Mouse Model of Endemic Burkitt Translocations Reveals the Long-Range Boundaries of Ig-Mediated Oncogene Deregulation

Mouse model of endemic Burkitt translocations reveals the long-range boundaries of Ig-mediated oncogene deregulation

Alexander L. Kovalchuka,1, Camilo Ansarah-Sobrinhob,1,Oﬁr Hakimc,1, Wolfgang Reschb, Helena Tolarováb, Wendy Duboisd, Arito Yamaneb, Makiko Takizawab, Isaac Kleinf, Gordon L. Hagerc, Herbert C. Morse IIIa, Michael Potterd,1, Michel C. Nussenzweige,f,2, and Rafael Casellasb,g,2

aLaboratory of Immunogenetics, National Institute of Allergy and Infectious Diseases, National Institutes of Health, Rockville, MD 20852; bGenomics and Immunity, National Institute of Arthritis and Musculoskeletal and Skin Diseases, National Institutes of Health, Bethesda, MD 20892; cLaboratory of Receptor Biology and Gene Expression, National Cancer Institute, National Institutes of Health, Bethesda, MD 20892; dLaboratory of Cancer Biology and Genetics, National Cancer Institute, National Institutes of Health, Bethesda, MD 20892; eLaboratory of Molecular Immunology, and fHoward Hughes Medical Institute, The Rockefeller University, New York, NY 10065; and gCenter of Cancer Research, National Cancer Institute, National Institutes of Health, Bethesda, MD 20892

Edited by Klaus Rajewsky, Max-Delbrück-Center for Molecular Medicine, Berlin, Germany, and approved May 22, 2012 (received for review January 5, 2012)

Human Burkitt lymphomas are divided into two main clinical human Burkitt lymphomas (BLs), mouse plasmacytomas (PCTs), variants: the endemic form, affecting African children infected with and rat immunocytomas (15, 16). malaria and the Epstein-Barr virus, and the sporadic form, distrib- Human BLs are divided into two main clinical variants: the uted across the rest of the world. However, whereas sporadic endemic form, affecting African children infected with malaria translocations decapitate Myc from 5′ proximal regulatory ele- and the Epstein-Barr virus, and the sporadic form, distributed ments, most endemic events occur hundreds of kilobases away across the rest of the world (17). Although both tumors are from Myc. The origin of these rearrangements and how they de- driven primarily by Igh/Myc rearrangements (18), the precise regulate oncogenes at such distances remain unclear. We here location of the breakpoints differs in the two subsets. In sporadic recapitulate endemic Burkitt lymphoma-like translocations in BLs (and mouse PCTs), translocations occur within Myc exon

