The genomic landscape of acute lymphoblastic leukemia with intrachromosomal amplification of chromosome 21.

Intrachromosomal amplification of chromosome 21 defines a subtype of high-risk childhood acute lymphoblastic leukemia (iAMP21-ALL) characterized by copy number changes and complex rearrangements of chromosome 21. The genomic basis of iAMP21-ALL and the pathogenic role of the region of amplification of chromosome 21 to leukemogenesis remains incompletely understood. In this study, using integrated whole genome and transcriptome sequencing of 124 patients with iAMP21-ALL, including rare cases arising in the context of constitutional chromosomal aberrations, we identified subgroups of iAMP21-ALL based on the patterns of copy number alteration and structural variation. This large data set enabled formal delineation of a 7.8 Mb common region of amplification harboring 71 genes, 43 of which were differentially expressed compared with non-iAMP21-ALL ones, including multiple genes implicated in the pathogenesis of acute leukemia (CHAF1B, DYRK1A, ERG, HMGN1, and RUNX1). Using multimodal single-cell genomic profiling, including single-cell whole genome sequencing of 2 cases, we documented clonal heterogeneity and genomic evolution, demonstrating that the acquisition of the iAMP21 chromosome is an early event that may undergo progressive amplification during disease ontogeny. We show that UV-mutational signatures and high mutation load are characteristic secondary genetic features. Although the genomic alterations of chromosome 21 are variable, these integrated genomic analyses and demonstration of an extended common minimal region of amplification broaden the definition of iAMP21-ALL for more precise diagnosis using cytogenetic or genomic methods to inform clinical management.

Whole genome sequencing provides comprehensive genetic testing in childhood B-cell acute lymphoblastic leukaemia.

Childhood B-cell acute lymphoblastic leukaemia (B-ALL) is characterised by recurrent genetic abnormalities that drive risk-directed treatment strategies. Using current techniques, accurate detection of such aberrations can be challenging, due to the rapidly expanding list of key genetic abnormalities. Whole genome sequencing (WGS) has the potential to improve genetic testing, but requires comprehensive validation. We performed WGS on 210 childhood B-ALL samples annotated with clinical and genetic data. We devised a molecular classification system to subtype these patients based on identification of key genetic changes in tumour-normal and tumour-only analyses. This approach detected 294 subtype-defining genetic abnormalities in 96% (202/210) patients. Novel genetic variants, including fusions involving genes in the MAP kinase pathway, were identified. WGS results were concordant with standard-of-care methods and whole transcriptome sequencing (WTS). We expanded the catalogue of genetic profiles that reliably classify PAX5alt and ETV6::RUNX1-like subtypes. Our novel bioinformatic pipeline improved detection of DUX4 rearrangements (DUX4-r): a good-risk B-ALL subtype with high survival rates. Overall, we have validated that WGS provides a standalone, reliable genetic test to detect all subtype-defining genetic abnormalities in B-ALL, accurately classifying patients for the risk-directed treatment stratification, while simultaneously performing as a research tool to identify novel disease biomarkers.

Hyperactive CREB subpopulations increase during therapy in pediatric B-lineage acute lymphoblastic leukemia.

Persistence of residual disease in acute lymphoblastic leukemia (ALL) during the initial stages of chemotherapy is associated with inferior survival. To better understand clonal evolution and mechanisms of chemoresistance, we used multiparameter mass cytometry, at single-cell resolution, to functionally characterize pediatric B-ALL cells at disease presentation and those persisting during induction therapy. Analysis of ALL cells from presentation samples (n=42) showed that the most abundant phosphosignals were pCREB, pH2AX and pHH3 and we identified JAK-STAT and RAS pathway activation in five of six patients with JAK or RAS genetic aberrations. The clonal composition of ALL was heterogeneous and dynamic during treatment but all viable cell clusters showed pCREB activation. Levels of pCREB in ALL cells were increased or maintained during therapy and high dimensional analysis revealed a subpopulation of ALL cells at presentation that was positive for pCREB/pHH3/pS6 which increased during treatment in some patients, implicating this signaling node in conferring a survival advantage to multi-agent induction therapy. The small molecule CREB inhibitor, 666-15, was shown to reduce CREB transcriptional activity and induce apoptosis in ALL patient-derived xenograft cells of varying cytogenetic subtypes in vitro, both in the presence and absence of stromal support. Together, these data suggest that the cAMP signaling pathway may provide an opportunity for minimal residual disease-directed therapy for many patients at high risk of relapse.

