PDX Models Recapitulate Genetics and Epigenetics of Pediatric T-Cell Leukemia

Patient-derived xenografts (PDX) emerged as important tools to study leukemia biology and potential therapeutic regimens. Here, we compared cellular, molecular and developmental features of PDX-derived blasts with primary pediatric precursor T-cell acute leukemia (T-ALL) cells. The cells were obtained at the times of initial diagnosis and relapse. We analyzed the samples at the genomic level (whole genome sequencing (WGS), whole exome sequencing (WES), multiplex ligation probe amplification (MLPA), targeted sequencing) and the epigenetic level (DNA methylation, Assay for Transposase-Accessible Chromatin sequencing (ATAC-seq)).

Altogether 75% of our samples engrafted successfully. Of these 12 were matched pairs of initial diagnosis and relapse as well as corresponding PDX. Of further 12 matched pairs either initial diagnosis (2/12) or relapse (7/12) or neither (3/12) engrafted indicating that relapse samples tend to engraft more successfully (p= 0.1824 (McNemar's test)), which is in agreement with observations in other cancer entities.

In 11 out of 12 matched pairs (initial and relapse) of primary and PDX genomic stability of the PDXs was assessed using WGS and WES. Ninety-two % of the mutations with an allele frequencies (AF) >30% were conserved. Of altogether 609 SNVs and InDels (AF>10%) detected in primary samples 68 % (417) were also detected in the corresponding PDXs. In some PDXs, mutations not identified by targeted sequencing in the primary sample (NOTCH1, PHF6, DNA11, USP9X) occured at a high AF (22-59%), which may have been the result of selection of a subclone or of an acquisition of the mutation in the process of PDX generation.

Thirty of 35 (86%) large CNA assessed by WGS were preserved in corresponding PDX models indicating that CNAs are likely earlier events in the leukemogenesis that are present at high AF in the primary samples. CNA analysis with MLPA confirmed that PDXs retained 96% (80/83) of the deletions and amplifications of primary leukemia.

In 18 out of the matched 22 PDXs all clonal mutations (AF >30%) detected in the primary sample have been preserved in the PDX. Seventeen out of 22 contained more than 1 clone. Fifteen out of these 17 retained the complexity of primary sample with different subclones successfully engrafted.

When considering the 11 matched pairs, we observed 4 different engraftment types of initial disease. Type 1 engrafted in a similar way in which they relapsed preserving all or nearly all mutations found in patients' sample (4 samples). Type 2: the relapsing clone that was not detected in initial diagnosis was present in PDX derived of initial diagnosis (PDX_INI) but it was not dominant (3). Type 3: the relapsing clone that was present at initial diagnosis did not engraft into PDX_INI (2). Type 4 preserved the relapsing clone as well as the mutations specific for initial diagnosis (2). Out of 11 PDXs of relapse 9 preserved all or nearly all mutations detected in primary sample. In 2/11 PDXs of relapse one of the clones preexisting in primary relapse was selected for, while the other one was lost.

Differential methylation analysis by EPIC array of 21,711 promoters of primary samples and PDXs revealed that the average methylation in primary samples (β=0.43) did not differ from that in PDXs (β=0.45). Principal Component Analysis and unsupervised hierarchical clustering showed that matched pairs, both of the primary and the PDX samples cluster together (ρ=0.9912). Similarly, analysis of chromatin accessibility by ATAC-seq showed that more than 92% of peaks of a height > 20 and more than 75% of the peaks of a height of > 10 were preserved between primary samples and corresponding PDXs.

In conclusion, PDX models largely reflect the heterogeneity and the complexity of the primary leukemic sample at the genomic and epigenomic level, although some clonal and particularly subclonal mutations can be lost. Differences in engraftment and various engraftment scenarios of initial diagnosis in contrary to relapse samples suggest that relapse samples are more preselected and stable in engraftment. Altogether, this overall genomic and epigenomic stability of the primary T-ALL in the PDX renders the models as a suitable source for material when primary samples are limiting and also to reduce uncontrollable pre-analytical changes in primary samples. Further, these models are suitable for the preclinical development of new treatment strategies for pediatric T-ALL.