Combined deletion of Xrcc4 and Trp53 in mouse germinal center B cells leads to novel B cell lymphomas with clonal heterogeneity

We established a new mouse model in which Xrcc4 (X) and Trp53 (P) genes were deleted at a later stage of mature B cell development using Cγ1Cre [49]. In Cγ1Cre knock-in mice, Cre expression is driven by the endogenous Iγ1 promoter and occurs in the majority of GC B cells [49]. Consistently, we found that GC B cells have the complete deletion of Xrcc4 (X) floxed allele in Cγ1Cre/X ^c/+ mice (c: floxed allele; +: wt allele) (Additional file 1: Figure S1A). Then, we examined whether deletion of Xrcc4 or Trp53 via Cγ1cre led to a high level of genomic instability in primary B cells. To do so, we assayed for Igh locus-specific genomic instability in B cells that can be induced via anti-CD40/IL4 activation in the in vitro culture of primary B cells [15, 37]. Using metaphase fluorescence in situ hybridization (FISH), we found that anti-CD40/IL4-stimulated Xrcc4/Trp53-double deficient (DKO) B cells harbored a high level of Igh locus instabilities, including breaks and translocations (Fig. 1a, b). Furthermore, our data showed that Xrcc4 deficiency alone resulted in a high level of Igh locus genomic instability; in contrast, Trp53 deficiency did not enhance the level of genomic instability in activated B cells compared to wt controls (Fig. 1a). Thus, the Igh locus genomic instability in the DKO B cells is largely attributable to Xrcc4 deficiency.

We conducted cohort studies by monitoring two cohorts of Cγ1CreX ^c P ^c mice for the occurrence of B cell lymphomas (X ^c : X ^c/c or X ^c/− ; P ^c : P ^c/c or P ^c/− ). Within the first cohort, we found that Cγ1CreX ^c P ^c/+ mice (Additional file 1: Figure S1B, n = 19, blue line) survived normally and were not cancer-prone. This phenotype is similar to that of CD21CreX ^c/- mice [47] in that both lines only have Xrcc4, but not Trp53, deleted. Cγ1CreX ^c/+ P ^c mice (Additional file 1: Figure S1B, n = 20, green line), which still express XRCC4, developed thymic lymphomas or solid tumors. This phenotype is likely due to p53 haploinsufficiency as the genotypes of all the tumors are P ^c/− . In contrast, 11 out of 47 Cγ1CreX ^c P ^c mice succumbed to B lineage lymphoma (Additional file 1: Figure S1B, n = 47, red line) as all 11 tumors expressed B lineage markers including B220 and CD19, while the rest of morbid mice died of thymic lymphomas or solid tumors. The floxed Xrcc4 allele was deleted in all of the B lineage lymphomas but not in thymic lymphomas and solid tumors (data not shown). Within the second cohort, we observed a survival curve closely resembling the first one, with a similar penetrance and latency for tumor development (Fig. 2a), albeit the second cohort included only wt controls (n = 37) and Cγ1CreX ^c P ^c mice (n = 51). We termed the Cγ1CreX ^c P ^c -derived B lineage tumors G1XP lymphomas.

We performed Southern blotting to assay G1XP tumor DNA for Igh rearrangements and found that 7 out of 11 analyzable primary G1XP tumors from the first cohort exhibited clonal Igh rearrangements, along with varying degrees of a germline J_H band (Additional file 1: Figure S1C). However, the germline band usually occurred in low levels indicating derivation from non-B lineage cells within the tumor. Consistently, the tumor samples analyzed from the second cohort also displayed clonal rearrangements of the J_H allele (Fig. 2b). Histological analysis of the tumor samples showed that the lymphoma cells possessed diffusely enlarged cytoplasm, salient nuclei, and nucleoli (Fig. 2c), consistent with immunoblastic B cells. Phenotypic characterization of G1XP lymphomas was performed by flow cytometry (Additional file 1: Table S1). We found that all of the lymphoma samples consistently expressed B220, CD24, CD38, CD43, CD93, and CD138. The majority of lymphomas also expressed PNA, a marker for GC B cells (see below), and about 40 % of them were surface IgG positive (Additional file 1: Table S1).

To test whether G1XP lymphomas derived from B cells that had undergone SHM/CSR, we cloned the VDJ_H exons from three G1XP lymphomas using a PCR approach [47]. We found that G1XP tumors contained structurally normal in-frame or out-of-frame V(D)J rearrangements; more importantly, we sequenced these VDJ_H exons and downstream J_H introns and found that these three tumors harbored 13, 1, and 3 mutations, respectively, in the J_H regions (Additional file 1: Figure S2). These results further confirm that inactivation of Xrcc4 and Trp53 via Cγ1cre leads to B cell lymphomas capable of undergoing SHM. Thus, we conclude that the deletion of both Xrcc4 and Trp53 genes via Cγ1Cre results in the development of novel B cell lymphomas.

