Compensatory IKKα activation of classical NF-κB signaling during IKKβ inhibition identified by an RNA interference sensitization screen

Abstract

A subtype of diffuse large B-cell lymphoma (DLBCL), termed activated B-cell-like (ABC) DLBCL, depends on constitutive nuclear factor-κB (NF-κB) signaling for survival. Small molecule inhibitors of IκB kinase β (IKKβ), a key regulator of the NF-κB pathway, kill ABC DLBCL cells and hold promise for the treatment of this lymphoma type. We conducted an RNA interference genetic screen to investigate potential mechanisms of resistance of ABC DLBCL cells to IKKβ inhibitors. We screened a library of small hairpin RNAs (shRNAs) targeting 500 protein kinases for shRNAs that would increase the killing of an ABC DLBCL cell line in the presence of a small molecule IKKβ inhibitor. Two independent shRNAs targeting IKKα synergized with the IKKβ inhibitor to kill three different ABC DLBCL cell lines but were not toxic by themselves. Surprisingly, IKKα shRNAs blocked the classical rather than the alternative NF-κB pathway in ABC DLBCL cells, as judged by inhibition of IκBα phosphorylation. IKKα shRNA toxicity was reversed by coexpression of wild-type but not kinase inactive forms of IKKα, suggesting that IKKα may directly phosphorylate IκBα under conditions of IKKβ inhibition. In models of physiologic NF-κB pathway activation by CARD11 or tumor necrosis factor-α, compensatory IKKα activity was also observed with IKKβ inhibition. These results suggest that therapy for ABC DLBCL may be improved by targeting both IKKα and IKKβ, possibly through CARD11 inhibition.

Cancer therapy is rapidly evolving toward the use of agents that specifically target signaling pathways controlling cancer cell proliferation and/or survival. Especially promising are agents that target pathways that are genetically altered during the generation of the malignant clone. However, resistance to such agents may arise by several mechanisms. Most commonly, mutations in the cancer genome may be clonally selected that blunt or abrogate the response to the therapeutic agent. Alternatively, cancer cells may engage compensatory signaling pathways that sustain proliferation and survival of the malignant clone.

Pathway-directed therapy holds promise for the treatment of DLBCL, the most common subtype of non-Hodgkin's lymphoma. This diagnostic category consists of at least three molecularly distinct subtypes that originate from different stages of normal B cell development, use distinct oncogenic mechanisms, and differ in clinical outcomes (1). The ABC subtype of DLBCL responds much less well to conventional multiagent chemotherapy than the germinal center B cell-like (GCB) subtype. The nuclear factor-κB (NF-κB) pathway is an attractive therapeutic target in ABC DLBCL, as this pathway is constitutively active and is required for the survival of this lymphoma subtype (2). A signaling complex involving CARD11, MALT1, and BCL10 is responsible for this constitutive NF-κB activation, in some cases because of somatic mutations in CARD11 (3, 4). These three proteins form a signaling scaffold at the plasma membrane that recruits a heterotrimeric protein complex consisting of IκB kinase β (IKKβ), IκB kinase α (IKKα), and a regulatory γ subunit. After membrane recruitment, IKKβ becomes active and phosphorylates IκB alpha (IκBα), leading to its destruction, and thereby allowing the NF-κB transcription factors to enter the nucleus. This signaling cascade is known as the “classical” NF-κB pathway.

Specific small-molecule inhibitors of IKKβ are toxic to ABC DLBCL cell lines but not to those of the GCB DLBCL subtype (5). Indeed, many IKKβ inhibitors are under development, raising the hope that one or more will be active in ABC DLBCL (6). In anticipation of this development, we investigated potential mechanisms of resistance to IKKβ inhibitors in ABC DLBCL. To identify genes in which activity might antagonize or synergize with the toxicity of IKKβ inhibition in ABC DLBCL, we performed an RNA interference (RNAi) sensitization screen using a retroviral library to express small hairpin RNAs (shRNA). We report here that killing of ABC DLBCL cell lines by an IKKβ inhibitor was enhanced by shRNAs targeting IKKα. Previous studies defined the importance of IKKα for the “alternative” NF-κB pathway, based on its ability to phosphorylate p100, leading to its proteolytic processing into p52, an NF-κB transcription factor subunit (7). The present work highlights an unexpected role for IKKα as an IκBα kinase in the classical NF-κB pathway under conditions of IKKβ inhibition, suggesting that targeting of both IKKα and IKKβ might improve the therapeutic response in patients with ABC DLBCL.

