Mitochondrial BAX Determines the Predisposition to Apoptosis in Human AML

Acute myeloid leukemia (AML) is a disease characterized by clinical and biological heterogeneity. Recently, genomics have facilitated insights into the genetic landscape of AML and identified novel mutations with functional implications (1–3). Genetic alterations have been described to be of major prognostic relevance (4–6). Genetic diversity, however, is not the sole determinant of relapse, drug resistance, and aggressiveness of leukemia biology (7). Large-scale genetic analysis approaches naturally lack insights on posttranscriptional protein expression, protein regulation, and localization, that is, the actual level of activation. Cell-to-cell variability in protein levels and in some cases subcellular localization induce large differences in the responsiveness to uniform stimuli between individual cells and descendants from different clones (8–10). Within a clonal cell population, some cells rapidly induce programmed cell death in response to death receptor–induced apoptosis, while other cells appear more resistant and survive (11). This “intratumoral” heterogeneity and the underlying differences in cell signaling critically influence cellular predisposition to apoptosis upon cytotoxic stress (e.g., induced by chemotherapy). Differential sensitivity of malignant cells to chemotoxic stress has become a central challenge in oncology. Therefore, the identification of prognostic factors and the analysis of individual cellular predisposition to apoptosis is a major focus of translational cancer research. Investigation of the apoptotic machinery may offer novel strategies to sensitize resistant cells to cytotoxic agents.

Analysis of signaling between prosurvival BCL-2 proteins, the proapoptotic BCL-2 proteins BAX and BAK, and BH3-only proteins (sharing only the BH3 domain with BCL-2) on the outer mitochondrial membrane has been proposed to predict cellular predisposition to chemotoxic cell death. Cellular cytotoxic response is suggested to depend on the presence of prosurvival BCL-2 proteins and the accumulated BH3-only proteins on the mitochondria, termed “mitochondrial priming,” resulting perhaps from prior sublethal cell stresses (12–15). Prosurvival BCL-2 proteins inhibit BAX and BAK via direct interactions or by sequestering “activator” BH3-only proteins, thereby preventing their interaction with BAX and BAK (16–21). In response to stress, BH3-only protein signaling is thought to activate BAX and BAK, which in turn permeabilize the outer mitochondrial membrane and release of cytochrome c (cyt c). Cytosolic cyt c results in mitochondrial dysfunction and initiates the caspase cascade that efficiently dismantles the cell (22, 23). The activation of BAX or BAK commits the cell to apoptosis necessitating a tight control of these proapoptotic BCL-2 proteins (24). In healthy cells, BAX and BAK are controlled by constant retrotranslocation of mitochondria-associated protein into the cytosol (25, 26). Permanent translocation of BAX and BAK to the mitochondria establishes an equilibrium between cytosolic and mitochondrial protein pools (25, 27). These processes are dependent on the porin VDAC2, which serves as a platform for BAX retrotranslocation (28). Differential rates of BAX and BAK shuttling determine localizations of BAX (predominantly in the cytosol) and BAK (largely on the mitochondria; ref. 26). Retrotranslocation of BAX and BAK depends on interactions with prosurvival BCL-2 proteins (25, 26). Importantly, BAX retrotranslocation rates determine the size of the mitochondrial BAX pool and cellular response to apoptosis stimulation (29). Moreover, BH3-only proteins reduce the rate of BAX retrotranslocation (25), suggesting a reflection of the BCL-2 protein signaling on the mitochondria (and “mitochondrial priming”) and perhaps other factors in the equilibrium between cytosolic and mitochondrial BAX.

Here, we provide first evidence for the correlation of the subcellular BAX (and BAK) localization with susceptibility to cytotoxic therapy in human AML cell lines and primary AML blasts. While BAX shift toward the mitochondria correlates with increased cellular sensitivity to chemotoxic stress and beneficial prognostic markers, BAX shift toward the cytoplasm rather correlates with cellular resistance to chemotherapeutic stress and adverse prognostic markers. Thus, relative BAX localization has the potential to combine all relevant factors to assess the cellular susceptibility to apoptosis induction as one of the “final effectors” of cell death.

