BCL6 breaks occur at different AID sequence motifs in Ig–BCL6 and non-Ig–BCL6 rearrangements

The CCND1 and BCL2 translocations in many human lymphomas appear to arise in B-cell progenitors in the bone marrow and to depend on the recombination activating gene (RAG) endonuclease complex for chromosome breaks at the J_H and D_H segments in the immunoglobulin heavy chain (IgH) locus. We previously showed that the IgH partner breakpoints in these translocations target the dinucleotide sequence CpG,^1,2  and we proposed a chromosome breakage model in which activation-induced cytosine deaminase (AID) creates T:G mismatches at methylated CpG sites. More recently, we showed that some mantle cell lymphomas have CCND1 breakpoints that occur near the motif WGCW (W = A or T) and that the MYC breakpoints in Burkitt lymphoma, which occur in germinal center B-cells, strongly prefer the same WGCW motif.³  We propose a model in which CpGs are targeted in translocations occurring in pro-B-cells, when AID levels are low, whereas WGCW is preferred when AID levels are higher, as in germinal center B-cells, where class-switch recombination (CSR), somatic hypermutation (SHM), and MYC translocations occur. We also found that most CpG breaks in pro-B translocations occurred at CGC, a known AID hotspot.³ 

BCL6 is the most commonly rearranged gene in diffuse large B-cell lymphoma, but mechanistic insight into the basis of the chromosomal breaks in BCL6 translocations has been scant (see supplemental Figure 1, available with other supplemental data on the Blood website). Here, we analyze a large set of BCL6 chromosomal translocations to the IgH switch (S_H) regions, Ig light chain (IgL) loci, and non-Ig partners, and we show that these breaks target CpG and CGC motifs, in addition to the canonical AID motifs WGCW, WGC, and WRC (R = A or G). Interestingly, the BCL6 breaks that partner with Ig loci occur near the WGCW, WGC, and WRC motifs, but not at CpG, whereas those fused to non-Ig partners occur near both CpG/CGC and WGCW/WGC/WRC motifs. Our findings raise the possibility that BCL6 breaks partner with Ig and non-Ig loci using somewhat different mechanisms.

A total of 137 unpublished and published junctional sequences were used in this study.⁴  Statistical methods are described in Tsai et al²  and supplemental Data.

Samples were collected from Kyoto University Hospital and affiliated hospitals. Approval was obtained from the Kyoto University Institutional Review Board for the study. Informed consent was provided according to the Declaration of Helsinki.

We gathered the breakpoint sequences of 137 known and previously unpublished BCL6 chromosomal translocations. We found that 120 (87.6%) of the breakpoints cluster within a 2156-bp region at the beginning of BCL6 intron 1 (Figure 1). We then analyzed these clustered breakpoints for their proximity to numerous sequence motifs.²  After separating the BCL6 breakpoints into those that partnered with Ig and non-Ig partner loci, 2 patterns emerged.

First, we noted that the 74 BCL6 breakpoints in Ig–BCL6 translocations show a highly significant proximity to the sequence motif WGCW (P = .004, binomial test; 28% of breaks within 4 nt of WGCW), which is highly enriched in S_H regions and targeted for deamination by AID during CSR (supplemental Table 1 and supplemental Figure 4).⁵  We also found significant proximity to the related sequence motifs WRC and WGC, which are targeted for deamination by AID during SHM in germinal center B cells (supplemental Table 1 and supplemental Figure 3). These proximities are significant by the binomial test, the Student t test, and the Mann-Whitney U-test (P < .005). In contrast, no significant proximity was identified to motifs such as GGC, GTC, GAC, and GCC, which are not efficiently targeted by AID in biochemical assays,^6,7  or to the CpG and CGC motifs used in pre-B/pro-B-type translocations (supplemental Tables 1-4 and supplemental Figures 5, 6, 9, and 10).

We recently reported that MYC breakpoints in human Burkitt lymphomas and murine IgH–MYC translocations also occur at WGCW motifs.³  The use of WGCW and other canonical AID motifs at the BCL6, MYC, and Ig loci in IgH–MYC and Ig–BCL6 translocations suggests a common germinal center origin for these rearrangements and strongly implicates AID in their pathogenesis.

Second, in marked contrast to the Ig–BCL6 translocations, the 46 BCL6 breakpoints that partner with non-Ig loci in non-Ig–BCL6 translocations show a highly significant proximity to CpG (P < .003; 15 [33%] of 46 breaks directly at CpG) and to the related motif CGC (P < .01; 12 [26%] of 46 breaks at CGC) (Figure 2A, supplemental Figure 7, and supplemental Table 2). This finding was unexpected, as we previously identified CpG and CGC breakpoints only in IgH translocations occurring at the pro-B/pre-B stage.^1,2,8,9  However, sequence analysis of the BCL6 breakpoint junctions strongly suggests that these translocations are not occurring in pro-B/pre-B cells (see supplemental Results and Discussion). In particular, we found no significant evidence of RAG or terminal deoxynucleotidyl transferase activity, which are characteristic of normal V(D)J junctions and pro-B/pre-B translocations such as IgH–BCL2 and IgH–CCND1 (supplemental Table 6). Moreover, 25 (44%) of the 57 breakpoints show evidence of junctional microhomology, a feature that is characteristic of non-homologous DNA end-joining repair but is infrequent in normal V(D)J rearrangements and pre-B/pro-B translocations.