Results

An RNAi Screen for Pathways that Sensitize ABC DLBCL Cells to an IKKβ Inhibitor.

Two structurally related small-molecule inhibitors of IKKβ, MLX105. and MLN120B block the NF-κB pathway and are toxic for ABC DLBCL cell lines (5, 8, 9). The specificity of these inhibitors is highlighted by their lack of toxicity for cells without NF-κB pathway activation and by the fact that their toxicity for ABC DLBCL cells can be prevented by expressing constitutively active forms of NF-κB (5). To identify genetic pathways that might potentiate or antogonize the toxic effect of these IKKβ inhibitors, we used a library of shRNA-expressing retroviruses to conduct an RNAi “sensitization” screen as depicted in Fig. 1A. The ABC DLBCL cell line OCI-Ly3 was transduced with a pool of retroviruses from this library and then treated with a submaximal lethal dose of MLX105 such that roughly 50% of the cells were killed after 3 days. Because each vector in the shRNA library possesses a unique 60-base pair molecular “bar code” sequence (3), we were able to compare the complement of shRNA vectors present in cultures treated with the IKKβ inhibitor to that in parallel untreated cultures. This allowed us to identify shRNAs that increased or decreased cell death in the presence of the IKKβ inhibitor but to ignore shRNAs that were toxic in a manner that was not synergistic with IKKβ inhibition. As we suspected that other kinases might impinge upon the NF-κB pathway in ABC DLBCLs, we screened a pool of shRNAs targeting 500 protein kinases, with at least of three shRNAs per gene.

Fig. 1.

Kinase shRNA library sensitization screen. (A) Kinase shRNA library screen for genes that can sensitize cancer cell proliferation and survival to IKKβ inhibitor (see text for details). (B) Barcode screen identified IKKα shRNAs that enhance IKKβ inhibitor killing in OCI-Ly3 cells. Depicted is the ratio of the shRNA abundance in untreated versus IKKβ inhibitor-treated cells as a log2 value. (C) Western blotting showing downregulation of IKKα protein expression with IKKα shRNAs in OCI-Ly3 ABC DLBCL cells. Dox: doxycycline-induction of shRNA expression.

This screen identified two independent shRNAs targeting IKKα (shIKKα1 and shIKKα2) that enhanced killing by the IKKβ inhibitor (Fig. 1B). Both shRNAs were effective in knocking down expression of IKKα protein, with shIKKα2 being more potent (Fig. 1C). Accordingly, shIKKα2 was also more toxic than shIKKα1 in the RNA interference screen (Fig. 1B). To confirm and extend these results, OCI-Ly3 cells were transduced with the individual IKKα shRNAs, exposed to a range of doses of the IKKβ inhibitor MLN120B, and assessed for viability after 2 days (Fig. 2A). Both IKKα shRNAs shifted the dose-response curve, but shIKKα2 again had a greater effect than shIKKα1 (Fig. 2A). The two IKKα shRNAs were not toxic by themselves, even after 10 days of induction, but again displayed synergistic toxicity with the IKKβ inhibitor MLN120B (Fig. 2B). IKKα inhibition also synergized with shRNA-mediated knockdown of IKKβ to kill ABC DLBCL cells, confirming the results with the small-molecule IKKβ inhibitors ([supporting information (SI) Fig. S1]). The synergistic toxicity of IKKα and IKKβ inhibition was apparently mediated by apoptosis, as shIKKα2 increased caspase 3/7 activation in the presence of the IKKβ inhibitor MLN120B but had no effect by itself (Fig. 2C).

Fig. 2.

IKKα shRNAs sensitize DLBCL cells to IKKβ inhibition. (A) Two different IKKα shRNAs (shIKKα1 and shIKKα2) were expressed via a doxycycline-regulated promoter in OCI-Ly3 cells for 4 days. Both IKKα shRNAs, in comparison to control shRNA targeting luciferase, sensitized OCI-Ly3 cells to the IKKβ inhibitor. (B) IKKα shRNA expression, not control DsRed shRNA expression, is toxic to OCI-Ly3 cells only in the presence of the IKKβ inhibitor. Cells were transduced with a retroviral vector co-expressing shIKKα2 and enhanced green fluorescent protein (EGFP), and the percentage of EGFP+ cells was monitored by flow cytometry. (C) Expression of IKKα shRNAs in OCI-Ly3 cells increases caspase 3/7 activity in the presence of IKKβ inhibitor. (D) Expression of IKKα shRNAs sensitizes other ABC DLBCL cell lines (HBL-1, SUDHL2, and OCI-Ly10) to the IKKβ inhibitor as compared with control Luc shRNA. Expression of IKKα shRNAs does not sensitize other cell lines (PMBL-U2940 and K1106; multiple myeloma-L363) to the IKKβ inhibitor as compared with control Luc shRNA.

