Relapse-Fated Latent Diagnosis Subclones in Acute B Lineage Leukemia Are Drug Tolerant and Possess Distinct Metabolic Programs

Home , MRPL17, MRPL3, MRPL33, MRPL40, MRPS22, NDUFA10

Published OnlineFirst February 21, 2020; DOI: 10.1158/2159-8290.CD-19-1059

Research Article

Stephanie M. Dobson1,2, Laura García-Prat2, Robert J. Vanner1,2, Jeffrey Wintersinger3, Esmé Waanders4,5,6, Zhaohui Gu6, Jessica McLeod2, Olga I. Gan2, Ildiko Grandal7, Debbie Payne-Turner6, Michael N. Edmonson8, Xiaotu Ma8, Yiping Fan8, Veronique Voisin1,9, Michelle Chan-Seng-Yue2,10, Stephanie Z. Xie2, Mohsen Hosseini2, Sagi Abelson2, Pankaj Gupta8, Michael Rusch8, Ying Shao11, Scott R. Olsen12, Geoffrey Neale12, Steven M. Chan2, Gary Bader1,9, John Easton11, Cynthia J. Guidos13,14, Jayne S. Danska7,13,14, Jinghui Zhang8, Mark D. Minden2,15, Quaid Morris1,3,9,16, Charles G. Mullighan6, and John E. Dick1,2

abstract Disease recurrence causes significant mortality in B-progenitor acute lymphoblastic leukemia (B-ALL). Genomic analysis of matched diagnosis and relapse samples shows relapse often arising from minor diagnosis subclones. However, why therapy eradicates some subclones while others survive and progress to relapse remains obscure. Elucidation of mechanisms underlying these differing fates requires functional analysis of isolated subclones. Here, large-scale limiting dilution xenografting of diagnosis and relapse samples, combined with targeted sequencing, identified and isolated minor diagnosis subclones that initiate an evolutionary trajectory toward relapse [termed diagnosis Relapse Initiating clones (dRI)]. Compared with other diagnosis subclones, dRIs were drug-tolerant with distinct engraftment and metabolic properties. Transcriptionally, dRIs displayed enrichment for chromatin remodeling, mitochondrial metabolism, proteostasis programs, and an increase in stemness pathways. The isolation and characterization of dRI subclones reveals new avenues for eradicating dRI cells by targeting their distinct metabolic and transcriptional pathways before further evolution renders them fully therapy-resistant.

Significance: Isolation and characterization of subclones from diagnosis samples of patients with B-ALL who relapsed showed that relapse-fated subclones had increased drug tolerance and distinct metabolic and survival transcriptional programs compared with other diagnosis subclones. This study provides strategies to identify and target clinically relevant subclones before further evolution toward relapse.

See related article by E. Waanders et al.

1Department of Molecular Genetics, University of Toronto, Toronto, of Medicine, University of Toronto, Toronto, Ontario, Canada. 16Vector Ontario, Canada. 2Princess Margaret Cancer Centre, University Health Institute, Toronto, Canada. Network, Toronto, Ontario, Canada. 3Department of Computer Science, Note: Supplementary data for this article are available at Cancer Discovery University of Toronto. Toronto, Ontario, Canada. 4Princess Máxima Center Online (http://cancerdiscovery.aacrjournals.org/). for Pediatric Oncology, Utrecht, the Netherlands. 5Department of Genetics, University Medical Center Utrecht, Utrecht, the Netherlands. 6Depart- L. García-Prat, R.J. Vanner, J. Wintersinger, and E. Waanders contributed ment of Pathology, St. Jude Children’s Research Hospital, Memphis, equally to this article. Tennessee. 7Genetics and Genome Biology, Hospital for Sick Children Corresponding Authors: John E. Dick, Princess Margaret Cancer Centre, Research Institute, Toronto, Ontario, Canada. 8Department of Computa- University Health Network, University of Toronto, Princess Margaret Cancer tional Biology and Bioinformatics, St. Jude Children’s Research Hospi- Research Tower, 101 College Street, Toronto, Ontario M5G 1L7, Canada. Phone: tal, Memphis, Tennessee. 9Donnelly Centre for Cellular and Biomolecular 416-581-7472; Fax: 416-581-7476; E-mail: [email protected]; Research, Toronto, Ontario, Canada. 10PanCuRx Translational Research and Charles G. Mullighan, St. Jude Children’s Research Hospital, 262 Danny Initiative, Ontario Institute for Cancer Research, Toronto, Ontario, Canada. Thomas Place, Mailstop 342, Memphis, TN 38105, Phone: 901-595-3387; 11Pediatric Cancer Genome Project Laboratory, St. Jude Children’s Fax: 901-595-5947; E-mail: [email protected] Research Hospital, Memphis, Tennessee. 12Hartwell Center for Bioinfor- Cancer Discov 2020;10:568–87 matics and Biotechnology, St. Jude Children’s Research Hospital, Memphis, Tennessee. 13Developmental & Stem Cell Biology Program, Hospital for doi: 10.1158/2159-8290.CD-19-1059 Sick Children Research Institute, Toronto, Ontario, Canada. 14Department ©2020 American Association for Cancer Research. of Immunology, University of Toronto, Toronto, Ontario, Canada. 15Faculty

568 | CANCER DISCOVERY April 2020 AACRJournals.org

from single-cell analysis, has largely substantiated these pre- Introduction dictions (8, 9). Functional studies to explain therapy failure Despite significant advancements in the treatment of acute have mainly been undertaken by comparing diagnosis cells lymphoblastic leukemia (ALL), the disease recurs in 15% to with those from relapse and identifying drug-resistance mecha- 20% of pediatric and 40% to 75% of adult patients, with the nisms present in relapse and absent at diagnosis. However, prognosis for patients who relapse being dismal (1–3). Analysis the properties of the relapse samples have been shaped by expo- of paired diagnosis and relapse ALL samples with increasingly sure to chemotherapy, causing further evolution and mutagen- broader and deeper genomic-sequencing methods has shown esis. Thus, two critical questions remain: What are the unique that classic Darwinian branching evolution of genomically properties and mechanisms that contribute to the relapse fate distinct subclones is a hallmark of disease recurrence (4, 5). of a particular diagnosis subclone prior to full evolution to At both diagnosis and relapse, a single neoplasm may con- drug-resistant relapse disease, and when does drug tolerance tain multiple genetic subclones related to each other through arise? Drug tolerance may arise stochastically through genetic complex evolutionary trajectories (4–7). Although relapse may or epigenetic mechanisms prior to exposure to therapy, and evolve from the predominant clone at diagnosis, in the major- be selected for by both cell-autonomous and non–cell autono- ity of patients relapse arises from preexisting minor subclones mous processes (10–13). Alternatively, therapy may induce within the diagnosis sample or from a rare ancestral clone (4–6). genomic aberrations that are then selected for during disease Although the population-level genetic analyses upon which progression, particularly if such alterations reduce leukemic these conclusions are based rely on computational inference cell fitness (14) during disease establishment and prior to the of their evolutionary relationships from analysis of bulk leu- administration of therapy (14, 15, 16). Without isolation of kemic cells, resolution of leukemic subclones at clonal lev- the subclones that contribute to disease progression from diag- els, either through isolation of subclones in xenografts or nosis samples, it is not possible to answer these questions and

April 2020 CANCER DISCOVERY | 569

RESEARCH ARTICLE Dobson et al. uncover the cellular and molecular properties that explain their B-ALL of varying genetic subtypes were undertaken to iden- differing subclonal fates and drug tolerance. tify somatic single-nucleotide variants (SNV), insertion– Many therapy resistance mechanisms have been implicated in deletion mutations (indels), and DNA copy-number altera- B-progenitor ALL (B-ALL), including acquisition of stemness tions (CNA). The patients encompassed a range of cytoge- programs, dormancy, the protective role of the niche, and the netic subtypes and varied in the length of their disease acquisition of resistance driver mutations (14, 17–21). How- remission (range 5.88–94.8 months; Supplementary Table ever, the ability to evade drug treatment is only one prerequisite S1). Diagnosis samples had a median of 24 somatic SNV/ of relapse; surviving cells must also possess significant clonal indels (range 7–100) and 13.5 CNA (range 1–51), whereas regenerative capacity to regrow or reproduce disease. For many relapse samples contained a median of 39.5 SNV/indels human cancers, only rare fractions of malignant cells are (range 22–405) and 16.5 CNA (range 2–58; Supplemen- capable of significant clonal propagation as detected by xen- tary Table S1). Leukemic variants were confirmed by tar- ograft-based cancer or leukemia-initiating cell (L-IC) assays geted sequencing using a custom capture array of the (10). Indeed, methods to propagate primary human leuke- variants identified by WES (Fig. 1A). Targeted sequencing mia samples were first undertaken in B-ALL with patient- and WES data were merged, resulting in a coverage of derived xenografts (PDX) recapitulating many features of approximately 350×. Computational analysis of the variant the patient’s disease (22, 23). Subsequent studies used a allele frequencies (VAF) of leukemic variants comparing limiting dilution approach to generate xenografts, thereby diagnosis and relapse samples predicted that the origin tracking the growth properties and genetic identity of single of relapse arose from a minor subclone present at diagno- L-ICs (9). Thus, depending on the cell dose transplanted, rare sis in 10 patients (patients 1, 3–7, 9, and 12–14; Fig. 1B; subclones with poorer competitive repopulation properties, Supplementary Table S1) and through further evolution compared with more aggressive or numerous L-ICs within of the major diagnostic subclone in 4 patients (patients the sample, could be identified. By genetic analysis of such 2, 8, 10, and 11; Supplementary Table S1). Broadly, these xenografts, evidence was found for the existence of ancestral findings of evolutionary origins are representative of the and/or minor subclones, proving that branching evolution much larger analysis of 92 paired samples that reports the and clonal diversity occur at the level of L-ICs; however, mutational landscape and patterns of clonal evolution of these studies did not test for relapse-fated subclones (8, 9). relapsed childhood ALL as described in Waanders and col- Conceptually, L-ICs with the capacity for clonal propagation leagues (7). should serve as the units of selection during disease progres- To gain insight into the genetic diversity at the level of sion because only mutations accumulating in clonal propa- L-ICs and uncover rare and/or outcompeted clones, puri- gating cells are relevant for further disease evolution. Thus, fied leukemic blasts from primary diagnosis and relapse the relationship between disease progression and properties samples were transplanted intrafemorally in a limiting dilu- of stemness is an active area of investigation (10, 24). The tion assay into 872 NOD.CB17-PrkdcscidIl2rgtm1Wjl/Szj (NSG) pairing of xenografting assays with genomic studies provides mice to generate primary PDXs. The frequency of L-ICs a unique opportunity to enrich for the cellular reservoirs of varied widely between samples, with L-IC ranges correspond- relapse. However, a direct test of this concept through paired ing to those described previously (31, 32); however, paired diagnosis and relapse xenografting in B-ALL has been limited analysis between diagnosis and relapse samples did not (23, 25, 26). Paired studies in acute myeloid leukemia (AML) show a consistent trend in L-IC enrichment at either time and T-lineage acute lymphoblastic leukemia (T-ALL) have point (Supplementary Table S2). This collection of PDXs demonstrated that xenografts can capture distinct subclones that were engrafted [n = 402, average 28 per patient (range present within the diagnosis samples of relapsing patients, some 11–58), 13 per sample (range 0–27)] was used to further genetically more closely related to the predominant diagnosis study the clonal landscape present in the patient samples. subclone and others to a corresponding relapse sample (27–30). To determine whether PDXs captured the clonal diversity These results suggest that latent relapse-initiating subclones and disease evolution present in the patient samples, human possess competitive growth advantages when assessed by xeno- cells were isolated from the bone marrow (BM), spleen, grafting, but remain suppressed by the predominant clone in the and central nervous system (CNS) of engrafted mice and patient, making them difficult to study. subjected to targeted sequencing using the custom capture Here, we undertook a combined functional and genomic array designed for the patient samples (Fig. 1A). PDXs with analysis of 14 genetically distinct paired diagnosis and relapse sufficient human engraftment for genotyping >( 10% human B-ALL patient samples to isolate latent subclones within chimerism, 372 PDXs total of 402 engrafted PDXs, average the diagnosis sample that initiate relapse. Isolation of sub- 26 PDXs per patient) were analyzed. Individual PDXs from clones enabled functional analysis of their growth and drug the same patient sample were often found to vary in their response properties as well as molecular analysis of their clonal composition in terms of the presence and frequency transcriptomic profiles. of variants, suggesting that the L-ICs initiating the grafts derived from genetically diverse subclones often capture the totality of the clonal diversity present in the diagnosis Results patient sample (Fig. 1C). Leukemic variants were classified on the basis of the VAF Isolation of Relapse-Initiating Subclones in B-ALL of the variants in the bulk patient diagnosis and relapse Whole-exome sequencing (WES) and SNP microarray samples from which the PDXs were generated: preserved vari- analysis of 6 adult patients and 8 pediatric patients with ants (VAF > 30% in both diagnosis and relapse samples, or

570 | CANCER DISCOVERY April 2020 AACRJournals.org

Characterization of Relapse-fated Clones in Diagnosis B-ALL RESEARCH ARTICLE

A CTCCCCTTGA Identify leukemic variants of patient samples Targeted sequencing of patient leukemic variants in Whole-exome human purified PDX cells sequencing CTCCCCTTGA

Patient leukemic cells Identify subclonal identities represented in distinct PDX PDX Sacrifice & harvest cells Intrafemoral transplantation 20–30 weeks Mouse cell depletion NSG mice or cell sorting for human cells n = 30/sample B 100 Common Diagnosis Relapse 80 d a Relapse specific b 60 d c a 40 c Minor clone evolution e OR 20 e Major clone evolution 0 b Diagnosis specific Relapse variant allele frequency (%) 0 20 40 60 80 100 Diagnosis variant allele frequency (%)

Pt. 1 Diagnosis xenografts Relapse xenografts Diagnosis Relapse

TBX5 FASN ZNF571 KCTD5 NBEA Disease time point PRSS16 Diagnosis SMARCAD1 MED12 Relapse Preserved RPL10 DOCK1 Sample source (a) PTPN11 Patient NOX4 HDAC9 PDX LPA MUC16 Cell dose transplanted CAMK2A Diagnosis LAMA2 n.a. TRMT1 250,000 specific (b) SLC5A12 10,000 SPANXN2 1,000 PLIN4 Latent ATG16L2 100 HIF3A 10 (c) ESRRG GPM6B EFCAB10 Variant class NUP188 DCDC2 Preserved CHD9 Diagnosis specific GTF2H4 MMP25 Latent Relapse POLR2A Relapse specific C2orf71 specific COIL ST3GAL3 (d and e) SCD5 FAT4 LOC284232 00.2 0.40.6 0.81 SLC19A1 ITPKB VAF PLCXD3 TFAP2E

Figure 1. PDXs capture clonal diversity present in paired diagnosis and relapse B-ALL samples. A, Experimental schematic. PDXs were generated for 6 adult and 8 pediatric B-ALL patient samples at diagnosis and relapse by intrafemoral transplantation of sorted leukemic blasts into 30 irradiated NSG mice in a limiting dilution assay. Mice were sacrificed 20 to 30 weeks post-transplant and their engraftment was assessed by flow cytometry. Patient samples were also subjected to genomic analysis including whole-exome sequencing (WES). Variants identified from the WES of patient samples at either time point were used to create custom capture baits for targeted sequencing at a deeper depth in the patient samples and their corresponding PDX. PDXs representing varying clones were identified. B, Schematic representation of the results obtained by mutational clustering of variants based on the variant allele frequencies (VAF) at diagnosis (x-axis) and relapse (y-axis) of patient 1 in 2-D VAF plots showing evolution from a minor subclone as depicted. Each dot represents a variant. Shared variants are shown in gray clusters (clusters a and c). Diagnosis- and relapse-specific variants are shown in the blue cluster (cluster b) and red clusters (clusters d and e), respectively. C, Heat map of the VAFs of leukemic variants at diagnosis, relapse, and in their corresponding PDXs for patient 1. Variant classes are labeled with their class and a letter corresponding to the clusters illustrated in B.