The BCL6 breaks in these non-Ig translocations also show proximity to WGCW (P < .05 for binomial, Student t, and Mann-Whitney U tests), WRC (P < .04, binomial and Mann-Whitney U), and WGC (P = .03, Mann-Whitney U) motifs, suggesting that these breaks also are initiated by AID. The statistical significance of these motifs appears to be weaker for non-Ig–BCL6 than Ig–BCL6 translocations (supplemental Tables 1 and 2), but this difference might reflect the fact that fewer non-Ig breaks were available for analysis and that a third of the non-Ig breaks occur at CpG/CGC.

Our findings strongly implicate AID in the pathogenesis of BCL6 translocations. Previous studies have shown that AID-dependent SHM causes mutations at BCL6.^10-13  Indeed, the distribution of BCL6 mutations is virtually identical to that of BCL6 translocation breakpoints (supplemental Figure 13). Aberrant SHM also causes mutations, to a lesser extent, at other genes actively transcribed in germinal center B-cells, including the BCL6 translocation partners PIM1, RHOH, POU2AF1, and CTIIA.^14-16  Other studies have shown that SHM causes not only mutations but also double-strand breaks.^17,18  In light of these studies, our results strongly suggest that BCL6 translocations are initiated by AID in germinal center B-cells (Figure 2B).

The basis for the highly focused clustering of BCL6 breakpoints and SHM mutations in the same 2156-bp region near the beginning of BCL6 intron 1 is unclear (supplemental Figure 13). This is likely a consequence of several factors that are yet to be defined, such as chromatin accessibility to AID. In this regard, it is important to note that the upstream portion of BCL6 intron 1 has a favorable H3K4me3 histone pattern for AID accessibility (Figure 1 and supplemental Figures 11 and 12).^16,19,20 

It also is unclear why BCL6 breakpoints occur at CpG/CGC motifs in non-Ig–BCL6 rearrangements (but not in Ig–BCL6 translocations), particularly because both rearrangement types appear to be initiated by AID in germinal center B-cells. One possibility is that differences in the kinetics of chromosomal breakage at Ig and non-Ig loci might somehow influence which sequence motif is used at BCL6. For example, AID is much less active at non-Ig than at Ig loci,^14,15  so the generation of non-Ig breaks is more likely to be rate-limiting than Ig breaks, which are efficiently introduced by CSR and SHM. In principle, the limited availability of non-Ig breaks should favor the selection of BCL6 breaks at methylated CpG/CGC motifs, where AID causes long-lived T:G mismatches (in comparison with short-lived U:G mismatches at nonmethylated WGCW/WRC/WGC sites).²  Alternatively, CSR-associated breaks at IgH (but not non-Ig breaks) might preferentially induce breaks at BCL6 by recruiting and activating the Artemis DNA nuclease.²¹  A third possibility is that the level of AID, which varies widely in different germinal center B-cell subsets,²²  determines the preferred motif.

In summary, we identify the first breakpoint sequence motifs in BCL6 translocations, and we show that these motifs are differentially used in Ig–BCL6 and non-Ig–BCL6 rearrangements. Our findings support previous analyses demonstrating potential biological and clinical differences between diffuse large B-cell lymphoma cases having Ig–BCL6 and non-Ig–BCL6 rearrangements.^23,24 

The work was supported in part by the National Institutes of Health (M.R.L.).

Contribution: Z.L., A.G.T., H.A.G., and M.R.L. designed the research; T.A. and H.O. cloned and sequenced 66 of the 137 breakpoints in this study (supplemental Table 8); Z.L., A.G.T., H.A.G., and M.R.L. analyzed and interpreted the data; Z.L. and A.G.T. performed statistical analyses; Y.J. and A.M.M. did the genome-wide analysis of histone marks; and Z.L., H.A.G., and M.R.L. wrote the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

The current affiliation for A.G.T. is Department of Pathology, Stanford University, Stanford, CA. The current affiliation for T.A. and H.O. is Department of Hematology, Tenri Hospital, Tenri, Nara, Japan.

Correspondence: Harvey A. Greisman, Department of Laboratory Medicine, University of Washington Medical Center, Box 357110, Seattle, WA 98195; e-mail: greisman@uw.edu.