Arsenic Trioxide Affects Signal Transducer and Activator of Transcription Proteins through Alteration of Protein Tyrosine Kinase Phosphorylation

Materials and Methods

Materials

All chemicals were purchased from Sigma Chemicals (St. Louis, MO) unless otherwise specified. Granulocyte colony-stimulating factor, granulocyte macrophage colony-stimulating factor, and arsenic trioxide were kindly provided by Amgen (Thousand Oaks, CA), Immunex (Seattle, WA), and Cell Therapeutic, Inc. (Seattle, WA), respectively.

Patient samples

Bone marrow samples were collected from four AML patients with >80% blasts at diagnosis. Two of the patients had complex karyotypes [02472 (French-American-British M6) and 02448 (French-American-British M2)], one had inv(16) [02537 (French-American-British M5a)], and one had normal karyotype with internal tandem duplication (ITD) of FLT3 [04153 (French-American-British M2)]. Light-density bone marrow cells were isolated by 1.077 g/mm³ Ficoll-Hypaque density gradient centrifugation and were cryopreserved using standard techniques. Cells were thawed before study. The study was approved by the Roswell Park Cancer Institute Scientific Review Committee and Institutional Review Board. All patients signed informed consent.

Cell lines

Cell lines used in the study included the AML cell line (HEL), a cytokine-independent human erythroleukemia cell line that carries the JAK2 V617F mutation (9) and has constitutive STAT activity, and the AML cell line constitutively expressing wild-type FLT3 (EOL-1). The human hepatoma cell line (HepG2) served as an example of a nonhematopoietic tumor line with gene induction and suppression of proliferation dependent on a cytokine-regulated STAT3 pathway (10). The human epithelial kidney 293 cell line was used for PTK expression by transfection (11).

Culture conditions

HEL and EOL-1 cells were exposed to arsenic trioxide at concentrations ranging from 0.25 to 4 μmol/L for 6 to 48 hours. HepG2 and 293 cells were exposed to arsenic trioxide at concentrations of 0.2 to 5 μmol/L for 6 to 24 hours to maximize the arsenic trioxide effect. HEL cells were also exposed to freshly prepared 100 μmol/L to 1 mmol/L sodium orthovanadate to inhibit protein-tyrosine phosphatase activities (12).

Cell proliferation and viability

Viability was determined by the trypan blue dye (Life Technologies, Gaithersburg, MD) exclusion assay. Apoptosis was analyzed with the Annexin-V-Fluos staining kit (Roche, Indianapolis, IN), following the manufacturer's directions; cells were evaluated on the FACSCalibur. Data analysis was done with WinList.

Morphologic analysis

To study the morphology of cells exposed to arsenic trioxide, cytospins were prepared at each time point. Following fixation with methanol, cells were stained with Giemsa May-Grünwald stain and analyzed by light microscopy (Nikon, Inc., Melville, NY).

Cell cycle analysis

DNA content was determined after ethanol fixation by staining the cells with propidium iodine in the presence of RNase. Cytofluorometric analysis was done using a FACScan (Becton Dickinson, San Jose, CA) flow cytometer equipped with a 488-nm argon line. Propidium iodine fluorescence was evaluated using linear amplification in the FL2 channel (585/42 band-pass filter). Data analysis was done with WinList and ModFit LT software for Windows (both from Verity Software House, Inc., Topsham, ME).

Transfection

To study the effect of arsenic trioxide of individual PTKs, 293 cells were transfected with expression vectors for the PTKs ZNF198/fibroblast growth factor receptor (FGFR1; ref. 11), green fluorescent protein–tagged ZNF198/FGFR1 (11), BCR/FGFR1 (13), p210^BCR/ABL (14), promyelocytic leukemia (PML; ref. 15), JAK1 (16), and JAK2 (17) using the calcium phosphate methods (18). After 24 hours of culturing, the transfected cells were subdivided for treatment. The gene regulatory effects of arsenic trioxide were studied in HepG2 cells transiently transfected with the STAT3-responsive reporter gene construct p(5xHPX-IL-6RE)-CAT (19) and the expression vectors for PTKs, as previously described (18). Subcultures of transfected HepG2 cells were treated with arsenic trioxide or interleukin-6 (IL-6), and the relative change in the expression of the CAT reporter gene was quantitated 24 hours later (18).

