Ex vivo drug testing is a promising approach to identify novel treatment strategies for acute myeloid leukemia (AML). However, accurate blast-specific drug responses cannot be measured with homogeneous “add-mix-measure” cell viability assays. In this study, we implemented a flow cytometry-based approach to simultaneously evaluate the ex vivo sensitivity of different cell populations in 34 primary AML samples to seven drugs and 27 rational drug combinations. Our data demonstrate that different cell populations present in AML samples have distinct sensitivity to targeted therapies. Particularly, blast cells of FAB M0/1 AML showed high sensitivity to venetoclax. In contrast, differentiated monocytic cells abundantly present in M4/5 subtypes showed resistance to Bcl-2 inhibition, whereas immature blasts in the same samples were sensitive, highlighting the importance of blast-specific readouts. Accordingly, in the total mononuclear cell fraction the highest BCL2/MCL1 gene expression ratio was observed in M0/1 and the lowest in M4/5 AML. Of the seven tested drugs, venetoclax had the highest blast-specific toxicity, and combining venetoclax with either MEK inhibitor trametinib or JAK inhibitor ruxolitinib effectively targeted all venetoclax-resistant blasts. In conclusion, we show that ex vivo efficacy of targeted agents and particularly Bcl-2 inhibitor venetoclax is influenced by the cell type, and accurate blast-specific drug responses can be assessed with a flow cytometry-based approach.
The treatment of AML with high-dose cytarabine and anthracycline-based intensive chemotherapy has remained the standard of care for the last four decades.1 Despite the increase in overall survival, only 35 to 40% of adult patients under 60 years are cured with chemotherapy and allogeneic stem cell transplantation.2 A number of novel targeted agents have been investigated in AML, but have usually generated clinical responses only in small patient subsets. Currently, genetic profiling is used for patient stratification and determination of treatment, evident by the recent approvals of midostaurin/gilteritinib and ivosidenib/enasidenib for treating AML patients with FLT3 or IDH1/IDH2 mutations, respectively.53 Furthermore, the Bcl-2 inhibitor venetoclax combined with a hypomethylating agent has recently been approved for AML with increased efficacy in patients with IDH1/2 and NPM1 mutations.76 However, the majority of AML patients lack actionable mutations and our understanding of the relationship between the cancer genotype, phenotype and drug function remains limited. Ex vivo drug testing with primary patient samples may help to identify novel treatment options and patient subgroups with sensitivity to a specific targeted therapy.
AML is diagnosed when the bone marrow (BM) contains at least 20% of myeloid lineage blast cells, and hematological relapse is defined when the BM exceeds 5% of blasts. The non-blast cells of the AML BM are comprised of other cell types, mainly lymphocytes and more mature leukemic cells (monocytes, granulocytes) or healthy cells. The BM content and the maturity level of leukemic cells is reflected in the French-American-British (FAB) subtypes.8 In FAB M0/1 subtypes, the differentiation blockade occurs at the early myeloid progenitor stage, whereas in FAB M4/5 subtypes the differentiation blockade is “leaky”. In addition to immature blasts in FAB M4/5 samples, leukemic cells often show myelomonocytic or monocytic differentiation, respectively. To achieve optimal response in patients, the drugs should target the less differentiated leukemic blasts.9 However, due to cellular heterogeneity, blast-specific drug responses are challenging to measure with conventional cell viability assays such as CellTiter-Glo (CTG) or tetrazolium reduction assays (MTT/MTS).10 Although enrichment of blasts is possible, this can be time consuming and enrichment might deplete cell populations such as monocytes that secrete cytokines important for blast cell survival and drug responses.1311
To evaluate the ex vivo sensitivity of AML patient samples at a cell population level, we applied a multiplexed, 96-well format flow cytometry (FC)-based drug sensitivity assay. We compared this approach with the CTG-based cell viability assay to study potential inconsistencies between these two methods. Furthermore, we aimed to identify drugs and drug combinations that could effectively target leukemic blasts in physiologically relevant concentrations. In addition to standard of care drugs, cytarabine and idarubicin, we selected five Food and Drug Administration approved targeted small molecule inhibitors that have shown AML-selective responses in our earlier studies:1514 MEK inhibitor (trametinib), JAK1/2 inhibitor (ruxolitinib), mTORC1 inhibitor (everolimus), FLT3/broad range tyrosine kinase inhibitor (TKI, sunitinib) and Bcl-2 inhibitor (venetoclax). Most importantly, we demonstrate that targeted agents, particularly venetoclax, have different efficacies towards AML cells at distinct stages of myeloid differentiation.
Methods are described in more detail in the Online Supplementary Material and Methods.
BM samples from 34 AML patients and three healthy volunteers were obtained from the Helsinki University Hospital Comprehensive Cancer Center after informed consent (permit numbers 239/13/03/00/2010, 303/13/03/01/2011, Helsinki University Hospital Ethics Committee) and in compliance with the Declaration of Helsinki. The patient characteristics are presented in the Online Supplementary Table S1.