HeLa cells were cultured in DMEM medium supplemented with 10 mmol/L HEPES and 10% heat-inactivated FBS in 5% CO₂ at 37°C. All AML cell lines and primary AML cells were cultured in RPMI1640 medium supplemented with 10 mmol/L Hepes, 10% heat-inactivated FBS, 4.5 g/L glucose, and 1 mmol/L sodium pyruvate in 5% CO₂ at 37°C. Cells were regularly (in 4-week intervals) tested for potential mycoplasma infection using the Venor GeM kit (Biochrom).

For the standard curve whole-cell lysates, HeLa cell lysate was used to establish a standardized mix of all analyzed proteins, ensuring similar standardization of all patient samples. HeLa cells were harvested, washed with PBS, and subsequently resuspended in cell lysis buffer (20 mmol/L Tris, 100 mmol/L NaCl, 1 mmol/L EDTA, 0.5% Triton X-100, pH 7.5) supplemented with protease inhibitors. Upon incubation on ice for 15 minutes, the samples were centrifuged at 15,000 × g for 10 minutes at 4°C. The supernatants were subjected to acetone precipitation, followed by resuspension in SDS sample buffer and storage at −80°C. To obtain mitochondrial and cytosolic fractions of patient bone marrow samples, mononuclear cells were isolated by Ficoll-Paque density gradient centrifugation. Cells were washed with PBS, resuspended in RPMI supplemented with 10% DMSO and 20% FCS, and stored in liquid nitrogen. Aspirates were then thawed and centrifuged for 5 minutes at 1,500 × g at 4°C. The mean percentage of blasts in the bone marrow aspirates (n = 44) was 64% (95% CI: 54%–74%). Upon washing with PBS, the cell pellet was resuspended in SEM buffer (10 mmol/L HEPES, 250 mmol/L sucrose, pH 7.2) supplemented with protease inhibitors and homogenized using the MINILYS (PeqLab) system. Subsequently, samples were centrifuged at 1,500 × g for 5 minutes at 4°C. The supernatant was transferred to a new tube and centrifuged for 30 minutes at 13,000 × g at 4°C. While sedimented mitochondria were washed two times, the supernatant of this step, the cytosolic fraction, was ultracentrifuged at 150,000 × g for 1 hour at 4°C. Finally, the samples were separated by SDS-PAGE and subjected to Western blot analysis.

Band intensities were captured using the Fujifilm LAS-4000 imager and quantified using ImageJ. On the basis of a whole-cell lysate standard curve within each individual blot, BAX (E63, Abcam) and BAK (Millipore) were then quantified using SigmaPlot. The amounts of mitochondrial and cytosolic protein were then determined as relative to COX IV (Invitrogen) and β-actin (Millipore), respectively. Finally, ratios were built from the mitochondrial and cytosolic values to obtain the relative protein localization. Accumulation of the cleaved form of the caspase substrate PARP was analyzed by Western blot (pAB, Cell Signaling Technology) to determine apoptosis progression.

Forty-eight bone marrow samples of the test cohort were obtained from the Hematology Tumor Bank Magdeburg (HTM) at the Department of Hematology and Oncology, Medical Center, Otto-von-Guericke University (Magdeburg, Germany). These samples were obtained after written informed consent of patients and in accordance with the Declaration of Helsinki. This study was approved by the institutional review board of the University Hospital Magdeburg and the local ethics committee (file #15/2008). The majority of patients were diagnosed with AML (n = 45; 93.8%). Other myeloid diagnoses included blast phase of chronic myeloid leukemia (n = 1), blastic plasmacytoid dendritic cell neoplasm (BPDCN; n = 1), and one patient was diagnosed with acute biphenotypic leukemia. Three of all samples (6.3%) were from relapsed patients. Analyzing the AML samples (n = 45) only, three were taken in relapse (6.7%), and 74.4% (32/43) were de novo AML. Of these AML samples, 20.9% (9/43) were FLT3-ITD (FMS-like tyrosine kinase 3 internal tandem duplication) positive, and 18.6% (8/43) were at good, 51.2% (22/43) were at intermediate, and 30.2% (13/43) were at poor risk according ELN criteria (6), respectively.