– ′ IMMUNOLOGY plasmacytomas from uracil N-glycosylase and activation-induced 1 intron 1, and as such they decapitate the oncogene from its 5 cytidine deaminase-deficient mice. Mapping of translocation regulatory elements (19). These canonical rearrangements result – breakpoints using an acetylated histone H3 lysine 9 chromatin from AID activity at IgS (during CSR) as well as at Myc (20 23). immunoprecipitation sequencing approach reveals Igh fusions up In contrast, endemic BL translocations are noncanonical in that ∼ they map hundreds of kilobases upstream from Myc basal to 350 kb upstream of Myc or the related oncogene Mycn.A – comprehensive analysis of epigenetic marks, PolII recruitment, promoters (24 28). Because these lesions leave Myc upstream and transcription in tumor cells demonstrates that the 3′ Igh en- promoter elements intact, it is unclear how they bring about hancer (Eα) vastly remodels ∼450 kb of chromatin into translo- oncogene deregulation. The etiology of these events is also un- cated sequences, leading to significant polymerase occupancy known, primarily because to date endemic BL-like translocations and constitutive oncogene expression. We show that this long- have not been systematically reproduced in the mouse. Because of their proximity to V(D)J genes, endemic BL range epigenetic reprogramming is directly proportional to the rearrangements have been proposed to arise as by-products of physical interaction of Eα with translocated sites. Our studies thus SHM but not of CSR (19, 27, 29). On the basis of this idea and in uncover the extent of epigenetic remodeling by Ig 3′ enhancers an attempt to recapitulate endemic translocations in the mouse, and provide a rationale for the long-range deregulation of translo- we combined a UNG null allele (30) with the Bcl-xL transgene − − cated oncogenes in endemic Burkitt lymphomas. The data also shed that produces plasma cell tumors (31). In UNG / B cells, AID light on the origin of endemic-like chromosomal rearrangements. hypermutation remains active as evidenced by high levels of C-to-T mutation (32). In contrast, CSR (32, 33) and canonical chromosome conformation capture | chromosome translocations | Igh/Myc translocations (34) are impaired because uracyl deglyco- activation-induced cytidine deamination | epigenetics sylation is an important intermediate step in double-strand DNA − − (dsDNA) break formation (12). The UNG / mouse thus might be uring B-cell responses to antigen, somatic hypermutation an ideal model to recapitulate rare, noncanonical translocations. D(SHM) introduces point mutations at variable (V) genes, a Consistent with a primary role for AID and UNG in the CSR process that, when coupled to clonal selection, can enhance the breaks mediating Igh/Myc translocations (20, 23, 34), we find that, antibody’saffinity for the antigen (1, 2). Class switch re- in the absence of UNG, the incidence of B-cell transformation is − − combination (CSR) is a second type of B-lymphocyte–specific markedly delayed. Strikingly, the vast majority of UNG / and − − DNA modification reaction that results in diversification of an- AID / tumors that do develop carry endemic BL-like trans- tibody effector functions by replacing the Igh constant domain locations with breakpoints accumulating over hundreds of kilo- Cμ with one of a set of downstream genes: Cγ,Cɛ,orCα (3–5). bases upstream of Myc or its homolog the Mycn oncogene. Both SHM and CSR are initiated by activation-induced cytidine Analysis of these tumors by deep-sequencing techniques sheds deaminase (AID) (6, 7), a B-cell–specific enzyme that deami- nates cytidine residues to produce uracils (8). Multiple repair pathways including uracil DNA glycosylase (UNG), mismatch Author contributions: A.L.K., G.L.H., H.C.M., M.C.N., and R.C. designed research; A.L.K., repair enzymes, and error-prone polymerases process U:G mis- C.A.-S., O.H., H.T., W.D., A.Y., M.T., and I.K. performed research; W.R., M.P., and R.C. matches into SHM or DNA double-strand breaks intermediate analyzed data; and M.P. and R.C. wrote the paper. to CSR (3–5, 9). The authors declare no conflict of interest. Although most AID-mediated lesions result in SHM or CSR, This article is a PNAS Direct Submission. they can occasionally generate chromosome deletions, trans- 1A.L.K., C.A.-S., O.H., and M.P. contributed equally to this work. – locations, and aneuploidy (10 14). These genomic abnormalities 2To whom correspondence may be addressed. E-mail: [email protected] or nussen@ have the potential to deregulate oncogenes and thereby promote rockefeller.edu. cell transformation. Perhaps the best-studied example is the This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. translocation that juxtaposes Igh to the Myc proto-oncogene in 1073/pnas.1200106109/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1200106109 PNAS Early Edition | 1of6 Downloaded by guest on October 2, 2021 light on both the origin of endemic-like translocations and the mouse lines for canonical T(12;15)s by long-distance PCR. As genomic features that impact their tumor capacity. previously described in wild-type mice (31), UNG+/+Bcl-xL and − UNG+/ Bcl-xL accumulated T(12:15)s following pristane injection. Results From 264 DNA samples analyzed for each strain, we isolated 39 − − +/+ UNG / PCTs Lack Canonical Igh/Myc Translocations. UNG deficiency (14.8%) and 36 (13.6%) distinct translocations from UNG +/− has been shown to profoundly impair CSR and canonical Igh/ Bcl-xL and UNG Bcl-xL mice, respectively (Fig. 1C). In con- Myc translocations in LPS+IL-4–stimulated B cells in vitro (34) trast, we failed to detect canonical T(12;15)s in any of the 216 −/− without significantly affecting SHM (32, 33) (Fig. S1). To in- DNAs isolated from UNG Bcl-xL lymphoid tissues or oil vestigate how this phenotype might impact the genesis of chro- granulomas (n = 216, Fig. 1C). This discrepancy, which is con- − − − mosomal translocations in vivo, UNG+/+, UNG+/ , and UNG / sistent with previous ex vivo translocation studies (34), may be mice bearing the Bcl-xL transgene were given one (day 0) or two explained by the fact that dsDNA break formation at switch (days 0 and 60) injections of the mineral oil pristane in the regions is decreased in the absence of UNG (35) (Fig. S2). The peritoneal cavity. Tumor development, which occurs in oil results therefore demonstrate that plasmacytomagenesis in granulomas, was diagnosed by the appearance of transformed Bcl-xL transgenic mice that lack UNG resembles that observed − −/− plasma cells in the ascites. UNG+/+ and UNG+/ mice rapidly in AID mice in that it is not associated with canonical Igh/ developed plasmacytomas with no apparent difference in kinet- Myc translocations. ics (Fig. 1A). In contrast, plasmacytomagenesis was inefficient in − − −/− the absence of UNG. Compared with wild-type controls, UNG / UNG Translocations Preferentially Localize Upstream of Myc Tran- −/− tumors developed with lower frequency (18% vs. 90% at day 220, scription Start Site. To explore whether UNG tumors carried P < 0.0005) and longer mean latency (136 vs. 73 d, Fig. 1A). noncanonical forms of T(12:15)s or other translocations, we next Consistent with previous findings (31), UNG+/+Bcl-xL mice applied FISH and spectral karyotyping (SKY). Despite our in- −/− developed multiple foci of premalignant, hyperchromatic plasma ability to amplify T(12:15)s by PCR, 9 of the 11 UNG tumors cells harboring Igh/Myc translocations [T(12;15)s] as early as 10 d carried rearrangements between chromosomes 12 and 15 (Fig. post pristane injection, whereas their appearance was delayed 2A and Fig. S3). To precisely map the translocation breakpoints, and their number reduced in the absence of UNG [7% (3 of 45 we made use of high-resolution array comparative genomic hy- mice analyzed), Fig. 1B]. Thus, by comparison with UNG- bridization (aCGH) and genome-wide chromatin immunopre- − − proficient littermates, UNG / Bcl-xL mice are relatively re- cipitation sequencing (ChIP-Seq) for acetylated histone H3 sistant to plasma cell tumor development. lysine 9 (H3K9Ac). H3K9Ac is a chromatin mark associated with To examine the nature of the translocations in premalignant active transcription (36, 37) that allows one to detect unexpected − − UNG / plasma cell foci, we assayed peritoneal oil granulomas discontinuities in active chromatin at translocation breakpoints. ’ For example, Fig. 2B shows H3K9Ac ChIP-Seq data for the (15), Peyer s patches, and mesenteric lymph nodes from all three − − UNG / PCT 6937. In this tumor, the H3K9Ac island at Eμ-Sμ- Cμ on chromosome 12 was truncated near Cμ (Fig. 2B, III), but resumed on chromosome 15, upstream of the Myc promoter [Fig. B ABFocus Atypical plasma cells 2 , II (red arrow)], an area lacking H3K9Ac in primary B cells fi fi 100 (Fig. 2B, I). As con rmed by PCR ampli cation and sequencing,

UNG+/+ P

90 =0.2 the T(12;15) breakpoint was indeed noncanonical in that it 80 mapped 7,542 bp 5′ of the Myc transcription start site (TSS) UNG+/- −/− 70 (Dataset S1). Notably, nearly all (8 of 10) UNG PCTs carried 60 P <0.00 70 translocation junctions from 0.8 to 314 kb upstream of the Myc 50 UNG+/+ TSS (Fig. 2C, Dataset S1, and Fig. S4 for a UCSC browser view).

