Acute myeloid leukaemia (AML) is a heterogeneous group of diseases with distinct clinicopathologic, cytogenetic and genetic characteristics. The heterogeneity has made unified and regimental approach unsuccessful for most patients, and outcome with standard chemotherapy and allogeneic haematopoietic stem cell transplantation (HSCT) was unsatisfactory with an overall cure rate of 30-40%. We hypothesised that an optimised in vitro drug screening platform for primary AML samples might help to identify the best personalised therapeutic agents for AML patients from whom the samples were obtained at real time. Primary mononuclear cells isolated from peripheral blood (PB) or bone marrow (BM) of AML patients at different stages of disease were seeded onto 96-well plates and treated with a panel of 25 selected drugs at 1000-fold concentration range for 3 days. The drugs included 17 tyrosine kinase inhibitors: axitinib, crizotinib, pazopanib, erlotinib, gefitinib, lapatinib, vandetanib, vemurafenib, sorafenib, quizartinib, ponatinib, lestaurtinib, nilotinib, dasatinib, ruxolitinib, TG101209, tofacitinib; 2 differentiation agents: arsenic trioxide, all-trans retinoic acid; a protein translation inhibitor: homoharringtonine; a proteasome inhibitor: bortezomib; a chemotherapy: cytarabine; a histone deacetylase inhibitor: vorinostat; a DNA methyltransferase inhibitor: azacitidine; and an mTOR inhibitor: rapamycin. The blast percentage in each sample was above 50% as confirmed by film review of the cytospin preparation. The inhibitory effect of each drug was evaluated by a high throughput PrestoBlue® fluorometric assay that was a measure of viable cell number. The results were expressed with reference to the vehicle control (0.1% DMSO) for each sample.

There was significant cell death when primary AML cells were cultured in IMDM medium supplemented with 10% FBS (IMDM10) for 3 days (Annexin V+ cells: 53.8% ± 4.5% , n=33). To identify the optimal culture condition, multiple culture conditions were tested for each sample (n=53 samples). The 3-day post-culture survival of primary AML cells improved significantly in a 1:1 mix of IMDM10 and medium conditioned by a mixture of mouse fibroblast lines engineered to produce human G-CSF, SCF, IL-3 and Flt3L (50%CM) (47.1% ± 6.5% improvement in Annexin V/7AAD -/- cell count normalised to day 0 input compared to IMDM10, p<0.0001). Assay readout was highly reproducible between replicates of the same samples (r=0.9681). There was significant correlation in drug response between PB and BM myeloblasts (r=0.9356, p<0.0001, 16 pairs), supporting the feasibility of drug screen based on PB samples. The readout might be predictive of clinical drug response in patients, as exemplified by the superior in vitro response to sorafenib in FLT3-ITD+ than FLT3 wild type AML samples (p<0.01 at 1 and 10 µM Sorafenib) that corroborated with the clinical observation for these patients treated with sorafenib. Furthermore, sorafenib-naïve FLT3-ITD+ AML samples were more sensitive to sorafenib in vitro than the resistant samples collected from the same patients during subsequent leukaemia progression.

To identify personalised agents effective for individual patients, a total of 85 samples from 60 AML patients were screened in 50%CM. Axitinib and vemurafenib had minimal cytotoxic effect on all samples whereas bortezomib, ponatinib, lestaurtinib, homoharringtonine and vorinostat exhibited significant anti-leukaemia effect (>75%) to over 90% samples at concentrations around or below the peak plasma concentrations in respective pharmacokinetic studies. The remaining 18 drugs exhibited variable effects on different samples. Further experiments and analyses are underway to identify distinct combinations and correlate with in vivo efficacy in xenotransplantation mouse model.

This optimised and clinically relevant in vitro drug screening platform might provide the laboratory basis for the selection of personalised treatment regime for AML patients with a short turnover time.


No relevant conflicts of interest to declare.

Author notes


Asterisk with author names denotes non-ASH members.