Western blotting

Tyrosine phosphorylated and nonphosphorylated STAT1, STAT3, STAT5A, STAT5B, JAK1, JAK2, extracellular signal-regulated kinase (ERK) 1/2, c-Jun NH₂-terminal kinase (JNK), ZNF198/FGFR1, BCR/FGFR1, BCR/ABL, and PML proteins were quantitated by Western blot analysis as previously described (2, 3). In brief, whole-cell extracts were separated on 6% or 7.5% polyacrylamide SDS gels, and the proteins were transferred onto nitrocellulose membranes. The membranes were incubated with antibodies against phosphorylated STAT1 (Y701), STAT3 (Y705), STAT5A/B (Y694/Y699), ERK1/2, JNK (Upstate Biotechnology, Lake Placid, NY), ABL (Cell Signaling Technology, Inc., Beverly, MA), and phosphotyrosine (PY20 and M1; Transduction Laboratories, Lexington, KY). To detect nonphosphorylated proteins, immunoblots were reacted with antibodies against the NH₂ termini of STAT1, STAT3, STAT5, ERK1/2, JNK (Transduction Laboratories), JAK1, JAK2, FGFR1, PML (sc-5621), and BCR (N-20; Santa Cruz Biotechnology, Santa Cruz, CA). The protein inhibitor of activated STAT3 (PIAS3) was detected with anti-PIAS3 (Abgent, San Diego, CA). To detect phosphorylated FLT3 expression, cell lysates were incubated at 4°C overnight with anti-FLT3 antibody (S18; Santa Cruz Biotechnology), then protein A agarose (Upstate Biotechnology) was added for an additional 2 hours (20). After electrophoresis and transfer to nitrocellulose membranes, immunoblotting was done using anti phosphotyrosine antibody 4G10 (Upstate Biotechnology) to assess phosphorylated FLT3 or with anti-FLT3 antibody to measure total FLT3. When necessary, anti-actin antibody (Santa Cruz Biotechnology) was used to show equal loading. The immune complexes were visualized by the enhanced chemiluminescence reaction (Amersham Life Science, Arlington Heights, IL).

Electrophoretic mobility shift assay

The DNA-binding activity of STAT3 and STAT5 in the absence of cytokine or growth factor treatment was assessed by electrophoretic mobility shift assay. Whole-cell extracts prepared from HEL cells as previously described (2, 3) were incubated with ³²P-labeled oligomers corresponding to the high-affinity binding element for STAT3, SIE (21), and STAT5 (TB2; ref. 22). The complexes were analyzed by 5% PAGE and autoradiography. Relative DNA-binding activity was determined by densitometry of the autoradiographs; the activity detected in extract from HEL cells served as standard and was defined as 100%. The detection limit of the electrophoretic mobility shift assay system was set at <2%. The identity of the STAT-containing complexes was determined by antibody supershift with the COOH-terminal–specific anti-STAT3 (C-20) or anti-STAT5 (C-17) monoclonal antibody (Santa Cruz Biotechnology).

In vitro arsenic trioxide reaction with ZNF198/FGFR1

ZNF198/FGFR1–transfected 293 cells were subdivided into two cultures. After an additional 1-day culture period, one culture was treated for 30 minutes at 37°C with medium alone and the other culture with medium containing 5 μmol/L arsenic trioxide. The cells were washed with ice-cold PBS and homogenized in the same buffer (1 × 10⁷/mL) using a tight-fitted Dounce homogenizer. Cell debris was removed by centrifugation for 10 minutes at 13,000 × g followed by preparation of the cytosolic fraction by centrifugation for 1 hour at 105,000 × g. PBS alone or PBS containing 5 μmol/L and 10-fold serially diluted arsenic trioxide were added to 4 μL of cytosol aliquots. These mixtures were incubated for 30 minutes at 37°C. The reactions were stopped by adding 15 μL of SDS sample buffer and boiling for 5 minutes. The apparent molecular size of ZNF198/FGFR1 was determined by SDS-PAGE and immunoblotting.

