IKZF1 Deletions with COBL Breakpoints Are Not Driven by RAG-Mediated Recombination Events in Acute Lymphoblastic Leukemia.

IKZF1 deletion (ΔIKZF1) is an important predictor of relapse in both childhood and adult B-cell precursor acute lymphoblastic leukemia (B-ALL). Previously, we revealed that COBL is a hotspot for breakpoints in leukemia and could promote IKZF1 deletions. Through an international collaboration, we provide a detailed genetic and clinical picture of B-ALL with COBL rearrangements (COBL-r). Patients with B-ALL and IKZF1 deletion (n = 133) were included. IKZF1 ∆1-8 were associated with large alterations within chromosome 7: monosomy 7 (18%), isochromosome 7q (10%), 7p loss (19%), and interstitial deletions (53%). The latter included COBL-r, which were found in 12% of the IKZF1 ∆1-8 cohort. Patients with COBL-r are mostly classified as intermediate cytogenetic risk and frequently harbor ETV6, PAX5, CDKN2A/B deletions. Overall, 56% of breakpoints were located within COBL intron 5. Cryptic recombination signal sequence motifs were broadly distributed within the sequence of COBL, and no enrichment for the breakpoint cluster region was found. In summary, a diverse spectrum of alterations characterizes ΔIKZF1 and they also include deletion breakpoints within COBL. We confirmed that COBL is a hotspot associated with ΔIKZF1, but these rearrangements are not driven by RAG-mediated recombination.

Introduction

B-cell precursor acute lymphoblastic leukemia (B-ALL) comprises multiple subtypes defined by structural and numerical chromosomal alterations. These initiating lesions include aneuploidy and chromosomal rearrangements, leading to expression of ectopic fusion proteins and/or to deregulation of gene expression. During the evolution of B-ALL, a series of secondary genomic alterations, such as DNA copy number alterations (CNAs) and sequence mutations, usually emerges [1]. These secondary events commonly involve deletions of genes that encode regulators of cell-cycle (CDKN2A and CDKN2B) and B-cell development (PAX5 and IKZF1) [2].

IKZF1 deletions occur in ~15% patients with B-ALL and are more recurrent in the BCR-ABL1 and Philadelphia-like subgroups, affecting more than 70% and 40% of the patients, respectively [3]. IKZF1 is located within chromosome band 7p12 and codifies the transcription factor IKAROS. It has two main domains: a DNA-binding zinc finger domain (encoded by exons 4-6) and a zinc finger dimerization domain (encoded by exon 8). Some intragenic IKZF1 deletions impair its DNA-binding domain resulting in a dominant-negative effect (e.g. Δ4-7), while deletions of the whole gene (Δ1-8), as well as deletions of exons 1-2 (e.g. Δ2-3) lead to the gene haploinsufficiency. IKZF1 deletions have been associated with dismal prognosis in both pediatric and adult B-ALL [4], [5], despite the type of deletion [6]. Interestingly, sequencing of IKZF1 breakpoints suggested that aberrant recombination activating gene (RAG)–mediated recombination is responsible for these deletions [7]. Due to their clinical relevance, the identification of IKZF1 deletions has become an essential prognostic biomarker over the last decade. Hence, aiming to detect these deletions, we developed a rapid and efficient polymerase chain reaction (PCR)–based method [8]. Recently, by delineating the genetic landscape of IKZF1 Δ1-8, we identified COBL (COBL cordon-bleu WH2 repeat protein) gene as an originating hotspot for IKZF1 deletions [9]. Here we established an international collaboration to enable a detailed genetic and clinical characterization of B-ALL with COBL rearrangements (COBL-r).