April 2020 CANCER DISCOVERY | 571

RESEARCH ARTICLE Dobson et al. preserved between samples); diagnosis-specific variants (pre- was instrumental in showing the existence of subclones sent at diagnosis and absent at relapse); latent variants (pre- at diagnosis that bear latent (patient 2) or relapse-specific sent at diagnosis with VAF < 30% and increasing at relapse); variants indicative of ancestral clones (patient 11) in relapse-specific variants (absent at diagnosis and present at 2 patients in which genetic analysis of bulk diagnosis and relapse; Supplementary Table S3). The limit of detection of relapse samples had predicted evolution from the major our combined sequencing was a VAF of 1%. This analysis diagnosis clone (Fig. 2B; Supplementary Table S3). Only one revealed three patterns of engraftment in PDXs derived from diagnosis patient sample (patient 10), a sample predicted to diagnosis samples (termed dPDXs). In 10 of 13 engrafting arise from evolution of the major diagnostic clone, did not diagnosis samples (76.9%, patients 1–7, 9, 12, and 14), latent engraft at the highest transplanted cell dose (250,000 cells). variants were enriched in dPDXs (>10% increase in VAF in Therefore, xenografting added considerable new insight into dPDXs) as compared with the diagnosis patient sample the subclonal repertoire of L-ICs in these patients and their demonstrating the regenerative potential of clones marked evolutionary fates and patterns. by these variants (Fig. 2A; Supplementary Fig. S1A and S1B). In 4 of 13 patients (patients 5, 6, 11, and 13), dPDXs Genetic Analysis of Xenografts Provides Insight were generated where leukemia cells bore relapse-specific into the Dynamics of Subclone Evolution variants establishing their existence within the diagnosis To gain insight into the evolutionary relationships and sample, despite being at levels below the detectable limit processes underlying the divergence of subclones, we under- of the sequencing in the patient sample (Fig. 2B). Finally, took two approaches: population phylogenetic analysis to in one patient (patient 8), whose relapse was predicted to examine the genetic similarity between xenografts and patient evolve from the major diagnosis clone based on analysis of samples; and generation of mutational trees to reconstruct the bulk patient samples, only diagnosis clones engrafted the clonal hierarchies. We began by using a population genet- in dPDXs (Supplementary Fig. S1C). This patient carried ics approach where phylogenetic analysis was performed an ETV6–RUNX1 translocation and had the longest remis- for the patient samples and the clones isolated from their sion, relapsing 8 years after the initial B-ALL diagnosis. PDXs to depict the evolutionary relationships between xeno- Our approach of generating PDXs with differing cell doses grafts and the patient samples (Fig. 3A and B; Supplementary

A B Diagnosis Relapse Pt. 9 xenografts xenografts Diagnosis Relapse Pt. 11

Diagnosis Relapse xenografts xenografts Relapse Diagnosis

KIF1A EPHA7 FMO4 n = 10 SHROOM1 THY1 GPR98 SPIN4 Disease time point YIPF7 n = 18 P2RY6 Diagnosis AMPD1 PHF19 ITFG2 C2orf15 Relapse RIMS2 SNX32 TRIL Sample source SOX9 HLF ADAMTS19 Patient OR8K3 FBN1 PDX ASH2L VPS26B AP3B2 ANKRD11 CREBBP_L1822L Cell dose transplanted ZNF655 PHRF1 n.a. CELSR3_L1837I TSC2 CELSR3_L1836L NRAP 250,000 INPPL1 PRINS 10,000 GABRA5 CREBBP_S1761 1,000 IL2 CREBBP_I2101M MC5R PLXNA4 PDE4D GOLGA8C Variant class VCX3B NUP205 MUC16 Preserved MYCBP2 SOX6 SFTA3 Diagnosis specific C16orf54 SGK1 Latent GNB4 DND1 Relapse specific RUNX1T1 TECTA KIAA0947 AVPR1A CMYA5 GRIK3 KLHL14 CLCNKB MDGA2 MAK ANO5 FAIM2_E12_C.T 0 0.2 0.4 0.6 0.8 1 CSN2 FAIM2_E12_C.G VAF AHNAK PRKAB2 TP53 SORCS1 VCP ATMIN SHPRH SMC1A EPHA3 TCAM1P

Figure 2. PDXs enrich for latent diagnosis clones. A and B, Heat maps of VAF of the SNV and indel leukemic variants identified by whole-exome and targeted sequencing in diagnosis/relapse patient samples and PDXs, respectively. Variants are clustered as preserved (present in diagnosis and relapse patient samples), diagnosis specific (present in diagnosis patient sample and absent in relapse patient sample), latent (present in diagnosis patient sample with VAF < 0.3 and expanding in relapse sample), and relapse specific (present in relapse patient sample and absent in diagnosis patient sample). PDXs are ordered in decreasing number of transplanted cell doses. A, Representative heat map for the selection of latent variants in diagnosis PDX as observed in patient 9. B, Representative heat map of patient 11 displaying the identification of a relapse-specific variant undetectable in the patient diagnosis sample but present in diagnosis PDX.

572 | CANCER DISCOVERY April 2020 AACRJournals.org

Characterization of Relapse-fated Clones in Diagnosis B-ALL RESEARCH ARTICLE

AB Pt. 11 Diagnosis Relapse Pt. 9 Diagnosis rPDX.18 dPDX.4 rPDX.14 dPDX.2 rPDX.12 dPDX.12 rPDX.6 rPDX.1 dPDX.5 rPDX.5 dPDX.8 rPDX.2 dPDX.1 rPDX.8 rPDX.1 rPDX.3 rPDX.11 rPDX.4 rPDX.4 rPDX.3 rPDX.13 rPDX.2 rPDX.10 rPDX.8 rPDX.7 rPDX.5 rPDX.15 rPDX.9 rPDX.7 dPDX.11 dRI Relapse dPDX.10 dPDX.9 dPDX.8 dPDX.7 dRI dPDX.6 dPDX.10 dPDX.3 dPDX.1 dPDX.6 dPDX.9 dPDX.3 dPDX.7 dPDX.11 dPDX.2 Diagnosis patient sample dPDX.5 dPDX.4 Diagnosis PDX dPDX.12 Relapse patient sample Relapse PDX

Figure 3. PDXs identify relapse-fated clones in diagnosis patient samples. Phylogenetic analysis showing relationship of patient 9 (A) and patient 11 (B) patient samples and xenografts, based on VAF of leukemic variants. The distances between symbols on the tree were estimated by a nearest neighbor joining method and represent the degrees of relation between them (Minkowski’s distance). Circles represent patient samples and triangles represent PDX; blue represents diagnosis and red represents relapse. Diagnosis clones on the trajectory to relapse were termed dRI and are indicated by a box with a hatched border.

Fig. S1D–S1G). These analyses demonstrate the ability of PDXs to the patient samples from which they were derived, such as to capture diagnosis subclones more closely related to the dPDX7 in patient 1 whose clonal composition is more closely diagnosis patient sample, as well as subclones present in the related to the relapse than the diagnosis sample (Fig. 4C and diagnosis patient sample that are on the trajectory to relapse. D). In addition, the PDXs aided in subdividing single muta- We term these partially evolved subclones that go on to cause tional populations into more than a single linearly related relapse diagnosis Relapse Initiating clones (dRI). dRIs were population or branch in 11 of 14 patients. For example, popu- identified in all engrafting diagnosis patient samples except lation G in patient 9 was expanded into two separate popula- for the patient with the longest time of remission (patient tions (populations 8 and 9) when PDXs were included (Fig. 8). Thus, our xenografting approach enabled the capture and 4E and F; Supplementary Figs. S2A–S2C, S3A–S3C, S4A–S4E, isolation of subclones, including those on the evolutionary tra- and S5A–S5C). jectory of relapse, allowing for further functional analysis and aiding in the determination of mutational acquisition. Genetically Diverse Subclones Have Differing For our second approach, we used the extensive mutational Xenograft Repopulation Kinetics information garnered from the 372 PDXs to supplement the The mutational analysis of the xenografts performed patient analysis. For each patient, computational analysis using PairTree helped illuminate the competitive differences using PairTree (Wintersinger and colleagues, in preparation) in xenograft repopulation kinetics of specific subclones. We was undertaken to group leukemic variants with similar VAFs compared the predominant diagnosis mutational popula- across numerous PDX and bulk samples to define geneti- tions in the patient sample to the mutational populations cally unique mutational populations. We then constructed that engrafted in the xenografts. In all but four patient sam- mutational trees to describe the evolutionary history of each ples, the dPDX captured the diversity of clones present in leukemia and the order of mutational acquisition (Fig. 4; the diagnosis samples (Fig. 4; Supplementary Figs. S2A–S2C, Supplementary Figs. S2A–S2C, S3A–S3C, S4A–S4E, and S5A– S3A–S3C, S4A–S4E, and S5A–S5C). In 2 of these 4 patients S5C; Supplementary Table S3). In 7 of 14 patients, the PDXs (patients 7 and 11), minor diagnosis-specific mutational pop- helped separate the mutational acquisition into multiple ulations, including a population harboring missense NRAS distinct populations that could not be resolved using the mutations in patient 7, did not engraft in the dPDX mice (Sup- sequencing data of the primary patient samples alone. For plementary Fig. S4A; Supplementary Table S3). In the other instance, in patient 1, analysis of the xenografts revealed the two patients (patients 9 and 12), all the dPDXs were initiated existence of four distinct subclonal populations (populations from a minor population in the diagnosis sample ancestral 10, 12, 14, and 15) at relapse that, when only the patient sam- to the relapse (Fig. 4E and F; Supplementary Figs. S4C and ples were considered, had been predicted to be only a single S5A). These data show the enhanced competitive repopula- subclone (population G; Fig. 4A–D). This analysis further tion properties of these dRI clones. In patient 9, all dPDXs were enabled comparisons of the state of the evolutionary trajec- initiated from mutational population 5 that corresponded to tory of the subclones captured within each individual PDX only 22% of the diagnosis leukemic cells (initially described (denoted by a numbered PDX referring to a specific mouse) as population E in the patient sample alone mutational tree)

April 2020 CANCER DISCOVERY | 573

RESEARCH ARTICLE Dobson et al.

1 1 A Diagnosis Relapse E subclonal subclonal B C D C D prevalence prevalence 0 0 Pt. 9 Pt. 1 A AB

EFG H E FG

B C D G F Pt. 1 345 10 11 A B

12 Pt. 9 diagnosis dPDx 8 dPDX 1 dPDX 5 dPDX 6 dPDX 12 dPDX 2 dPDX 4 dPDX 3 dPDX 7 dPDX 10 dPDX 9 dPDX 11 Pt. 9 Relapse EF Pop. 2 6 789 12 13 BCD Pop. 3 Pt. 9 234 Pop. 4 Pop. 5 14 A 1 15 16 17 G H E 8910 567 F 11 C 1 dPDX 7 3451011 subclonal 1 Subclonal Pt. 1 0 prevalence 0 prevalence 12 678912 13 14 G 15 16 17 B Pt. 11 A D CD E Pt. 1 Diagnosis dPDx 26 dPDX 8 dPDX 2 dPDX 15 dPDX 7 dPDX 14 dPDX 20 dPDX 29 rPDX 25 rPDX 2 rPDX 17 rPDX 21 Pt. 1 Relapse

Pop. 1 Pop. 2 H D 56 Pop. 3 Pt. 11 B A 2 Pop. 4 1 C Pop. 5 34 E Pop. 6 7 8 Pop. 7 Pop. 8 Pop. 3 9

Pop. 9 Diagnosis Relapse

Pop. 10 Pt. 11 dPDx 12 dPDX 4 dPDX 6 dPDX 5 dPDX 9 dPDX 1 dPDX 7 dPDX 8 dPDX 2 dPDX 10 dPDX 11 Pt. 11 Pop. 3 1 Pop. 11 Subclonal 0 prevalence Pop. 12 Pop. 13 Pop. 14 Pop. 15 Pop. 16 Pop. 17