We next investigated whether the synergism between IKKα and IKKβ inhibition extended to other IKKβ-dependent cell lines. The IKKα2 shRNA enhanced killing by the IKKβ inhibitor in three additional ABC DLBCL cell lines (OCI-Ly10, HBL-1, and SUDHL-2) but did not affect two primary mediastinal B-cell lymphoma cell lines (K1106 and U2940) or four multiple myeloma cell lines (L363, KMS28, LP1, and MM1) (Fig. 2D and data not shown). These data suggest that the mechanism responsible for IKKα engagement after IKKβ inhibition is restricted to ABC DLBCL cells.

IKKα Participates in the Classical NF-κB Pathway during IKKβ Inhibition.

The toxicity of IKKβ inhibition in ABC DLBCL cell lines is associated with loss of classical NF-κB pathway activity, as judged by phosphorylation and degradation of IκBα (2, 5). Although the effects of IKKα on NF-κB activity are mainly thought to involve the alternative pathway (7) or histone phosphorylation (10), we hypothesized that IKKα might instead contribute to the classical NF-κB pathway in ABC DLBCL cells under conditions of IKKβ inhibition. To test this possibility, we used a cell-based assay for classical NF-κB signaling in which a fusion protein between IκBα and Photinus luciferase is expressed together with Renilla luciferase as a control (5). This IκBα-Photinus luciferase reporter is phosphorylated and degraded similarly to IκBα, and thus the ratio of Photinus to Renilla luciferase activity increases as IKK activity decreases. This reporter system was introduced into OCI-Ly3 cells, which were subsequently transduced with shRNAs targeting IKKα, GFP, or DsRed. After induction of shRNA expression for 2 days by doxycycline, the IKKβ inhibitor was added at various concentrations for 4 hours. As expected, the IκBα-Photinus reporter level rose after exposure to the IKKβ inhibitor in a dose-dependent fashion in the cells harboring control shRNAs (Fig. 3A). Induction of the IKKα shRNAs had no effect on this reporter in the absence of the IKKβ inhibitor, indicating that IKKα does not contribute to basal IκBα kinase activity in these cells. However, the IKKα shRNAs augmented the response to the IKKβ inhibitor, with shIKKα2 having a somewhat greater effect than shIKKα1, indicating that IKKα participates in the classical NF-κB pathway under these conditions (Fig. 3A).

Fig. 3.

IKKα participates in the classical NF-κB pathway during IKKβ inhibition. (A) IKKα knockdown intensifies the effect of IKKβ inhibition on IκBα kinase activity in ABC DLBCL cells. OCI-Ly3 cells harboring the IκBα-Photinus luciferase and control Renilla luciferase reporters were induced to express IKKα or control (GFP, DsRed) shRNAs for 3 days before treatment with 0-25 μM IKKβ inhibitor for 4 hours. Shown is relative expression of IκBα-Photinus luciferase, normalized to cells expressing DsRed shRNA without shRNA induction and IKKβ inhibitor treatment. (B) Gene expression profiling of ABC DLBCL cells treated with IKKβ inhibitor. Shown are genes that were downregulated in OCI-Ly3 cells expressing IKKα2 shRNA treated with or without 12.5 μM IKKβ inhibitor for the indicated times. (C) Average NF-κB target gene expression in cells with or without IKKβ inhibitor treatment and IKKα knockdown.

These data suggest that IKKα plays a “compensatory” role in the classical pathway signaling: IKKα made a larger contribution to IκBα kinase activity when IKKβ was inhibited than when cells were untreated. This compensatory model predicts that the IKKα shRNA should have a greater influence on NF-κB target gene expression when IKKβ is inhibited than when cells are untreated. To test this, we used DNA microarrays to measure expression of a set of previously defined NF-κB target genes in ABC DLBCLs (5). Induction of the IKKα shRNA alone for 2 or 3 days had a measurable but modest effect on NF-κB target gene expression in OCI-Ly3 cells compared with uninduced cells (Figs. 3B, 3C). However, when cells were also treated with the IKKβ inhibitor, the influence of the IKKα shRNA on the NF-κB signature was more pronounced, consistent with the compensatory model (Figs 3B, 3C).