Material and Methods

Patients

This study included 146 patients diagnosed with B-ALL and IKZF1 deletions. This patient cohort comprised two independent groups. The first one included 133 patients with 1) IKZF1 complete deletion, Δ1-8 (n = 104); or 2) intragenic deletion of IKZF1 including exon 8, Δn-8 (n = 29). The second group referred to 13 patients already diagnosed with IKZF1 deletions and COBL-r at collaborating centers [8], [9], [10]. Patient samples and information were obtained through the cooperative efforts of nine centers worldwide, and the project was conducted in accordance with the declaration of Helsinki.

Detection of IKZF1 Deletions and CNAs

IKZF1 deletions and CNAs in EBF1, JAK2, CDKN2A, CDKN2B, PAX5, ETV6, BTG1, RB1, and the pseudoautosomal region 1-PAR1 (SHOX, CRLF2, C2F2RA, IL3RA, and P2RY8) were determined by either multiplex ligation-dependent probe amplification (MLPA), or single nucleotide polymorphism (SNP) array analyses. The SALSA MLPA P335-(A3-B2) and/or SALSA MLPA P202-B1 (MRC Holland) were used for MLPA experiments, and the data analyses were performed using Coffalyser software. The CytoScan HD Array (Affymetrix) assessed the occurrence of CNAs in the Brazilian (partially), British, and French cohorts. Data were analyzed with Chromosome Analysis Suite software, version 3.2 (Applied Biosystems), and the GRCh38/hg38 build of the Human Genome Assembly.

Characterization of COBL Breakpoints

We developed customized MLPA assays for the screening of CNAs within chromosome 7. The probes’ design and assay conditions have been previously described [9]. To validate our SNP array and MLPA findings, we performed either multiplexed long-distance PCR (M-PCR) or long-distance inverse (LDI)-PCR. These PCR approaches allowed us to confirm COBL-r and to determinate the breakpoints at nucleotide level.

Patient Data and Survival Analyses

For the comparison of laboratory and clinical data between patients with or without COBL-r, we used the Pearson χ2 test. Overall survival (OS) was defined as the time in months from the date of diagnosis to death or to the last follow-up assessment for patients alive. Kaplan-Meier method was used to estimate OS rates of patients according to COBL status with differences compared by the log-rank test. Statistical analyses were performed using R software, version 3.5.2, and P values < .05 were considered statistically significant.

Analysis of Activation-Induced Deaminase (AID) or RAG Recognition Signal Sequences in Recombined COBL Alleles

An agnostic search for motifs located within fragments spanning 50 bp from the breakpoint junctions of COBL-r was performed using MEME. The limit of output motifs was set to 5, and width ranged from 2 to 15 bp. The FIMO (Find Individual Motif Occurrences) tool was used for the analysis of WGCW and CG sequences, and the recombination signal sequence (RSS) consensus sequence was used to search for cryptic RSS sequences [11], [12].

Results

Most of the deletions encompassing IKZF1 exon 8 were classified as whole-gene deletions (Δ1-8: 78% vs. Δn-8: 22%). The CNAs found in the cohort were associated with monosomy 7 (14%; n = 16); isochromosome 7q (6%; n = 7): 7p loss (15%; n = 17), and interstitial deletions within 7p (66%; n = 76) (Figure 1A). Among the Δn-8 group, most of the alterations (93%) were IKZF1 deletions detectable by the M-PCR, i.e., Δ2-8 or Δ4-8 (Figure 1B). Alternatively, 47% of the Δ1-8 group presented with aberrations involving loss of the whole 7p, such as monosomy 7 (18%; n = 19), isochromosome 7q (10%; n = 10), and 7p loss (19%; n = 20). The remaining 55 patients (53%) harbored interstitial deletions; COBL-r were found in 12% of the Δ1-8 cohort (Figure 1C).