Preparation of drug plates
The compounds (Online Supplementary Table S2) were dispensed on 96-well V-bottom plates (Thermo Fisher Scientific, Carlsbad, CA) and 384-well plates (Corning, Corning, NY, USA) using an acoustic liquid handling device Echo 550 (Labcyte, Sunnyvale, CA). Drug plate layouts and concentrations are presented in Online Supplementary Figure S1. BM mononuclear cells (BM-MNC) were isolated using Ficoll-Paque Premium (GE Healthcare, Little Chalfont, Buckinhamshire, UK) density gradient centrifugation. Fresh or frozen BM-MNC were suspended in mononuclear cell medium (MCM; PromoCell, Heidelberg, Germany) supplemented with 10 μg/mL gentamicin and 2.5 μg/mL amphotericin B and plated in parallel on pre-drugged 96-well plates (100,000 cells/well in 100 μl) for FC analysis and 384-well plates (10,000 cells/well in 25 μl) for CellTiter-Glo (CTG)-based cell viability assay. The cells were incubated with the drugs for 72 hours at 37°C and 5% CO2.
Following the 72-hour incubation with the drugs, cells were stained with an antibody mix (CD33, CD45, CD14, CD38 and CD34) followed by apoptosis (Annexin-V) and dead (7-AAD) cell staining. A detailed description of the methods is presented in Online Supplementary Material and Methods and the gating strategy is illustrated in Figure 1B.
Cell viability analysis using CellTiter-Glo®
Parallel to FC analysis, cell viability was measured with CellTiter-Glo (CTG; Promega, Madison, WI) in 384-well plates as described earlier.14 After the 72-hour incubation with the drugs, 25 μL CTG was added to each well. The luminescence signal was measured using a PHERAstar plate reader (BMG LABTECH, Ortenberg, Germany).
Calculation of the drug sensitivity (DSS) and drug combination scores
Ex vivo drug sensitivity of AML and healthy BM cells to the tested drugs was calculated using a DSS as previously described.16 Drug combination efficacies were calculated as the difference between observed and expected values. The expected value is computed using the Bliss independence model17 as reference, which assumes that two drugs exhibit their effect independently.18
Gene expression and pathway analysis
Publicly available microarray data from the Hemap data set2019 (http://hemap.uta.fi/) and RNA-seq data (RSEM values) from the TCGA Research Network21 (http://cancergenome.nih.gov/) also included in the Hemap resource were used for gene expression and pathway analysis. Beat AML data22 was used to assess the correlation between venetoclax drug sensitivity and BCL2 family and monocytic/granulocytic differentiation marker gene expression. For the analysis of gene expression in healthy hematopoietic cell types Differentiation Map data was used.23 Detailed methods are described in the Online Supplementary Material and Methods.
Statistical analysis was conducted with Graph Prism version 7.0 (GraphPad Software, San Diego, CA). Differences between drug responses were analyzed by Mann-Whitney U test, and for multiple t-tests P-values were adjusted using the Benjamin-Hochberg method (P<0.10 used to determine significance). The Kruskal-Wallis test was used when more than two groups were tested and significant comparisons were validated with post-hoc analysis (Dunn’s test). Statistical dependence between two variables was assessed by Spearman’s rank correlation.
Analysis of the AML bone marrow compartment
To measure blast-specific drug responses in mononuclear cell (MNC) enriched BM AML samples, we tested 34 AML samples collected at diagnosis or relapse with seven drugs. Following a 72-hour drug treatment we analyzed the samples by both FC and CTG-based cell viability assays (Figure 1A). With the CTG assay we measured the overall BM-MNC sensitivity, while with the FC analysis the number of viable cells in different cell populations was measured. We used four cell surface markers (CD45, CD34, CD33, CD14) to identify the major leukocyte populations present in the AML BM: leukemic blasts, immature granulocytes, promonocytes/monocytes and lymphocytes (Figure 1B). In the studied samples, the fraction of blasts out of CD45 positive leukocytes varied between 17-92% and the lymphocyte population ranged from 1-49% (Online Supplementary Table S3). As expected, we observed high numbers of monocytic cells in FAB M4/5 samples, whereas M0/1 samples mainly consisted of blasts and lymphocytes (Figure 1C). After 72-hours in culture, we observed monocytic maturation in several M5 samples,24 and in many samples the granulopoietic cell population diminished or was completely lost (Online Supplementary Figure S2).
FC versus homogeneous cell viability assay-based drug sensitivity profiling
In order to determine the correlation between drug sensitivity of the samples measured by FC or CTG-based methods, we converted the cell viability readouts from each assay to DSS (a drug sensitivity metric based on area under the dose-response curve, higher DSS indicates higher sensitivity).16 We observed a strong correlation between CTG and FC viability derived DSS when all live CD45 leukocytes were used as the FC readout (R=0.64, P<0.0001, Figure 2A), and when the blast-specific drug responses were exclusively taken as the FC readout from samples with blast counts over 50% (R=0.75, P<0.0001, Figure 2B). However, we observed poor correlation between the FC and CTG results in a sample cohort with blast counts below 50% (R=0.24, P=0.05, Figure 2C). The most prominent differences were seen in the response to trametinib and venetoclax (Online Supplementary Figure S3). The poor correlation was partly due to highly different drug sensitivities of the non-blast cell populations compared to blasts as demonstrated in two samples with low blast counts (Figure 2D-G). Our data shows that AML BM subpopulations have heterogeneous drug responses that confound the assessment of blast specific drug sensitivities when using homogenous cell viability assays in unsorted BM-MNC samples.
