18780261, 2022, 16, Downloaded from https://febs.onlinelibrary.wiley.com/doi/10.1002/1878-0261.13276 by Readcube (Labtiva Inc.), Wiley Online Library on [23/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Clonal heterogeneity and rates of specific chromosome gains are risk predictors in childhood high-hyperdiploid B-cell acute lymphoblastic leukemia Mireia Ramos-Muntada1, Juan L. Trincado2,3,4, Joan Blanco1, Clara Bueno2,4,5, Virginia C. Rodrı́guez-Cortez2,4, Alex Bataller2,4,6, Belén López-Millán2,4, Claire Schwab7, Margarita Ortega8, Pablo Velasco9, Maria L. Blanco10, Josep Nomdedeu10, Manuel Ramı́rez-Orellana11, Alfredo Minguela12, Jose L. Fuster13, Esther Cuatrecasas14, Mireia Camós15,16,17, Paola Ballerini18, Gabriele Escherich19, Judith Boer20,21, Monique DenBoer20,21,22, Jesús M. Hernández-Rivas23, Maria J. Calasanz24, Giovanni Cazzaniga25, Christine J. Harrison7, Pablo Menéndez2,4,5,26 and Oscar Molina2,4 1 Cell Biology, Physiology and Immunology Department, Genetics of Male Fertility Group, Universitat Autònoma de Barcelona, Spain 2 Department of Biomedicine, School of Medicine, Josep Carreras Leukemia Research Institute, University of Barcelona, Spain 3 CNAG-CRG, Centre for Genomic Regulation (CRG), Barcelona Institute of Science and Technology (BIST), Spain 4 Red Española de Terápias Avanzadas (TERAV), ISCIII, Barcelona, Spain 5 Centro de Investigación Biomédica en Red de Cáncer (CIBER-ONC), ISCIII, Barcelona, Spain 6 Hematology Department, Hospital Clı´nic de Barcelona, IDIBAPS, University of Barcelona, Spain 7 Wolfson Childhood Cancer Research Centre, Newcastle University, UK 8 Hematology Service, Experimental Hematology, Vall d’Hebrón Institute of Oncology (VHIO), Vall d’Hebrón Hospital Universitari, Barcelona, Spain 9 Pediatric Oncology and Hematology Department, Vall d’Hebrón Hospital, Barcelona, Spain 10 Hematology Laboratory, Hospital Sant Pau, Barcelona, Spain 11 Hematology Diagnostic Laboratory, Hospital Niño Jesús, Madrid, Spain 12 Immunology Service, Clinic University Hospital Virgen de la Arrixaca and Instituto Murciano de Investigación Biomédica (IMIB), Murcia, Spain 13 Pediatric Hematology and Oncology Department, Hospital Clı´nico Universitario Virgen de la Arrixaca, Instituto Murciano de Investigación Biosanitaria (IMIB), Murcia, Spain 14 Hematology Laboratory, Institut de Recerca Hospital Sant Joan de Déu, Barcelona, Spain 15 Hematology Laboratory, Hospital Sant Joan de Déu, University of Barcelona, Spain 16 Leukemia and Other Pediatric Hemopathies, Developmental Tumor Biology Group, Institut de Recerca Hospital Sant Joan de Déu Barcelona, Spain 17 Centro de Investigación Biomédica en Red de Enfermedades Raras (CIBERER), Instituto de Salud Carlos III, Madrid, Spain 18 AP-HP, Service d’Hématologie Pédiatrique, Hôpital A. Trousseau, Paris, France 19 Department of Pediatric Hematology and Oncology, University Medical Center Hamburg, Germany 20 Princess Máxima Center for Pediatric Oncology, Utrecht, The Netherlands 21 Oncode Institute, Utrecht, The Netherlands 22 Department of Pediatric Oncology/Hematology, Erasmus MC – Sophia Children’s Hospital, Rotterdam, The Netherlands 23 Departamento de Hematologı´a, Salamanca-IBSAL, Hospital Universitario de Salamanca, Spain 24 CIMA Lab Diagnostics, Universidad de Navarra, Pamplona, Spain 25 Fondazione Tettamanti, Monza, Italy 26 Institució Catalana de Recerca i Estudis Avançats (ICREA), Barcelona, Spain Keywords B-cell acute lymphoblastic leukemia (B-ALL) is the commonest childhood chromosomal gains; clonal heterogeneity; cancer. High hyperdiploidy (HHD) identifies the most frequent cytogenetic computational modeling; high-hyperdiploid subgroup in childhood B-ALL. Although hyperdiploidy represents an B-ALL; risk predictors; sequential iFISH important prognostic factor in childhood B-ALL, the specific chromosome Correspondence gains with prognostic value in HHD-B-ALL remain controversial, and the O. Molina and P. Menéndez, Josep Carreras current knowledge about the hierarchy of chromosome gains, clonal Leukemia Research Institute, School of heterogeneity and chromosomal instability in HHD-B-ALL remains very Medicine, University of Barcelona, Casanova limited. We applied automated sequential-iFISH coupled with single-cell Molecular Oncology 16 (2022) 2899–2919 Ó 2022 The Authors. Molecular Oncology published by John Wiley & Sons Ltd on behalf of 2899 Federation of European Biochemical Societies. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. 18780261, 2022, 16, Downloaded from https://febs.onlinelibrary.wiley.com/doi/10.1002/1878-0261.13276 by Readcube (Labtiva Inc.), Wiley Online Library on [23/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Novel risk predictors in hyperdiploid B-ALL M. Ramos-Muntada et al. 143, 08036 Barcelona, Spain computational modeling to identify the specific chromosomal gains of the E-mail: omolina@carrerasresearch.org (OM); eight typically gained chromosomes in a large cohort of 72 primary diag- E-mail: pmenendez@carrerasresearch.org nostic (DX, n = 62) and matched relapse (REL, n = 10) samples from (PM) HHD-B-ALL patients with either favorable or unfavorable clinical out- Mireia Ramos-Muntada and Juan L. come in order to characterize the clonal heterogeneity, specific chromosome Trincado contributed equally to this work gains and clonal evolution. Our data show a high degree of clonal hetero- geneity and a hierarchical order of chromosome gains in DX samples of (Received 24 January 2022, revised 7 April HHD-B-ALL. The rates of specific chromosome gains and clonal hetero- 2022, accepted 17 June 2022, available geneity found in DX samples differ between HHD-B-ALL patients with online 19 July 2022) favorable or unfavorable clinical outcome. In fact, our comprehensive anal- yses at DX using a computationally defined risk predictor revealed low doi:10.1002/1878-0261.13276 levels of trisomies +18+10 and low levels of clonal heterogeneity as robust relapse risk factors in minimal residual disease (MRD)-negative childhood HHD-B-ALL patients: relapse-free survival beyond 5 years: 22.1% versus 87.9%, P < 0.0001 and 33.3% versus 80%, P < 0.0001, respectively. More- over, longitudinal analysis of matched DX-REL HHD-B-ALL samples revealed distinct patterns of clonal evolution at relapse. Our study offers a reliable prognostic sub-stratification of pediatric MRD-negative HHD-B- ALL patients. ~ 30% of cases, and is typically associated with favor- 1. Introduction able prognostic features [11]. Typically, children with B-cell acute lymphoblastic leukemia (B-ALL) is the HHD-B-ALL achieve negative minimal residual dis- most common cancer in children, characterized by the ease (MRD) after induction treatment and have excel- accumulation of B-cell progenitors in the bone marrow lent cure rates, with 5-year event-free survival (EFS) (BM) [1]. Over the past 20 years, much progress has and overall survival (OS) rates of ~ 75%, and ~ 90%, been made in understanding the biology of the disease respectively [11]. providing significant progress in molecular diagnosis The distribution of chromosomal gains in HHD-B- and risk stratification for treatment, leading to consid- ALL is non-random, with gains of chromosomes X, 4, erable improvements in disease management and clini- 6, 10, 14, 17, 18, and 21 being the most frequent [12]. cal outcome [2–4]. Despite these encouraging Despite chromosomal gains per se being accepted as a advances, the outcome of patients with refractory/re- prognostic factor, specific trisomies have been impli- lapsed (R/R) B-ALL remains dismal, resulting in the cated as better indicators of outcome than MN. The most common cause of death from malignancy in chil- Pediatric Oncology Group showed that trisomies 4 dren [5–7]. and 10 were associated with a good prognosis [13]. Aneuploidy has long been considered a significant Similarly, the Children’s Cancer Group reported that prognostic factor in childhood B-ALL, with early stud- trisomies 10 and 17 conferred a superior outcome in ies suggesting that B-ALL with chromosomal gains HHD-B-ALL [10], while the Children’s Oncology (hyperdiploidy) is associated with a favorable outcome Group and others identified that combined trisomies 4, [8]. Subsequent studies refined the prognostic value of 10, and 17 were the strongest indicator of favorable hyperdiploid B-ALL, demonstrating that patients with outcome in HHD-B-ALL [14–16]. Moreover, Moor- a modal chromosome number (MN) of > 50 (high man et al. showed superior outcomes for HHD-B- hyperdiploidy; HHD) had the most favorable outcome ALL patients with trisomies 4, 10, or 18, with trisomy [9,10]. Indeed, HHD defines the most frequent cytoge- 18, together with patient age, being the strongest inde- netic subgroup in childhood B-ALL, accounting for pendent prognostic indicator [17,18]. Overall, although Abbreviations B-ALL, B-cell acute lymphoblastic leukemia; BM, bone marrow; CIN, chromosome instability; CR, complete remission; DX, diagnostic; EFS, event-free survival; HHD, high-hyperdiploidy; HSPC, hematological stem/progenitor cells; iFISH, interphase fluorescence in situ hybridization; MN, modal number; MRD, minimal residual disease; OS, overall survival; PMC, percentage of major clone; R/R, refractory/relapse; REL, relapse; RFS, relapse-free survival; Seq-iFISH, sequential interphase fluorescence in situ hybridization. 