Figure 1

Genetic alterations associated with IKZF1 deletions in B-ALL. (A) Patients with IKZF1 deletions spanning its exon 8 presented four types of aberrations on chromosome 7: monosomy 7; isochromosome 7q, i(7q); 7p loss; and interstitial deletions within 7p arm. *One cohort of 17 cases comprising only complete deletions of IKZF1 was excluded from this analysis because they would lead to a biased estimative of CNAs' frequencies among patients with IKZF1 deletions overall. Then, patients were divided into two groups: (B) intragenic deletions of IKZF1 spanning its exon 8 (n = 29) and (C) complete deletions of IKZF1 (n = 104). (D) The COBL rearrangements were identified in 25 patients. Most of the deletions spanned between IKZF1 exon 1 and COBL intron 5, although they presented variable size. (E) These patients had additional alterations, such as ETV6, PAX5, CDKN2A, and CDKN2B deletions. NA, not available. (F) Distribution of molecular subtypes of B-ALL among patients with or without COBL-r. Legend colors illustrate alterations associated with high risk (red), intermediate risk (blue), and good risk (green). HH, high hyperdiploidy. (G) The frequency of deletions in genes associated with B-ALL was compared according to COBL status. PAX5, CDKN2A, CDKN2B, and ETV6 deletions were recurrently observed; however, the frequency of these CNAs was similar for all these genes, as indicated by P values above each bar. PAR1, pseudoautosomal region 1. (H) No significant difference was observed when comparing the overall survival of B-ALL patients with COBL-r vs. COBL wild type (P = .65).

We identified COBL-r in 25 B-ALL cases with IKZF1 deletions (Figure 1D). COBL-r were detected by MLPA screening in 133 patients with B-ALL and IKZF1 deletions (n = 12) or SNP array/NGS investigation performed at collaborating centers (n = 13), as described in the methodology. The loss of genes located within 9p locus — PAX5 (n = 11), CDKN2A (n = 10), and CDKN2B (n = 7) — and ETV6 deletions (n = 10) were the most recurrent additional alterations among these patients (Figure 1E). Demographic and laboratory data for the 25 patients with COBL-r showed they included 17 males and 8 females who were mainly children and adolescents (n = 21), with a median age at diagnosis of 5.5 years (range 1-59 years) and a median white blood cell count of 7.5 × 109/l (range 1.5-459.6 × 109/l) (Table 1). The patients were treated on diverse therapy protocols (Table 2). Seven patients relapsed within a median of 5.5 years (range 1.4-16.3 years), and six of them experienced isolated disease recurrence in the bone marrow (n = 5) or testes (n = 1). One patient relapsed at both sites: bone marrow and central nervous system. The median follow-up was 5.1 years (range 0.3-16.3 years), and the five deaths (mainly in patients with early relapse) occurred at a median 4.2 years (range 0.3-5.5 years) following diagnosis. In addition, the comparison between patients with COBL-r vs. COBL wild-type revealed that both groups presented similar laboratory and clinical characteristics (Table 3). Among the B-ALL cytogenetic abnormalities, TCF3-PBX1 and ETV6-RUNX1 were exclusively found in patients without or with COBL-r, respectively (Figure 1F). Although we did not observe any significant difference in the frequency of additional gene deletions when comparing patients with vs. without COBL-r, it is worthy of note that CDKN2A, ETV6, and PAX5 deletions were more frequent in patients with COBL-r (Figure 1G). Follow-up data were available for 111 B-ALL patients with (n = 20) or without COBL-r (n = 91). The OS of patients with COBL-r was similar to those with IKZF1 deletion only (hazard ratio, 1.278; 95% CI, 0.35-4.68; P = .646) (Figure 1H).

