐κB activities (top panel) induced upon stimulation of WT MEFs (F) or BLS4 and BLS12 cells (G) were resolved in EMSA. The arrow and the arrowhead indicate the NF‐κB DNA binding complex composed of RelA and RelB heterodimers, respectively. Supershifting RelA with an anti‐RelA antibody, the residual RelB DNA binding activities were revealed (middle panel). Oct1 DNA binding served as a loading control (bottom panel). L represents cells stimulated with αLTβR alone for 48 h. The signals corresponding to the nuclear DNA binding activities of RelB (RelBn) were quantified from three independent experiments.
• H
  The RelB DNA binding activity induced in the nucleus of WT MEFs upon indicated treatments was assessed by TransAM DNA‐binding ELISA. Data represent four biological replicates.

Data information: In all panels, data are means ± SEM. ***P < 0.001; **P < 0.01; *P < 0.05 (paired two‐tailed Student's t‐test). Source data are available online for this figure.

Figure EV2. Investigating NF‐κB signaling in WT MEFs and in BLS4 cells as well as BLS12 cells.

1. Gene expression analysis showing the relative abundances of CXCL13 and CXCL12 mRNAs in WT and Relb ^−/− MEFs subjected to treatment with 0.3 μg/ml of αLTβR for 36 h or 48 h or with 1 ng/ml of TNF for 12 h. Data are means ± SEM from three independent experiments. ***P < 0.001; **P < 0.01 (paired two‐tailed Student's t‐test).
2. WT MEFs, BLS4 cells, and BLS12 cells were stimulated in a time course with 0.1 μg/ml of αLTβR. Cells were harvested, and nuclear extracts were analyzed by EMSA for the presence of NF‐κB DNA binding activity. The arrow and the arrowhead indicate the NF‐κB DNA binding complexes composed of RelA and RelB heterodimers, respectively. The signals corresponding to LTβR‐induced nuclear RelB activity (RelBn) were quantified from three experiments and presented as means ± SEM.
3. The composition of NF‐κB DNA binding complexes induced upon 48 h of αLTβR treatment in WT MEFs (top panel), as well as in BLS4 (middle) and BLS12 (bottom) cells were examined by supershift analysis using the indicated antibodies. “#” represent antibody supershifted complexes. The arrow and the arrowhead indicate the NF‐κB DNA binding complexes composed of RelA and RelB heterodimers, respectively.
4. WT MEFs, BLS4 cells, and BLS12 cells were stimulated in a time course with 1 ng/ml of TNF. Cells were harvested, and nuclear extracts were analyzed by EMSA for the presence of NF‐κB DNA binding activity. The signals corresponding to TNF‐induced nuclear RelA activity (RelAn) were quantified from three experiments and presented as means ± SEM. The arrow indicates the NF‐κB DNA binding complexes composed of the RelA heterodimers.
5. ChIP analyses, representative of four biological replicates, revealing recruitment of RelB to Cxcl13 promoter in WT MEFs subjected to the indicated treatments. Fold enrichment of Cxcl13 promoter in RelB immunopellet relative to control IgG was determined by qPCR. Binding to Gapdh promoter served as negative control. Data are means ± SEM. *P < 0.05 (paired two‐tailed Student's t‐test).

We then captured the nuclear NF‐κB activities induced in these cells in the electrophoretic mobility shift assay (EMSA), and distinguished between RelA‐ and RelB‐containing DNA binding dimers in the supershift analysis. We found that LTβR stimulation of WT MEFs, BLS4 cells, and BLS12 cells induced a similar, delayed RelB activity, which persisted in the nucleus with comparable levels at 36 and 48 h post‐stimulation, and was composed of mostly the RelB:p52 dimer (Fig EV2B and C). TNF treatment alone rapidly induced the RelA:p50 activity, which was subsequently attenuated by the IκBα negative feedback (Fig EV2D). Our time course analysis revealed that the nuclear RelB activity present in LTβR‐stimulated MEFs was removed within 8 h of TNF treatment despite the continuing presence of αLTβR, and was restored by 48 h (Fig 2F). However, RelB activity was unaffected at the early 0.5‐h time point, which corresponds to the maximal RelA activity induced by TNF. Similarly, TNF treatment drastically diminished the RelB activity in LTβR‐stimulated BLS4 cells and BLS12 cells, and this inhibition was apparent at the late 12‐h time point (Fig 2G). Furthermore, our DNA‐binding ELISA analyses confirmed that TNF treatment for 12 h abrogated the nuclear RelB activity in LTβR‐stimulated MEFs (Fig 2H). As reported earlier (Bonizzi et al, 2004), LTβR stimulation of MEFs induced the recruitment of RelB to the promoter of Cxcl13 gene in our chromatin immunoprecipitation analyses (Fig EV2E). Corroborating our nuclear DNA binding analyses, TNF treatment of LTβR‐stimulated MEFs abolished this RelB binding to the Cxcl13 promoter. We conclude that TNF abrogates noncanonical RelB NF‐κB signaling in LTβR‐stimulated cells and downregulates the expressions of RelB‐target homeostatic chemokines involving a cell‐autonomous mechanism.