1 Patient sample Source Subclonal Diagnosis Patient prevalence Relapse Xenograft 0

574 | CANCER DISCOVERY April 2020 AACRJournals.org

Characterization of Relapse-fated Clones in Diagnosis B-ALL RESEARCH ARTICLE in comparison with the mutational population 2 lineage notype were observed for dRI-PDXs enriched for these latent that represented 71% of the diagnosis leukemic cells (initially variants that were present at low VAFs in the diagnosis patient described in population B; Fig. 4E and F). In the dPDXs, sample (1%–11%; Fig. 5A and B). The dRI-PDXs exhibited population 5 displayed a selective advantage representing at a CD45−CD34+ immunophenotype reflective of the relapse least 93% of the leukemic cells, outcompeting the population sample, whereas others contained both a CD45−CD34+ and 2 lineage (encompassing populations 2–4) and recapitulat- a CD45dimCD34+ population as observed in the diagno- ing the evolutionary dynamics of the patient relapse, where sis patient sample (Fig. 5A). Relapse PDX (rPDX), like the 98% of the leukemic cells came from the population 5 lineage relapse sample itself, were CD45− (Fig. 5A). dPDXs that were (Fig. 4F). In contrast, in 5 patients (patients 1, 4, 6, 7, and 11), the engrafted primarily with CD45dim or both immunopheno- relapse-fated mutational population represented only a small typic populations had a genotype that was more closely percentage of the engrafting cells in most dPDXs and only related to the diagnosis sample (Fig. 5B). The differences rose to predominance in rare mice. For example, in patient noted in immunophenotype appeared to reflect the rise of 11, relapse-fated mutational population 3, initially described the relapse-fated clone and aided in identifying different sub- in population C, corresponded to only 1% of leukemic cells at clones present within a single xenograft. To validate that the diagnosis and remained at similarly low levels in most dPDXs. difference in immunophenotype segregated the predominant However, in two dPDXs transplanted with a low, near limiting diagnosis clone from the minor subclone that seeded the cell dose (10,000 cells), population 3 rose to 92% and 74% of relapse, we isolated the two immunophenotypic populations leukemic cells, distinguishing it from population 4 and high- by flow-sorting cells from 6 dPDXs and subjected the popula- lighting the importance of the limiting dilution approach to tions to targeted sequencing using the custom capture array capture rare subclones by isolating them from other subclones designed for the patient samples. This analysis confirmed with higher competitive capacity (Fig. 4G and H; Supplemen- the enrichment of the dRI clone in the CD45− population, tary Fig. S4E). Population 3 was defined by a single variant whereas latent variants were absent or very rare (<6%) in in the 3′UTR of AHNAK, a gene whose expression has previ- the CD45dimCD34+ population (Fig. 5B). Of note, change in ously been implicated in disease relapse in T-ALL (33). Thus, CD45 expression between diagnosis and relapse timepoints the xenografting strategy permitted integration of functional was identified in 5 of our patient samples. PDX from patient and genomic information, providing insight into the dis- 9 showed markedly reduced leukemic dissemination from the tinct competitive growth properties of functionally defined injected femur to other hematopoietic sites (Supplementary L-IC subclones. Taken together, the phylogenetic analysis and Fig. S6A). mutational evolutionary trees provide strong evidence that The ability to use differences in immunophenotype to relapse-fated subclones were already present in the diagnosis segregate clones enabled us to select evolutionarily related sample, confirming prior studies of several human acute leu- dRI subclones for RNA-sequencing (RNA-seq) analysis from kemic diseases and in silico predictions (4, 5, 28, 29). patient 9. We compared the expression profiles of two dRI clones, an ancestral and a daughter clone (clones 1 and 2, Genetically Diverse Subclones Have Differing respectively), to a representative relapse clone (clone 3). The Immunophenotypic and Migratory Properties dRI clone most genetically similar to the relapse, bearing To further examine whether genetically distinct subclones latent leukemic variants including 3 variants in CREBBP (2 also possessed variation in their immunophenotype or func- missense and 1 silent) as well as variants in TCS2, NRAP, tional properties that might explain why some are fated PLXNA4, and PRINS (dRI-PDX clone 2) had an expression to contribute to relapse and others are not, we examined profile very closely related to representative relapse clones their differentiation, growth, and migration properties. We with only 24 differentially expressed genes (Fig. 5C and first interrogated the differential properties of the diagnosis D; Supplementary Table S4). In comparison, the ancestral clones isolated from patient 9 in which the presence of latent clone not harboring CREBBP (dRI-PDX clone 1CREBBP_WT) variants, including the known relapse driver CREBBP (21), showed 479 differentially expressed genes compared with the was segregated between clones. Differences in immunophe- relapse clones, suggesting that the majority of transcriptional

Figure 4. Generation of mutational trees from the combined genomic data of xenografts and bulk diagnosis and relapse patient samples. Mutational trees of variants clustered to form populations using the PairTree algorithm. Nodes in mutational trees are divided in half, with the intensity of blue in the left half indicating the frequency of the population’s variants at diagnosis, and the intensity of red in the right half showing the frequency of the population’s variants at relapse. Color intensity shows subclonal prevalence as noted in the legend of A and applicable to all other trees except C. A, Mutational tree of patient 1 derived by analysis of the patient samples (diagnosis and relapse) alone. Mutational populations identified from bulk patient samples alone are denoted by a square node labeled with an alphabetical letter. B, Combined mutational tree derived from the variant analysis of both patient 1 patient samples and all their generated xenografts. Mutational populations derived from combined patient and xenograft analysis are represented by circular nodes labeled with numerals. The mutational populations identified using the patient samples alone inA are overlaid on the tree as boxes labeled with their corresponding alphabetical letter. This identifies instances where single populations inA correspond to multiple populations when xenografts are included (i.e., mutational population G). C, Combined mutational tree of patient 1 shaded to indicate the prevalence of variants in dPDX 7 (instead of diagnosis and relapse) demonstrating that this PDX is composed primarily of variants of the relapse lineage. D, Presence of identified mutational populations in patient samples and representative xenografts. Mutational populations (Pop.) are displayed on the y-axis and individual patient samples or xenografts are displayed on the x-axis. The height of the population bar represents the prevalence of the lineage in the patient sample (Pt.) or xenograft. E, Mutational tree, similar to A, derived from patient samples alone of patient 9. F, Combined mutational tree, similar to B, derived jointly from patient 9 patient samples and xenografts. Subclonal prevalences of populations 2 to 5 are shown, indicating the absence of diagnosis populations 2 to 4 and presence of population 5 (the first node in the relapse lineage branch) in all dPDX. G, Mutational tree, similar to A, of patient 11 derived from patient samples alone. H, Combined mutational tree, similar to B, of patient 11 derived from patient samples and xenografts. The prevalence of mutational population 3 is displayed, highlighting its absence in the diagnosis patient and its detection in only two dPDXs.

April 2020 CANCER DISCOVERY | 575

RESEARCH ARTICLE Dobson et al.

A Patient 9 Pt. 9 dPDXPt. 9 rPDX Diagnosis Relapse dPDX 1 dPDX 2 dPDX 3 dPDX 4 dPDX 5 dPDX 7 dPDX 12 rPDX 1rPDX 2

105 8.88 77.5 45.2 46.687.511 53.235.34248.297.641.57.49 0.9971.0296.30.648

104

103

0102 103 104 105

CD34 APC-Cy7 CD34 CD45 PE-Cy7

B C

dPDX 1 dPDX 2 dPDX 3 dPDX 4 dPDX 5 dPDX 7dPDX 12 dRI-PDX clone1CREBBP_WT Diagnosis Relapse dRI-PDX clone2 rPDX clone 3 PHF19 C2orf15 RIMS2 Relative expression TRIL 2 ADAMTS19 FBN1 1 ASH2L AP3B2 0 CREBBP_L1822L PHRF1 −1 Latent variants TSC2 −2 NRAP PRINS CREBBP_L1761 CREBBP_I2101M PLXNA4

Unsorted CD45dimCD34+ 00.1 0.20.3 0.40.5 CD45negCD34+ VAF

D Lymphocyte Negative regulation of activation epithelial proliferation Negative regulation IL12–STAT4 pathway B-lymphocyte pathway TCR signaling of DDR by p53 CSK CFTR

EGFR signaling Leukocyte cell–cell adhesion HIF1α UPR

Alcohol biosynthesis IFNγ Estrogen response

Chromatin remodeling Cell-cycle transition CTCF KRAS signaling Activation of immune response Neural action potential FOXM1 KIT signaling MAPKK activation E2F targets Morphogenesis Antigen processing E CGMP effects dPDX 7 and presentation Response to calcium dRI PDX 11 DCC mediated attractive signaling TOB1 pathway 40

Development TGFβ pathway SPRY pathway NO stimulates guanylate cyclase 30 ) in PB (%) dRI-PDX clone 1CREBBP_WT vs. dRI-PDX clone 2 FDR q-value ≤0.05 + dRI-PDX clone 1CREBBP_WT vs. rPDX clone 3 FDR q-value ≤0.05 20 Edges: Overlapping genes between gene sets or hCD10 + Jaccard overlap combined: 0.375 10

Gene sets enriched in clone 1 clones (hCD19 Gene sets enriched in clone 2 and/or clone 3 Percentage of human engraftment 0 0510 15 20 25 AutoAnnotate clusters Weeks

576 | CANCER DISCOVERY April 2020 AACRJournals.org

Characterization of Relapse-fated Clones in Diagnosis B-ALL RESEARCH ARTICLE changes that are seen at relapse had already occurred in dRI- combined functional and genotyping studies show that indi- PDX clone 2 at diagnosis prior to exposure to chemotherapy vidual subclones possess distinct immunophenotypic, com- (Supplementary Table S4). Broadly, the changes in gene petitive repopulation, proliferative, and migration properties expression of dRI-PDX clone 2 in relation to dRI-PDX clone and that dRI subclones can already possess distinct biological 1CREBBP_WT centered on dysregulation of histone variants and properties at diagnosis, before exposure to chemotherapeutic inflammation-related genes as well as downregulation of agents. genes involved in morphogenesis (TGFβ signaling, GATA3), the T-cell receptor and B-cell receptor pathways (CD19, CD79A, dRI Subclones Display Differential Response CD4, etc.) and antigen processing and presentation pathway, to Chemotherapeutic Agents while heatshock response (HSR) and unfolded protein respone To directly test the functional properties of dRI subclones (UPR) pathways were enriched (Supplementary Table S4). for their ability to survive and contribute to relapse, we com- Thus, the early acquisition of additional leukemic variants pared the drug sensitivities of individual subclones for 5 of in this relapse-fated subclone, including the CREBBP muta- the genetically distinct patients. Multiple secondary PDXs tions, caused significant changes in gene expression. Inter- were generated from dPDXs with known dRI clones (dRI- estingly, tyrosine kinases such as EPHB2, LTK, and ERBB2 PDX), predominant diagnosis clones (dPDX), or representa- were upregulated after acquisition of CREBBP variants in tive relapse clones (rPDX). Treatment of patients with B-ALL both dRI-PDX clone 2 and the relapse clone, suggesting pos- includes combination chemotherapy with supportive care; sible vulnerabilities to tyrosine kinase inhibitors. Despite this however, it is not possible to replicate human therapy pre- striking change, the dRI subclone remained dormant within cisely in xenografts. Our interest was to determine whether the patient for many years, as this patient displayed a long there was any variation in the responsiveness of different sub- remission (4 years). clones to individual drugs used in these chemotherapeutic PDXs from the majority of samples showed extensive migra- protocols. Following engraftment, PDXs received single-agent tion of leukemic cells to other hematopoietic sites and other treatments of dexamethasone, vincristine, l-asparaginase, or tissues including the spleen, CNS, and peripheral blood (Sup- saline for 4 weeks (Fig. 6A and B; Supplementary Fig. S7A– plementary Fig. S6B and S6C). Because patients with B-ALL S7F). Differences in therapeutic responses to one or more may present with leukemic dissemination to the CNS and tes- drugs between dRI and representative diagnosis clones were tes where they can provide a sanctuary for relapse, we inves- observed for four patient samples. In three patient samples tigated whether there were differences in the dissemination (patient 1, 6, 7), dPDXs harboring a dRI clone (dRI-PDX) were of subclonal populations to each site. Targeted sequencing compared with dPDXs repopulated with the predominant analysis showed that there was genetic discordance between diagnosis clone and demonstrated reduced sensitivity to 2 or CNS and bone marrow in 40% (44 of 111) of xenografts and 3 of the 3 chemotherapy agents tested (Fig. 6A and B; Sup- between spleen and BM in 17.8% (28 of 157) of xenografts plementary Fig. S7A–S7D). Reduced sensitivity to a single (Supplementary Table S5). In one patient (patient 7), our chemotherapeutic agent was also observed in one additional analysis of the genetic discordance revealed the presence of a patient (patient 4; dexamethasone, significant in injected dRI clone engrafting in the CNS (VAF > 39%) of a xenograft femur and trend in BM and spleen; Supplementary Fig. transplanted with a high cell dose (250,000 cells), in which S7E). In contrast, there was no difference in the therapeutic the clone was a minor clone barely detectable in the BM (VAF sensitivity of two dRI-PDXs from patient 11 defined by the of < 3%) and outcompeted by a more predominant diagnosis presence of the AHNAK 3′UTR variant. This patient has a clone (Supplementary Fig. S6B). The identification of the strong driver KMT2A (MLL) translocation, and the presence dRI subclone in the CNS of dPDX is consistent with the of the single relapse-specific variant does not appear to confer ability of relapse-fated cells to disseminate and cause disease sufficient evolution of the leukemia to alter the therapeutic recurrence in the CNS of patients with B-ALL. We also noted sensitivity in this context (Supplementary Fig. S7F). Purifica- the occurrence of a difference in peripheral blood dissemination of human cells from the secondary PDXs post-therapy tion of dRI in an additional patient (patient 4), with delayed and targeted sequencing confirmed their genotype and did mobilization of the dRI-PDX subclone as compared with not reveal the selection of any further relapse-specific variants the representative diagnosis clone (Fig. 5E). Collectively, the (Supplementary Fig. S8A–S8C). The observed differences in

Figure 5. Competitive dRI-PDX clones identified in diagnosis PDXs. A, Flow cytometry analysis of patient samples and representative dPDX and rPDX of patient 9 display the presence of two different immunophenotypic populations: a CD45dimCD34+ and a CD45−CD34+. B, Targeted sequencing of the dPDX revealed variability in the VAF of latent variants that corresponded with the shift in immunophenotypic populations. Cell sorting for immunophenotypic populations followed by targeted sequencing revealed the isolation of the latent variants in the CD45−CD34+ population. C, RNA-seq analysis of differentially expressed genes (adjusted p-value of <0.05 and absolute log2 fold change of >1) between the two dPDX (dRI-PDX clone 2; dRI-PDX clone 1CREBBP_WT) versus rPDX for Patient 9. Relative expression was generated from variance stabilized normalized counts. D, Enrichment map of gene sets differentially enriched in patient 9 dRI-PDX clone 1CREBBP_WT versus dRI-PDX clone 2 (FDR q value ≤ 0.05) and dRI-PDX clone 1CREBBP_WT versus rPDX clone 3 (FDR q value ≤ 0.05). Node size is proportional to the number of genes included in each gene set (minimum 10 genes/gene set). Gray and red edges indicate gene overlap. Green node: enrichment in dRI-PDX clone 1CREBBP_WT (positive NES). Purple node: enrichment in dRI-PDX clone 2 and/or rPDX clone 3 (negative NES). Autoannotate app in Cytoscape was used to automatically annotate clusters (black squares). NES, normalized enrichment score; DDR, DNA damage response; TCR, T-cell receptor; UPR, unfolded protein response; cGMP, cyclic guanosine monophosphate; DCC, deleted in colorectal cancer gene; NO, nitric oxide; CSK, C-terminal Src kinase; CFTR, cystic fibrosis transmembrane conductance regulator; SPRY, Sprouty gene. E, Human purified cells from primary dPDX and dRI-PDX from patient 4 were transplanted into secondary NSG recipients. Mice were monitored for peripheral blood human chimerism until mean blood levels reached greater than 10%, revealing different kinetics of chimerism between dPDX 7 and dRI-PDX 11. Symbols represent the mean chimerism value of dPDX 7 (n = 20 mice) and dRI-PDX 11 (n = 16 mice) and bars represent SD.

