Quantitative assessment of the sensitivity of dormant AML cells to the BAD mimetics ABT-199 and ABT-737.

Abstract Cells from patients with acute myeloid leukemia (AML) that remain dormant and protected by stromal cells may escape effects of chemotherapy. We modeled dormancy in vitro and investigated the ability of Bcl-2 inhibitors ABT-199 and ABT-737 to overcome chemoprotection of dormant cells. CD34-enriched primary AML cells with aberrant leukemia-associated phenotypes (LAPs) were cultured on stromal cells. The chemosensitivity of dormant (PKH26high), CD34+, LAP+ cells was ascertained by 5-colour flow cytometric counting after 12 d. The PKH26high, CD34+, LAP + subset retained clonogenic capacity. The dormant fraction was completely resistant to Ara-C (p = .007). However, ABT-199 and ABT-737 were able to reduce the dormant fraction by 84% and 80%, respectively, of their effects on proliferating counterparts. In conclusion, we have elaborated a system for quantifying chemosensitivity in LAP+ dormant leukemia cells, thought to contribute to disease relapse, and shown sensitivity of dormant LAP+ cells to ABT-199 and ABT-737 in this system.


Introduction
Relapse is common in patients with acute myeloid leukemia (AML) after remission-induction with nucleoside analogue/anthracycline-based chemotherapy. The cells which give rise to relapse are likely to have evaded the effects of treatment by lying dormant [1,2]. The concept of a dormant cancer cell can initially seem counter-intuitive, but in AML the presence of these cells has been substantiated by both functional and phenotypic evidence [1][2][3][4][5][6][7]. Dormant AML cells are concentrated adjacent to stromal cells [1,8]. Protection of AML cells from chemotherapeutic drugs by proximity to stromal cells has been shown for several compounds, including the standard chemotherapeutics daunorubicin and cytarabine [1,9,10] as well as FLT3 inhibitors [5,11]. Alvares and colleagues combined dormancy and niche in a single model and were able to demonstrate resistance to a FLT3 inhibitor in a dormant AML subpopulation cultured in vitro on stromal cells [5].
A potential novel approach to AML therapy is to target the B-cell lymphoma 2 (BCL2) protein family and thus directly induce apoptosis [12]. Anti-apoptotic members of the BCL2 family, i.e. BCL2, BCLX L , BCLW, BFL1 (A1), BCLB and MCL1, are considered as guardians of mitochondrial integrity by opposing pro-apoptotic BCL2 family proteins. In theory, oncogenic stresses may upregulate pro-apoptotic pathways, and therefore the successful cancer cell needs to have just enough functioning anti-apoptotic molecules to hold apoptosis in check. Cancer cells may therefore be dependent on ("addicted to") survival gene expression and/or be "primed" by pro-apoptotic BH3 only proteins to readily undergo apoptosis once the anti-apoptotic family members are inhibited [13][14][15]. ABT-737 and ABT-199 (a reverse engineering version of ABT-737's oral form ABT-263 that has high affinity for BCL-2 and lower affinity for BCLX L ), have exhibited promising anti-cancer activities in vitro and in vivo [16,17] including activity against AML cells [18][19][20][21][22]. Furthermore, these agents are found to significantly potentiate certain other chemotherapeutics [22][23][24][25][26][27][28][29], suggesting a promising future for novel drug combinations.
In the current study, we compare the sensitivities of proliferating and dormancy-enriched primary AML cells to ABT-199 and ABT-737. We show that aberrant phenotypes can persist through several days of culture on MS-5 stromal cells and we were able to use this finding in order to quantify dormant AML cells labeled with the proliferation marker PKH26 with and without the addition of ABT-199 and ABT-737.

Materials
Antibodies were from BD Biosciences (Cowley, UK) and other reagents were from Sigma-Aldrich (Poole, UK), unless otherwise stated. TG02, originally designated SB1317, was kindly provided by Tragara Pharmaceuticals (Carlsbad, CA). ABT-737 was from Sequoia Research products (Pangbourne, UK). ABT-199 was from Active BioChem (Kowloon, Hong-Kong). All three agents were reconstituted in dimethyl sulfoxide (DMSO) and stored at 10 mM.

AML patient samples
Fresh or cryopreserved bone marrow or peripheral blood samples of patients presenting with de novo or relapsed AML were collected after written consent at diagnosis at the Nottingham University Hospitals. Use of these samples was approved by the Nottingham Research Ethics Committee and the samples were anonymised. Mononuclear cells were isolated using a standard density gradient/centrifugation method with Histopaque and red blood cells depleted from the sample using ammonium chloride. Cryopreserved samples kept in liquid nitrogen were thawed and rested in RPMI-1640 medium (Sigma-Aldrich, Poole, UK) enriched with 20% fetal calf serum (FCS) (First Link, Birmingham, UK), for 60-90 min before experimental handling. Only samples with >85% post-rest viability, assessed by trypan blue exclusion, were used. Leukemia-associated phenotypes (LAPs) were assessed by published methods [30] prior to culture.

