Transcriptional and genomic characterization of measurable residual disease in acute myeloid leukaemia.

Failure to achieve complete response (CR) and persistence of measurable residual disease (MRD) are associated with inferior survival in acute myeloid leukaemia (AML).1 Poor outcomes are attributed to primary chemotherapy resistance, as well as to high relapse rates among AML patients achieving CR.2, 3 A better understanding of why therapies are unable to eradicate these residual leukaemic cells could be relevant, particularly in elderly AML for which outcomes remain dismal despite increasing CR rates with novel combinations. However, there are virtually no data on the molecular traits of tumour cells persisting in patients achieving less than CR and those in CR but positive MRD (CR/MRD+) because it requires patient-matched longitudinal samples, as well as the ability to detect and isolate resistant cells after pre-specified time points and in a setting of homogenous treatment.4 Here, we analysed mechanisms of resistance by comparing the transcriptional and genomic profile of paired leukaemic cells at diagnosis and after treatment with semi-intensive chemotherapy or a hypomethylating agent, in patients achieving partial remission (PR) and CR/MRD+. Our results identified deregulated genes and recurrent mutations with limited or previously unknown role in AML pathobiology, which were associated with treatment resistance and inferior survival. The phase 3 PETHEMA-FLUGAZA clinical trial (NCT02319135) was the framework for this study (Figure 1A, Methods S1). Two-hundred and eighty-three patients aged over 65 with newly diagnosed AML (excluding those with acute promyelocytic leukaemia according to World Health Organization [WHO] criteria5) and with an Eastern Cooperative Oncology Group PS <4, were enrolled. All patients provided written informed consent, and the trial was approved by institutional review boards or ethics committees at all sites. The study was conducted according to the Declaration of Helsinki. Patients were randomized 1:1 to receive open-label treatment with either azacitidine (AZA) or low-dose Ara-C (cytarabine) plus fludarabine (FLUGA). Patients in CR with or without incomplete blood count recovery (CR/CRi), PR, haematology improvement or stable disease after three induction cycles continued with consolidation, which consisted of six additional cycles. Response was assessed according to International Criteria.6 At the end of the ninth cycle, patients in CR/CRi had bone marrow (BM) aspirates for MRD assessment. Those with MRD levels ≥0.01% continued treatment (AZA or FLUGA) until relapse or progressive disease were documented. Patients whose MRD levels were <0.01% suspended treatment and proceeded to follow-up. MRD was assessed after induction and consolidation using multidimensional flow cytometry as described elsewhere.7 The main outcomes of the FLUGAZA clinical trial were reported recently.8 Patients' clinical data and sample’ disposition are detailed in Table S1. After induction, 35 patients were in PR, 51 attained CR/CRi but showed persistent MRD (CR/MRD+), whereas 13 achieved CR/CRi and undetectable MRD (CR/MRD−). There were no differences in overall survival (OS) between patients in PR versus CR/MRD+ (median of 21 months vs. 20 months, respectively; p = 0.603), and both subgroups showed a trend of inferior OS when compared to patients in CR/MRD− (median of 27 months, p = 0.097) (Figure 1B). In line with previous studies, the OS of patients with undetectable MRD was poor.9 These findings urge a more sensitive assessment of MRD in elderly AML, which may potentially benefit from the combination of immunophenotypic and molecular methods as shown in patients eligible for intensive therapy.10 The patient-specific aberrant phenotypes of leukaemic cells persisting after treatment guided their subsequent isolation using fluorescence-activated cell sorting (Figure 1A). A 3′-end RNAseq method (MARS-seq) optimized for generating libraries from low-input starting material adapted for bulk RNA-seq11 was used to analyse transcriptomes of resistant leukaemic cells from 22 patients (10 achieving PR and 12 attaining CR/MRD+). The same RNAseq method was performed in leukaemic cells at diagnosis from 220 