April 2020 CANCER DISCOVERY | 577

RESEARCH ARTICLE Dobson et al.

Pt. 7 diagnosis

dPDX.20 dRI + **** *** *** Relapse 1.2 rPDX.20 Pt. 7 dPDX 8 rPDX.5 1.0 Pt. 7 dRI dPDX 20 rPDX.12 Pt. 7 rPDX 5 rPDX.9 0.8 rPDX.15 rPDX.11 0.6 rPDX.8 rPDX.6 rPDX.7 0.4 rPDX.10 0.2

rPDX.2 with drug vs. saline

rPDX.4 Ratio of human CD45 rPDX.3 0.0 rPDX.1 engraftment in PDX treated dPDX.7 dPDX.4 dPDX.13 dPDX.2 Vincristine dPDX.3 dPDX.10 L-Asparaginase dPDX.8 Dexamethasone dPDX.5 dPDX.6 dPDX.11 dPDX.14 Diagnosis patient sample dPDX.18 dPDX.12 Diagnosis PDX dPDX.9 Relapse patient sample Relapse PDX

B **** ****

+ ** *** *** Pt. 1 dPDX 2 1.0 Pt. 1 dPDX 5 0.8 Pt. 7 dRI dPDX 7 0.6 Pt. 1 rPDX 2 Pt. 1 rPDX 22 0.4 0.2 with drug vs. saline

Ratio of human CD45 0.0 engraftment in PDX treated

Vincristine -Asparaginase Dexamethasone L

C Patient 1 dRI-PDX Patient 1 dPDX Saline Dexamethasone Saline Dexamethasone 105 99.8 105 72.6 105 99.7 105 97.3

104 104 104 104

103 103 103 103

0 0 0 0 0.0477 26.3 0.0139 1.82 0102 103 104 105 CD19 PE CD33 PE-Cy7

Figure 6. dRI subclones display decreased sensitivity to commonly used chemotherapeutic drugs. A, Phylogenetic analysis showing clonal relationships in patient 7 based on VAF of leukemic variants shows clear evidence of the isolation of a relapse-fated, dRI clone in dPDX 20. The distances between symbols on the tree were estimated by a nearest neighbour joining method and represent the degrees of relation between them (Minkowski’s distance). Circles represent patient samples and triangles represent PDXs; blue represents diagnosis and red represents relapse. dRI-PDX 20 is indicated by a hatched border box. Purified human cells from primary dPDX 7, dRI-PDX 20, and rPDX 5 (representative relapse genetics) xenografts were injected into secondary NSG mice and allowed to engraft. Mice were randomized into 4 groups (with 4 to 5 mice per group) and treated with either saline, dexamethasone, l-asparagine, or vincristine. After 4 weeks of treatment mice were sacrificed and engraftment in the IF, BM, and spleen were analyzed by flow cytometry. Ratio of human chimerism in the BM of drug-treated mice in comparison with saline controls is shown. B, Ratio of human chimerism in the BM of drug-treated mice in comparison with saline controls of dPDX, dRI-PDX, and rPDX of patient 1. C, Representative flow plots of the CD19 and CD33 immunophenotype of dPDX and dRI-PDX dexamethasone-treated mice from patient 1. Lines represent mean and SD. Only significance between dPDX and dRI-PDX are shown. **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; unpaired two-sided t tests. drug response could not be accounted for by any consistent lation upon dexamethasone treatment but not in saline con- changes in L-IC frequency (Supplementary Table S6). trols or upon treatment with the other two drugs (Fig. 6C; Unexpectedly, in 2 of 5 patient samples used for drug Supplementary Fig. S9A and S9B). This population was testing, we found that the dRI subclones exhibited immu- very rare or not observed in the primary recipients. The nophenotypic plasticity as compared with the predominant CD33+CD19dim/− cells resembled myeloid cells with respect diagnosis clone when exposed to drugs. In the secondary to their cell surface marker expression, size, and granularity recipients of dRI-PDX of patient 1 and dRI-PDX of patient 7, (Supplementary Fig. S9C). An immunophenotypic shift both patients harboring DUX4 translocations at diagnosis, toward the myeloid lineage especially with steroid chal- there was the emergence of a distinct CD33+CD19dim/− popu- lenge has previously been reported in ERG/DUX4 patients,

578 | CANCER DISCOVERY April 2020 AACRJournals.org

Characterization of Relapse-fated Clones in Diagnosis B-ALL RESEARCH ARTICLE but our finding of a subclone-specific switch is of interest checkpoints, DNA replication, DNA repair, and microtu- (34). The emergence of CD33+CD19dim/− cells was not due bule organization (Fig. 7A; Supplementary Table S7). to the outgrowth of a different subclone, as these cells were To validate the metabolic signature observed in dRIs that genetically identical to their CD19+ counterpart (Supple- was often further enhanced at relapse, we performed flow mentary Fig. S9D and S9E). We speculate that this propen- cytometry and immunostaining analysis in the PDX from sity for immunophenotypic plasticity might be linked to five patient samples (Fig. 7C and D; Supplementary Fig. S10E chromatin remodeling and the ability to evade therapy, as and S10F). This analysis confirmed the convergence of simi- other studies have shown (35–38). Overall, our data provide larities in metabolic rewiring of dRI-PDX/rPDX in compari- strong evidence that prior to exposure to chemotherapy, dRI son with dPDX in all 5 patients with B-ALL. This functional subclones already possess distinct preexisting functional validation showed that there was an increase in total mito- properties including differential sensitivity to standard chondrial mass, with individual mitochondria having similar chemotherapy agents. mitochondrial membrane potential in most of the patients analyzed. However, despite this increase in absolute mito- Transcriptional Analysis of dRIs Reveals Metabolic chondria abundance, the levels of reactive oxygen species Rewiring and Enrichment of a Stem-Cell State in (ROS) were unexpectedly found to be lower in dRI-PDX and Progression to Relapse rPDX cells as compared with dPDX cells. The reduced ROS To gain mechanistic insight into the molecular pathways levels are suggestive of the presence of increased antioxidant present in dRIs, RNA-seq analysis was performed on cells defense upon progression toward relapse. This interpretation from dRI-PDX, dPDX, or rPDX of 4 patients (patients 1, 4, 6, was concordant with our gene-expression data that showed 7; n = 14 dPDXs, 15 dRI-PDXs, and 13 rPDXs; Supplementary an enrichment of ROS-defense and peroxisomal activity Table S7). This analysis confirmed the placement of dRIs as genes (i.e., catalase), which degrade toxic hydrogen peroxide intermediates between diagnosis and relapse, sharing tran- and metabolize drugs, in the dRI-PDX and rPDX samples scriptional programs with both timepoints (Supplementary (Fig. 7A and C; Supplementary Fig. S10B, S10D–S10F; Sup- Fig. S10A). Given the distinct B-ALL subtypes/cytogenetics of plementary Table S7). Significant enrichment for chromatin the patient samples analyzed, as expected only a few (n = 23) remodeling and cell stress response (such as the UPR) were differentially expressed genes reached significance when com- also identified as dRI-PDX/rPDX common pathways (Fig. paring dPDX versus dRI-PDX across all samples (Supplemen- 7A; Supplementary Fig. S10B–S10D). The chromatin remod- tary Table S7). Surprisingly, one of the genes upregulated in eling pathways including expression of histone variants and the dRI-PDX is asparaginase (ASPG), a catalytic enzyme that isoforms were likely instrumental in the plasticity observed hydrolyzes asparagine to aspartic acid, albeit nonrobustly in in the progression to relapse, including the immunophe- human cells. The human ASPG protein can display cytotoxic notypic plasticity we observed in dRI-PDXs of patients 1, activity in human leukemic cell lines (39), suggesting that the 7, and 9 (Figs. 4A and 5C), whereas the expression of the dRI-PDX cells may have altered their response to cytotoxic cell stress response pathways identified could contribute to stress or metabolic requirements, thereby explaining why enhanced survival of dRI subclones. To independently vali- some of the dRI clones are less sensitive to l-asparaginase date the results obtained from the PDX-identified pathways, treatment than dPDX clones. we directly evaluated bulk diagnosis and relapse patient sam- As pathway enrichment analysis is more sensitive than ples. We found enrichment (FDR q-value ≤ 0.05) of pathways differential gene expression for finding differences between involved in metabolism, mitochondrial regulation, and cell populations, we undertook this approach to uncover sig- cycle at relapse (Fig. 7E; Supplementary Fig. S10G; Supple- nificantly differentially enriched pathways that were shared mentary Table S7). Furthermore, we found significant enrich- among all patients. Gene set enrichment analysis (GSEA) ment of a signature present at minimal residual disease (18) comparing the pathways present in dRI-PDX versus dPDX in dRI-PDX and rPDX compared with dPDX, lending further revealed that most pathways significantly enriched in dRI- support for the relevance of the dRI-PDX/rPDX pathways PDX (FDR q-value ≤ 0.05) were also present or further for understanding the mechanisms of relapse disease initia- enriched in the comparison of rPDX to dPDX; these are tion and their presence prior to chemotherapeutic challenge termed dRI-PDX/rPDX common pathways (Fig. 7A and B; (Supplementary Fig. S10H). Collectively, our transcriptional Supplementary Fig. S10B–S10D; Supplementary Table S7). analysis has for the first time uncovered the utilization of Network analysis revealed that these pathways were cen- chromatin remodeling, stress responses, and metabolic path- tered on a metabolic signature composed of genes involved ways in dRI subclones whose activity could serve to protect in cellular and mitochondrial metabolism including amino them during chemotherapy treatment and contribute to their acid metabolism, tricarboxylic acid cycle, oxidative phos- ability to further progress to relapse disease. As such, these phorylation, mitochondrial translation and transport, newly identified pathways represent rich areas to investigate and lipid metabolism (Fig. 7A and B; Supplementary Fig. for new therapeutic strategies to target dRI specifically. S10B–S10D). In concordance with a previous study on ALL To investigate whether the broader cellular state of the dRIs relapse (26), the central regulator of growth and metabo- contributes to their functional differences, we performed gene lism, mTOR, was enriched in dRI and relapse clones (Sup- set variation analysis (GSVA) comparing the transcriptomic plementary Fig. S10B; Supplementary Table S7). Pathways profiles of representative clones with normal hematopoietic identified as uniquely enriched in rPDX versus dPDX (rPDX- cell populations isolated from human umbilical cord blood unique) included a large network of highly interconnected using two independent datasets, one from our own newly nodes involved in cell-cycle regulation such as cell-cycle generated data and the other published previously (ref. 40;

April 2020 CANCER DISCOVERY | 579

RESEARCH ARTICLE Dobson et al.

