Synergistic cooperation and crosstalk between MYD88^L265P and mutations that dysregulate CD79B and surface IgM

Diffuse large B cell lymphoma (DLBCL) is one of the most frequent and aggressive B cell malignancies (Lenz and Staudt, 2010). The activated B cell type of DLBCL (ABC-DLBCL) represents a particularly aggressive form, distinguished by constitutive activation of the canonical NF-κB transcription factor family and by poor patient survival and response to the standard treatment regimen of R-CHOP (Lenz and Staudt, 2010). NF-κB transcription factors are normally activated by two key receptors for microbes on B cells, the B cell antigen receptor (BCR) and the TLRs, and serve as essential inducers of normal B cell survival, growth, and differentiation (Thome, 2004; Gerondakis and Siebenlist, 2010; Hayden and Ghosh, 2012). Somatic mutations in CD79B and MYD88, affecting critical proteins for BCR and TLR signaling, respectively, are recurrently found together in ABC-DLBCLs (Lenz et al., 2008; Davis et al., 2010; Ngo et al., 2011). It is not known if these mutations act cooperatively to dysregulate B cell survival, growth, or other functions or how cooperation and crosstalk might occur.

Missense mutations in MYD88 occur in 39% of cases of ABC-DLBCLs, with a single L265P amino acid substitution accounting for 75% of the mutations (Ngo et al., 2011). The same MYD88^L265P mutation occurs in almost 100% of Waldenström macroglobulinemia (WM), 47% of IgM monoclonal gammopathy of undetermined significance, and 3–10% of chronic lymphocytic leukemia (Puente et al., 2011; Wang et al., 2011; Treon et al., 2012; Xu et al., 2013). MYD88 is an essential cytoplasmic adaptor protein, downstream from most TLRs and the IL-1/18 receptor, required to activate the IL-1 receptor–associated kinases (IRAKs) and NF-κB (Akira and Takeda, 2004). MYD88 has two distinct domains. A Toll/IL-1R domain (TIR) promotes homotypic and heterotypic multimerization of MYD88 proteins upon recruitment to dimerized TIR domains in the cytoplasmic tail of TLRs that have been engaged by their microbial ligands (Vyncke et al., 2016). A death domain forms a helical multimeric signaling complex known as the Myddosome comprising six MYD88 molecules, four IRAK4 molecules, and four IRAK2 molecules (Akira and Takeda, 2004; Lin et al., 2010). The MYD88^L265P mutation in the TIR domain is predicted to cause allosteric changes in two binding surfaces and has been shown to promote multimerization with wild-type MYD88 and spontaneous formation of the MYD88-IRAK signaling complex, resulting in elevated NF-κB activity (Ngo et al., 2011; Avbelj et al., 2014; Vyncke et al., 2016). When introduced into mature mouse B cells by retroviral transduction, MYD88^L265P is sufficient to initiate mitogen and T cell independent B cell proliferation that is terminated after several cell divisions, in part by feedback inhibition of NF-κB (Wang et al., 2014). More recently, a mouse model bearing a conditional Myd88^L252P allele has been described to develop lymphoproliferative disease with occasional transformation into clonal lymphomas (Knittel et al., 2016). Conversely, knockdown of MYD88 potently kills ABC-DLBCL cell lines, establishing that these tumors are addicted to MYD88 activation for survival (Ngo et al., 2011).

Somatic mutations in CD79B occur in 21% of ABC-DLBCLs (Davis et al., 2010). CD79B and CD79A associate noncovalently with membrane immunoglobulin, serving as the signal-transducing subunits of the BCR through an immunoreceptor tyrosine-based activation motif (ITAM) in the CD79B and CD79A cytoplasmic tails (Reth and Wienands, 1997). Upon antigen binding, the two tyrosines in each ITAM are phosphorylated by LYN and other SRC-family kinases, providing a docking site for the paired SH2 domains of SYK, activating SYK, and initiating the intracellular signaling cascade (Cambier et al., 1994). 85% of CD79B mutations substitute the first ITAM tyrosine residue at position 196 (Y196) to another amino acid, most frequently histidine (Davis et al., 2010). Unlike MYD88 mutations, CD79B ITAM mutations do not spontaneously activate NF-κB in ABC-DLBCL cell lines (Lenz et al., 2008; Davis et al., 2010). Instead, CD79B ITAM mutations cause elevated surface BCR expression, possibly by inhibiting Lyn-mediated receptor internalization, resulting in higher surface BCR expression on ABC-DLBCLs but not in other tumors absent for CD79B mutations (Davis et al., 2010). In mice with a targeted mutation substituting alanine in place of both tyrosine residues in the CD79B ITAM, the mature B cells displayed more BCRs on their surface, had delayed BCR internalization after antigen binding, and had exaggerated BCR signaling to calcium, extracellular signal-regulated kinase (ERK), and AKT but normal signaling to NF-κB (Gazumyan et al., 2006). Therefore, it is speculated that the likely role of CD79B mutation in the pathogenesis of ABC-DLBCL is by allowing B cells to respond inappropriately to BCR stimulation by foreign or self-antigens (Rui et al., 2011). However this hypothesis remains to be tested experimentally.

One third of ABC-DLBCLs bearing the MYD88^L265P mutation also have CD79B ITAM mutations (Ngo et al., 2011). shRNA knockdown of CD79B and other components in BCR signaling in ABC-DLBCLs synergizes with shRNA against MYD88 in killing ABC-DLBCL cell lines in vitro, suggesting that BCR signaling through CD79B plays an important and nonredundant role in supplementing the growth and survival of MYD88^L265P tumors (Ngo et al., 2011). In a phase 1/2 clinical trial of the Bruton tyrosine kinase (BTK) inhibitor ibrutinib, complete or partial responses occurred in four of five patients with ABC-DLBCL bearing both CD79B and MYD88^L265P mutations, contrasting with zero of seven patients with MYD88^L265P ABC-DLBCL without CD79B mutations (Wilson et al., 2015). However, it is not known if CD79B and MYD88 mutations cooperate to dysregulate normal B cell survival and growth or how such cooperation might impose dependency on BTK. Addressing this question may also illuminate fundamental mechanisms of synergistic and antagonistic crosstalk between antigen receptor and TLR signals in B cells (Rui et al., 2003, 2006).