A hallmark of alternative NF-κB signaling is activation of IKKα by the kinase NIK, leading to phosphorylation of p100 by IKKα (11–15). To test whether NIK participates in IKKα activation in ABC DLBCL cells, we used two NIK shRNAs (shNIK1 and shNIK2) that were previously shown to effectively silence NIK and to kill myeloma cell lines with high NIK activity (9). The expression of these NIK shRNAs did not enhance killing of OCI-Ly3 cells by the IKKβ inhibitor and did not affect the activity of the IκBα-Photinus reporter (Figs. S2A, S2B). These results are consistent with involvement of IKKα in classical rather than alternative NF-κB signaling in ABC DLBCL cells upon IKKβ inhibition.

IKKα Kinase Activity Is Required for Activating the Classical NF-κB Pathway.

We considered two general mechanisms to explain the role of IKKα in classical pathway activation under conditions of IKKβ inhibition. IKKα might be active as a kinase to phosphorylate IκBα during IKKβ inhibition in ABC DLBCL lines. In support of this possibility, IKKα is capable of phosphorylating IκBα in vitro and MLN120B does not inhibit IKKα phosphorylation of IκBα in vitro at the doses used to inhibit IKKβ, as shown in Fig. S3 (16, 17). Alternatively, IKKα kinase activity might not be required for its effect in ABC DLBCL cells, but instead IKKα might perform a scaffolding function that is required for optimal IKKβ activity.

To distinguish between the two possibilities, we devised a complementation assay in which we knocked down expression of the endogenous IKKα in ABC DLBCL cells using an shRNA directed against the IKKα 3′ untranslated region (3′UTR), and then introduced either wild-type IKKα or a catalytically inactive form (K44A) (16) using cDNAs lacking the 3′UTR. Fig. S4 shows that an shRNA targeting the IKKα 3′ UTR (shIKKα3) sensitized OCI-Ly3 cells to IKKβ inhibition. We introduced shIKKα3 together with expression vectors for wild-type or kinase-dead IKKα, or a control empty vector, into OCI-Ly3 cells harboring the IκBα-Photinus luciferase reporter. In empty vector control cells, IKKα knockdown augmented the effect of the IKKβ inhibitor by roughly 50% (Fig. 4A). In cells complemented with wild-type IKKα, IKKα knockdown had no effect on the IκB-luciferase reporter. By contrast, in cells complemented with kinase-dead IKKα, the effect of IKKα knockdown was similar to that in control cells. Moreover, we measured cellular viability under the same conditions and observed that wild-type IKKα could prevent the toxicity of the (shIKKα3) whereas kinase-dead IKKα had little or no ability to rescue the cells (Fig. 4B). We conclude that the kinase activity of IKKα is required for the compensatory activity of the classical NF-κB pathway under conditions of IKKβ inhibition in ABC DLBCL cells.

Fig. 4.

IKKα kinase activity contributes to the activity of the classical NF-κB pathway. (A) Wild-type but not kinase-dead IKKα complements IKKα knockdown in ABC DLBCL cells treated with an IKKβ inhibitor. OCI-Ly3 cells harboring the IκBα-Photinus luciferase reporter were engineered to inducibly express shIKKα3 (targeting the IKKα3′UTR) and coding region cDNAs for either wild-type or kinase-dead IKKα (K44A). Relative activity of the IκBα-Photinus luciferase reporter in shIKKα3-induced versus uninduced cells is shown. Values above 100% indicate an inhibitory effect of shIKKα3 on IκBα kinase activity. Results were normalized to cells transduced with a control DsRed shRNA without shRNA induction or IKKβ inhibitor treatment. (B) IKKα kinase activity enhances ABC DLBCL cell viability during IKKβ inhibition. OCI-Ly3 expressing wild type or kinase-dead IKKα were induced to express shIKKα3 for 2 days and then treated with the IKKβ inhibitor (0.78–12.5 μM) for 3 days. Viability was measured in shIKKα3-induced cells relative to uninduced cells.

Diverse Stimuli Evoke Compensatory IKKα Activity during IKKβ Inhibition.