DNA from 40 available samples (low n = 8; medL n = 9; medH n = 11; high n = 12) were isolated and genotyped using the Illumina OmniExpressExome 1.2 Chip. The data were analyzed using Illumina Genome Studio 2011.1 Genotyping Module to obtain normalized genotype data. CNV analysis was carried out using the CNVpartition 3.2.0 algorithm within Genome Studio using default parameters. Somatic copy-number alterations were determined using GISTIC 2.0. Identification of differential single nucleotide variants (SNV) was performed using PLINK (P < 0.001).

For the main statistical analysis, Sigma Plot was used. The logistic regression and ROC calculation was done using SAS Software 9.4. A significance level of α = 5% was used for testing; however, all P values have explorative character only. For the ROS, curve area under curve and 95% confidence intervals were calculated using the Mann–Whitney method.

For caspase assays, cells were washed with ice-cold 1× PBS and resuspended in ice-cold Cell Lysis Buffer (20 mmol/L TRIS pH 7.4, 100 mmol/L NaCl, 1 mmol/L EDTA, including protease inhibitor cocktail and 0.5 % Triton X-100). Whole-cell lysate was incubated with Caspase-3/7 substrate (BD Pharmingen) for 60 minutes at 37°C, and protein concentration was determined by a Bradford Assay (Roth). Substrate cleavage was measured for 50 cycles with 10-second delay (excitation at 380 nm, emission at 430–460 nm). Kinetics were determined and calculated to the amount of protein per sample.

Cells were seeded in 96-well plate format with mock-treated samples and lysis control on the same plate. Cells were treated with 1 or 20 μmol/L daunorubicin, 1 μmol/L actinomycin D, 100 μmol/L etoposide, 5 μmol/L doxorubicin, 10 μmol/L ABT-737, 10 μmol/L UMI-77, or 1 μmol/L staurosporin for 24 hours. Then, 50 μL supernatant was transferred to fresh a 96-well plate and mixed with substrate (Promega). Lactate dehydrogenase (LDH) activity was monitored over 25 cycles with a 25-second delay, and kinetics were calculated.

The regulation of BAX and BAK by steady translocation to the mitochondria and retrotranslocation back into the cytosol of nonapoptotic cells raises the question whether BAX/BAK localization could predict cellular predisposition to apoptosis. Therefore, the relative distribution of BAX and BAK between cytosolic and mitochondrial pools in primary bone marrow samples of human AML was analyzed in a test cohort by fractionating 48 patient samples into cytosol and mitochondria (Fig. 1A). To allow comparison of samples not analyzed on the same Western blot, relative mitochondrial and relative cytosolic protein was determined using the fractionation loading controls COX IV and β-actin, respectively. All detected proteins were normalized using a titration of the same HeLa cell extract and thus the same amounts of all detected proteins for the evaluation of every analyzed sample. Both relative protein pools were then combined to calculate the relative protein localization (relative BAX localization = mitochondrial BAX/cytosolic BAX), with high relative protein localization values pointing to increased mitochondrial protein and low values to predominantly cytosolic protein independent of cellular protein levels (Fig. 1B). Surprisingly, the analysis of patient bone marrow samples revealed a high interindividual variability of relative BAX and BAK localizations (Fig. 1C; Supplementary Table S1; Supplementary Fig. S1). Nonetheless, relative BAX localizations show a general trend towards cytosolic localizations, whereas BAK is present in larger mitochondrial pools (Supplementary Fig. S2A). Comparison of cytosolic and mitochondrial pools of BAX or BAK shows no apparent correlation (Fig. 1D; Supplementary Fig. S2B). The absence of correlations between the levels of mitochondrial and cytosolic BAX suggests that BAX localization does not depend on cellular BAX levels. Also the maintenance of a constant ratio between mitochondrial and cytosolic BCL-2 protein pools is not apparent.