Clone category A B Mitochondrial translation OXPHOS Pathway name NES dPDX dRI-PDX HALLMARK_UNFOLDED_PROTEIN_RESPONSE CELLULAR RESPONSE TO TOPOLOGICALLY INCORRECT PROTEIN rPDX CELLULAR RESPONSE TO UNFOLDED PROTEIN ASPARTATE FAMILY AMINO ACID CATABOLIC PROCESS MRPL33 COX5A Relative expression HALLMARK_MYC_TARGETS_V2 MRPS2 1 MRPL23 NDUFA4 VALIDATED TARGETS OF C-MYC TRANSCRIPTIONAL ACTIVATION MRPS6 ATP5J HALLMARK_P53_PATHWAY MRPS31 UQCRQ MRPS26 0.5 EPIGENETIC REGULATION OF GENE EXPRESSION MRPS18B NDUFB9 NDUFA3 RMTS METHYLATE HISTONE ARGININES MRPS23 MRPL22 ATP5H 0 CHROMATIN SILENCING AT RDNA MRPL10 MRPL49 NDUFB1 HALLMARK_REACTIVE_OXIGEN_SPECIES_PATHWAY MRPL17 NDUFC1 dRI dPDX-specific MRPL35 STOML2 −0.5 0.00.5 1.01.5 2.02.5 MRPS11 NDUFS1 MRPL36 ERAL1 NDUFS2 MITOCHONDRIAL TRANSLATION ELONGATION APEH COX10 −1 MITOCHONDRIAL TRANSLATION MRPL40 NDUFB3 MRPL3 SURF1 TRICARBOXYLIC ACID METABOLIC PROCESS MRPL15 MRPS18C COX15 HALLMARK_OXIDATIVE_PHOSPHORYLATION MRPL18 NDUFAF1 DNA METHYLATION MRPS14 COA6 MRPL13 PRC2 METHYLATES HISTONES AND DNA MRPL47 SDHA CHROMOSOME MAINTENANCE MRPL14 SDHC MRPL46 ATP5A1 G1 S TRANSITION MRPS10 ATP5F1 common EIF2D DNA REPLICATION MRPL34 UQCRC1 HALLMARK_INTERFERON_ALPHA_RESPONSE MRPL16 COX8C DAP3

dRI dPDX/rPDX NDUFB10 PEROXISOME ORGANIZATION PELO MRPL19 COX5B NDUFV3 0 1 23 MRPL57 MRPS22 ATP5J2 MRPL42 UQCR10 HALLMARK_E2F_TARGETS MRPS18A MRPL50 NDUFA10 CELL CYCLE, MITOTIC MRPS28 ATP5C1 MCTS1 NDUFS4 DNA STRAND ELONGATION MRPL2 NDUFA5 MITOTIC METAPHASE AND ANAPHASE MRPL28 MRPS12 NDUFB6 MITOTIC SISTER CHROMATID SEGREGATION ABCE1 CYC1 MRPL1 G2 M CHECKPOINTS%REACTOME MRPS16 CYCS DOUBLE-STRAND BREAK REPAIR MRPS34 SDHD MRPL32 NDUFA8 ACTIVATION OF ATR IN RESPONSE TO REPLICATION STRESS MRPS36 NDUFAB1 MRPL9 rPDX-specific CELL CYCLE DNA REPLICATION MRPS15 NDUFS8 MRPL52 ATP5B 0 1 23 MRPL51 NDUFA12 MRPL11 CHCHD1 UQCC3 dPDX vs. dRI-PDX FDR q-value < 0.05 GADD45GIF COX6A1 MRPL44 NDUFB4 MRPS35 NDUFB5 dPDX vs. rPDX FDR q-value ≤ 0.05 MRPL4 CD Patient 1 Patient 4 Patient 7 2.5 2.0 1.5 dPDX * ** 2.0 1.5 1.5 1.0 1.0 1.0 0.5 dRI-PDX

Ratio MFI MitoTracker 0.5 0.0 0.5

1.5 1.5 1.5 rPDX SOD1 TOMM20 MRPS18b DAPI 2 µm

1.0 1.0 SOD1 TOMM20 MRPS18b Patient 1 1.0 2.0 2.0 2.0 ** Patient 3 * *** Patient 7 0.5 0.5 1.5 1.5 1.5 dPDX

Ratio CellRox MFI dRI-PDX * ** 1.0 1.0 1.0 0.5 0.0 **** 0.0 rPDX Ratio IntDen dPDX dRI-PDX rPDX 0.5 0.5 0.5

E F dPDX vs. dRI-PDX dPDX vs. rPDX Dx vs. Rel Pt. samples

0.0 0.0 0.0 −0.1 −0.1 −0.1 0.2 − −0.2 NES −2.5342 NES −2.085 0.2 NES −1.9456 −0.3 − 100 −0.3 100 −0.4 −0.3 −0.4 0.5 1 − −0.4 −0.5 1 −0.6 −0.5 −0.6 −0.7 Enrichment score (ES) 100 −0.6 100 dPDXdRI-PDX dPDX rPDXDx Rel 0.8

‘na_pos’ (positively correlated) ‘na_pos’ (positively correlated) 7.5 ‘na_pos’ (positively correlated) 0.8 5 10 5.0 5 0 2.5 100 Zero cross at 25670 0 0.0 translation Zero cross at 20577 Zero cross at 10661 0.6 −5 −5 −2.5 −5.0 Mitochondrial −10 −10 −7.5 ‘na_neg’ (negatively correlated) −15 ‘na_neg’ (negatively correlated) ‘na_neg’ (negatively correlated) 0.6 0.05 05,000 10,000 15,000 20,000 25,000 30,000 35,000 40,000 45,000 05,000 10,000 15,000 20,000 25,000 30,000 35,000 40,000 45,000 02,500 5,0007,500 10,000 12,500 15,000 17,500 20,000 Rank in ordered dataset Rank in ordered dataset Rank in ordered dataset

Ranked list metric (PreRanked) Enrichment score (ES) Enrichment profile Hits Ranking metric scores Ranked list metric (PreRanked) Enrichment score (ES) Enrichment profile Hits Ranking metric scores Ranked list metric (PreRanked) Enrichment profile Hits Ranking metric scores 0.4 B genes 0.05

0.05 HSC genes 0.0 0.0 0.0 0.4 −0.1 −0.1 NES 2.0108 −0.1 NES 2.1054 0.2 0.05 0.2 − − − NES −2.4413 −0.2 −0.2 −0.3 −0.3 −0.3 −0.4 0.05 −0.4 −0.5 −0.4 0.0 −0.5 0.0 −0.6 0.5

Enrichment score (ES) Enrichment score (ES) Enrichment score (ES) − −0.6 Dx Rel dPDx dRI rPDX aggregated enrichment score aggregated enrichment score Dx Rel dPDx dRI rPDX dPDXdRI-PDX dPDX rPDXDxRel PDX PDX ‘na_pos’ (positively correlated) ‘na_pos’ (positively correlated) ‘na_pos’ (positively correlated) 5 10 5 5 Pt samples Pt samples 0

OXPHOS 0 Zero cross at 25670 Zero cross at 20577 0 Zero cross at 10661 −5 −5 −10 −5 −10 ‘na_neg’ (negatively correlated) −15 ‘na_neg’ (negatively correlated) ‘na_neg’ (negatively correlated) 05,000 10,000 15,000 20,000 25,000 30,000 35,000 40,00045,000 05,000 10,000 15,000 20,000 25,000 30,000 35,000 40,000 45,000 02,500 5,000 7,500 10,000 12,500 15,000 17,500 20,000 Rank in ordered dataset Rank in ordered dataset Rank in ordered dataset Ranked list metric (PreRanked) Ranked list metric (PreRanked) Ranked list metric (PreRanked) Enrichment profile Hits Ranking metric scores Enrichment profile Hits Ranking metric scores Enrichment profile Hits Ranking metric scores

G Diagnosis Relapse Chemotherapy

dRI

Stemness Metabolic rewiring

580 | CANCER DISCOVERY April 2020 AACRJournals.org

Characterization of Relapse-fated Clones in Diagnosis B-ALL RESEARCH ARTICLE

Supplementary Fig. S11A; Supplementary Table S7). dRI- analysis of bulk tissues, it also extends our understanding of PDX and rPDX exhibited a transcriptome profile significantly leukemia evolution by revealing a richer diversity of branching enriched for hematopoietic stem cell (HSC) genes and a slight evolution and additional subclones. Our functional studies reduction in B-cell genes as compared with dPDX (Fig. 7F; Sup- of subclones analyzed either alone or in the context of com- plementary Fig. S11B and S11C). Depletion of B-cell genes and petitive repopulation demonstrate that dRI subclones can enrichment of HSC genes at relapse were also observed in the be found to display increased clonal repopulation kinetics, bulk patient samples from our cohort (Fig. 7F). Furthermore, immunophenotypic plasticity, and differences in leukemic the leading-edge genes of common pathways enriched in both dissemination such as dissemination to the CNS, a known dRI-PDX and rPDX are upregulated in HSCs as compared with disease sanctuary site in patients. Importantly, dRI showed an lymphoid cells (Supplementary Fig. S11D). The enrichment intrinsic difference in tolerance to standard induction chemo- of mitochondrial metabolism and stemness at relapse were therapeutic agents. Moreover, dRI subclones were typically also validated in the larger Waanders and colleagues cohort very minor latent subclones in the diagnostic sample, suggest- of paired diagnosis-relapse patient B-ALL samples (7) encom- ing that they are restrained within the patient in comparison passing several genetic subtypes (Supplementary Fig. S11E and to the predominant subclone, perhaps due to dormancy, S11F). These findings are in line with previous studies in the timing, subclonal cross-talk, or slow cycling, although this normal and leukemia stem-cell fields (25, 41, 42) where several was not directly measured in our studies. Our study resolved metabolic and stress response pathways such as those identified the theoretical question as to whether functional properties here (mitochondria metabolism, UPR, antioxidant defense) have of relapse clones are present prior to chemotherapeutic chal- been described to be crucial to the maintenance of stem-cell lenge or are induced by chemotherapy. Although some muta- homeostasis and function. In addition, enrichment of stemness tions not present at diagnosis are recurrently acquired after signatures is also a hallmark of high-risk B-ALL (43). Thus, our chemotherapy (e.g., USH2A, NT5C2), our study documents findings report a link for the acquisition of HSC stemness prop- the preexistence of dRI subclones with intrinsic tolerance erties in combination with metabolic rewiring in dRI as part of to clinical drugs. This data provides direct evidence that the progression to relapse in lymphoid B-ALL (Fig. 7G). Luria–Delbrück principle that resistance in a cell population may be intrinsic occurs in human leukemia (44). We speculate that the dRI subclones arise stochastically and these partially Discussion evolved subclones then become selected for by chemotherapy. Our study provides new insight into the leukemogenic Gene-expression and pathway analysis showed that even at process of human B-ALL through a deep analysis of the func- this early stage where relapse fate arises, there was consider- tional and molecular properties of genetically diverse diagno- able metabolic rewiring within dRI subclones that was often sis subclones isolated through xenografting from pediatric further enhanced in relapse clones. Recurrent pathways were and adult patients with B-ALL. By combining xenografting identified across a spectrum of different patients with B-ALL with sequencing, broad clonal structures were unambiguously with distinct genetic drivers, suggestive of convergent evolu- demonstrated and additional subclones were uncovered. We tion. Broadly, we found that adult and pediatric dRI had identify relapse-fated dRI subclones, prior to chemothera- similar properties in terms of their functional xenografting peutic exposure, with the capacity for clonal propagation and properties, therapy response, transcriptional pathways, and leukemia initiation in B-ALL that are both genetically and metabolic rewiring, also supporting the concept of conver- transcriptionally related to the relapse. Our data extends prior gent evolution. Pathways regulating mitochondrial dynamics B-ALL studies that relied on genetic analysis and computa- and proteostasis are particularly relevant because they are tional methods of bulk leukemia samples without functional linked to survival under stress, processes that may underpin studies, and that could only infer the presence and survival the ability of dRI cells to adapt and survive chemotherapy. of minor subclones that were ancestral to the relapse driv- Further evolution and mutagenesis of such subclones during ing clone and could not speak to the functional properties of the treatment and/or remission phases ultimately results in these cells at diagnosis (4, 5). As shown by Waanders and col- disease recurrence. Expression of the dRI-PDX/rPDX common leagues (7), although our xenografting approach substantiates pathways converged on an HSC signature, shared by both dRI the predicted lineage relationships of the traditional genetic and relapse cells, analogous to findings from recent studies

Figure 7. dRI subclones share a common metabolic and stem cell profile. A, Plot showing the normalized enrichment score (NES) for the top differentially enriched gene sets (FDR q value ≤ 0.05) of dRI-PDX unique, dRI-PDX/rPDX common, and rPDX unique groups from GSEA of the following comparisons: dPDX vs. dRI-PDX and dPDX vs. rPDX. B, Heat maps showing the expression of leading-edge genes (subset of genes found in the ranking at or just before the maximal enrichment score in GSEA) for selected gene sets enriched in dRI-PDX and rPDX from enrichment map in A. Relative expression was generated from variance stabilized normalized counts C, dPDX, dRI-PDX, and rPDX from Patient 1, Patient 4, and Patient 7 were stained with MitoTracker and CellROX dyes and analyzed by flow cytometry. MFI for each sample and dye is represented as ratio to dPDX for each patient (Patient 1: dPDX n = 5, dRI-PDX n = 5, rPDX = 4; Patient 4: dPDX n = 5, dRI-PDX n = 4, rPDX = 4; Patient 7: dPDX n = 5, dRI-PDX n = 5, rPDX = 5). D, Immunostaining analysis for TOMM20, MRPS18B, and SOD1 in dPDX, dRI-PDX, and rPDX cells from Patients 1, 3, and 7. The Integrated Density (IntDen = Area × MFI) for 40 to 50 cells from each clone was analyzed using Fiji. The mean for each clone was normalized and calculated as a ratio to the dPDX for each patient separately. Representative images for Patient 7 are shown. (Patient 1: dPDX n = 1, dRI-PDX n = 2, rPDX = 1; Patient 7: dPDX n = 2, dRI-PDX n = 2, rPDX = 1; Patient 3: dPDX n = 2, dRI-PDX n = 2, rPDX = 1). E, GSEA enrichment plots from the following comparisons: dPDX vs. dRI-PDX (n = 4 pts); dPDX vs. rPDX (n = 4 pts); and diagnosis (Dx) versus relapse (Rel) patient samples from our cohort were generated for mitochondrial translation and oxidative phosphorylation gene sets. F, Bar plot of the aggregated GSVA scores for B-cell genes and HSC genes in each sample. GSVA scores for samples in each category were summed and scaled from 0 to 1. The numbers above the bars represent how many times the observed score was higher than random scores obtained in 1,000 permutations using a list of 1,000 random genes. G, Schematic diagram of dRI with altered metabolic and stem-cell programs preexisting in diagnosis patient samples that survive chemotherapy and seed the relapse disease. *, P < 0.05; **, P < 0.01; ***, P < 0.001; unpaired two-sided t test.