05 60 40 UNG-/- 50 In stark contrast, 11 of 12 translocations isolated from tumors of 30 Plasmacytomas (%) -/- ﬁ UNG typical foci 40 UNG-pro cient mice were located downstream of the Myc TSS 20 a 10 30 in exon 1 or intron 1 (Fig. 2C, Dataset S1, and Fig. S4). Con- −/− 0 20 sistent with impaired CSR in UNG B cells, only 6 of 11 of the 0 50 100 150 200 250 − − 10 / days post-pristane %micewith UNG Bcl-xL translocations were directly within S domains, 0 days 5-9 10-14 15-19 20-24 whereas all translocations isolated from wild-type tumors involved these sequences (Dataset S1). Despite these differences, +/+ P =0.37P < 0.0005 Myc expression was comparable between Ung Bcl-xL and C Canonical 45 −/− Igh/Myc translocations 40 39/264 UNG Bcl-xL tumors (Fig. 2D). 36/264 − − 35 / s Finally, two UNG Bcl-xL tumors did not bear rearrange- UNG+/+ n

o 30 i t 25 ments between chromosomes 12 and 15. One displayed a T(6;15) oca 20 κ +/- sl juxtaposing Ig to the Myc-Pvt1 locus [tumor 6200 (Fig. S5) UNG n 15

Tra 10 and +74 kb (Fig. 2C)], and a second carried a chromosome -/- 5 12 paracentric inversion (Inv12, tumor 7210, Fig. S6) that UNG 0/216 0 μ ′ UNG+/+ UNG+/- UNG-/- fused S to the Mycn 3 UTR, leading to overexpression of Mycn and transcriptional repression of Myc (Fig. S6) as previously Fig. 1. Impaired PCT development and absence of canonical T(12:15) in described in similar tumors (31, 38). On the basis of these find- UNG-deficient mice. (A) Incidence of plasmacytomas in pristane-injected − − − ings we conclude that, in addition to decreasing the incidence UNG+/+ (magenta diamonds), UNG+/ (yellow squares), and UNG / (black and increasing the latency of malignant transformation, UNG triangles) mice expressing the Bcl-xL transgene. Tumors were diagnosed by deletion results in a shift of translocation breakpoints in chro- the appearance of 10 or more atypical plasma cells in the ascites. Statistical mosome 15 from the 3′ to the 5′ side of Myc TSS. significance for A was calculated via Fisher’s exact test. (B)(Upper) Photo- micrographs showing an example of atypical plasma cell foci developing in − − UNG+/+ and UNG / Tumors Recapitulate Sporadic and Endemic BL oil granulomas of pristane-treated mice. (Lower) Atypical focus formation in −/− +/+ Translocations, Respectively. The segregation of Igh/Myc trans- UNG mice (black bars) and littermate UNG controls (magenta bars) at − − locations in UNG+/+ and UNG / mice resembles the trans- various times following pristane injection. (C) Igh/Myc canonical chromo- fi somal translocations were amplified by PCR from DNA samples isolated from location pro les observed in sporadic and endemic BLs in − − − UNG+/+,UNG+/ , or UNG / mice 25–35 d following pristane injection using humans. Translocations isolated from PCTs of wild-type mice are nested primers specificforMyc intron 1 and Igh Cμ,Cγ1, Cγ2b, and Cα. akin to those in sporadic BL in that both involve Igh rearrange- (Right) Total number of PCR products identified in multiple tissues and ments at a hotspot centered in the first intron of Myc (Fig. 3, IV confirmed by sequence analysis. and V). This well-defined translocation hotspot corresponds to

2of6 | www.pnas.org/cgi/doi/10.1073/pnas.1200106109 Kovalchuk et al. Downloaded by guest on October 2, 2021 Myc A B −/− Fig. 2. Mapping of noncanonical UNG -/- I UNG tumor 6937 H3K9Ac translocations by SKY and deep sequencing. WT Chr15 ﬁ 7542 bps (A) SKY identi es reciprocal T(12;15) and −/− 12345 II T(15:12) translocations in UNG tumor 6937. Other abnormalities include trisomy 1, 16, Chr15 67 8 910 and 19. (B) H3K9Ac genome-wide mapping III reveals the precise Igh/Myc translocation breakpoint (red arrow) joining chromosomes 11 12;15 13 14 15;12 Tumor 6937 Chr12 − − 12 (III) and 15 (II) in UNG / tumor 6937. For IV comparative purposes, the same proﬁle is 16 17 18 19 X WT Chr12 provided for primary B cells activated by LPS +IL-4 (I and IV). Myc and Eμ-Sμ-Cμ genes are Cμ Sμ Eμ +0.1Kb depicted below each lane. ChIP-Seq peaks C -0.8Kb D represent the number of sequence tags for -/- 6 UNG translocations -1Kb every 200-bp genomic window and denote -4.8Kb the relative magnitude of H3K9 acetylation.