2900 Molecular Oncology 16 (2022) 2899–2919 Ó 2022 The Authors. Molecular Oncology published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies. 18780261, 2022, 16, Downloaded from https://febs.onlinelibrary.wiley.com/doi/10.1002/1878-0261.13276 by Readcube (Labtiva Inc.), Wiley Online Library on [23/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License M. Ramos-Muntada et al. Novel risk predictors in hyperdiploid B-ALL specific trisomies have been associated with survival, and revealed specific chromosomal gains and clonal the exact combination of chromosomal gains with the heterogeneity to be potent predictors of relapse in highest and most reliable predictive value remains HHD-B-ALL patients who were MRD-negative after inconclusive. induction treatment. Cytogenetic analysis of metaphase chromosomes has been the gold standard technique used to assess the 2. Material and methods prognostic value of chromosomal gains in HHD-B- ALL [19,20]. However, it has several limitations ham- pering investigation of the genomic heterogeneity and 2.1. Patient and donor samples complexity of the disease, including: (a) low number of metaphases for analysis, due to failure of blast growth B-ALL diagnosis was based on French-American- ex vivo, (b) biased selection of clones through cell cul- British and World Health Organization classifications. ture, and (c) poor chromosome quality, especially in BM samples (n = 82) from pediatric patients with con- HHD-B-ALL, preventing accurate assessment of chro- firmed HHD-B-ALL were obtained from collaborating mosome number and identity due to chromosome con- hospitals. A number of samples (n = 32; discovery densation defects [21]. Thus, fluorescence in situ cohort) were available for seq-iFISH and computa- hybridization on interphase nuclei (iFISH) in combina- tional modeling (Table 1). Among these samples were tion with conventional cytogenetics has proved to be DX samples (n = 22) obtained at disease presentation useful for the accurate detection of chromosomal from patients who either remained alive in CR, with- gains, as iFISH does not require dividing cells and it out relapses after a minimum of 5-year follow-up (CR, can detect hidden clones and mosaicism within the n = 10), or who relapsed within this timeframe (except samples [22]. Indeed, the use of iFISH by independent one patient who relapsed after 7 years; REL, n = 12). groups has revealed high levels of clonal heterogeneity The remaining samples (n = 10) were relapse samples in HHD-B-ALL samples [23–25], with multiple sub- with matched corresponding DX samples (Table 1, clones containing different combinations of chromoso- Fig. 1). An independent cohort of 50 DX HHD-B- mal gains in individual cells, suggesting chromosomal ALL samples (validation cohort) was used for blind instability (CIN) within HHD-B-ALL. Notably, validation of computational modeling data (Table S1). sequential iFISH analysis (seq-iFISH), which permits Fetal liver (FL) CD34+ hematopoietic/stem progenitor identification of the eight typically gained chromo- cells (HSPC; n = 2) and peripheral blood mononuclear somes in HHD-B-ALL, showed changes in chromo- cells (PBMNC), obtained from a healthy donor, were some number to be hierarchical with sequential used as technical controls. Fetal tissue was collected chromosomal gains, rather than losses, retained from from the Medical Research Council (MRC)/Wellcome lower to higher MNs [25]. Notably, while CIN has Trust Human Developmental Biology Resource with been proposed as an underlying mechanism in HHD- informed consent and approval by our local ethics B-ALL, its association with patient outcome has not committee. been investigated so far. The study was approved by the Barcelona Clı́nic In this study, we sought to investigate the poten- Hospital Institutional-Review Ethics Board (HCB/ tial prognostic impact of clonal heterogeneity in 2014/0687). The study methodologies conformed to HHD-B-ALL. To this end, we investigated the pres- the standards set by the Declaration of Helsinki. The ence of eight typically gained chromosomes in experiments were undertaken with the understanding HHD-B-ALL patients. We used automated seq- and written consent of each subject. iFISH coupled with single-cell computational model- ing in individual cells from a large cohort of pri- 2.2. Sequential fluorescence in situ hybridization mary samples obtained at disease presentation (diagnostic, DX). HHD-B-ALL patients were either Seq-iFISH analyses were performed on cell suspen- in complete remission (CR), disease-free and without sions from HHD-B-ALL BM samples fixed in metha- relapse after a minimum of 5 years after treatment, nol : acetic acid (3 : 1). Three successive hybridization or have relapsed (REL) within this timeframe. In rounds were performed on the same sample to assess addition, we performed longitudinal analyses to the presence of the eight chromosomes typically gained study clonal evolution during disease progression in HHD-B-ALL (Fig. 2A,B). In the first hybridization within matched DX-REL HHD-B-ALL samples. Our round, triple-color FISH was performed with cen- data showed a high degree of clonal heterogeneity tromere enumeration probes (CEP; Abbott Molecular and hierarchical chromosomal gains in DX samples Inc., Des Plaines, IL, USA) for chromosomes 4 (4p11- Molecular Oncology 16 (2022) 2899–2919 Ó 2022 The Authors. Molecular Oncology published by John Wiley & Sons Ltd on behalf of 2901 Federation of European Biochemical Societies. Table 1. Cytogenetic and clinical data of all the childhood HHD-B-ALL samples used for seq-iFISH analyses. DX, Diagnosis; RX: Relapse. CR: patients who remained disease-free after a 2902 minimum of 5 years of follow-up after treatment; REL: patients who relapsed within this timeframe (except REL07, relapse after 7 years). WBC dx Code Event Karyotype Age Gender % blastsa (×109L−1) Follow-upb,c Treatment protocol Relapse Exitus CR01 Dx 58,XY,+X,−Y,+4,+6,+8,+10,+11,+12,+14,+17,+18,+21,+21,+22 3 M 91.6 195 14 COALL03-HR-S No No (SNP6) CR02 Dx 53,XX,+X,+X,+6,+14,+17,+21,+21 (SNP6) 2 F 89.2 9 9.6 COALL03LR-R No No CR03 Dx 55 XXYY,+4,+6,+12,+15,+18,+21 8 M 98.6 - 5.5 SEHOP/PETHEM2013 No No CR04 Dx 54,XX,+X,+6,+8,+14,+17,+18,+21,+21[15]/54,idem,−13,+mar[11]/ 4 F 89 8.2 8.2 ALL10-MR No No 46,XX[19] CR05 Dx 57,XX,+X,+X,+4,+6,+8,+10,+14,+17,+18,+21,+21 (SNP6) 2 F 99 4 13.3 COALL03-LR-R No No CR06 Dx 54,X,+X,Y,+6,+10,+14,+17,+18,+21,+mar[30] 3 M 94.2 33.9 16 PETHEMA LAL96 No No CR07 Dx 55,XXY,+3,+4,+6,+10,+14,+18,+21,+21/46,XY 2.3 M 93.9 25.4 9 SHOP-2005 No No Novel risk predictors in hyperdiploid B-ALL CR08 Dx 60,XY,+X,+4,+6,+7,+8,+9,+14,+17,+18,+21,+21,+3mar/46,XY 6 M 82.4 2.67 11 SHOP-2005 No No CR09 Dx 57,XXY,+4,+6,+8,+10,+10?,+13,+14,+17,+18,+21[14]/46,XY[16] 13 M 90.1 4.2 7 SHOP-2005 No No CR10 Dx 54,XXX,+4,+6,+8,inv(9)(p11q12),+14,+17,+18,+21[22]/46,XX,inv 3 F 95.6 9.7 7 SHOP-2005 No No (9)(p11q12)[2] Mean  SEM 4.6  1.1 6/4 31.4  19.7 10.1  1.1 REL01 Dx 59,XY,+?X,+4,+6,+7,+8,+12,+14,+?add(17) 6 M 80.6 4.2 4.4 UKALL2003-SR Yes Yes (p1),+18,+18,+21,+der(?)t(?;5)(?;q?13),+mar[6]/63,idem,dup(1) (q2q3),+22,inc[7]/46,XY[5] Rx 55-57,X,−Y,+X,+8,+?10,+12,+14,+?add(17)(p1),+18,+21,+der(?)t (?;5)(?;q?13),inc[cp7]/46,XY[25] REL02 Dx 51,XX,+X,+8,+14,+21,+21[5]/46,XX[9] 2 F 95.9 119 3.6 UKALL2003-B Yes No Rx 51,XX,+X,+8,del(11)(q2?2),+14,+21,+21[12]/46,XX[8] REL03 Dx 54,XY,+X,+4,+6,+10,+14,+add(19)(p13),+20,+21[cp8] 13 M 98 1.2 4.2 UKALL2003-B Yes No Rx 54 ~ 55,XY,+X,+4,+5,+6,+10,+14,add(19) (p13),? + 20,+21,+21,+mar[cp9]/46,XY[1] REL04 Dx 46,XX[20]/High hyperdiploid by FISH 13 F 72.9 4.1 2.6 UKALL2003-A Yes Yes Rx 55,XX,+X,+1,+4,+8,+14,+15,+21,+21,+mar[cp4]/46,XX[16] REL05 Dx 52-56,XY,+X,dup(1) 8 M 96.1 47 3.5 UKALL2003-A Yes Yes (q25q44),+4,+8,+10,+11,+14,+18,+21,+21,+mar[cp7]/46,XY[3] Rx High hyperdiploid by FISH REL06 Dx 46,XY[20]/High hyperdiploid by FISH 8 M 50.2 7.8 4.2 UKALL2003-A Yes Yes Rx High hyperdiploid by FISH REL07 Dx 56,XY,+X,ins(1;?), 2 M 86.5 55 7.2 UKALL2003-C Yes No (q21.3;?),+4,+6,+14,+17,+18,+21,+21,+22,+mar[5]/46,XY[5] Rx 46,XY[30]/High hyperdiploid by FISH REL08 Dx High hyperdiploid by FISH 8 M 79.3 44.6 4.6 UKALL2003-A Yes No Rx 52,XY,+X,+Y,+14,+21,+21[1]/46,XY[9] REL09 Dx 51,XX,+X,+6,+14,+21,+21[16]/51,idem,add(9)(q34)[2]/46,XX[2] 3 F 77.2 31.7 3.3 UKALL2003-A Yes Yes Rx 52,XX,+X,+6,+14,+17,+21,+21[13]/46,XX[3] Molecular Oncology 16 (2022) 2899–2919 Ó 2022 The Authors. Molecular Oncology published by John Wiley & Sons Ltd on behalf of M. Ramos-Muntada et al. Federation of European Biochemical Societies. 18780261, 2022, 16, Downloaded from https://febs.onlinelibrary.wiley.com/doi/10.1002/1878-0261.13276 by Readcube (Labtiva Inc.), Wiley Online Library on [23/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 18780261, 2022, 16, Downloaded from https://febs.onlinelibrary.wiley.com/doi/10.1002/1878-0261.13276 by Readcube (Labtiva Inc.), Wiley Online Library on [23/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License M. Ramos-Muntada et al. Novel risk predictors in hyperdiploid B-ALL q11 alpha-satellite DNA; Spectrum Green), 6 (D6Z1; Exitus Yes Yes No Spectrum Aqua), and 10 (10p11.1–q11.1 alpha-satellite DNA; Spectrum Orange). In the second round, triple- Relapse color FISH was performed with CEP probes for chro- Yes Yes Yes mosomes X (DXZ1; Spectrum Green), 17 (D17Z1; Spectrum Orange) and 18 (D18Z1; Spectrum Aqua). In Treatment protocol the third round, dual-color FISH was performed with PETHEMA LAL 96 PETHEMA LAL 96 LAL-SHOP-2005 a locus-specific probe for chromosome 21 (21q21.1, 5-Fluorescein; Empire Genomics) and a subtelomeric probe for chromosome 14 (D14S1420; Spectrum Orange; Abbott Molecular Inc.). Seq-iFISH was per- formed following standard procedures [26,27]. In brief, slides were dehydrated in an ascending ethanol series Follow-upb,c 3.5  0.5 and denatured in 70% formamide/2 × SSC at 73 °C for 2 min. DNA probes mixed in hybridization buffer 0.5 (Abbott Molecular Inc.) were denatured at 73 °C for 2 2 5 min before hybridization in a humid chamber at (×109L−1) 32.6  9.9 WBC dx 37 °C overnight. Slides were then washed twice in 0.4 × SSC containing 0.03% NP40 at 73 °C and then 43.9 0.54 in 2 × SSC at room temperature for 2 min. Slides were mounted with 40 ,6-diamino-2-phenylindole (DAPI) % blastsa solution for DNA counterstaining. DNA probes were 71.1 34.7 65.5 removed between hybridization rounds by washing the slides in 0.0625 × SSC and incubating at 73 °C for Gender 5 min. 8/5 M M F 6.5  1.1 2.3. Fluorescence microscopy and nuclei re- Age localization 2 5 5 Automated FISH evaluation was performed using an 59,XXY,+Y,der(1)(q?),+4,+5,+6,+8,+10,+11,+18,+18,+21,+22,mar Olympus BX-61 epifluorescence microscope coupled to SPOT AX software (Applied Imaging, Newcastle, UK), and equipped with a BX-UCB motorized stage (Olym- pus), a 60× objective and narrow band-pass fluores- 53,XX,+4,+6,+14,+17,+18,+2mar[36%]/46,XX[64%] cence filters specific for DAPI, Aqua, fluorescein 57,XY,der(1)(p?q?),+4,10mar[56%]/46,XY[44%] 53,XX,+4,+6,+14,+17,+18,+2mar[8]/46,XX[22] isothiocyanate (FITC), and cyanine 3 (Cy3). After the first hybridization round, slides were scanned using nuclei recognition morphometric parameters and FISH Karyotype Indicates follow-up in CR for non-relapsing patients. signal capture [28]. Merged images (overlay of DAPI with fluorescence signals) and the exact coordinates of Indicates time-to-relapse for relapsing patients. a minimum of 500 nuclei per sample were recorded and evaluated. After the second and third hybridiza- tion rounds, slides were evaluated by semi-automated As observed by FISH (% HHD cells). re-localization of previously recorded cells, overall [9]/46,XY[41] Mean  SEM allowing the assessment of the ploidy status of the eight chromosomes at a single-cell level. Cells with no hybridization signals in one or more channels were Table 1. (Continued). excluded from further study. Analysis of all nuclei was performed using strict criteria relating to intensity, Event size, and distribution of FISH signals, as previously Dx Dx Dx Rx described by our group [28]. Only nuclei with at least REL10 REL11 REL12 one chromosome gain (hyperdiploid blasts) were used Code for further analysis. A minimum of 200 hyperdiploid b a c Molecular Oncology 16 (2022) 2899–2919 Ó 2022 The Authors. Molecular Oncology published by John Wiley & Sons Ltd on behalf of 2903 Federation of European Biochemical Societies. 