Statistical methods

The following variables of arsenic trioxide effect on HEL cells were evaluated: total number of cells, percentage of necrotic and apoptotic cells, and activity levels of phosphorylated STAT1, STAT3, STAT5 and ERK1/2, and expression levels of total STAT3 protein. The mean and variance were calculated for each set of replicates within each experimental condition. The effects of dose and time were evaluated using the Hill model (23, 24). Model goodness-of-fit was checked by plotting the actual versus predicted values, and by evaluating the 95% confidence intervals (95% CI) for variable estimates.

Due to the limited range of dose levels, only exploratory analyses were employed using nonparametric methods to compare phosphorylated and total STAT3 expression levels among the dose levels. To control the overall type I error rate at 10%, a Bonferonni adjustment (25) was made so that the level of significance for each statistical test was set at 0.008.

Pharmacodynamic models. The inhibitory effects of arsenic trioxide on cell proliferation and the phosphorylation of STAT1, STAT3, and STAT5 were analyzed using the Hill model equation (Eq. A) below (24). In this equation, C is the drug concentration, E is the measured effect, E_con is the control level observed when no drug is present, B is the background level observed at high drug concentration, and IC₅₀ is the drug concentration inducing 50% inhibition of the maximal effect (E_con − B). The slope variable γ corresponds to a shallow concentration-effect curve when its absolute value is <1 and corresponds to a steeper curve as its absolute value becomes >1. A negative sign of γ indicates an inhibitory effect.

Due to the limited number of replicates, the background variable B was fixed at its mean value, so that the other variables could be estimated with more precision.

A modification was made in the model for phosphorylated STAT1 activity, in which a fixed background level was not assumed due to wide variation in the data at 1 μmol/L arsenic trioxide. Instead, background was estimated as a separate variable in this model.

The effect of arsenic trioxide concentration on apoptosis was of a stimulatory nature, whereby increased drug concentrations were associated with increased percentages of apoptotic cells. The structural equation providing the best fit for these data was a modified version of the E_max model. In this structural equation, the minimum level is the control level when no drug is present (E_con), and the maximal effect is equal to the effect seen at the highest dose level (E_max). Setting E_con equal to its mean value in the equation yielded a more parsimonious model (Eq. B).

Equation B was also used to model the effect of arsenic trioxide dose on the percentage of dead (both necrotic and apoptotic) cells.

Results

Arsenic trioxide decreases phosphorylated STAT3 activity in patient samples. We first tested the effect of arsenic trioxide on STAT phophorylation in primary AML blasts with constitutively activated STAT3. Cells from three AML patients were treated with two doses of arsenic trioxide for 6 hours and the level of phosphorylated STAT3 was determined (Fig. 1A ). In each case, a reduction of tyrosine phosphorylated STAT3 was evident. Of note is that arsenic trioxide had no appreciable effect on the total STAT3 protein level, suggesting that arsenic trioxide specifically decreases the relative amount of activated STAT3. To ensure that the effect was STAT3-specific, we studied the effect of arsenic trioxide on phosphorylated and total ERK1/2 activity in the patient samples. As shown in Fig. 1A (bottom), arsenic trioxide did not reduce the basal level phosphorylation of ERK1/2 in the patient samples. In fact, the cells from one patient, 02448, which showed a low basal ERK phosphorylation compared with the cells from other two patients, responded to arsenic trioxide by an increased ERK phosphorylation that may be related to a particularly strong stress response. To evaluate the effect of arsenic trioxide on PTKs upstream of the STAT proteins, we studied a sample from an AML patient with normal karyotype but with FLT3-ITD. As shown in Fig. 1B (left), arsenic trioxide induced down regulation of FLT3-ITD and STAT5. Similar results were observed in the AML cell line, EOL-1, expressing consitutively active FLT3 (Fig. 1B, right).