The analysis of test cohort quartiles [low, mostly cytosolic; high, largely mitochondrial, medium high (medH) and medium (medL)] of relative BAX localization reveals a strikingly better survival probability among patients with high relative BAX localization (Fig. 1E). Low relative BAX localization values increase the odds of death during treatment by 11.5 times. The same difference is apparent analyzing only patients from the test cohort with de novo AML (Supplementary Fig. S3). On the basis of the similar response of patients with intermediate or low relative BAX localization, the use of relative BAX localization as continuous value is not appropriate to distinguish responder and nonresponder (Supplementary Fig. S4). However, categorizing patients with AML according to their relative BAX localization could predict treatment response.

Among test cohort patients, relative BAX localization shows a strong positive correlation with relative BAK localization (Fig. 2A). A strong correlation between cytosolic BAX and relative BAX localization suggests high variability of cytosolic BAX pools compared with similar mitochondrial levels detected among the patients with AML (Supplementary Fig. S5A and S5B). These results indicate that cells tolerate a larger variety of BAX concentrations in the cytosol, probably explained by the fact that cytosolic BAX is monomeric and lacks apparent protein interactions (30). Of note, the strong positive correlation between cytosolic BAX and BAK suggest that cells also tolerate a large variety of cytosolic BAK despite a greater tendency of this BCL-2 protein to localize on the mitochondria (Supplementary Fig. S5C). The subcellular distribution of BAX and BAK is clearly regulated by the same retrotranslocation process in pretreatment bone marrow samples of patients with AML. Therefore, the localization analysis of BAX is sufficient to assess the cellular BAX/BAK regulation.

The potential of relative BAX localization as a diagnostic tool was analyzed in the test cohort of 48 patients with AML (Supplementary Table S2.) and correlated with clinical data. Exemplarily, FMS-like tyrosine kinase 3 internal tandem duplications (FLT3-ITD), a bona fide mutational marker of negative prognostic impact in AML (31, 32), were investigated. Strikingly, FLT3-ITD–positive patients show low relative BAX localization values, suggesting increased cytosolic BAX and increased protection against apoptosis (Fig. 2B). Accordingly, FLT3-ITD is associated with a nonfavorable prognosis (Supplementary Fig. S6A). FLT3-ITD has been associated with activation of both prosurvival BCL-2 proteins and BH3-only proteins and could therefore induce changes in the relative BAX localization (33–36). Analysis of patient collective quartiles [low, mostly cytosolic; high, largely mitochondrial, medium high (medH) and medium (medL)] of relative BAX localization confirms predominant occurrence of FLT3-ITD in samples with low relative BAX localization values (Fig. 2C). The relative risk of FLT3-ITD in this group is at least 6.25 times of that in the groups with medium and high relative BAX localization values. The same coherence is apparent based on the analysis of the relative BAK localization, again suggesting a BAX localization–based analysis is sufficient for the regulatory status of both prosurvival BCL-2 proteins (Supplementary Fig. S6B). On the other hand, the analysis based on BAX levels instead of BAX localization shows FLT3-ITD occurrence in patients with high cellular BAX levels (Supplementary Fig. S7). Remarkably, patients with intermediate BAX localization values show inferior survival compared with high relative BAX localization patients despite a similar chance of FLT3-ITD (Figs. 1E and 2B). Of note, one FLT3-ITD–positive individual with high relative BAX localization value shows prolonged survival following chemotherapy. FLT3-ITD thus may influence relative BAX localization, but many other factors (e.g., genetic or epigenetic) also determine BAX localization and the predisposition to apoptosis.

Consistently, single-nucleotide variants (SNVs) occurring in patient samples of high or other relative BAX localization were analyzed. Overall, distribution and frequency of SNVs was similar across the different BAX localizations reflecting a quite homogenous pattern of genetic alterations and very little genomic diversity among the different patient groups (Fig. 2D). Notably, patient samples with high relative BAX localization displayed a significantly lower number of genetic losses compared with all other samples. GISTIC analyses further confirmed a high concordance of driving copy-number alterations across the different groups and very few specific alterations (Supplementary Table S3). Notably, while no recurrent losses were detected in high BAX localization, both intermediate and low relative BAX localization quartiles showed losses on 5q31.3. Nevertheless, comparison of the differences between SNVs in BAX high versus the other subgroups showed little overlap among the different subgroups, indicating the different BAX localizations might be determined by diverse molecular mechanisms and cell stress rather than common genetic alterations (Supplementary Fig. S8).