h ) 4 -314Kb -241Kb -180Kb -140Kb -24Kb +74Kb d (C) Myc mouse locus (including the Pvt1 gene) Pvt1 2 showing the distribution of translocations −/− +/+ -403Kb ( Myc / Gap from PCT of UNG and UNG mice. The 2 ) Myc 0 (3 E log distance of each translocation breakpoint (arrows) relative to Myc TSS is provided in -2 kilobases. (D) Myc transcript levels in PCT of -0.4Kb to +1.4Kb UNG−/− and UNG+/+ mice carrying T(12;15) or 2) 5) +/+ 1 UNG translocations Tumors 2;15) 2; 1 (12;1 T(12-12) (Mycn driven tumor) translocations. T(1 T( T -/- -/- /+ G G + Values were normalized based on Gapdh NG UN UN U transcripts and plotted as the log2 ratio.

one of ∼150 sites in the genome that suffer frequent AID-mediated the profiles of translocations found in cases of sporadic and DNA damage (39–41). These data are consistent with the notion endemic BLs, respectively. IMMUNOLOGY that most Igh/Myc translocations in PCTs in wild-type mice and in sporadic BL are the result of AID activity (23, 39, 40, 42). 3′Eα Functional Interactions Are Asymmetrical and Reprogram ∼450 kb In contrast to wild-type tumors, Igh translocations in UNG- of Chromatin. Translocations between Igh and sequences down- deleted PCTs recapitulated the profile of rearrangements in stream of Myc intron 1 (Fig. 3, right dotted line), including a sec- human endemic BL in that breakpoints did not accumulate at ond AID-dependent hotspot at Pvt1 exon 5 (Fig. 3, VI), truncate well-defined hotspots but were instead distributed within a ∼350-kb Myc-coding sequence or result in acentric chromosomes that are domain 5′ of Myc (Fig. 3, II and III). Within this window, the incompatible with cellular viability. Consistent with this view, such pattern closely resembled the random distribution of Igh trans- events have been detected only in short-term B-cell cultures under − − locations isolated from AID / primary B cells under non- nonselective conditions (Fig. 3, I and VII) (39). However, these ′ selective conditions (Fig. 3, I) (39). These observations raise the features are unlikely to explain the 5 boundary of rearrangements − − −/− possibility that translocations in tumors in UNG / Bcl-xL mice from UNG PCT or endemic BL (depicted by the left dotted result from AID-independent dsDNA breaks. To directly test line in Fig. 3). Instead, we reasoned that this boundary might re- −/− flect a threshold beyond which the Igh 3′Eα enhancer can no this idea, we induced PCTs in AID Bcl-xL mice with pristane. −/− −/− longer deregulate Myc or Mycn at a distance. In PCT from AID As previously shown (31), AID mice rarely develop plasma- − − − − / ′ α cytomas. Nevertheless, we were able to isolate four AID / PCTs or UNG mice, these oncogenes were separated from 3 E by from 20 pristane-injected mice after a median latency of 158 d as up to 400 kb (Fig. 2C and Dataset S1), an observation that is compared with 73 d in AID wild-type mice (Fig. 1A). Similar to highly consistent with distances observed in endemic BLs (28). −/− The 3′Eα regulatory region consists of multiple enhancer PCT from UNG-deficient mice, three of the four AID PCTs elements that mediate Igh gene expression in its physiological carried translocations upstream of Myc or Mycn TSS (1.9, 54, and context (9, 43) and oncogene deregulation in translocated 233 kb, Dataset S1). In conclusion, Igh/Myc rearrangements in +/+ −/− chromatin (44, 45). To explore the extent and nature of 3′Eα PCTs of UNG Bcl-xL and UNG Bcl-xL mice recapitulate long-range activity in PCTs, we first mapped the 3′Eα interactome by chromosome conformational capture coupled with deep sequencing (4C-Seq) (41, 46). As bait, we used a DNA Igh translocations SSeeellleeecccttteeeddd(((~~~333555000KKKbbb)) sequence within 3′Eα (Materials and Methods) and constructed

I AID-/- primary B four independent 4C libraries from Myc and Mycn-driven PCTs. Resting and activated primary B cells were also included in the II Endemic Burkitt analyses. Remarkably, we found 3′Eα local interactions to be III UNG-/- PCT ′ Pvt1 100Kb asymmetrically biased to 5 sequences in all cases (Fig. 4, I and

Myc II). This unequal distribution can be explained by the presence of IV UNG+/+ PCT a documented CTCF-binding domain ∼4kb3′ of the enhancer V Sporadic Burkitt (Fig. 4, III), which has been proposed to insulate downstream ′ α VI AID+/+ primary B non-Ig genes from 3 E activity (47). In both tumors, the average contact frequency for 3′Eα across the entire cis-chromosome declined monotonically as a function of distance (Fig. S7). Be- Fig. 3. Segregation of sporadic and endemic BL-like translocations as a function of UNG or AID. Comparative analysis of chromosomal trans- tween 0.1 and 10 Mb from the bait, for example, 4C-Seq reads − − − locations in mouse primary B cells (I: AID / ; VI: AID+/+) activated ex vivo with showed a power law scaling with an exponent of 1.19 (Fig. S7), LPS+IL-4, as determined by TC-Seq; human BL (II: endemic; V: sporadic), which is in good agreement with that calculated in primary B − − compiled from the literature (27); mouse PCT (III: UNG / ; IV: UNG+/+) is from cells using the same bait (−1.06) (41). This correlation indicates Fig. 2C. Physical boundaries of selected translocations in human BL and that chromosomal translocations do not appreciably alter the mouse PCTs are depicted with dotted lines. local contact proﬁles of 3′Eα.