Arsenic trioxide inhibits cell proliferation. HEL cells, like 50% of AML cases, express constitutive STAT activities and therefore were used here as a tissue culture model (9). When treated with arsenic trioxide for 6 or 24 hours, no appreciable effect on cell number was detected. Following 48 hours, HEL cells showed decreased proliferation (Fig. 2A ). The arsenic trioxide IC₅₀ concentration was calculated to be 1.23 μmol/L (95% CI, 0.94-1.53). To understand the mechanism leading to the reduced cell counts at 48 hours, percentages of necrotic and apoptotic cells were analyzed at 48 hours. The IC₅₀ for necrotic cells was equal to 0.98 μmol/L (95% CI, 0.26-1.69), whereas the IC₅₀ for apoptotic cells was 1.71 μmol/L (95% CI, −0.19 to 3.61). This indicated that arsenic trioxide caused necrosis at a lower concentration level during the 2-day treatment period in HEL cells.

Arsenic trioxide induces mitotic arrest. HEL cells were exposed to 4 μmol/L arsenic trioxide for 6, 24, and 48 hours. Mitotic arrest with chromosome condensation and disappearance of the nuclear envelope was first detected at 24 hours (Fig. 2B). This was associated with a decrease in cells entering the cycle with an increase in G₂-M phase cells detected by flow cytometry (Fig. 2C and D).

Arsenic trioxide modulates signal transduction pathways. HEL cells exposed to arsenic trioxide for 6 hours showed a sharp decrease in constitutive STAT1, STAT3, and STAT5 activities but not in the total level of these proteins, suggesting an arsenic trioxide–mediated reduction of activation and/or enhanced dephosphorylation (deactivation; Fig. 3 ). The results obtained for the DNA-binding activity of the STAT proteins by electrophoretic mobility shift assay and for relative level of tyrosine phosphorylation by Western blotting were highly complementary (Fig. 3A, top). The IC₅₀ for phosphorylated STAT1 was 0.72 μmol/L (95% CI, 0.19-1.25), and the slope variable was −2.24 (95% CI, −5.30 to 0.83), indicating a relatively steep decline in phosphorylated STAT1 as the arsenic trioxide dose level increased (Fig. 3B). Similarly, the IC₅₀ for phosphorylated STAT3 at 6 hours was 0.91 μmol/L (95% CI, 0.19-1.63), and the slope variable γ was −1.19 (95% CI, −1.96 to −0.43; Fig. 3C). Finally, the IC₅₀ for phosphorylated STAT5 was 0.32 μmol/L (95% CI, 0.20-0.44), and the slope variable was estimated to be −2.50 (95% CI, −3.95 to −1.05; Fig. 3D). The effect of arsenic trioxide was most pronounced on phosphorylated STAT5. The relative level of tyrosine phosphorylated STAT proteins continued to decrease during treatments longer than 6 hours, along with the decrease in survival. Removal of arsenic trioxide after 6 hours resulted in recovery of the phosphorylated STAT3 and STAT5 levels to baseline at 24 hours (data not shown). Similarly, there was no effect on cell survival after arsenic trioxide removal.

