Retinoic Acid and Arsenic Synergize to Eradicate Leukemic Cells in a Mouse Model of Acute Promyelocytic Leukemia

Leukemic cells were isolated from bone marrow and spleen of leukemic hMRP8-PML/RARα transgenic mice (leukemia 935) as described (10), by flushing RPMI medium through long bones and collecting exudates from spleen. In vitro, spleen cells were cultured in RPMI medium supplemented with 10% FCS and 2% pockweed mitogen spleen–conditioned medium and were left untreated or were treated with 1 μM RA, 1 μM As₂O₃ (Sigma Chemical Co.), or both.

Leukemias were propagated by injecting blasts (10⁷ viable hematopoietic cells) into the tail vein of 6–7-wk-old syngenic FVB-NICO mice. Animal handling was done according to the guidelines of institutional animal care committees. Mice implanted with leukemic cells were randomly assigned to either type of treatment. RA was administrated to leukemic mice by subcutaneous implantation of a 21-d release pellet containing 10 mg ATRA (Innovative Research of America). A stock solution of 330 mM As₂O₃ was prepared by diluting the powder in 1 M NaOH, then a dilution in Tris-buffered saline (TBS) was administered by daily intraperitoneal injection at the concentration of 5 μg/g mice. Control mice were treated with placebo pellets or intraperitoneal injections of TBS.

Specimens of spleen, liver, and lung were cut into three parts and immediately processed for snap freezing in liquid nitrogen or fixations. Specimens of long bones were fixed in formaldehyde, decalcified in 10% nitric acid, and further processed for paraffin embedding. Spleen, liver, and lung were either fixed in alcohol-formaldehyde-acetic acid reagent (AFA; Carlo Erba Laboratories), paraffin embedded and stained with hematoxylin-eosin and May-Grünwald-Giemsa, or fixed in 2.5% glutaraldehyde in cacodylate buffer and epon embedding for electron microscopic examination. The extent of the leukemic infiltrate was assessed on paraffin sections. The differentiation of the leukemic cells was assessed by combining cytological and histological stains, immunofluorescent staining of cryocut sections with a rat anti–mouse CD11b antibody (PharMingen) and electron microscopic analysis. In situ cell death was studied by morphological analysis on paraffin sections, electron microscopic grids, and by terminal deoxynucleotidyltransferase– mediated dUTP nick end labeling (TUNEL) assays (reagents from Boehringer Mannheim), both on paraffin and cryocut sections.

hMPR8-PML/RARα transgenic mice develop transplantable leukemias which differentiate both in vivo and in vitro upon RA exposure (10). To test their sensitivity to arsenic in vitro, leukemic cells were isolated from spleen or bone marrow of moribund animals and cultured in the presence or absence of arsenic. Little apoptosis and no differentiation were observed by TUNEL or cytological examinations. Conceivably, growth factors present in conditioned media may block apoptosis, as demonstrated in other cellular settings. However, both arsenic and RA induced PML/RARα degradation (data not shown), as shown previously in APL cell lines (25, 26), confirming that degradation of the fusion protein does not suffice to trigger arsenic-induced apoptosis (30).

Syngenic FVB mice were then injected with 10⁷ leukemic cells. Transplantation was always successful, as all animals died with an intraexperimental variation of <1 wk, generally in 30–50 d. In dose–response experiments, mice were treated for 1 mo with daily injections of arsenic or TBS 4 d after leukemia engraftment. Although 1 μg/g body wt arsenic daily yielded no tumor regression upon killing, 10 μg/g led to many early deaths, presumably of toxic origin (pathological examination revealed some hepatic toxicity and widespread pulmonary edema). However, with 5 μg/g, arsenic-treated animals showed greatly reduced leukemic infiltrate of the organs analyzed. As nontransplanted mice treated with this same dose for the same length of time also showed no evidence for toxicity, a daily dose of 5 μg/g was used thereafter. Despite the much higher doses used in mice compared with humans, the circulating arsenic levels were in the range of those present in arsenic-treated APL patients (31; data not shown). In pilot survival experiments where mice were treated 4 d after transplantation for 38 d, the 10 arsenic-treated mice lived significantly longer than the 10 controls (mean: 124 ± 6 vs. 50 ± 4 d). Altogether, our results demonstrate that leukemic cells from PML/ RARα transgenic mice are arsenic sensitive in vivo.

We have previously shown in cell lines that arsenic and RA appear to synergize for both differentiation and apoptosis (30), although this has been disputed (22, 28). To test the possible synergy between these two agents in vivo, we evaluated their effects on the regression of established leukemias. Hence, for this set of experiments, leukemias were allowed to develop for 20–25 d before therapy. Leukemic mice were then randomly assigned to treatment with arsenic, RA, both, or vehicle for 4 or 8 d and killed (two mice per treatment and time point). In three different experiments, RA or arsenic treatments reduced spleen weight and liver infiltration, whereas their association completely normalized the macroscopic appearance of these organs (not shown).