TNF signaling inhibits NIK:IKK1 activity and induces the accumulation of inhibitory p100‐IκBδ in LTβR‐stimulated cells

Mathematical reconstruction of the NF‐κB network has led to the identification of emergent properties in prior studies (Basak et al, 2012). To understand the mechanism underlying TNF‐mediated suppression of noncanonical signaling, we utilized a previously published NF‐κB mathematical model (version 5.0‐MEF, Shih et al, 2012) subsequent to necessary revisions (Fig EV3A and Appendix Supplementary Methods). Our kinase assay revealed that LTβR induced a persistent NIK:IKK1 activity, which lasted even after 48 h of cell stimulation (Fig EV3B) (Banoth et al, 2015). When we fed this experimental NIK:IKK1 activity as input, computational simulations captured various features of LTβR‐induced noncanonical signaling, such as sustained nuclear activation of the RelB dimers, progressive accumulation of p52, and gradual disappearance of p100 (Figs 3A and EV3C–E). Model simulations involving the short‐lived NEMO:IKK2 activity, which was experimentally measured in a TNF time course (Fig EV3F) (Banoth et al, 2015), similarly recapitulated transient RelA activation observed during TNF signaling (Fig EV3C). To simulate the combinatorial regime, we first plugged the NIK:IKK1 activity in the model and after 36 h additionally fed the NEMO:IKK2 input. However, TNF signaling in LTβR‐stimulated system further augmented the nuclear RelB activity in silico that was accompanied by an increased accumulation of p52, but a relatively unaltered p100 level (Fig 3A). Corroborating the earlier report (Shih et al, 2009), TNF induced RelA‐dependent expression of Nfkb2 mRNA in naïve as well as LTβR‐stimulated MEFs in both our experiments and simulations (Figs 3B and EV3G). We reasoned that the enduring NIK:IKK1 activity efficiently converted p100 produced from TNF‐induced Nfkb2 mRNA into p52 and thereby reinforced the RelB:p52 activity in our computational analyses.

Figure EV3. Combined experimental and mathematical modeling studies to dissect the crosstalk between the LTβR pathway and TNF signaling.

1. Relative affinities of the RelA and the RelB heterodimers for binding to the C‐terminal inhibitory domain of p100 were biochemically assessed. Briefly, nuclear extracts derived from LTβR‐stimulated MEFs were used as a source of the DNA binding RelA and RelB heterodimers. These endogenous NF‐κB dimers, mostly consisting of the RelA:p50 dimer and the RelB:p52 dimer, were incubated with a gradient of recombinant p100_406–899, which contains the inhibitory domain as well as the signal‐responsive serines, for 30 min to facilitate the formation of IκB:NF‐κB complex, which does not bind DNA. Subsequently, the abundances of the residual unbound NF‐κB dimers were scored in EMSA. Incubation with 5 nM of p100_406–899 was sufficient for completely abrogating the RelB DNA binding. However, more than 500 nM of p100_406–899 was required for preventing the RelA DNA binding. The data, representing three experimental replicates, also indicated that the C‐terminal inhibitory domain of p100 acted in trans to abrogate the DNA binding activity of the pre‐existing NF‐κB dimers, and that it preferentially sequestered RelB heterodimers, particularly the RelB:p52 dimer, compared to the RelA:p50 dimer.
2. Kinase assays revealing NIK:IKK1 activity during LTβR signaling in a time course in WT MEFs. Cytoplasmic extracts were prepared from stimulated cells and were subjected to immunoprecipitation using an anti‐NIK antibody. The immunoprecipitates were examined for the presence of NIK:IKK1 activity using recombinant p100_406–899 as substrate. The bar graph represents corresponding quantified kinase activities. Data are mean ± SEM of three independent experiments.
3. Experimentally derived NIK:IKK1 activity profile or NEMO:IKK activity profile was used as model input to simulate LTβR‐induced nuclear RelB activity or TNF‐induced nuclear RelA activity, respectively.
4. WT MEFs were stimulated with αLTβR in a time course before the cells were lysed and analyzed by immunoblotting with an anti‐p52/p100 antibody. The data are means ± SEM from three independent biological replicates.
5. Computational simulation recapitulating p100 and p52 levels observed in LTβR‐stimulated MEFs.
6. Kinase assays revealing NEMO:IKK activity induced during TNF signaling in a time course in WT MEFs. Cytoplasmic extracts were prepared from stimulated cells and were subjected to immunoprecipitation using an anti‐NEMO antibody. The immunoprecipitates were examined for the presence of NEMO:IKK activity using recombinant IκBα_1‐54 as substrate. The bar graph represents corresponding quantified kinase activities. Data are mean ± SEM of three independent experiments.
7. Computational simulation charting the expressions of Nfkb2 mRNA in response to 48 h of LTβR stimulation (L) or 12 h of TNF treatment (T). T_mut indicates TNF signaling in a cell system defective of RelA‐induced expressions of Nfkb2 mRNA.
8. The bar graph, related to Fig 3G, represents quantified signals corresponding to nuclear RelB, nuclear p100, and nuclear p52 from four independent experiments in MEFs subjected to indicated treatments. Data are means ± SEM. ***P < 0.001; *P < 0.05; ns, not significant (paired two‐tailed Student's t‐test).
9. The relative abundances of p52 and p100 in the RelB‐immunoprecipitates obtained from MEFs subjected to indicated treatments were quantified from three independent experiments and presented in the bar graph, which relates to Fig 3H. Data are means ± SEM. ***P < 0.001.

Figure 3. Molecular mechanism underlying TNF‐mediated inhibition of noncanonical NF‐κB signaling in LTβR‐stimulated cells.

