We discovered that besides activating canonical NF-κB, expression of API2-MALT1 in HEK293T or SSK41 B-lymphoma cells induced proteasome-dependent processing of the NF-κB precursor, p100, to its mature form, p52, and stimulated nuclear translocation of p52/RelB (Fig. 1, A–D and fig. S1). These findings indicate that API2-MALT1 activates the “noncanonical” NF-κB pathway, which requires NIK-dependent phosphorylation and activation of IκB kinase-α (IKKα). This in turn triggers proteasome-mediated partial degradation of p100 to p52 and generation of transcriptionally active p52/RelB NF-κB dimers (7). Consistent with this notion, dominant negative NIK mutants (8, 9) blocked API2-MALT1-dependent p100 processing and p52 nuclear translocation (Fig. 1, E and F).

cIAP1 and cIAP2 (API2) associate with NIK and promote NIK degradation via RING domain ubiquitin ligase activity (10–12). We hypothesized that API2-MALT1, which lacks the cIAP2 RING domain, stimulates noncanonical signaling through competitive inhibition of cIAP-mediated NIK degradation. In testing this, we discovered that expression of API2-MALT1 instead induced proteolytic cleavage of NIK, generating ~37 kD N-terminal and ~70 kD C-terminal NIK fragments (Fig. 2, A and B and fig. S2).

API2-MALT1 fusion transcripts invariably contain three intact baculoviral IAP repeat (BIR) domains from API2 and an intact “caspase-like” domain from MALT1, suggesting that these domains are critical for lymphomagenesis (1). The caspase-like domain of wild-type MALT1 possesses proteolytic activity, and Bcl10 and the NF-κB inhibitor, A20, are the only known substrates (13, 14). We therefore investigated whether the caspase-like domain within API2-MALT1 is also able to cleave NIK. Deletion mutants of API2-MALT1 lacking portions of the caspase-like domain and API2-MALT1-C678A, in which the catalytic cysteine within the MALT1 proteolytic domain is replaced with alanine, were unable to induce NIK cleavage (Fig. 2C and fig. S3, A–C). Furthermore, treatment with z-VRPR-fmk, a MALT1 protease inhibitor (14), blocked API2-MALT1-induced NIK cleavage, whereas z-IETD-fmk, a caspase-8 inhibitor, had no effect (fig. S3, D and E). Finally, an in vitro cleavage reaction using purified recombinant proteins showed that NIK is a direct substrate of the MALT1 protease domain (Fig. 2D and fig. S4).

Cellular expression of the MALT1 moiety alone was unable to induce NIK cleavage, suggesting that the API2 moiety also contributes in some way (Fig. 2E). Indeed, analyses revealed that NIK physically associates with API2-MALT1 via the API2 moiety (Fig. 2F), and that the region within the API2 moiety that mediates auto-oligomerization of API2-MALT1 (AA 49–98) (5) is required for efficient API2-MALT1-dependent NIK cleavage and p100 processing (Fig. 2F and fig. S5). The collaborative relationship of the API2 and MALT1 moieties in achieving NIK cleavage was further underscored in several different contexts. First, unlike API2-MALT1, induced expression of wild-type MALT1 in BJAB B-cells did not result in NIK cleavage (fig. S6A). Second, NIK cleavage was not observed in SSK41 B-cells, which are characterized by MALT1 gene amplification, overexpression of MALT1 and constitutive MALT1 protease activity (13, 15, 16). In contrast to A20, NIK was cleaved only if SSK41 cells were engineered to express API2-MALT1 (fig. S6B). Third, co-expression of wild-type MALT1 with Bcl10, which triggers MALT1 oligomerization and activation (3), did lead to cleavage of A20 but not of NIK (fig. S6, C and D). Fourth, ligand induced B-cell receptor stimulation, which similarly activates the MALT1 protease (13), did not trigger NIK cleavage (fig. S6E). Furthermore, although MALT1 oligomerization and activation requires Bcl10 (3), API2-MALT1-mediated NIK cleavage occurred in the absence of Bcl10 (fig. S6F). Together, these findings suggest that NIK is a substrate for the MALT1 protease domain, but only when this domain is present within the context of API2-MALT1.