Kovalchuk et al. PNAS Early Edition | 3of6 Downloaded by guest on October 2, 2021 bait

) 200 I 3 -

150 Myc PCT 4C-Seq

100

50 ∼460Kb

Mean contact frequency (x10 0

Non-Ig genes 3'E Myc Pvt1 breakpoint ) 200 II -3

150 Mycn PCT 4C-Seq

100

∼ ′ α 50 480Kb Fig. 4. The 3 E local interactions at Myc-driven PCT TEPC 1165 (I) (62) or the Mycn-driven tumor 4961 (II) (31); 5′ and Mean contact frequency (x10 0 3′ interactions are depicted with grey and black circles, respectively. Average Non-Ig genes 3'E 2a 2b 1 3 μ Mycn DV652083 interaction frequencies are represented breakpoint with lines using the same color codes. 0.6 III The vertical dotted line shows the lo- e CTCF cation of the 4C bait. The horizontal 0.4 dotted lines demarcate the overall mean interaction frequency of 3′Eα with 0.2 downstream non-Ig genes. The in- RPM/nucleoti d tersection of this mean with the 5′ in- 0 teraction average is shown with an -300 -200 -100 0 100 200 300 400 500 arrow. (III) CTCF-binding proﬁles at Igh/ Distance (Kb) Mycn Inv12 as determined by ChIP-Seq.

To estimate the functional genomic reach of 3′Eα, we com- these modifications were not present at the Mycn locus in pared its average contact frequency at translocated Myc and primary activated B cells (Fig. 5, III). The findings are thus Mycn loci with that measured downstream of the putative in- consistent with the notion that 3′Eα remodels ∼0.45 Mb of sulator where 3′Eα is not known to modulate gene expression. translocated chromatin, an estimate that agrees well with its Despite the fact that the distance between 3′Eα and the onco- predicted functional contact profile (Fig. 4) and the maximum genes was markedly different in the two tumors (15 kb for Myc distance separating 3′Eα from Myc or Mycn in PCTs and en- and 226 kb for Mycn, Dataset S1), the analysis showed very demic BLs (Fig. 3 and Dataset S1). similar results. In the Myc PCT, the 5′ and 3′ average contact frequencies intersected ∼460 kb 5′ of the 4C-Seq bait, whereas in Discussion the Mycn tumor, the intersection occurred at ∼480 kb (Fig. 4, I Early cytogenetic analyses of sporadic BL demonstrated that and II, respectively). These data predict that 3′Eα must influence translocation breakpoints at Myc accumulate at a well-defined gene expression over distances of at least 450 kb. To directly test hotspot comprising promoter-proximal sequences and most of this idea, we used deep sequencing to measure PolII, mRNA, intron 1 (24, 27, 28, 48). Several lines of evidence have since and nine activating chromatin modifications (H2AZ, H2BK5Ac, ascribed translocations at this site to AID activity. First, the Myc H2BK120Ac, H3K4me3, H3K36me2, H3K9Ac, H3K27Ac, hotspot physically overlaps with the recruitment profile of AID H4K8Ac, and H4K91Ac) in primary activated B cells and the (22) and its cofactor Spt5 (49). Spt5 plays an active role in RNA Mycn tumor. This PCT was chosen for further study because, in PolII stalling, and, as such, it is believed to facilitate AID in- contrast to Myc, the Mycn locus is transcriptionally silent in pe- teraction with single-stranded DNA (its physiological target) at ripheral B cells (22). As expected, Mycn was inactive in cultured PolII-stalling sites both in Ig and non-Ig genes (50). Consistent B cells stimulated with LPS+IL4 as evidenced by the absence of with this idea, the Myc translocation hotspot in stimulated B cells Pol II recruitment and mRNA synthesis at the locus (Fig. S8,I displays extensive PolII stalling (49) and accumulates substantial and II). In contrast, we found high levels of PolII and mRNA at hypermutation and dsDNA breaks in an AID-dependent manner Mycn and the nearby DV652083 EST in the translocated tumor (22, 41, 51). Formal proof that these lesions directly promote cells (Fig. S8, III and IV). Consistent with these findings, the Igh/Myc rearrangements was recently provided by gene-targeting translocated locus was associated with activating epigenetic studies (23) and genome-wide maps of B-cell translocations modifications, which could be clustered into two distinct groups. generated under nonselective conditions (39, 40). In both cases, H3K4me3, H2BK5Ac, H3K9Ac,andH3K27Acwerecircum- the sporadic BL translocation hotspot at Myc was recapitulated in − − scribed to Mycn, DV652083,andthe3′Eα enhancer (Fig. 5, I). AID+/+ but not in AID / primary mouse B cells, supporting the In contrast, H2AZ, H2BK120Ac, H3K36me2, H4K8Ac, and contention that AID directly targets Myc. H4K91Ac extended as a group uninterruptedly for about 450 kb In contrast to translocations found in sporadic BL, trans- from 3′Eα across the entire Ch locus into the translocated do- locations characteristic of endemic BL do not accumulate pref- main (Fig. 5, II). Beyond this point, no substantial differences erentially at Myc promoter-proximal sequences. Instead, they are were found in PolII recruitment, gene expression, or epigenetic scattered across several hundred kilobases 5′ of Myc TSS with no modifications between primary and tumor cells. Importantly, obvious bias for switch recombination homology sequences or

4of6 | www.pnas.org/cgi/doi/10.1073/pnas.1200106109 Kovalchuk et al. Downloaded by guest on October 2, 2021 Breakpoint I H3K4me3 H2BK5Ac H3K9Ac Plasmacytoma H3K27Ac