April 2020 CANCER DISCOVERY | 581

RESEARCH ARTICLE Dobson et al. we have undertaken from purified AML stem-cell populations ing antibodies: anti-CD19 PE (BD Biosciences, clone 4G7), anti-CD3 (24, 45). The finding of stemness signatures already present in FITC (BD Biosciences, clone SK7) or anti-CD3 APC (Beckman Coul- dRI subclones, together with the identified metabolic changes, ter clone UCHT11), anti-CD45 APC (BD Biosciences, clone 2D1) provides the molecular basis to explain why dRI subclones or anti-CD45 FITC (BD Biosciences, clone 2D1), and anti-CD34 may both survive therapy and possess the regenerative capacity APC-Cy7 (clone 581). Each sample was sorted on a FACSAria III (BD Biosciences) for leukemic blasts (CD19+CD45dim/−) and T cells (CD3+ to initiate disease relapse after a period of dormancy. Future CD45hi). NSG mice were bred according to protocols established and studies that explore the transcriptome of dRI subclones from approved by the Animal Care Committee at the University Health across an even wider spectrum of B-ALL patient samples would Network. Eight- to 12-week-old mice were sublethally irradiated at give insight into how the timing of acquisition of these molec- 225 cGy 24 hours before transplant. Only female mice were used in ular pathways leads to relapse. these studies. Intrafemoral injections of 10 to 250,000 sorted leuke- The ability to isolate and characterize dRI subclones pro- mic blasts were performed as described previously (48). Mice were vides an important first step in understanding the basis for sacrificed 20 to 30 weeks post-transplant or at the onset of disease therapy resistance and clonal propagation. Such subclones symptoms. Human cell engraftment in the injected femur, bone isolated in this way are not simply an in silico depiction; rather, marrow (noninjected bones, left tibia, right tibia, left femur), spleen, they can be viably preserved cells or serially propagated xeno- and CNS were accessed using human specific antibodies for CD45 (PE-Cy7, BD Biosciences, clone HI30; v500, BD Biosciences, clone grafts for future studies. Whole-genome sequencing, methyla- HI30), CD44 (PE, BD Biosciences, clone 515; FITC, BD Biosciences, tion, and chromatin accessibility studies could be undertaken clone L178), CD3 (APC, BD Biosciences, clone UCHT1), CD19 (PE- to build upon our transcriptomic analysis, thereby explor- Cy5, Beckman Coulter, clone J3-119), CD33 (PE-Cy7, BD Biosciences, ing more deeply the mechanisms that drive their functional clone P67-6; APC, BD Biosciences, P67-6), and CD34 (APC-Cy7, BD properties. The signatures developed from isolated dRI clones Biosciences, clone 581) analyzed on an LSRII (BD Biosciences). Mice could yield biomarkers to improve the classification of patients were considered to be engrafted when >0.1% of cells in the injected who are at increased risk of relapse and to better monitor femur were positive for one or more human B-ALL specific cell sur- residual disease. The immunophenotypic plasticity we linked face markers (CD45, CD44, CD19, CD34). The confidence intervals to different genetic subclones points to the need to under- for the frequency of L-ICs was calculated using ELDA software (49). stand the breadth of cell-surface phenotypes present to ensure WES that all leukemic subclones are properly tracked during flow cytometry–based residual disease monitoring. Finally, further DNA from the adult samples was extracted from sorted leu- + dim/− investigation of the dRI transcriptomic profile and metabolic kemic blasts (CD19 CD45 , purity >90%) and sorted T cells (CD3+CD45hi, purity >90%) using the Gentra Puregene Cell Kit rewiring may be used to uncover the vulnerabilities of dRI (Qiagen). T cells from patients 2, 3, and 5 were whole-genome ampli- subclones, resulting in new therapeutic targets. Improved fied (REPLI-g Mini Kit, Qiagen) due to limited material. DNA from eradication of dRIs during early treatment phases before the the pediatric samples with >90% leukemic blasts was extracted from subclones evolve would prevent progression to more aggressive, diagnosis, remission, and relapse samples using phenol–chloroform. therapy-resistant disease. Prior studies from us and others have Exomes were captured using the TruSeq Exome Library Prep Kit shown that xenografting of a wide spectrum of primary patient (67 Mb, 1 μg DNA input) or the Nextera Rapid Capture Expanded samples provides a powerful tool to evaluate novel therapeu- Exome (62 Mb, 50 ng DNA input; Illumina). Paired-end sequenc- tics and develop biomarkers (23, 25, 46, 47). Our study sug- ing was performed with the HiSeq 2500 genome sequencer (Illu- gests that extending the xenograft-based drug development mina). The data was mapped to human reference genome hg19 and paradigm by including genetic analysis to uncover subclonal variant calling was performed using the Bambino variant detector as described previously (50). Briefly, leukemia and germline files were responses to drug treatment will open up avenues to evaluate combined by the program and aligned against the reference genome. whether relapse-fated clones are effectively targeted. Putative sequence variants including SNVs and indels were detected by running the variation detection module of Bambino. The output Methods contained detailed read counts for each variant with columns for tumor/normal status, allele, and strand. Variants were not filtered Patient Samples for coverage prior to combination with targeted-sequencing results. Patient samples were obtained at diagnosis and relapse from 6 adult patients and 8 pediatric patients with B-ALL according to Copy-Number Analysis preestablished guidelines approved by the Research Ethics Board of Patient copy-number aberrations were determined using SNP6.0 the University Health Network and the St. Jude Institutional Review microarrays according to manufacturer’s instructions (Affymetrix). Board, respectively, and were conducted in accordance with recog- Data was analyzed as described previously (51) using optimal refer- nized ethical guidelines. Adult samples were collected at the Princess ence normalization (52) and circular binary (53, 54) segmentation Margaret Cancer Centre (Toronto, Ontario, Canada) and pediatric with Genotyping Console (Affymetrix) and dCHIP (build Apr 2010; samples were collected at St. Jude Children’s Research Hospital ref. 55). Detection of loss of heterozygosity and allelic ratios were (Memphis, TN). Written informed consent was obtained from all performed using Nexus 7.5.2 software (BioDiscovery Inc). All seg- patients or patient families. All samples were frozen viably and stored ments were manually curated. long term at −150°C. Samples were selected retrospectively based on sample and paired sample availability. Recovery of Human Cells and DNA Isolation from Xenografts PDX Generation and Limiting Cells from the injected femur (IF), BM, and spleen were frozen via- Dilution Analysis bly after sacrifice. The IF and BM of mice engrafted with> 10% human Twenty-nine clinical samples obtained from the 14 patients cells were combined. These cells as well as cells from diagnosed PDX (1 patient having two relapse samples) were stained with the follow- spleens were depleted of mouse cells using the Miltenyi Mouse Cell

582 | CANCER DISCOVERY April 2020 AACRJournals.org

Characterization of Relapse-fated Clones in Diagnosis B-ALL RESEARCH ARTICLE

Depletion Kit (Miltenyi Biotec; samples with >20% engraftment) or their mutual “A and B occurred in the same cell” probabilities were by cell sorting with human CD45 and human CD19 and or CD34 high. cell-surface antibodies to a purity of >90% as determined by post- For each cluster, a representative “supervariant” was created. To processing flow cytometry. CNS cells from mice with greater than compute a precise estimate of supervariant CCF, we summed the 60% engraftment were used directly for DNA isolations. DNA was read counts of all variants within the supervariant’s cluster, then isolated using the QIAamp DNA Blood Mini or Micro Kit (Qiagen). computed the pairwise ancestral relationship probabilities between supervariants. These probabilities guided expert-driven tree con- Targeted Sequencing struction, in which trees were built according to how well their implied pairwise relationships between clusters matched the com- All somatic SNVs and indels identified by WES were validated in puted pairwise probabilities. Confirmation that these expert-driven the patient samples using NimbleGen SeqCap Target Enrichment trees yielded data likelihoods as good as or better than fully auto- according to the manufacturer’s instructions (Roche, NimbleGen). mated tree reconstructions produced by a more mature version of the Library preparation was completed using 250 to 500 ng of DNA PairTree algorithm (Wintersinger and colleagues, in preparation) was using the NEXTflex DNA-SEQ Library Prep Kit (BiooScientific) with performed. The tree-constrained lineage frequencies of each popula- NEXTflex-96 DNA Barcodes (BiooScientific). Sequencing was per- tion in each sample were then computed using PairTree. The variants formed on a HiSeq 2500 genome sequencer to a mean coverage included in the populations defined by the patient-only analysis were >350× for patient samples and >200× for PDXs. then compared to the variants included in the populations defined Targeted-Sequencing Data Analysis for the patient and xenograft combined analysis to reveal the overlap, differences, and additional clonality described with inclusion of the Final patient variant calls used the combined results of WES xenograft samples. and capture validation. Variants were filtered out if the VAF in the germline was greater than 10% or if there was a dbSNP frequency of greater than 1%. Variants were classified based on the VAFs in the RNA-seq bulk patient diagnosis and relapse samples as: preserved variants For PDX. Cell pellets for PDX RNA extraction were frozen in (VAF > 30% in both diagnosis and relapse samples, or preserved between 1 mL of TRIzol (Invitrogen) and kept at −80°C. Total RNA was purified samples); diagnosis-specific variants [present at diagnosis >( 1%) and by phenol/chloroform and integrity and concentration were verified absent at relapse (<1%)]; latent variants (present at diagnosis with on a Bioanalyzer Pico Chip (Agilent Technologies). cDNA conver- VAF < 30% and increasing at relapse); relapse-specific variants [absent sion was performed using SMART-Seq v4 Ultra Low Input RNA Kit at diagnosis (<1%) and present at relapse (>1%)]. For xenograft analy- for Sequencing (Takara; 1 ng total RNA input) and libraries were sis, variants with less than 5× coverage or uncovered in numerous prepared using Nextera XT DNA Library Preparation Kit (Illumina; xenografts were removed. Results were analyzed by visualization in 1 ng input of cDNA). Equimolar quantities of libraries were pooled heat maps (i.e., Fig. 1C). Phylogenetic analysis showing genetic rela- and sequenced 4 cDNA libraries per lane on a High Throughput Run tionships of patient samples and xenografts were estimated using Mode Flowcell with v4 sequencing chemistry on the Illumina HiSeq Minkowski’s distance calculated from the VAFs and represented 2500 following manufacturer’s protocol generating paired-end reads visually using a nearest neighbor joining algorithm. Genetic concord- of 126-bp in length to reach depth of 65 million reads per sample. ance between different tissues of the same xenograft were determined by visual assessment by three independent and blinded individuals. For Patient Samples. RNA was extracted from sorted leukemic Discordance was called only when all three investigators were in blasts (CD19+CD45dim/−, purity >90%) using RNeasy Micro Kit agreement. (Qiagen) for adult patient samples. From these samples approximately 10 ng of total RNA was processed using the SMART cDNA synthesis Generation of Mutational Trees from Patient protocol including SMARTScribe Reverse Transcriptase (Clontech) Samples and Xenografts as per the manufacturer’s instructions. The amplified cDNA was sub- Two independent computational analyses were performed—first ject to automated Illumina paired-end library construction using the for patient-only tissue samples, and then for patient samples aug- NEBNext paired-end DNA sample Prep Kit (NEB) and libraries were mented with xenografts (BM and spleen)—using an early version of sequenced on HiSeq2000 (Illumina) to an average of approximately the PairTree algorithm (Wintersinger and colleagues, in preparation). 161 million Chastity-passed paired reads of 75 bp in length per sam- PairTree uses variant read counts to estimate the posterior probabil- ple. For pediatric samples RNA was extracted from patient samples ity distribution over four possible ancestral relationships between with >90% leukemic blasts using TRIzol (Life Technologies) and every variant pair (A, B). The four ancestral relationships are as fol- quality and quantity assessed by Qubit (Thermo Fisher Scientific) lows: variant A and variant B occurred in the same cells, such that no and RNA6000 Chip (Agilent). One microgram of RNA was used for cell possessed one variant but not the other; variant A is ancestral to library preparation with TruSeq RNA Library Prep Kit v2 (Illumina) variant B, such that some cells have A but not B; variant B is ancestral and 2 × 100 bp paired-end sequencing was performed on a HiSeq to variant A, such that some cells have B but not A; or neither is the 2500 (Illumina). Patient RNA-seq samples were aligned against the ancestor of the other, such that A and B are on different branches of human genome (hg19) using STAR v2.3 with default parameters. All the evolutionary tree. PDX samples were aligned with STAR v2.5.2b (56) against the human To permit temporal ordering of mutations, the infinite sites genome build version GRCh38 and the ensembl v90 gtf file. Default assumption was made, such that variant A could never be the ances- parameters were used except chimeric segments were screened with tor of variant B if A’s cancer cell frequency (CCF) was lower than B’s a minimum size 12, junction overlap 12, and segment reads gap in any sample, after their CCFs were estimated from each mutation’s maximum 3; splice junction overlap 10, aligned mates gap maximum VAF. To simplify the estimation of CCF from VAF, variants were 100,000, aligned intron maximum 100,000 and alignSJstitchMis- discarded if they lay in CNA-affected regions determined by SNP6.0 matchNmax 5 -1 5 5. For both patient samples and PDX, transcript analysis of the patient samples, or if their VAFs were suggestive of an counts were obtained using HTSeq v0.7.2 (57). Data was library size uncalled CNA region. X-chromosome variants in male patients had normalized using RLE, followed by a variance stabilizing transforma- their VAFs halved when estimating CCF to correct for their haploid tion using DESeq2 v1.22.1 (58). Principal component analysis plots nature. Distinct mutational populations were defined by semiauto- were generated on a per sample basis using the top 1,000 variable matically clustering variants, with variants clustered together when genes. For downstream visualization, differential expression, and

April 2020 CANCER DISCOVERY | 583

RESEARCH ARTICLE Dobson et al. pathway analysis, the mean expression of each sample clone condi- custom made by BD Biosciences), and sorted on FACS Aria III (Bec- tion was utilized. On a per patient level, differentially expressed genes ton Dickinson), consistently yielding >95% purity. HSCs were sorted were identified. Genes with adjustedP value of < 0.05 and absolute on the basis of the following markers: CD34+CD38−CD45RA−CD90+ + + log2 fold change of > 1 were considered significant. All visualizations CD49f from CD34 CB cells as described previously (60). B cells were were generated using R v3.5.1 and the pheatmap v1.0.10 and lattice sorted as CD34−CD38+CD19+CD33−CD3−CD56−. v0.20–38 packages and ggplot2 v3.1.0 packages. GSVA Pathway Enrichment Analysis and Visualization Read counts from 5 B-cell and 3 HSC-sorted populations from Pathway enrichment analysis and visualization was performed normal human umbilical cord blood were normalized using the as described previously (59). Briefly, a score to rank genes from top trimmed of M values using the edgeR_3.24.3 R package, and gene upregulated to downregulated was calculated using the formula differential expression was calculated using the quasi likelihood ratio

