Stimulus-Dependent Inhibitor of Apoptosis Protein Expression Prolongs the Duration of B Cell Signalling

Different dynamic behaviours of signalling activity can induce distinct biological responses in a variety of cells. However, the molecular mechanisms that determine the dynamics of kinase activities in immune cells are not well understood. In this study, we showed that the duration of both IκB kinase (IKK) and extracellular signal-regulated kinase (ERK) activities in B cell receptor (BCR)- and CD40-signalling pathways in B cells were regulated by transcriptional feedback loops. We conducted a time-course transcriptome analysis after BCR or CD40 stimulation and identified the following four candidate genes as feedback regulators for IKK and ERK: inhibitor of apoptosis protein (IAP), TNF alpha-induced protein 3, dual-specificity phosphatase 5, and sprouty homolog 2. Quantitative experiments and mathematical modelling suggested that IAP inhibition shortened the duration of IKK and ERK activity following both BCR and CD40 pathway stimulation, indicating a positive role for IAP in B cell signalling. Furthermore, transient kinase activities induced by IAP blockage reduced the levels of delayed expression genes. Together, our findings suggest that IKK and ERK activity durations can be fine-tuned by the coordinated regulation of positive and negative transcriptional feedback and that these network properties determine the biological output of B cells.

The duration of kinase activity is crucial in deciding the fate of mammalian cells. It is widely known that the transient and sustained activation of extracellular signal-regulated kinase (ERK) induces the proliferation and differentiation of many types of cells [1][2][3][4][5][6] and that prolonged Iκ B kinase (IKK) activity is required for stimulus-specific gene expression programs 7 . Furthermore, in the immune system, sustained signalling activities are important for efficient T cell activation and differentiation [8][9][10][11][12][13][14][15] . Although the kinetics of different molecules are involved in determining the duration of signalling activities 16,17 , feedback regulation in the signalling network also plays a role in duration control. Two types of feedback regulation are known to determine the duration of signalling activities: rapid post-translational feedback (mainly caused by protein phosphorylation or dephosphorylation) and slow transcriptionally induced feedback. The ERK system is a typical example wherein activity is dynamically controlled by these two types of feedback loops through post-translational modifications 4,5 and transcription-induced regulation 6,18 . Transcriptionally induced feedback regulators modulate signalling activity on a longer timescale; therefore, their functions in the control of cell fate appear to be of particular importance 18 .
B cells play an important role in humoral adaptive immunity. The recognition of foreign antigens by B cell receptors (BCRs) initiates multiple signalling cascades for the induction of B cell activation, expansion, differentiation, and cell survival, which result in mobilisation aimed at antigen clearance 19 . Importantly, dysregulation of these processes leads to immunodeficiency, lymphoid tumours, or autoimmune disease 20 . After the engagement of a BCR with an antigen, secondary signals mediated by co-stimulatory molecules such as CD40, a tumour necrosis factor (TNF) receptor family member, promote adequate development of the adaptive immune response. The association of CD40 with its ligand, CD40L/CD154 (which is transiently expressed on T cells) functions to permit and promote B cell survival, proliferation, and development 21 . Thus, the general activation scheme of B