Since the synergism between IKKα and IKKβ inhibition was evident in ABC DLBCLs but not in other NF-κB-dependent cell lines (Fig. 2D), we suspected that this phenomenon was dependent on the particular mechanism by which NF-κB was activated. We therefore examined several models of classical NF-κB pathway activation for evidence of compensatory IKKα activity upon inhibition of IKKβ. Constitutive IKK activation in ABC DLBCL requires the CARD11, BCL10, and MALT1 (CBM) complex (3), sometimes as a consequence of somatic mutations in CARD11 (4). The CBM complex is engaged after B cell receptor (BCR) stimulation as a result of CARD11 phosphorylation by protein kinase C β (18, 19). This signaling pathway can be mimicked by treatment of B cells with phorbol myristate ester plus ionomycin (P/I). In the GCB DLBCL cell line BJAB, the NF-κB pathway is relatively quiescent, but treatment with P/I induced expression of the NF-κB target gene CD83 roughly 10-fold (Fig. 5A). The IKKβ inhibitor MLN120B moderately suppressed this CD83 induction (Fig. 5A, right), and knockdown of IKKα by shRNA did not have any effect by itself (Fig. 5A). However, combined inhibition of IKKβ and IKKα markedly suppressed CD83 induction, suggesting that IKKβ inhibition led to compensatory IKKα activity under these conditions. To test whether IKKα kinase activity was required for this effect, we knocked down the expression of endogenous IKKα in BJAB with shIKKα3 and introduced expression vectors for either wild-type IKKα or kinase-dead (K44A) IKKα (Fig. 5B). In the absence of IKKβ inhibitor, the CD83 response to P/I treatment was comparable in cells with wild-type and kinase-dead IKKα (Fig. 5B). Again, the IKKβ inhibitor suppressed CD83 induction, but this suppression was much more pronounced in cells with kinase-dead IKKα (Fig. 5B). Thus, in the setting of BCR pathway stimulation and IKKβ inhibition, the compensatory function of IKKα in the NF-κB pathway requires its kinase activity, consistent with the complementation experiments in ABC DLBCL cells described above (Fig. 4 A and B).

Fig. 5.

IKKβ inhibition during activation of NF-κB by diverse stimuli elicits compensatory IKKα activity. (A) Flow-cytometric measurement of expression of the NF-κB target gene CD83 expression in cells with or without expression of IKKα shRNA or control luciferase shRNA and in the presence or absence of IKKβ inhibition after PMA/ionomycin stimulation of the GCB DLBCL cell line BJAB. (B) CD83 in BJAB cells expressing IKKα shRNA and either wild-type or kinase-dead IKKα, in the presence or absence of IKKβ inhibition after PMA/ionomycin stimulation. (C) CD83 in BJAB cells expressing the exogenous mutated form of CARD11 with expression of IKKα shRNA or luciferase shRNA and in the presence or absence of IKKβ inhibition. (D) NF-κB-driven expression of EGFP in the Jurkat cell line JPM50.6 expressing a control or IKKα shRNA, with or without stimulation using TNFα, in the presence or absence of IKKβ inhibition.

Mutations in the coiled-coil domain of CARD11 occur in roughly 10% of primary ABC DLBCL tumors, and introduction of these mutant CARD11 isoforms into GCB DLBCL cell lines activates the classical NF-κB pathway constitutively (4). We therefore tested whether a mutant CARD11 isoform could evoke compensatory IKKα activity in BJAB cells, as did P/I treatment. BJAB cells were engineered to express a mutated form of CARD11 in a doxycycline-inducible fashion. These cells were superinfected with retroviral vectors to express IKKα shRNA or a control shRNA. CD83 was upregulated upon expression of mutant CARD11, and this induction was modestly suppressed by the IKKβ inhibitor in cells bearing the control shRNA (Fig. 5C, left). However, in cells bearing the IKKα shRNA, the effect of the IKKβ inhibitor on CD83 expression was much more profound (Fig. 5C, right). These findings suggest that CARD11 is involved in the compensatory activation of IKKα in ABC DLBCLs.