-sign(logFC) × -log10(P). The rank file from each comparison was method and included cord blood batch correction in the experimen- used in GSEA (http://software.broadinstitute.org/gsea/index.jsp) tal design. The top 1,000 genes enriched in B cells (B-cell genes) and using 2,000 permutations and default parameters against a pathway the top 1,000 genes enriched in HSCs (HSC genes) were selected to be database containing Msigdb c2 and c3, NCI, IOB, NetPath, Human- used as the reference signature. Two mixed profiles were used: dPDX, Cyc, GO BP, Reactome, and Panther (http://baderlab.org/GeneSets, dRI-PDX and rPDX clones and 14 bulk diagnosis and relapse patient version June 2018). GSEA progressively examines genes from the top samples. The mixed profiles read counts were variance stabilized and to the bottom of the ranked list, increasing the enrichment score (ES) library size normalized using DESeq_1.34.1. The gsva() function if a gene is part of the pathway and decreasing the score otherwise. from the R GSVA_1.30.0 package was employed using a Gaussian These running sum values are weighted, so that enrichment in the kernel to estimate enrichment of reference signatures in each sample very top-ranking (and bottom-ranking) genes is amplified, whereas of the mixed profile dataset. GSVA score for each patient and each enrichment in genes with more moderate ranks is not amplified. The category were plotted on a box plot and a strip chart. GSVA scores ES is calculated as the maximum value of the running sum and nor- were summed for each mixed profile category and standardized from malized relative to pathway size, resulting in a normalized ES (NES) 0 to 1. One thousand permutations with a random gene list of size that reflects the enrichment of the pathway in the list. Positive and 1,000 were performed on the mixed profile and percentages calcu- negative NES values represent enrichment at the top and bottom lated to indicate how many times the observed score was higher than of the list, respectively. A permutation-based P value is computed the random scores. These results were confirmed using the HSC and and corrected for multiple testing to produce a permutation-based B-cell expression profiles from Novershtern and colleagues (ref. 40; FDR q-value that ranges from zero (highly significant) to one (not Supplementary Table S7). significant). EnrichmentMap version 3.1.0 in Cytoscape 3.7.0 was used to visualize enriched gene sets at FDR q-value ≤ 0.05 for each Transcriptomic Validation Experiments comparison with a Jaccard coefficient set to 0.375. The Enrichment Staining for mitochondria content, ROS, mitochondrial mem- Map software takes as input a text file containing pathway enrich- brane potential, and mitochondrial superoxide levels was performed ment analysis results (from GSEA) and another text file contain- by incubating thawed PDX cells at 37°C with 1 mmol/L of MitoTrack- ing the pathway gene sets used in the original enrichment analysis erGreen (M7514; dilution 1:10,000); 5 μmol/L CellROX deep red (http://baderlab.org/GeneSets, version June 2018). An enrichment (C10422; dilution 1:500); 1 μmol/L TMRE (T668; dilution 1:1,000), map is a network representing overlaps among enriched pathways or 5 mmol/L MitoSOX Red (M36008; dilution 1:1,000) following with pathways represented as circles (nodes) that are connected with the manufacturer’s instructions (Thermo Fisher Scientific) and directly lines (edges) to other pathways with overlapping genes. The network analyzed on a BD LSRII cytometer. Mean fluorescence intensity layout and clustering algorithm Autoannotate app in Cytoscape (MFI) for each sample and dye is represented as ratio to dPDX for was used to automatically display and group similar pathways as each patient. Immunostaining analysis for TOMM20, MRPS18B, major biological themes. Pathways differentially enriched in dPDX and SOD1 were performed on PDX cells. Briefly, cells were spun onto versus dRI-PDX and/or dPDX versus rPDX clones were classified as: poly-l-lysine (Sigma)–coated slides (Ibidi, 200 × g, 10 minutes), fixed dRI-PDX-unique [pathways significantly enriched (FDRq -value ≤ with 4% paraformaldehyde (Sigma) and permeabilized with 0.5% 0.05) in dRI-PDX in comparison with dPDX], dRI-PDX/rPDX com- Triton (Sigma) before blocking (PBS, 10% FBS, 5% BSA). Slides were mon [pathways significantly enriched (FDRq -value ≤ 0.05) in both incubated with primary antibodies (TOMM20: ab56783, dilution dRI-PDX and rPDX in comparison with dPDX] or rPDX-unique 1/200; SOD1: ab8866, dilution 1/100; and MRPS18B: ab191891, [pathways significantly enriched (FDRq -value ≤ 0.05) in rPDX in dilution 1/200) in blocking solution overnight at 4°C. Secondary comparison with dPDX]. anti-mouse AF568, anti-rabbit AF488, and anti-sheep AF647 (Inv- itrogen, 1:400) antibodies were added (PBS, 0.025% Tween, Sigma, Isolation of HSCs and B Cells from Human 1.5 hours, room temperature). After washing, nuclei were stained Cord Blood Samples with 1 μg/mL DAPI (Invitrogen) and slides were mounted (Fluoromount Human cord blood samples were obtained with informed con- G, Invitrogen). Images were captured by a Zeiss LSM700 Confocal sent from Trillium and Credit Valley Hospital according to proce- (oil, 63×/1.4NA, Zen 2012) and analyzed with ImageJ/Fiji. The Inte- dures approved by the University Health Network Research Ethics grated Density (IntDen = Area × MFI) for 40 to 50 cells from each Board. Mononuclear cells were obtained by centrifugation on Lym- clone was analyzed. The mean for each clone was normalized and phoprep medium (StemCell Technologies) followed by red blood cell calculated as a ratio to the dPDX for each patient separately. lysis using ammonium chloride (StemCell Technologies). Human CD34+ and CD34− CB cells were separated using CD34 Micro- Secondary Transplantations for Drug Assays beads Kit (Miltenyi Biotec) according to manufacturer’s protocol and Limiting Dilution Assays and stored at −150°C. Cells were stained with antibodies (all from Human purified cells from the primary recipients were thawed and BD Biosciences, unless stated otherwise): FITC–anti-CD45RA (1:50, transplanted into 8- to 12-week-old female NSG mice sublethally 555488), PE–anti-CD90 (1:50, 555596), PECy5–anti-CD49f (1:50, irradiated as described for the primary recipients. The number of 551129), V450–anti-CD7 (1:33.3, 642916), PECy7–anti-CD38 (1:100, mice used for secondary transplantation experiments/drugs was 335790), APC–anti-CD10 (1:50, 340923), APCCy7–anti-CD34 (1:200, determined by cell and mouse availability and feasibility. Intrafemoral

584 | CANCER DISCOVERY April 2020 AACRJournals.org

Characterization of Relapse-fated Clones in Diagnosis B-ALL RESEARCH ARTICLE injections of 26,000 to 100,000 leukemic blasts were performed for Authors’ Contributions drug assays and a range of 10,000 to 100,000 leukemic blasts were Conception and design: S.M. Dobson, C.G. Mullighan, J.E. Dick injected for secondary limiting dilution assays. After 4 weeks, mice Development of methodology: S.M. Dobson, J. Wintersinger, were bled from the saphenous vein and human chimerism was evalu- I. Grandal, G. Bader, J. Easton, J.S. Danska, Q. Morris, C.G. Mullighan, ated by flow cytometry. Once human engraftment in the peripheral J.E. Dick blood reached between 1% and 10%, or after 10 to 14 weeks for those Acquisition of data (provided animals, acquired and managed samples in which leukemic blasts did not mobilize to the periph- patients, provided facilities, etc.): S.M. Dobson, L. García-Prat, eral blood, mice were randomized and single-agent treatments were R.J. Vanner, E. Waanders, J. McLeod, O.I. Gan, I. Grandal, S.Z. Xie, started. Dexamethasone (15 mg/kg), l-asparaginase (1,000 kU/kg), M. Hosseini, S.R. Olsen, G. Neale, S.M. Chan, J. Easton, C.J. Guidos, and saline were given daily by intraperitoneal injection 5 days a week. J.S. Danska, M.D. Minden, C.G. Mullighan, J.E. Dick Vincristine (0.5 mg/kg) was given once a week by intraperitoneal injec- Analysis and interpretation of data (e.g., statistical analysis, bio- tion. All four arms of the drug treatment were performed for 4 weeks statistics, computational analysis): S.M. Dobson, L. García-Prat, and mice were sacrificed the following day (vincristine/dexametha- R.J. Vanner, J. Wintersinger, E. Waanders, Z. Gu, M.N. Edmonson, sone/saline) or 1 week after the last treatment (vincristine). Analysis X. Ma, Y. Fan, V. Voisin, M. Chan-Seng-Yue, S. Abelson, M. Rusch, of the secondary limiting dilution assay was performed 16 weeks S.M. Chan, G. Bader, J. Zhang, M.D. Minden, Q. Morris, C.G. Mul- posttransplant. Human cell engraftment in the injected femur, bone lighan, J.E. Dick marrow, and spleen were assessed using human-specific antibodies for Writing, review, and/or revision of the manuscript: S.M. Dobson, CD45 (FITC, BD Biosciences, clone 2D1; v500, BD Biosciences, clone L. García-Prat, R.J. Vanner, J. Wintersinger, E. Waanders, O.I. Gan, HI30), CD19 (BD Biosciences, PE, clone 4G7), CD33 (PE-Cy7, BD S.Z. Xie, S. Abelson, J. Easton, J.S. Danska, M.D. Minden, Q. Morris, Biosciences, clone P67-6), CD3 (APC, BD Biosciences, clone UCHT1), C.G. Mullighan, J.E. Dick CD10 (Qdot605, BD Biosciences, clone HI10A), CD38 (BV421, Bio- Administrative, technical, or material support (i.e., report- Legend, clone HIT2), and CD34 (APC-Cy7, BD Biosciences, clone ing or organizing data, constructing databases): S.M. Dobson, 581). Mice were considered to be engrafted when >0.1% of cells in L. García-Prat, J. Wintersinger, E. Waanders, J. McLeod, I. Grandal, the injected femur were positive for one or more human B-ALL– D. Payne-Turner, P. Gupta, Y. Shao specific cell-surface markers (CD45, CD44, CD19, CD34). Response Study supervision: Q. Morris, C.G. Mullighan, J.E. Dick to treatment was analyzed as a ratio of human engraftment of drug- treated versus saline-treated mice to eliminate interclonal differences Acknowledgments in engraftment levels. Ratio of human engraftment in each individual drug-treated PDX to the average engraftment level of all saline We thank the patients and their families and physicians who made controls was calculated. Lineage stains were performed on xenografts this study possible. We also thank N. Simard, S. Laronde, L. Jaimeson, expressing CD33 (APC, BD Biosciences clone P67-6) and CD19 A. Khandani, T. Velanthapillai, and S. Zhao at the UHN/SickKids (PE, BD Biosciences, clone 4G7) including CD14 (PE-Cy7, Beckman Flow Cytometry Facility; P. Lo and R. Lopez from the UHN Animal Coulter, clone 52), CD15 (v450, BD Biosciences, clone MMA), CD10 Resource Centre; the Genome Sequencing Facility of the Hartwell (Qdot605, BioLegend, clone HIT2), and CD34 (APC-Cy7, BD Bio- Centre for Bioinformatics and Biotechnology, and the Biorepository sciences, clone 581). The confidence intervals for the frequency of of St. Jude Children’s Research Hospital. We thank J. Ho, E. Lechman, A. L-ICs was calculated using ELDA software (49). Statistical analysis Mitchell, L. Jin, M. Doedens, J. Loo-Yong-Kee, K.L. Woo, M.C. Shoong, was performed using PRISM 6 (GraphPad Software). and J. Roth for their technical assistance; J.A. Kennedy for clinical annotation; and K. Kaufmann for data analysis script. This work FACS from Xenografts was supported by funds to J.E. Dick from the Princess Margaret Cancer Centre Foundation, Ontario Institute for Cancer Research, Leukemic cells from primary xenografts were sorted for immu- with funding from the Province of Ontario, Canadian Institutes for nophenotypic populations on a FACSARIAIII (BD Biosciences). Cells Health Research, Canadian Cancer Society Research Institute, Terry from the IF and BM were pooled and stained with CD19, CD34, Fox Research Institute Program Project, Genome Canada through CD45, CD10, and CD33 and collected at a sort purity of >99%. the Ontario Genomics Institute, and a Canada Research Chair; and to C.G. Mullighan from the the American Lebanese Syrian Associated Data Availability Statement Charities of St. Jude Children’s Research Hospital, a St. Baldrick’s The datasets generated during this study are included in this pub- Foundation Robert J. Arceci Innovation Award, the Henry Schueler lished article and its supplementary information files. 41&9 Foundation, the NCI grants P30 CA021765 (St. Jude Cancer Center Support Grant) and Outstanding Investigator Award R35 Code Availability Statement CA197695. Q. Morris was supported by funds from the Natural Sci- Code used in this study is available at https://www.github.com/ ences and Engineering Research Council, a Compute the Cure award morrislab/pairtree from the NVIDIA charitable foundation, and the Vector Institute. E. Waanders is funded by the Dutch Cancer Society (KUN2012-5366). Disclosure of Potential Conflicts of Interest Received September 11, 2019; revised December 21, 2019; accepted J.S. Danska reports receiving commercial research support from February 18, 2020; published first February 21, 2020. Trillium Therapeutics, Inc. and has ownership interest (including patents) in the same. M.D. Minden reports receiving commercial research support from Kura and has received speakers bureau hono- References raria from Astellas. C.G. Mullighan is a scientific advisory board . 1 Forman SJ, Rowe JM. The myth of the second remission of acute member for Illumina, reports receiving commercial research grants leukemia in the adult. Blood 2013;121:1077–82. from AbbVie, Pfizer, and Loxo Oncology, and has received speak- 2. Liew E, Atenafu EG, Schimmer AD, Yee KW, Schuh AC, Minden MD, ers bureau honoraria from Illumina, Amgen, Pfizer, and Aptitude et al. Outcomes of adult patients with relapsed acute lymphoblastic Health. J.E. Dick served on the SAB at Trillium Therapeutics, reports leukemia following frontline treatment with a pediatric regimen. receiving a commercial research grant from Celgene, and has owner- Leuk Res 2012;36:1517–20. ship interest (including patents) in Trillium Therapeutics Inc. No 3. Hunger SP, Mullighan CG. Acute lymphoblastic leukemia in chil- potential conflicts of interest were disclosed by the other authors. dren. N Engl J Med 2015;373:1541–52.

April 2020 CANCER DISCOVERY | 585

RESEARCH ARTICLE Dobson et al.