Dynamics of BCR-and CD40-induced IKK and ERK activation in B cells.
We examined the dynamics of ERK and IKK activities over time (up to 120 min) following BCR or CD40 stimulation with anti-IgM or CD40 ligands, respectively (Fig. 1, Supplementary Fig. S1). Following BCR application, B cells from a mouse spleen (Splenic B) showed rapidly decaying IKK activation and sustained activation of ERK. On the other hand, CD40 stimulation induced sustained IKK and transient ERK activities (Fig. 1). Among these findings, we observed a characteristically prolonged duration of BCR-stimulated ERK activity at later time points. Similar activation patterns of IKK and ERK were observed in the DT40 B cell line ( Supplementary Fig. S1), which has been shown to be suitable for the comprehensive analysis of regulatory networks 27 . Gene expression profiles of BCR and CD40-stimulated B cells. The distinct duration of kinase activities at late phases after stimulation suggests regulation via transcriptionally induced feedback. To identify genes that potentially modulate kinase activity kinetics, we performed a comprehensive gene expression analysis via microarray using wild type, MEK inhibitor-pre-treated (WT + Inh), or IKKβ inactive (IKKβ SA) cells following stimulation with BCR or CD40 for 0, 15, 30, 45, 60, or 90 min. We selected genes that showed significantly altered expression levels when compared to control cells (time 0). From this, 578 genes were selected (False Discovery Rate (FDR) < 0.05) ( Fig. 2a and Supplementary Fig. S2). To explore the biological function of the selected genes, we employed a Gene Ontology (GO) enrichment analysis ( Table 1). As expected, many genes that were significantly enriched exhibited functions known to contribute to immune-and signal-regulation. The most popular GO term was related to the positive and negative regulation of signals, which suggested possible roles of the selected genes in the regulation of signalling pathways. In our experimental setting, only three CD40-specific genes were identified (Fig. 2a), that is LOC768982, cation transport regulator homolog 1 (CHAC1), and TNF receptor-associated factor 5 (TRAF5), and that are not related to the regulation of IKK and ERK signalling 21 . In contrast, BCR-stimulation induced the up-and down-regulation of a large number of genes (511) over 90 min (Fig. 2a,b) and shared 13 genes in common with CD40 stimulation. Therefore, we performed k-means clustering  Fig. 2b were subjected to k-means clustering (k = 8) and the average was plotted. on the BCR-regulated genes, which resulted in their subdivision into eight clusters based on their time-course profiles (Fig. 2c). Clusters 1 (26 genes), 2 (52 genes), 3 (93 genes), and 4 (56 genes) showed increasing gene expression with slightly different expression peak times. Notably, A20 28 , a typical feedback regulator of IKK and an early stimulus response gene in NF-κ B signalling, was included in cluster 1, which showed an immediate increase after cell stimulation. Thus, we explored other potential feedback regulators by focusing on cluster 1 (the full list of genes is provided in Supplementary Dataset 1).

Identification of feedback regulators.
Based on the literature and previously reported primary B cell gene expression data 29 , we identified four candidate genes in cluster 1 as potential feedback regulators for ERK or IKK signalling in BCR-and CD40-stimulated B cells (Fig. 3). In addition to A20, we identified two other well-known negative feedback regulators, sprouty homolog 2 (SPRY2) and DUSP5. A20 is a ubiquitin editing enzyme and is known to negatively control the activation dynamics of IKK-NF-κ B signalling 30,31 . DUSP5 is a MAPK phosphatase that negatively regulates ERK activity and modulates both innate and adaptive immunity 32-34 . Sprouty proteins have been shown to inhibit ERK signalling 35 and a member of the Sprouty family, mammalian SPRY2, has been suggested to regulate BCR-mediated ERK activity 36 . In addition to these protein genes, we identified an E3 ubiquitin ligase, IAP, as a novel feedback regulator. IAP has been suggested to positively regulate CD40-signalling-mediated IKK-NF-κ B and ERK activation 37,38 . However, the role of IAP in BCR signalling remains unknown. Therefore, we examined the regulatory functions of IAP in BCR-signalling using IAP-deficient DT40 cells. Using this approach, our data clearly showed that IAP-deficient DT40 cells failed to sufficiently activate IKK and ERK following BCR engagement ( Supplementary Fig. S3). Together, these findings suggest that IAP acts as a positive regulator in BCR-signalling.

Mathematical modelling explains IAP feedback effects on the duration of IKK and ERK activity.
Genetic and biochemical approaches have limitations in identifying regulatory feedback loops, as genetic deletion in particular often ignores the development of biological networks over time 26 . In addition, as in the case under study, multiple positive (IAP) and negative (A20, DUSP, and SPRY2) feedback regulators combine to determine kinase activities. Therefore, we used a computational approach to evaluate the mechanisms underlying the IAP-mediated control of kinase activation dynamics. First, we constructed an ordinary differential equation (ODE)-based mathematical model (Figs 4 and 5, Supplementary Table S1; details of the model are described in the Methods section). Using this model (hereafter called model (A)), complied with Michaelis-Menten kinetics, we were able to describe the active states of kinases associated with three types of regulatory networks: those without feedback loops (Fig. 4a), those without a positive feedback loop (Fig. 4b), and those with positive and negative feedback loops (Fig. 4c). Kinetics parameters were obtained by fitting experimental time-course profiles of IKK and ERK activities in cycloheximide (CHX)-treated cells, IAP inhibitor-treated cells, and wild type cells (Figs 4d and 5c, and Supplementary Fig. S4), which corresponded to the three abovementioned regulatory networks, respectively ( Fig. 4a-c). Since the effect of positive feedback is considered to be mainly dependent on IAP expression, we assumed that the reaction rate associated with this type of network would be proportional to the amount of newly synthesized IAP (Fig. 4e, red). The amount of synthesized IAP was obtained by subtracting the CHX-treated IAP protein value from the CHX-nontreated value ( Supplementary Fig. S5). Further, we assumed that the negative feedback reaction rate was proportional to the averaged expression kinetics (Fig. 3) Table 1. Gene enrichment analysis. The representative results of Gene Ontology (GO) analysis providing the terms of biological processes. The complete list of GO terms is presented in Supplementary Dataset 2. FDR, false discovery rate. BIRC2, also known as cIAP1, is called IAP in the text.
Scientific RepoRts | 6:27706 | DOI: 10.1038/srep27706 of the negative regulators (A20 for BCR-signalling and DUSP and SPRY2 for CD40-signalling) (Fig. 4e, blue). As shown in Fig. 5a, our simulated results fit well with the experimental data with/without transcriptional feedback (Fig. 4d). Furthermore, when we ran the simulation without the positive feedback loop, we observed activated kinase profiles that were similar to those observed experimentally in the presence of the IAP inhibitor (AT; AT-406) (Fig. 5b,c and Supplementary Fig. S4). Together, these results indicate that the dynamics of IKK activation in BCR-signalling is primarily shaped by negative feedback, while IKK activation dynamics in CD40-signalling is mainly determined by positive feedback. In contrast, our results on ERK activation dynamics implicate the involvement of two feedback loops in either BCR-or CD40-signalling. Thus, we demonstrated that the transcriptional feedback of IAP regulates kinase activity duration in B cells.

