WT1 and DNMT3A Play an Essential Function and Represent Therapeutic Vulnerabilities in Certain AML Samples, As Shown By CRISPR/Cas9 Mediated Knockout in PDX Models In Vivo

Background:

The prognosis of patients with acute myeloid leukemia (AML) remains poor and novel therapeutic options are intensively needed. Targeted therapies specifically address molecules with essential function for AML and deciphering novel essential target genes is of utmost importance. Functional genomics via CRISPR\Cas9 technology paves the way for the systematic discovery of novel essential genes, but was so far mostly restricted to studying cell lines in vitro, lacking features of, e.g., primary tumor cells and the in vivo tumor microenvironment. To move closer to the clinical situation in patients, we used the CRISPR\Cas9 technology in patient-derived xenograft (PDX) models of AML in vivo.

Methods:

Primary tumor cells from seven patients with AML were transplanted into immunocompromised NSG mice and serially transplantable PDX models derived thereof. PDX models were selected which carry the AML specific mutations of interest at variant allele frequencies close to 0.5. PDX cells were lentivirally transduced to express the Cas9 protein and a sgRNA; successfully transduced PDX cells were enriched by flow cytometry gating on a recombinant fluorochrome or by puromycin. The customized sgRNA library was designed using the CLUE (www.crispr-clue.de) platform and cloned into a lentiviral vector with five different sgRNAs per target gene, plus positive and negative controls (Becker et al., Nucleic Acids Res. 2020). PDX cells were lentivirally transduced with the CRISPR/Cas9 sgRNA library, transplanted into NSG mice, grown in vivo and cells re-isolated at advanced AML disease. sgRNA distribution was measured by next generation sequencing and compared to input control using the MAGeCK pipeline.

Interesting dropout hits from PDX in vivo screens were validated by fluorochrome-guided competitive in vivo experiments in the PDX models, comparing growth of PDX AML cells with knockout of the gene of interest versus control knockout in the same mouse. PDX cells were transduced with lentiviral vectors expressing a single sgRNA, using in parallel three different sgRNAs per target gene. Targeting and control sgRNAs were marked by different fluorochromes; PDX cells expressing targeting or control sgRNA were mixed at a 1:1 ratio, injected into NSG mice and PDX models competitively grown until advanced disease stage, when cell distributions was determined by flow cytometry.

Human AML cell lines were studied in vitro for comparison.

Results:

In search for genes with essential function in AML, we cloned a small customized sgRNA library targeting 34 genes recurrently mutated in AML and tested the library in two PDX AML models in vivo. From the dropouts, we validated most interesting target genes using fluorochrome-guided competitive in vivo assays.

Knockout of NPM1 abrogated in vivo growth in all PDX AML models tested, reproducing the known common essential function of NPM1. KRAS proved an essential function in PDX AML models both with and without an oncogenic mutation in KRAS, although with a stronger effect upon KRAS mutation, suggesting that patients with tumors both with and without KRAS mutation might benefit from treatment inhibiting KRAS.

Surprising results were obtained for WT1 and DNMT3A. Both genes are frequently mutated in AML, but most AML cell lines tested in vitro do not show an essential function of any of the two genes, in published knockdown or knockout data, including from the Cancer Dependency Map database. On the contrary, knockout of either WT1 or DNMT3A was shown to enhance growth of AML cell lines and increase leukemogenesis in certain models.

In PDX models in vivo, we found a clearly essential function for DNMT3A in all AML samples and WT1 in most samples tested and PDX in vivo results were discordant to cell line in vitro data, suggesting that cell line inherent features and/or the in vivo environment influence the function of WT1 and DNMT3A.

Conclusion:

We conclude that functional genomics in PDX models in vivo allows discovering essentialities hidden for cell line in vitro approaches. WT1 and DNMT3A harbor the potential to represent attractive therapeutic targets in AML under in vivo conditions, warranting further evaluation.