Microenvironmental Hypoxia regulates FLT3 expression and biology in AML

Fms-like tyrosine kinase 3 (FLT3) is a receptor tyrosine kinase constitutively expressed by acute myeloid leukaemia (AML) blasts. In addition, 25% of AML patients harbour a FLT3-ITD mutation, associated with inferior outcome due to increased relapse rate. Relapse might be propagated by interactions between AML blasts and the bone marrow microenvironment. Besides cellular elements of the microenvironment (e.g. mesenchymal stromal cells), bone marrow hypoxia has emerged as an additional crucial component. Hence, effects of hypoxia on FLT3 expression and biology could provide novel insight into AML biology. Here we show that 25% of AML patients down-regulate FLT3 expression on blasts in response to in vitro hypoxia (1% O2), which was independent of its mutational state. While virtually no AML cell lines regulate FLT3 in response to hypoxia, the down-regulation could be observed in Ba/F3 cells stably transfected with different FLT3 mutants. Hypoxia-mediated down-regulation was specific for FLT3, reversible and proteasome-dependent; with FLT3 half-life being significantly shorter at hypoxia. Also, PI-3K inhibition could partially abrogate down-regulation of FLT3. Hypoxia-mediated down-regulation of FLT3 conferred resistance against cytarabine in vitro. In conclusion, FLT3 expression in AML is dependent on the oxygen partial pressure, but response to hypoxia differs.

FLT3 have been extensively studied but have been by and large disappointing 10 . Interestingly, recent evidence suggests that the addition of sorafenib to chemotherapy improves event-and relapse-free survival but not overall survival in younger patients 11 . Relapse develops in the bone marrow where LIC are obviously able to survive anti-leukemic therapy 12 by interacting with components of the bone marrow microenvironment 13 . As relapse rates in FLT3-ITD patients are higher than in FLT3-WT patients, one might assume a role for FLT3 in microenvironmental interactions, conferring e.g. survival advantage by rendering leukemic blasts less sensitive towards anti-leukemic drugs or TKI. The bone marrow microenvironment comprises cellular (e.g. perivascular and mesenchymal stromal cells) 14 and non-cellular components (e. g. cyto-and chemokines) 14 , which usually work in concert for the maintenance of the normal haematopoiesis. Moreover, several studies have implied that the bone marrow is hypoxic, and characterised by heterogeneous areas of low oxygen levels that can vary from 6% O 2 to 1% O 2 15,16 . These low oxygen areas have been associated with physiological functions like differentiation of haematopoietic stem cells 17 . Hypoxia might influence AML development, and also alter biological features of AML like susceptibility towards chemotherapy 18 and biology of cytokine receptor CXCR4 15 . CXCR4 is known to have a fundamental role in haematopoiesis (e.g. homing of haematopoietic stem cells towards the bone marrow niche) and it has additionally been linked to FLT3 in AML 19 . Hence, it seemed worthwhile to investigate the effects of hypoxia (as a physiological mainstay of the bone marrow microenvironment) on FLT3, in AML as well.

Methods and Materials
Patient samples. Primary samples were collected from AML patients who underwent routine diagnostic bone marrow aspiration or from peripheral blood, after informed consent was obtained. All were diagnosed with AML according to WHO criteria. Mononuclear cells were separated by Ficoll-Hypaque (Sigma-Aldrich, St Louis, MO, USA) density-gradient centrifugation and cultured at a density of 10 6 /ml in RPMI 1640 containing 10% FCS, 2 mM glutamine and 1% penicillin-streptomycin. All experiments were performed on fresh or frozen cells and approved by the local ethics committee (Ethic Committee of the University Hospital Großhadern, University of Munich) and carried out in accordance with the approved guidelines.

Western blotting. For Western Blot analysis, cells were lysed in 1× Cell lysis buffer (Cell Signaling
Technology, Danvers, MA, USA) in either normoxia or hypoxia. Lysis buffer was supplemented with a protease inhibitor cocktail (Roche, Mannheim, Germany). Lysates were then separated on a 8% polyacrylamide gel, transferred to Hybond-P membranes (GE Healthcare, Little Chalfont, United Kingdom), probed with the appropriate antibodies FLT3 (c-20), EGF-R (1005), Epo-R (M-20) (all from Santa Cruz Biotechnology, Santa Cruz, CA, USA), c-KIT, p-STAT5, p-AKT (Ser473), GAPDH (all from Cell Signaling Technology, Danvers, MA, USA), β -actin (Sigma-Aldrich, St Louis, MO, USA) and visualized using an enhanced chemiluminescence kit (Thermo Scientific, Rockford, IL, USA). Relative quantification of protein bands from western blot films was performed by using ImageJ software and calculated as optical density (OD). OD was calculated as a ratio of each protein band relative to the lane's loading control.