12-Day leukemic cultures on stroma
MS-5 stromal cells [31] were maintained at a density of 5 Â 10 4 /mL in a-minimum essential medium (a-MEM; Gibco, Paisley, UK), supplemented with 10% heat-inactivated FCS, 1% penicillin/streptomycin and 2 mmol/L L-glutamine. For 12-d co-culture studies, MS-5 cells were sub-cultured at 2 Â 10 4 /mL on a gelatincoated 12-well flat-bottom plate for 24 to 48 h to form a monolayer before addition of AML cells. CD34 þ cells from fresh or cryopreserved AML samples were enriched by magnetic bead separation using Mini-MACS CD34 isolation kit (MiltenyiBiotec, Bisley, UK) and then labeled with the permanent cell membrane dye PKH26 according to manufacturers' instructions. Labelled cells (at 2 Â 10 6 cells/mL) were added to the MS-5 feeder layer and co-cultured in Gartner medium [32]. Cells were kept in culture for 12 d and demipopulated on day 7. Drugs were added to cells on day 8 of the cell culture. On day 12, cells were harvested and processed for clonogenic assay and flow cytometric assessment as detailed below.

Methylcellulose colony formation (CFU) assay
Following 12-d culture, cells were labeled as above and sorted on a MoFlo TM XDP Cell Sorter (Beckman Coulter, Fullerton, CA). Sorted cells were washed, resuspended in StemSpan Expansion medium with 1% Stemspan CC100 cytokine cocktail for 3-4 d and transferred to Methocult H4034 colony medium (all from Stem Cell Technologies, Grenoble, France). Cells were plated onto 96-well flat-bottom plates at 2 Â 10 4 cells/ 100 mL in triplicates and were incubated at 37 C in 5% CO 2 for 14 d. Colonies of >50 cells were counted.

Statistics
Data were analyzed using SPSS 21 software package (Chicago, IL).

Results
Dormant primary leukemia cells retaining clonogenic capacity can be identified after 12 d of co-culture and are effectively targeted by ABT-737 and ABT-199 We established and validated a system for measuring the chemosensitivity of dormant primary leukemia cells (Figure 1). After labeling CD34 þ primary cells with PKH26 and culturing on MS-5 stromal cells for 12 d, we ascertained that the leukemic nature of the cells following culture could be assured through the selection of samples with a major LAP. The aberrant CD34/CD7 and CD34/CD19 status of cells identified in presentation samples was maintained in the dormant subset throughout the 12 days of culture (Table 1). We further ascertained the clonogenicity of sorted CD45 dim CD34 þ LAP þ PKH26 high (dormant) subpopulation following the 12-d culture. We found that this subpopulation retained almost all of the clonogenic capacity compared to the CD45 dim CD34 ± LAP þ PKH26 low cells that had already divided in vitro (  samples, a comparison of stromal support by MS-5 and by mesenchymal stem cells was made, with healthy human adult BM-MSCs cultured as described [33]. Clonogenicity of cells cultured by both methods was measured, but major differences were not observed ( Table 2). Drugs were added after 8 d to primary AML cells in the MS-5 co-culture system. The drug concentrations used were chosen on the basis that these were the average 72-h IC 40 values determined in preliminary assays in suspension culture (data not shown). Chemosensitivity was ascertained by 5-colour flow cytometric counting using a suspension of CytoCount TM fluorospheres as a reference population. Six samples were studied with this protocol as shown in Table 3. We have previously shown that the multikinase inhibitor TG02, which targets transcriptional cyclin-dependent kinases, has efficacy against dormant AML cells [7] and we used this agent as a positive control. 200 nM Ara-C reduced the cell number by 49% in the cycling fraction, but there was no cell reduction at all in the dormant fraction (p ¼ .007, Figure 2). In contrast, exposure to ABT-199, ABT-737 and TG02 markedly reduced the number of both dormant and cycling cells. Although there is some favoring of cycled cells with these agents (Table 3), we found that ABT-199 and ABT-737 were able to reduce the dormant cell fraction by 84% and 80%, respectively, of their effects in proliferating counterparts.