B cells actively acquire tolerance to self-antigens through a series of bone marrow and peripheral lymphoid tissue checkpoints that inhibit B cell survival, growth, and plasma cell differentiation (Goodnow, 2007). B cells undergo multiple rounds of clonal proliferation upon binding to foreign antigen, but undergo anergy or apoptosis when their receptors are repeatedly engaged by self-antigens. Many human DLBCLs, and to a lesser extent follicular, marginal zone, and Burkitt’s lymphoma cells, express BCRs that are self-reactive because they use the VH4-34 element (Silberstein et al., 1991; Zhu et al., 1994; Hsu and Levy, 1995; Ottensmeier et al., 1998; Lossos et al., 2000; Young et al., 2015), which binds to human erythrocyte I-i polysaccharide antigen and human B lymphocyte CD45 O-linked polysaccharide antigen (Pascual et al., 1991; Grillot-Courvalin et al., 1992). A recent study showed that self-antigen engagement of the BCR is required for the survival of some ABC-DLBCL cells, notably in a VH4-34⁺ lymphoma in which point mutations in the binding site that disrupt self I-i binding diminished the capacity of that BCR to support survival or growth of the lymphoma in tissue culture (Young et al., 2015). In healthy people, VH4-34⁺ B cells with intact I-i binding sites are negatively selected against accumulating in the germinal center, memory B cell, or plasma cell repertoires (Pugh-Bernard et al., 2001; Reed et al., 2016). Therefore, a key question is whether CD79B mutations disrupt normal peripheral B cell tolerance checkpoints and switch the B cell response to self-antigen from apoptosis into proliferation, as has been shown for CARD11 mutations (Jeelall et al., 2012).

Here, we analyze the consequences of CD79B mutations in normal B cells and test if cooperation between MYD88 and CD79B somatic mutations disrupts peripheral self-tolerance checkpoints in B cells. We show that CD79B mutation on its own does not promote spontaneous growth or survival, nor does it block self-antigen–induced peripheral tolerance. However, extraordinary epistasis between CD79B and MYD88^L265P occurs when coexpressed, blocking self-tolerance checkpoints and promoting spontaneous plasmablast differentiation. We show that MYD88^L265P spontaneously triggers a block in intracellular processing and surface expression of BCRs similar to that observed in self-reactive B cells, and this block is relieved either by CD79B ITAM mutation or by increasing expression of wild-type CD79B. These findings reveal an unexpected, potent crosstalk and cooperation between mutations affecting the BCR and TLR signaling pathways and provide an explanation for the frequent co-occurrence of MYD88 and CD79B mutations in ABC-DLBCLs.

To model the acquisition by activated B cells of somatic mutations in MYD88 and CD79B, we transiently activated normal spleen B cells with a 6-h exposure to hen egg lysozyme (HEL) antigen in vivo followed by 24-h culture with anti-CD40 antibody but no further exposure to BCR ligand. The activated B cells were then transduced with retroviral vectors during a further 36-h culture with anti-CD40 antibody but not BCR ligand, after which the cells were washed free of anti-CD40 and cultured in medium alone. When activated B cells are transduced with vectors encoding MYD88^L265P or CARD11^L232LI and then placed back into culture with no mitogen, they are induced to divide repeatedly, in contrast to control cells transduced with empty vector or vectors encoding normal versions of these proteins, which do not divide but instead rapidly die (Jeelall et al., 2012; Wang et al., 2014). We therefore used the same assay to test if retroviral expression of CD79B bearing ITAM mutations found in lymphomas could also drive mitogen-independent cell proliferation in otherwise normal B cells. We transduced activated mature B cells with bicistronic EGFP vectors encoding either wild-type CD79B or CD79B with missense mutations in the first ITAM tyrosine, corresponding to the recurring human Y196H or Y196F mutations (Davis et al., 2010), or a Y207F substitution in the second ITAM, or Y196F combined with Y207F (Fig. 1, A and B). After transduction, the cells were washed and placed back in culture for 3 d. In the absence of mitogen, the numbers of live EGFP⁺ B cells steadily declined at the same rate as empty vector transduced B cells that expressed only EGFP, and at the same rate as EGFP⁻ cells in the same culture (Fig. 1 B). Thus, enforced expression of CD79B wild-type or mutants alone conferred no measurable survival advantage to primary B cells. Analysis of cell trace violet dilution also showed no induction of cell division (not depicted).

The cell surface proteins CD23 and CD25 are encoded by genes directly induced by NF-κB, and both are markedly increased on primary B cells transduced with retroviral vectors encoding mutant alleles of CARD11 or IκB kinase β (IKK) found in human lymphomas (Wang et al., 2014). Using the same assay, expression of CD79B ITAM mutant proteins did not induce surface CD23 on EGFP⁺ cells relative to nontransduced EGFP⁻ cells in the same culture, relative to EGFP⁺ cells expressing wild-type CD79B or empty vector (Fig. 1 C). Nor did the expression of CD79B ITAM mutant proteins induce surface CD25 relative to EGFP⁺ cells expressing wild-type CD79B (Fig. 1 C).

Because Y196H is the most frequent CD79B mutation co-occurring with MYD88^L265P (Davis et al., 2010), we next analyzed the consequences of CD79B^Y196H coexpressed with MYD88^L265P. Coexpression of CD79B^Y196H and MYD88^L265P failed to extend the transient phase of B cell proliferation induced by MYD88^L265P alone in vitro (Fig. 1 D). The lack of any effect contrasts with the extended proliferation when MYD88^L265P was combined with loss-of-function mutations in Tnfaip3 (A20) or Bcl2l11 (BIM) or gain-of-expression of Bcl2 (Wang et al., 2014). To assess whether CD79B mutation affected the negative feedback of NF-κB by MYD88^L265P, we also measured NF-κB–inducible protein markers on doubly transduced B cells as previously described (Wang et al., 2014). As a positive control, we also transduced B cells with IKK bearing an activating mutation found in human lymphoma, which strongly induced CD23, CD25, and CD95 (Jeelall et al., 2012). In contrast, all markers were down-regulated relative to empty vector controls on B cells expressing MYD88^L265P alone and were comparably down-regulated on B cells coexpressing CD79B^Y196H and MYD88^L265P (Fig. 1 E). The lack of any measurable effect of CD79B in these assays contrasts with our previous findings, in which loss-of-function mutations in Tnfaip3 (A20) blocked MYD88^L265P-induced down-regulation of CD23, CD25, CD86, and CD95 (Wang et al., 2014).

