IκBε is a key regulator of B-cell expansion by providing negative feedback on cRel and RelA in a stimulus-specific manner

Cell isolation and culture

Spleens were harvested from C57Bl6 mice wild-type (Jackson Labs, Bar Harbor, MN), and C57Bl6 IκBε^−/− mice (2). The collected spleens were homogenized using frosted glass slide grinding. For B cell isolation, homogenized splenocytes were incubated with anti-CD43 (Ly-48) microbeads for 15 minutes at room temperature. Following this incubation, the cells were washed with Hanks Buffered salt solution (HBSS) (Gibco 14170) containing 1% FCS, 10mM HEPES (Gibco 15630) and 1% FCS (Sigma F2442) and separated over a magnetic column (LS column MiltenyiBiotec 130-042-401). For B cells, purity was determined by flow cytometry using PE anti-B220 (eBioscience 12-0452-83), FITC anti-CD3 (eBioscience 11-0031-82), APC anti-CD4 (eBioscience 17-0041-83) and PerCP anti-CD8 (ebioscience 46-0081-82). Purity was consistently found to be between 92% and 95% (data not shown). For experiments where separation of marginal zone (MZ) B cell and follicular B cell was performed, anti-CD43 magnetically separated B cells were stained with APC anti-CD9 (eBioscience 17-0091-82) and B cell population separation was performed using a FACS-ARIA II cell sorter (BD Bioscience). Complete media consisting of RPMI-1640 (Gibco 11875), 10 mM HEPES (Gibco 15630), 1 mM Sodium Pyruvate (Gibco 11360), 1 mM non-essential amino acids (Gibco 11140), 0.055 mM β-mercaptoethanol (Gibco 21985), 100 units Penicillin/Streptomycin (Gibco 10378016) and 0.3 mg/ml glutamine was used to culture either B cells. B cells were stimulated with either 10 μg/ml anti-IgM (Jackson laboratories 115-006-020) or 10 μg/ml LPS (Sigma L2630).

Flow cytometry analysis of cell proliferation and survival

Purified B cells were stained with 5 μM CFSE (Invitrogen C1157) and culture in complete media with previously mentioned stimuli (see above). At various time points B cells were collected and stained with 7AAD (Invitrogen A1310). The B cells were analyzed for proliferation and survival using a C6 Accuri flow cytometer (BD biosciences). Gating of live B cells was determined by forward scatter and side scatter properties. Differences in cell proliferation was measured using FlowJo (Tree star inc.), and FlowMax (26), which determines maximum-likelihood cellular parameter sets for fraction of responding cells, times to division, and times to death for generations 0 to 10.

EMSA and Supershifts

Nuclear extracts were generated from B cells using high salt extraction. In brief, purified B cells were incubated with a low salt buffer (10mM HEPEs pH 7.9 (Gibco), 10 mM KCl (Thermo Fisher Scientific P217), 0.1 mM EGTA (Sigma E-4378), 0.1 mM EDTA (Thermo Fisher Scientific S312), 1mM DTT (Thermo Fisher Scientific BP172-5), 1 mM PMSF (Sigma P7626), 5 μg/ml apoprotein (Sigma A1153), 5 μg/ml leupeptin (Sigma L2884), 1 μM pepstatin A (Sigma P5318)) for 10 minutes on ice. Following this incubation, the cells were disrupted through the addition of NP-40 (US Biological N3500) to a final concentration of 0.5% and votexing for 15 seconds. Nuclei were pelleted away from the cytoplasmic fraction by centrifugation at 15,000 rpm for 1 minute and the cytoplasmic fraction was pipetted into a separate tube. The remaining nuclei were disrupted by a 20 minute incubation at 4°C in a high salt buffer (20 mM HEPES pH 7.9 (Gibco), 400 mM NaCl (Thermo Fisher Scientific S671), 1 mM EGTA (Thermo Fisher Scientific), 1 mM EDTA (Thermo Fisher Scientific), 20% Glycerol (Thermo Fisher Scientific), 1 mM DTT (Thermo Fisher Scientific), 1 mM PMSF(Sigma)). The nuclear fraction was collected following centrifugation at 15,000 rpm for 5 minutes. Equal amounts of nuclear extracts (1 μg) were pre-incubated for 20 minutes on ice in the presence or absence of antibodies specific for RelA (Santa Cruz Biotechnology sc-372), RelB (Santa Cruz Biotechnology sc-226), cRel (Santa Cruz Biotechnologies sc-71) or in combinations as follows; RelA/RelB, RelA/cRel and RelB/cRel. Following the pre-incubation with antibodies, [γ–³²P]ATP (GE health) radio-labeled probe derived from HIV-κB sequence: 5′-GCTACAAGGGACTTTCCGCTGGGGACTTTCCAGGGAGG-3′ was added and incubated at room temperature for an additional 15 minutes. The resulting DNA/protein/antibody complexes were resolved by electrophoresis on a 5% non-denaturing polyacrylamide gel and exposed to storage phosphor screen (GE healthcare) overnight before image development on at Typhoon 9200 Variable Mode Imager (GE healthcare). Images were analyzed and quantitated using ImageQuant™ (GE Health).