Fig. 5. Extent of epigenetic reprog- Non-Ig genes 3'E 2a 2b 1 3 μ Mycn DV652083 II ramming by 3′Eα. Chromatin modifications at the wild-type or rearranged Mycn locus in primary or tumor cells. Plasmacytoma (I) Composite analysis of H3K4me3, H2BK5Ac, H3K9Ac, and H3K27Ac de- ∼ 450Kb position at the Igh/Mycn inverted locus. H2Az (II) The same analysis for modifications H2BK120Ac H3K36me2 H2AZ, H2BK120Ac, H3K36me2, H4K8Ac, H4K8Ac H4K91Ac and H4K91Ac. The black arrow denotes III the extent of epigenetic remodeling above background signal, which was Primary cells (reconstructed translocation) measured in primary activated B cells (III). Although the Mycn locus is not rearranged in primary cells, for comparative purposes the compiled epigenetic modifications in III are shown -300 -200 -100 0 100 200300400500 over a genome carrying the same Distance (Kb) translocation as in I and II.

known AID hotspots, although these are not as well defined in more than 2 Mb away (57–59). This stage-specific locus contrac- humans as in mice (24, 27, 28, 48). Also in contrast to sporadic tion occurs via DNA looping and appears to be mediated by IMMUNOLOGY BL translocations, endemic events rarely involve IgS domains, specialized transcription factors, cohesin, and CTCF, all of which but appear to accumulate in the vicinity of Jh sequences (19). A assemble at dedicated domains within the heavy chain locus fi key aspect of our ndings is that we have recapitulated endemic (57, 58, 60, 61). For the most part, these sequences are deleted BL-like translocations in the absence of UNG. Analogous to the − − by functional V(D)J joints and thus have not been reported human tumors, translocation breakpoints in UNG / PCTs are in translocated Igh alleles. One exception is hs5-7, which spans distributed over hundreds of kilobases upstream of Myc TSS and ∼ ′ α do not display the strict bias for IgS domains seen in wild-type 10 kb of DNA downstream of 3 E and displays insulating mice. Importantly, a similar distribution of translocation break- functions in reporter constructs (47). Consistent with such activity, − − points was isolated from rare AID / tumors that occur in Bcl-xl we now show that 3′Eα local interactions are asymmetrically transgenic mice injected with pristane. Considering these sim- biased to 5′ sequences. Presumably, 3′ non-Ig genes are protected ilarities, it is likely that endemic-like rearrangements in both by hs5-7 from 3′Eα potent transcriptional enhancing activity. Al- mouse strains are generated by the same nontargeted mecha- though unproven, this idea is attractive because it helps explain nism, which remains to be characterized. how 3′Eα deregulates 5′-translocated oncogenes at very long During CSR, 3′Eα elements regulate transcription of targeted distances whereas proximal (3′)non-Ig genes are unaffected. By Ch domains at large distances (52). Mechanistically, this activity the same token, the presence of an insulator downstream of Igh correlates with direct physical interactions between 3′Eα, I pro- μ would also explain the observation that, even though Myc trans- moters, and E (53, 54), analogous to nuclear contacts observed locates on both sides of 3′Eα in primary B cells (39), only 5′ events at the β-globin loci and other enhancer-promoter systems (55). In B-cell tumors, 3′Eα is not needed for the development of are selected during tumorigenesis. A direct test of these ideas chromosomal translocations per se, but it physically associates awaits genetic deletion of the putative hs5-7 insulator. with rearranged oncogenes (56) and is required for their long- range activation (45). To measure the functional genomic reach of 3′Eα, we have mapped its interactome and the epigenetic alterations induced at translocated chromatin via 4C-Seq and ChIP-Seq. Our results demonstrate that 3′Eα interacts with and reprograms up to 0.45 Mb of translocated chromatin. Notably, the span of this interaction was comparable regardless of the distance separating 3′Eα from Myc or Mycn. This observation suggests that the topological constraints imposed by 3′Eα on chromatin are mostly independent of gene composition or location. Under this scenario, we find it useful to envision 3′Eα as “warping” nearly half a megabase of chromatin space, with oncogenes being deregulated by translocations that occur only within such a genomic window (Fig. 6). In the context of endemic −/− ′ BL and UNG PCT, this scenario explains the 5 boundary of ~0.5Mb selected translocation breakpoints, which reflect the upper limit of 3′Eα long-range oncogenic activity. It is important to point out that, under different developmental Fig. 6. “Warping” model of 3′Eα activity. Our studies indicate that the 3′Eα settings, 3′Eα may engage in substantially longer functional con- acts over half a megabase of chromatin space, with oncogenes such as Myc tacts. During V(D)J recombination in immature B cells, for ex- being deregulated by translocations that fall within this genomic window. ample, 3′Eα forms a chromatin hub with recombining diversity Non-Ig genes downstream of the enhancer might be protected from 3′Eα and variable domains, which are positioned in some instances activity by the hs5-7 insulator.

Kovalchuk et al. PNAS Early Edition | 5of6 Downloaded by guest on October 2, 2021 Materials and Methods ACKNOWLEDGMENTS. We thank Barbara Birshtein, Beverly Mock, and the Casellas laboratory for helpful comments; Jim Simone for assistance with cell SI Materials and Methods contains detailed information on plasmacytoma- sorting; Gustavo Gutierrez and Hong-Wei Sun for assistance with the Genome genesis induction; microscopy characterization of plasma cell foci; quantita- Analyzer and deep-sequencing pipeline; Val Bliskovsky for assistance with tive PCR measurements of Myc and Mycn expression; ChIP-Seq of chromatin sequencing and primer design;Zohreh Naghashfar for performing gene expres- modiﬁcations (Fig. S9); ampliﬁcation of translocations by PCR; ex vivo B-cell sion microarrays; and Kelly Nelson for assistance with FISH. The study made use ’ activation conditions; FISH, SKY, and aCGH analysis of tumors; 4C- and mRNA- of the National Institutes of Health s high-performance computational capabil- ities of Helix and Biowolf clusters. This research was supported in part by the Seq experiments; and bioinformatics analyses of all genomic datasets. Details Intramural Research Program of the National Institutes of Health; the National on spectral karyotyping are provided. Sequences of primers used in the Institute of Arthritis, Musculoskeletal and Skin Diseases; the National Institute various PCR reactions are also included. of Allergy and Infectious Diseases; and the National Cancer Institute.