Next we considered whether CD79B ITAM mutations affected surface BCR expression on retrovirally transduced primary B cells, because lymphomas with CD79B mutations have elevated BCR expression (Davis et al., 2010). Activated mature B cells were transduced with retroviral vectors, cultured in the absence of mitogen for 24 h, and analyzed by flow cytometry for surface CD79B and IgM. Compared with B cells transduced with empty vector, B cells expressing the wild-type CD79B retroviral vector had 150% increased surface CD79B and surface IgM, while this was increased to 200% of empty vector controls in cells expressing CD79B^Y196H (Fig. 2 A). In contrast, surface CD79B and IgM were dramatically decreased on EGFP⁺ cells expressing MYD88^L265P alone, representing only 30% of the levels on empty vector EGFP⁺ cells (Fig. 2 B).

These convergent but opposing effects of mutant MYD88 and CD79B on BCR surface levels were then analyzed in more detail. Surface IgM and CD79B were decreased by expression of the L265P mutant but not wild-type MYD88 (Fig. 2 C). A cell surface protein not part of the BCR complex, CD44, was unaffected. Dual expression of CD79B^WT MYD88^L265P or CD79B^Y196H MYD88^L265P in B cells opposed the reduction of surface CD79B and IgM caused by MYD88^L265P alone, resulting in receptor levels comparable to empty vector controls but lower than cells with the CD79B vectors but without MYD88^L265P (Fig. 2 C).

These results revealed an unexpected inhibitory effect of MYD88^L265P on surface IgM receptor levels. To test if this occurred at the level of gene transcription, we measured mRNAs encoding the four protein subunits required for surface IgM display: Ighm encoding IgM μ-heavy chains, Igk encoding κ-light chains, Cd79a, and Cd79b. These were analyzed in microarray datasets of whole transcriptomes from EGFP⁺ B cells sorted from two independent cultures expressing empty vector, MYD88^L265P, CARD11^L232LI, or IKK^K171E as previously described (Wang et al., 2014). As validation of the microarray data, Myd88 mRNA was increased 800% in samples of MYD88^L265P transduced cells relative to all the other samples (Fig. 3 A). Consistent with the flow cytometric data in Fig. 1 E, the NF-κB target gene Fcer2a encoding CD23 was decreased in MYD88^L265P-transduced cells relative to empty vector but increased in IKBKB^K171E-transduced cells. In contrast, MYD88^L265P did not decrease Ighm or Igk mRNA levels and only marginally decreased Cd79a and Cd79b (Fig. 3 A). To verify this result, EGFP⁺ cells expressing empty vector or MYD88^L265P were sorted from three independently transduced cultures of each, and cDNA was analyzed by quantitative PCR. This revealed increased Myd88 mRNA in MYD88^L265P-transduced cells but no significant difference in Ighm, Cd79a, or Cd79b mRNA (Fig. 3 B).

Because the decreased surface IgM could not be explained by changes at the mRNA level, we used Western blotting of sorted EGFP⁺ cell lysates to measure the relative amounts of CD79B and IgM proteins in the pre-Golgi or post-Golgi compartments. N-linked glycans on newly synthesized (immature) CD79B and IgM μ-chains constitute high-mannose chains that can be cleaved by endoglycosidase H (Endo H), resulting in a shift to a lower molecular weight in SDS-PAGE. The N-linked carbohydrates are modified and become resistant to Endo H once the mature IgM-CD79A/B complex is assembled and transported through the trans-Golgi (Bell and Goodnow, 1994). Protein extracts from transduced and sorted EGFP⁺ B cells were treated with Endo H for 1 h at 37°C or mock-treated, followed by separation on SDS-PAGE and Western blotting with antibodies against CD79B, μ-chain, and tubulin. In B cells expressing empty vector, most of the CD79B and IgM μ-chain was in the mature, slowly migrating, Endo H–resistant form (Fig. 3 C). In contrast, in B cells expressing MYD88^L265P alone, IgM μ-chains were mostly immature and Endo H–sensitive, and these accumulated markedly (Fig. 3, C and D). In the same lysates, there was little change in the pool of immature CD79B, but there was a decrease in mature Endo H–resistant CD79B (Fig. 3, C and D). The pool of immature CD79B protein was increased in cells transduced with the CD79B vector, and coexpression of CD79B^WT with MYD88^L265P resulted in a higher proportion of μ-chains in the mature form and a decrease in the immature μ-chain pool (Fig. 3, C and D). Collectively, the mRNA and Western blotting analysis indicates that MYD88^L265P triggers a post-translational block in IgM receptor trafficking to the trans-Golgi, resembling the trafficking block that explains low surface IgM in anergic self-reactive B cells (Bell and Goodnow, 1994). This IgM trafficking “checkpoint” could be relieved by increasing the pool of nascent CD79B protein, although future studies will need to determine how the increase in CD79B reverses the effects of MYD88^L265P.

The wider transcriptional effects of MYD88^L265P were explored by analyzing the microarray datasets described above using the limma package (Ritchie et al., 2015), focusing on well-annotated genes with Entrez IDs. Using CAMERA, a competitive test of enrichment suitable for small sample sizes and resistant to intergene correlation (Wu and Smyth, 2012), we first queried for enrichment of differentially expressed genes against predefined sets of hallmark, curated, and immunological signature gene sets defined in MSigDB (Table S1; Subramanian et al., 2005). We focused on evidence for differential expression between MYD88^L265P- and CARD11^L232LI-transduced B cells because both cell populations were actively proliferating at the time of RNA isolation, unlike empty vector–transduced cells, and because CD79B mutations appear to be coselected with MYD88^L265P in human lymphomas but not obviously with CARD11 mutations.

Hallmark gene set M5890, comprising 207 genes regulated by NF-κB in response to TNF, was one of three hallmark gene sets most significantly enriched for differentially expressed genes between MYD88^L265P- and CARD11^L232LI-expressing B cells (false discovery rate [FDR] = 7.78 × 10⁻⁶; Table S2). The majority of NF-κB–induced mRNAs in this set were decreased in MYD88^L265P-expressing B cells (Fig. 4 A), consistent with our previous evidence that MYD88^L265P in transduced B cells induces negative feedback inhibition of NF-κB through TNFAIP3 (Wang et al., 2014).