Table 1

Demographic and Laboratory Characteristics of B-ALL Cases with COBL Rearrangements.*

Patient	Age (Years)	Gender	WBC (×109/l)	Blasts at BM	5′ Breakpoint	3′ Breakpoint	Detection Method	Karyotype
P004	1	F	96.000	95%	ELMO1 intron 14	Upstream COBL	SNP array	46,XX,+2,t(3;16)(p21;p13),del(4)(q22),−7,der(12)t(7;12)(q11;p13),−13,+16,−18,−19,+mar1,+mar2
P034	1	M	3.400	98%	RAD50 intron 5	COBL intron 5	LDI-PCR	NA
P086	1	M	5.000	95%	7pter	COBL intron 7	MLPA/LDI-PCR	NA
P120	1	F	459.600	80%	7p14	COBL intron 5	SNP array/M-PCR	NA
P056	2	F	1.600	98%	SPATA intron 1	COBL intron 5	LDI-PCR	46,XX
P122	2	M	13.400	40%	7pter	COBL intron 7	MLPA	NA
P145	2	F	92.000	98%	TCRGC2 intron 1	COBL intron 13	SNP array	47,XX,del(7)(p11),+10,add(12)(p13),del(12)(p11.2)[4]
P022	4	F	7.500	85%	7p12	COBL intron 2-5	MLPA	46,XX
P003	5	M	10.000	95%	IKZF1 intron 3	COBL intron 5	SNP array/M-PCR	46,XY
P074	5	M	16.400	78%	7p12.1	COBL intron 5‡	LDI-PCR	NA
P079	5	F	5.700	88%	7p	COBL intron 8-13	MLPA	NA
P121	5	F	7.470	50%	7p12.2	COBL intron 5	SNP array/M-PCR	NA
P144	5	M	31.000	NA	7p14.2	COBL intron 5	SNP array	47,XY,+X,del(16)(q13),i(17)(q10),ider(21)(q10)dup(21)(q?)[3]/ 47,idem,add(7)(p1)[3]
P002	6	M	115.000	97%	7p12.2	COBL intron 5	SNP array/M-PCR	46,XY,del(6)(q21q25),der(12)del(12)(p11p13)t(12;17)(p13;q11),der(17)t(12;17)
P069	10	F	204.000	89%⁎	IKZF1 intron 1	COBL intron 5	LDI-PCR	NA
P146	11	M	1.500	90%	SPATA intron 2	COBL intron 5	SNP array	48,XY,+X,?t(6;20)(p1;q1),?t(7;9)(p1;p2),i(9)(q10),+12,der(21)dup(21)(q?)[9]
P068	12	M	NA	NA	IKZF1 intron 3	COBL intron 5	RNA-seq/LDI-PCR	48,XY,+X,del(4)(q25),−7,del(9)del(9)t(4;9)(q25;p13),+21,+mar[cp8]/
P104	13	M	1.680	84%	7p14	COBL intron 5	MLPA	46,XY,del(7)(p14),der(9;12)(q10;q10),+mar[6]
P090	15	M	4.500	83%	7p22	COBL intron 5	MLPA	NA
P114	16	M	55.000	45%	7p	COBL intron 2-5	MLPA	NA
P143	16	M	6.200	96%	COBL intron 1†	7p12.1	SNP array	46,XY,idic(9)(p13)[1]/ 45,idem,−7[3]/ 45,idem,add(2)(p25),−7[9]
P147	Adolescent	M	NA	NA	SUN3 intron 5	COBL intron 5	NGS	NA
P005	21	M	2.820	90%	7pter	Upstream COBL	SNP array	46,XY,t(9;22)(q34;q11)
P006	38	M	3.040	67%	VWC2 intron 3	Upstream COBL	SNP array	46,XY
P001	59	M	58.000	75%	7p12	COBL intron 7§	SNP array	45,XY,−7,t(9;22;15)(p24;q11;q21),add(9)(p13)

BM, bone marrow; LDI-PCR, long-distance inverse PCR; M-PCR, multiplex PCR; MLPA, multiplex ligation-dependent probe amplification; NA, not available; WBC, white blood cell count.

In this case, blast percent was estimated from peripheral blood.

Patient had COBL rearrangement and 7p12.3::GRB10 intron 5 deletion.

Patient had an inversion within COBL intron 5.

Patient had COBL rearrangement and monosomy 7.