1. Di Noia JM, Neuberger MS (2007) Molecular mechanisms of antibody somatic hy- 33. Imai K, et al. (2003) Human uracil-DNA glycosylase deficiency associated with pro- permutation. Annu Rev Biochem 76:1–22. foundly impaired immunoglobulin class-switch recombination. Nat Immunol 4: 2. Peled JU, et al. (2008) The biochemistry of somatic hypermutation. Annu Rev Immunol 1023–1028. 26:481–511. 34. Ramiro AR, et al. (2006) Role of genomic instability and p53 in AID-induced c-myc-Igh 3. Stavnezer J, Guikema JE, Schrader CE (2008) Mechanism and regulation of class switch translocations. Nature 440:105–109. recombination. Annu Rev Immunol 26:261–292. 35. Schrader CE, Linehan EK, Mochegova SN, Woodland RT, Stavnezer J (2005) In- 4. Chaudhuri J, et al. (2007) Evolution of the immunoglobulin heavy chain class switch ducible DNA breaks in Ig S regions are dependent on AID and UNG. JExpMed recombination mechanism. Adv Immunol 94:157–214. 202:561–568. 5. Honjo T, Kinoshita K, Muramatsu M (2002) Molecular mechanism of class switch re- 36. Kuchen S, et al. (2010) Regulation of microRNA expression and abundance during combination: Linkage with somatic hypermutation. Annu Rev Immunol 20:165–196. lymphopoiesis. Immunity 32:1–12. 6. Muramatsu M, et al. (2000) Class switch recombination and hypermutation require 37. Barski A, et al. (2007) High-resolution profiling of histone methylations in the human activation-induced cytidine deaminase (AID), a potential RNA editing enzyme. Cell genome. Cell 129:823–837. 102:553–563. 38. Ma A, et al. (1991) Mechanism of endogenous myc gene down-regulation in E mu-N- 7. Revy P, et al. (2000) Activation-induced cytidine deaminase (AID) deficiency causes the myc tumors. Mol Cell Biol 11:440–444. autosomal recessive form of the Hyper-IgM syndrome (HIGM2). Cell 102:565–575. 39. Klein IA, et al. (2011) Translocation-capture sequencing reveals the extent and 8. Di Noia JM, et al. (2007) Dependence of antibody gene diversification on uracil ex- nature of chromosomal rearrangements in B lymphocytes. Cell 147:95–106. cision. J Exp Med 204:3209–3219. 40. Chiarle R, et al. (2011) Genome-wide translocation sequencing reveals mechanisms of 9. Khamlichi AA, Pinaud E, Decourt C, Chauveau C, Cogné M (2000) The 3′ IgH regula- chromosome breaks and rearrangements in B cells. Cell 147:107–119. tory region: A complex structure in a search for a function. Adv Immunol 75:317–345. 41. Hakim O, et al. (2012) DNA damage defines sites of recurrent chromosomal trans- 10. Küppers R, Dalla-Favera R (2001) Mechanisms of chromosomal translocations in B cell locations in B lymphocytes. Nature 484(7392):69–74. lymphomas. Oncogene 20:5580–5594. 42. Takizawa M, et al. (2008) AID expression levels determine the extent of cMyc onco- 11. Casellas R, Yamane A, Kovalchuk AL, Potter M (2009) Restricting activation-induced genic translocations and the incidence of B cell tumor development. J Exp Med 205: cytidine deaminase tumorigenic activity in B lymphocytes. Immunology 126:316–328. 1949–1957. 12. Nussenzweig A, Nussenzweig MC (2010) Origin of chromosomal translocations in 43. Birshtein BK, et al. (1997) Murine and human 3′IgH regulatory sequences. Curr Top lymphoid cancer. Cell 141:27–38. Microbiol Immunol 224:73–80. 13. Zhang Y, et al. (2010) The role of mechanistic factors in promoting chromosomal 44. Janz S (2006) Myc translocations in B cell and plasma cell neoplasms. DNA Repair translocations found in lymphoid and other cancers. Adv Immunol 106:93–133. (Amst) 5:1213–1224. 14. Tsai AG, et al. (2008) Human chromosomal translocations at CpG sites and a theo- 45. Gostissa M, et al. (2009) Long-range oncogenic activation of Igh-c-myc translocations retical basis for their lineage and stage specificity. Cell 135:1130–1142. by the Igh 3′ regulatory region. Nature 462:803–807. 15. Potter M (2003) Neoplastic development in plasma cells. Immunol Rev 194:177–195. 46. Simonis M, et al. (2006) Nuclear organization of active and inactive chromatin do- 16. Klein G (1999) Immunoglobulin gene associated chromosomal translocations in B-cell mains uncovered by chromosome conformation capture-on-chip (4C). Nat Genet 38: derived tumors. Curr Top Microbiol Immunol 246:161–167. 1348–1354. 17. Burkitt DP (1983) The discovery of Burkitt’s lymphoma. Cancer 51:1777–1786. 47. Garrett FE, et al. (2005) Chromatin architecture near a potential 3′ end of the igh 18. Hecht JL, Aster JC (2000) Molecular biology of Burkitt’s lymphoma. J Clin Oncol 18: locus involves modular regulation of histone modifications during B-Cell de- 3707–3721. velopment and in vivo occupancy at CTCF sites. Mol Cell Biol 25:1511–1525. 