Integrative genomic analysis of childhood acute lymphoblastic leukaemia lacking a genetic biomarker in the UKALL2003 clinical trial.

Incorporating genetics into risk-stratification for treatment of childhood B-progenitor acute lymphoblastic leukaemia (B-ALL) has contributed significantly to improved survival. In about 30% B-ALL (B-other-ALL) without well-established chromosomal changes, new genetic subtypes have recently emerged, yet their true prognostic relevance largely remains unclear. We integrated next generation sequencing (NGS): whole genome sequencing (WGS) (n = 157) and bespoke targeted NGS (t-NGS) (n = 175) (overlap n = 36), with existing genetic annotation in a representative cohort of 351 B-other-ALL patients from the childhood ALL trail, UKALL2003. PAX5alt was most frequently observed (n = 91), whereas PAX5 P80R mutations (n = 11) defined a distinct PAX5 subtype. DUX4-r subtype (n = 80) was defined by DUX4 rearrangements and/or ERG deletions. These patients had a low relapse rate and excellent survival. ETV6::RUNX1-like subtype (n = 21) was characterised by multiple abnormalities of ETV6 and IKZF1, with no reported relapses or deaths, indicating their excellent prognosis in this trial. An inferior outcome for patients with ABL-class fusions (n = 25) was confirmed. Integration of NGS into genomic profiling of B-other-ALL within a single childhood ALL trial, UKALL2003, has shown the added clinical value of NGS-based approaches, through improved accuracy in detection and classification into the range of risk stratifying genetic subtypes, while validating their prognostic significance.

hiPSC-derived bone marrow milieu identifies a clinically actionable driver of niche-mediated treatment resistance in leukemia.

Leukemia cells re-program their microenvironment to augment blast proliferation and enhance treatment resistance. Means of clinically targeting such niche-driven treatment resistance remain ambiguous. We develop human induced pluripotent stem cell (hiPSC)-engineered niches to reveal druggable cancer-niche dependencies. We reveal that mesenchymal (iMSC) and vascular niche-like (iANG) hiPSC-derived cells support ex vivo proliferation of patient-derived leukemia cells, affect dormancy, and mediate treatment resistance. iMSCs protect dormant and cycling blasts against dexamethasone, while iANGs protect only dormant blasts. Leukemia proliferation and protection from dexamethasone-induced apoptosis is dependent on cancer-niche interactions mediated by CDH2. Consequently, we test CDH2 antagonist ADH-1 (previously in Phase I/II trials for solid tumors) in a very aggressive patient-derived xenograft leukemia mouse model. ADH-1 shows high in vivo efficacy; ADH-1/dexamethasone combination is superior to dexamethasone alone, with no ADH-1-conferred additional toxicity. These findings provide a proof-of-concept starting point to develop improved, potentially safer therapeutics targeting niche-mediated cancer dependencies in blood cancers.

Clonal heterogeneity and rates of specific chromosome gains are risk predictors in childhood high-hyperdiploid B-cell acute lymphoblastic leukemia.

B-cell acute lymphoblastic leukemia (B-ALL) is the commonest childhood cancer. High hyperdiploidy (HHD) identifies the most frequent cytogenetic subgroup in childhood B-ALL. Although hyperdiploidy represents an important prognostic factor in childhood B-ALL, the specific chromosome gains with prognostic value in HHD-B-ALL remain controversial, and the current knowledge about the hierarchy of chromosome gains, clonal heterogeneity and chromosomal instability in HHD-B-ALL remains very limited. We applied automated sequential-iFISH coupled with single-cell computational modeling to identify the specific chromosomal gains of the eight typically gained chromosomes in a large cohort of 72 primary diagnostic (DX, n = 62) and matched relapse (REL, n = 10) samples from HHD-B-ALL patients with either favorable or unfavorable clinical outcome in order to characterize the clonal heterogeneity, specific chromosome gains and clonal evolution. Our data show a high degree of clonal heterogeneity and a hierarchical order of chromosome gains in DX samples of HHD-B-ALL. The rates of specific chromosome gains and clonal heterogeneity found in DX samples differ between HHD-B-ALL patients with favorable or unfavorable clinical outcome. In fact, our comprehensive analyses at DX using a computationally defined risk predictor revealed low levels of trisomies +18+10 and low levels of clonal heterogeneity as robust relapse risk factors in minimal residual disease (MRD)-negative childhood HHD-B-ALL patients: relapse-free survival beyond 5 years: 22.1% versus 87.9%, P < 0.0001 and 33.3% versus 80%, P < 0.0001, respectively. Moreover, longitudinal analysis of matched DX-REL HHD-B-ALL samples revealed distinct patterns of clonal evolution at relapse. Our study offers a reliable prognostic sub-stratification of pediatric MRD-negative HHD-B-ALL patients.