Ex vivo drug screening predicts induction therapy response
Next, we evaluated whether incomplete BM blast clearance at day 14 and day 28 after induction treatment was associated with decreased ex vivo drug sensitivity. We evaluated BM samples from 15 patients collected prior to anthracycline+cytarabine induction chemotherapy. Amongst these patients, five had >10% blast cells at day 14 and/or day 28 and were defined as chemoresistant as described in the Online Supplementary Table S1. Additionally, we included samples from two patients resistant to induction (collected at the time of resistant disease) in the chemoresistant group. A combined DSS of cytarabine and idarubicin showed significantly lower values for the resistant patients both with FC and (P<0.05, 2H) and CTG (P<0.01, Figure 2I). Furthermore, we observed a significant difference between responders and non-responders when blast-specific idarubicin response was measured with FC (P<0.05, Figure 2H) or total sample sensitivity was measured with CTG (P<0.05, Figure 2I). These results are in line with a recent study demonstrating that a similar FC-based platform can predict induction therapy response in a larger AML cohort.25
Blasts are highly sensitive to Bcl-2 inhibition whereas monocytes and granulocytes are resistant
Using the FC approach, we were able to evaluate blast-specific drug responses and compare them to other cell types within the same or between samples. Amongst the seven tested drugs, venetoclax (IC50=3.0nM) and idarubicin (IC50=28.7nM) showed the highest toxicity against blasts (Table 1). However, between these two drugs venetoclax showed the most selective efficacy against blasts when compared to other cell populations and healthy CD34 cells (Figure 3A, IC50 values in the Online Supplementary Figure S4). Moreover, venetoclax was also effective against CD34CD38 cells, which suggests activity against leukemic stem cells (Online Supplementary Figure 5). Compared to blasts, monocytic cells (CD14) were highly resistant to Bcl-2 inhibition (P<0.001, Mann-Whitney U test), but sensitive to MEK and JAK inhibition (P<0.001, Figure 3A). The phenomenon was clearly observed in samples from patients diagnosed with acute monocytic leukemia (M5) that contained substantial fractions of both cell types (Figure 3B-C).
Overall BM AML sample sensitivity to venetoclax is associated with FAB subtype
To follow-up on our findings, we hypothesized that AML samples with a high monocytic cell content should have a distinct drug response profile when overall BM-MNC sensitivity is measured with the CTG assay. We re-analyzed our earlier published CTG-based drug sensitivity data of 37 AML samples comprised of FAB M1, M2, M4 and M5 samples that were screened with 296 compounds.1514 Amongst the 296 compounds, venetoclax showed the largest drug sensitivity difference between M1 and M5 AML (P<0.001, Online Supplementary Table S4, Figure 4A). Similarly, the CTG-based sensitivity of the AML sample cohort studied here showed a gradual decrease in venetoclax sensitivity from M0 towards M5 subtype (Figure 4B). When we limited our FC analysis to diagnostic samples, a significant but smaller difference in blast-specific venetoclax sensitivity was also associated with FAB subtype (P<0.05, Figure 4C). This significance was not observed when we also included relapse and chemorefractory samples in the analysis (Figure 4D) largely due to a high number of chemorefractory M1/2 samples in our cohort that were more resistant to venetoclax (P<0.001, Figure 4D-E). Taken together, monocytic cells blur the high blast specific venetoclax effect in Ficollenriched M4/5 samples when measured with CTG but FAB subtype still has a significant effect on venetoclax response in blasts in our diagnosis AML sample cohort.
FAB subtype is associated with BCL2 and MCL1 gene expression
Anti-apoptotic Mcl-1 and Bcl-2 are considered the most important pro-survival factors in AML.2726 Furthermore, their expression and phosphorylation has been shown to be regulated through the Ras/Raf/MEK/ERK, PI3K/PTEN/AKT and JAK/STAT signal transduction pathways in different leukemias.3128 To study whether the expression of BCL2 family members and activity of signal transduction pathways is associated with FAB subtypes, we analyzed gene expression data of MNC of diagnosis AML samples using publicly available microarray and RNA-seq data. BCL2 was highly expressed in M0/1 AML and gradually decreased towards M5 samples and healthy monocytes (Figure 5A, Online Supplementary Figure S6). Notably, MCL1 showed an opposite trend in expression and was most highly expressed in healthy monocytes (Figure 5A). We also detected higher expression of BCL2A1, BCL2L11 (BIM), BID and JAK2 in M4/5 AML. A more detailed analysis of the healthy myeloid compartment revealed that BCL2 family expression is highly dependent on differentiation stage, which likely also influences the expression patterns seen between the different FAB subtypes (Figure 5B). Interestingly, high BCL2 and low MCL1 expression was also observed in FAB M3 AML and their healthy counterparts, colony forming unit (CFU) granulocytes (Figure 5A-B). High BCL2/MCL1 expression ratio in CFU granulocytes might explain the neutropenia seen in venetoclax treated patients.