Finally, we investigated whether other stimuli known to activate the classical NF-κB pathway are also associated with compensatory IKKα activation during IKKβ inhibition. The Jurkat T cell line JPM50.6 possesses an NF-κB-driven GFP reporter that responds to tumor necrosis factor-α (TNFα) stimulation. JPM50.6 cells were infected with vectors expressing IKKα shRNA or a control CARD11 shRNA, and the effect of TNFα on GFP was determined in the presence or absence of the IKKβ inhibitor. TNFα markedly enhanced GFP expression, and this was suppressed by the IKKβ inhibitor (Fig. 5D, left). Expression of the IKKα shRNA by itself did not alter GFP levels (Fig. 5D, compare right and left panels), but these cells had a much greater response to the IKKβ inhibitor than cells bearing the control shRNA (Fig. 5D, right). Thus, compensatory IKKα activity is apparently also invoked during TNFα stimulation of T cells when IKKβ is inhibited.

Discussion

Using an RNAi-based sensitization screen, we have uncovered a role for IKKα in the classical NF-κB pathway under conditions in which IKKβ is inhibited. We characterize this phenomenon as compensatory, as we observed little or no effect of knocking down IKKα without concurrent inhibition of IKKβ in a variety of assays including cell death, caspase 3/7 activation, NF-κB target gene expression and IκBα kinase activity in ABC DLBCL cells, expression of the NF-κB target gene CD83 in a P/I-stimulated B cell line, and NF-κB reporter activity in a TNFα-stimulated T-cell line. Under conditions of IKKβ inhibition, IKKα knockdown had a clear effect on each of these assays. Previous biochemical and genetic studies demonstrated that IKKα can serve as an IκBα kinase (20, 21, 22). However these studies were not designed to reveal a compensatory role of IKKα since they did not measure the function of IKKα as an IκBα kinase quantitatively in the presence and absence of acute IKKβ inhibition, as we did.

This compensatory IKKα mechanism is engaged in lymphomas of the ABC DLBCL subtype, which have constitutive activity of the classical NF-κB pathway that relies on signaling through CARD11. This phenomenon is not restricted to “pathological” NF-κB signaling but was also observed during inducible activation of NF-κB by treatment with P/I or TNFα. It is important to emphasize that IKKα did not play a compensatory role in all cells with classical NF-κB signaling: cell line models of primary mediastinal B cell lymphoma and multiple myeloma were dependent upon IKKβ for survival, yet IKKα shRNAs did not synergize in killing these cells (Fig. 2D). We conclude that the compensatory role of IKKα during IKKβ inhibition is NF-κB stimulus-dependent.

In certain cellular contexts, IκBα kinase activity can be completely dependent on IKKα. For example, knock-in mice with a catalytically inert form of IKKα fail to produce milk, which was traced to a role for IKKα as an IκBα kinase after RANK ligand stimulation of primary mammary epithelial cells (23). Likewise, IKKα acted as an IκBα kinase during CD40 and CD27 stimulation of a Burkitt lymphoma cell line (12), and during lymphotoxin α stimulation of fibroblasts (24). Although IKKβ is required for the maintenance of all mature B cells (8, 25, 26), the role of IKKα in B lymphocyte development is more restricted. B-cell-intrinsic IKKα kinase activity is required for the development of germinal center B cells and their differentiation into long-lived plasma cells (27). This study made use of a knock-in mouse strain in which the serines in the IKKα activation loop were mutated to alanine, resulting in a form of the kinase that could not be activated by NIK. NIK function in B cells is also required for their differentiation to the germinal center stage in response to antigenic challenge (28), suggesting that the IKKα function in germinal center B cells may depend on NIK. By contrast, the function of IKKα in ABC DLBCL cells does not apparently depend on NIK, as it was impervious to almost complete knockdown of NIK expression. Therefore, it is possible that some other kinase besides NIK may activate IKKα in ABC DLBCL cells and possibly also during CBM-dependent and TNFα-dependent signaling. Because TAK1 phosphorylates IKKβ during CBM and TNFα signaling, it might also activate IKKα, which is in the same protein complex as IKKβ (29, 30, 31, 32).