• A
  Computational simulations charting the relative abundances of the indicated species in a time course in LTβR‐stimulated cells (blue line), and in cells that were additionally treated with TNF during the last 12 h of LTβR stimulation (dashed orange line).
• B
  Naïve or LTβR‐stimulated WT MEFs were treated with TNF before being subjected to qRT–PCR analysis of Nfkb2 mRNA abundance. Data represent three independent experiments.
• C
  WT MEFs, BLS4, and BLS12 cells were subjected to the indicated treatments, and whole‐cell extracts were analyzed by immunoblotting. Densitometric analysis of the relative abundances of p52 and p100 quantified from three independent experiments has been presented in the bar graph.
• D, E
  Kinase assay comparing the NIK:IKK1 activity present in cells stimulated through LTβR alone for 48 h (L), and in those stimulated through LTβR for 36 h and then additionally treated with TNF (L‐T) in the continuing presence of αLTβR for the indicated time periods (D) or for another 12 h (E). MEFs (D) and BLS4 as well as BLS12 (E) cells were treated before being subjected to immunoprecipitation with an anti‐NIK antibody. The immunoprecipitate was examined for the presence of NIK:IKK1 activity using recombinant p100_406–899, which contains the signal‐responsive serines, as substrate. Actin detected in the input extracts by immunoblot served as loading control. The bar graph in (D) represents quantified signals corresponding to phosphorylated p100 from three independent experiments.
• F
  Computational simulations charting the relative abundances of the indicated species in a time course in LTβR‐stimulated cells (blue line), and in cells that were additionally treated with TNF during the last 12 h of LTβR stimulation (dashed orange line). Dotted orange line represents a cell system devoid of RelA‐induced, but not constitutive, synthesis of Nfkb2 mRNA.
• G
  MEFs were stimulated with αLTβR for 36 h (L) or with TNF for 12 h (T) or were subjected for 12 h to TNF treatment in the continuing presence of αLTβR subsequent to 36 h of stimulation through LTβR alone (L‐T). Nuclear extracts were prepared and analyzed by immunoblot with antibodies against the indicated proteins. Nuclear export inhibitor LMB was added to the culture media 3 h before harvesting the cells. TFIID served as loading control. Data represent four independent experiments.
• H
  Immunoblot of RelB‐immunoprecipitates obtained using whole‐cell extracts derived from stimulated MEFs and normalized for the RelB content. Data represent three biological replicates.

Data information: In all panels, data are means ± SEM. **P < 0.01; *P < 0.05 (paired two‐tailed Student's t‐test). Source data are available online for this figure.

To address the discrepancy between experimental and computational studies, we expanded our biochemical analyses in MEFs, BLS4 cells, and BLS12 cells. As such, LTβR stimulation induced almost equivalent NIK:IKK1 activity at 36 and 48 h post‐stimulation and promoted progressive accumulation of p52 at the expense of p100 between these time points (Fig EV3B and D). In contrast to the prediction by our computational model, our immunoblot analysis showed that unprocessed p100 potently accumulated with an accompanying modest decrease in the abundance of p52 in MEFs, which were treated with TNF for 12 h in the continuing presence of αLTβR subsequent to 36 h of stimulation through LTβR alone (Fig 3C). The cellular level of RelB, if anything, was increased upon TNF treatment. Strikingly, TNF treatment gradually reduced the NIK:IKK1 activity present in these LTβR‐stimulated cells to basal levels (Fig 3D). We also stimulated BLS4 cells and BLS12 cells for 36 h through LTβR and then treated these cells with TNF for 12 h in the continuing presence of αLTβR. TNF treatment of these lymphoid stromal cells for 12 h similarly abolished the LTβR‐induced NIK:IKK1 activity (Fig 3E).

Use of this attenuated NIK:IKK1 activity in the combinatorial regime readily improved the computational model performance: simulations closely mirrored the decrease in the RelB activity and the increase in the level of p100 experimentally observed in LTβR‐stimulated cells during TNF signaling (Fig 3F). Modeling studies further predicted that unprocessed p100, accumulated in response to TNF, would translocate into the nucleus and would sequester the RelB dimers as multimeric IκBδ. Interestingly, despite the attenuated NIK:IKK1 activity as input, an absence of RelA‐induced Nfkb2 mRNA synthesis in our computational model prevented TNF from inducing the accumulation of p100 and inhibiting the LTβR‐stimulated RelB activity.

Our immunoblot analyses consistently showed that TNF induced the nuclear accumulation of p100 and concomitantly depleted RelB and p52 from the nucleus of MEFs subjected to prior stimulation through LTβR for 36 h, despite continuing presence of αLTβR (Figs 3G and EV3H). The nuclear export inhibitor Leptomycin B (LMB) further elevated the nuclear level of p100 and trapped RelB as well as p52 in the nucleus. These data indicated that p100‐IκBδ actively exported these NF‐κB subunits from the nucleus of LTβR‐stimulated cells in response to TNF to abrogate the RelB DNA binding activity. Immunoprecipitated RelB complexes revealed that LTβR signal promoted formation of the RelB:p52 dimer at the expense of the association between RelB and p100. Indeed, TNF treatment of LTβR‐stimulated cells restored the binding of p100 to RelB, while preserving the interaction between RelB and p52 (Figs 3H and EV3I). Our biochemical and mathematical studies illustrate a multitier inhibition mechanism, which involves distinct tiers of the NF‐κB network, employed by TNF to disrupt noncanonical NF‐κB signaling. TNF treatment of LTβR‐stimulated cells induces the synthesis of Nfkb2 mRNA engaging the transcriptional regulatory module, and also inhibits the activity of NIK:IKK1 involving receptor proximal regulatory mechanisms. These processes collectively accumulate unprocessed p100, which exports the pre‐existing RelB dimers from the nucleus as IκBδ.

TNF stabilizes TRAF2 and TRAF3 to deplete NIK from LTβR‐stimulated cells

In resting cells, a complex composed of TNF receptor‐associated factor 2 (TRAF2), TRAF3, and cellular inhibitor of apoptosis 1 or 2 (cIAP1/2) mediates K48‐linked polyubiquitination of NIK that leads to its proteasomal degradation (Vallabhapurapu et al, 2008; Zarnegar et al, 2008). Activated LTβR recruits this complex through its interaction with TRAF2 and TRAF3 (Sanjo et al, 2010). In the LTβR‐associated complex, TRAF2 ubiquitinates TRAF3 and undergoes auto‐ubiquitination that promotes the degradation of both TRAF3 and TRAF2. Therefore, LTβR activation rescues NIK from TRAF2:TRAF3‐dependent constitutive degradation. Because the cellular abundance of NIK primarily determines the activity of the NIK:IKK1 complex, we examined NIK levels in our stimulation regimes by immunoblotting. NIK was almost undetectable in untreated cells, but accumulated upon LTβR stimulation, and TNF treatment completely depleted NIK from LTβR‐stimulated MEFs (Fig 4A). Interestingly, TNF depleted NIK in LTβR‐stimulated MEFs despite inducing the expression of TRAF1 (Fig EV4A), which was shown to stabilize NIK from the degradation by TRAF2 and TRAF3 (Choudhary et al, 2013). On the other hand, chronic LTβR stimulation of MEFs led to protracted reduction in the levels of TRAF2 and TRAF3 that were gradually restored upon TNF treatment of these LTβR‐stimulated MEFs (Figs 4B and EV4B). Likewise, TNF restored the levels of TRAF2 and TRAF3 in LTβR‐stimulated lymphoid stromal cells (Fig 4C). However, TNF treatment did not significantly alter the levels of LTβR (Fig EV4C) or TRAF2 as well as TRAF3 mRNAs (Fig EV4D).

