Establishment and Characterization of a Panel of Patient-Derived Acute Leukemia Mouse Models As Emerging Platform for Translational Research

Many efforts are made to treat acute leukemia effectively, but still; new treatment options are urgently needed. In the past, mainly chemically induced and cell line-derived mouse models were used to evaluate the efficacy of new drugs. The high failure rate of 96% in early clinical development illustrates the urgent need for more predictive preclinical models as a major prerequisite for rapid bench-to-bedside translation of investigational anticancer therapies. Transplantable patient-derived xenograft (PDX) models of leukemia offer a strong preclinical tool for drug screening and biomarker development as they represent the complex clinical tumor heterogeneity and molecular diversity of human cancer.

Between 2011 and 2015 we collected 91 samples of peripheral blood (PB) or bone marrow (BM) from patients diagnosed with acute myeloid or lymphoblastic leukemia (AML, ALL) and injected them intratibially (i.t.) into NOG (NOD/Shi-scid/IL-2Rγ^null) mice (n=1-5/patient sample). Infiltration of human leukemic cells was determined by flow cytometry in murine PB, BM and spleen. If more than 5% of hCD45+, hCD33+, hCD34+, hCD38+, and/or HLA-ABC+ cells were detectable in one sample, it was classified as engrafted. Results were confirmed by histo-pathological examination. FISH analyses confirmed the cytogenetical concordance with the donor patient where available. 8 models were treated with cytarabine and results were compared to patient´s outcome and treatment experiments using cell line derived models of AML. Whole exome sequencing analyses of the transplantable models are underway for a deeper characterization of the respective leukemic clone which adapted to the murine microenvironment.

Specimens of 44 female and 47 male patients (median age: 59 years, median BM infiltration: 39%) were collected. The donor patient cohort covered a broad range of different molecular subtypes: amongst others 5 MLL-rearrangements, 1 MLL-Deletion, 2 t(8;21) translocations, 20 cytogenetically normal and 7 complex karyotypes were included. Up to now 12 PDX models were in passage (P) 5 and higher, where P1 represents first implantation of human cells into the murine BM. 15 PDX models grew in P3 - P4, 21 in P2, 36 in P1 and 7 samples showed no engraftment. Engraftment capacity of the leukemic cells did not correlate significantly with any of the patient characteristics. BM engraftment ranged from 30% - 80%. Spleen and PB depicted 5 - 30% of leukemic cells. Infiltration rate in different organs and immuno-phenotypic characteristics of the human cells were specific for a defined model and preserved during serial transplantation. Take rates within one mouse cohort in serial transplantation were ≥98% for all transplantable PDX, similar to cell line derived models thereby qualifying the PDX approach as a suitable preclinical platform. Overall survival (OS) in P1 ranged from 52 to > 310 days (d). Models which could be serially transplanted showed model specific median OS ranging from 32 - 150 d. Comparable cell line derived models depicted median OS of 13 - 45 d. Of note, cell line derived models (KG-1, NOMO-1, MOLM-13, THP-1, MV4-11 or HL60, n ≥ 10 mice/cell line) induced hind limb paresis in all recipient mice. These symptoms could not be observed in PDX and most important are not part of the clinical picture. Cytarabine induced a significantly prolonged OS in 8/8 tested PDX models. Respective donor patients showed hematologic response under cytarabine based therapy highlighting the excellent predictivity of the in vivo platform. In contrast none of the investigated cell line derived models showed sensitivity towards cytarabine, although representing similar subtypes of AML as the investigated PDX models.

Taken together a constantly expanding panel of well characterized AML/ALL PDX covering a broad range of different subtypes of the disease is available for drug development and tumorbiology studies. The comparison with cell line derived in vivo models revealed significant advantages of the PDX approach as the latter represents the molecular diversity more in detail, mimics the clinical signs of leukemia more realistic and most important mirrors sensitivity towards standard of care in a direct comparison with the donor patient´s clinical outcome. Therefore, we strongly believe that the AML/ALL PDX platform is a robust and predictive tool to address translational challenges in oncology research.