Flow cytometry. Flow cytometric data were acquired using a FACSCalibur cytometer (BD Biosciences, San Jose, USA). Assessment of apoptotic cells were performed with the FITC Annexin V Apoptosis Detection Kit I (BD Pharmingen, Heidelberg, Germany). Analysis of Annexin-V expression was performed using BD CellQuest Pro software. After digest with 1 μ g/ml RNAse A (Sigma-Aldrich, St Louis, MO, USA) cells were stained with 50 μ g/ml PI (Sigma-Aldrich, St Louis, MO, USA) and DNA content was quantified by flow cytometry.
Real-Time PCR. Total RNA was isolated from 10 6 cells using MagNA Pure LC mRNA HS Kit (Roche, Basel, Switzerland) according to the manufacturer's instructions. cDNA was synthesized from 1 μ g aliquots of total RNA in a 20 μ L standard reaction mixture using SuperScript ® III First-Strand Synthesis System for RT-PCR (Invitrogen, Camarillo, CA, USA) according to the manufacturer's instructions. Quantitative Real-time polymerase chain reaction (RT-PCR) was performed using the 7900HT Fast Real-Time PCR System (Applied Biosystems, Waltham, MA, USA) with 2 μ L of cDNA, Fast SYBR ® Green Master Mix (Applied Biosystems, Waltham, MA USA), and the following primers TGCCCGTCTGCCTGTAAAAT (forward) and AACCGGAATGCCAGGGTAAG (reverse) for FLT3. β -actin (Eurofins Scientific, Luxembourg, Luxembourg), was used as a reference for normalization of Δ Ct values. All experiments were performed in triplicates.
WST-Assay. Cytotoxycity was assessed by WST-1 assay according to the manufacturer's instructions.
After 48 hours 2.5 × 10 4 Ba/F3 cells were exposed for 24 hours to increasing concentrations of cytarabine (Sigma-Aldrich, St Louis, MO, USA) or sorafenib (LTK Laboratories, St. Paul, MN, USA). 20 μ l of WST-1 reagent were then added for a further incubation period of 3 hours. The ELISA reader (OptiMax, Molecular Devices, Sunnyvale, CA, USA) was set at a wavelength of 450 nm with a reference wavelength of 690 nm. Viability of cells after drug exposure was calculated as a percentage relative to untreated controls. CD34+ cell isolation. CD34+ cells were purified from not infiltrated bone marrow by magnetic bead separation using the human CD34 MicroBead kit and the MACS Pro separator (Miltenyi Biotec, Bergisch Gladbach, Germany) as instructed by the manufacturer. The purity of the CD34+ fraction was assessed by flow cytometry using an anti-CD34− PE antibody (Miltenyi Biotec, Bergisch Gladbach, Germany), and CD34+ fraction showing purity higher than 75% was used.

Statistical analysis.
Results are shown as the mean or median ± standard deviation of the mean (SEM) or standard deviation (SD) of at least 3 experiments each. Paired data were analysed using the paired Student t test. For survival analyses, the method of Kaplan-Meier using Log Rank test was used. Optical density (OD) was analysed using Image J Software. Microsoft Excel was used for data acquisition, storage and analysis, with additional calculations done with SPSS Statistics ® (Version 22, IBM). A P value < .05 was considered statistically significant.