Discussion
The presence of dormant cells within a leukemic clone presents the chemoresistance researcher with a distinct challenge, as most anti-neoplastic drugs have been designed to target proliferating cells as a surrogate for tumor cells, and therefore highly chemoresistant dormant cancer cells do not fit into the mainstream chemotherapeutic paradigm.
We built on a model previously described by Alvares and colleagues for identifying dormant cells grown on MS-5 stromal cells [5]. Additionally, we used LAP markers to identify leukemic cells against a background of potentially normal hematopoietic cells. We chose 12 d to approximate the length of time an AML cell would need to remain dormant to resist a course of chemotherapy. Guan and colleagues had demonstrated that the majority of ex vivo dormant AML cells rapidly re-enter the cell cycle in vitro [4], and thus phenotypic labeling of freshly isolated or thawed cells with dormancy markers might greatly overestimate the proportion of cells that would remain dormant, whereas the 12-d culture allowed us to be more confident that we were studying a clinically relevant subset of these cells. Our evaluation of this model indicated that not only could the expression of some LAP markers persist in the 12-d culture, but also that some dormant LAP þ leukemia cells maintained their clonogenic potential, thus alleviating our concern that the in vitro conditions may have rendered the cells senescent. A second argument against the in vitro induction of senescence in our model is that senescent cells are characterized by increased size and granularity and therefore have increased low-angle and orthogonal light scatter (forward scatter and side scatter) on flow cytometric analysis [34], whereas the CD34þ/PKH high cells selected for analysis in our experiments retained low scatter propertiesthis is illustrated in the AML: acute myeloid leukemia. a Although MS-5 cells were used in all cases, in two cases (#603 and #596) we also used primary mesenchymal stem cells (MSC) so that the clonogenic yield from each support system could be compared. AML: acute myeloid leukemia; LAP; leukemia-associated phenotypes. Figure 1 gating. The variability in proliferative abilities between cells within a single sample has been appreciated for several decades [35,36], but the reasons why clonogenic cells can remain dormant in culture with stromal cells and then form colonies in semi-solid medium suggests a contribution from the stromal cells to the state of dormancy. The retention of clonogenic potential in PKH high, but not PKH low cells was very marked, and the use of LAP markers in the sorting process ensured that non-leukemic cells were not affecting the output. The efficacy of ABT-199 and ABT-737 in dormant cells cultured on MS-5 was at least 80% of their efficacy against cycled cells and the contrast to Ara-C, which completely failed to eradicate dormant primary cells when cultured on stromal cells, was pronounced. Primary samples differ enormously in their rates of cycling and therefore the proportion of cells likely to be vulnerable to Ara-C [37]. Morevoer, the osteoblastic niche has been shown to protect a dormant subset of AML cells from this drug [1]. The in vitro model system used thus enabled us to incorporate a protective stromal microenvironment, to discriminate between leukemic and residual normal cells based on LAPs and to contrast the effects of novel agents with the established drug Ara-C, which is selective for proliferating cells. The success of this model in assessing drug sensitivity in both cycling and dormant cell compartments, while unlikely to eliminate the need for in vivo models, has the potential to refine the number and scale of such studies.
Two of the six samples used in this study were from patients with complex cytogenetics. A further patient with normal cytogenetics was studied with refractory disease. A fourth patient had transformed chronic myelomonocytic leukemia and was unsuitable for intensive treatment. There were no good cytogenetic risk samples included. The high proportion of  CD7 þ aberrant phenotypes probably contributed to this bias. There have been at least 20 studies of aberrant CD7 expression in AML with very mixed results as to its prognostic value. However, the majority consensus appears to be that it is associated with normal or adverse cytogenetics and with multi-drug resistance [38,39]. This is therefore an interesting cohort to study with a view to identifying therapies from which hardto-treat patients might benefit. Evidence that BCL2 is therapeutically relevant in AML was established in our laboratory and by others nearly two decades ago [40,41], but has been refined recently with the emergence of first ABT-737 and then ABT-199 as research tools and as drugs. An additional important discovery is that cells with low levels of reactive oxygen species (ROS-low), functionally defined as leukemic stem cells, aberrantly over-express BCL-2 and are selectively targeted by ABT-737 [20]. ROS-low AML cells are enriched for dormant (Ki-67 negative) cells and therefore overlap the subpopulation we have investigated in the current work. An observation that AML cells with high BCL2 expression tend to have an impaired ability to enter the cell cycle has also been documented [42]. ABT-737 causes prolonged thrombocytopenia associated with BCLX L targeting, but ABT-199, which is more specific for Bcl-2, is of huge clinical interest: one report has been published [43] and 12 clinical studies are currently (September 2017) listed for ABT-199 (also known as venetoclax) in AML by clinicaltrials.gov. Our data suggest that this agent may have efficacy against cells protected by niche and dormancy. Analysis of post-treatment biopsies would be of great help to establish the cell cycle status of residual cells and thus whether venetoclax can eradicate dormant leukemic cells in the clinic.
In summary, we have established and validated a system for the direct comparison of the chemosensitivity of intra-sample cycling and dormant leukemic cells. Using this methodology we have shown a differential response to Ara-C in cycling versus dormant leukemia cells; however, ABT-199 and ABT-737 demonstrated effective targeting of both cellular compartments emphasizing the potential importance of these agents in the clinical eradication of all leukemic cells.