19. Cory S (1986) Activation of cellular oncogenes in hemopoietic cells by chromosome 48. Guzmán E, Lis JT (1999) Transcription factor TFIIH is required for promoter melting translocation. Adv Cancer Res 47:189–234. in vivo. Mol Cell Biol 19:5652–5658. 20. Ramiro AR, et al. (2004) AID is required for c-myc/IgH chromosome translocations 49. Pavri R, et al. (2010) Activation-induced cytidine deaminase targets DNA at sites of in vivo. Cell 118:431–438. RNA polymerase II stalling by interaction with Spt5. Cell 143:122–133. 21. Ramiro A, et al. (2007) The role of activation-induced deaminase in antibody di- 50. Pavri R, Nussenzweig MC (2011) AID targeting in antibody diversity. Adv Immunol versification and chromosome translocations. Adv Immunol 94:75–107. 110:1–26. 22. Yamane A, et al. (2011) Deep-sequencing identification of the genomic targets of the 51. Liu M, et al. (2008) Two levels of protection for the B cell genome during somatic cytidine deaminase AID and its cofactor RPA in B lymphocytes. Nat Immunol 12:62–69. hypermutation. Nature 451:841–845. 23. Robbiani DF, et al. (2008) Activation induced deaminase is required for the chromo- 52. Pinaud E, et al. (2011) The IgH locus 3′ regulatory region: Pulling the strings from somal translocations in c-myc that lead to c-myc/IgH translocations. Cell 135:1028–1038. behind. Adv Immunol 110:27–70. 24. Gutiérrez MI, et al. (1992) Molecular epidemiology of Burkitt’s lymphoma from South 53. Ju Z, et al. (2007) Evidence for physical interaction between the immunoglobulin America: Differences in breakpoint location and Epstein-Barr virus association from heavy chain variable region and the 3′ regulatory region. JBiolChem282: tumors in other world regions. Blood 79:3261–3266. 35169–35178. 25. Pelicci PG, Knowles DM, II, Magrath I, Dalla-Favera R (1986) Chromosomal breakpoints 54. Wuerffel R, et al. (2007) S-S synapsis during class switch recombination is promoted by and structural alterations of the c-myc locus differ in endemic and sporadic forms of distantly located transcriptional elements and activation-induced deaminase. Immu- Burkitt lymphoma. Proc Natl Acad Sci USA 83:2984–2988. nity 27:711–722. 26. Neri A, Barriga F, Knowles DM, Magrath IT, Dalla-Favera R (1988) Different regions of 55. de Laat W, et al. (2008) Three-dimensional organization of gene expression in ery- the immunoglobulin heavy-chain locus are involved in chromosomal translocations in throid cells. Curr Top Dev Biol 82:117–139. distinct pathogenetic forms of Burkitt lymphoma. Proc Natl Acad Sci USA 85: 56. Duan H, Xiang H, Ma L, Boxer LM (2008) Functional long-range interactions of the IgH 2748–2752. 3′ enhancers with the bcl-2 promoter region in t(14;18) lymphoma cells. Oncogene 27: 27. Shiramizu B, et al. (1991) Patterns of chromosomal breakpoint locations in Burkitt’s 6720–6728. lymphoma: Relevance to geography and Epstein-Barr virus association. Blood 77: 57. Guo C, et al. (2011) Two forms of loops generate the chromatin conformation of the 1516–1526. immunoglobulin heavy-chain gene locus. Cell 147:332–343. 28. Joos S, et al. (1992) Variable breakpoints in Burkitt lymphoma cells with chromosomal 58. Guo C, et al. (2011) CTCF-binding elements mediate control of V(D)J recombination. t(8;14) translocation separate c-myc and the IgH locus up to several hundred kb. Hum Nature 477:424–430. Mol Genet 1:625–632. 59. Jhunjhunwala S, et al. (2008) The 3D structure of the immunoglobulin heavy-chain 29. Magrath I (1990) The pathogenesis of Burkitt’s lymphoma. Adv Cancer Res 55: locus: Implications for long-range genomic interactions. Cell 133:265–279. 133–270. 60. Ebert A, et al. (2011) The distal V(H) gene cluster of the Igh locus contains distinct 30. Endres M, et al. (2004) Increased postischemic brain injury in mice deficient in uracil- regulatory elements with Pax5 transcription factor-dependent activity in pro-B cells. DNA glycosylase. J Clin Invest 113:1711–1721. Immunity 34:175–187. 31. Kovalchuk AL, et al. (2007) AID-deficient Bcl-xL transgenic mice develop delayed 61. Degner SC, et al. (2011) CCCTC-binding factor (CTCF) and cohesin influence the ge- atypical plasma cell tumors with unusual Ig/Myc chromosomal rearrangements. J Exp nomic architecture of the Igh locus and antisense transcription in pro-B cells. Proc Natl Med 204:2989–3001. Acad Sci USA 108:9566–9571. 32. Rada C, et al. (2002) Immunoglobulin isotype switching is inhibited and somatic 62. Bauer SR, et al. (1989) Altered myc gene transcription and intron-induced stabiliza- hypermutation perturbed in UNG-deficient mice. Curr Biol 12:1748–1755. tion of myc RNAs in two mouse plasmacytomas. Oncogene 4:615–623.

6of6 | www.pnas.org/cgi/doi/10.1073/pnas.1200106109 Kovalchuk et al. Downloaded by guest on October 2, 2021