Results
FLT3 expression in AML subgroups is regulated by the oxygen partial pressure and might affect prognosis. In a first step, primary AML cells (n = 8) were analysed by western blot for their basal total FLT3 protein expression from uncultured cells. FLT3 expression levels were heterogeneous but independent from the mutational state of FLT3 (Fig. 1A,B). Next, we investigated by western blot analysis the effects of in vitro hypoxia of 1% O 2 on FLT3 expression in AML patient samples (n = 33, 11 ITD, 22 WT). As expected, expression of FLT3 was highly variable: in 10 patients no FLT3 expression was observed (30%), while in 23 patients (70%), FLT3 levels could be detected. Surprisingly, in the latter group, we found two distinct behaviours concerning the effect of hypoxia on FLT3 expression ( Fig. 1C): while 65% showed no difference in FLT3 expression after 48 hours of exposure to 1% O 2 compared to 21%, in 35% of samples a significant down-regulation of FLT3 expression was observed (80% reduction at 1% O 2 , p = 0,01; Fig. 1D). The down-regulation of FLT3 was independent from the mutational state of FLT3 (down-regulators: ITD 57%, WT 42%, non-regulators: ITD 33%, WT 67%, not significant). Clinical data was available for 24 of these patients. Median age was 61.9 years [23.2-82.0] with 62.5% female patients. Seven samples showed in-vitro down-regulation of FLT3, 9 were non-regulators and in 8 patients no FLT3 was detected by western blotting. Adequate blast clearance (defined as < 10% myeloblasts 7-10 days after completion of induction chemotherapy) was achieved in 100% of patients whose blasts showed hypoxia-mediated down-regulation of FLT3 in vitro and by 66% of not-regulating patients (not significant). Kaplan-Meier analysis revealed that patients with a regulating phenotype tended to live longer than patients without in vitro regulation of FLT3, however the difference was not significant (p = 0.09, Fig. 1E). The reason for the possible difference in survival was due to 3 early deaths, and interestingly, FLT3 non-regulators tended to have higher leukocyte count in the peripheral blood at the time point of initial diagnosis (107 G/l versus 35 G/l; p = 0.08). Additionally, FLT3 expression was analysed in normal CD34+ cells after 48 hours of exposure to 1% O 2 and showed down-regulation of FLT3 as well (Fig. 1F).   Fig. 2A). We than switched to Ba/F3 cells and analysed FLT3 expression in cells stably transfected with either FLT3-WT or FLT3-ITD construct W51. In this model, hypoxia induced down-regulation of FLT3 was reliably observed in all Ba/F3 cells independently from the mutational state of the receptor (Fig. 2B). As we were unable to detect significant differences between FLT3-WT and the FLT3-ITD mutant with respect to FLT3 down-regulation, Ba/F3 W51 (FLT3-ITD) cells were chosen for subsequent experiments due to their long-term IL-3 independent in vitro growth. With this in vitro model of hypoxia-mediated down-regulation of FLT3, we established the oxygen threshold required for down-regulation between 1% and 6% O 2 (Fig. 2C) and the time window for down-regulation between 48 and 72 hours of hypoxia (Fig. 2D). FLT3 down-regulating phenotype was maintained after 9 and also after 12 days of hypoxia, suggesting that hypoxia-mediated FLT3 down-regulation is not a short-term stress reaction of AML cells towards 1% O 2 (Fig. 2E). However, FLT3 down-regulation is reversible. Ba/F3 cells cultured for 3 days in hypoxic conditions showed a restoration of total protein within the first 24 hours after re-oxygenation with 21% O 2 (Fig. 2F,G) and also exposure to fresh medium (data not shown). As expected, exposure to 1% O 2 decreases proliferation of Ba/F3 cells (cell counts after 72 hours: 2.1 × 10 5 /ml for 21% O 2 , 1.3 × 10 5 /ml for 1% O 2 ; p = 0.007, Fig. 2H), but viability was unaltered (88.7% viable cells at 21% O 2 versus 85.4% at 1% O 2 ; not significant, Figure I).

Mechanism of FLT3 down-regulation.