Table 2

Clinical Characteristics and Outcome of B-ALL Cases with COBL Rearrangements

Patient	Clinical Trial	CNS Disease	MRD Day 33	MRD Day 78	PR	CMR Day 33	Relapse	Outcome	Follow-Up (Months)
P147	AALL0232_1	NA	NA	NA	NA	NA	NA	NA	NA
P068	AALL0331/UKALLR3	No	Negative	Negative	NA	Yes	BM	Dead	45
P034	AIEOP-BFM ALL2000	No	Negative	Negative	Good	Yes	No	1st CR	96
P022	AIEOP-BFM ALL2000	No	Negative	Negative	Good	Yes	No	1st CR	79
P114	AIEOP-BFM ALL2000	No	NA	NA	Good	Yes	NA	Dead	NA
P120	AIEOP-BFM ALL2000	No	Positive	Positive	Poor	No	No	Dead	4
P104	ALL IC 2009	Yes	Negative	Negative	Good	Yes	No	1st CR	38
P145	UKALL2003	No	Negative	Negative	NA	Yes	No	1st CR	105
P146	UKALL2003	No	Negative	NA	NA	Yes	BM, CNS	Dead	66
P144	UKALL97	No	NA	NA	NA	Yes	No	1st CR	26
P143	UKALL97	No	NA	NA	NA	Yes	No	1st CR	167
P003	CAALLF01	No	Negative	NA	Good	Yes	NA	1st CR	4
P079	COALL 05-92	No	NA	NA	NA	Yes	No	1st CR	161
P086	COALL 06-97	No	Negative	NA	NA	Yes	Testes	2nd CR	195
P074	COALL 06-97	No	Negative	NA	NA	Yes	BM	2nd CR	154
P069	COALL 07-03	No	Positive	NA	NA	No	No	1st CR	123
P004	EORTC 58081	No	Negative	Negative	Good	Yes	No	1st CR	49
P090	EsPhALL	No	Positive	NA	Good	Yes	Isolated BM	Dead	56
P006	FRALLE 93	No	Positive	Positive	Poor	Yes	Yes	Relapse	192
P002	FRALLE BT	No	Positive	Negative	Poor	Yes	No	1st CR	15
P122	GBTLIALL99	No	NA	NA	Good	Yes	NA	Alive	NA
P121	GBTLIALL99	No	Negative	Negative	Good	Yes	No	Alive	NA
P005	GRAAPH	No	Positive	Positive	Not evaluable	Yes	No	1st CR	27
P056	NA	No	Positive	NA	Good	Yes	BM	Lost follow-up	17
P001	NA	No	Positive	NA	Good	Yes	No	SCT; alive	9

BM, bone marrow; CNS, central nervous system; CMR, complete morphological remission; MRD, minimal residual disease; NA, not available; PR, prednisone response; SCT, stem-cell transplantation; WBC, white blood cell count. MRD-negative status was defined as <0.01% leukemic cells in bone marrow and peripheral blood.

Table 3

Demographic and Laboratory Data of Patients with B-ALL*

	COBL Rearrangement		P Value⁎
	Non = 121	Yesn = 25
	n (%)	n (%)
Gender			0.473
Male	48 (39.7)	8 (32.0)
Female	73 (60.3)	17 (68.0)
Age at diagnosis			0.701
<1 year	2 (1.7)	0 (0.0)
1-9 years	49 (41.5)	12 (48.0)
≥10 years	67 (56.8)	13 (52.0)
WBC (×109/L)			0.948
<50	85 (70.2)	16 (69.6)
≥50	36 (29.8)	7 (30.4)
NCI risk			0.797
Standard	69 (58.0)	14 (60.9)
High	50 (42.0)	9 (39.1)
CNS disease			0.741
No	96 (94.1)	23 (95.8)
Yes	6 (5.9)	1 (4.2)
Prednisone response			0.143
Good	66 (91.7)	11 (78.6)
Poor	6 (8.3)	3 (21.4)
Relapse			0.607
No	75 (70.8)	13 (65.0)
Yes	31 (29.2)	7 (35.0)
Outcome			0.604
Alive	94 (77.7)	20 (83.3)
Dead	27 (22.3)	4 (16.7)

WBC, white blood cell count; NCI, National Cancer Institute of US; CNS, central nervous system.