Western blot analysis

Whole cell lysates were prepared using RIPA buffer lysis of B cells. Cytoplasmic extracts and nuclear extracts were prepared as previously described (27,28). The resulting lysates and extracts were run on either 10% SDS-PAGE gels or 5%–14% Criterion Tris-HCl Gel (Bio Rad). The following antibodies were used to identify the protein of interest: p65 (Santa Cruz Biotechnology), IκBα (Santa Cruz Biotechnology), cRel (Santa Cruz Biotechnology) and actin (Santa Cruz Biotechnology). The resulting proteins were detected using the Bio-Rad ChemiDoc XRS System and SuperSignal West Femto Substrate Maximum Sensitivity Substrate (Thermo Scientific) to detect chemiluminescence released by HRP-labeled secondary antibodies.

Cytokine Neutralization and Receptor blocking

CFSE labeled B cells isolated from either wild type or IκBε^−/− B cells were stimulated with either 10 μg/ml IgM or 10 μg/ml LPS in the presences of 2 μg/ml of the following antibodies: anti-mouse IL-1 alpha/IL-1F1 AF-400-NA (R&D systems), anti-mouse IL-1 beta/IL-1F2 AF-401-NA (R&D systems), anti- mouse IL-6 AF-406-NA (R&D systems), anti-mouse IL-1 receptor AF-480-NA (R&D systems), anti-mouse IL-6 receptor alpha AF1830 (R&D systems). Cell proliferation was measure using a C6 accuri flow cytometer (BD biosciences). Differences in cell proliferation was measured using FlowJo (Tree star inc.).

ELISA assays

Isolated B cells were plated at a concentration of 2 million cells per ml in the presences of 10 μg/ml IgM or 10 μg/ml LPS. At 2, 4, 8, 22, 34 and 48 hours cells were harvested, spun down and the supernatant was collected. The resulting supernatant was test for measured release of IL-1α (Mouse IL-1α ELISA MAX™ Deluxe, Biolegend) IL-1β (Mouse IL-1β ELISA MAX™, Biolegend) and Deluxe, IL-6 (Mouse IL-6 ELISA MAX™ Deluxe, Biolegend).

Transcriptome Analysis

Total RNA was isolated from IgM or LPS stimulated B cells isolated from wild type or IκBε^−/− B cells over a four point time course. mRNA was extracted from 2 μg total RNA using oligo(dT) magnetic beads and fragmented at high temperature using divalent cations. Next cDNA library were generated using the IlluminaTruSeq kits and quantitation was performed using the Roche Light Cycler 480. Sequencing was performed on Illumina’s HiSeq 2000, according to the manufacturer’s recommendations by the BIOGEM core facility located at University of California, San Diego. Reads were aligned to the mouse mm10 genome and RefSeq genes (PMID 12045153, PMID 12466850) with Tophat (PMID 19289445). Cufflinks and CummRbund were used to ascertain differential expression of genes. Gene differential FPKMs were obtained from the cuffdiff program in the Tuxedo RNA-seq analysis suite. In R, we took the log base 2 of the FPKM values and used the wild type zero hour to normalize. The subset of genes that were induced by two fold in at least one of the wild type time points were identified and used to identify gene induced by IgM or LPS stimulation. Hyperinduced genes of the IκBε^−/− samples were identified as being two fold greater than their corresponding wild type timepoint in two or more of the non-zero time points. Using R’s gplots package heat maps were created for fold change for all genes that made the above induction cut offs. Additionally, heat maps for latent genes induced no more than 0.3 log base 2 above their corresponding wild type time point were created. For both heat maps, the lists of genes mapped were recorded. The above process was done for both IgM and LPS stimulated conditions. Determination of NFκB motifs containing genes was performed using the NFκB PWMs previously identified by Bulyk ML et. al. (29) and converted into formats for HOMER motif discovery software (30). The promoter regions of genes found to be either hyper or latent induced were searched for the occurrence of these motifs. The percentage of genes from the identified gene lists that contained each motif versus the percent of promoters containing the NFκB motifs from genes identified as induced were graphed. Significance was determined using R’s stats package to do a Pearson’s Chi-squared test using the background counts as the expected values versus the percentage of hyper or latent induced genes containing the NFκB motifs.