Half-life (t½) of FLT3 was reported as 2-3 hours 20 , however we found t½ of FLT3 at 1% O 2 to be significantly shorter: as shown in Fig. 3A,B, degradation of FLT3 protein was fastened by 1 h at 1% O 2 , suggesting that hypoxia increases the turn-over of the protein. Treatment with the proteasome inhibitor MG132 (5 μ M) reversed FLT3 degradation both at 1% O 2 and at standard laboratory conditions (Fig. 3C), indicating that proteasome dependent degradation of FLT3 is also active during hypoxic conditions. Total levels of CBL were unaltered by hypoxia (data not shown). To investigate whether transcriptional mechanisms might also be involved in the different expression of FLT3 under hypoxic conditions, real-time PCR was performed for FLT3 mRNA levels at 1% O 2 and 21% O 2, but no significant differences were observed in FLT3 transcript level at the different oxygen conditions (Fig. 3D). ELISA-based analysis of the supernatant of FLT3 down-regulating cells did not show detectable levels of FLT3, making extracellular shedding of the protein, as a potential mechanism of loss of FLT3 protein by hypoxia, highly unlikely (data not shown). To further identify possible other molecular mechanisms, we investigated the effects of phosphatidylinositide 3-kinase (PI3-K), Heat shock Protein 90 (Hsp90) inhibition and HIF-hydroxylase inhibitor (i.e. HIF1alpha stabilization) on hypoxia-mediated FLT3 down-regulation. As shown in Fig. 3E, FLT3 down-regulation could partially be abrogated by PI3-K inhibition but was neither affected by Hsp90 inhibition nor by HIF-1 alpha activation, suggesting that hypoxia-mediated FLT3-down-regulation might be partially regulated by the PI3-K pathway, but independent from HIF1alpha.
Hypoxia mediated down-regulation is restricted to FLT3 protein, but independent from its mutational state. Results in primary AML blasts and in Ba/F3 WT and W51 cells suggested that FLT3 down-regulation is independent from the mutational state of the receptor. To further validate this hypothesis, we investigated the expression of FLT3 at 1% O 2 in 2 additional Ba/F3 FLT3-ITD mutants (Ba/F3 W78 and Ba/F3 NPOS) and in the kinase-inactive Ba/F3 K644R mutant. The hypoxia mediated down-regulation was also observed in each of these Ba/F3 cells (Fig. 4A,B). Next, we investigated whether hypoxia-mediated down-regulation is a general phenomenon applying to other tyrosine kinase receptors (TKR) and cytokine receptors as well; especially receptors that have a key role in haematopoiesis and leukaemogenesis as EGF-R, c-KIT and Epo-R. Ba/F3 cells stably transfected with EGF-R and Epo-R, were analysed for a 3-day time kinetic at 21% O 2 and 1% O 2 and, as shown in Fig. 4C,D, we failed to observe any hypoxia-induced regulation of the respective receptors.
Expression levels of EGF-R and KIT receptor were checked in primary samples (n = 4) and results are shown in Fig. 4F: there was no regulation of EGF-R and KIT receptor under 1% O 2 , regardless of FLT3 regulation. In particular, individual differences between different patients were observed for KIT receptor expression under 1% O 2 compared to 21% O 2 , nevertheless no significant down-regulation was observed in any of the patient sample tested (Fig. 4G).
Functional consequences of FLT3 down-regulation. The further characterize functional consequences of FLT3 hypoxia-mediated down-regulation, we tested the susceptibility of Ba/F3 cells towards the tyrosine kinase inhibitor sorafenib, according to the expression level of FLT3 (i.e. 21% O 2 versus 1% O 2 ). Interestingly, no differences in the anti-leukemic efficacy of the drug were observed between hypoxia and normoxia (IC50 5.06 μ M at 21%O 2 versus 3.93 μ M at 1% O 2 , not significant, Fig. 5A). In line with this finding was the observation that both canonical down-stream pathways of FLT3 (STAT5 and AKT) were unaltered by hypoxia as shown in Fig. 5B. However, FLT3

Discussion
FLT3 has been well studied in AML and mutations of this receptor belong to the oldest known genetic aberrations. The robust prognostic impact of FLT3 has been shown beyond doubt and a multitude of in vitro data has unravelled its biological function. However, therapies aimed at FLT3 have failed on various levels, and FLT3 targeted therapy has not entered standard therapy of AML. This suggests that large parts of FLT3 biology are still not understood. Several studies have implied in a direct or indirect way 16,21,22 , that the physiological state of the bone marrow is hypoxic and characterised by heterogeneous areas of low oxygen levels that can vary from 6% O 2 to 1% O 2 . We have recently reported that normal and leukemic bone marrow do not differ in their low oxygen levels 23 , suggesting that the definition of hypoxia in the context of AML differs from the ones used in other malignancies, where it is used to describe a non-physiologic condition, e.g. a non-physiologic hypoxic or anoxic core arising from insufficient