Figure 4. TNF‐induced K63‐linked ubiquitination of TRAF2 disrupts the LTβR signaling complex.

• A
  Immunoblot analysis of NIK in MEFs subjected to indicated treatments. Fivefold more extracts were used for detecting NIK.
• B, C
  Immunoblot depicting the abundances of TRAF2 and TRAF3 in MEFs (B) and in BLS4 as well as BLS12 cells (C).
• D, E
  MEFs were subjected to the indicated treatments and additionally treated with MG132 for 3 h before harvesting the cells. Cell extracts were subjected to immunoprecipitation under denaturing condition with an anti‐TRAF2 (D) or an anti‐TRAF3 (E) antibody, and the immunoprecipitates were examined by immunoblotting for the presence of K48‐linked and K63‐linked polyubiquitin chains.
• F
  MEFs were treated before being subjected to immunoprecipitation of TRAF2 under native condition. TRAF2 coimmunoprecipitates were analyzed by immunoblotting.

Data information: In (A–C), data are means ± SEM. *P < 0.05 (paired two‐tailed Student's t‐test). The bar graphs represent corresponding signal intensities quantified from three independent experiments.Source data are available online for this figure.

Figure EV4. Biochemical analyses of LTβR‐stimulated MEFs treated with TNF .

• A
  WT MEFs were stimulated through LTβR for 36 h and then were treated with TNF for the indicated times in the continuing presence of αLTβR. Subsequently, the cells were lysed and analyzed by immunoblotting using an anti‐TRAF1 antibody. The data represent three biological replicates.
• B
  WT MEFs were stimulated with αLTβR in a time course before being analyzed by immunoblotting for TRAF2 and TRAF3. The quantified data represent means ± SEM from three biological replicates. ***P < 0.001; *P < 0.05 (paired two‐tailed Student's t‐test).
• C, D
  WT MEFs were subjected to the indicated stimulations before the cells were lysed and analyzed by qRT–PCR for the abundance of mRNA encoding LTβR (C) or those encoding TRAF2 and TRAF3 (D). Data are means from three (C) or four (D) independent experiments ± SEM.
• E
  Immunoblot demonstrating TRAF2 and TRAF3 levels in WT MEFs subjected to αLTβR treatment for 48 h in the absence or the presence of MG132. MG132 was added to the culture media at 3 h prior harvesting cells. Our data, representative of two experimental replicates, confirm that TRAF2 and TRAF3 are degraded by proteasome in response to LTβR stimulation.
• F
  Immunoblot revealing TRAF3 level in WT MEFs infected with lentivirus particles expressing shRNA targeting Traf3 expression. A panel of lentiviral constructs were tested, and RMM4431‐200349061 and RMM4431‐200398580 (GE Dharmacon, Lafayette, CO, USA) were used for studies described in the main text and in the Appendix, respectively.
• G
  NIK:IKK1 activity measured during signaling in WT MEFs subjected to knockdown of Traf3 expression using shRNA RMM4431‐200398580.
• H
  Relative levels of CXCL13 and CXCL12 mRNAs during signaling in WT MEFs subjected to TRAF3 knockdown using shRNA RMM4431‐200398580. The abundances of mRNAs were measured from three experimental replicates. Data are means ± SEM.
• I
  The parental Nfkb2 ^−/− MEFs were transduced with retrovirus generated using either empty pBabe‐puro vector (EV) or with a pBabe‐based plasmid expressing the mature p52 subunit (p52‐Tg) before being subjected to puromycin selection and immunoblot analysis. Two separate cell lines expressing p52‐Tg in the Nfkb2 ^−/− background were generated (#1 and #2).
• J, K
  EMSA demonstrating the nuclear NF‐κB activity induced in Nfkb2 ^−/− MEFs expressing p52 from a transgene in response to the indicated cell treatments (top, J). By ablating the RelA DNA binding activity using an anti‐RelA antibody, RelB complexes were revealed (middle panel, J). The DNA binding activity of Oct1 served as a loading control. U denotes untreated cells; L represents cells treated with αLTβR for 48 h. For the combinatorial stimulation regime, cells were stimulated using αLTβR for 36 h and then were additionally treated with TNF for either 0.5 h (L‐T 0.5 h) or 12 h (L‐T 12 h). The data represent two biological replicates. The composition of NF‐κB DNA binding complexes induced in p52‐Tg/Nfkb2 ^−/− MEFs subjected to the indicated stimulation regime were examined by supershift analysis using the indicated antibodies (K). “#” represent antibody supershifted complexes.