Next, we investigated whether common cytogenetic abnormalities (RUNX1-RUNX1T1, CBFB-MYH11, MLL, PML-RARA) or mutations (FLT3, NPM1, RUNX1, CEBPA) explain some of the variations we observed in MCL1, BCL2 or BCL-xL gene expression within FAB subgroups (Online Supplementary Figure S7). AML samples with RUNX1T1-RUNX1T1 fusions showed significantly different gene expression exclusively in the M2 subgroup while samples with MLL or CBFB-MYH11 fusions showed significantly different gene expression exclusively in the M4 subgroup (Figure 5C, Online Supplementary Table S5). Particularly, M4 samples with MLL fusions had high BCL2 but low MCL1 expression levels compared to other M4 samples. To assess whether major signal transduction pathways are differentially active in FAB subtypes, we performed gene set enrichment analysis (GSEA). The analysis revealed significant enrichment of gene sets associated with inflammatory signaling and IL6/JAK/STAT pathway in M4/5 AML (Figure 5D-E, Online Supplementary Table S6).
To study whether ex vivo venetoclax response is associated with differentiation markers and BCL2 family expression, we analyzed the published Beat AML data set which includes data from 562 AML patients.22 Supporting our previous findings, samples that had high expression of monocytic/granulocytic cell markers (CD14, CD11b, CD86, CD68) were resistant to venetoclax (Figure 5F). High BCL2 expression was associated with venetoclax sensitivity whereas high MCL1 and BCL2A1 expression was associated with resistance (Figure 5F). These findings were also presented earlier by two different research groups.3332
Taken together, the gene expression data of mononuclear cell enriched AML samples indicate that M4/5 AML have low BCL2 but high MCL1 and BCL2A1 expression and increased inflammatory signaling. Thus, the data support the decreased venetoclax sensitivity we observed with the total mononuclear cell fraction of M4/5 samples.
MEK and JAK inhibitors sensitize venetoclax-resistant blast cells to venetoclax
Next, we studied whether mutations might explain the observed differences in blast specific venetoclax responses, but no significant correlation between genetic lesions and venetoclax response in our limited patient cohort was found (Online Supplementary Table S7). However, as demonstrated earlier, we detected decreased venetoclax sensitivity in chemorefractory and M5 samples (Online Supplementary Table S7, Figure 4A-E). When we divided the AML samples into two subgroups (sensitive DSS 21-43, IC50<20nM and resistant DSS 0-21, IC50>20nM) from the mid-point of the venetoclax response range, we noticed that resistant blasts were sensitive to either MEK and/or JAK inhibitors (Figure 6A). This finding suggests that venetoclax resistant blasts rely on either JAK/STAT and/or MAPK pathways. Furthermore, venetoclax sensitive blasts were enriched for NPM1 (8 of 25, 32% in sensitive and 1 of eight, 12.5% in resistant) and IDH1/IDH2 (10 of 25, 40% in sensitive and 1/8, 12.5% in resistant) mutations supporting the good clinical activity of venetoclax seen in this patient group (Figure 6A, Online Supplementary Table S7).
To assess the efficacy and clinical relevancy of 27 drug combinations against blasts, we used concentrations achieved in patients’ plasma during treatment. The results demonstrated prominent inter-patient variability with the most synergistic drug combinations when blast-specific drug responses were measured by FC (Online Supplementary Figure S8). Of the 27 tested drug combinations, venetoclax plus kinase inhibitors showed the highest average synergistic and blast killing effect (Table 2, higher BLISS score and lower mean % live blasts). Importantly, blasts were highly sensitive to single-agent venetoclax in 76% (25 of 33) of the samples with IC50<20nM. Thus, we did not observe synergy in the majority of the samples with a single venetoclax concentration of 50nM as this concentration alone was sufficient to kill the blasts (Figure 5A). To study the drug combination effect in more detail, we conducted additional drug testing of venetoclax with a more detailed concentration range on four AML samples. We observed that with lower venetoclax concentrations (10nM) a synergistic effect with MEK and/or JAK inhibitors was also detected in samples that were sensitive to single agent 50nM venetoclax treatment (Online Supplementary Figure S9).
Importantly, venetoclax (50nM) plus ruxolitinib (300nM) showed high efficacy (apoptosis/death>70%) and synergism in 6 of eight venetoclax resistant samples (Figure 5A-B). Strikingly, by combining venetoclax (50nM) with trametinib (25nM), all venetoclax resistant blasts were effectively targeted (Figure 6A-B). Although the combinations showed substantial toxicity to healthy CD34 cells, they targeted most effectively leukemic blasts (Figure 6B). As a comparison, a drug combination used during induction treatment (cytarabine+idarubicin) showed remarkable inter-patient differences in blast toxicity and it was also toxic to healthy CD34 cells (Figure 6B). Furthermore, the broad-spectrum tyrosine kinase and FLT3 inhibitor sunitinib (100nM) or mTOR inhibitor everolimus (10nM) were not as effective when combined with venetoclax (Figure 6A, Table 2). Our data demonstrate that by simultaneously inhibiting JAK and/or MEK signaling and Bcl-2, blast cells involving chemorefractory AML cells, can be effectively targeted ex vivo in physiologically relevant concentrations.
