1. Identification of mutations prognostic of relapse after allogeneic transplantation and novel clone emergence at disease recurrence: implications for strategies to prevent post transplant relapse. P Ferguson, L Quek, M Metzner, I Ahmed, C Garnett, S Jeffries, K Piechocki, R Danby, M Raghavan, A Peniket, M Griffiths, A Bacon, J Ward, K Wheatley, P Vyas, C Craddock

2. Introduction The main cause of treatment failure following transplantation for AML remains disease relapse. With the advent of next generation DNA sequencing (NGS) we are uncovering the complex mutational landscape present in AML. NGS studies have been applied to patients receiving induction chemotherapy and demonstrated that specific mutations, mutation number and kinetics of mutation clearance can predict for response to induction therapy. Patel et al. New England Journal of Medicine. 2012; 366;12: 1079-1089 Klco et al. Journal of the American Medical Association. 2015; 314: 811-822 Wakita S et al. Leukemaia. 2016; 30: 545-554 Ivey et al. New England Journal of Medicine. 2016; 374:422-433

3. To date no study has performed a detailed mutational analysis of AML and correlated this with clinical outcomes following allogeneic transplantation. Furthermore the changes in clonal structure that occur through treatment and relapse following transplantation have not been elucidated. Strategies which reduce the risk of disease relapse after transplantation are urgently required but are dependant on a deeper understanding of disease biology post relapse.

4. Aims To determine whether mutations present at diagnosis could predict for disease relapse following allogeneic transplantation. To understand the changes in clonal structure that occur between diagnosis and disease relapse post transplant and how this might impact the development of strategies to reduce relapse risk after allograft.

5. Methods Performed targeted resequencing of 34 AML-associated genes on diagnostic bone marrow samples of 113 adults who underwent allogeneic transplantation. Analysed bone marrow aspirates from 29/49 patients who relapsed post-allograft, including 20 CR samples taken prior to transplant. Libraries constructed using Fluidigm Access Array were sequenced on a MiSeq platform. Variant calling was performed using Varscan, GATK and Pindel.

6. Recurrent somatic variants previously associated with AML were included in the analysis. Novel variants were validated by Sanger sequencing. Mutational analysis was compared at diagnosis, remission and relapse. Performed multivariate analysis of relapse risk, relapse free survival and overall survival.

7. Molecular predictors of outcome post transplant

8. Distribution of mutations Ley et al. New England Journal of Medicine. 2013; 368: 2059-2074

9. Relapse risk IDH1 - improved relapse risk (p=0.03) WT1 (p=0.03) DNMT3A (p=0.05) FLT3-ITD (p=0.06) TP53 (p=0.07) Overall survival TP53 (p=0.027) IDH1 - improved OS (p=0.058) DNMT3A (p=0.082) Multivariate analysis of relapse risk and overall survival in all patients

10. Relapse Risk DNMT3A (p=0.016) FLT3-ITD (p=0.021) TP53 (p=0.071). Overall survival TP53 (p=0.027) Multivariate analysis of relapse risk and overall survival in patients with non- adverse cytogenetics

11. Analysis of clonal architecture at diagnosis and relapse

12. Analysis of paired diagnostic/Relapse samples 17/29 had no change in mutational profile at relapse compared with diagnosis. 12/29 had changes in mutational profile between diagnosis and relapse – 9/12 patients acquired new mutations at relapse not detected at diagnosis with or without an associated loss of mutation. – 3/12 patients had a loss of mutations between diagnosis and relapse.

13. Evolution of pre-existing clones at relapse UPN 56UPN 11 Variant Allele Frequency %

14. Emergence of de novo clones at relapse UPN 20 UPN 112 Variant Allele Frequency %

15. No evidence of clonal evolution and mutations not detectable at CR UPN 1UPN 10 Variant Allele Frequency %

16. No evidence of clonal evolution but persistence of mutations at CR UPN 64UPN 61 Variant Allele Frequency %

17. Acquisition and loss between diagnosis and relapse with mutations present at relapse also detected at CR UPN 35 Variant Allele Frequency %

18. Conclusion Demonstrated similar background mutational patterns to that previously published by the TCGA group with enrichment for FLT3-ITD, KIT, SRSF2 and TET2 mutants and depletion of NPM1 mutants in our transplant patient cohort. We have shown that mutations in IDH1, WT1, FLT3-ITD and DNMT3A can predict for relapse following allogeneic transplantation for AML. Mutations in TP53 predict for reduced overall survival in patients transplanted for AML.

19. We demonstrate, for the first time, that the dominant AML clone at relapse post allograft is molecularly distinct from clones at presentation in a substantial number of patients. Our data shows that this clonal evolution involves multiple genetic mechanisms including loss of pre-existing mutations and acquisition of new mutations. The emergence of novel mutations at relapse and loss of pre- existing ones has important implications for the design of post-transplant maintenance strategies in patients allografted for AML.

20. These observations throw doubt on the utility of maintenance strategies utilising targeted therapeutics which are likely to have minimal activity in emerging clones at relapse in some patients. Interventions with a broader anti-leukaemic activity, such as donor lymphocyte infusions, or alternative pharmacological agents with a broader up-regulation of the GVL response may be preferable. These data now require validation in a larger patient cohort.

21. Thank you

22. Patient demographics Age range 16-70 yrs Median 49 yrs Cytogenetics – Favourable/Intermediate 91 (80.5%) – Adverse 22 (19.5%) Status at transplant – CR 105 (93%) CR1 78 CR2 27 – NCR 8 (7%)

23. Transplant conditioning – Myeloablative 38 (34%) – Reduced intensity 75 (66%) Donor source – Sibling 49 (43%) – Unrelated 61 (54%) – Cord 3 (3%)

24. TP53 mutations are associated with a decreased OS (p=0.027). Trend towards significance – IDH1 mutations for improved OS (p=0.058) – DNMT3A mutations for reduced OS (p=0.082) Multivariate analysis of overall survival in all patients

25. Multivariate analysis of relapse risk in all patients

26. Multivariate analysis of overall survival in all patients

27. Multivariate analysis of relapse risk in patients with non-adverse cytogenetics

28. Multivariate analysis of overall survival in patients with non-adverse cytogenetics

29. Analysis of paired diagnostic/Relapse samples Molecular characterization of paired diagnosis and relapse samples showed novel mutations in 20/30 patients. 11/20 (55%) had changes in mutation number, with or without cytogenetic change. 9/20 (45%) patients had new mutations at relapse not detected at diagnosis (VAF sensitivity of <0.5%). Multiple distinct mutational events occurred in 5/9 patients.

30. In contrast, 5/30 (16%) had no detectable changes in either mutations or karyotype at relapse. 7 patients had cytogenetic changes only. In 4 patients changes in clonal structure were identified at relapse by mutational profiling only, with no cytogenetic changes detected. We could deduce clonal evolution from pre-existing clones by mutation profile in 6 patients; or from de novo clones with loss of pre-existing clones in 3 patients

31. No evidence of clonal evolution but persistence of mutations at CR

32. De novo mutations acquired at relapse and not detected at CR