Next, we immunoprecipitated TRAF2 and TRAF3 proteins and examined their ubiquitination status by immunoblotting. As proteasome degrades these TRAFs in LTβR‐stimulated cells (Fig EV4E), we supplemented the cell culture media with the proteasome inhibitor MG132 to ensure uniform abundances of TRAF2 and TRAF3 under various treatment regimes. Corroborating the earlier study (Sanjo et al, 2010), LTβR stimulation induced degradative K48‐linked polyubiquitination of TRAF2 and TRAF3 (Fig 4D and E). Intriguingly, TNF treatment of LTβR‐stimulated MEFs triggered the assembly of non‐degradative K63‐linked polyubiquitins on TRAF2 at the expense of K48‐linked chains (Fig 4D). As reported (Habelhah et al, 2004; Li et al, 2009), TNF treatment by itself also induced K63‐polyubiquitination of TRAF2. Although TNF treatment alone did not induce K63‐linked polyubiquitination of TRAF3, it prevented K48‐polyubiquitination of TRAF3 in LTβR‐stimulated cells (Fig 4E). Our coimmunoprecipitation analyses ascertained that TRAF2 interacted with TRAF3 in untreated cells, and was recruited to LTβR in response to αLTβR stimulation (Fig 4F). TNF treatment obstructed the recruitment of TRAF2 to the activated LTβR, but did not impede TRAF2 binding to TRAF3 (Fig 4F). Taken together, our data indicate that TNF signal dominantly mediates K63‐polyubiquitination of TRAF2, which interacts with TRAF3, but does not bind to LTβR. This impedes LTβR‐associated K48‐polyubiquitination and degradation of these TRAFs. Our data suggest that TNF‐modified TRAF2, in association with TRAF3, instead promotes the degradation of NIK in LTβR‐stimulated cells.

Depletion of TRAF3 or deficiency of p100 mitigates TNF‐mediated inhibition of noncanonical NF‐κB signaling

Our biochemical analyses implied that stabilization of TRAF2 and TRAF3, and accumulation of p100‐IκBδ accounted for TNF‐mediated suppression of noncanonical signaling. To genetically substantiate the proposed mechanism, we first subjected WT MEFs to shRNA‐mediated depletion of TRAF3 (Fig EV4F), which anchors NIK to the degradative polyubiquitinating complex. As suggested earlier (He et al, 2006; Shih et al, 2012), TRAF3 depletion raised the basal NIK:IKK1 activity by threefold (Figs 5A and EV4G) that promoted p100 processing into p52, and produced constitutive RelB activity (Fig 5B and C). LTβR stimulation of TRAF3‐depleted cells only modestly augmented this constitutive noncanonical signaling. TNF treatment of LTβR‐stimulated TRAF3‐depleted MEFs failed to attenuate the NIK:IKK1 activity; as also predicted by our computational model (Fig 3A), this unrestricted NIK:IKK1 activity led to subtly enhanced RelB activity in response to TNF (Fig 5A–C). TRAF3 depletion resulted in the increased basal expressions of the RelB‐target homeostatic chemokines, whose abundances were not discernibly altered upon LTβR stimulation (Figs 5D and EV4H). Indeed, TNF treatment was unable to suppress the expressions of these chemokines in TRAF3‐depleted MEFs.

Figure 5. Noncanonical NF‐κB signaling is resilient to TNF in TRAF3‐depleted or p100‐deficient cells.

1. NIK:IKK1 activity measured during signaling in WT MEFs subjected to shRNA‐mediated knockdown of Traf3 expression. WT MEFs expressing a control shRNA served as a control.
2. TRAF3‐depleted MEFs were subjected to indicated treatments, and extracts were analyzed by immunoblotting using antibodies against p52/p100.
3. The nuclear NF‐κB activities induced during signaling in WT MEFs subjected to Traf3 knockdown were examined by EMSA.
4. Relative levels of CXCL13 and CXCL12 mRNAs during signaling in WT MEFs subjected to TRAF3 knockdown. The abundances of mRNAs were measured from four independent experiments.
5. Immunoblot revealing TRAF2 and TRAF3 levels in Nfkb2 ^−/− MEFs subjected to indicated treatments.
6. NIK:IKK1 activity measured during signaling either in Nfkb2 ^−/− MEFs. Unstimulated WT MEFs served as a control.
7. The nuclear NF‐κB activities induced during signaling in Nfkb2 ^−/− MEFs were examined by EMSA.
8. Relative levels of CXCL13 and CXCL12 mRNAs during signaling in Nfkb2 ^−/− MEFs. The abundances of mRNAs were measured from four independent experiments.

Data information: In all panels, data are means ± SEM. **P < 0.01; *P < 0.05 (paired two‐tailed Student's t‐test). The bar graphs represent corresponding signal intensities quantified from three independent experiments. Source data are available online for this figure.

We then examined Nfkb2 ^−/− MEFs, which is devoid of both p52 and p100‐IκBδ. LTβR stimulation degraded TRAF2 and TRAF3, and induced the NIK:IKK1 activity in Nfkb2 ^−/− MEFs (Fig 5E and F). TNF treatment restored the levels of TRAFs and efficiently attenuated the NIK:IKK1 activity in LTβR‐stimulated Nfkb2 ^−/− cells. Consistent to previous reports (Lo et al, 2006; Basak et al, 2008), Nfkb2 ^−/− MEFs displayed constitutive RelB:p50 activity, which resulted in the elevated basal expressions of CXCL13 as well as CXCL12, and these did not substantially change upon LTβR stimulation (Fig 5G and H). Despite the attenuated NIK:IKK1 activity, a lack of p100‐IκBδ alleviated TNF‐mediated suppressions of the RelB activity and chemokine gene expressions. In fact, we detected an increase in the RelB activity in response to TNF in Nfkb2 ^−/− MEFs that was noted earlier and attributed to the autoregulatory RelB synthesis (Roy et al, 2017). We also expressed the mature p52 subunit from a transgene in Nfkb2 ^−/− MEFs that led to a constitutive RelB activity mostly composed of the RelB:p52 dimer (Fig EV4I–K). Consistent to our analyses involving the parental Nfkb2 ^−/− MEFs, LTβR stimulation did not enhance and subsequent TNF treatment was unable to downregulate this constitutive RelB:p52 activity in the absence of inhibitory p100 in this engineered cell line (Fig EV4J). These studies causally link TRAFs and p100‐IκBδ to TNF‐mediated subversion of noncanonical RelB NF‐κB signaling.

p100/Nfkb2 is required for suppressing the expression of homeostatic chemokines in OVA–CFA‐immunized mice