Genetic characterisation of childhood B-other-acute lymphoblastic leukaemia in UK patients by fluorescence in situ hybridisation and Multiplex Ligation-dependent Probe Amplification.

While next-generation sequencing technologies provide excellent strategies to screen for newly defined genetic abnormalities of prognostic or therapeutic significance in patients with B-other-acute lymphoblastic leukaemia (ALL), they are not widely available. We used a dual screening approach, incorporating fluorescence in situ hybridisation (FISH) and Multiplex Ligation-dependent Probe Amplification (MLPA), to establish the frequency and long-term outcome of a representative cohort of specific subgroups of B-other-ALL recruited to the childhood ALL trial, UKALL2003. We focussed on abnormalities of known prognostic significance, including ABL-class fusions and ERG deletions, as a surrogate marker for DUX4-rearranged ALL. ABL-class fusions accounted for ~4% of B-other-ALL and were associated with high levels of minimal residual disease (MRD; 14/23 with MRD >5%) and a high relapse rate (55·7%) following treatment without tyrosine kinase inhibitor (TKI), confirming the importance of prospective screening with a view to incorporating TKI into therapy. Patients with deletions of ERG (~10% of B-other-ALL) had a 10-year event-free-survival of 97·2%, validating previous reports of their excellent outcome. Rearrangements of ZNF384, MEF2D and NUTM1 were observed at low frequencies. Here, we estimate that approximately one third of B-other-ALL patients can be reliably classified into one of the known genetic subgroups using our dual screening method. This approach is rapid, accurate and readily incorporated into routine testing.

Prognostic impact of chromosomal abnormalities and copy number alterations in adult B-cell precursor acute lymphoblastic leukaemia: a UKALL14 study.

Chromosomal abnormalities are established prognostic markers in adult ALL. We assessed the prognostic impact of established chromosomal abnormalities and key copy number alterations (CNA) among 652 patients with B-cell precursor ALL treated on a modern MRD driven protocol. Patients with KMT2A-AFF1, complex karyotype (CK) and low hypodiploidy/near-triploidy (HoTr) had high relapse rates 50%, 60% & 53% and correspondingly poor survival. Patients with BCR-ABL1 had an outcome similar to other patients. JAK-STAT abnormalities (CRLF2, JAK2) occurred in 6% patients and were associated with a high relapse rate (56%). Patients with ABL-class fusions were rare (1%). A small group of patients with ZNF384 fusions (n = 12) had very good survival. CNA affecting IKZF1, CDKN2A/B, PAX5, BTG1, ETV6, EBF1, RB1 and PAR1 were assessed in 436 patients. None of the individual deletions or profiles were associated with survival, either in the cohort overall or within key subgroups. Collectively these data indicate that primary genetic abnormalities are stronger prognostic markers than secondary deletions. We propose a revised UKALL genetic risk classification based on key established chromosomal abnormalities: (1) very high risk: CK, HoTr or JAK-STAT abnormalities; (2) high risk: KMT2A fusions; (3) Tyrosine kinase activating: BCR-ABL1 and ABL-class fusions; (4) standard risk: all other patients.

Single nucleotide polymorphism array-based signature of low hypodiploidy in acute lymphoblastic leukemia.