patients, including the 22 mentioned above. Differential gene expression analysis was performed in R using DESeq2 (Methods S1).12 There was only one gene (PIEZO2) differentially expressed between leukaemic cells at diagnosis versus after treatment in patients achieving PR. By contrast, there were 117 differentially expressed genes (adjp <0.05, log2FoldChange > |2|) between leukaemic cells at diagnosis vs after treatment in CR/MRD+ patients (Figure S2, Table S2). Only one of the 117 genes showed differential expression specific to treatment: AHSP was overexpressed in patients receiving AZA but not FLUGA (Figure S1A). There was a trend of prolonged OS with AZA versus FLUGA among patients overexpressing AHSP (Figure S1B). Furthermore, upregulation of this gene has been previously associated with favourable prognosis in FLT3-ITD negative AML patients.13 To confirm that gene deregulation was intrinsically associated with treatment resistance, the prognostic value of gene over- or underexpression in accordance with the differential observed between diagnostic and MRD leukaemic cells was subsequently analysed by RNAseq in the 220 patients at diagnosis. Under- and overexpression of 15 genes among the 117 identified above was significantly associated with inferior OS (Figure 1C, Table S3). Only a few of these genes had a previously ascribed role in AML pathogenesis and/or chemoresistance. Namely, CASC15 which may act as a proto-oncogene affecting proliferation and differentiation,14 KDM7A that is a lysine-specific histone demethylase involved in cell differentiation and resistance to chemotherapy,15 and IGFBP2 that is associated with chemoresistance.16 Interestingly, none of the 15 genes differentially expressed in persisting leukaemic cells and predictive of inferior OS in the FLUGAZA trial was prognostic in the BeatAML cohort (Table S3).17 This finding suggests that the transcriptome of MRD resistance may be treatment-specific. In addition to RNAseq, whole-exome sequencing (WES) was performed using molecular barcoding in paired diagnostic-MRD leukaemic cells from 14 patients (6 of them having paired RNAseq and WES data) (Table S1), all of whom achieving CR/MRD+ (Figure 1A). Variants were annotated using HD Genome One (DREAMgenics), based on several databases containing functional (Ensembl, CCDS, RefSeq, Pfam), populational (dbSNP, 1000 Genomes, ESP6500, ExAC) and cancer-related (COSMIC, ICGC) data (Methods S1). Variants were considered if the allele frequency was ≥0.05 in leukaemic blasts and ≤0.2 in T cells, providing the coverage was greater than 9 reads in each cell type being sequenced and compared. A total of 6054 mutations were detected after WES of matched leukaemic cells at diagnosis and after induction. Among these, 4708 (78%) were detected at both time points, 354 (6%) were present at diagnosis while absent in MRD blasts, and 992 (16%) emerged during MRD resistance (p ≤ 0.011). The possibility that some of the later mutations are due to contamination with clonal haematopoiesis cannot be excluded.18, 19 There were no differences between treatment arms in the number of shared and private mutations in diagnostic versus MRD cells (Figure 2A, Figure S1A). Shared mutations comprised mutations in well-known AML driver genes, for example, ASXL1, DNMT3A and RUNX1. Among the 1346 mutations that either became undetectable or present at MRD, recurrence in 3 or more patients was observed in 48 genes and time points exclusivity were observed in 20 genes (Figure S3). Recurrent mutations at diagnosis that became undetectable in MRD cells were exclusively observed in the ADAMTSL5, KIAA1522 and KRTAP4-11 genes (Figure 2B), while recurrent mutations emerging de novo in MRD cells were exclusively observed in 17 genes; of note, AVP and CACNG4 were mutated after treatment in 6 and 5 of the 14 patients, respectively (Figure 2B). Interestingly, mutations in MFSD4A and R3HDM2 genes were only present in MRD cells from patients treated with FLUGA (Figure S3). In conclusion, this study showed that while elderly AML patients have similarly dire survival regardless of achieving PR or CR/MRD+, the former appear to be characterized by primary resistance, whereas CR/MRD+ is associated with the emergence of molecular traits of acquired