Gene set M1487, comprising 55 genes that are up-regulated in plasma cells relative to other B cell developmental stages (Mori et al., 2008), was one of two curated gene sets most significantly enriched for differentially expressed genes between MYD88^L265P- and CARD11^L232LI-expressing B cells (FDR = 1.68 × 10⁻¹²; Table S3). The majority of the plasma cell up-regulated gene set was increased in MYD88^L265P-expressing cells (Fig. 4 B). This set was one of the four most significantly enriched gene sets of all queried sets in MSigDB (Table S1). The three other gene sets with comparably small FDR values all constitute genes increased in myeloid cell differentiation, which were decreased in MYD88^L265P-expressing cells.

When the transcriptome of MYD88^L265P-expressing B cells was compared with empty vector–expressing B cells using limma, 111 genes had strong evidence for differential expression with 88 increased in MYD88^L265P cells (Fig. 4 C and Table S4). Compared with CARD11^L232LI-expressing B cells, 84 genes had strong evidence for differential expression in MYD88^L265P B cells, with 30 increased, including the plasma cell–induced genes Prdm1 and Igj (J-chain; Fig. 4 D). Comparing MYD88^L265P B cells with all the other groups yielded 22 probe sets with strong evidence of differential expression in MYD88^L265P-transduced cells relative to all the other cell types examined, and all were increased in MYD88^L265P cells. Ranked by fold change relative to CARD11^L232LI-expressing cells, they were Hnf1b, Tmed6, Myd88, Mybl1, Pcbd1, Map2, Nqo1, Gm10558, Prdm1, Chac2, Rnd3, Ccl3, Tnip3, Fam46c, Ccr9, Ehf, Cdc14b, Slco4c1, Depdc7, Sfmbt2, Fam81a, and Fam135a. Myd88 was expected on the basis of the analysis above (Fig. 3), validating the limma analysis. Tnip3 encodes a negative regulator of NF-κB that interacts with TNFAIP3 (Verstrepen et al., 2014). Fam46c is a member of the curated plasma cell up-regulated gene set M1487 (Mori et al., 2008) and is frequently deleted or mutated in human multiple myeloma (Lohr et al., 2014).

Because B cell receptor signaling is critical for B cell survival and differentiation in vivo, we next asked if the CD79B^Y196H mutations would allow differentiation of MYD88^L265P-expressing B cells into plasmablasts when transplanted into Rag1-deficient mice, as observed for cells expressing single gain-of-function CARD11 mutations (Jeelall et al., 2012). Plasmablast differentiation from polyclonal C57BL/6 B cells could theoretically have been driven by BCRs that bound to self- or microbe antigens; therefore for these experiments we used HEL-specific B cells from MD4 transgenic mice that received no BCR stimulation after retroviral transduction and transplantation in vivo. HEL and anti-CD40 activated B cells were transduced with CD79B and/or MYD88 vectors, washed, and transplanted into Rag1^−/− recipient mice and analyzed in the spleen after 11 d.

Gating on the EGFP⁺ subset of transferred B cells revealed that those expressing CD79B^WT MYD88^L265P or CD79B^Y196H MYD88^L265P vectors accumulated in greater numbers and had lower CD19 and B220 expression compared with B cells expressing MYD88^L265P alone (Fig. 5, B–E). CD19 and B220 are two B cell lineage markers that are repressed upon differentiation into plasma cells (Nutt et al., 1998). CD19 and B220 down-regulation on CD79B^Y196H MYD88^L265P B cells paralleled the phenotype of Tlr9^−/− MYD88^L265P B cells that undergo plasmablast differentiation after transplantation to Rag1^−/− recipients (Wang et al., 2016) and B cells transduced with gain-of-function CARD11 mutation (Jeelall et al., 2012). After correcting for the number of EGFP⁺ cells transplanted on day 0 (Fig. 5 A), MYD88^L265P B cells were present in 5-fold higher number than empty vector B cells on day 11, and CD79B^WT MYD88^L265P or CD79B^Y196H MYD88^L265P B cells accumulated to 8- and 15-fold higher numbers, respectively (Fig. 5 C).

Plasma cell differentiation was further analyzed by transducing polyclonal C57BL/6 B cells, transplanting them into Rag1^−/− recipients, and analyzing the recipients 11 d later (Fig. 5, F–I). CD19 was significantly decreased on IgM⁺ EGFP⁺ cells coexpressing MYD88^L265P with CD79B or CD79B^Y196H compared with control cells expressing only one or other protein (Fig. 5 F). In contrast, surface IgM was increased on the cells coexpressing MYD88^L265P with CD79B or CD79B^Y196H compared with cells expressing MYD88^L265P alone, the latter having lower surface IgM than empty vector–expressing B cells (Fig. 5 G). Down-regulation of CD19 was accompanied by induction of CD138 and by secretion of large amounts of serum IgM in recipients of cells coexpressing MYD88^L265P with CD79B or CD79B^Y196H (Fig. 5, H and I; and Fig. S2). Interestingly, there was no detectable secretion of IgG, indicating that the cooperation between MYD88^L265P and CD79B drives plasma cell differentiation but not isotype switching to IgG. These results are similar to the down-regulation of CD19 and B220, induction of CD138, and secretion of IgM but not IgG by C57BL/6 B cells expressing MYD88^L265P combined with TLR9-deficiency (Wang et al., 2016).

Gain-of-function CARD11 mutations derived from human lymphomas disrupt several B cell tolerance checkpoints when introduced into activated mouse B cells, switching the effect of self-antigen from inducing apoptosis into induction of proliferation and plasmablast differentiation and autoantibody secretion (Jeelall et al., 2012). Using the same experimental system, MYD88^L265P did not protect B cells from self-antigen–induced deletion unless combined with enforced BCL2 expression, and the MYD88^L265P BCL2 combination remained insufficient to break tolerance checkpoints that oppose plasmablast differentiation and autoantibody secretion (Wang et al., 2014). Hence we used the identical experimental model to test if cooperation between CD79B and MYD88 mutations would override peripheral tolerance checkpoints in self-antigen–binding B cells transferred to antigen-expressing recipient mice. Mature MD4 Ig^HEL-transgenic B cells bearing HEL-specific BCRs were activated, transduced with the CD79B and MYD88 single or combined retroviral vectors, and transplanted into ML5-transgenic Rag1^−/− recipient mice that expressed HEL protein in their circulation as self-antigen (Goodnow et al., 1988).