18780261, 2022, 16, Downloaded from https://febs.onlinelibrary.wiley.com/doi/10.1002/1878-0261.13276 by Readcube (Labtiva Inc.), Wiley Online Library on [23/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Novel risk predictors in hyperdiploid B-ALL M. Ramos-Muntada et al. Fig. 1. Clinical outcome and biological features of the high-hyperdiploid B-cell acute lymphoblastic leukemia (HHD-B-ALL) patients analyzed by seq-iFISH. (A) Time of follow-up after diagnosis (DX) of the HHD-B-ALL patients analyzed by seq-iFISH. Complete remission (CR) denotes those patients who remained disease-free after a minimum of 5 years of follow-up after treatment denoted as a dashed red line (n = 10). Relapse (REL) denotes those patients who relapsed within this timeframe (except one patient who relapsed after 7 years; n = 12). Light and dark gray bars represent follow-up in CR and after relapse, respectively. Arrow heads depict the time of relapse. † denotes patient’s death. (B,C) age (B) and WBC (C) of CR and REL patients. Bars represent the mean values of each group and the error bars represent the standard error of the mean (SEM). Each dot represents an individual patient. Two-sided unpaired t-test. nuclei with informative data for all hybridization chromosomes analyzed were used as subclone identi- rounds were analyzed. Cut-off levels for FISH signal fiers (Subclone ID) for each individual cell. Entropy positivity were determined by Binomial distribution in values were calculated for each DX sample using sub- the three diploid control samples described above clone formulas, as previously reported [29]. Hierarchi- using GRAPHPAD PRISM (San Diego, CA, USA). In cal clustering of both cell clones and individual patient samples, aneuploidy combinations were consid- chromosomes was applied to infer the ordering of ered relevant if each aneusomy level proved to be chromosomal gains throughout leukemia evolution, beyond the assigned cut-off value. based on the frequency of specific chromosomal gains. The Euclidean method was used to compute distance matrices and the agglomeration method was applied to 2.4. Single-cell computational analyses assess complete linkage [30]. Shannon entropy indices were obtained for each DX For modeling the chromosomal gains in DX sam- HHD-B-ALL sample to assess the differences between ples from CR and REL HHD-B-ALL patients, a Ran- CR and REL patients using the formula dom Forest decision tree-based approach was used to HðP1 . . . Pn Þ ¼ ∑ni¼1 Pi log2 Pi , where Pi is the probabil- model the relationship of chromosome gains and ity of value and n is the number of possible values. response to therapy [31]. Gini importance was used to The number of gains for each of the eight assess which chromosomes were the most relevant for 2904 Molecular Oncology 16 (2022) 2899–2919 Ó 2022 The Authors. Molecular Oncology published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies. 18780261, 2022, 16, Downloaded from https://febs.onlinelibrary.wiley.com/doi/10.1002/1878-0261.13276 by Readcube (Labtiva Inc.), Wiley Online Library on [23/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License M. Ramos-Muntada et al. Novel risk predictors in hyperdiploid B-ALL predicting response to therapy [32]. Gini importance clonal heterogeneity together with other clinically rele- measures feature the importance across all the sub- vant variables including MRD status after induction trees that the Random Forest algorithm generates on treatment, treatment protocol, gender, age and white a leave-one-out cross-validation strategy (for each blood cell count (WBC) at presentation, were per- sample, we used all the remaining samples for training formed using the Cox model [33]. and the selected instance as test). 3. Results 2.5. Stress tests for prognostic predictors We performed a stress test for each of the three mod- 3.1. High degree of clonal heterogeneity in HHD- els derived by Random Forest analysis. Random devi- B-ALL at DX ates from a uniform distribution between 0 and each noise level are randomly added or subtracted from the Retrospective DX samples from patients with HHD-B- original input data 100 times and predicted again their ALL were sub-grouped according to their clinical out- CR or REL condition. The real conditions were com- come as (a) patients (n = 10) who remained disease- pared with the predictions on the noisy data. The free without relapse after > 5-year follow-up, or (b) robustness of each model was evaluated by assessing patients (n = 12) who relapsed (REL) within this time- the average number of patients correctly or erro- frame (average time to relapse of 3.5 years; range: 2– neously predicted along the 100 repetitions and after 7 years). Seven out of these 12 patients (60%) died increasing noise levels. after the first or second relapse (Fig. 1A and Table 1). No differences in gender distribution, age and WBC were observed between CR and REL HHD-ALL 2.6. Validation analyses patients (Fig. 1B,C and Table 1). The prognostic predictor value obtained by computa- To assess clonal heterogeneity in the HHD-B-ALL tional modeling of seq-iFISH data was validated in DX samples, we performed seq-iFISH for single-cell a blind analysis using an independent cohort of 50 analysis of the eight chromosomes typically gained in BM samples obtained at disease presentation from HHD-B-ALL (4, 6, 10, 14, 17, 18, 21, and X, Fig. 2A, the corresponding number of patients with HHD-B- B). To first establish the cut-off levels for chromosome ALL. Validation analysis was performed by three- gains and losses, we optimized the seq-iFISH analysis color FISH using CEP10 (Spectrum Orange), CEP18 on euploid CD34+ HSPCs (n = 2) and PBMNCs, (Spectrum Aqua), and the LSI21 (Spectrum Green) which showed consistent rates of chromosome gains probes, following standard FISH protocols. Nuclei < 0.37% and chromosome losses < 5% (Fig. 2C–E with no gains for any probed chromosome were con- and Table S2). As expected, diploid clones were sidered normal/healthy hematopoietic cells. A mini- observed in ~ 90% of the cells analyzed (range 88– mum of 200 hyperdiploid nuclei were analyzed per 90.6%) with only minor (range: 0.38–5.2%) subclones sample. showing mainly ‘false’ chromosomal losses (Fig. 2E,F). Remarkably, the order of hybridization steps and the ploidy status of the analyzed chromosomes did not 2.7. Statistical analyses affect the read-out accuracy of our experimental design R-statistics version 4.0.0 (R Foundation for Statistical (Fig. S1). These results validate the high reliability of Computing, Vienna, Austria) was used to perform all our seq-iFISH analysis of clonal composition in single-cell computational analyses, stress tests and mul- HHD-B-ALL. Seq-iFISH analysis of the HHD-B-ALL tivariate analyses. Patients that remained disease-free samples at DX showed high levels of clonal hetero- after a minimum of 5-year follow-up were compared geneity in relation to chromosome copy-number alter- with those that relapsed within this timeframe. For ations, as previously observed [25], with major analysis of clonal evolution, DX-REL patient-matched subclones representing between 2.5% and 30.3% of longitudinal HHD-ALL samples were compared. cells (Fig. 2G and Table 2). Notably, whereas the P < 0.05 was considered statistically significant. major subclone observed by seq-iFISH corresponded Relapse-free survival (RFS) was calculated from the to that observed by conventional karyotyping analysis date of DX to the date of either REL or death, and in 7 of 22 HHD-B-ALL cases (32%), it was different was estimated with the Kaplan–Meier method and in most cases (15/22, 68%; Table 2). These results compared with the log-rank test. Univariate and multi- revealed high clonal ‘chromosomal’ heterogeneity in variate analyses using specific chromosomal gains and HHD-B-ALL DX samples. Molecular Oncology 16 (2022) 2899–2919 Ó 2022 The Authors. Molecular Oncology published by John Wiley & Sons Ltd on behalf of 2905 Federation of European Biochemical Societies. 