Computational modeling of NFκB dimer activity

The computational model appends the previously published IκB models with the reactions that govern the generation of RelA- and cRel-containing NFκB dimers (Supplemental Figure 3). The model contains 36 species and 146 reactions (Supplemental Model Equations) governed by 74 parameters (Supplemental Table 1). The ODEs were solved numerically using MATLAB version R2013a (The MathWorks, Inc.) with subroutine ode15s, a variable order, multi-step solver. Prior to stimulation, the system was allowed to equilibrate from starting conditions to a steady state, defined as showing no concentration changes greater than 1% over a period of 4000 minutes. Stimulus-induced perturbation from the steady state was accomplished by direct modulation of IKK activity via a numerical input curve representing IgM or LPS stimulation (adapted from Werner 2008)(31). MATLAB model codes are available upon request.

Animal use

The animal protocols for this study were approved by the University of California, San Diego Animal Care and Use Committee.

IκBε-deficiency in B cell subsets results in increased stimulus-responsive proliferation and survival

As NFκB controls B cell expansion, we sought to determine IκB regulators that limit NFκB activity in B-cells and thus B cell proliferation. Examining whole-cell extracts, we found that while IκBα protein levels rapidly decrease upon B-cell stimulation with IgM or LPS, IκBε protein levels decrease only slightly after stimulation and increase at late timepoints (Supplementary Figure 1A), suggesting a role in post-induction attenuation of NFκB.

Using B cells magnetically purified from mixed splenocytes collected from wild type or IκBε^−/− mice, we examined B cell expansion following ex vivo stimulation with 10 μg/ml IgM or 10 μg/ml LPS using CFSE dye dilution. We found that B cells lacking IκBε displayed increased expansion with either stimulus compared to wild type B cells (Figure 1A). Several repeats of the CFSE experiments yielded highly reproducible results (Figure 1B). To determine if these differences were the result of a proliferation or survival defect we used the computational phenotyping tool FlowMax (26), which parameterizes a modified cyton model to CFSE time courses and yields maximum likelihood non-redundant cellular parameters, such as the percentage of cells entering the proliferative program (pF), the time to the first division (Tdiv0), and the time to death (Tdie0) of cells not entering the proliferative program (Figure 1C). Using FlowMax, we found that the CFSE data indicates that IκBε^−/− B cells are more likely to respond to the stimulus (pF₀) than wild type cells under the same conditions (Figure 1D). In response to IgM, 51.7% of IκBε^−/− B cells entered division compared to only 24% of wild type B cells. Following LPS stimulation 64% of the IκBε^−/− B cells entered division, compared to 52% of their wild type counterparts. As only non-responding cells are susceptible to death (32–34), and the Tdie0 and Tdiv0 parameters showed little change in the knockout, this suggested that the death rates in IκBε^−/− would be lower. Testing this prediction with 7AAD staining of responding cells at 24 hours, we indeed found lower percentages of dying cells in the IκBε^−/− than wild type populations (Figure 1E and Figure 1F), 33.3% vs. 54.8% in response to IgM, and 21.9 vs. 37.7% in response to LPS).

Examining the B-cell subsets in the spleen, we observed a higher percentage of marginal zone B cells as opposed to follicular B cell populations in IκBε^−/− mice compared to wild type controls (Figure 2A). As marginal zone B cell maturation is more sensitive to NFκB activity than follicular B cell maturation, this observation is consistent with elevated NFκB activity following IκBε ablation. It also prompted us to question if the skewed distributions were responsible for the ex vivo expansion phenotype. To address this concern, we purified wild type and IκBε^−/− marginal zone and follicular B cells away from each other, using APC anti-CD9 staining in conjunction with FACs cell sorting (35). The purity was determined to be consistently between 90% and 100% (Figure 2B). Purified follicular B cells whose physiological role is to respond to antigens, showed increased B-cell expansion in response to IgM stimulation when derived from IκBε^−/− mice. Similarly, marginal zone B cells whose physiological role is to monitor for circulating endotoxin showed increased proliferation when derived from IκBε^−/− mice (compared to wild type) and stimulated with LPS (Figure 2C and Figure 2D). These results indicate that the difference in proliferation and survival is a cell intrinsic B cell phenotype rather than a result of developmental changes that result in differences in the B cell subpopulation distributions.