The number of surface HEL-binding IgM^a+EGFP⁺ B cells was measured at the time of transplantation (input) and in the spleen of the recipients 11 d after transplantation (Fig. 6, A and C). Retrovirally transduced EGFP⁺ B cells expressing no additional protein or only MYD88^L265P were eliminated by self-tolerance, as were EGFP⁺ B cells expressing only CD79B^WT or CD79B^Y196H (Fig. 6, B and C), indicating that the CD79B ITAM mutation on its own did not disrupt this tolerance checkpoint. In contrast, B cells coexpressing CD79B^WT or CD79B^Y196H with MYD88^L265P were protected from self-antigen-induced deletion (Fig. 6, B and C). On average, 10 to 30 times more EGFP⁺ cells accumulated when the transplanted B cells were CD79B^WT MYD88^L265P or CD79B^Y196H MYD88^L265P, respectively, compared with recipient mice that received B cells expressing only one or other mutant protein (Fig. 6 C). Inhibiting the peripheral deletion checkpoint therefore required cooperation between dysregulated CD79B and MYD88^L265P. Moreover, self-reactive B cells coexpressing CD79B^WT MYD88^L265P or CD79B^Y196H MYD88^L265P differentiated into CD19^low plasmablasts (Fig. 6 B). Analysis of the serum from the recipients by ELISA showed secreted anti-HEL autoantibodies only in recipients of B cells with dysregulated CD79B and mutant MYD88 (Fig. 6 D), indicating that cooperation between the mutations caused a potent disruption to B cell tolerance checkpoints.

Transcription of CD79B^WT from the retroviral promoter increased CD79B and IgM on the surface of transduced B cells to a similar amount compared with the CD79B ITAM mutation (Fig. 2, A and B) and increased the pool of pre-Golgi Endo H–sensitive CD79B (Fig. 3 C). We therefore sought to diminish translation of the retroviral CD79B transcript with the aim of distinguishing the effect of CD79B^Y196H from that of CD79B^WT. CD79B translation was diminished by introducing one of six different upstream open reading frames into the 5′ untranslated region (5′ UTR) of the CD79B coding sequence (Fig. S1; Calvo et al., 2009). When cell surface CD79B was measured on the transduced B cells with the same level of EGFP expression, the 5′ UTR variants decreased the CD79B mean fluorescence intensity compared with B cells expressing the CD79B^WT with the typical Kozak sequence (Fig. 7 A). Two 5′ UTR sequences were also tested for their effects on surface IgM expression. Whereas surface IgM and CD79B were twice as high on EGFP⁺ cells expressing CD79B^WT cDNA compared with cells expressing the control EGFP vector alone, both were decreased to levels comparable with the control cells when the CD79B^WT cDNA carried the UTR4 insertion and were increased by only 35% when the UTR6 insertion was present (Fig. 7 B).

To test the effects of these translationally dampened CD79B vectors on peripheral B cell tolerance checkpoints, we transduced HEL-specific B cells with the CD79B^WT MYD88^L265P or CD79B^Y196H MYD88^L265P vectors preceded by UTR4 or UTR6 and transplanted the B cells into HEL-transgenic mice as before (Fig. 7 C). After 5 d, HEL-specific EGFP⁺ B cells expressing empty vector, MYD88^L265P alone, CD79B^WT alone, or CD79B^Y196H alone did not persist in the recipient mice expressing HEL antigen, but B cells expressing the CD79B^WT MYD88^L265P or CD79B^Y196H MYD88^L265P vectors resisted deletion to varying extents (Fig. 7 D). On average, twice the number of EGFP⁺ B cells persisted when they expressed the higher expression UTR6 vector compared with the UTR4 vector (Fig. 7 D). Furthermore, B cells expressing the UTR6 CD79B^Y196H MYD88^L265P vector preferentially accumulated on average to twice the number of B cells expressing UTR6 CD79B^WT MYD88^L265P (Fig. 7 D). Although the number of self-reactive EGFP⁺ B cells resisting deletion was variable, it was strongly correlated with their surface CD79B and IgM expression in the recipient mice (Fig. 7 E), indicating that escape from self-tolerance by MYD88^L265P B cells required increased BCR expression through cooperation with alterations in CD79B.

CD79B and MYD88 mutations frequently occur together in mature B cell malignancies. Here we reveal that they cooperate synergistically to induce B cell differentiation into plasmablasts and break self-tolerance checkpoints in B cells within peripheral lymphoid tissue. What might be the basis for the striking cooperation between CD79B and MYD88 mutations in our experiments?

Several lines of evidence here support a conclusion that surface BCR density is governed by CD79B and represents a critical, rate-limiting factor for B cells expressing MYD88^L265P. First, CD79B appears to be the rate-limiting component for surface BCR expression in normal B cells, because retrovirally increasing the expression of CD79B increased the mean surface IgM to 150% of controls, and this could be further increased if the retrovirally expressed CD79B carried tyrosine substitutions in the membrane proximal ITAM found recurrently in human lymphomas (Davis et al., 2010). Second, B cells expressing MYD88^L265P down-regulate their surface CD79B and IgM despite having no decrease in mRNA for all four essential components of surface IgM BCRs. Low surface IgM expression in MYD88^L265P B cells is instead explained by accumulation of immature, pre-Golgi, Endo H–sensitive IgM, representing a “checkpoint” remarkably similar to the posttranslational block to surface IgM expression induced by chronic binding to self-HEL antigen (Bell and Goodnow, 1994). Hence it is conceivable that the expression of MYD88^L265P and chronic exposure to self-antigen act additively or cooperatively to block surface BCR trafficking and accumulation on the cell surface, which may deprive the cells of tonic BCR survival signals to counter B cell apoptosis in vivo. The third line of evidence that CD79B represents a critical limiting factor in B cells expressing MYD88^L265P comes from the experiment introducing 5′ UTR open reading frames that compete with the normal translation initiation of retrovirally encoded CD79B protein. The decrease in CD79B and IgM surface display was correlated with a decrease in the capacity of CD79B to block deletion of HEL-specific B cells when combined with MYD88^L265P. In this case, the rescue of mature B cells from self-antigen induced deletion was enhanced by Y196H mutation and correlated with increased surface CD79B and IgM.