18780261, 2022, 16, Downloaded from https://febs.onlinelibrary.wiley.com/doi/10.1002/1878-0261.13276 by Readcube (Labtiva Inc.), Wiley Online Library on [23/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Federation of European Biochemical Societies. M. Ramos-Muntada et al. Molecular Oncology 16 (2022) 2899–2919 Ó 2022 The Authors. Molecular Oncology published by John Wiley & Sons Ltd on behalf of Novel risk predictors in hyperdiploid B-ALL 2906 18780261, 2022, 16, Downloaded from https://febs.onlinelibrary.wiley.com/doi/10.1002/1878-0261.13276 by Readcube (Labtiva Inc.), Wiley Online Library on [23/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License M. Ramos-Muntada et al. Novel risk predictors in hyperdiploid B-ALL Fig. 2. Reliability of the seq-iFISH analyses employed for detection of clonal heterogeneity in high-hyperdiploid B-cell acute lymphoblastic leukemia (HHD-B-ALL). (A) Scheme depicting the seq-iFISH stepwise analysis. (B) Representative images of consecutive FISH rounds using the indicated chromosomes in the same cell. Top and bottom panels show interphase and metaphase cells, respectively. Scale bar = 10 μ m. (C) Karyotypes of the indicated control (ctrl) samples; n = 20 metaphases per sample. (D) Frequency of chromosomal gains (top) and losses (bottom) observed for each chromosome by seq-iFISH analysis in control samples. (E) Modal number (MN) of chromosomal gains/ losses in control samples. Graphs represent the mean value of three independent experiments and error bars represent the SEM; n = 200 nuclei were analyzed per experiment. (F) Single-cell analysis showing the size of diploid and ‘false aneuploid’ clones observed in the control samples. Subclone formulas/codes on the left indicate the number of gains (1 or 2) or losses (−1) for the chromosomes X, 4, 6, 10, 14, 17, 18, and 21, respectively; n = 150 nuclei, 265 nuclei and 191 nuclei for controls 1, 2, and 3, respectively. (G) Single-cell analysis showing the 10 major clones observed at diagnosis (DX) in all HHD-B-ALL patients; n = a minimum of 200 nuclei per sample. The size of the black circles in (F and G) represents the proportion of cells with the indicated subclone in each sample. Table 2. Major subclones identified by conventional cytogenetics and seq-iFISH analyses in HHD-B-ALL patients. FISH subclones depict the number of gains observed for chromosomes X, 4, 6, 10, 14, 17, 18, and 21, respectively. Chromosomes identified by seq-iFISH are labeled in bold. Cases with karyotypes representing the major clone observed by Seq-iFISH were labeled in blue. Seq-iFISH HHD-B-ALL Karyotype Major clone % CR01 58,XY,+X,−Y,+4,+6,+8,+10,+11,+12,+14,+17,+18,+21,+21,+22 (SNP6) 01110111 4.74 CR02 53,XX,+X,+X,+6,+14,+17,+21,+21 (SNP6) 10001102 7.95 CR03 55,XXYY,+X,+Y,+4,+6,+12,+15,+18,+21 11111111 7.12 CR04 54,XX,+X,+6,+8,+14,+17,+18,+21,+21[15] / 54,idem,−13,+mar[11]/46,XX[19] 10101102 7.39 CR05 57,XX,+X,+X,+4,+6,+8,+10,+14,+17,+18,+21,+21 (SNP6) 21111111 10.85 CR06 54,X,+X,Y,+6,+10,+14,+17,+18,+21,+mar[30] 10111111 12.95a CR07 55,XXY,+X,+3,+4,+6,+10,+14,+18,+21,+21/46,XY 11111012 8a CR08 60,XY,+X,+4,+6,+7,+8,+9,+14,+17,+18,+21,+21,+3mar/46,XY 11112122 3.96 CR09 57,XXY,+X,+4,+6,+8,+10,+10?,+13,+14,+17,+18,+21[14]/46,XY[16] 11121111 2.55a CR10 54,XXX,+X,+4,+6,+8,inv(9)(p11q12),+14,+17,+18,+21[22]/46,XX,inv(9)(p11q12)[2] 11101111 7.72a REL01 59,XY,+?X,+4,+6,+7,+8,+12,+14,+?add(17)(p1),+18,+18,+21,+der(?)t(?;5)(?;q?13),+mar[6]/63,idem, 11121121 3.61 dup(1)(q2q3),+22,inc[7]/46,XY[5] REL02 51,XX,+X,+8,+14,+21,+21[5]/46,XX[9] 10001002 30.32a REL03 54,XY,+X,+4,+6,+10,+14,+add(19)(p13),+20,+21[cp8] 11111102 10.36 REL04 46,XX[20]/High hyperdiploid by FISH 01112002 22.3 REL05 52-56,XY,+X,dup(1)(q25q44),+4,+8,+10,+11,+14,+18,+21,+21,+mar[cp7]/46,XY[3] 11011002 9.09 REL06 46,XY[20]/High hyperdiploid by FISH 11111111 5.83 REL07 56,XY,+X,ins(1;?),(q21.3;?),+4,+6,+14,+17,+18,+21,+21,+22,+mar[5]/46,XY[5] 11101112 12a REL08 High hyperdiploid by FISH 10001002 11.57 REL09 51,XX,+X,+6,+14,+21,+21[16]/51,idem,add(9)(q34)[2]/46,XX[2] 10101002 10.85a REL10 53,XX,+4,+6,+14,+17,+18,+2mar[8]/46,XX[22] 11101111 7.86 REL11 59,XXY,+X,der(1)(q?),+4,+5,+6,+8,+10,+11,+18,+18,+21,+22,mar[9]/46,XY[41] 12120022 8.45 REL12 57,XY,der(1)(p?q?),+4,10mar[56%]/46,XY[44%] 11121012 10.61 a Major clone identified by seq-iFISH corresponds with the clone detected by conventional karyotyping. between CR and REL HHD-B-ALL patients. 3.2. Differential rates of specific chromosome Increased rates of gains of chromosomes 17 and 18, gains and clonal heterogeneity at DX between and moderate increased rates of gains of chromosomes CR and REL HHD-B-ALL patients 6 and 10, were found in DX samples from CR HHD- Seq-iFISH analysis performed on DX samples revealed B-ALL patients. In contrast to the increased rates of no differences in the number of total chromosomal chromosomal gains in CR HHD-B-ALL patients, gains observed in CR versus REL HHD-B-ALL moderately higher rates of chromosome 21 gains were patients, with a peak between four and six chromoso- found in REL HHD-B-ALL patients (Fig. 3B,C). mal gains in both groups (Fig. 3A). However, different These chromosomal gains were mainly trisomies of all rates of specific chromosome gains were observed chromosomes analyzed, except for chromosome 21, Molecular Oncology 16 (2022) 2899–2919 Ó 2022 The Authors. Molecular Oncology published by John Wiley & Sons Ltd on behalf of 2907 Federation of European Biochemical Societies. 18780261, 2022, 16, Downloaded from https://febs.onlinelibrary.wiley.com/doi/10.1002/1878-0261.13276 by Readcube (Labtiva Inc.), Wiley Online Library on [23/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Novel risk predictors in hyperdiploid B-ALL M. Ramos-Muntada et al. Fig. 3. Differential rates of specific chromosome gains and clonal heterogeneity in diagnostic (DX) samples from complete remission (CR) and relapsed (REL) high-hyperdiploid B-cell acute lymphoblastic leukemia (HHD-B-ALL patients). (A) Number of total chromosomal gains in DX samples from CR and REL HHD-B-ALL patients. Graphs represent the mean value and error bars represent the SEM. (B) Frequency of chromosomal gains for the indicated chromosomes in DX samples from CR and REL HHD-B-ALL patients. Graphs represent the median val- ues of total gains for each chromosome and dots represent the values obtained for individual patients. (C) Chromosomal gains as observed by density values in single-cell computational analysis for the indicated chromosomes distinguishing the contribution of trisomies and tetrasomies. Red boxes indicate the differences observed in DX samples from CR and REL HHD-B-ALL patients. (D) Box plots comparing the frequency of cells harboring triple trisomies 4, 10, and 17 (left) and single trisomy 18 (right) between DX samples from CR and REL HHD-B-ALL patients. (E) Box plots comparing the clonal heterogeneity as observed by Shannon entropy values of data between controls (ctrl) and DX samples from CR and REL HHD-B-ALL patients. Boxes represent the quartiles 25–75 and horizontal lines represent the mean value. Error bars represent the standard deviation (SD); n = 3 ctrl, n = 10 CR, and n = 12 REL HHD B-ALL patients; two-sided unpaired t-test. *P-value < 0.05. 2908 Molecular Oncology 16 (2022) 2899–2919 Ó 2022 The Authors. Molecular Oncology published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies. 18780261, 2022, 16, Downloaded from https://febs.onlinelibrary.wiley.com/doi/10.1002/1878-0261.13276 by Readcube (Labtiva Inc.), Wiley Online Library on [23/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License M. Ramos-Muntada et al. Novel risk predictors in hyperdiploid B-ALL which showed similar trisomy rates between groups REL HHD-B-ALL patients (Fig. 4A). Three major and higher tetrasomy rates in REL HHD-B-ALL sam- clusters were distinguished in DX samples from CR ples (Fig. 3C). HHD-B-ALL patients at a Euclidean distance of 48.3. Specific trisomies, such as triple trisomies for chro- Chromosome 21 lies at the base of the tree, followed mosomes 4, 10, 17, and single trisomy for chromosome by chromosomes 14 and X that cluster separately from 18 have been reported to have prognostic value in the remaining chromosomes which, in turn, also form HHD-ALL based on conventional cytogenetics analy- smaller sub-clusters (Fig. 4A,B). Two major clusters sis [14,17,34]. We next analyzed the percentage of DX were observed in DX samples from REL HHD-B- blasts bearing either trisomies 4, 10, 17, or trisomy 18 ALL patients with chromosomes 21 and 14 at the base in our cohort, irrespective of other chromosomal of the tree, separated from the remaining chromo- gains. In line with previous studies, we found that the somes by a Euclidean distance of 11.6 (Fig. 4A,B). frequency of DX blasts bearing either trisomies 4, 10, Hierarchical cluster analyses in individual samples 17 (P > 0.05), or trisomy +18 (P = 0.025) was higher showed variability between individuals but with a in CR than in REL HHD-B-ALL samples (Fig. 3D), marked tendency for chromosomes 21, 14 or X to suggesting that at least trisomy 18 may have prognos- cluster together at the base of the tree (Figs S2 and tic value for HHD-B-ALL. Beyond specific chromoso- S3). In addition, chromosomes 4 and 6 clustered mal gains, the levels of CIN have been associated with together as a unique sub-cluster separated at higher tumor progression in different types of cancer [35–39]. Euclidean distances from the others, suggesting that To assess whether CIN levels are associated with these chromosomes were gained later during leukemo- HHD-B-ALL outcome, we next analyzed the levels of genesis both in CR and REL HHD-B-ALL. Our clonal heterogeneity in our samples, which is directly results confirm a hierarchical ordering in chromosomal associated with CIN [40], based on the analysis of the gains in HHD-B-ALL, with chromosomes 21, 14, and entropy values obtained by computational analysis of X gained earlier and chromosomes 4 and 6 later dur- single-cell data [29]. We found significantly higher ing leukemogenesis. levels of clonal heterogeneity in DX samples from CR To assess whether the hierarchical order of chromo- than from REL HHD-B-ALL patients, as revealed by somal gains results in the formation of specific clones increased entropy values (Fig. 3E), suggesting that clo- that can be associated with REL in HHD-B-ALL, we nal heterogeneity may also represent a favorable prog- next performed a hierarchical cluster analysis of those nostic factor in HHD-B-ALL at DX. Collectively, our unique subclones observed at DX in each analyzed data show that HHD-B-ALL patients show a high and sample. Results highlighted an extensive clonal hetero- variable degree of clonal heterogeneity with differential geneity in HHD-B-ALL DX samples with no specific rates of specific chromosomal gains between CR and clones (clustering) associated with REL (Fig. 4C). REL HHD-B-ALL patients. Notably, the higher levels of clonal heterogeneity observed in DX samples from 3.4. Computational modeling identifies trisomies CR HHD-B-ALL