IAP-mediated kinase activity duration is critical for late-phase gene expression. Quantitative
information about signalling activity undergoes transcriptional regulation. Thus, to understand how IAP-regulated IKK and ERK activity durations affect downstream gene expression, we conducted a computational analysis using a simple transcriptional model 39 (hereafter called model (B)) (Fig. 6a). Since we observed that IAP controlled IKK and ERK activation dynamics, signal inputs (Kinase) in model (B) were set as the IKK or ERK activities (Fig. 5b, right) used in model (A) with and without a positive feedback. Using this model, we observed that IKK input without positive feedback induced a low output (gene X expression) (Fig. 6b), whereas ERK signal without positive feedback showed transient gene expression (Fig. 6c). In comparison, input controlled by both positive and negative feedback led to a consistent level of amplitude and duration of gene expression (Fig. 6b,c,    in murine splenic B cells (Fig. 6d). To dissect any effects of positive feedback, we employed the following three conditions: CD40 stimulation alone, CD40 stimulation pre-treated with IAP inhibitor, and the addition of IAP inhibitor 10 min after CD40 stimulation. Expression of the two genes showed attenuated profiles over time in IAP inhibitor-pre-treated cells, although the difference between pre-and post-treated cells was not significant for Pik3r5. In comparison, IAP inhibition after CD40 treatment resulted in significantly lowered gene expression in the late phase of the examined time-course (Fig. 6d). Thus, our results indicated that the kinase activation dynamics that are controlled by IAP positive feedback are critical for late phase gene expression.