Previous studies have shown that the deletion of HEL-specific B cells in HEL-transgenic mice is preceded by down-regulation of surface IgM and is mediated by apoptosis that is inhibited by Bcl2 overexpression (Cyster et al., 1994) and requires the BCL2 antagonist BIM (Enders et al., 2003). When the capacity to synthesize surface IgM is experimentally ablated by conditional gene deletion, mature B cells undergo apoptosis in FOXO1-dependent and BIM-dependent manner that is also inhibited by enforced expression of BCL2 (Lam et al., 1997; Srinivasan et al., 2009). B cells that lose surface IgM have decreased phosphorylation of AKT, an intermediate in phosphoinositide 3-kinase (PI3K) inhibition of FOXO1, and can be protected against apoptosis by mutations that constitutively activate PI3K (Srinivasan et al., 2009). Conversely, exaggerated surface IgM accumulation and increased AKT activation occurs in B cells with substitutions in both ITAM tyrosines in CD79B (Gazumyan et al., 2006). These results are consistent with the view that surface IgM paired with CD79A and CD79B, in the absence of antigen, continuously signals through the ITAMs in CD79A to activate PI3K at a tonic level and thereby activate AKT, inhibit FOXO1 accumulation, and repress BIM formation in B cells to promote their survival. Thus, the BIM-dependent deletion of HEL-specific B cells that follows self-antigen-induced BCR down-regulation may therefore result, at least in part, from the loss of surface BCR CD79A tonic signaling to PI3K. Because MYD88^L265P decreases surface BCR even in the absence of self-antigen, any prosurvival signal induced by MYD88^L265P, for example through the activation of NF-κB, may be overwhelmed by the loss of surface BCR and its PI3K survival signal. Our previous work showing the striking cooperation between Bcl2 overexpression and MYD88^L265P for enhancing B cell survival both in the absence or presence of self-antigen supports the idea that the capacity of MYD88^L265P to promote B cell survival and proliferation is normally overwhelmed by BIM-mediated apoptosis (Wang et al., 2014).

Notably, overexpression of neither CD79B^WT nor CD79B^Y196H alone was sufficient to block self-antigen-induced deletion. One explanation is that preserving tonic BCR signaling to PI3K alone is insufficient to counter other inducers of apoptosis, including the elevation of BIM and Nur77 (Healy et al., 1997; Glynne et al., 2000; Zikherman et al., 2012). When MYD88^L265P is then coexpressed with CD79B vectors in self-reactive B cells, MYD88^L265P-IRAK signaling to activate NF-κB may counter additional proapoptotic mediators by inducing prosurvival proteins such as BCL-XL and A1, which are not normally induced by self-antigen activation of the BCR (Glynne et al., 2000).

In addition to cooperating to block self-antigen–induced B cell deletion, the combination between CD79B and MYD88 mutations also overrides the B cell tolerance checkpoint that normally inhibits TLR9- or TLR4-driven self-reactive B cell differentiation into CD19^low plasmablasts (Rui et al., 2003, 2006). This checkpoint is induced by chronic self-antigen engagement and BCR signaling to ERK (Rui et al., 2003, 2006). ERK phosphorylates and promotes the activity of an essential inhibitor of plasmablast differentiation, ETS1 (Bories et al., 1995; Hollenhorst et al., 2011). The promotion of plasmablast differentiation by CD79B was not simply a result of enhanced survival of self-reactive B cells, because overexpressing BCL2 enhances the accumulation of MYD88^L265P-expressing B cells but not their differentiation into CD19^low plasmablasts (Wang et al., 2014).

The microarray data identified increased expression of plasma cell up-regulated genes in MYD88^L265P-expressing B cells compared with CARD11^L232LI-expressing B cells, raising the possibility that MYD88^L265P alone initiates an abortive plasma cell differentiation program that requires an additional signal from the BCR to be completed. Weaker gain-of-function alleles of CARD11 do not spontaneously drive B cell differentiation into plasmablasts unless the cells receive an additional BCR stimulus by binding self-antigen (Jeelall et al., 2012). Elevated surface BCR caused by CD79B mutations or overexpression could provide an additional signal through PI3K activation to drive plasmablast differentiation by down-regulating ETS1. Several independent observations support this hypothesis: (1) the activation of PI3K p100 delta subunit regulates autoantibody formation by marginal zone and B1 B cells (Durand et al., 2009), (2) constitutively active PI3K ensures mature B cell survival even in the absence of IgM (Srinivasan et al., 2009), and (3) PI3K activation is involved in reducing Ets1 levels during plasma cell differentiation (Luo et al., 2014). Similar to enhanced BCR signaling by CD79B mutations, B cell differentiation into plasmablasts in mice is exaggerated by other mutations that dysregulate BCR signaling, including loss-of-function mutations in Lyn, Ptpn6 (SHP1), and Cd22, resulting in lupus-like symptoms such as pathogenic autoantibody production (Chan et al., 1997; Gross et al., 2009; Luo et al., 2014). These observations are likely due to the loss of Ets1 expression, because Ets1 is controlled by BCR-mediated signals through PI3K, BTK, IKK, and JNK (Luo et al., 2014). Additionally, Ets1 deficiency also leads to B cell hyper-responsiveness to TLR9 stimulation and TLR9-dependent autoimmunity in mice (Wang et al., 2005). Therefore, in future studies, it will be interesting to test if the observed phenotype in CD79B and MYD88 double-mutant B cells results from PI3K activation to down-regulate ETS1.