To test whether arsenic trioxide affected the mitogen-activated protein kinase pathway, arsenic trioxide activity on ERK1/2 phosphorylation level was determined. As shown in Fig. 4 (top), there was no appreciable specific effect of arsenic trioxide on ERK1/2 in HEL cells following 6 hours of incubation. The Hill model could not be derived. Because arsenic trioxide treatment could conceivably introduce a chemical stress response, the activation of the stress mitogen-activated protein kinase was also analyzed by immunoblotting for phosphorylated JNK. A prominent immune signal for phosphorylated JNK was observed that did not change as a function of arsenic trioxide treatment (Fig. 4, middle). A modification of STAT function through enhanced expression of PIAS3 by arsenic trioxide was ruled out by identifying a constant level of PIAS3 protein on immunoblots (Fig. 4, bottom). These data suggest, at least, relative specificity of arsenic trioxide for the STAT pathway.

It is conceivable that arsenic trioxide achieved reduced phosphorylation of STATs by enhancing the activities of protein phosphatases. To test this, HEL cells were treated for 6 hours with pervanadate to achieve a broad range inhibition of protein phosphatases. In such cells, the level of tyrosine phosphorylated STATs was moderately increased. Concomitant treatment with arsenic trioxide, still resulted in reduced phosphorylation of STATs (data not shown), suggesting that arsenic trioxide did not exert its inhibitory effect through vanadate-sensitive phosphatase activity.

Arsenic trioxide modulates PTKs, including ZNF198/FGFR1, leading to decreased phosphorylation of STATs. The attenuating effect of arsenic trioxide on STATs (Figs. 1 and 3) has been interpreted to mean that arsenic trioxide alters the activities of one or more PTKs that are able to phosphorylate STAT. The proposed causal relationship between arsenic trioxide–sensitive PTK action and STAT phosphorylation could not be defined in primary AML cells because we do not know the identity of the PTKs that are responsible for the constitutive activation of STATs. Therefore, we chose to test the kinase inhibitory effects of arsenic trioxide on several defined oncogenic PTKs in cell culture models. The first model consisted of cells expressing ZNF198/FGFR1, a novel PTK cloned by us (26). This is a 150-kDa protein resulting from fusion between the zinc-finger protein ZNF198 on chromosome 13q12 and FGFR1 on chromosome 8p11. The translocation has been found to be the sole genomic alteration that leads to a myeloproliferative disorder that can progress to either B- or T-lineage acute lymphoblastic leukemia. The resulting chimeric gene product exhibits constitutive kinase activity of the FGFR1 cytoplasmic domain, and this activity, when expressed in cells, confers a continuous stimulation of the STAT pathway (Fig. 5A ; ref. 11). Remarkably, arsenic trioxide treatment of cells expressing ZNF198/FGFR1 altered the phosphorylation of the kinase, as shown by increased molecular size and increased reaction with anti-phosphotyrosine, although the total amount of the fusion kinase, as detected by anti-FGFR1 immunoblotting, decreased. During arsenic trioxide treatment, the kinase changed its cytoplasmic localization from predominant concentration in aggregates to a more uniform distribution (Fig. 5B). Although the phosphorylation of the kinase was elevated, the kinase activity decreased, resulting in an arsenic trioxide dose-dependent reduction of STAT3 protein phosphorylation (Fig. 5A) but retaining constant total STAT3 protein level. Finally, the arsenic trioxide–dependent inhibition of protein tyrosine phosphorylation by FGFR1 kinase could also be shown with a second oncogenic fusion protein that combined BCR with FGFR1 (Fig. 5C). However, the arsenic trioxide–induced decrease of the electrophoretic mobility noted for the ZNF198/FGFR1 was not observed with the BCR/FGFR1 chimera or any other PTKs that were tested.