Structural analyses predict that the MALT1 protease should show specificity for substrates with a basic or uncharged residue at P1 (N-terminal to the cleavage site) (17). Furthermore, the MALT1 cleavage sites of Bcl10 and A20 both contain a P2-serine preceding a P1-arginine (13, 14, 18). Thus, we identified candidate P2-Ser/P1-Arg MALT1 cleavage sites within NIK that would generate fragments of ~37 and 70 kD (fig. S7A), and we individually changed each candidate P1-Arg(R) to Ala(A). The R366A and R368A NIK mutants were readily cleaved by API2-MALT1; however, the R325A mutant was resistant (Fig. 3A), suggesting that API2-MALT1-dependent cleavage of NIK occurs at R325 (fig. S7B). Expression of the resulting C-terminal NIK cleavage fragment, NIK(326–947), which retains the kinase domain, induced robust p100 processing (fig. S7C), p52 nuclear translocation (Fig. 3B), and noncanonical NF-κB target gene expression (fig. S7D). NIK(326–947) also induced nuclear translocation of the p65 NF-κB subunit, indicating activation of the canonical NF-κB pathway as well (Fig. 3B). Conversely, API2-MALT1-C678A, the catalytically inactive mutant that cannot induce NIK cleavage to produce NIK(326–947), failed to stimulate p100 processing (Fig. 3C) and p52 nuclear translocation (Fig. 3D and fig. S8).

NIK associates with the adaptor protein, TRAF3, via an N-terminal NIK domain (AAs 78–84), and this interaction targets NIK for proteasomal degradation (19–21). Because cleavage of NIK at R325 separates this TRAF3-binding site from the NIK kinase domain, we hypothesized that the active C-terminal NIK cleavage product would be resistant to TRAF3-directed degradation. Indeed, NIK(326–947) retained binding to IKKα but not to TRAF3 (Fig. 3E) and, unlike full-length NIK, was resistant to TRAF3-dependent proteasomal degradation (Fig. 3F). We also demonstrated the unique stability of the API2-MALT1-generated C-terminal NIK cleavage fragment in SSK41 B-lymphoma cells. In the absence of MG132, expression of full-length NIK was very low regardless of whether API2-MALT1 was present, consistent with the fact that NIK is subject to constitutive proteasomal degradation (19, 22) (Fig. 3G). In contrast, high amounts of endogenous C-terminal NIK cleavage fragment were detected in SSK41 cells expressing API2-MALT1 (Fig. 3G).

API2-MALT1-dependent generation of the C-terminal NIK cleavage fragment in SSK41 cells was associated with enhanced transcription of noncanonical NF-κB target genes, including Pim-2, an oncogenic kinase that blocks apoptosis by phosphorylating the pro-apoptotic Bcl-2 family member, BAD (figs. S9, A and B) (23–27). RNAi-mediated knock-down of NIK in API2-MALT1 expressing SSK41 cells led to loss of the 70 kD NIK fragment, loss of p100 processing, and loss of API2-MALT1-dependent induction of Pim-2 kinase and BAD phosphorylation (Fig. 4, A and B). In accordance with its impact on anti-apoptotic signal transduction, API2-MALT1 expression protected SSK41 cells from dexamethasone-induced cell death, which was reversed by NIK knock-down (Fig. 4, B and C and fig. S10). Knockdown of IKKα impaired API2-MALT1-dependent protection, supporting a role for the noncanonical NF-κB pathway in mediating this effect of API2-MALT1-induced NIK cleavage (Fig. S11). IKKβ knockdown impaired API2-MALT1-dependent protection as well, implying that the canonical pathway may also contribute (Fig. S11).