We performed whole genome next generation sequencing (NGS) to identify the inter-chromosomal translocations (CTX) involving Ig loci in 6 G1XP lymphomas. NGS allows us to molecularly characterize the CTX junctions at the base pair (bp) level, which may provide insights into the molecular mechanisms leading to these translocations. NGS identified 25 CTXs involving Ig genes, 20 of which were confirmed by manual alignment to reference databases (mm9), while the remainder was excluded due to alignment artifacts or sequencing limitations. Among the 20 CTXs, 16 of them occurred at the Igh locus, with 10 breakpoints present within or around S regions on chromosome 12q13.2 (Fig. 3a, Additional file 1: Table S2), indicating that the translocations originated from CSR process; the other 6 CTXs occurred in close proximity to various V_H gene segments of the Igh locus (Fig. 3a, Additional file 1: Table S2), suggesting that these CTXs were mediated by RAGs. The additional 4 CTXs targeted Igκ or Igλ loci on chromosome 6 or 16, respectively, with the breakpoints occurring in close proximity to Vλ or Vκ gene segments, consistent with RAG-mediated DSBs (Fig. 3a, Additional file 1: Table S2).

Notably, an interesting pattern emerged among these CTXs that apparently separated into two categories. First, all of the 10 CTXs occurring around S or C regions of the Igh locus were translocated to the c-myc or plasmacytoma variant translocation 1 (Pvt1) locus on chromosome 15 (8 vs 2, respectively) (Fig. 3a). The translocation breakpoints at the c-myc locus clustered in the 5′ non-coding regions on chromosome 15qD1 (Fig. 3b, Additional file 1: Table S2). Second, all of the 10 CTXs occurring close to V gene segments of Igh, Igκ, and Igλ were translocated to various chromosomes involving random genetic or intergenic regions (Fig. 3a) (Additional file 1: Table S2). This intriguing pattern of translocation partners suggests that distinct mechanisms operate to mediate these two categories of translocations, probably involving AID vs RAGs, respectively. NGS data revealed a novel translocation partner of Igh locus, which targeted the Pvt1 locus (Fig. 3c), located about 30 kb telomeric of the c-myc locus. The Pvt1 locus contains a long non-coding RNA gene, which is a frequent translocation partner in mouse plasmacytomas and variants of human Burkitt’s lymphomas (BL) [50, 51]. NGS data identified Igh breakpoints located within Sμ or Sα regions, strongly implicating the involvement of AID (Fig. 3d). The junctional sequences harbored micro-homology (MH), an indicator of alternative end-joining (A-EJ) [35, 52]. Taken together, our data suggest that Ig loci translocations are likely caused by AID or RAG activity.

Apart from CTX junctions, we also identified V(D)J recombination junctions from our NGS data. Since Xrcc4 was deleted in the GC B cells, these lymphoma cells harbored normal D-J or V-D-J recombination junctions at the Igh locus (Additional file 1: Figure S3, junction 1 and 2, respectively). In contrast, we detected aberrant Igλ locus V-J rearrangements, which harbored large deletions of Vλ1 or Jλ3 exon including 41 bp of Vλ1 exon and the entire Jλ3 exon (Additional file 1: Figure S3, junction 3). These data are consistent with our previous results showing that CD21cre-mediated Xrcc4 deletion led to aberrant Igλ rearrangements [15, 47]. We identified MH at the V-J junction (Additional file 1: Figure S3, junction 3), consistent with the involvement of A-EJ. Taken together, our data suggest that these aberrant Igλ rearrangements occurred in the absence of Xrcc4, probably in the context of secondary V(D)J recombination.

To further validate the observed genomic instability, we performed metaphase FISH assay to detect locus-specific clonal translocations, which were also validated via independent methods (Additional file 1: Table S3). Of note, all of the detailed analysis utilized the tumor samples from the second cohort of Cγ1CreX ^c P ^c mice. Among the six tumors sequenced by NGS, we confirmed that four out of six samples harbored clonal c-myc translocations while five out of six had Igh locus translocation (Fig. 4a). Clonal translocations were defined as more than 80 % of tumor metaphases harboring such translocations. Furthermore, we found that six out of nine analyzed G1XP lymphoma samples harbored clonal c-myc translocations while seven out of nine contained clonal Igh translocations. Next, we performed FISH analysis using 3′Igh (red) and 3′c-myc (green) probes or 5′c-myc (red) and 5′Igh probes (green) to determine whether these are reciprocal translocations occurring between Igh and c-myc loci. Indeed, we found that Igh and c-myc probes were juxtaposed on the translocated chromosomes (Fig. 4b). Thus, we conclude that the majority of G1XP lymphomas harbor Igh-c-myc reciprocal translocations. In addition to these translocations, we detected clonal translocations involving the Igλ locus (Additional file 1: Figure S4), one of the Ig light chain gene loci, in the majority of G1XP lymphomas analyzed.