Pearson χ2 calculation.

The breakpoints of the COBL-r were determined at nucleotide level in 11 of the 25 cases, and 56% of breakpoints were located within COBL intron 5 (Figure 2B). To address the possible causes of these breakpoints, we first performed an agnostic motif search. This analysis identified the motif CASWGTGG (E-value = 0.87) within all 22 breakpoint sequences of COBL-r (Figure 2A; Supplementary Table S1). CASWGTG is similar to the heptamer of the RAG RSS, which is composed of a heptamer (5′-CACAGTG-3′) and a nonamer (5′-ACAAAAACC-3′) sequence, interspaced by 12 or 23 random nucleotides associated to RAG-type rearrangements. Since the nonamer was absent in our first analysis, we then investigated the presence of complete motifs associated with the occurrence of rearrangements in leukemia: cryptic RSS sequence (RAG-type fusions), WGCW (AID-type fusions), and CG sequences. WGCW and CG sequences were not found; however, we identified cryptic RSS in 5 out of 22 breakpoint regions, although none of them were spanning breakpoints within COBL (Supplementary Table S2). Additionally, we performed a robust analysis for the identification of cryptic RSSs within the whole sequence of COBL (Figure 2, C-D). The results revealed a broad distribution of this motif throughout the gene and no enrichment for the breakpoint cluster located in intron 5 of COBL.

Figure 2

Identification of motifs within the breakpoint sequences. (A) An agnostic motif search using MEME identified the sequence CASWGTGG (E-value = 0.87) among 22 breakpoint sequences. (B) The map of 19 deletion breakpoints (red triangles) within COBL revealed a hotspot located at intron 5. Three breakpoints were detected within a downstream region of 7p12.1. The sequences highlight two breakpoint clusters located at COBL intron 5. The mapped cryptic recombination signal sequences were not statistically significant. The cryptic recombination signal sequences (cRSS) with a spacer of 12-bp (c) and 23-bp (d) were mapped along COBL gene. The highest RIC scores represent cRSS (blue dots) associated with RSS functionality. The gray area highlights the COBL intron 5 and breakpoint cluster regions.

Discussion

In this study, among the Δn-8 group, the majority of the alterations were IKZF1 deletions which had already been detected by the M-PCR method [8]. Since the remaining Δn-8 samples harbored deletions restricted to IKZF1, either they had DNA fusions outside the breakpoint cluster region of IKZF1 or M-PCR failed to detect them. Among the Δ1-8 group, COBL-r were found in 12% of the patients. This result revealed that COBL-r are more frequent among Δ1-8 but rarely related to Δn-8. Although COBL-r were not detected within the Δn-8 group for the cases included in the current proposal, we found these deletions in patients with COBL-r from a previous study [9]. These patients had IKZF1-COBL fusions, which involved IKZF1 intron 1 (n = 1) or intron 3 (n = 2) and COBL intron 5.

Confirming the idea that COBL represents a genomic hotspot for IKZF1 deletions in B-ALL [9], we identified COBL-r in 25 B-ALL cases with IKZF1 deletions. Although either good (ETV6-RUNX1, high hyperdiploid) or high-risk (BCR-ABL1) cytogenetic groups have been observed in some of these patients, most of them had the so-called “B-other” subclassification with either normal or other abnormal (BCR-JAK2, IGH-EPOR) cytogenetic profile. In the current genetic risk stratification, many cases with COBL-r would then be classified as intermediate cytogenetic risk. We found that these patients also presented other secondary abnormalities commonly identified in B-ALL, such as PAX5 and CDKN2A/B and ETV6 deletions. This combination is of special interest because these patients could benefit from the newly proposed combined risk stratification strategies, such as IKZF1plus or the UKALL-CNA classifiers [13], [14], [15]. Considering that patients with B-ALL share similar laboratory and clinical characteristics regardless of COBL involvement, IKZF1 deletions may play a major role on risk stratification for these patients.