neo-angiogenesis. The data presented here demonstrate that the oxygen partial pressure (pO 2 ), significantly influences expression and biology of FLT3. Hypoxia of 1% O 2 but not 6% O 2 down-regulates FLT3 expression in (I) appr. 25% of AML patients, (II) in normal CD34+ haematopoietic progenitors and (III) in the Ba/F3 cell line stably transfected with different FLT3 constructs. Down-regulation of FLT3 by hypoxia was irrespective whether the protein was mutated or not and of the carried mutation. In contrast, hypoxia-induced down-regulation of FLT3 does not occur in appr. 75% of AML patients and in virtually none of the investigated AML cell lines. The question that arises from this finding is whether down-regulation of FLT3 is a normal physiological consequence of hypoxia. Concluding from our data, we would support this hypothesis: (I) normal CD34+ positive cells showed FLT3 down-regulation, (II) the down-regulation was "specific" to FLT3 and not observed in other TKR and cytokine receptor relevant in haematopoiesis and (III) AML patients with a down-regulating phenotype tend to have a better prognosis (longer survival, higher CR rates and lower peripheral blast counts). This hypothesis is supported by the observation that in AML, high FLT3 expression levels confers a tendency towards worse survival 24 . Moreover the finding that 12 AML cell lines studied (which are usually established from more "aggressive" types of leukaemia with autonomous in vitro growth) do not regulate FLT3, further suggest that the non-regulating phenotype is not physiological, and that some AML blasts lose the property of a physiological, hypoxia-mediated FLT3 regulation. We chose to further characterize the phenomenon of hypoxia-mediated down-regulation of FLT3 -but admittedly the identification of loss of this feature could be rewarding in the context of AML as well. As only primary AML samples and FLT3 transfected Ba/F3 cells showed the hypoxia-mediated down-regulating phenotype, the latter murine system was chosen as a model for further experiments. It was found that the down-regulation of FLT3 by hypoxia follows the "normoxic" conditions, as the down-regulation remained proteasome dependent but with a t½ that was 1 hour shorter at hypoxia of 1% O 2 . Transcriptional mechanisms seemed not to be involved in FLT3 down-regulation. Interestingly, PI3-K might act as a possible relevant pathway connected to hypoxia-mediated FLT3 down-regulation. PI3-K pathway is one of the most frequently activated in AML and also activated by hypoxia, potentially by increased lipid raft formation 22 . Additionally, binding of FLT3 ligand to FLT3 activates PI3-K 6 , whereas mutated FLT3 signals through STAT5 and JAK pathways 25,26 . Surprisingly, in the murine system, none of these latter pathways were altered in every aspect we investigated, neither by hypoxia nor by the expression levels of FLT3. By inhibiting PI3-K pathway we could partially reverse the FLT3 down-regulation, indicative for a negative feedback loop between FLT3 and PI3-K at hypoxic conditions. The question how AML blasts and cell lines lose their ability to regulate FLT3 in response to pO 2 remains elusive. A "simple" mutation in a protein involved in FLT3 metabolism (e.g. in degradation as CBL 27 seems unlikely in the light of the fact that the un-regulating phenotype was observed in 75% of patients. A mutation with that frequency would have certainly been already detected. Hence, other mechanisms like different expression of degrading proteins or epigenetic mechanisms could be responsible. The effects however may be more obvious; increased expression of FLT3 is known to be associated with an inferior prognosis 24 , and presumably the increased expression of FLT3 receptor might result in increased FLT3-L binding and hence increased FLT3 signalling and activation of PI3-K -comparable to a constitutive activation. Hence, a constitutive expression of that receptor that eludes physiological regulation might result in the escape from environmental control. However, FLT3 might have other, hitherto unknown functions as well: surprisingly, we found that FLT3 itself affected the susceptibility of cells to cytarabine, as cells lacking FLT3 showed no difference in response to cytarabine according to oxygen in contrast to cells expressing FLT3. Whether and how FLT3 is involved in the pharmacodynamics of cytarabine certainly deserves further investigations, given the importance of cytarabine in the treatment of AML. In conclusion, these data show that FLT3 is regulated in hematopoietic cells by the oxygen partial pressure (as observed in the hematopoietic stem cell niche in the bone marrow) and that a substantial proportion of AML cells lose this phenotype. Whether this is an AML initiating event or a side effect of the e.g. well established hematopoietic differentiation block in AML needs to be further investigated, as well as the functional consequences.