Interestingly, we identified a clonal Igh translocation in one G1XP lymphoma that did not involve c-myc locus since c-myc FISH probes had intact signals (Fig. 4c). Notably, this particular G1XP lymphoma (119 J) is aneuploid as evidenced by five copies of broken Igh locus and five copies of intact c-myc locus (Fig. 4c), suggesting the existence of novel translocation partners. Indeed, NGS data showed that this particular lymphoma harbored an Igh-Pvt1 translocation (Fig. 3d). Next, we performed FISH analysis to validate the occurrence of this translocation. Surprisingly, we found that almost all metaphases harbored an intact Pvt1 locus with 5′ and 3′ Pvt1 probes always co-localized (Fig. 5a, left panel, and 5b), suggesting the absence of gross chromosomal structural alterations at the Pvt1 locus. To reconcile the apparent discrepancy between NGS and FISH data, we used the 3′Igh and 3′Pvt1 probes for FISH assay and found that these two probes co-localized in 100 % of metaphases analyzed (Fig. 5a, middle panel and 5b). In contrast, the 5′Igh and 5′Pvt1 probes were not co-localized (Fig. 5a, right panel, and 5b). Thus, these data demonstrated that a piece of chromosome 12 was inserted into the Pvt1 locus, which contained the centromeric portion of the Igh locus encompassing the region from Sα to Sμ (Fig. 5c). More importantly, these data implicate that G1XP lymphomas harbor complicated genomes including segmental translocations.

We hypothesize that G1XP lymphomas probably harbor ongoing genomic instability that mediates the acquisition of chromosomal structural rearrangements. To test this hypothesis, we examined the DNA damage level in G1XP lymphomas via detecting the phosphorylation of H2AX (at Ser 139), a variant of H2A. H2AX phosphorylation occurs at sites flanking DSBs minutes after their induction, leading to the formation of distinct gamma-H2AX foci (γ-H2AX) [53]. Thus, γ-H2AX foci reflect the level of ongoing DSBs and are often used as an indicator of DSB formation. We used a flow cytometry-based method to measure γ-H2AX level in the G1XP lymphomas and in wt naïve (B220⁺PNA^low) and GC (B220⁺PNA^high) B cells as controls. Our data showed that G1XP lymphoma cells displayed a remarkably high level of γ-H2AX staining (Fig. 6a, right panel). In addition, this particular lymphoma expressed the GC B cell marker, PNA (Fig. 6a, left panel). In contrast, both the wt naïve and GC B cells exhibited a minimal level of γ-H2AX staining (Fig. 6b). Notably, the γ-H2AX level is consistently higher in GC B cells than naïve B cells (Fig. 6b). Thus, we conclude that G1XP lymphomas harbor ongoing DNA damage.

We predict that G1XP lymphomas may display a higher level of clonal heterogeneity. To test this possibility, we employed FISH assays to further analyze the lymphomas harboring the c-myc translocation in detail. Our analysis of these lymphomas showed that all metaphases indeed contained a c-myc translocation as shown by the split signals of 5′ and 3′ c-myc probes. Unexpectedly, we found that the configuration of FISH signals was highly heterogeneous since we observed an array of configurations that showed various numbers of intact, green (G) only or red (R) only FISH signals (Fig. 7a). It appeared that there were a few dominant subclones, for example, clone (1,3,2) and clone (2,2,2) (intact, G only, R only) (Fig. 7b), which occurred at a relatively high frequency, and many minor subclones carried various configurations of FISH signals (Fig. 7a). Thus, our G1XP lymphomas may provide a unique and novel model to study the mechanism of ongoing genomic instability and clonal heterogeneity.

Cγ1Cre knock-in (KI) mice [49], Xrcc4 [37], or Trp53 [81] conditional knock-out (KO) mice were generated previously. These mice were in mixed genetic background of C57BL/6, 129/Ola, and FVB/N [37, 49, 81]. Animal work was approved by the Institutional Animal Care and Use Committee of University of Colorado Anschutz Medical Campus (Aurora, CO), National Jewish Health (Denver, CO), and Children’s Hospital in Boston (Boston, MA).

Splenic B cells were isolated from naïve mice, purified by negative selection kit (Stem Cell Technologies, Canada), activated with anti-CD40 and IL4 as described previously [82], and collected 4 days after culture for metaphase preparation and FISH analysis. FISH analysis was performed with specific BAC probes as previously described [82] (see details in Additional file 1). Genomic DNA was isolated from tumor masses or normal tissues from control mice, and Southern blotting was performed as previously described [41].

Tumor masses usually presented in the abdomen and were adjacent to gut-associated lymphoid tissues such as mesenteric lymph nodes or Peyer’s patch. Tumors were dissected, fixed in 10 % formalin, and processed for H&E histology staining. Tumor single-cell suspensions were prepared and subjected to flow cytometry for phenotypic characterization. Samples were analyzed using a FACSCalibur (BD Bioscience), and FACS analysis was performed with FlowJo software. Single-cell suspensions prepared from tumors or wt naïve or GC B cells were stained for γ-H2AX foci according to the manufacturer’s instructions (BD Bioscience). The generation of splenic GC B cells was induced by in vivo immunization as described previously [82].

Tumor DNA samples were employed to generate the NGS paired-end library using the standard TruSeq DNA library preparation kit (Illumina, San Diego, CA). The libraries were subjected to whole genome sequencing on the Illumina Hi-Seq 2000 platform (pair-ended, 2 × 100 bp per read). Details of NGS analysis are provided in the Additional file 1 including usage of CREST software [83] and generation of Circos plots [84].