We further analyzed Nfkb2 ^−/− mice to determine whether p100‐IκBδ downregulated the expressions of RelB‐target homeostatic chemokines in inflamed SLOs. Because Nfkb2 ^−/− mice lack popliteal LNs (Carragher et al, 2004), we measured the splenic abundances of these chemokines. Administration of OVA–CFA through intraperitoneal route significantly decreased the splenic abundances of these chemokine proteins and mRNAs in WT mice (Fig 6A and B). As reported earlier (Lo et al, 2006), the compensatory RelB:p50 activity mediated the expressions of CCL21, CXCL13, and CXCL12 in unimmunized Nfkb2 ^−/− mice, albeit at a relatively reduced level (Fig 6A). Remarkably, immunization of Nfkb2 ^−/− mice failed to suppress the expressions of these chemokines. Importantly, OVA–CFA immunization led to equivalent accumulation of TNF in the spleen of WT and Nfkb2 ^−/− mice (Fig EV5). Analyses of reciprocal bone marrow chimeras, created by using WT and Nfkb2 ^−/− mice, confirmed a stromal requirement of Nfkb2 for downregulating homeostatic chemokines in inflamed SLOs (Fig 6C).

Figure 6. Immunization of Nfkb2 ^−/− mice does not deplete homeostatic chemokines or impede naïve lymphocyte ingress in the spleen.

• A, B
  WT and Nfkb2 ^−/− (n = 6 for each) mice were administered with OVA–CFA through intraperitoneal route, and the splenic abundances of the indicated chemokines at day 2 post‐immunization were determined by ELISA (A) or the relative levels of chemokine mRNAs were estimated by qRT–PCR (B). WT (n = 6) and Nfkb2 ^−/− (n = 4) mice administered with saline were used as respective controls. ns, not significant.
• C
  ELISA showing levels of CCL21 and CXCL13 in the indicated bone marrow chimeras (n = 5) at day 2 post‐immunization.
• D
  Naïve CD45.1⁺ B or T lymphocytes were transferred into unimmunized or day 2 immunized WT or Nfkb2 ^−/− mice (n = 3 for each). After another 8–12 h, spleens were harvested, spleen sections were stained for ER‐TR7 (green) and CD45.1 (red), and visualized through fluorescence microscope. The scale bar indicates 100 μm. Objective magnification, 20×; RP, red pulp; WP, white pulp. The bar graph shows the quantification of CD45.1⁺ cells in the white pulp area of the spleen. Five fields/section were used for quantification and presented as mean ± SEM.

Data information: In all panels, data are means ± SEM. ***P < 0.001; **P < 0.01; *P < 0.05 (paired two‐tailed Student's t‐test).

Figure EV5. The abundance of TNF in the spleen of immunized WT and Nfkb2 ^−/− mice.

WT and Nfkb2 ^−/− mice (n = 6 for each) were administered with OVA–CFA through intraperitoneal route, spleens were harvested at the indicated times post‐immunization, and the abundance of TNF was measured by ELISA. Data are means ± SEM. **P < 0.01 (paired two‐tailed Student's t‐test).

Migration of lymphocytes into the white pulp area of the spleen is dictated by homeostatic chemokines (Cyster, 2005). We adoptively transferred naïve CD45.1⁺ B or T lymphocytes into untreated or immunized mice, and probed the presence of these transferred cells in the white pulp. Indeed, prior immunization with OVA–CFA diminished the abundance of these transferred cells in the white pulp in WT, but not Nfkb2 ^−/− mice (Fig 6D). As reported earlier (Acton et al, 2014), the numbers and the morphology of FRCs were not discernibly altered at day 2 post‐immunization with OVA–CFA. These analyses suggest that p100‐IκBδ is necessary for downregulating homeostatic chemokines in inflamed SLOs and prevents continuing ingress of naïve lymphocytes.

Discussion

Disruption of lymphoid stromal cells by microbial pathogens is thought to abrogate the expression of homeostatic chemokines and thereby prevent the ingress of naïve lymphocytes in reactive SLOs (Chang & Turley, 2015). We utilized non‐infectious immunogen OVA–CFA, which elicits potent inflammatory responses and triggers well‐characterized lymphatic drainage. Our study revealed similar downregulation of homeostatic chemokines in OVA–CFA‐immunized mice. TNF promotes the recruitment of immune cells to the infected tissue by inducing the expression of pro‐inflammatory chemokines. However, TNF also specifically accumulates in draining SLOs, presumably from local cellular sources (McLachlan et al, 2003; Tumanov et al, 2010). We found that TNF subverted LTβR‐stimulated noncanonical RelB NF‐κB signaling and restricted largely the expression of homeostatic chemokines as well as the trafficking of lymphocytes in inflamed SLOs of OVA–CFA‐immunized mice. Therefore, our study indicated that host‐derived cytokine TNF, independent of infectious pathogens, might modulate these immune homeostatic processes.

Based on the measurement in day 8 infected IFNγ‐null mice, Mueller et al (2007) earlier reported that IFNγ is involved in downregulating homeostatic chemokines and in limiting the ingress of naïve lymphocytes during LCMV infection. Our study revealed that these chemokines were depleted within 2 days of OVA–CFA immunization in both WT and IFNγ‐null mice, although restoration of their expressions was somewhat accelerated in IFNγ‐null mice. It appears that potent pro‐inflammatory responses elicited by OVA–CFA favors largely TNF‐dependent rapid downregulation of homeostatic chemokines in reactive LNs, while IFNγ‐mediated mechanisms predominantly function during virus infections in a slower timescale. Etanercept restored completely the expressions of B‐cell chemokines CXCL13 as well as CXCL12, and rescued fully the frequency of adoptively transferred B cells in reactive LNs of WT mice. However, etanercept was only moderately effective in preventing the downregulation of the T‐cell chemokine CCL21 and in restoring the frequency of transferred T lymphocytes. Therefore, it is possible that other mechanisms, albeit in part, contribute to the downregulation of CCL21 mRNA in reactive LNs. Future studies should also focus on distinguishing between different B‐ and T‐cell subsets with respect to their ingress in draining LNs.