CD79B mutations are seen in only 30% of MYD88^L265P ABC-DLBCL and 15% of WM (Ngo et al., 2011; Poulain et al., 2013), suggesting that other mechanisms may exist to antagonize the MYD88^L265P effect on CD79B and IgM trafficking through the Golgi to the cell surface. Indeed, Lyn copy number loss is found in 60% of patients with WM (Hunter et al., 2014). Lymphoma B cells with the CD79B ITAM mutation show reduced Lyn activity, suggesting that the effect of CD79B ITAM mutation could also be achieved by Lyn deficiency (Davis et al., 2010). In mice, Lyn deficiency results in exaggerated B lymphoblast expansion, autoantibody production, and elevated sensitivity to BCR stimulation (Chan et al., 1997; Cornall et al., 1998). Additionally, Lyn deficiency in mice causes hypersensitivity to MYD88 signaling, and this MYD88-dependent signaling is crucial for autoimmunity (Lamagna et al., 2013, 2014). However, one might question why ABC-DLBCLs with CD79B and MYD88 mutations resist differentiation and cell cycle arrest. One clue comes from clinical observations that 23% of human DLBCL cases have chromosome 11q24.3 gain to elevate Ets1 and Fli1 expression to antagonize B cell differentiation (Bonetti et al., 2013). Additional mechanisms of differentiation block in DLBCL come from the frequent inactivation of PRDM1 in up to 24% of ABC-DLBCLs by chromosome 6q21-q22.1 deletion (Pasqualucci et al., 2006). Furthermore, 77% of these tumors lack the BLIMP1 protein despite the presence of PRDM1 mRNA (Pasqualucci et al., 2006). Therefore, different disease outcomes could result from additional mutations among the average 20 or more protein-altering mutations B cells acquire, which includes CD79B and MYD88 mutations.

Splenic B cells were collected from C57BL6 mice or MD4 mice bearing rearranged transgenic immunoglobulin gene encoding for HEL-specific antibodies (Ig^HEL transgenic; Goodnow et al., 1988). All mice were generated on a C57BL6 background or backcrossed to that background for greater than 10 generations and were housed in specific pathogen–free environment at the Australian Phenomics Facility, Australian National University. Donor mice were aged 8–16 wk at the time of the experiments. Rag1^−/− and Rag1^−/− ML5 transgenic mice that express soluble HEL (Goodnow et al., 1988) were used as recipients at ages 8–14 wk. All animals used in this study were cared for and used in accordance with protocols approved by the Australian National University Animal Experimentation Ethics Committee and the current guidelines from the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes.

Mouse Cd79b was amplified by Platinum Pfx DNA polymerase (Invitrogen) from mouse splenic cDNA and subcloned into pcDNA3.1(+) vector. PCR-based site-directed mutagenesis was used to introduce Cd79b mutations. Sequence-confirmed Myd88 and Cd79b genes were then transferred into pMXs-IRES-GFP vectors (Kitamura et al., 2003) either individually or as dual-expression vectors (Fig. 1 A). To modulate protein translation, short open reading frames were inserted into the 5′ UTRs of Cd79b coding sequence (Fig. S1). Retroviral vectors containing Myd88 or Cd79b or dual Myd88 and Cd79b were transfected into Phoenix ecotropic packaging cells (ATCC) using calcium phosphate precipitation (Swift et al., 2001). Supernatants containing retroviral particles were collected 48 and 72 h after transfection and frozen at −80°C until use.

MD4 transgenic mice were given 5 mg HEL to provide a pulse of antigen to activate HEL-specific B cells 6 h before removing the spleen and preparing a single-cell suspension and washing twice with fresh complete RPMI by centrifugation at 300 g for 5 min at 8°C. Splenocytes were cultured at 4 × 10⁶ cells/ml in complete RPMI containing 10 µg/ml anti-CD40 antibody (FGK4.5; Bio X Cell) for 24 h. For splenocytes from nontransgenic C57BL/6 mice, 7 µg/ml anti-IgM (Jackson ImmunoResearch Laboratories, Inc.) was also included during the initial 24-h activation culture. After 24 h, the cells were washed twice in RPMI and spin-infected in 6-well plates at 920 g for 90 min at room temperature with retrovirus supernatant containing 10 µl/ml DOTAP (Roche). The cells were then placed back in fresh RPMI containing 10 µg/ml anti-CD40 but no BCR ligands for HEL-specific B cells, or anti-CD40 plus 10 µg/ml anti-IgM for nontransgenic B cells for 36 h. The transduced cells were then washed three times with fresh complete RPMI by centrifugation at 300 g for 5 min at 8°C, and any remaining media containing trace amount of anti-CD40 and/or anti-IgM antibodies was carefully removed. The number of live cells was determined by hemocytometer counting of Trypan blue–negative cells, and the percentage of EGFP⁺ 7AAD⁻ cells was determined by flow cytometry. Washed, transduced cells were then cultured in 24-well plates at 10⁶ cells/ml in fresh complete RPMI without anti-CD40 or BCR ligand, with the start of the mitogen-free cultures designated “day 0.”

Retrovirally transduced B cells were recovered 36 h after transduction and washed, and live cells were enumerated by Trypan blue exclusion. Live cells (2–10 × 10⁶), consisting of a mixture of transduced and nontransduced B cells, were transferred into each recipient mouse through lateral tail vein injection. Recipient mice were sacrificed after 5–11 d and their spleens analyzed for the amount of EGFP⁺ HEL-binding mature B cells. For cell culture experiments, cells were washed and cultured at 10⁶ cells/ml fresh RPMI media supplemented with 10% vol/vol FCS and 2% vol/vol penicillin, streptomycin, and glutamine without B cell mitogens in 24-well plates. The start of the cultures was designated “day 0.” The number of EGFP⁺ cells on the subsequent days in culture was then determined by hemocytometer counting of Trypan blue–negative cells and flow cytometric analysis of the percentage of EGFP⁺ 7AAD⁻ cells and expressed as relative amount compared with day 0.

Single-cell suspensions were prepared and transferred to 96-well round-bottom plates. Cells were then incubated for 30 min at 4°C with antibody cocktails containing appropriate combination and dilution of antibodies in flow cytometry buffer (PBS with 2% vol/vol bovine serum and 0.1% wt/vol sodium azide). Cells were then washed twice and resuspended in flow cytometric buffer containing 7AAD for analysis on an LSR II or LSR Fortessa flow cytometer (BD). The following antibodies were used: anti-B220, CD19, IgM^a, CD79B, Qdot 605–streptavidin, and 7AAD. HyHEL9 antibody conjugated to Alex Fluor 647 was performed in-house with a monoclonal antibody labeling kit (Molecular Probes). Cells were first stained with 1 µg/ml HEL protein before staining with the HyHEL9 antibody. FlowJo (Tree Star) software was used to analyze the flow cytometry data.