Low hypodiploidy (30-39 chromosomes) is one of the most prevalent genetic subtypes among adults with ALL and is associated with a very poor outcome. Low hypodiploid clones can often undergo a chromosomal doubling generating a near-triploid clone (60-78 chromosomes). When cytogenetic techniques detect a near triploid clone, a diagnostic challenge may ensue in differentiating presumed duplicated low hypodiploidy from good risk high hyperdiploid ALL (51-67 chromosomes). We used single-nucleotide polymorphism (SNP) arrays to analyze low hypodiploid/near triploid (HoTr) (n = 48) and high hyperdiploid (HeH) (n = 40) cases. In addition to standard analysis, we derived log2 ratios for entire chromosomes enabling us to analyze the cohort using machine-learning techniques. Low hypodiploid and near triploid cases clustered together and separately from high hyperdiploid samples. Using these approaches, we also identified three cases with 50-60 chromosomes, originally called as HeH, which were, in fact, HoTr and two cases incorrectly called as HoTr. TP53 mutation analysis supported the new classification of all cases tested. Next, we constructed a classification and regression tree model for predicting ploidy status with chromosomes 1, 7, and 14 being the key discriminators. The classifier correctly identified 47/50 (94%) HoTr cases. We validated the classifier using an independent cohort of 44 cases where it correctly called 7/7 (100%) low hypodiploid cases. The results of this study suggest that HoTr is more frequent among older adults with ALL than previously estimated and that SNP array analysis should accompany cytogenetics where possible. The classifier can assist where SNP array patterns are challenging to interpret.

MYB rearrangements and over-expression in T-cell acute lymphoblastic leukemia.

We investigated MYB rearrangements (MYB-R) and the levels of MYB expression, in 331 pediatric and adult patients with T-cell acute lymphoblastic leukemia (T-ALL). MYB-R were detected in 17 cases and consisted of MYB tandem duplication (tdup) (= 14) or T cell receptor beta locus (TRB)-MYB (= 3). As previously reported, TRB-MYB was found only in children (1.6%) while MYB tdup occurred in both age groups, although it was slightly more frequent in children (5.2% vs 2.8%). Shared features of MYB-R T-ALL were a non-early T-cell precursor (ETP) phenotype, a high incidence of NOTCH1/FBXW7 mutations (81%) and CDKN2AB deletions (70.5%). Moreover, they mainly belonged to HOXA (=8), NKX2-1/2-2/TLX1 (=4), and TLX3 (=3) homeobox-related subgroups. Overall, MYB-R cases had significantly higher levels of MYB expression than MYB wild type (MYB-wt) cases, although high levels of MYB were detected in ~ 30% of MYB-wt T-ALL. Consistent with the transcriptional regulatory networks, cases with high MYB expression were significantly enriched within the TAL/LMO subgroup (P = .017). Interestingly, analysis of paired diagnosis/remission samples demonstrated that a high MYB expression was restricted to the leukemic clone. Our study has indicated that different mechanisms underlie MYB deregulation in 30%-40% of T-ALL and highlighted that, MYB has potential as predictive/prognostic marker and/or target for tailored therapy.

Adjuvant tyrosine kinase inhibitor therapy improves outcome for children and adolescents with acute lymphoblastic leukaemia who have an ABL-class fusion.

Patients with an ABL-class fusion have a high risk of relapse on standard chemotherapy but are sensitive to tyrosine kinase inhibitors (TKI). In UKALL2011, we screened patients with post-induction MRD ≥1% and positive patients (12%) received adjuvant TKI. As the intervention started during UKALL2011, not all eligible patients were screened prospectively. Retrospective screening of eligible patients allowed the outcome of equivalent ABL-class patients who did and did not receive a TKI in first remission to be compared. ABL-class patients who received a TKI in first remission had a reduced risk of relapse/refractory disease: 0% vs. 63% at four years (P = 0·009).

A validated novel continuous prognostic index to deliver stratified medicine in pediatric acute lymphoblastic leukemia.

