AML Xenografts Undergo Extensive Clonal Competition and Unmask Rare Clones in Patient Samples

The biology of acute myeloid leukemia (AML) is largely dictated by the complex combinations of mutations that initiate and propagate the disease. It is therefore critical that models accurately reflect the mutation patterns of the specific leukemias under investigation. In recent years, patient-derived xenografts have emerged as one of the most faithful models to study AML and evaluate potential treatments in vivo. However, little is still known about how leukemias are affected by transplantation and consecutive passaging in vivo. To address this issue, we characterized the mutational patterns and clonal dynamics in 19 AML patient samples and 65 corresponding xenografts from primary and secondary recipients by whole-exome sequencing.

We found that xenografts from each patient generally displayed the same mutational profile but with striking differences from the corresponding patient sample as a result of extensive clonal competition. Only 1 of 19 patient samples (5%) retained its mutation pattern in vivo and an additional 3 of 19 samples (16%) displayed gains or losses of non-recurrent presumed passenger mutations. However, the remaining 15 of 19 samples (79%) were shown to contain multiple clones with different recurrent AML mutations that engrafted according to one of the four identified patterns of clonal dynamics: loss, expansion, burst and stable. In 5 of 19 samples (26%), we observed a loss or reduction of subclones containing either NRAS, KRAS, CDKN2A, RUNX1T1 or del(19)(p13), giving way to the respective parental clones, despite allele frequencies in the diagnostic samples suggesting these mutations to be present in the entire cell populations. Seven of 19 samples (37%) showed an expansion of a minor subclone in the patient sample to the point where it made up the entire xenograft. This was paralleled by the complete loss of parental clones as well as sibling clones carrying other recurrent AML mutations such as BCOR, NRAS and WT1. Furthermore, two of the expanding subclones (with TET2 loss of heterozygosity and monosomy 7) were present but not detected by sequencing (100x) of the diagnostic samples, showing that xenografts can unmask unexpected clonal complexity and rare subclones in patient samples. Notably, certain expanding subclones were defined by mutations in NRAS and CDKN2A, which were both lost in other patient samples, suggesting that the interplay between driver mutations is complex and that engraftment cannot be predicted solely based on single mutations. In an additional 2 of 19 samples (11%), a subclonal burst was observed. In these cases, subclones with mutations in SMC3 and PTPN11, respectively, expanded in primary recipients but were lost upon secondary transplantation, indicating that subclones may differ not only in their ability to initiate but also to sustain leukemia in vivo. This notion is further supported by the observation that certain subclones that decreased in vivo only did so in secondary recipients. Notably, only 1 of 19 patient samples (5%) displayed a pattern of stable clonal dynamics, where a subclone was retained at its original frequency throughout primary and secondary passaging.

In conclusion, patient-derived xenografts are highly representative of the corresponding patient samples, in that the cells maintain their mutation patterns and gain very few, if any, novel driver mutations throughout consecutive passaging in vivo. However, most AML patient samples comprise multiple clones that differ in their ability to initiate and propagate leukemia in vivo. This leads to xenografts with clonal compositions and mutation patterns that are significantly different from the patient samples, demonstrating the need for genetic profiling of individual xenografts for precise studies of AML biology and treatment.