In an attempt to identify further factors that are reflected by the relative BAX localization, we analyzed potential and known predictors of response to therapy and survival in AML. In the test cohort, patient age seems to have no influence on relative BAX localization (Supplementary Fig. S9A). Surprisingly, the established overall risk rating shows no striking congruence with relative BAX localization (Supplementary Fig. S9B). However, patients with therapy-associated AML present with elevated mitochondrial BAX levels (Supplementary Fig. S10). This association is conceivable, as prior nontoxic cellular stress could be reflected in the relative BAX localization (29). Strikingly, the initial white blood count is significantly increased in patients with low relative BAX localization compared with patients with BAX shifted to their mitochondria (Fig. 2E). Relative BAX localization shows a negative correlation with initial leukocytosis (Fig. 2F). A shift of the BAX population towards the mitochondria and high predisposition to commitment to mitochondrial apoptosis thus is associated with lower initial leukocyte counts.

Treatment of primary AML samples of seven patients with six different apoptotic stimuli shows reduced apoptotic response by AML cells with low relative BAX localization (Fig. 2G). Patient samples with high relative Bax localization show increased apoptotic response, resulting in a correlation of relative BAX localization and apoptotic response, particularly for doxorubicin- (and daunorubicin-) treated cells (Fig. 2H). Relative Bax localization can identify cellular predisposition to apoptosis.

The link between BAX localization and predisposition to apoptosis in AML was further tested by determining the relative BAX localization in eight different AML cell lines (Fig. 3A; Supplementary Fig. S11). Consistent with the AML patient data, the analysis shows increased pools of mitochondrial BAK compared with BAX (Supplementary Fig. S12A). However, in AML cell lines, this difference is much more pronounced, while relative cytosolic BAX and relative mitochondrial BAX also lack an apparent correlation (Supplementary Fig. S12B).

The effects of relative BAX localization in AML cell lines on the induction of cell death were investigated in response to cytotoxic stress induced by daunorubicin. Application of 1 μmol/L daunorubicin for 24 hours resulted 2%–16% cell death depending on the cell line. The signal-to-noise ratio was increased using 20 μmol/L daunorubicin (37), resulting in robust cell death induction in seven of the eight tested AML cell lines (Fig. 3B). However, HEL cells did not induce cell death even at 20 μmol/L daunorubicin, revealing remarkably reduced predisposition to cell death compared with other cell lines. Noteworthy, HEL cells show a particular low relative BAX localization (Fig. 3C). Treatment with cytarabine (ara-C) for 24 hours did not induce significant apoptosis in HEL or OCI-AML3 cells (Supplementary Fig. S13). Comparing the apoptotic response of HEL cells and OCI-AML3 cells with increased mitochondrial BAX pools in response to daunorubicin treatment demonstrates a robust activation of caspases-3/7 in OCI-AML3 cells, whereas HEL cells show low levels of active caspase after 24 hours (Fig. 3D). This difference is corroborated by accumulation of the cleaved form of the caspase substrate PARP in OCI-AML3 cells but not HEL cells (Fig. 3E). These results corroborate the correlation between BAX localization and apoptotic response, resulting from altered BAX retrotranslocation rates, as previously shown in HCT116 and MEF cells (29). In parallel, HEL cells show increased resistance against different apoptotic stimuli and the BH3 mimetic UMI-77 (Fig. 3F; Supplementary Fig. S14). Among the 8 AML cell lines, MV4-11 cells contain the FLT3-ITD. FLT3 inhibition results in a shift of BAX towards the mitochondria suggesting an increase of apoptosis predisposition of MV4-11 cells but not THP-1 cells (Fig. 3G). However, MV4-11 cells are not particularly protected against apoptosis induction, suggesting an influence on relative BAX localization by FLT3 activity among other factors. The responses of AML cell lines to apoptotic stimuli suggest that relative BAX localization could predict predisposition to apoptosis.