Risk stratification is essential for the delivery of optimal treatment in childhood acute lymphoblastic leukemia. However, current risk stratification algorithms dichotomize variables and apply risk factors independently, which may incorrectly assume identical associations across biologically heterogeneous subsets and reduce statistical power. Accordingly, we developed and validated a prognostic index (PIUKALL) that integrates multiple risk factors and uses continuous data. We created discovery (n = 2405) and validation (n = 2313) cohorts using data from 4 recent trials (UKALL2003, COALL-03, DCOG-ALL10, and NOPHO-ALL2008). Using the discovery cohort, multivariate Cox regression modeling defined a minimal model including white cell count at diagnosis, pretreatment cytogenetics, and end-of-induction minimal residual disease. Using this model, we defined PIUKALL as a continuous variable that assigns personalized risk scores. PIUKALL correlated with risk of relapse and was validated in an independent cohort. Using PIUKALL to risk stratify patients improved the concordance index for all end points compared with traditional algorithms. We used PIUKALL to define 4 clinically relevant risk groups that had differential relapse rates at 5 years and were similar between the 2 cohorts (discovery: low, 3% [95% confidence interval (CI), 2%-4%]; standard, 8% [95% CI, 6%-10%]; intermediate, 17% [95% CI, 14%-21%]; and high, 48% [95% CI, 36%-60%; validation: low, 4% [95% CI, 3%-6%]; standard, 9% [95% CI, 6%-12%]; intermediate, 17% [95% CI, 14%-21%]; and high, 35% [95% CI, 24%-48%]). Analysis of the area under the curve confirmed the PIUKALL groups were significantly better at predicting outcome than algorithms employed in each trial. PIUKALL provides an accurate method for predicting outcome and more flexible method for defining risk groups in future studies.

Design of a Comprehensive Fluorescence in Situ Hybridization Assay for Genetic Classification of T-Cell Acute Lymphoblastic Leukemia.

T-cell acute lymphoblastic leukemia (T-ALL) results from deregulation of a number of genes via multiple genomic mechanisms. We designed a comprehensive fluorescence in situ hybridization (CI-FISH) assay that consists of genomic probes to simultaneously investigate oncogenes and oncosuppressors recurrently involved in chromosome rearrangements in T-ALL, which was applied to 338 T-ALL cases. CI-FISH provided genetic classification into one of the well-defined genetic subgroups (ie, TAL/LMO, HOXA, TLX3, TLX1, NKX2-1/2-2, or MEF2C) in 80% of cases. Two patients with translocations of the LMO3 transcription factor were identified, suggesting that LMO3 activation may serve as an alternative to LMO1/LMO2 activation in the pathogenesis of this disease. Moreover, intrachromosomal rearrangements that involved the 10q24 locus were found as a new mechanism of TLX1 activation. An unequal distribution of cooperating genetic defects was found among the six genetic subgroups. Interestingly, deletions that targeted TCF7 or TP53 were exclusively found in HOXA T-ALL, LEF1 defects were prevalent in NKX2-1 rearranged patients, CASP8AP2 and PTEN alterations were significantly enriched in TAL/LMO leukemias, and PTPN2 and NUP214-ABL1 abnormalities occurred in TLX1/TLX3. This work convincingly shows that CI-FISH is a powerful tool to define genetic heterogeneity of T-ALL, which may be applied as a rapid and accurate diagnostic test.

Concordance of copy number abnormality detection using SNP arrays and Multiplex Ligation-dependent Probe Amplification (MLPA) in acute lymphoblastic leukaemia.

In acute lymphoblastic leukaemia, MLPA has been used in research studies to identify clinically relevant copy number abnormality (CNA) profiles. However, in diagnostic settings other techniques are often employed. We assess whether equivalent CNA profiles are called using SNP arrays, ensuring platform independence. We demonstrate concordance between SNP6.0 and MLPA CNA calling on 143 leukaemia samples from two UK trials; comparing 1,287 calls within eight genes and a region. The techniques are 99% concordant using manually augmented calling, and 98% concordant using an automated pipeline. We classify these discordant calls and examine reasons for discordance. In nine cases the circular binary segmentation (CBS) algorithm failed to detect focal abnormalities or those flanking gaps in IKZF1 probe coverage. Eight cases were discordant due to probe design differences, with focal abnormalities detectable using one technique not observable by the other. Risk classification using manually augmented array calling resulted in four out of 143 patients being assigned to a different CNA risk group and eight patients using the automated pipeline. We conclude that MLPA defined CNA profiles can be accurately mirrored by SNP6.0 or similar array platforms. Automated calling using the CBS algorithm proved successful, except for IKZF1 which should be manually inspected.

SH2B3 inactivation through CN-LOH 12q is uniquely associated with B-cell precursor ALL with iAMP21 or other chromosome 21 gain.