With FC-based drug testing we were able to simultaneously measure drug sensitivities of different cell populations in primary AML BM samples. Monocytic cells abundantly present in FAB M4/5 AML were markedly resistant to the Bcl-2 inhibitor venetoclax, while less differentiated blast cells in the same M4/5 samples or in M0/1/2 samples were sensitive. Accordingly, the overall BM-MNC sensitivity to venetoclax was strongly influenced by FAB subtype. Our study shows that FC-based, phenotypic drug testing can improve the current understanding of ex vivo drug effects and may help to identify blast-specific treatments for AML patients.
Along with our previous studies, several other groups have evaluated ex vivo drug responses of Ficoll-enriched AML mononuclear cells using high-throughput CTG or MTS based cell viability assays.363414 While these assays provide fast and robust readouts they fail to accurately measure blast specific drug responses. By using more accurate microscopy based screening, Snijder et al. have recently demonstrated that blast specific or relative blast fraction-based readouts increase predictive accuracy to treatment outcome.37 Similarly, Martinéz-Cuadrón et al. showed that a FC-based platform measuring blast specific effect in whole BM without MNC enrichment, predicted clinical response to induction therapy.25 We also showed earlier that in chronic myeloid leukemia, CD34-depleted cells (mature granulopoietic cells) were insensitive to BCR-ABL-1 inhibitors ex vivo whereas CD34 progenitor cells showed good sensitivity.38 In accordance, we demonstrate here with a FC-based approach that blasts differ in their drug sensitivities in comparison to other cell populations in the same AML samples. The highest blast-specific efficacy was observed with venetoclax, whereas ruxolitinib and trametinib showed increased activity towards monocytic cells. Importantly, we demonstrate that in samples with a low blast count, the overall mononuclear cell fraction sensitivity does not correlate well with the blast-specific drug sensitivity.
Consistent with our results, earlier studies have shown that primary AML samples are sensitive to venetoclax ex vivo.403915 Most of the studies have used mononuclear cell fractions to assess cell viability and to measure protein and gene expression levels. We observed that mononuclear cells of M0/1 samples that mainly consisted of blasts, were sensitive to venetoclax compared to mononuclear cells of M4/5 samples when using a homogenous CTG-based cell viability assay. Earlier, high ex vivo sensitivity to Bcl-2 inhibition has been associated with M3 AML in a study by Niu et al., whereas Pan et al. found no associations with FAB subtypes.4039 Importantly, both study cohorts lacked comprehensive spectra of different subtypes, with none or only one M0/1 AML case. To support our observation, mononuclear cells of M0/1 samples had a high BCL2/MCL1 gene expression ratio whereas M4/5 samples had a low ratio. Increased Bcl-2 protein expression has also been reported in M0/1 AML,41 and increased Mcl-1 expression in M4/5 AML26 of which the latter has been linked to elevated Mcl-1 expression in differentiating monocytes.42 Accordingly, we observed high MCL1 and BCL2A1 but low BCL2 expression in healthy monocytic and granulocytic cell populations.
By using a FC-based approach, we observed that several M5 samples contained venetoclax-sensitive blasts and a resistant monocytic cell fraction. This observation raises the question whether drug sensitivity profiling and gene/protein expression studies should focus on the immature blast cells and not the total MNC fraction especially in in M4/5 samples and samples with low blast count. When we compared the FC measured blast-specific venetoclax response between FAB subtypes, we observed a smaller but still significant difference between diagnosis M1 versus M5 subgroups. In clinical trials, NPM1, IDH1/2 and RUNX1 mutations have shown to be promising biomarkers for venetoclax+HMA treatment.437 Based on a study analyzing genotype and FAB subtype-specific patterns of 4,373 adult de novo AML cases,44 both IDH1/2 and RUNX1 mutations are enriched in M0/1/2 AML whereas NPM1 mutations are common in FAB M1/2/4/5 subtypes. Therefore, patient cohorts with mutated IDH1/2 or RUNX1 may be skewed to contain larger numbers of FAB M0/1/2 samples. To identify responders, it might be useful to evaluate the combined genetic and cell phenotype/FAB subtype information in a clinical setting.
With the FC method we also looked for effective combinations, since an overall response rate of only 19% was observed with venetoclax monotherapy in patients with high-risk relapsed/refractory (R/R) AML.6 In our study, all venetoclax-resistant blasts showed sensitivity to MEK and/or JAK inhibitors suggesting that JAK/STAT and MAPK pathways play a major role in venetoclax resistance. We showed earlier that stromal cell secreted cytokines such as GM-CSF mediate resistance to venetoclax, which can be counteracted by JAK inhibition.45 Moreover, the MAPK pathway plays a critical role in resistance through the proposed upregulation of MCL1.28 Both of these studies also demonstrated remarkable antileukemic activity in murine xenograft models when inhibiting JAK or MEK kinases together with Bcl-2. In agreement with the good synergism between ruxolitinib or trametinib with venetoclax observed here and in a recent study by the Beat AML study group,46 Kurtz et al. additionally showed that several different kinase inhibitors exhibited good synergism with venetoclax in AML samples.47 However, a recent clinical study with MEK inhibitor cobimetinib and venetoclax in R/R AML was closed due to limited clinical activity demonstrating that ex vivo drug screening results might not directly translate into a clinical setting.48
Inflammatory pathways are more active in M4/5 AML based on GSEA, consistent with the observed high sensitivity of monocytic cells to ruxolitinib and trametinib. Earlier studies have demonstrated that leukemic cells of patients with M4/5 AML produce IL1/IL613 and have a higher proliferative activity in cytokine-free medium.49 Thus, secreted cytokines and culturing conditions may have a big impact on the drug sensitivity profiles. While further investigation is warranted, results suggest that the JAK/STAT and MEK pathways are more active in differentiated monocytic cells as well as in venetoclax resistant blasts.