4. Mullighan CG, Phillips LA, Su X, Ma J, Miller CB, Shurtleff SA, et al. 24. Ng SW, Mitchell A, Kennedy JA, Chen WC, McLeod J, Ibrahimova Genomic analysis of the clonal origins of relapsed acute lymphoblas- N, et al. A 17-gene stemness score for rapid determination of risk in tic leukemia. Science 2008;322:1377–80. acute leukaemia. Nature 2016;540:433–7. 5. Ma X, Edmonson M, Yergeau D, Muzny DM, Hampton OA, Rusch 25. Liem NL, Papa RA, Milross CG, Schmid MA, Tajbakhsh M, Choi S, M, et al. Rise and fall of subclones from diagnosis to relapse in et al. Characterization of childhood acute lymphoblastic leukemia pediatric B-acute lymphoblastic leukaemia. Nat Commun 2015; xenograft models for the preclinical evaluation of new therapies. 6:6604. Blood 2004;103:3905–14. 6. Oshima K, Khiabanian H, da Silva-Almeida AC, Tzoneva G, Abate F, 26. Meyer LH, Eckhoff SM, Queudeville M, Kraus JM, Giordan M, Ambesi-Impiombato A, et al. Mutational landscape, clonal evolution Stursberg J, et al. Early relapse in ALL is identified by time to leukemia patterns, and role of RAS mutations in relapsed acute lymphoblastic in NOD/SCID mice and is characterized by a gene signature involv- leukemia. Proc Natl Acad Sci U S A 2016;113:11306–11. ing survival pathways. Cancer Cell 2011;19:206–17. 7. Waanders E, Gu Z, Dobsom SM, Antic Z, Crawford JC, Ma X, et al. 27. Shlush LI, Mitchell A, Heisler L, Abelson S, Ng SWK, Trotman-Grant Mutational landscape and patterns of clonal evolution in relapsed A, et al. Tracing the origins of relapse in acute myeloid leukaemia to pediatric acute lymphoblastic leukemia. Blood Cancer Discov stem cells. Nature 2017;547:104–8. 2020;1:1–16. 28. Clappier E, Gerby B, Sigaux F, Delord M, Touzri F, Hernandez L, 8. Anderson K, Lutz C, van Delft FW, Bateman CM, Guo Y, Colman SM, et al. Clonal selection in xenografted human T cell acute lymphoblas- et al. Genetic variegation of clonal architecture and propagating cells tic leukemia recapitulates gain of malignancy at relapse. J Exp Med in leukaemia. Nature 2011;469:356–61. 2011;208:653–61. 9. Notta F, Mullighan CG, Wang JC, Poeppl A, Doulatov S, Phillips LA, 29. Klco JM, Spencer DH, Miller CA, Griffith M, Lamprecht TL, et al. Evolution of human BCR-ABL1 lymphoblastic leukaemia- O’Laughlin M, et al. Functional heterogeneity of genetically defined initiating cells. Nature 2011;469:362–7. subclones in acute myeloid leukemia. Cancer Cell 2014;25:379–92. 10. Kreso A, Dick JE. Evolution of the cancer stem cell model. Cell Stem 30. Richter-Pechanska P, Kunz JB, Bornhauser B, von Knebel Doeberitz Cell 2014;14:275–91. C, Rausch T, Erarslan-Uysal B, et al. PDX models recapitulate the 11. Foo J, Michor F. Evolution of acquired resistance to anti-cancer genetic and epigenetic landscape of pediatric T-cell leukemia. EMBO therapy. J Theor Biol 2014;355:10–20. Mol Med 2018;10:e9443. 12. Kim C, Gao R, Sei E, Brandt R, Hartman J, Hatschek T, et al. Chemo 31. Rehe K, Wilson K, Bomken S, Williamson D, Irving J, den Boer ML, resistance evolution in triple-negative breast cancer delineated by et al. Acute B lymphoblastic leukaemia-propagating cells are present single-cell sequencing. Cell 2018;173:879–93. at high frequency in diverse lymphoblast populations. EMBO Mol 13. Iqbal Z, Aleem A, Iqbal M, Naqvi MI, Gill A, Taj AS, et al. Sensitive Med 2013;5:38–51. detection of pre-existing BCR-ABL kinase domain mutations in 32. Schmitz M, Breithaupt P, Scheidegger N, Cario G, Bonapace L, Meissner CD34+ cells of newly diagnosed chronic-phase chronic myeloid leu- B, et al. Xenografts of highly resistant leukemia recapitulate the clonal kemia patients is associated with imatinib resistance: implications in composition of the leukemogenic compartment. Blood 2011;118: the post-imatinib era. PLoS One 2013;8:e55717. 1854–64. 14. Tzoneva G, Dieck CL, Oshima K, Ambesi-Impiombato A, Sanchez- 33. Chiaretti S, Li X, Gentleman R, Vitale A, Vignetti M, Mandelli F, Martin M, Madubata CJ, et al. Clonal evolution mechanisms in NT5C2 et al. Gene expression profile of adult T-cell acute lymphocytic leuke- mutant-relapsed acute lymphoblastic leukaemia. Nature 2018; mia identifies distinct subsets of patients with different response to 553:511–4. therapy and survival. Blood 2004;103:2771–8. 15. Hunter C, Smith R, Cahill DP, Stephens P, Stevens C, Teague J, et al. 34. Slamova L, Starkova J, Fronkova E, Zaliova M, Reznickova L, van Delft A hypermutation phenotype and somatic MSH6 mutations in recur- FW, et al. CD2-positive B-cell precursor acute lymphoblastic leukemia rent human malignant gliomas after alkylator chemotherapy. Cancer with an early switch to the monocytic lineage. Leukemia 2014;28: Res 2006;66:3987–91. 609–20. 16. Alexandrov LB, Nik-Zainal S, Wedge DC, Aparicio SA, Behjati S, 35. O’Connell MP, Marchbank K, Webster MR, Valiga AA, Kaur A, Vultur Biankin AV, et al. Signatures of mutational processes in human can- A, et al. Hypoxia induces phenotypic plasticity and therapy resistance cer. Nature 2013;500:415–21. in melanoma via the tyrosine kinase receptors ROR1 and ROR2. 17. Churchman ML, Low J, Qu C, Paietta EM, Kasper LH, Chang Y, et al. Cancer Discov 2013;3:1378–93. Efficacy of retinoids in IKZF1-mutated BCR-ABL1 acute lymphoblas- 36. Kemper K, de Goeje PL, Peeper DS, van Amerongen R. Phenotype tic leukemia. Cancer Cell 2015;28:343–56. switching: tumor cell plasticity as a resistance mechanism and target 18. Ebinger S, Ozdemir EZ, Ziegenhain C, Tiedt S, Castro Alves C, for therapy. Cancer Res 2014;74:5937–41. Grunert M, et al. Characterization of rare, dormant, and therapy- 37. Chaidos A, Barnes CP, Cowan G, May PC, Melo V, Hatjiharissi E, resistant cells in acute lymphoblastic leukemia. Cancer Cell 2016;30: et al. Clinical drug resistance linked to interconvertible phenotypic 849–62. and functional states of tumor-propagating cells in multiple mye- 19. Polak R, de Rooij B, Pieters R, den Boer ML. B-cell precursor acute loma. Blood 2013;121:318–28. lymphoblastic leukemia cells use tunneling nanotubes to orchestrate 38. Pisco AO, Huang S. Non-genetic cancer cell plasticity and therapy- their microenvironment. Blood 2015;126:2404–14. induced stemness in tumour relapse: ‘What does not kill me strength- 20. Iwamoto S, Mihara K, Downing JR, Pui CH, Campana D. Mesenchy- ens me’. Br J Cancer 2015;112:1725–32. mal cells regulate the response of acute lymphoblastic leukemia cells 39. Belviso S, Iuliano R, Amato R, Perrotti N, Menniti M. The human to asparaginase. J Clin Invest 2007;117:1049–57. asparaginase enzyme (ASPG) inhibits growth in leukemic cells. PLoS 21. Mullighan CG, Zhang J, Kasper LH, Lerach S, Payne-Turner D, Phillips One 2017;12:e0178174. LA, et al. CREBBP mutations in relapsed acute lymphoblastic leukae- 40. Novershtern N, Subramanian A, Lawton LN, Mak RH, Haining mia. Nature 2011;471:235–9. WN, McConkey ME, et al. Densely interconnected transcriptional 22. Kamel-Reid S, Letarte M, Sirard C, Doedens M, Grunberger T, Fulop G, circuits control cell states in human hematopoiesis. Cell 2011;144: et al. A model of human acute lymphoblastic leukemia in immune- 296–309. deficient SCID mice. Science 1989;246:1597–600. 41. Jones CL, Stevens BM, D’Alessandro A, Reisz JA, Culp-Hill R, 23. Lock RB, Liem N, Farnsworth ML, Milross CG, Xue C, Tajbakhsh M, Nemkov T, et al. Inhibition of amino acid metabolism selectively tar- et al. The nonobese diabetic/severe combined immunodeficient gets human leukemia stem cells. Cancer Cell 2018;34:724–40. (NOD/SCID) mouse model of childhood acute lymphoblastic leuke- 42. van Galen P, Mbong N, Kreso A, Schoof EM, Wagenblast E, Ng SWK, mia reveals intrinsic differences in biologic characteristics at diagno- et al. Integrated stress response activity marks stem cells in normal sis and relapse. Blood 2002;99:4100–8. hematopoiesis and leukemia. Cell Rep 2018;25:1109–17.

586 | CANCER DISCOVERY April 2020 AACRJournals.org

Characterization of Relapse-fated Clones in Diagnosis B-ALL RESEARCH ARTICLE

43. Mullighan CG, Su X, Zhang J, Radtke I, Phillips LA, Miller CB, et al. 52. Pounds S, Cheng C, Mullighan C, Raimondi SC, Shurtleff S, Deletion of IKZF1 and prognosis in acute lymphoblastic leukemia. Downing JR. Reference alignment of SNP microarray signals for copy N Engl J Med 2009;360:470–80. number analysis of tumors. Bioinformatics 2009;25:315–21. 44. Luria SE, Delbruck M. Mutations of bacteria from virus sensitivity to 53. Olshen AB, Venkatraman ES, Lucito R, Wigler M. Circular binary virus resistance. Genetics 1943;28:491–511. segmentation for the analysis of array-based DNA copy number data. 45. Eppert K, Takenaka K, Lechman ER, Waldron L, Nilsson B, van Biostatistics 2004;5:557–72. Galen P, et al. Stem cell gene expression programs influence clinical 54. Venkatraman ES, Olshen AB. A faster circular binary segmenta- outcome in human leukemia. Nat Med 2011;17:1086–93. tion algorithm for the analysis of array CGH data. Bioinformatics 46. Jin L, Hope KJ, Zhai Q, Smadja-Joffe F, Dick JE. Targeting of CD44 2007;23:657–63. eradicates human acute myeloid leukemic stem cells. Nat Med 2006; 55. Lin M, Wei LJ, Sellers WR, Lieberfarb M, Wong WH, Li C. dChipSNP: 12:1167–74. significance curve and clustering of SNP-array-based loss-of-heterozy- 47. Chen WC, Yuan JS, Xing Y, Mitchell A, Mbong N, Popescu AC, gosity data. Bioinformatics 2004;20:1233–40. et al. An integrated analysis of heterogeneous drug responses in acute 56. Dobin A, Davis CA, Schlesinger F, Drenkow J, Zaleski C, Jha S, et al. myeloid leukemia that enables the discovery of predictive biomarkers. STAR: ultrafast universal RNA-seq aligner. Bioinformatics 2013;29: Cancer Res 2016;76:1214–24. 15–21. 48. Mazurier F, Doedens M, Gan OI, Dick JE. Rapid myeloerythroid 57. Anders S, Pyl PT, Huber W. HTSeq–a Python framework to work with repopulation after intrafemoral transplantation of NOD-SCID mice high-throughput sequencing data. Bioinformatics 2015;31:166–9. reveals a new class of human stem cells. Nat Med 2003;9:959–63. 58. Love MI, Huber W, Anders S. Moderated estimation of fold change 49. Hu Y, Smyth GK. ELDA: extreme limiting dilution analysis for and dispersion for RNA-seq data with DESeq2. Genome Biol 2014; comparing depleted and enriched populations in stem cell and other 15:550. assays. J Immunol Methods 2009;347:70–8. 59. Reimand J, Isserlin R, Voisin V, Kucera M, Tannus-Lopes C, Rosta- 50. Edmonson MN, Zhang J, Yan C, Finney RP, Meerzaman DM, Buetow mianfar A, et al. Pathway enrichment analysis and visualization of KH. Bambino: a variant detector and alignment viewer for next- omics data using g:Profiler, GSEA, Cytoscape and EnrichmentMap. generation sequencing data in the SAM/BAM format. Bioinformatics Nat Protoc 2019;14:482–517. 2011;27:865–6. 60. Notta F, Doulatov S, Laurenti E, Poeppl A, Jurisica I, Dick JE. Isola- 51. Mullighan CG. Single nucleotide polymorphism microarray analysis tion of single human hematopoietic stem cells capable of long-term of genetic alterations in cancer. Methods Mol Biol 2011;730:235–58. multilineage engraftment. Science 2011;333:218–21.

April 2020 CANCER DISCOVERY | 587

Relapse-Fated Latent Diagnosis Subclones in Acute B Lineage Leukemia Are Drug Tolerant and Possess Distinct Metabolic Programs

Stephanie M. Dobson, Laura García-Prat, Robert J. Vanner, et al.

Cancer Discov 2020;10:568-587. Published OnlineFirst February 21, 2020.

Updated version Access the most recent version of this article at: doi:10.1158/2159-8290.CD-19-1059

Supplementary Access the most recent supplemental material at: Material http://cancerdiscovery.aacrjournals.org/content/suppl/2020/02/21/2159-8290.CD-19-1059.DC1

Cited articles This article cites 59 articles, 19 of which you can access for free at: http://cancerdiscovery.aacrjournals.org/content/10/4/568.full#ref-list-1

Citing articles This article has been cited by 12 HighWire-hosted articles. Access the articles at: http://cancerdiscovery.aacrjournals.org/content/10/4/568.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerdiscovery.aacrjournals.org/content/10/4/568. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.