Next, we focused on the patients with an age of 60 years and older who are frequently not eligible for intensive chemotherapy or hematopoietic stem cell transplantation. Association of the susceptibility to cytotoxic therapy with relative BAX localization could facilitate treatment decisions toward intensive treatment versus nonintensive approaches such as targeted- or immune-therapies and best supportive care. Among elderly patients (n = 25), a pronounced difference in relative BAX localization became evident when comparing FLT3-ITD–positive to FLT3-ITD–negative patients (Fig. 4A). While 60% of elderly patients with low relative BAX localization levels harbor FLT3-ITD mutations, no elderly patient with high relative BAX localization values is FLT3-ITD–positive (Fig. 4B). These observations are supported by the analysis based on relative BAK localization (Supplementary Fig. S15). In addition, different blast counts between samples with low and high relative BAX localization values were detected (Fig. 4C). Therefore, it is tempting to speculate on a potential use of relative BAX localization to guide clinical decisions as diagnostic or prognostic tool for elderly patients with AML.

To evaluate our previous insights into the localization of BAX and BAK in human AML and correlation of BAX localization with clinical markers, we analyzed a validation cohort of 80 elderly patients with AML treated with myelosuppressive chemotherapy within the AMLSG. Consistently with the test cohort, correlation between mitochondrial and cytosolic protein levels is not apparent (Fig. 5A; Supplementary Table S4; Supplementary Fig. S16). As observed in the test cohort, a wide variety of BAX/BAK localizations were detected with a general trend toward cytosolic BAX localization and mitochondrial BAK (Fig. 5B). A strong correlation also exists between the relative localizations of BAX and BAK in the validation cohort (Fig. 5C), demonstrating similar regulation of both proapoptotic BCL-2 proteins in human AML.

Different relative BAX/BAK localizations were detectable in the validation cohort depending on the mutational status of FLT3, although the difference was less pronounced compared with the test cohort (Fig. 5D). The same effect could be observed for the subgroup of intermediate risk patients according to ELN risk criteria (Supplementary Fig. S17). When applying the test cohort criteria and categorizing patients according to low (mostly cytosolic), medium low (medL) medium high (medH), and high (largely mitochondrial) BAX localization, decreased risk of patients with high mitochondrial BAX for FLT3-ITD positivity can be confirmed (Fig. 5E). Relative BAX localization also correlates with leukocytosis in the validation cohort (Fig. 5F; Supplementary Fig. S18), suggesting a similar correlation between relative BAX localization and predisposition to apoptosis in both patient cohorts.

Next, we analyzed the potential link between relative BAX localization and therapeutic response. Of note, 16 patients from the validation cohort that underwent allogeneic stem cell transplantation were excluded, as their therapeutic response is rather related to immunogenic control than to susceptibility to cytotoxic therapy. The analysis shows a tendency of mitochondrial BAX localization towards positive response to chemotherapy (Supplementary Fig. S19). Moreover, patients with high relative BAX localization are less likely refractory to chemotherapy (Fig. 6A). This group of patients also shows a trend toward improved survival (Fig. 6B). However, elderly patients from the test cohort with high relative BAX localization values show strikingly improved survival (Fig. 6C). The effects observed in the intensively treated validation cohort (Fig. 6B) are less pronounced compared with the test cohort, where patients received nonmyelosuppressive regimens (cytoreductive therapies or best supportive care; Figs. 1E and 6C). Again, the analysis of relative BAX localization as a continuous value provides no significant difference in receiver operating curve for validation cohort patients (Supplementary Fig. S20). The analysis basing on categorized BAX levels is showing a clear trend in both patient cohorts but results are still suffering from the small number of participants. Logistic regression on survival as a function of relative BAX localization, sex, FLT3-ITD, ELN-risk, and total-risk showed a significant influence of increased mitochondrial BAX levels OR = 1.61 (95% confidence interval, 1.1–2.4), while all other showed trends on survival influence but failed to reach significance (Supplementary Table S5).