suggest its potential as a biomarker 18 and 10 as a relapse predictor in MRD-negative for HHD-B-ALL outcome, suggesting that one of the HHD-B-ALL subclones present at DX emerges as the dominant clone at REL. We next examined whether specific combinations of chromosomal gains in DX samples could distinguish MRD-negative HHD-B-ALL patients with favorable 3.3. HHD-B-ALL shows hierarchical chromosome (CR) versus unfavorable (REL) clinical outcome. We gains without specific clones associated with employed a machine learning approach, which ranks REL the relative contribution of each chromosome analyzed A sequential and ordered acquisition of chromosomal in distinguishing between CR and REL HHD-B-ALL gains has been reported in HHD-B-ALL [25,41]. To [32]. This analysis highlighted trisomies 18, X, 17, 21, study the potential order of chromosomal gains in DX and 10 as potential informative parameters classifying samples from both CR and REL HHD-B-ALL samples as CR or REL (Fig. 5A). Using these data, patients, we next performed a hierarchical cluster anal- we next used decision trees to test which combinations ysis of the acquisition of chromosomes using the num- of these five chromosomal gains best classified HHD- ber of overall gains in all the cells analyzed for each B-ALL patients. Among all possible permutations, group at DX. Such hierarchical cluster analysis only three displayed significant predictor potential: revealed specific associations regarding chromosome chromosomes 21 and 14, chromosomes 21 and 10, and gains that did not substantially differ between CR and chromosomes 18 and 10 (Fig. 5B). The reliability of Molecular Oncology 16 (2022) 2899–2919 Ó 2022 The Authors. Molecular Oncology published by John Wiley & Sons Ltd on behalf of 2909 Federation of European Biochemical Societies. 18780261, 2022, 16, Downloaded from https://febs.onlinelibrary.wiley.com/doi/10.1002/1878-0261.13276 by Readcube (Labtiva Inc.), Wiley Online Library on [23/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Novel risk predictors in hyperdiploid B-ALL M. Ramos-Muntada et al. Fig. 4. High-hyperdiploid B-cell acute lymphoblastic leukemia (HHD-B-ALL) shows hierarchical chromosome gains without specific clones associated with relapse. (A) Heatmaps depicting the acquisition of chromosomal gains in all the cells analyzed in diagnostic (DX) samples from complete remission (CR; left; n = 2983 cells) and relapse (REL; right; n = 3142 cells) HHD-B-ALL patients who had achieved minimal residual disease (MRD)-negativity post-induction therapy. Lines represent individual clones and columns represent individual chromosomes. The number (value) of chromosomal gains per clone is represented as different red color intensities. (B) Summarized dendrograms using data from (A). Euclidean distances are shown for each cluster identified. (C) Heatmap depicting the unique subclones in control (Ctrl), CR and REL HHD-B-ALL samples. The color code represents the row Z-score for individual clones. these combinations of chromosomal gains found in (18 of 22 cases were correctly classified; Fig. 5B). We DX samples in predicting favorable or unfavorable thus established a risk predictor for HHD-B-ALL clinical evolution of the HHD-B-ALL patients was patients based on a two-step classification using chro- confirmed by ‘stressing’ the data through a gradual mosomes 18 and 10 trisomy rates. Patients were first increase in the percentage of noise [31,32]. Stress tests classified by the percentage of chromosome 18 gains, allowed assessment of the robustness of each combina- with a threshold of 40%. Cases below this threshold tion of chromosomal gains in relation to biological were accurately classified as unfavorable-risk (REL) and technical variability. These analyses revealed that patients. Next, cases above this threshold were further the predictor-3, based on trisomies 18 and 10, was the refined based on the percentage of chromosome 10 most stable risk predictor associated with a lower rate gains, with a threshold of 40%. Those HHD-B-ALL of misclassification events, a stable reproducibility, patients below this second threshold were also classi- even at high noise levels, and with an accuracy of 82% fied as unfavorable risk and cases above were regarded 2910 Molecular Oncology 16 (2022) 2899–2919 Ó 2022 The Authors. Molecular Oncology published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies. 18780261, 2022, 16, Downloaded from https://febs.onlinelibrary.wiley.com/doi/10.1002/1878-0261.13276 by Readcube (Labtiva Inc.), Wiley Online Library on [23/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License M. Ramos-Muntada et al. Novel risk predictors in hyperdiploid B-ALL as favorable prognosis (Fig. 5C). This computational RFS of 87.9% and 22.1%, respectively (P < 0.0001; modeling-based risk predictor correctly classified in Fig. 5F). DX samples HHD-B-ALL patients with favorable and Using seq-iFISH and computational modeling we unfavorable outcome with a 70% and 100% accuracy, have established that HHD-B-ALL DX samples with respectively. smaller relative representation (percentage) of the major clone (PMC) display higher entropy, indicative of higher clonal heterogeneity (Table 2 and Fig. 3E). 3.5. An independent cohort of DX samples The levels of clonal heterogeneity analyzed in DX validates the high reliability of low levels of samples were, in fact, higher in CR than REL HHD- trisomy +18+10 and low levels of clonal B-ALL patients, suggesting that the levels of clonal heterogeneity as relapse predictors in MRD- heterogeneity may also predict the clinical evolution of negative HHD-B-ALL HHD-B-ALL patients. To assess the prognostic value To further validate our computationally defined risk of clonal heterogeneity, we used the PMC obtained in predictor in relation to clinical outcome, we performed a our three-color iFISH analyses. To obtain a cut-off blind validation test using a larger independent cohort value for the PMC, we first dichotomized this variable of HHD-B-ALL DX samples from 50 patients (48/50 using maximally-selected rank statistics, finding that (96%) MRD-negative after induction therapy) selected the best threshold for PMC was 50%. Our results to include both favorable (remain disease-free after revealed, within our cohort, a significantly higher RFS 5 years) and unfavorable (relapsed within 5 years) cases in patients with PMC ≤ 50% (RFS beyond 5 years: (66% versus 34%, respectively; Table S1). For this vali- 79.8% versus 33.3%, P = 0.0001, Fig. 5G), thus dation, we analyzed the aneuploidy rates for chromo- demonstrating that lower levels of clonal heterogeneity somes 18, 10, and 21 by three-color iFISH. represent a risk factor associated with REL HHD-B- Chromosome 21 was used as a control to discriminate ALL. Importantly, the rates of trisomies 18 and 10 hyperdiploid leukemic cells from normal hematopoietic were associated with the PMC, as revealed by the neg- cells disomic for chromosomes 18 and 10. Consistent ative correlation between both variables for each with previous data from seq-iFISH, these validation patient (P < 0.0001, Fig. S4), suggesting that the levels results showed significantly higher frequencies of both of aneuploidy correlate with CIN in HHD-B-ALL. chromosome 18 and 10 trisomies in DX samples from These results validate our computationally defined risk those HHD-B-ALL patients who remained disease-free predictors in HHD-B-ALL, and demonstrate that low as compared with those that relapsed after treatment rates of trisomies 18 and 10, and low levels of clonal (Fig. 5D). Remarkably, HHD-B-ALL patient classifica- heterogeneity are risk factors in HHD-B-ALL. tion at DX using the trisomies 18- and 10-based risk pre- dictor was highly reliable, with 42/50 (84%) of the 3.6. Longitudinal analysis of matched DX-REL patients correctly classified as either disease-free or HHD-B-ALL samples reveals distinct patterns of relapsed (Fig. 5E and Table S1). clonal evolution at REL To evaluate the prognostic value of this computation- ally defined risk predictor in the outcome of HHD-B- We next sought to analyze the clonal evolution during ALL patients (48/50 (96%) MRD-negative after induc- disease progression through longitudinal analyses of tion therapy), we next performed univariate and multi- available matched DX-REL samples from 10 HHD-B- variate statistical analyses of different clinically relevant ALL patients (Table 1). Results showed no major dif- parameters, including age, gender, WBCs, MRD status, ference in the number of total chromosomal gains treatment protocol and the trisomies 18- and 10-based between matched DX and REL samples, peaking risk predictor using a Cox regression model. Of note, between four and five chromosomal gains in both both univariate and multivariate analyses showed that groups (Fig. 6A). The rates of individual chromosomal the relapse predictor based on low levels (cut-off, 40%) gains were consistently reduced for all chromosomes in of trisomies 18 and 10 was the only independent risk fac- the REL samples compared with the matched DX tor associated with higher REL rates (hazard ratio = samples (Fig. 6B,C). Analysis of clonal heterogeneity 11.1; 95% confidence interval = 3.55–34.8, P < 0.001; in matched DX-REL samples showed a trend towards Table 3). In fact, within our validation cohort, statisti- reduced entropy levels in REL samples (P = 0.064), cally significant higher relapse-free survival (RFS) rates with six of 10 patients showing lower entropy levels in were observed in patients with a favorable (trisomies the REL samples (Fig. 6D), supporting previous data +18+10 > 40%) than in patients with unfavorable (tri- on DX samples of high levels of clonal heterogeneity somies +18+10 < 40%) risk predictor, with a 10-year as a favorable prognostic factor in HHD-B-ALL. Molecular Oncology 16 (2022) 2899–2919 Ó 2022 The Authors. Molecular Oncology published by John Wiley & Sons Ltd on behalf of 2911 Federation of European Biochemical Societies. 