Discussion
Using computational simulation and quantitative experiments, we revealed that IAP transcriptional feedback regulated the dynamics of ERK and IKK activity in B cells. In addition, we found that DUSP, A20, and SPRY2 might contribute to signalling and gene expression kinetics in B cells.
DUSPs are major players in the regulation of both BCR and CD40 signalling via negative feedback. Negative transcriptional regulation of ERK by DUSPs has been intensively characterized in previous studies 6,40,41 . In B cells, prolonged ERK activation has been reported to inhibit cell differentiation 34 ; however, BCR signalling without cytokine or toll-like receptor co-stimulation, which induces ERK activation, cannot induce proper cell differentiation [42][43][44][45] . Consistent with this, ERK activation induced solely by BCR engagement showed sustained ERK activation in our system (Fig. 1), indicating that BCR-dependent DUSP5 induction might not be sufficient for the attenuation of ERK activity. Co-stimulatory cytokines (e.g., interleukin 2) might also be required to induce DUSP family gene expression that is sufficient enough to effect B cell differentiation 34 . (e) Red circles depict the rate of newly synthesized IAP (Positive) relative protein amounts that were analysed by immuno blotting and calculated as described in Supplementary Fig. S4. The data shown represent the means ± s.d. (n = 2). The blue circle (Negative) indicates the averaged gene expression kinetics of A20 for BCRsignalling and of DUSP5 and SPRY2 for CD40-signaling, respectively as displayed in Fig. 3. The values shown indicate values relative to the maximum for each gene expression. The dotted line represents the fitted curve as described in Methods. It is known that the NF-κ B pathway is highly regulated via the attachment of K63-linked polyubiquitin chains, which lead to the recruitment of signalling molecules to induce signal transduction 46 . A20, CYLD, and A20-binding inhibitor of NF-κ B (ABIN1) are ubiquitin-related enzymes or co-factors that terminate this signal through the disassembling of ubiquitin-anchored signal complexes 30,47 . In particular, the negative regulation of IKK's activation dynamics appeared to be largely dependent on A20; both CYLD and ABIN1 gene transcripts were expressed at much later time points (Supplementary Dataset 1). In fact, B cell-specific deletion of A20 in mice has been shown to sustain IKK activation in response to BCR and CD40 ligand binding, resulting in plasma cell hyperplasia and higher subsequent levels of serum immunoglobulins 32,48,49 . Together, these observations indicate that the signalling dynamics that are shaped by A20 are crucial for B cell biology.
Our mathematical model for kinase reactions supported the concept that IAP works as a positive regulator of IKK and ERK activation and that newly synthesized IAP prolongs their activities (Fig. 7). Furthermore, using experimentation and mathematical simulations, we demonstrated that IAP inhibition after CD40 stimulation lowered gene expression levels after 1 h. In the case of Bcl-x L , peak expression emerged after 1 h, and it is important to note that this kind of early gene expression might not be affected by newly expressed IAP. Accordingly, post-treatment with IAP inhibitor likely only affected Pik3r5 expression (which was delayed) and later Bcl-x L expression. Of note, we identified two patterns of effects of IAP on gene expression, one that involved elevated levels of expression (high amp-type) and one that involved sustained gene expression (sustained-type). Our simulation implied that the former was dependent on IKK input, while the latter relied on ERK input. In fact, Pik3r5 was not induced in IKKβ -inactive cells (Supplementary Fig. S6). Moreover, up-regulation of Bcl-x L , which appeared to be dependent on ERK input, has been suggested to require ERK activation 50 . To confirm these findings, we tested whether forced expression of the dominant-positive forms of the IKKβ and MEK1, could restore the abovementioned patterns of mRNA expression. Indeed, high amp-type Pik3r5 and sustained-type Bcl2A1 expression, which were reduced in IAP-deficient DT40 cells, were rescued by the dominant-positive forms of IKKβ and MEK1, respectively ( Supplementary Fig. S7). The importance of IAP in B cell physiology has been highlighted by the finding that IAP-deficient B cells fail to mobilize B cell immunity 37 . Indeed, our current study revealed that IAP was required for ERK and IKK activation, which are involved in the regulation of BCR-and CD40-signalling via positive feedback; however, the site of IAP action remains unclear. IAP can interact with several proteins including ABINs, IKK, A20, NIK, and TRAFs 51 , and also mediates the canonical NF-κ B pathway in TNF receptor signalling via K63 ubiquitination of receptor (TNFRSF)-interacting serine-threonine kinase 1 (RIP1) with the E2 ubiquitin ligase UBC5 46,51,52 . During CD40 signalling, IAPs induce K48 ubiquitination and degradation of the TRAF3 protein that initiates the release of TRAF2-MEKK1 and TRAF6-TAK1 signalling complexes into the cytoplasm. This process then leads to the activation of MAPK and NF-κ B pathways. In the current study, CD40 signalling depicted a faster decrease in activation kinetics even though no transcriptional feedback effects were observed when compared to BCR signalling (Fig. 4d). Thus, we assumed that the ubiquitin-dependent degradation of the signal module progressed independently of gene expression, since IAP showed rapid decay in CHX-treated cells (Fig. 4d). Consistent with this assumption, proteasome inhibition has been reported to sustain activation of MAPK 38 .
Regarding BCR-signalling, the molecules which interact with IAP remain elusive. Here, we revealed that IAP-mediated inhibition of the IKK pathway had slightly different effects on BCR-and CD40-signalling, suggesting that IAP complex formation and its targets might differ (Fig. 5). Therefore, we predicted IAP-interacting proteins via network analysis of a public database (Supplementary Fig. S8). Since IAP is a K48 ubiquitin ligase, negative regulators of signalling, such as TRAF3 in CD40-signaling, could act as direct targets of the pathway 52 . In this context, A20, a negative regulator of IKK 28 could be regarded as one potential target of BCR-signalling. Finally, although TRAF3 is associated with both IKK and ERK pathways in CD40-signalling, we failed to find a molecule which interacts with A20 in the ERK pathway. Thus, determining the link between IAP and the ERK cascade will require further study.
Since IAP positive feedback is critical for establishing gene expression profiles, information regarding feedback-specific IAP target molecules or modification sites will permit the control of gene expression dynamics. Furthermore, the ability to regulate gene expression is considered a promising approach in identifying pharmaceutical targets with minimal side effects 53 .
In conclusion, in this study we employed mathematical modelling to analyse the effects of transcriptional feedback -a key determinant of gene expression kinetics -on generating distinct dynamics during BCR-and CD40-mediated signalling.

