Combining Whole Exome Sequencing and Rnaseq to Provide a Comprehensive Landscape of the Mechanisms of Post-Transplantation Leukemia Relapse

Allogeneic hematopoietic stem cell transplantation (allo-HSCT) represents an effective treatment for many hematological malignancies, but post-transplantation relapses remain frequent, and their biological bases poorly understood.

Here we combined Whole Exome Sequencing (WES) and RNA-Seq to compare the features of 15 cases of Acute Myeloid Leukemia before allo-HSCT and at post-transplantation relapse. Leukemic blasts were collected and purified at the two timepoints, and compared between each other and with patient and donor germline controls. Three donors were HLA-identical siblings, 7 were matched unrelated volunteers (all of which were matched to the patient for 10/10 HLA loci) and 5 were HLA-haploidentical relatives. Median time from allo-HSCT to relapse was 144 days (range 33-574).

The average coverage for WES, consistent between all cases, was 120x for leukemia samples and 60X for germline controls.

Analysis of copy number alterations did not reveal major genomic alteration acquired at relapse, except for one patient who gained trisomy for chr. 6 and 21.

Also the transitions/transversions ratio remained constant between the pre-Tx and post-Tx samples.

We detected an average of 15 somatic variants (SNV and small InDel) per leukemia sample, and the overall mutational burden increased significantly between pre- and post-Tx samples (Wilcoxon test, pvalue = 0.0088). We evidenced 4 major patterns of clonal evolution. In pattern 1 (n=3), the pre-Tx clones persisted unchanged at relapse, in pattern 2 (n=6) and 3 (n=2), subclones were gained or lost, respectively, whereas in pattern 4 (n=4) we found a mixed scenario.

Sixty somatic variants were present only in relapsed samples, encompassing known AML driver genes, including KRAS and WT1 (de novo mutated at relapse in 2 patients). Unexpectedly, WES analysis did not detect relapse-specific mutations in genes related to immune function, and even an ad hoc developed pipeline for the analysis of somatic mutations in HLA class I and class II genes did not detect any denovo acquired sequence abnormality.

Conversely, a linear model analysis of RNA-seq showed ~800 genes significantly deregulated in AML blasts at post-transplantation relapse. The down-regulated genes were mainly immune-related, encompassing in particular those involved in HLA class II antigen presentation. Upregulated genes comprised genes relevant to DNA replication and cell cycle control (including several component of Minichromosome Maintanance Complex). Of interest, these two processes appeared also to cluster independently in our patient series, suggesting the presence of different transcriptional mechanisms of relapse.

Of interest, we observed HLA class II downregulation also in several cases in which donor and patient were matched for those loci. Thus, to understand whether the decreased expression of antigen presentation molecules could in these cases be driven by an increase in the levels of presented antigens, we extracted from the RNA-seq dataset information regarding known leukemia associated antigens, finding several of them upregulated at post-transplantation relapse (MPO, TERT, PRTN3)

Moreover, we combined WES and RNA-seq data to predict the number of patient-specific neoantigens and minor histocompatibility antigens (MiHAgs) presented on leukemia blast before and after allo-HSCT. In line with the low overall burden of mutations, we predicted a very low number of neoantigens per case (on average 3 per sample, ranging from 0 to 20), increasing at relapse in 5/15 patients. The number of predicted MiHAgs was sizably higher, and varied considerably in relation to donor-recipient matching (on average 481 in haploidentical HSCTs, 874 in HSCTs from HLA-identical siblings, and 1435 in unrelated donor HSCTs). However, the overall level of expression of MiHAgs did not vary between pre- and post-Tx samples

Our results provide for the first time a detailed landscape of the many features that shape the interplay between immune system and leukemia in allo-HSCT. The resulting picture is composite, suggesting that mechanisms of relapse are highly patient-specific and combine genomic and non-genomic, immunological and non-immunological changes. For this reason we modeled our data in a comprehensive framework that we termed "relapsogram", that might help in elucidating the peculiarity of each case of relapse and in customizing the therapy.