Transduced B cells were cultured without anti-CD40 for 24 h and EGFP⁺ cells sorted using FACSAria I and II (BD) cell sorter and washed twice with cold PBS. For Western blot analysis, cytoplasmic proteins were extracted from 10⁶ cells using the NE-PER protein extraction kit (78833; Thermo Fisher Scientific) per the manufacturer’s protocol. For Endo H treatment, 5 × 10⁵ cell equivalents in 9-µl protein lysates were mixed with 1 µl of 10× glycoprotein denaturing buffer, heated at 100°C for 10 min, and incubated at 37°C with the addition of 2 µl 10× G5 reaction buffer and 1 µl Endo H in a total volume of 20 µl for 1 h (P0702; New England Biolabs, Inc.). At the end of the reaction, 10 µl 3× SDS sample buffer (0.2 M Tris-HCl, pH 6.8, 10% wt/vol SDS, 30% vol/vol glycerol, 10% vol/vol β-mercaptoethanol, and bromophenol blue) was added to stop the reaction, and samples were stored at −20°C until use.

Lysates were heated briefly at 100°C, microcentrifuged for 1 min at 13,000 rpm, and resolved on a 15% polyacrylamide gel. Resolved proteins were then transferred onto a polyvinylidene fluoride membrane (Bio-Rad Laboratories). Nonspecific antibody binding to the membrane was blocked with 5% wt/vol skim milk in TBST (10 mM Tris-HCl, pH 7.5, 150 mM NaCl, and 0.1% vol/vol Tween 20) for 30 min at room temperature and then probed overnight at 4°C with antibodies to IgM and CD79B. After primary antibody incubation, the membrane was washed three times with TBST and incubated with HRP-conjugated anti-rabbit IgG secondary antibody for 60 min at room temperature. Membranes were then washed at least five times with TBST before detection using enhanced chemiluminescence detection reagent (PerkinElmer) and developed in a darkroom onto Kodak films. Membranes were reprobed with anti-α/β-tubulin antibodies as loading control.

For microarray analysis, sorted cells were resuspended in TRIzol. Phase separation was performed by the addition of chloroform and centrifugation at 12,000 g for 15 min, followed by isopropanol RNA precipitation. The air-dried RNA pellet was dissolved in RNase-free water, and mRNA expression was analyzed on Affymetrix mouse ST 1.0 arrays as per the manufacturer’s instructions. Microarray data were subjected to quality control, normalization, and differential expression analysis in limma (Ritchie et al., 2015). Quality control reviewed reconstructed array images, distribution of probe intensities, M(log ratio) versus A(mean average) plots, fitting of a probe level model in the oligo package (Carvalho and Irizarry, 2010) to generate box plots of relative log expression and normalized unscaled standard errors. One of three independent samples of sorted cells in each experimental group was excluded on the basis of insufficient quality. The final set of eight samples was normalized using PLM (Carvalho and Irizarry, 2010) and resulting probe weights were used to derive weights for transcript expression estimates on each array suitable for use with limma in a linear model, comparing gene expression in MYD88^L265P cells with each of the other groups. CAMERA (Wu and Smyth, 2012) was used to query for enrichment of differentially expressed genes against predefined sets of hallmark, curated, and immunological signature gene sets defined in MSigDB (Subramanian et al., 2005). Microarray data that support the findings of this study have been deposited in Gene Expression Omnibus (accession no. GSE99775).

For quantitative RT-PCR, total RNA was extracted from sorted EGFP⁺ cells with TRIzol (Thermo Fisher Scientific). For each sample, an equal amount of total RNA was reverse-transcribed to cDNA with SuperScript II reverse transcription (Invitrogen) for 2 h at 42°C. Dilution (1:100) of the cDNA stock was used to set up TaqMan quantitative RT-PCR on an ABI7500 instrument (Applied Biosystems) using the following mouse TaqMan probes (Thermo Fisher Scientific): Myd88 (Mm00440338_m1), Cd79a (Mm00432423_m1), Cd79b, (Mm00434143_m1), and Ighm (Mm01718955_g1). Data were normalized to β-actin expression: ACTB-F, 5′-GCCTTCCTTCTTGGGTATGGAA-3′, ACTB-R, 5′-GGCGGACTGTTACTGAGCTG-3′.

Data were generated and analyzed using Prism version 5.0 (GraphPad Software). Statistically significant p-values of <0.05, <0.01, and <0.001, determined by two-tailed unpaired Student’s t test or ANOVA, are indicated where applicable.

Fig. S1 shows open reading frames inserted into Cd79b 5′ UTRs. Fig. S2 shows CD138 expression on EGFP⁺ cells transduced with the indicated vector and analyzed 11 d after adoptive transfer into Rag1^−/− recipient mice. Table S1 shows gene set enrichment analysis of MYD88^L265P- versus CARD11^L232LI-expressing B cells. Table S2 shows hallmark gene sets enriched in MYD88^L265P- versus CARD11^L232LI-expressing B cells. Table S3 shows curated gene sets enriched in MYD88^L265P- versus CARD11^L232LI-expressing B cells. Table S4 shows a complete microarray dataset comparing B cells expressing MYD88^L265P with cells expressing empty vector, CARD11^L232LI, or IKBKB^K171E. Tables S1–S4 are provided as Excel files.

We thank the Australian Phenomics Facility for expert care and genotyping of animals, the Australian Cancer Research Foundation Biomolecular Resource Facility for sequencing and microarray, and the John Curtin School of Medical Research image and cytometry facility for cell sorting.

J.Q. Wang was supported by an Australian Postgraduate Award and an Australian National University Postgraduate Scholarship. H.M. Yoo was supported by a Korean Visiting Scientist Training Award Fellowship from the Korea Health Industry Development Institute (HI15C1075). C.C. Goodnow was supported by National Health and Medical Research Council (NHMRC) Australia Fellowship 585490 and NHMRC Senior Principal Research Fellowship 1081858. The research was supported by NHMRC grants to C.C. Goodnow (1016953 and 1113904) and to K. Horikawa (1086770).

The authors declare no competing financial interests.

Author contributions: J.Q. Wang designed and performed experiments, analyzed data, and drafted the manuscript. Y.S. Jeelall, E.L. Batchelor, and S.M. Kaya assisted with experiments. P. Humburg analyzed the microarray data. H.M. Yoo performed and analyzed the quantitative PCR experiments. K. Horikawa conceived the study and constructed the vectors. C.C. Goodnow and K. Horikawa initiated, designed, supervised, and interpreted the experiments and revised the manuscript.