TRAF2 and TRAF3 have non‐redundant roles in resting cells in suppressing the activity of NIK, the upstream kinase of the noncanonical pathway (He et al, 2006). Activated LTβR recruits TRAF2 and TRAF3 that promotes their K48‐linked polyubiquitination and degradation. TRAF2 also reversibly associates with the complex1 formed by TNFR1 and transduces signal to the canonical NF‐κB pathway, although this TRAF2 function is redundant to TRAF5 (Wertz & Dixit, 2010). Consistently, TNF treatment rapidly activated RelA in our LTβR‐stimulated cells, despite the reduced level of TRAF2 at the early time points. TNFR1 was also shown to stimulate K63‐polyubiquitination of TRAF2 (Li et al, 2009), but the precise role of this polyubiquitination event remains unclear. Our results demonstrated that K63‐polyubiquitinated TRAF2 interacted with TRAF3, but this complex was unable to bind to LTβR. As a consequence, TNF treatment of LTβR‐stimulated cells led to the slow accumulation of de novo synthesized TRAF2 in its modified form, which inactivated NIK in cooperation with TRAF3 (Fig 7).

Figure 7. A mechanistic model explaining TNF‐mediated suppression of noncanonical NF‐κB signaling.

In resting cells, TRAF2:TRAF3:cIAP complex promotes the degradation of NIK, and p100‐IκBδ retains RelB and other NF‐κB monomers. Activation of LTβR leads to the proteolysis of TRAF2 and TRAF3 that rescues NIK from the constitutive degradation. NIK in association with IKK1 induces the processing of p100 into p52 that liberates the RelB heterodimers. TNF signal renders TRAF2 and TRAF3 resistant to degradation in LTβR‐stimulated cells resulting in the inactivation of NIK. In addition, TNF‐activated RelA dimers induce the transcription of Nfkb2 mRNA, which encodes p100. Therefore, TNF treatment of LTβR‐stimulated cells potently accumulates the precursor p100, which sequesters the pre‐existing nuclear RelB dimers as inhibitory IκBδ and terminates RelB‐mediated expressions of homeostatic chemokines. The proposed mechanism impedes continuing ingress of naïve lymphocytes in reactive SLOs.

Our computational studies indicated that collaboration between NIK‐inactivation and p100‐dependent mechanisms was essential for TNF to abrogate LTβR‐stimulated RelB activity. TNF attenuated the NIK:IKK1 activity and concurrently induced the expression of Nfkb2 mRNA through the canonical pathway; these together accumulated unprocessed p100 in LTβR‐stimulated cells. It was shown that p100 produced in LPS‐stimulated fibroblasts inhibits the RelA activity as IκBδ (Shih et al, 2009). Biochemical studies confirmed that p100‐IκBδ sequesters RelB and other NF‐κB subunits in resting cells (Tao et al, 2014). Signal‐induced processing of p100 promotes the noncanonical RelB activity. Our analyses identified an additional regulatory role of p100 in abrogating the pre‐existing RelB activity induced by noncanonical signaling in LTβR‐stimulated cells (Fig 7). In a parallel to the inhibition by the classical NF‐κB inhibitor IκBα, p100‐IκBδ not only sequestered RelB in the cytoplasm of resting cells, but also exported the RelB dimers from the nucleus of LTβR‐activated cells in response to TNF. The nuclear export also facilitated the reactivation of RelB; the NIK:IKK1 activity, restored by LTβR upon cessation of TNF signaling, liberated the RelB dimers from this p100 inhibited cytoplasmic complex. A constitutive RelB:p50 nuclear activity functionally compensates for the absence of the LTβR‐stimulated RelB:p50 dimer in Nfkb2 ^−/− cells. Indeed, TNF was ineffective in curtailing this RelB:p50 activity in Nfkb2 ^−/− MEFs; an absence of p100‐IκBδ in stromal cells prevented the downregulation of homeostatic chemokines and the reduction of naïve lymphocytes in inflamed SLOs of OVA–CFA‐immunized mice.

Initial antigen recognition in SLOs is followed by intranodal repositioning of activated lymphocytes that facilitate the interaction between T and B lymphocytes, and promotes the formation of germinal centers (GCs), which are important for humoral responses (Cyster, 2005). Interestingly, both TNF and Nfkb2 were shown to be important for establishing GCs and for the humoral immunity (Caamano et al, 1998; Tumanov et al, 2010). Recent investigation indicated that the Nfkb2 expression in B cells is dispensable for the GC formation (De Silva et al, 2016). We hypothesize that the inhibition of noncanonical NF‐κB signaling in stromal cells by the newly described TNF‐p100 pathway orchestrates ongoing immune response; downregulation of homeostatic chemokines relieves the retention of lymphocytes within their respective compartments and thereby favors interstitial movement of activated T and B lymphocytes, GC formation, and humoral responses. Consistently, TNF was shown to inhibit the production of CXCL12 by bone marrow stromal cells and mobilize osteoclast precursors into the circulation in the animal model of inflammatory arthritis (Zhang et al, 2008). By blocking the ingress of naïve lymphocytes in reactive SLOs, this inhibitory mechanism likely also preserves local resources for rapid expansion of activated lymphocytes. Therefore, it appears that acute TNF signal, generated upon immune activation, reinforces ongoing adaptive immune responses by modulating homeostatic chemokines. However, suppression of homing of naïve lymphocyte generates potential vulnerability to subsequent microbial infections. Indeed, previous studies demonstrated that immune challenge of mice 8 days after LCMV infection fails to produce sufficient CD8⁺ T cells and neutralizing IgG against the secondary antigen (Mueller et al, 2007; Scandella et al, 2008).