We next investigated the impact of API2-MALT1-dependent NIK cleavage on B-cell adhesion because we had observed that API2-MALT1 induced the expression of B-cell integrins (fig. S9A), known noncanonical NF-κB gene targets (25, 26). API2-MALT1 expression was associated with increased B-cell adhesion to plates coated with the endothelial protein, VCAM-1, and this pro-adhesive phenotype was fully dependent on NIK (Fig. 4D). Lymphocyte adhesion is thought to play role in lymphoma dissemination, thus NIK-cleavage-dependent API2-MALT1-induced adhesion may contribute to the higher rate of tumor spread among t(11;18)-positive MALT lymphomas (1). Again, knockdown of IKKα or IKKβ impaired API2-MALT1-dependent adhesion, suggesting that both noncanonical and canonical pathways contribute to the proadhesive phenotype following API2-MALT1-dependent NIK cleavage (fig. S12).

The striking pattern of NIK cleavage and stability observed in API2-MALT1-expressing B-lymphoma cell lines was recapitulated in MALT lymphoma patient specimens. Full-length NIK levels were relatively low in all lymphomas, whereas an endogenous 70 kD C-terminal NIK fragment was detected only in t(11;18)-positive MALT lymphomas expressing API2-MALT1 (Fig. 4E). The presence of the NIK cleavage product was associated with an elevated p52/p100 ratio in the t(11;18)-positive tumors, indicating enhanced noncanonical NF-κB activation (Fig. 4E). We performed absolute gene set enrichment analysis (GSEA) comparing the expression of NF-κB target genes between t(11;18)-positive MALT lymphomas (n=9) vs. those with no translocation (n=8). Analysis revealed a significant difference in the pattern of expression, with most known noncanonical NF-κB target genes over-represented in the t(11;18)-positive cases (Fig. 4F and fig. S13, tables S1 and S2). One exception is cyclin D1, although its categorization as a noncanonical NF-κB target gene is controversial (28). We then compared the expression of these noncanonical target genes in a completely separate group of six t(11;18)-positive and eight t(11;18)-negative MALT lymphomas that were collected at a different institution, and again found that many noncanonical NF-κB target genes were more highly expressed in the t(11;18)-positive tumors (Fig. 4G). CXCR4, which encodes a chemokine receptor whose expression is associated with widespread lymph node involvement in B-lymphomas (29–31), is one intriguing example of a noncanonical gene target that is upregulated in t(11;18)-positive tumors in both tumor collections (Fig. 4F and G).

Deregulated NIK activity has been increasingly implicated in the pathogenesis of B-cell neoplasms. For example, an EFTUD2-NIK fusion oncoprotein that retains the NIK kinase domain but lacks the TRAF3 binding site, and is resistant to proteasomal degradation, was recently identified in a case of multiple myeloma (22). Another recent report described a murine model of B-lymphoproliferative disease in which a NIK mutant lacking the TRAF3 binding domain (NIKΔT3) was expressed in B-cells (32). Compared to control transgenic mice expressing full-length NIK, the NIKΔT3 mice demonstrated increased NIK levels with enhanced p100 processing in B-cells. They also showed expanded MALT and profound splenic marginal zone B-cell hyperplasia, a phenotype that bears similarity to the Eµ-API2-MALT1 transgenic mouse (33). Together with our results, these findings suggest that separating the TRAF3 binding site on NIK from the kinase domain, either through aberrations of the NIK gene or through proteolytic cleavage of NIK protein, may represent a common mechanism for deregulating NIK activity in B-cell neoplasms. Our findings suggest that in API2-MALT1-expressing MALT lymphomas, the API2 moiety mediates auto-oligomerization of API2-MALT1 and recruitment of NIK, and the MALT1 protease domain cleaves NIK, leading to degradation-resistant NIK kinase and deregulated noncanonical NF-κB signaling (see model, fig. S14). Data suggest that NIK cleavage protects API2-MALT1-expressing B-cells from apoptosis and promotes B-cell adhesion, both of which could contribute to the more aggressive phenotype of t(11;18)-positive MALT lymphomas (1). Disrupting the API2-NIK interaction and/or blocking MALT1 protease or NIK kinase activity could represent new treatment approaches for refractory t(11;18)-positive MALT lymphoma.

Supplement