The modulatory action of arsenic trioxide on PTKs could also be confirmed for the oncogenic BCR/ABL (Fig. 5C), FLT3 (Fig. 1B), v-SRC, and FMS (data not shown) and even the normal forms of JAK1 and JAK2 (Fig. 5D). The modulating effect of arsenic trioxide on JAKs suggested that arsenic trioxide could attenuate the STAT-dependent signaling reactions of cytokine receptors that involve JAK1 or JAK2. This was shown by a decreased transcription of a STAT3-responsive gene in IL-6-treated cells (Fig. 6, left ). Moreover, the arsenic trioxide–dependent inhibition of the constitutively active oncogenic kinases similarly lowered the transcriptional stimulation of the STAT-responsive reporter gene (Fig. 6, right). The time-dependent reduction of STAT activity correlated with an apparent loss of protein in the case of PML and BCR/ABL (Fig. 7A ) but, interestingly, not for BCR/FGFR1 and the JAKs. In the latter two cases, the inhibition of the enzymatic activity of the kinases seems sufficient to explain the reduction in STAT signaling (Fig. 5C and D; data not shown).

PML is known to undergo degradation by sumolation following exposure to arsenic trioxide (27). We therefore compared the effect of arsenic trioxide on PML to its effect on phosphorylated ZNF198/FGFR1, STAT3, and BCR/ABL over time (Fig. 7A). In the first 15 to 30 minutes of incubation with arsenic trioxide, PML underwent close to complete degradation. In contrast, the loss of the BCR/ABL and the molecular weight changes of ZNF198/FGFR1 occurred with much slower kinetics. This suggests that arsenic trioxide's mechanism of action is target dependent.

Finally, to determine whether arsenic trioxide's effect is kinase selective, such as through a direct chemical reaction of arsenic trioxide with the kinase protein or whether it depends on the intracellular environment, we tested arsenic trioxide's effect on ZNF198/FGFR1 in cell-free conditions. As shown in Fig. 7B, arsenic trioxide, even at high concentrations, did not modify the electrophoretic mobility of ZNF198/FGFR1 protein when incubated in cell-free extracts, suggesting that arsenic trioxide exerts its effect through cellular mediators.

Taken together, these data support the conclusion that arsenic trioxide acts as a broad-spectrum PTK inhibitor. The mechanism of action does not seem to include phosphatase activation but involves altered PTK activities, changes in protein modification, inactivation, and decreased expression of the PTK protein.

Discussion

We have shown that arsenic trioxide inhibits PTKs and thereby indirectly decreases activation of STAT proteins. Arsenic trioxide caused a change in the molecular weight of ZNF198/FGFR1, altered its intracellular distribution, and induced progressive loss of the protein. Arsenic trioxide modulation of ZNF198/FGFR1 was reproduced for hematopoietic oncogenic (FLT3 and BCR/ABL) and non-oncogenic (JAK1 and JAK2) PTKs. To the best of our knowledge, such information has not been previously reported. Moreover, our model provides measurable readouts for arsenic trioxide effects.

Arsenic trioxide has been found to induce growth inhibition through mitotic arrest before apoptosis in cancer cells (28, 29). The cells are arrested in the early phases of mitosis prophase, premetaphase, and metaphase (28); this effect is distinct from anti-tubulin agents, such as paclitaxel (29). It is not clear why in HEL cells we detected more necrosis than apoptosis.

Arsenic trioxide inhibition of STAT3, preceding inhibition of cellular proliferation, has been described by Hayashi et al. (8) in multiple myeloma cells. In that system, STAT3 activity was induced by IL-6. Although Schuringa et al. (30) suggested that constitutive STAT3 activity in AML blasts resulted from autocrine secretion of IL-6, IL-6 is known to be inhibitory for AML proliferation (31). Therefore, the role of IL-6-induced STAT3 activation in leukemogenesis remains unclear. Other growth factors, including thrombopoietin and granulocyte colony-stimulating factor, are also known to activate STAT3 (32, 33). In that regard, in view of our demonstration of modulation of JAK1 and JAK2 activity by arsenic trioxide, it is likely that arsenic trioxide will affect resident PTKs upstream of the STATs, if induced by a ligand in AML.