IκBε provides negative feedback on both RelA- and cRel-containing dimers, albeit stimulus-specifically

Given the B cell proliferation phenotype of IκBε deficiency, we sought to determine how the loss of IκBε affected NFκB dimer activation. In previous studies, IκBε had been shown to bind preferentially to cRel (22,24), suggesting that IκBε-deficiency may preferentially affect cRel-containing dimer activity. We prepared nuclear extracts from wild type and IκBε^−/− B cells stimulated with IgM or LPS as previously described, and used them for electrophoretic mobility shift assays (EMSA) with a κB-site containing probe. NFκB heterodimers (all containing p50, but containing either RelA, cRel or RelB, see below) were found elevated in IκBε^−/−B cells at later time points (18 and 24 hours) when compared to wild type B cells (Figure 3A), consistent with the high level of induction of this inhibitor seen in immunoblots of wild type cells (Supplementary Figure 1A). Supershift analysis of these B cells nuclear extracts using specific antibodies for the three activation domain-containing Rel proteins, RelA, cRel and RelB, were employed to quantitate both the supershifted complex observed with one antibody, as well as the remaining non-supershifted complex when two antibodies were used to ablate the activities of their cognate Rel proteins. We found greatly increased levels of the cRel:p50 dimer activity in IκBε^−/−B cells extracts stimulated with either IgM or LPS (Figure 3B and 3C). Interestingly, examining RelA activity, we found that RelA nuclear activity was seen to be significantly increased in LPS-stimulated IκBε^−/−B cells when compared to wild type B cells, but not in the IgM stimulation condition (Figure 3B and 3C). No difference was seen the basal levels of RelA and cRel (Supplementary Figure 1B), suggesting that this affect is induced during stimulation of the B cells. We employed immunoblots of nuclear extracts to examine the results further. Following three biological repeats we found statistically significant differences only for the hyperactivation of cRel in response to LPS, while other conditions showed the same trend as the EMSA results but did not pass the statistical test of significance (Figure 3D). Together, these biochemical data suggest that IκBε provides a key function in limiting cRel-containing dimer activity via a negative feedback loop, while it is critical for limiting RelA activity in response to only some stimuli but not others.

A mathematical model of RelA and cRel dynamics suggests a kinetic basis for IκBε’s stimulus-specific functions

The observation that the signaling phenotypes are stimulus-specific may suggest that there are underlying stimulus-specific biochemical mechanisms such as a co-stimulatory signaling pathway that is activated by one stimulus but not another. An alternative, more parsimonious explanation may be that differential signaling kinetics may account for the stimulus-specific phenotypes. Using a mathematical modeling approach we sought to test the latter hypothesis. To begin, we summarized the known relative relationships between the two potential negative feedback regulators IκBα and IκBε in terms of their interactions with RelA and cRel-containing dimers, their differential responsiveness to IKK-induced degradation and the stimulus-specific dynamics of IKK activity (Figure 4A). Interestingly, we found that whereas NFκB-responsive IκBα gene expression required RelA, NFκB-responsive IκBε gene expression could be mediated by either RelA or cRel (Supplementary Figure 2). Next, we constructed a mathematical model with these parameters by adapting a previously established mathematical model (3) to include the cRel dimers and to recapitulate B cell-specific dynamic control of RelA- and cRel-containing dimers (see Supplementary Methods). Simulations of this model with the IgM-induced transient IKK activity showed that RelA:p50 dimer is barely affected by IκBε-deficiency, but that in response to LPS-induced long lasting IKK activity, RelA:p50 remains hyperactivated at late times (Figure 4B, top panels). In contrast, cRel:p50 was hyperactivated in either condition (Figure 4b, bottom panels). By quantitating the time course at 24hrs, the stimulus-agnostic effect on cRel and stimulus-specific effect on RelA is readily appreciated (Figure 4C); this graph in fact closely resembles the experimental results obtained biochemically (Figure 3C).

These simulation results demonstrate that the kinetic argument is a sufficient explanation for the stimulus-specific phenotype seen in IκBε-deficient B-cells. We may summarize the kinetic argument as follows: In response to transient IKK signals, IκBα is capable of providing post-induction repression on its high affinity target RelA:p50, but less effectively on its low affinity target cRel:p50, which requires IκBε for complete suppression. However, in response to long lasting IKK signals, IκBα’s responsiveness to IKK renders it effectively neutralized; thus in this stimulation condition IκBε, which has lower responsiveness, plays an important role not only for its high affinity target cRel:p50 but also its low affinity target RelA:p50. Interestingly, the single specificity of the IκBα negative feedback loop for RelA:p50 and the dual specificity of the IκBε negative feedback loop for RelA;p50 and cRel:p50 are reflected in the dimer requirements for IκBα and IκBε inducible expression (Supplementary Figure S1). We note, that although reported kinetic relationships (Supplementary Methods) are consistent with this sufficiency argument, we cannot rule out that a stimulus-specific signaling pathway also plays a role in the described phenotype.