18780261, 2022, 16, Downloaded from https://febs.onlinelibrary.wiley.com/doi/10.1002/1878-0261.13276 by Readcube (Labtiva Inc.), Wiley Online Library on [23/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Federation of European Biochemical Societies. M. Ramos-Muntada et al. Molecular Oncology 16 (2022) 2899–2919 Ó 2022 The Authors. Molecular Oncology published by John Wiley & Sons Ltd on behalf of Novel risk predictors in hyperdiploid B-ALL 2912 18780261, 2022, 16, Downloaded from https://febs.onlinelibrary.wiley.com/doi/10.1002/1878-0261.13276 by Readcube (Labtiva Inc.), Wiley Online Library on [23/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License M. Ramos-Muntada et al. Novel risk predictors in hyperdiploid B-ALL Fig. 5. Trisomies 18 and 10 and clonal heterogeneity are robust relapse risk predictors in high-hyperdiploid B-cell acute lymphoblastic leuke- mia (HHD-B-ALL) who had achieved minimal residual disease (MRD)-negativity post-induction therapy. (A) Contribution of the indicated tri- somies at predicting clinical outcome; complete remission (CR) versus relapse (REL). Dots represent individual patient Gini importance values. Boxes represent the quartiles 25–75 and horizontal lines represent the mean value. Error bars represent the SD; n = 10 CR and 12 REL patient samples. (B) Stress tests on the indicated predictors for risk classification at diagnosis (DX) of HHD-B-ALL patients (n = 22). Blue bars represent the number of patients properly classified according to disease outcome (CR versus REL) and red bars the number of patients erroneously classified. (C) Classification based on DX samples of HHD-B-ALL patients according to risk predictor 3 as either CR (fa- vorable, % trisomies 18 and 10 > 40%) or REL (unfavorable, % trisomies 18 and 10 < 40%) risk groups. Number and percentage of prop- erly classified patients are indicated in the right. (D) Frequency of the indicated chromosome gains in an independent and blind validation cohort of DX samples of favorable (CR, n = 33) and unfavorable (REL, n = 17) HHD-B-ALL patients by iFISH analysis of chromosomes 18 and 10. Graphs represent the median values and dots represent the values obtained for individual patients; n = 200 nuclei per sample. (E) Classification of HHD-B-ALL patients as CR or REL using the chr18-chr10 risk predictor (predictor 3) after blind FISH analyses. The DX sam- ples from the validation cohort were initially blind-grouped as favorable or unfavorable risk based on chr18–chr10 risk predictor results, and then correlated with relapse information (No versus Yes) available from the clinic. 96% (48/50) of patients had achieved MRD negativity after induction therapy. The total number of patients for each group is indicated outside the quadrant. Note that 84% of the 50 HHD-B-ALL patients used in this blind and independent validation cohort were properly classified using predictor 3. (F,G) Kaplan–Meier curves for relapse-free survival (RFS) of HHD-B-ALL patients grouped according to the chr18–chr10 risk predictor (F) and clonal heterogeneity defined by the percentage of major clone (PMC) (G) after blind FISH analyses; n = 50 HHD-B-ALL samples collected at diagnosis (DX). Analyses for panels (F and G) were performed with the hazard ratios obtained with cox multivariate analyses. Table 3. Evaluation of the FISH risk predictor as a prognostic factor in HHD-B-ALL. Univariate and multivariate analyses comparing different clinical parameters and the chromosome 18–10 trisomies-based risk predictor prognostic value in a blind validation of an independent HHD- B-ALL cohort (n = 50 patients). CI, 95% confidence interval; MRD, MRD-positive after induction; WBC, white blood cell count. Univariate Multivariate Variable Hazard ratio CI P-val Hazard ratio CI P-val Agea 1.47 0.52–4.2 0.47 1.16 0.48–3.98 0.5 Sex (male) 2.85 0.65–12.5 0.2 2.04 0.45–9.29 0.4 WBC 1.01 0.99–1.01 0.9 1 0.99–1.02 0.8 MRD 3.54 0.80–15.64 0.09 4.45 0.36–55.1 0.2 Unfavorable FISH risk predictor 11.9 3.82–36.9 < 0.001 11.1 3.55–34.8 < 0.001 Treat (PETHEMA) Reference Reference Treat (SHOP) 0.74 0.15–3.67 0.7 1.73 0.24–12.3 0.6 Treat (AIEOP-BFM) 2.14 0.74–6.19 0.2 1.57 0.47–5.23 0.5 Treat (UKALL2003) 10.2 1.11–93.54 0.04 0.87 0.04–20.1 0.9 a Per 5-year increase. To study the clonal evolution in HHD-B-ALL dur- 4. Discussion ing disease progression, we next performed a hierar- chical cluster analysis of those unique subclones In this study, we took advantage of primary DX and observed in matched DX-REL samples and found REL BM samples from a large cohort of patients with that in 7/10 (70%) of the HHD-B-ALL patients the HHD-B-ALL with either favorable or unfavorable DX-REL samples clustered together (Fig. 6E). Inter- clinical outcome (after MRD negativity post-induction estingly, analyses of clonal composition in matched treatment), in order to comprehensively characterize DX-REL samples revealed two different patterns of clonal heterogeneity, specific chromosomal gains and chromosomal clonal evolution: (a) a pattern where clonal evolution. An important goal was to assess the the major leukemic clones are shared in DX and potential prognostic impact of specific chromosomal REL (7/10 of the patients), and (b) a pattern with a gains and clonal heterogeneity in HHD-B-ALL. This large clonal replacement from DX to REL (3/10 of study is technically sound because seq-iFISH analysis the patients; Fig. 6F and Fig. S5). Collectively, our was coupled with single-cell computational modeling longitudinal analysis of matched DX-REL HHD-B- allowing for a comprehensive investigation of the eight ALL samples reveals distinct patterns of clonal evolu- chromosomes typically gained in HHD-B-ALL at a tion at REL. single-cell level. Molecular Oncology 16 (2022) 2899–2919 Ó 2022 The Authors. Molecular Oncology published by John Wiley & Sons Ltd on behalf of 2913 Federation of European Biochemical Societies. 18780261, 2022, 16, Downloaded from https://febs.onlinelibrary.wiley.com/doi/10.1002/1878-0261.13276 by Readcube (Labtiva Inc.), Wiley Online Library on [23/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Federation of European Biochemical Societies. M. Ramos-Muntada et al. Molecular Oncology 16 (2022) 2899–2919 Ó 2022 The Authors. Molecular Oncology published by John Wiley & Sons Ltd on behalf of Novel risk predictors in hyperdiploid B-ALL 2914 18780261, 2022, 16, Downloaded from https://febs.onlinelibrary.wiley.com/doi/10.1002/1878-0261.13276 by Readcube (Labtiva Inc.), Wiley Online Library on [23/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License M. Ramos-Muntada et al. Novel risk predictors in hyperdiploid B-ALL Fig. 6. Longitudinal analysis of matched diagnostic-relapse high-hyperdiploid B-cell acute lymphoblastic leukemia (HHD-B-ALL) samples reveals distinct patterns of clonal evolution at REL. (A) Number of total chromosomal gains in matched diagnostic (DX) and relapse (REL) HHD-B-ALL samples. Graphs represent the mean value and error bars represent the SEM; n = 20 samples (10 DX and 10 REL). (B) Fre- quency of chromosomal gains for the indicated chromosomes in matched DX and REL HHD-B-ALL samples. Graphs represent the median values and dots represent the values obtained for each single sample. (C) Number of chromosomal gains in matched DX and REL HHD-B- ALL samples as observed by density values in single-cell computational analysis for the indicated chromosomes. (D) Box plot comparing the clonal heterogeneity as observed by Shannon entropy values between matched DX and REL HHD-B-ALL samples. Boxes represent the quartiles 25—75 and horizontal lines represent the mean value. Error bars represent the SD; two-sided paired t-test. (E) Heatmap depicting the unique subclones in matched DX and REL HHD-B-ALL samples. The color code represents the row Z-score for individual clones. On the top, dendrogram with hierarchical cluster analyses of samples regarding clonal composition similarity. (F) Single-cell analysis of the major clones observed in matched DX and REL HHD-B-ALL samples. Representative analyses of the two different patterns of chromosomal clonal evolution observed, with the major leukemic clones being either shared (top) or distinct (bottom) in matched DX and REL samples. The sub- clone formulas on the left indicate the number of gains for chromosomes X, 4, 6, 10, 14, 17, 18, and 21, respectively. The size of the black circles represents the proportion of cells showing that indicated subclone in each sample. Our data showed a high degree of clonal ‘chromoso- clinical outcome. To assess the prognostic value of clo- mal’ heterogeneity in HHD-B-ALL patients at DX, nal heterogeneity, we used the PMC and confirmed in with major subclones commonly representing < 30% the validation cohort a significantly higher RFS in of the leukemic cells. The major subclones identified patients with PMC ≤ 50%, thus demonstrating that by seq-iFISH usually differed from those observed lower levels of clonal heterogeneity represent a risk using conventional karyotyping. Earlier studies have factor associated with unfavorable outcome. reported contradictory data on the presence of clonal Contradictory data exist regarding the predictive heterogeneity in HHD-B-ALL which depend on the power of different chromosomal abnormalities on methodology of analysis used [9,23,25,41]. Consistent HHD-B-ALL outcome [14,16–18,34]. Our results with our data, clonal heterogeneity was observed by revealed that the rates of specific chromosomal gains, iFISH [23–25], spectral-karyotyping [42], single-cell similar to the levels of clonal heterogeneity, differ next-generation sequencing [43] and by cell division between HHD-B-ALL patients with favorable or unfa- studies directly assessing chromosome segregation in vorable clinical outcomes. Our results are in line with primary leukemic blasts [21]. Our results reinforce that previous data suggesting trisomy 18 as a favorable conventional karyotyping fails to reflect the actual clo- prognostic factor in HHD-B-ALL [17], as rates of tri- nal heterogeneity in HHD-B-ALL patients, suggesting somy 18 in DX samples were significantly higher in that clonal selection seen in conventional karyotyping CR than in REL patients. Moreover, computational analysis likely arises from the cell culture needed for modeling of seq-iFISH single-cell data allowed us to karyotyping, which may mask the actual clonal com- define a risk predictor based on the rates of trisomies position of HHD-B-ALL samples. 