Several biochemical mechanisms can be envisioned to account for the compensatory IKKα activity that we observed. It is first important to note that there was no change in the abundance of either IKKα or IKKβ protein upon IKKβ inhibition (Fig. S5), supporting mechanisms that alter the ability of IKKα to serve as an IκBα kinase. Consideration of the targets of IKKβ phosphorylation suggests several mechanisms to account for the compensatory IKKα activity. First, because IKKβ and IKKα are in the same high-molecular-weight complex in the cytoplasm, it is conceivable that by inhibiting the ability of IKKβ to phosphorylate IκBα, IκBα may be more available to IKKα as a substrate. In this regard, it is notable that IKKα is considerably less active than IKKβ as an IκBα kinase in vitro (33), and thus if IκBα is brought in proximity to the IKKαβγ complex during cellular signaling, it may be preferentially phosphorylated by IKKβ. During antigen receptor stimulation, BCL10 is an additional substrate of IKKβ (34, 35). Phosphorylation of BCL10 by IKKβ attenuates NF-κB signaling either by “remodeling” the CBM signaling complex or by promoting BCL10 degradation. Indeed, treatment of ABC DLBCL cells with the IKKβ inhibitor causes dramatic upregulation of BCL10 protein expression (V. Ngo, unpublished results). Thus, IKKβ inhibition could enhance CBM complex formation by stabilizing BCL10, conceivably leading to exaggerated activation of IKKα as a kinase. Finally, phosphorylation of serines in the carboxy-terminal region of IKKβ, presumably by autophosphorylation, decreases its kinase activity (36). Interestingly, IKKα has an analogous serine-rich region in its carboxy-terminus (36). It is therefore possible that phosphorylation of this region by IKKβ might inhibit IKKα kinase activity, thereby accounting for the compensatory IKKα activity that we observed.

Our observations suggest that therapeutic targeting of both IKKα and IKKβ may have a more potent and synergistic effect on the classical NF-κB pathway than targeting IKKβ alone. A potential dual inhibitor of both kinases is the NEMO binding domain peptide, comprising an amino acid sequence that is identically conserved in both IKKα and IKKβ that is responsible for interaction with the NEMO subunit of the IKK complex (37). We treated ABC DLBCL cells harboring the IκBα-luciferase reporter system with a cell-permeable version of this peptide, and observed that it inhibited IκBα kinase activity as a single agent and added to the effect of the IKKβ inhibitor MLN120B (Fig. S6). Most pharmaceutical development has been aimed at IKKβ because of its prominent role in the pro-inflammatory functions of the classical NF-κB pathway. As IKKβ inhibition may have undesired side effects such as increased secretion of IL-1β during sepsis (38), it is possible that inhibition of IKKα might allow use of lower doses of IKKβ inhibitors to achieve comparable blockade of the classical NF-κB pathway. Because IKKα kinase activity appears to be required only for lactation and germinal center B cell differentiation in mouse models (23, 27), combined pharmacologic inhibition of this kinase and IKKβ should be considered for therapy in patients with ABC DLBCL and potentially other NF-κB-mediated diseases.

Materials and Methods

RNA Interference Library Screen.

The ABC DLBCL cell line OCI-Ly3 was infected with a library of retroviruses expressing shRNAs, as described (3). Briefly, after selecting for infected cells with puromycin, doxycycline was added to induce shRNA expression for 2 days before treating cells with 25 μmol/l MLX105. Control infected cells were treated with doxycycline but not MLX105. Cells were harvested after 3 days, genomic DNA was isolated, and the unique 60-base pair molecular bar code in each shRNA vector was amplified by PCR. DNA microarrays consisting of the bar code sequences were used to compare the relative abundance of each shRNA vector in the MLX105-treated and control cell populations. Four biological replicates were performed, and statistical analysis identified shRNAs that were significantly depleted or enriched in the MLX105-treated cells versus control (P < 0.005).

Analysis of Individual shRNAs.

shRNA sequences were as follows: GCCAGATACTTTCTTTACTAA (shIKKα1), GGTGGAAAGATAATACATAAA (shIKKα2), and GAGGGCTGGTTAATGTAGTAT (shIKKα3). IKKβ shRNA (3) and NIK shRNA (9) sequences were described previously. IKKα knockdown was assessed by Western blotting (39).

Cell Line Assays.

Flow cytometry for analysis of NF-κB activation was performed 2 days after transgene and shRNA induction with doxycycline (20 ng/ml). CD83 expression in BJAB cells and NF-κB-driven EGFP expression in the Jurkat T-cell line JPM50.6 (40) were determined as previously described (4). The IκB-luciferase reporter was assayed as described (5). To assess synergy between IKKα knockdown and IKKβ inhibition, OCI-Ly3 cells harboring IKKα shRNAs were induced with doxycycline for 2 days and treated with various concentrations of MLN120B (0–25 μmol/l) for another 2 days before being assayed for viability using MTT as described (41). Caspase 3/7 activity was measured using the Caspase-Glo 3/7 assay (Promega).

Supplemental methods are available as SI Text.