IκBε^−/−B cells show increased expression of NFκB target genes

To further characterize the phenotype at the molecular level, we examined how the gene expression programs induced by IgM or LPS were affected by the loss of IκBε. To this end, we employed high-throughput sequencing of polyA RNA isolated from wild type and IκBε^−/− B cells at 0, 2, 8 and 24 hours following stimulation with either IgM or LPS. Sequence data was converted into transcriptome levels using CummRbund (36), and these were normalized to the wild type 0 hour timepoint. We selected for genes induced in wild type B-cells by at least 2-fold by IgM and LPS in at least one timepoint. This resulted in 881 and 846 genes, respectively. We identified hyperinduced genes as those that showed at least a 2-fold change in the IκBε^−/− B-cell data at two timepoints when compared to the corresponding wild type timepoints, resulting in 56 and 106 genes, respectively (Figure 5A).

To determine if the identified hyperinduced genes are under the control of NFκB dimers, we utilized the HOMER motif discovery software adapted to perform searches of the known NFκB dimer motifs previously identified (29) and summarized here (Figure 5B top). We found enrichment for these κB-motifs within the promoter regions of genes hyperinduced in IκBε^−/−B cells when compared to controls that were not hyperinduced (Figure 5B). Interestingly, we found that the enrichment of the cRel:p50 motif in hyperinduced genes was statistically significant in both the IgM and LPS condition; however, the RelA:p50 motif enrichment was statistically significant only when B-cells were activated with LPS, and not IgM. These results reflect the prior biochemical data that showed that whereas cRel:p50 is hyperactivated in IκBε^−/− B-cells in both IgM and LPS conditions, RelA:p50 hyperactivation occurs primarily in response to LPS (Figure 3C).

Increased expression of IL-6 mediates the enhanced proliferation of IκBε-deficient B cells

Among the NFκB target genes that were hyperinduced in LPS-stimulated IκBε^−/− B-cells (Supplementary Table), we found inflammatory IL-6, which was initially discovered as a B cell stimulating and differentiation factor. Subsequent studies of IL-6’s effects on B cells have found that IL-6 allows for increased proliferation, enhanced differentiation and reduced apoptosis in B cells (37–40). The literature is unclear whether IL-6’s expression is cRel- or RelA-dependent (41–44). In our stimulation conditions, we found that IL-6 is robustly induced in response to LPS in wild-type cells both at the level of mRNA (Figure 6A) and secreted cytokine measured in supernatants (Figure 6B). As expected, IκBε^−/− B cells show hyperinduction at both early and late times, with the strongest effect at the protein level being at the late timepoint of 24 hrs. Interestingly, cRel-deficient B-cells showed no mRNA reduction at early timepoints (2 and 8 hrs), though a significant deficiency was observed at 24 hrs. Further, IL-6 hyperinduction appeared to be LPS-specific, correlating with RelA-hyperactivity, and prior observations in other cell types that pointed to RelA-dependent expression (41–43). Together, these data suggest that IL-6 induction is first triggered by RelA and may then be enhanced by cRel-containing dimers.

We next asked if the enhanced proliferation of the LPS stimulated IκBε^−/− B cells is potentially mediated by increased autocrine IL-6 co-stimulation of these cells. To test this hypothesis, we isolated B cells from wild type or IκBε^−/− mice and stimulated them with LPS in the presence of either 2 μg/ml cytokine-neutralizing or 2 μg/ml receptor-blocking antibodies for IL-6. We used antibodies neutralizing IL-1ε and IL-1β as controls, as these pro-proliferative cytokines were not found to be hyperinduced in our transcriptomic profiling. Effects on B cell proliferation was measured using CFSE staining. We found that antibodies blocking IL-6 signaling had little effect on B cell expansion from wild type mice; however, using B cells from IκBε^−/− mice we found a reduction in the B cell expansion to almost wild type levels (Figure 6C). Neither of the control antibodies had an effect on the LPS-triggered expansion of wild type or IκBε^−/− B cells. These findings suggest that the increased production of IL-6 is at least in part, responsible for the increased proliferative capacity of IκBε^−/− B cells.