18 and 10. This computationally defined risk predictor Importantly, the clonal heterogeneity revealed by was validated in an independent cohort of HHD-B- our seq-iFISH analysis suggested the presence of CIN ALL patients and correctly classified 84% of patients in HHD-B-ALL. CIN has emerged as a prognostic according to clinical outcome, confirming the high reli- factor in different types of cancer, typically being asso- ability of the trisomies 18 and 10 risk predictor. In ciated with unfavorable clinical outcomes [44,45]. fact, and similar to clonal heterogeneity, univariate However, there is an extensive body of evidence and multivariate analyses revealed high levels (> 40%) suggesting that high CIN levels can also be found of trisomies 18 and 10 as an independent favorable associated with favorable outcomes due to a reduced prognostic factor in childhood HHD-B-ALL with a tumor-cell viability [35,46]. Indeed, increasing CIN 10-year RFS of 88% versus 22%. Importantly, the levels represent a strategy to selectively target cancer rates of trisomies 18 and 10 were inversely correlated cells [47]. To assess whether CIN levels are associated with the PMC, which further suggested that the levels with HHD-B-ALL outcome, we investigated the levels of aneuploidy correlate with CIN in HHD-B-ALL. Of of clonal heterogeneity based on analysis of the note, almost all (96%) patients in the validation cohort entropy values obtained by computational analysis of had reached MRD negativity after induction treat- single-cell data [29]. Our results show variable levels of ment, thus highlighting the reliability of both risk pre- clonal heterogeneity in DX samples of HHD-B-ALL dictors in stratifying those HHD-B-ALL patients who, patients, with significantly higher levels in DX samples even having achieved MRD negativity are at risk of from HHD-B-ALL patients showing a favorable progression/relapse. Furthermore, the multivariate Molecular Oncology 16 (2022) 2899–2919 Ó 2022 The Authors. Molecular Oncology published by John Wiley & Sons Ltd on behalf of 2915 Federation of European Biochemical Societies. 18780261, 2022, 16, Downloaded from https://febs.onlinelibrary.wiley.com/doi/10.1002/1878-0261.13276 by Readcube (Labtiva Inc.), Wiley Online Library on [23/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Novel risk predictors in hyperdiploid B-ALL M. Ramos-Muntada et al. analysis confirmed the power of both risk predictors in Acknowledgements stratifying regardless of the treatment protocol. Single-cell computational modeling also provided The authors thank Anthony V Moorman (University of information about clonal evolution in HHD-B-ALL Newcastle, UK) and Francesc Solé (Josep Carreras samples. On one hand, and consistent with previous Leukemia Research Institute) for critical discussions. data [25], our analysis revealed a high-order hierarchy Special thanks to Francisco Gutierrez-Agüera, Paola in relation to chromosomal gains in HHD-B-ALL, Romecı́n and Talı́a Velasco-Hernández (PM’s lab) for with specific chromosomes being gained together at technical help. We thank the CERCA program (Gener- different stages of disease initiation/evolution. A hier- alitat de Catalunya) and the Josep Carreras archical cluster analysis of those unique subclones Foundation-Obra Social la Caixa for institutional sup- observed at DX in each analyzed sample highlighted port. This work is supported by the Spanish Ministry of an extensive clonal heterogeneity in HHD-B-ALL Science and Innovation (PID2019-108160RB-I00/AEI/ DX samples with no specific clones associated with 10.13039/501100011033, the European Research Coun- relapse, suggesting that specific chromosomal gains cil (CoG-2014-646903), the Spanish Ministry of Econ- rather than specific clones were associated with dis- omy and Competitiveness (SAF2016-80481-R and ease outcome in HHD-B-ALL. On the other hand, SAF-2019-108160-R), the Leo Messi Foundation, the longitudinal analyses using paired DX-REL HHD-B- Health Institute Carlos III (ISCIII) (RICORS, RD21/ ALL samples revealed two different patterns of 0017/0029) within the Next Generation EU program chromosomal clonal composition at relapse: (a) major (plan de recuperación, transformación y resilencia), and leukemic clones shared in DX and REL, and (b) the Fundación Uno entre Cienmil (PM). Additional large clonal replacement from DX to REL. This funding was provided by the ISCIII (FEDER PI17/ latter clonal evolution pattern may reflect the emer- 01028 and PI20/00822) (CB), the Generalitat de Catalu- gence of leukemic subclones resistant/adapted to nya (2017/SGR-503) and the Universitat Autònoma de chemotherapy-induced clonal pressure. Although the Barcelona (UAB/CF-180034) (JB) and the Blood Can- number of matched DX-REL samples is relatively cer UK (CJH). JLT was supported by a Juan de la limited to reach definitive conclusions, it is notewor- Cierva fellowship by the Spanish Ministry of Science thy that the patients with such clonal replacement and Innovation (FJC2019-040868-I), BL-M was sup- from DX to REL had a worse outcome and they all ported by the Asociación Española Contra el Cáncer succumbed to the disease. (AECC; INVES20011LÓPE). AB was supported by the HHD-B-ALL is one of the most common malignan- Fundación Española de Hematologı́a y Hemoterapia cies in children [12] and up to 20% of these patients (FEHH). OM was supported by a Beatriu de Pinós eventually relapse. In fact, in absolute numbers there Postdoctoral Fellowship (BP2016-00048) from are more cases of relapsed HHD-B-ALL than de novo AGAUR (Generalitat de Catalunya), a 2017 Lady Tata diagnoses of other unfavorable molecular-subtypes of Memorial Trust International Award, and the AECC B-ALL [48]. Therefore, an improved risk-stratification (INVES211226MOLI). PM is an investigator of the of HHD-B-ALL patients is crucial to prospectively Spanish Cell Therapy cooperative network (TERCEL). identify those patients, who, having achieved MRD negativity after induction treatment, remain at high Conflict of interest risk of progression/relapse. The authors declare no conflict of interest. 5. Conclusions Author contributions Here, we report the levels of clonal heterogeneity and the rates of trisomies 18 and 10 as robust and inde- MR-M, JLT, VCR-C, AB, and BL-M performed the pendent relapse risk predictors in HHD-B-ALL. FISH experiments. CS, MO, PV, MLB, JN, MR-O, AM, analyses are commonly used in hemato-oncology clini- JLF, EC, MC, PB, GE, JB, MB, JMH-R, MJC, GC, cal laboratories for B-ALL diagnosis, as a comple- and CJH: provided clinical samples and biological ment for refining diagnosis after karyotyping or in data. JB and CB: designed experiments and interpreted cases with failed cytogenetics [49]. Our results provide data. PM: conceived the study, designed experiments, a panel of chromosomes for application of routine interpreted the data, financially supported the study FISH testing of HHD-B-ALL and offer a reliable and wrote the manuscript. OM: conceived the study, prognostic sub-stratification of HHD-B-ALL patients designed and performed experiments, analyzed/inter- at DX. preted the data, and wrote the manuscript. 2916 Molecular Oncology 16 (2022) 2899–2919 Ó 2022 The Authors. Molecular Oncology published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies. 18780261, 2022, 16, Downloaded from https://febs.onlinelibrary.wiley.com/doi/10.1002/1878-0261.13276 by Readcube (Labtiva Inc.), Wiley Online Library on [23/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License M. Ramos-Muntada et al. Novel risk predictors in hyperdiploid B-ALL 11 Paulsson K, Johansson B. High hyperdiploid childhood Peer review acute lymphoblastic leukemia. Genes Chromosomes The peer review history for this article is available at Cancer. 2009;48:637–60. https://publons.com/publon/10.1002/1878-0261.13276. 12 Bizard AH, Hickson ID. Anaphase: a fortune-teller of genomic instability. Curr Opin Cell Biol. 2018;52:112–9. 13 Harris MB, Shuster JJ, Carroll A, Look AT, Borowitz Data availability statement MJ, Crist WM, et al. Trisomy of leukemic cell The datasets used and/or analyzed during the current chromosomes 4 and 10 identifies children with B- progenitor cell acute lymphoblastic leukemia with a study are available from the corresponding author on very low risk of treatment failure: a pediatric oncology reasonable request. group study. Blood. 1992;79:3316–24. 14 Sutcliffe MJ, Shuster JJ, Sather HN, Camitta BM, References Pullen J, Schultz KR, et al. High concordance from 1 Pui CH, Mullighan CG, Evans WE, Relling MV. independent studies by the Children’s cancer group Pediatric acute lymphoblastic leukemia: where are we (CCG) and pediatric oncology group (POG) associating going and how do we get there? Blood. 2012;120:1165– favorable prognosis with combined trisomies 4, 10, and 74. 17 in children with NCI standard-risk B-precursor acute 2 Schmiegelow K, Forestier E, Hellebostad M, Heyman lymphoblastic leukemia: a Children’s oncology group M, Kristinsson J, Soderhall S, et al. Long-term results (COG) initiative. Leukemia. 2005;19:734–40. of NOPHO ALL-92 and ALL-2000 studies of 15 Sharathkumar A, DeCamillo D, Bhambhani K, childhood acute lymphoblastic leukemia. Leukemia. Cushing B, Thomas R, Mohamed AN, et al. Children 2010;24:345–54. with hyperdiploid but not triple trisomy (+4,+10,+17) 3 Maloney KW, Devidas M, Wang C, Mattano LA, acute lymphoblastic leukemia have an increased Friedmann AM, Buckley P, et al. Outcome in children incidence of extramedullary relapse on current with standard-risk B-cell acute lymphoblastic leukemia: therapies: a single institution experience. Am J Hematol. results of Children’s oncology group trial AALL0331. J 2008;83:34–40. Clin Oncol. 2020;38:602–12. 16 Paulsson K, Forestier E, Andersen MK, Autio K, 4 Pui CH, Yang JJ, Hunger SP, Pieters R, Schrappe M, Barbany G, Borgstrom G, et al. High modal number Biondi A, et al. Childhood acute lymphoblastic and triple trisomies are highly correlated favorable leukemia: progress through collaboration. J Clin Oncol. factors in childhood B-cell precursor high hyperdiploid 2015;33:2938–48. acute lymphoblastic leukemia treated according to the 5 Smith MA, Altekruse SF, Adamson PC, Reaman GH, NOPHO ALL 1992/2000 protocols. Haematologica. Seibel NL. Declining childhood and adolescent cancer 2013;98:1424–32. mortality. Cancer. 