In summary, we show that ex vivo sensitivity of AML patient samples to venetoclax is associated with cell composition. Furthermore, we demonstrate that FC-based drug screening could be implemented to identify effective targeted drugs and drug combinations against immature blasts, accelerating drug discovery and individualizing therapy for AML patients.
We thank the patients and donors who participated in the study. We appreciate critical comments and input from colleagues at FIMM (Dimitrios Tsallos, Jarno Kivioja, Riku Turkki, Komal Javarappa, Joseph Saad, Samuli Eldfors, Muntasir Mamun Majumder, Aleksandr Ianevski and Krister Wennerberg). We would like to thank the FIMM High Throughput Biomedicine Unit for preparing the drug plates (Laura Turunen, Jani Saarela), and the FIMM Sequencing Lab for preparing the exome sequence data. We also thank Alun Parsons, Minna Suvela and Siv Knaappila for their help in patient sample processing, and clinicians Riikka Räty, Eeva Martelin, Tuija Lundán, Sanna Siitonen, Minna Lehto, Juha Lievonen and Sari Kytölä for providing the samples.
- This study was supported by grants from the European Research Council (M-IMM), Academy of Finland, Finnish Cancer Organizations, Finnish Cancer Institute, Emil Aaltonen Foundation, Sigrid Juselius Foundation, Orion Research Foundation, Instrumentarium Science Foundation, State Funding for University-Level Health Research in Finland, Relander Foundation, Gyllenberg Foundation, Ida Montin Foundation, Finnish Hematology Association and TEKES - the Finnish Funding Agency for Technology and Innovation.
- Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/105/3/708
- Received December 17, 2018.
- Accepted July 8, 2019.
- Tauro S. The blind men and the AML elephant: can we feel the progress?. Blood Cancer J. 2016; 6(5):e424. Google Scholar
- Döhner H, Weisdorf DJ, Bloomfield CD. Acute meloid leukemia. N Engl J Med. 2015; 373(12):1136-1152. PubMedhttps://doi.org/10.1056/NEJMra1406184Google Scholar
- Stone RM, Mandrekar SJ, Sanford BL. Midostaurin plus chemotherapy for acute myeloid leukemia with a FLT3 mutation. N Engl J Med. 2017; 377(5):454-464. PubMedhttps://doi.org/10.1056/nejmoa1614359Google Scholar
- DiNardo CD, Stein EM, de Botton S. Durable Remissions with ivosidenib in IDH1-mutated relapsed or refractory AML. N Engl J Med. 2018; 378(25):2386-2398. PubMedhttps://doi.org/10.1056/NEJMoa1716984Google Scholar
- Stein EM, DiNardo CD, Pollyea DA. Enasidenib in mutantIDH2relapsed or refractory acute myeloid leukemia. Blood. 2017; 130(6):722-731. PubMedhttps://doi.org/10.1182/blood-2017-04-779405Google Scholar
- Konopleva M, Pollyea DA, Potluri J. Efficacy and biological correlates of response in a Phase II Study of venetoclax monotherapy in patients with acute myelogenous leukemia. Cancer Discov. 2016; 6(10):1106-1117. PubMedhttps://doi.org/10.1158/2159-8290.CD-16-0313Google Scholar
- DiNardo CD, Pratz K, Pullarkat V. Venetoclax combined with decitabine or azacitidine in treatment-naive, elderly patients with acute myeloid leukemia. Blood. 2019; 133(1):7-17. PubMedhttps://doi.org/10.1182/blood-2018-08-868752Google Scholar
- Bennett JM, Catovsky D, Daniel MT. Proposals for the classification of the acute leukaemias. French-American-British (FAB) co-operative group. Br J Haematol. 1976; 33(4):451-458. PubMedhttps://doi.org/10.1111/j.1365-2141.1976.tb03563.xGoogle Scholar
- Bonnet D, Dick JE. Human acute myeloid leukemia is organized as a hierarchy that originates from a primitive hematopoietic cell. Nat Med. 1997; 3(7):730-737. PubMedhttps://doi.org/10.1038/nm0797-730Google Scholar
- Riss TL, Moravec RA, Niles AL. Cell viability assays. Eli Lilly & Company and the National Center for Advancing Translational Sciences.Google Scholar
- Panoskaltsis N, Reid CDL, Knight SC. Quantification and cytokine production of circulating lymphoid and myeloid cells in acute myelogenous leukaemia1. Leukemia. 2003; 17(4):716-730. PubMedhttps://doi.org/10.1038/sj.leu.2402835Google Scholar
- Carey A, Edwards DK, Eide CA. Identification of interleukin-1 by functional screening as a key mediator of cellular expansion and disease progression in acute myeloid leukemia. Cell Rep. 2017; 18(13):3204-3218. PubMedhttps://doi.org/10.1016/j.celrep.2017.03.018Google Scholar