BAX and BAK share common regulation of their localization in primary AML patient samples and AML cell lines. In healthy cells, both proapoptotic BCL-2 proteins are retrotranslocated from the mitochondria into the cytosol by the same prosurvival BCL-2 protein-dependent mechanism (25, 26). Shuttling at similar rates, BAX and BAK share similar apoptotic activity and subcellular localization (26). Therefore, the analysis of the relative BAX localization is sufficient to determine the status of BAX/BAK signaling. The total cellular protein level is not linked to BAX regulation. Increased BAX level could even be associated with more apoptosis-resistant scenarios (Supplementary Fig. S6). This argues against previous attempts to predict cellular survival or chemotherapy response solely by analyzing cellular BAX levels (38, 39). Also, the determination of the BAX/BCL-2 ratio, as previously suggested (40, 41), considers a potentially important interaction for BAX retrotranslocation, but neglects BAK-dependent apoptosis and potential cell protection from BAX/BAK activity by other prosurvival proteins, such as Bcl-x_L and Mcl-1 (25). The analysis of relative BAX/BAK localization includes the combined activities of prosurvival BCL-2 proteins and potential influence of BH3-proteins, like Bim, on BAX/BAK retrotranslocation (25, 26, 29). The equilibrium between cytosolic and mitochondrial BAX pools displays a surprising diversity in AML patient samples that reflects the heterogeneous response to chemotoxic stress. Significant differences mainly in the cytosolic BAX pool hold the promise of discriminating between clinically different AML.

Cellular survival requires fast BAX retrotranslocation from the mitochondria (26, 27). Consequently, a shift of the equilibrium between cytosolic and mitochondrial BAX toward the mitochondria results in increased apoptosis induction in response to apoptotic stimuli (29). In parallel to other cell cultures, reduced predisposition to apoptosis is observed in HEL cells with cytosol-shifted BAX compared with other AML cell lines. The other tested cell lines show a similar response to apoptotic stimuli, corroborating previous results from BH3 profiling (42). While BH3 profiling has the potential to dissect contributions of different key proteins, cell permeabilization, required for this type of analysis, removes cytosolic protein pools and triggers reequilibration potentially interfering with the determination of cellular predisposition to apoptosis. Relative BAX localization reveals the status of cellular BAX/BAK regulation, including contributions of known and unknown factors but cannot dissect their contributions.

Relative BAX localization itself is influenced by a plethora of factors, probably including genetic aberrations, for example, FLT3-ITD. However, genetic determinants of BAX regulation are heterogeneous in the analyzed patients' samples. The use of genomic data for risk stratification is probably prone to underscore the effect of mutational events on transcriptional, translational, and posttranslational consequences. They may be further influenced by the microenvironment or epigenetic alterations. Presence of complex aberrant cytogenetic changes has been described to correlate with resistance to low-dose cytarabine treatment before (43). However, even patients without any cytogenetic aberrations do not uniformly respond to chemotherapy. Most recently, an integrative genomic and clinical analysis of 1,540 AML patients enrolled in three prospective clinical trials and treated with myelosuppressive chemotherapy identified a set of 5,234 driver mutations with more than one driver mutation occurring in the vast majority of patients (86%; ref. 44). Moreover, cooccurring mutations influenced outcome and survival also beyond the previously described categories (6), highlighting the complexity of risk-assessment in AML. Here, determination of cellular susceptibility toward induction of apoptosis could potentially serve as a rapid and suitable assessment to predict therapeutic response and to stratify patients towards myelosuppressive chemotherapy versus rather nonintensive therapy such as targeted agents, epigenetic treatment, or immunotherapy.

Clinically, the predictive value of certain prognostic markers has a distinct impact, depending on the patient's fitness and comorbidities. AML of the elderly or frail patient requires detailed assessment of physical status and past medical history before patients are stratified toward intensive treatment versus nonintensive therapies or best supportive care. The majority of patients diagnosed with AML is above the age of 60 years and many of them do not have any curative options due to comorbidities or lack of response. Achievement of remission leads to prolonged survival and improved quality of life of the patient; however, which patients will respond is difficult to predict. When focusing on these patient-oriented aspects, prediction of response by using one major regulator of cellular resistance could facilitate stratification of elderly and frail patients to chemotherapeutic approaches versus targeted therapies or best supportive care.