Cellular transformation triggered by various tyrosine kinase oncoproteins, such as v-SRC and LCK, has also been associated with constitutive activation of STAT3 (34, 35), but involvement of these oncoproteins has not been shown in AML and cannot explain the high proportion of cases with constitutive STAT3 activity in this disease. Other, yet unidentified, oncogenic PTKs may be involved in inducing constitutive STAT3 activity in AML. In that regard, ZNF198/FGFR1 offered a model for understanding the mechanism of action of arsenic trioxide on oncogenic PTKs.

STAT5 activation by oncogenic tyrosine kinases has been described in BCR/ABL–expressing (36, 37) and FLT3-ITD–expressing (38, 39) cells. Targeted down-regulation of these two oncogenic proteins resulted in down-regulation of STAT5 [BCR/ABL (40) and FLT3-ITD (20, 41)]. Therefore, the effect of arsenic trioxide on these two oncogenic tyrosine kinases and their downstream pathways (e.g., STAT5) lend credence to our hypothesis that arsenic trioxide's effect on STAT proteins is attributed to its effect on PTKs.

Several mechanisms are proposed to explain the increased molecular weight of ZNF198/FGFR1 following exposure to arsenic trioxide. Reactive oxygen species generated in response to arsenic trioxide lead to accumulation of hydrogen peroxide (42). These will, in turn, affect sulfhydryl groups, which are known to be involved in refolding of proteins (43). However, treatment of cells with redox reactions, including H₂O₂, UV radiation, and photodynamic therapy, were unable to reproduce the modification. This suggests an arsenic trioxide-specific event. Sumolation, or covalent binding of SUMO-1, a ubiquitin-like polypeptide, to its target is one possible mechanism to explain the increased molecular weight of ZNF198/FGFR1 following exposure to arsenic trioxide. This is supported by our previous findings that ZNF198/FGFR1 binds to another ubiquitin-like protein HHR6 (44). Arsenic trioxide has been shown to induce binding of SUMO-1 to PML (27) and IκB (45), leading to changes in conformation. Of interest is that both PML and ZNF198 contain zinc finger-like domains, and that they both undergo changes in response to arsenic trioxide within 15 to 30 minutes.

Redistribution of proteins in response to arsenic trioxide has been well recognized previously. Arsenite (NaAsO₂) has been shown to induce aggregation of THY-1, a glycoprotein located on the membrane of early progenitor cells (46). This effect required interactions with sulfhydryl groups (46). Furthermore, in acute promyelocytic leukemia cells, arsenic trioxide was shown to target PML proteins into nuclear bodies, leading to their degradation (47). This effect is believed to result from sumolation of PML (47). Our demonstration of redistribution of ZNF198/FGFR1 from concentrated cytoplasmic aggregates to a more uniform pattern upon exposure to arsenic trioxide is therefore in line with the published literature. Additional work will determine whether this effect of arsenic trioxide is dependent on sulfhydryl groups or on other mechanisms.

Arsenic trioxide was shown to induce degradation of other proteins, in addition to PML/retinoic acid receptor α (8, 42) (e.g., AKT; named after the acute transforming retrovirus) protein in the absence of an effect on phosphorylation (8). More difficult to explain is why some proteins are degraded (ZNF198/FGFR1 and BCR/ABL), whereas others (BCR/FGFR1, FLT3, JAK1, JAK2, and STATs) are not. One possible explanation may be the half-life of the proteins. For example, the half-life of BCR/ABL was estimated to be ∼48 hours (48), and at least 48 hours were required to notice a change in its phosphorylation. Others have shown similar results and the mechanism to be arsenic trioxide–dependent reduction of mRNA translation (49). Therefore, at this point, it is unclear whether arsenic trioxide primarily reduces synthesis of the kinases or enhances their degradation.

In summary, our work supports previous findings that arsenic trioxide induces death of non–acute promyelocytic leukemia AML cells (5, 6) and partially elucidates its mechanism of action. However, arsenic trioxide given as a single agent did not significantly improve the outcome of patients with relapsed or refractory AML (50). We therefore propose phase I clinical trials of arsenic trioxide to study the in vivo arsenic trioxide effects on the signaling pathways.