In more than 30% of B-cell precursor acute lymphoblastic leukaemia (B-ALL), chromosome 21 sequence is overrepresented through aneuploidy or structural rearrangements, exemplified by intrachromosomal amplification of chromosome 21 (iAMP21). Although frequent, the mechanisms by which these abnormalities promote B-ALL remain obscure. Intriguingly, we found copy number neutral loss of heterozygosity (CN-LOH) of 12q was recurrent in iAMP21-ALL, but never observed in B-ALL without some form of chromosome 21 gain. As a consequence of CN-LOH 12q, mutations or deletions of the adaptor protein, SH2B3, were converted to homozygosity. In patients without CN-LOH 12q, bi-allelic abnormalities of SH2B3 occurred, but only in iAMP21-ALL, giving an overall incidence of 18% in this sub-type. Review of published data confirmed a tight association between overrepresentation of chromosome 21 and both CN-LOH 12q and SH2B3 abnormalities in B-ALL. Despite relatively small patient numbers, preliminary analysis linked 12q abnormalities to poor outcome in iAMP21-ALL (p = 0.03). Homology modelling of a leukaemia-associated SH2 domain mutation and in vitro analysis of patient-derived xenograft cells implicated the JAK/STAT pathway as one likely target for SH2B3 tumour suppressor activity in iAMP21-ALL.

Validation of the United Kingdom copy-number alteration classifier in 3239 children with B-cell precursor ALL.

Genetic abnormalities provide vital diagnostic and prognostic information in pediatric acute lymphoblastic leukemia (ALL) and are increasingly used to assign patients to risk groups. We recently proposed a novel classifier based on the copy-number alteration (CNA) profile of the 8 most commonly deleted genes in B-cell precursor ALL. This classifier defined 3 CNA subgroups in consecutive UK trials and was able to discriminate patients with intermediate-risk cytogenetics. In this study, we sought to validate the United Kingdom ALL (UKALL)-CNA classifier and reevaluate the interaction with cytogenetic risk groups using individual patient data from 3239 cases collected from 12 groups within the International BFM Study Group. The classifier was validated and defined 3 risk groups with distinct event-free survival (EFS) rates: good (88%), intermediate (76%), and poor (68%) (P < .001). There was no evidence of heterogeneity, even within trials that used minimal residual disease to guide therapy. By integrating CNA and cytogenetic data, we replicated our original key observation that patients with intermediate-risk cytogenetics can be stratified into 2 prognostic subgroups. Group A had an EFS rate of 86% (similar to patients with good-risk cytogenetics), while group B patients had a significantly inferior rate (73%, P < .001). Finally, we revised the overall genetic classification by defining 4 risk groups with distinct EFS rates: very good (91%), good (81%), intermediate (73%), and poor (54%), P < .001. In conclusion, the UKALL-CNA classifier is a robust prognostic tool that can be deployed in different trial settings and used to refine established cytogenetic risk groups.

Dynamic clonal progression in xenografts of acute lymphoblastic leukemia with intrachromosomal amplification of chromosome 21.

Intrachromosomal amplification of chromosome 21 is a heterogeneous chromosomal rearrangement occurring in 2% of cases of childhood precursor B-cell acute lymphoblastic leukemia. These abnormalities are too complex to engineer faithfully in animal models and are unrepresented in leukemia cell lines. As a resource for future functional and preclinical studies, we have created xenografts from the leukemic blasts of patients with intrachromosomal amplification of chromosome 21 and characterized them by in-vivo and ex-vivo luminescent imaging, flow immunophenotyping, and histological and ultrastructural analyses of bone marrow and the central nervous system. Investigation of up to three generations of xenografts revealed phenotypic evolution, branching genomic architecture and, compared with other B-cell acute lymphoblastic leukemia genetic subtypes, greater clonal diversity of leukemia-initiating cells. In support of intrachromosomal amplification of chromosome 21 as a primary genetic abnormality, it was always retained through generations of xenografts, although we also observed the first example of structural evolution of this rearrangement. Clonal segregation in xenografts revealed convergent evolution of different secondary genomic abnormalities implicating several known tumor suppressor genes and a region, containing the B-cell adaptor, PIK3AP1, and nuclear receptor co-repressor, LCOR, in the progression of B-cell acute lymphoblastic leukemia. Tracking of mutations in patients and derived xenografts provided evidence for co-operation between abnormalities activating the RAS pathway in B-cell acute lymphoblastic leukemia and for their aggressive clonal expansion in the xeno-environment. Bi-allelic loss of the CDKN2A/B locus was recurrently maintained or emergent in xenografts and also strongly selected as RNA sequencing demonstrated a complete absence of reads for genes associated with the deletions.