In principle, BAX localization can be analyzed within a short period of time (less than 72 hours) and requires low amounts of cells (<1 × 10⁷ cells) derived from the peripheral blood or bone marrow. Our analysis has shown that the use of relative BAX localization as categorical but not continuous value identifies patients that are likely to have favorable therapeutic response. The strength of the analysis of the relative BAX localization in comparison with other available markers is the direct detection of the cellular predisposition to apoptosis in response to cytotoxic stress or specific inhibitors (e.g., BH3 mimetics). This final stage of cell fate decision could incorporate various effectors of leukemia development and maintenance: genomic and epigenetic aberrations, metabolic changes, cell-intrinsic and cell-extrinsic signaling, and modifications/homeostasis of the proteome.

In our analysis, determination of relative BAX localization shows that BAX shift toward the mitochondria is associated with predisposition to apoptosis and improved patient survival. These findings indicate increased susceptibility of the predominant AML clone to cytotoxic therapies. BAX localization shift toward the cytosol, on the other hand, leads to a resistance phenotype and is associated with a high tumor burden and FLT3-ITD, described to coincide with poor prognostic outcome. These correlations are apparent, although not homogeneous, in both cohorts of patients with AML analyzed in this article. Of note, small sample size and heterogeneity of consolidation (e.g., allogeneic stem cell transplantation) or second-line therapies (demethylating agents, targeted therapies) may significantly influence analysis of survival as readout for response. Still, BAX localization may predict response and outcome and its predictive value is even better detectable in analysis of rather small but more homogeneous subgroups (Fig. 6B and C). Prospective analysis of response rates to first-line treatment especially in a higher number of older patients with AML is necessary to validate its predictive value in the future.

Our findings indicate the potential use of BAX localization to assess for response to cytotoxic therapy in AML but also in the treatment of other tumor entities, as the underlying BAX regulation is a general mechanism that prevents BAX activity in human cells. Nevertheless, this analysis identifies also the limits of our analysis, as we assess biologic response of the dominant AML clone at diagnosis. Subclones may be selected or emerge during chemotherapy. Overall, we provide first evidence that BAX localization is associated with response to chemotherapy.

No potential conflicts of interest were disclosed.

Conception and design: F. Reichenbach, C. Wiedenmann, F. Edlich

Development of methodology: F. Reichenbach, C. Wiedenmann, F. Edlich

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): F. Reichenbach, C. Wiedenmann, E. Schalk, D. Wolleschak, K. Döhner, J.U. Marquardt, F. Heidel

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): F. Reichenbach, C. Wiedenmann, K. Funk, P. Scholz-Kreisel, F. Todt, K. Döhner, J.U. Marquardt, F. Heidel, F. Edlich, D. Becker

Writing, review, and/or revision of the manuscript: F. Reichenbach, C. Wiedenmann, E. Schalk, P. Scholz-Kreisel, K. Döhner, J.U. Marquardt, F. Heidel, F. Edlich

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): F. Reichenbach, F. Heidel

Study supervision: F. Edlich

This work was supported by the DFG Emmy Noether program, the Collaborative Research Cluster (CRC) 746, the Else Kröner-Fresenius-Stiftung, the Wilhelm Sander-Stiftung, the Spemann Graduate School of Biology and Medicine (SGBM, GSC-4) and the Centre for Biological Signalling Studies (BIOSS, EXC-294) funded by the Excellence Initiative of the German Federal and State Governments (F. Reichenbach, C. Wiedenmann, F. Todt, F. Edlich). F. Heidel was supported by a grant of the German Research Foundation, Collaborative Research Cluster (CRC) 854 (Project A20), the ProExcellence Research Initiative “RegenerAging” (State of Thuringia), the Thuringian country programme ProExzellenz (RegenerAging - FSU-I-03/14) of the Thuringian Ministry for Research (TMWWDG), and the German Jose-Carreras Leukemia Society (DJCLS SP12/08). K. Döhner was supported by a grant of the German Research Foundation, Collaborative Research Cluster (CRC) 1074 (Project B3) and the AMLSG study group. J.U. Marquardt is supported by the German Cancer Aid (DKH 110989), the Wilhelm Sander-Stiftung, and the Volkswagen Foundation (Lichtenberg program).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.