Mutational and Clonal Dynamics in Patient-Derived Xenografts of Acute Myeloid Leukemia

Patient-derived xenografts have emerged as an attractive model to faithfully recapitulate acute myeloid leukemia (AML) in vivo and to test the therapeutic efficacy of new treatment regiments. The most efficient host for AML cells appears to be the NSGS mouse strain engineered to express the human cytokines SCF, GM-CSF and FLT3L. However, little is known about how the underlying genomics affect engraftment in this model and how the genomics and clonality of the patient samples are affected by the successional passaging in vivo.

To address these questions, we transplanted cells from 27 AML patients to NSGS mice. First, we correlated engraftment to leukemia-associated mutations determined by whole-exome sequencing. The frequency of engraftment in primary recipients for the most common mutations in AML were as follows; FLT3-ITD: 29% (2/7), NPM1: 75% (9/12), DNMT3A: 55% (6/11), IDH1/2: 54% (7/13), TET2: 100% (5/5), all samples: 54% (13/24). Although samples carrying the FLT3-ITD mutation engrafted with relatively low frequency, they generated more prolific disease, with cell numbers several fold higher than for any other patient. Samples transplanted to multiple mice showed strikingly similar characteristics.

Next, we determined changes in mutation patterns and clonal composition by whole-exome sequencing of human myeloid cells sorted from primary and secondary recipient mice. For all the 11 patients analyzed, the variant allele frequencies (VAF) of the mutations found in the patient sample were increased to or maintained at around 50% by the first passage in vivo. This corresponds to a heterozygous mutation being present in the whole cell population and indicates that the xenotransplantation model enriches the leukemic cells. Importantly, no novel mutations in known AML-associated genes were detected after either the first or second passage in mice, demonstrating that the genotype of the patient sample is preserved during expansion in vivo.

We then studied the clonal evolution in the 3 patients who presented with multiple clones. Notably, all 3 cases displayed drastic changes in the allele frequencies of specific mutations. One of the patients had what appeared to be 2 clones at diagnosis, a major clone with 8 AML-associated mutations (VAF 30-55%) and a minor clone with an additional NRAS mutation (VAF 5%). After one passage in vivo, the BCOR mutation in the major clone had disappeared (VAF 52% to 0%), while the other mutations in the major clone were maintained or slightly increased and the NRAS mutation had increased drastically (VAF 5% to 46%). This shows that the minor clone containing the NRAS mutation must have branched from a cell with all the presumed major clone mutations except that in BCOR and that this subclone completely outcompeted the major clone in vivo. This clonal composition also remained upon secondary transplantation. The second patient presented with a major clone carrying 9 mutations (VAF 35-55%) and a small subclone containing an SMC3 mutation (VAF 2%). Also in this case, the small subclone vastly outgrew the presumed founding clone in multiple mice to a VAF of 25-35%. The third patient carried mutations in 11 genes at diagnosis. Upon in vivo passaging, the lowest-frequency mutation, in IDH2, had markedly increased (VAF 10% to 36%), whereas the 3 second-lowest frequency mutations had completely disappeared (VAF 20-23% to 0%), while the high-frequency mutations remained at close to 50%. This development reveals 3 subclones at diagnosis; a founding clone that had given rise to 2 independent subclones, the largest of which was lost upon transplantation and the smallest of which vastly expanded. Hence, for all 3 patients with multiple clones, the smallest subclone drastically expanded in vivo at the expense of the others. This may reflect the biology in the patient, where subclones can only reach detectable levels by expanding much more rapidly than the founding clone. We show that this process continues in the xenografts and may thus model the evolution from diagnosis to relapse.

In conclusion, our results suggest that AML patient cells generally maintain their genotype during passaging in vivo but that clonal competition drastically alters the mutational landscape, emphasizing the need for genetic characterization of patient-derived xenografts.