2014;120:2497–506. 17 Moorman AV, Ensor HM, Richards SM, Chilton L, 6 Bhojwani D, Pui CH. Relapsed childhood acute Schwab C, Kinsey SE, et al. Prognostic effect of lymphoblastic leukaemia. Lancet Oncol. 2013;14:e205–17. chromosomal abnormalities in childhood B-cell 7 Schrappe M, Hunger SP, Pui CH, Saha V, Gaynon PS, precursor acute lymphoblastic leukaemia: results from Baruchel A, et al. Outcomes after induction failure in the UKMedical Research Council ALL97/99 childhood acute lymphoblastic leukemia. N Engl J Med. randomised trial. Lancet Oncol. 2010;11:429–38. 2012;366:1371–81. 18 Moorman AV, Richards SM, Martineau M, Cheung 8 Bloomfield CD, Goldman AI, Alimena G, Berger R, KL, Robinson HM, Jalali GR, et al. Outcome Borgstrom GH, Brandt L, et al. Chromosomal heterogeneity in childhood high-hyperdiploid acute abnormalities identify high-risk and low-risk patients lymphoblastic leukemia. Blood. 2003;102:2756–62. with acute lymphoblastic leukemia. Blood. 1986;67: 19 Hann I, Vora A, Harrison G, Harrison C, Eden O, Hill 415–20. F, et al. Determinants of outcome after intensified 9 Raimondi SC, Pui CH, Hancock ML, Behm FG, therapy of childhood lymphoblastic leukaemia: results Filatov L, Rivera GK. Heterogeneity of hyperdiploid from Medical Research Council United Kingdom acute (51–67) childhood acute lymphoblastic leukemia. lymphoblastic leukaemia XI protocol. Br J Haematol. Leukemia. 1996;10:213–24. 2001;113:103–14. 10 Heerema NA, Sather HN, Sensel MG, Zhang T, 20 Moorman AV, Harrison CJ, Buck GA, Richards SM, Hutchinson RJ, Nachman JB, et al. Prognostic impact Secker-Walker LM, Martineau M, et al. Karyotype is of trisomies of chromosomes 10, 17, and 5 among an independent prognostic factor in adult acute children with acute lymphoblastic leukemia and high lymphoblastic leukemia (ALL): analysis of cytogenetic hyperdiploidy (> 50 chromosomes). J Clin Oncol. data from patients treated on the Medical Research 2000;18:1876–87. Council (MRC) UKALLXII/eastern cooperative Molecular Oncology 16 (2022) 2899–2919 Ó 2022 The Authors. Molecular Oncology published by John Wiley & Sons Ltd on behalf of 2917 Federation of European Biochemical Societies. 18780261, 2022, 16, Downloaded from https://febs.onlinelibrary.wiley.com/doi/10.1002/1878-0261.13276 by Readcube (Labtiva Inc.), Wiley Online Library on [23/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Novel risk predictors in hyperdiploid B-ALL M. Ramos-Muntada et al. oncology group (ECOG) 2993 trial. Blood. classification of spectral data. BMC Bioinformatics. 2007;109:3189–97. 2009;10:213. 21 Molina O, Vinyoles M, Granada I, Roca-Ho H, 33 Chillon MC, Gomez-Casares MT, Lopez-Jorge CE, Gutierrez-Aguera F, Valledor L, et al. Impaired Rodriguez-Medina C, Molines A, Sarasquete ME, et al. condensin complex and Aurora B kinase underlie Prognostic significance of FLT3 mutational status and mitotic and chromosomal defects in hyperdiploid B-cell expression levels in MLL-AF4+ and MLL-germline ALL. Blood. 2020;136:313–27. acute lymphoblastic leukemia. Leukemia. 2012;26:2360– 22 Harrison CJ, Moorman AV, Barber KE, Broadfield ZJ, 6. Cheung KL, Harris RL, et al. Interphase molecular 34 Schultz KR, Pullen DJ, Sather HN, Shuster JJ, Devidas cytogenetic screening for chromosomal abnormalities of M, Borowitz MJ, et al. Risk- and response-based prognostic significance in childhood acute classification of childhood B-precursor acute lymphoblastic leukaemia: a UKcancer cytogenetics lymphoblastic leukemia: a combined analysis of group study. Br J Haematol. 2005;129:520–30. prognostic markers from the pediatric oncology group 23 Talamo A, Chalandon Y, Marazzi A, Jotterand M. (POG) and Children’s cancer group (CCG). Blood. Clonal heterogeneity and chromosomal instability at 2007;109:926–35. disease presentation in high hyperdiploid acute 35 Weaver BA, Silk AD, Montagna C, Verdier-Pinard P, lymphoblastic leukemia. Cancer Genet Cytogenet. Cleveland DW. Aneuploidy acts both oncogenically and 2010;203:209–14. as a tumor suppressor. Cancer Cell. 2007;11:25–36. 24 Talamo A, Marazzi A, Rovo A, Schanz U, Tichelli A, 36 Murugaesu N, Wilson GA, Birkbak NJ, Watkins T, Chalandon Y, et al. High hyperdiploid acute McGranahan N, Kumar S, et al. Tracking the genomic lymphoblastic leukemia in adults shows clonal evolution of esophageal adenocarcinoma through heterogeneity and chromosomal instability at diagnosis neoadjuvant chemotherapy. Cancer Discov. 2015;5:821–31. and during the course of the disease. Ann Hematol. 37 Bolhaqueiro ACF, Ponsioen B, Bakker B, Klaasen SJ, 2012;91:793–6. Kucukkose E, van Jaarsveld RH, et al. Ongoing 25 Alpar D, Pajor G, Varga P, Kajtar B, Poto L, Matics chromosomal instability and karyotype evolution in R, et al. Sequential and hierarchical chromosomal human colorectal cancer organoids. Nat Genet. changes and chromosome instability are distinct 2019;51:824–34. features of high hyperdiploid pediatric acute 38 Vargas-Rondon N, Perez-Mora E, Villegas VE, Rondon- lymphoblastic leukemia. Pediatr Blood Cancer. Lagos M. Role of chromosomal instability and clonal 2014;61:2208–14. heterogeneity in the therapy response of breast cancer cell 26 Godo A, Blanco J, Vidal F, Anton E. Accumulation of lines. Cancer Biol Med. 2020;17:970–85. numerical and structural chromosome imbalances in 39 Bakhoum SF, Cantley LC. The multifaceted role of spermatozoa from reciprocal translocation carriers. chromosomal instability in cancer and its Hum Reprod. 2013;28:840–9. microenvironment. Cell. 2018;174:1347–60. 27 Catalina P, Montes R, Ligero G, Sanchez L, de la 40 Bakhoum SF, Landau DA. Chromosomal instability as Cueva T, Bueno C, et al. Human ESCs predisposition a driver of tumor heterogeneity and evolution. Cold to karyotypic instability: is a matter of culture Spring Harb Perspect Med. 2017;7:a029611. adaptation or differential vulnerability among hESC 41 Heerema NA, Raimondi SC, Anderson JR, Biegel J, lines due to inherent properties? Mol Cancer. 2008;7:76. Camitta BM, Cooley LD, et al. Specific extra 28 Molina O, Sarrate Z, Vidal F, Blanco J. FISH on chromosomes occur in a modal number dependent sperm: spot-counting to stop counting? Not yet. Fertil pattern in pediatric acute lymphoblastic leukemia. Steril. 2009;92:1474–80. Genes Chromosomes Cancer. 2007;46:684–93. 29 Niida A, Nagayama S, Miyano S, Mimori K. 42 Elghezal H, Le Guyader G, Radford-Weiss I, Perot C, Understanding intratumor heterogeneity by combining Van Den Akker J, Eydoux P, et al. Reassessment of genome analysis and mathematical modeling. Cancer childhood B-lineage lymphoblastic leukemia karyotypes Sci. 2018;109:884–92. using spectral analysis. Genes Chromosomes Cancer. 30 de Vienne DM, Aguileta G, Ollier S. Euclidean nature 2001;30:383–92. of phylogenetic distance matrices. Syst Biol. 43 Bakker B, Taudt A, Belderbos ME, Porubsky D, 2011;60:826–32. Spierings DC, de Jong TV, et al. Single-cell sequencing 31 Liaw A, Wiener M. Classification and regression by reveals karyotype heterogeneity in murine and human randomForest. R News. 2002;2:18–22. malignancies. Genome Biol. 2016;17:115. 32 Menze BH, Kelm BM, Masuch R, Himmelreich U, 44 Vishwakarma R, McManus KJ. Chromosome Bachert P, Petrich W, et al. A comparison of random instability; implications in cancer development, forest and its Gini importance with standard progression, and clinical outcomes. Cancers (Basel). chemometric methods for the feature selection and 2020;12:824. 2918 Molecular Oncology 16 (2022) 2899–2919 Ó 2022 The Authors. Molecular Oncology published by John Wiley & Sons Ltd on behalf of Federation of European Biochemical Societies. 18780261, 2022, 16, Downloaded from https://febs.onlinelibrary.wiley.com/doi/10.1002/1878-0261.13276 by Readcube (Labtiva Inc.), Wiley Online Library on [23/01/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License M. Ramos-Muntada et al. Novel risk predictors in hyperdiploid B-ALL 45 Carter SL, Eklund AC, Kohane IS, Harris LN, Fig. S1. (Related to Fig. 2). Read-out accuracy and Szallasi Z. A signature of chromosomal instability reliability of Seq-iFISH analyses. inferred from gene expression profiles predicts clinical Fig. S2. (Related to Fig. 4). Hierarchical chromosomal outcome in multiple human cancers. Nat Genet. gains in the indicated complete remission (CR) high- 2006;38:1043–8. hyperdiploid B-cell acute lymphoblastic leukemia 46 Silk AD, Zasadil LM, Holland AJ, Vitre B, Cleveland (HHD-B-ALL) patients. DW, Weaver BA. Chromosome missegregation rate Fig. S3. (Related to Fig. 4). Hierarchical chromosomal predicts whether aneuploidy will promote or suppress gains in the indicated relapsed (REL) high-hyper- tumors. Proc Natl Acad Sci USA. 2013;110:E4134–41. diploid B-cell acute lymphoblastic leukemia (HHD-B- 47 Tanaka K, Hirota T. Chromosomal instability: a ALL) patients. common feature and a therapeutic target of cancer. Fig. S4. (Related to Fig. 5). Aneuploidy levels are Biochim Biophys Acta. 2016;1866:64–75. associated with chromosome instability (CIN). 48 Sanjuan-Pla A, Bueno C, Prieto C, Acha P, Stam RW, Fig. S5. (Related to Fig. 6). Individual longitudinal Marschalek R, et al. Revisiting the biology of infant t (4;11)/MLL-AF4+ B-cell acute lymphoblastic leukemia. analysis of matched diagnostic-relapse (DX-REL) Blood. 2015;126:2676–85. high-hyperdiploid B-cell acute lymphoblastic leukemia 49 Schwab C, Harrison CJ. Advances in B-cell precursor (HHD-B-ALL). acute lymphoblastic leukemia genomics. Hemasphere. Table S1. Cytogenetic and clinical data of all the child- 2018;2:e53. hood high-hyperdiploid B-cell acute lymphoblastic leu- kemia (HHD-B-ALL) samples used for blind validation analysis. Supporting information Table S2. Levels of ‘false’ gains and losses observed Additional supporting information may be found by Seq-iFISH analysis in the indicated control (Ctrl) online in the Supporting Information section at the end samples. of the article. Molecular Oncology 16 (2022) 2899–2919 Ó 2022 The Authors. Molecular Oncology published by John Wiley & Sons Ltd on behalf of 2919 Federation of European Biochemical Societies.