Materials and Methods
Mice, cells, antibodies (Abs), and reagents. C57BL/6 mice from Charles River Laboratories International, Inc. were maintained under specific pathogen-free conditions and used at 8-12 wk of age. All protocols were approved by the RIKEN Animal Committee and all experiments were carried out in accordance with the approved guidelines. For western blots, splenic B cells were purified by depleting CD43 + cells with magnetic beads using AutoMACS (Miltenyi Biotec) and were cultured in Iscove's modified Dulbecco's medium supplemented with 10% foetal bovine serum and 1% penicillin-streptomycin.
Abs specific for ERK, anti-phospho-ERK, anti-phospho-Iκ Bα , and anti-phospho-IKKα /β were purchased from Cell Signaling Technology. The anti-chicken IgM mAb M4 was used for BCR stimulation 55 . Anti-mouse IgM mAb (α -μ ) was obtained from Jackson Immuno Research, and anti-mouse CD40 from BD Biosciences. A GST-chicken-sCD154 fusion protein was used as the CD40 ligand. The mitogen activated kinase kinase (MEK) inhibitor, U0126 (Wako), was used at a concentration of 5 μ M and the IAP inhibitor, AT, (AT-406; Active Biochem) was used at concentration of 1 μ M.
IKK kinase assays were performed as described previously 55 . For protein detection, the ECL Plex fluorescent western blotting system and ImageQuant LAS 4000 (GE Healthcare) were used. Kinase activity was quantified from the intensities of the protein and phosphorylated protein bands using a Multi Gauge version2.2 (Fujifilm) densitometer as described previously 29 .
Generation of IAP-deficient DT40 cells. A genomic clone of chicken IAP was obtained by polymerase chain reaction (PCR) using oligonucleotides designed by the NCBI database (Gene ID: 374012) as primers and chicken DT40 B cell genomic DNA as a template. The targeting constructs for chicken IAP were designed for the neomycin and histidinol drug-resistant gene cassettes to replace exons 1-8 of the chicken IAP gene, as shown in Supplementary Fig. 3a. These targeting vectors were sequentially transfected into DT40 cells, resulting in generation of chicken IAP-deficient DT40 cells as described elsewhere 54,55 . Microarray analysis. Cells were stimulated with 10 μ g/ml anti-IgM (M4) or 6 μ g/ml CD40 ligand for 15, 30, 45, 60, or 90 min using wild type DT40 cells. To identify ERK-or IKK-dependent transcripts, DT40 or IKKβ SA cells, respectively, were pre-treated with MEK inhibitor 30 min prior to ligand stimulation. Untreated cells were used as the control. Total RNA was purified with the Qiagen RNeasy mini kit. GeneChip (Affymetrix GeneChip Chicken Genome Array) experiments were carried out according to the manufacturer's protocol. Microarray data analysis was performed as described previously 29 . The microarray data were deposited at the NCBI Gene Expression Omnibus database (GEO ID: GSE59002).  Fig. 4c), and comprised of the following equations:

Gene expression analysis.
where k i is a reaction rate constant, km i is a Michaelis constant, Kinase is neutral IKK or ERK, Kinase* is active IKK or ERK, signal(t) is the time-course of upstream signalling activity, PF(t) is the time-course of the positive feedback regulator, IAP, and NF(t) is the averaged gene expression kinetics of the negative feedback regulators (A20 for BCR-signalling and DUSP and SPRY2 for CD40-signalling). Here, the expression of each gene is used relative to their own maximal values. The factor signal(t) was employed as a function according to our previous study 29  To reflect the gene expression kinetics and IAP synthesis data to the rate of feedback, time-courses of NF(t) and PF(t) were employed as interpolations of the relative expression profiles of the negative regulators (Fig. 4e) and newly synthesized IAP protein expression data ( Fig. 4e and Supplementary Fig. S4), respectively, using the method "pchip" of the function interp1 in MATLAB R2014a (MathWorks, Natick MA). The simulations for Fig. 4a,b changed the parameters k 4 and k 5 to zeros and the parameter k 4 to zero, respectively.
Model (B) of the simple motif for gene expression, as shown in the diagram of Fig. 6a, comprised the following equation: where X is a gene and n is the Hill coefficient. Simulations were performed using the function ode15s of MATLAB R2014a. All parameters can be found in Supplementary Tables S1 and S2.