- van der Schoot C, Jansen P, Poorter M. Interleukin-6 and interleukin-1 production in acute leukemia with monocytoid differentiation. Blood. 1989; 74(6):2081-2087. PubMedGoogle Scholar
- Pemovska T, Kontro M, Yadav B. Individualized systems medicine strategy to tailor treatments for patients with chemorefractory acute myeloid leukemia. Cancer Discov. 2013; 3(12):1416-1429. PubMedhttps://doi.org/10.1158/2159-8290.CD-13-0350Google Scholar
- Kontro M, Kumar A, Majumder MM. HOX gene expression predicts response to BCL-2 inhibition in acute myeloid leukemia. Leukemia. 2017; 31(2):301-309. Google Scholar
- Yadav B, Pemovska T, Szwajda A. Quantitative scoring of differential drug sensitivity for individually optimized anticancer therapies. Sci Rep. 2014; 4:5193. PubMedhttps://doi.org/10.1038/srep05193Google Scholar
- BLISS CI. The toxicity of poisons applied jointly11. Ann Appl Biol. 1939; 26(3):585-615. https://doi.org/10.1111/j.1744-7348.1939.tb06990.xGoogle Scholar
- Zhao W, Sachsenmeier K, Zhang L, Sult E, Hollingsworth RE, Yang H. A new bliss independence model to analyze drug combination data. J Biomol Screen. 2014; 19(5):817-821. PubMedhttps://doi.org/10.1177/1087057114521867Google Scholar
- Pölönen P, Mehtonen J, Lin J. Hemap: An interactive online resource for characterizing molecular phenotypes across hematologic malignancies. Cancer Res. 2019; 79(10):2466-2479. PubMedhttps://doi.org/10.1158/0008-5472.CAN-18-2970Google Scholar
- Mehtonen J, Pölönen P, Häyrynen S. Data-driven characterization of molecular phenotypes across heterogeneous sample collections. Nucleic Acids Res.Google Scholar
- Genomic and epigenomic landscapes of adult de novo acute myeloid leukemia. N Engl J Med. 2013; 368(22):2059-2074. PubMedhttps://doi.org/10.1056/NEJMoa1301689Google Scholar
- Tyner JW, Tognon CE, Bottomly D. Functional genomic landscape of acute myeloid leukaemia. Nature. 2018; 562(7728):526-531. Google Scholar
- Novershtern N, Subramanian A, Lawton LN. Densely interconnected transcriptional circuits control cell states in human hematopoiesis. Cell. 2011; 144(2):296-309. PubMedhttps://doi.org/10.1016/j.cell.2011.01.004Google Scholar
- Salem M, Delwel R, Mahmoud LA, Clark S, Elbasousy EM, Löwenberg B. Maturation of human acute myeloid leukaemia in vitro: the response to five recombinant haematopoietic factors in a serum-free system. Br J Haematol. 1989; 71(3):363-370. PubMedGoogle Scholar
- Martínez-Cuadrón D, Gil C, Serrano J. A precision medicine test predicts clinical response after idarubicin and cytarabine induction therapy in AML patients. Leuk Res. 2019; 76:1-10. Google Scholar
- Teh T-C, Nguyen N-Y, Moujalled DM. Enhancing venetoclax activity in acute myeloid leukemia by co-targeting MCL1. Leukemia. 2018; 32(2):303-312. Google Scholar
- Glaser SP, Lee EF, Trounson E. Anti-apoptotic Mcl-1 is essential for the development and sustained growth of acute myeloid leukemia. Genes Dev. 2012; 26(2):120-125. PubMedhttps://doi.org/10.1101/gad.182980.111Google Scholar
- Konopleva M, Milella M, Ruvolo P. MEK inhibition enhances ABT-737-induced leukemia cell apoptosis via prevention of ERK-activated MCL-1 induction and modulation of MCL-1/BIM complex. Leukemia. 2012; 26(4):778-787. PubMedhttps://doi.org/10.1038/leu.2011.287Google Scholar
- Wang JM, Chao JR, Chen W, Kuo ML, Yen JJ, Yang-Yen HF. The antiapoptotic gene mcl-1 is up-regulated by the phosphatidylinositol 3-kinase/Akt signaling pathway through a transcription factor complex containing CREB. Mol Cell Biol. 1999; 19(9):6195-206. PubMedhttps://doi.org/10.1128/MCB.19.9.6195Google Scholar
- Shenoy AR, Kirschnek S, Häcker G. IL-15 regulates Bcl-2 family members Bim and Mcl-1 through JAK/STAT and PI3K/AKT pathways in T cells. Eur J Immunol. 2014; 44(8):2500-2507. PubMedhttps://doi.org/10.1002/eji.201344238Google Scholar
- Faderl S, Harris D, Van Q, Kantarjian HM, Talpaz M, Estrov Z. Granulocyte-macrophage colony-stimulating factor (GM-CSF) induces antiapoptotic and proapoptotic signals in acute myeloid leukemia. Blood. 2003; 102(2):630-637. PubMedhttps://doi.org/10.1182/blood-2002-06-1890Google Scholar