resistance. These findings could help explaining why in the PETHEMA-FLUGAZA clinical trial, additional treatment with the same schema was unable to overcome the poor prognosis of persistent MRD in elderly AML patients achieving CR/CRi. On genomic grounds, this study identified a few recurrently mutated genes with limited or previously unknown role in AML pathobiology. Taken together, it could be hypothesized that mechanisms of resistance may be treatment-specific and, therefore, should be analysed in well-defined protocols, larger cohorts and using MRD assessments beyond its prognostic utility towards the isolation and characterization of resistant leukaemic cells. Catia Simoes, Jesús F. San-Miguel, Pau Montesinos and Bruno Paiva involved in study conception and design. Catia Simoes, Sara Villar, Beñat Ariceta, Juan-José Garcés and Bruno Paiva involved in analysis and data interpretation. Catia Simoes and Bruno Paiva involved in statistical analysis. Sara Villar, Leire Burgos, Diego Alignani, Sarai Sarvide, David Martínez-Cuadrón, Juan-Miguel Bergua, Susana Vives, Lorenzo Algarra, Mar Tormo, Pilar Martinez, Josefina Serrano, Pilar Herrera, Fernando Ramos, Olga Salamero, Esperanza Lavilla, Cristina Gil, Jose-Luis Lopez-Lorenzo, Maria-Belen Vidriales, Carmen Chillon, Jorge Labrador, Jose-Francisco Falantes, María-José Sayas, Rosa Ayala, Joaquin Martinez-Lopez, Ana Alfonso Pierola, Maria-Jose Calasanz, Felipe Prosper, Jesús F. San-Miguel, Miguel Á. Sanz and Pau Montesinos involved in the provision of study material and/or patients. Catia Simoes and Bruno Paiva wrote the manuscript. All authors reviewed and approved the manuscript. We acknowledge patients, caregivers, Biobank of the University of Navarra and supported centers: Centro de Investigación Biomédica en Red – Área de Oncología - del Instituto de Salud Carlos III, Instituto de Salud Carlos III/Subdirección General de Investigación Sanitaria, Plan de Investigación de la Universidad de Navarra and Cancer Research UK FC AECC and AIRC under the Accelerator Award Program (EDITOR). This study was supported by the Centro de Investigación Biomédica en Red – Área de Oncología - del Instituto de Salud Carlos III (CIBERONC; CB16/12/00369, CB16/12/00233, CB16/12/00489 and CB16/12/00284), Instituto de Salud Carlos III/Subdirección General de Investigación Sanitaria (FIS No. PI16/01661 and PI16/00517 and PI19/01518) and the Plan de Investigación de la Universidad de Navarra (PIUNA 2014–18). This study was supported internationally by the Cancer Research UK [C355/A26819] FC AECC and AIRC under the Accelerator Award Program (EDITOR). F.P.: Honoraria and research funding: Oryzon, Janssen, BMS-Celgene. R.A.: Membership on an entity's Board of Directors advisory committees: Incyte Corporation, Astellas; Honoraria: Novartis, Celgene and Incyte. J.A.P.S.: Honoraria or budget for research projects and/or participation in advisory board and learning activities or conferences: Janssen, Takeda, Pfizer, Jazz, BMS, Amgen, Gilead. J.F.S.M.: Consultancy, membership on an entity's Board of Directors advisory committees: AbbVie, Amgen, Bristol-Myers Squibb, Celgene, GlaxoSmithKline, Janssen, Karyopharm, Merck Sharpe & Dohme, Novartis, Regeneron, Roche, Sanofi, SecuraBio, Takeda. P.M.: Consultancy, membership on an entity's Board of Directors advisory committees, research funding, speaker's bureau: Celgene, Sanofi, Incyte, Karyopharm, Novartis, Stemline/Menarini, Agios, Astellas Pharma, Daiichi Sankyo; Membership on an entity's Board of Directors advisory committees: Pfizer, Teva, AbbVie; Research Funding, Speakers Bureau: Janssen; Consultancy: Tolero Pharmaceutical, Forma Therapeutics, Glycomimetics. B.P.: served as a consultant for and received honoraria from Adaptive, Amgen, Becton Dickinson, Bristol Myers Squibb/Celgene, GSK, Janssen, Roche, Sanofi, and Takeda; and received research support from Bristol Myers Squibb/Celgene, GSK, Roche, Sanofi, and Takeda. The remaining authors declare that they have no competing interests. Not applicable. Phase 3 PETHEMA-FLUGAZA clinical trial (NCT02319135). The sequencing data of this manuscript is available in the European Genome-phenome Archive database (accession number EGAS00001006766). Data S1. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.