Regarding the breakpoints of the COBL-r, it is remarkable that 56% of them were located within COBL intron 5, although this region represents only 16% (47,770 out of 300,587 bp) of the entire gene. Based on this observation, we formulated two hypotheses to explain the presence of a hotspot for breakpoints within COBL: 1) the production of a truncated COBL protein, encoded by exons 1-5 only, could have a role on leukemogenesis, or 2) there is a breakage mechanism involving COBL intron 5, thus enriching this area for gene rearrangements.

COBL protein has three Wiskott-Aldrich syndrome protein homology 2 domains for actin binding. It shows substantial expression in neurons and muscle cells, although levels are low in blood [16]. COBL functions as an actin nucleator, controlling neuronal morphology and development [17]. Considering that COBL does not play a direct role in lymphoid development, the enrichment of COBL-r in B-ALL is more likely to be related to mechanisms controlling DNA breakage and promotion of genetic fusions.

Usually, genetic rearrangements in lymphoid malignancies are caused by either AID or the RAG complex. Therefore, we searched for motifs located within the breakpoint junctions of COBL-r, which could potentially provide a rational explanation for the observed chromosomal rearrangements. RSSs are recognized by RAG enzymes during V(D)J recombination, and previous studies have located cryptic RSS immediately internal to the breakpoints of intragenic deletions of IKZF17. Although the mutual motif within sequences spanning the breakpoints was similar to the heptamer sequence, our results do not support the idea that aberrant RAG-mediated recombination is the mechanism responsible for IKZF1 and COBL codeletions.

In summary, our results highlight COBL as a hotspot for interstitial deletions within chromosome 7, especially for deletions including the IKZF1 gene. Most of the COBL-r arose within COBL intron 5, leading to complete deletion of IKZF1; nevertheless, we also observed fusions between both genes. The analysis of the breakpoint sequences revealed a common motif resembling the heptamer of recombination signal sequence, but the analysis of the whole consensus sequence did not provide evidence for RAG-mediated recombination. Lastly, our study demonstrates that patients with IKZF1 deletions are associated with worse outcome regardless of COBL-r.

The following are the supplementary data related to this article.

Supplementary Table S1

Map of the Agnostically Identified Motif Within Breakpoint Sequences

Supplementary Table S2

Identification of Cryptic Recombination Signal Sequences Spanning Breakpoint Sequences

Funding

This research was funded by CNPq (PQ-2017#305529/2017-0) and FAPERJ-JCNE (E_26/201.539/2014 and E_26/203.214/2017) awarded to M. E. B. A. L. was supported by CAPES (Coordenação de Aperfeiçoamento de Pessoal de Nível Superior) and the Alexander von Humboldt Foundation. G. C. was supported by the Italian Association for Cancer Research (AIRC).

Conflict of Interest

The authors declare no conflict of interest.

Author Contribution

Conception and design: B. A. L., C. M., R. M., M. E. Experiments, data analysis, and interpretation: B. A. L., C. M., T. C. B., C. P. P., N. D., M. B., U. S., C. P., N. C. V., R. M., M. E. Provision of samples, data collection, and assembly: B. A. L., M. B. M., N. D., C. J.H., U. S., M. H., M. S. P. O., G. C., R. S., C. N. A., G. T., S. G., S. B. Writing and/or revision of the manuscript: B. A. L., R. M., M. B. M., C. J. H., G. C., R. S., C. N. A., M. E. Final manuscript approval: all authors.