- Zhang H, Wilmot B, Bottomly D. Biomarkers predicting venetoclax sensitivity and strategies for venetoclax combination treatment. Blood. 2018; 132(Suppl 1):175. https://doi.org/10.1182/blood-2018-175Google Scholar
- White BS, Khan SA, Ammad-ud-din M. Comparative Analysis of independent ex vivo functional drug screens identifies predictive biomarkers of BCL-2 inhibitor response in AML. Blood. 2018; 132(Suppl 1):2763. https://doi.org/10.1182/blood-2018-99-111916Google Scholar
- Tyner JW, Yang WF, Bankhead A. Kinase Pathway Dependence in Primary Human Leukemias Determined by Rapid Inhibitor Screening. Cancer Res. 2013; 73(1):285-296. PubMedhttps://doi.org/10.1158/0008-5472.CAN-12-1906Google Scholar
- Dietrich S, Oleś M, Lu J. Drug-perturbation-based stratification of blood cancer. J Clin Invest. 2018; 128(1):427-445. Google Scholar
- Swords RT, Azzam D, Al-Ali H. Ex-vivo sensitivity profiling to guide clinical decision making in acute myeloid leukemia: A pilot study. Leuk Res. 2018; 64:34-41. Google Scholar
- Snijder B, Vladimer GI, Krall N. Image-based ex-vivo drug screening for patients with aggressive haematological malignancies: interim results from a single-arm, open-label, pilot study. Lancet Haematol. 2017; 4(12):e595-e606. Google Scholar
- Pietarinen PO, Eide CA, Ayuda-Durán P. Differentiation status of primary chronic myeloid leukemia cells affects sensitivity to BCR-ABL1 inhibitors. Oncotarget. 2017; 8(14):22606-22615. Google Scholar
- Niu X, Wang G, Wang Y. Acute myeloid leukemia cells harboring MLL fusion genes or with the acute promyelocytic leukemia phenotype are sensitive to the Bcl-2-selective inhibitor ABT-199. Leukemia. 2014; 28(7):1557-1560. PubMedhttps://doi.org/10.1038/leu.2014.72Google Scholar
- Pan R, Hogdal LJ, Benito JM. Selective BCL-2 inhibition by ABT-199 causes on-target cell death in acute myeloid leukemia. Cancer Discov. 2014; 4(3):362-375. PubMedhttps://doi.org/10.1158/2159-8290.CD-13-0609Google Scholar
- Bogenberger JM, Kornblau SM, Pierceall WE. BCL-2 family proteins as 5-Azacytidine-sensitizing targets and determinants of response in myeloid malignancies. Leukemia. 2014; 28(8):1657-1665. PubMedhttps://doi.org/10.1038/leu.2014.44Google Scholar
- Yang T, Kozopas KM, Craig RW. The intracellular distribution and pattern of expression of Mcl-1 overlap with, but are not identical to, those of Bcl-2. J Cell Biol. 1995; 128(6):1173-1184. PubMedhttps://doi.org/10.1083/jcb.128.6.1173Google Scholar
- DiNardo CD, Rausch CR, Benton C. Clinical experience with the BCL2-inhibitor venetoclax in combination therapy for relapsed and refractory acute myeloid leukemia and related myeloid malignancies. Am J Hematol. 2018; 93(3):401-407. Google Scholar
- Rose D, Haferlach T, Schnittger S, Perglerová K, Kern W, Haferlach C. Subtype-specific patterns of molecular mutations in acute myeloid leukemia. Leukemia. 2017; 31(1):11-17. Google Scholar
- Karjalainen R, Pemovska T, Popa M. JAK1/2 and BCL2 inhibitors synergize to counteract bone marrow stromal cell– induced protection of AML. Blood. 2017; 130(6):789-802. PubMedhttps://doi.org/10.1182/blood-2016-02-699363Google Scholar
- Kurtz SE, Eide CA, Kaempf A. Dual inhibition of JAK1/2 kinases and BCL2: a promising therapeutic strategy for acute myeloid leukemia. Leukemia. 2018; 32(9):2025-2028. Google Scholar
- Kurtz SE, Eide CA, Kaempf A. Molecularly targeted drug combinations demonstrate selective effectiveness for myeloid- and lymphoid-derived hematologic malignancies. Proc Natl Acad Sci U S A. 2017; 114(36):E7554-E7563. PubMedhttps://doi.org/10.1073/pnas.1703094114Google Scholar
- Daver N, Pollyea DA, Yee KWL. Preliminary results from a Phase Ib study evaluating BCL-2 inhibitor venetoclax in combination with MEK inhibitor cobimetinib or MDM2 inhibitor idasanutlin in patients with relapsed or refractory (R/R) AML. Blood. 2017; 130(Suppl 1):813. Google Scholar
- Lowenberg B, van Putten W, Touw IP, Delwel R, Santini V. Autonomous proliferation of leukemic cells in vitro as a eeterminant of Prognosis in adult acute myeloid leukemia. N Engl J Med. 1993; 328(9):614-619. PubMedhttps://doi.org/10.1056/NEJM199303043280904Google Scholar