In sum, we illustrate that an inhibitory mechanism involving TNF and p100 controls homeostatic chemokine expressions and naïve lymphocyte ingress during immune responses. While autoimmune and neoplastic diseases are associated with the chronically elevated expression of TNF, therapeutic application of the TNF‐inhibitor etanercept in psoriasis led to lymphadenopathy (Hurley et al, 2008). Of note, prevalence of lymphadenopathy in individuals with Sjögren's syndrome correlated with the increased levels of CCL21 (Lee et al, 2017). Future studies ought to examine whether the proposed regulations of homeostatic chemokines by the newly described TNF‐p100 pathway contribute to the pathogenesis of human ailments associated with inflammation.

Materials and Methods

Mice, cells, and plasmids

WT and gene‐deficient C57BL/6 mice were housed at SAF, NII, and used adhering to the Institutional guidelines (approval no. IAEC 401/16). MEFs were obtained from E12.5 to E14.5 embryos. BLS4 and BLS12 cell lines were kind gift from Tomoya Katakai, Kyoto University. αLTβR was a generous gift from Jeff Browning, Boston University and Adrian Papandile, Biogen. Lentivirus particles were produced in 293T cells using shRNA constructs from GE Dharmacon, USA.

In vivo studies

Eight‐ to 10‐week‐old mice were administered with 100–200 μg of OVA–CFA (1:1 emulsion), and SLOs were harvested. For certain experiments, 500–1,000 μg of etanercept (Pfizer, UK) was administered. To score lymphocyte ingress, naïve CD45.1⁺ splenocytes (5–20 × 10⁶) were transferred retro‐orbitally into the recipient mice. Hematopoietic cells were harvested from pLNs, the frequency of transferred CD45.1⁺ B or T cells in pLNs was measured in FACS (BD Biosciences) using anti‐CD45.1 Ab and anti‐CD45R or anti‐CD90 Ab (eBioscience, USA), and the data were analyzed using FlowJo v 9.5. Alternately, CD45.1⁺ B or T cells were first purified using lymphocyte isolation kits (Miltenyi Biotec), 10–20 × 10⁶ cells were transferred, and subsequently cryosections were obtained from the spleen. Stromal reticular fibre network was visualized using anti‐ERTR7 Ab (Abcam, UK) and Alexa flour 488 conjugated secondary Ab (Invitrogen). The presence of transferred lymphocytes in the white pulp was scored using anti‐CD45.1 Ab conjugated to Alexa flour 594 (Biolegend, USA). The image was captured in Axioimager Z1 fluorescence microscope (Zeiss, USA). The bone marrow chimeras were subjected to immunization after ~8 weeks of reconstitution.

Gene expression analyses and ELISA

Total RNA was isolated using RNeasy kit (Qiagen, Germany) from tissues or cultured cells, and was subjected to qRT–PCR (see Appendix Table S1 for primer descriptions). Also tissue homogenates were prepared, normalized for the total protein content (BCA kit; Thermo Fisher, USA), and utilized in ELISA for detecting CCL21, CXCL13, CXCL12, TNF (DuoSet kit; R&D Systems, USA), and IFNγ (BD Bioscience, USA).

Biochemical analyses

Cells were treated with 0.1 μg/ml of αLTβR, 1 ng/ml of TNF (Roche, Switzerland) or subjected to combinatorial stimulations. Nuclear, cytoplasmic or whole‐cell extracts were used in EMSA, supershift analysis, immunoblotting, and immunoprecipitation‐based studies, as described (Banoth et al, 2015). RelB DNA binding activity was also measured using TransAM flexi NF‐κB family kit (Cat. No. 43298, Active Motif, USA) according to the manufacturer's instructions. For detecting ubiquitinylated proteins, cells extract was prepared in denaturing buffer (Sanjo et al, 2010). NF‐κB/IκB antibodies have been described (Roy et al, 2017). Antibodies against TRAF2 (sc876), TRAF3 (sc949), LTβR (sc398929) were from Santa Cruz Biotechnology, USA; antibodies against TRAF1 (4710), p52/p100 (4882), K48‐linked polyubiquitin (8081), and K63‐linked polyubiquitin (5621) were from Cell Signaling Technology, USA. For immunoblotting the coimmunoprecipitates, TrueBlot (18‐8816‐33, Rockland) was used. The gel images were acquired using PhosphorImager (GE Amersham, UK) and quantified in ImageQuant 5.2. For certain experiments, cells were treated with 20 μM of MG132 (Sigma‐Aldrich, USA) or 20 nM of Leptomycin B (Santa Cruz Biotechnology). NIK coimmunoprecipitates derived from the cytoplasmic extracts were incubated with γ³²P‐ATP and recombinant p100_406–899 (BioBharati Life Sciences, India) for measuring the NIK:IKK1 activity (Banoth et al, 2015).

Computational modeling

Based on our experimental data and the literature, we included the description of p100‐IκBδ‐mediated inhibition of the RelB:p52 dimer into the mathematical model published earlier by Shih et al (2012) and accordingly revised the model (see Appendix Supplementary Methods for details). We used numerical input derived from experimental kinase activities in our simulations. The model was simulated using ode15s in MATLAB (2012b, Mathworks, USA).

Statistical analysis

Error bars are shown as SEM of three to five replicates for biochemical experiments and 4–14 mice for animal studies. Quantified data are means ± SEM, and paired two‐tailed Student's t‐test was used for calculating statistical significance, where “*”, “**”, and “***” indicate P < 0.05, P < 0.01, and P < 0.001, respectively.

Author contributions

TM conducted the cell‐based analyses with the assistance from SSB, MC, PR, and the guidance from SB. BC carried out the in silico studies under the supervision of SB. TM and AD performed animal experiments under the supervision of AG, VB, SR, and SB. TM wrote the manuscript with SB.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Expanded View Figures PDF

Review Process File

Source Data for Figure 2

Source Data for Figure 3

Source Data for Figure 4

Source Data for Figure 5

The EMBO Journal (2017) 36: 3501–3516