Microscopic examination of hematoxylin-eosin–stained sections of bone marrow, spleen, and liver from these animals confirmed this observation. In the absence of therapy, massive leukemic infiltration was evident in all three organs. In particular, the bone marrow was strictly monomorphic, consisting of promyelocyte-like cells that retained immature features such as basophilic cytoplasm (Fig. 1 A).

As reported previously, RA caused the rapid differentiation of leukemic cells into polymorphonuclear leukocytes. In the bone marrow, 4 d of RA treatment induced a drastic reduction of the cellular density with reappearance of some adipocytes (Fig. 1 B, and data not shown). Nevertheless, the marrow remained monomorphic, almost exclusively composed of polymorphonuclear cells (arrowheads, Fig. 1 B). After 8 d of RA, normal hematopoiesis was restored, with a large number of erythroblasts and a decrease in granulocytes compared with nonleukemic bone marrow (Fig. 1 A). In the liver of animals treated with RA for 4 d, small remaining tumor masses consisting of maturating myeloid cells were found around vessels of the portal tracts or centrilobular veins (see arrows in Fig. 2 and Fig. 5 A). Leukemic infiltration of the parenchyme was dramatically reduced at day 8 (not shown). The spleen contained a large number of granulocytes at both 4 and 8 d, but the leukemic infiltrate rapidly diminished (not shown). These observations confirm previous analyses of these animals (10).

4 d after arsenic treatment, some cells with a condensed nucleus have apoptotic-like features (arrows, Fig. 1 B), while partly differentiated cells with indented nuclei were also observed (arrowheads, Fig. 1 B). At 8 d, the bone marrow remained quite monomorphic, consisting of myeloid cells with an altered chromatin clearly distinct from that of untreated blasts (Fig. 1 A). In the liver of untreated animals, leukemic blasts infiltrate the parenchyme as very large perivascular masses associated with smaller aggregates of leukemic cells that obstructed sinusoids (Fig. 2, and see Fig. 5 A). In the leukemic blasts from the small intrasinusoid aggregates, arsenic induced morphological changes such as the appearance of indented nuclei and apoptosis-like nuclear condensation (arrowheads and arrows, Fig. 2). As a result of arsenic therapy, only large perivascular masses consisting of differentiated/apoptotic cells remain after the first week (not shown). Nevertheless, at both time points, the reduction in tumor mass was less drastic than that observed with RA.

Treatment with both RA and arsenic led to a much faster decrease in the leukemic population. In the marrow, islets of normal erythroblasts were already clearly visible 4 d after treatment, which was not the case with the single agent treatments (arrows, Fig. 1 B). After 8 d, the bone marrow was normal, with abundant erythroblasts and megakaryocytes (Fig. 1 A). Interestingly, we found numerous activated phagocytes with internalized granulocytes, which could account for the relative deficit in granulocytes compared with nonleukemic marrow. 4 d after treatment, the liver presented only very small remaining aggregates of leukemic cells around large vessels (Fig. 2, and see Fig. 5 A). At 8 d, both liver and spleen appeared tumor-free (not shown).

Ultrastructural analysis of liver sections was undertaken to analyze the morphology of leukemic cells after 4 d of therapy (Fig. 3). In livers of untreated leukemic animals, blasts (with lobulated nuclei and dense cytoplasm with some granulations) were clearly visible among hepatocytes and endothelial cells. Upon RA treatment, differentiating myeloid cells resembling granulocytes with fragmented nuclei and dense chromatin were found in the vascular space. Interestingly, arsenic treatment led to the appearance of many cell remnants, often consisting of naked nuclei, or with profound cytoplasmic alterations including large vacuoles and disrupted plasma membrane. However, the chromatin appeared moderately condensed at the nuclear periphery. The nuclear indentations and the presence of cytoplasmic granulations are strongly suggestive for the leukemic origin of these cells. We have recently demonstrated that PML triggers a caspase-independent cell death (7). The aspects of arsenic-treated APL blast unraveled here (Fig. 3) are highly reminiscent of PML-induced death, consistent with the idea that one of the effects of arsenic is to trigger PML- mediated death. Dual-treated specimens harbored very few hematopoietic cells, but on some occasions, images of apoptotic granulocyte phagocytosis were observed (Fig. 3). Altogether, these analyses confirm that RA induces differentiation whereas arsenic triggers a cell death process not associated with major nuclear alterations.

To quantify differentiation and apoptosis, sections were stained with CD11b for assessment of differentiation (22, 29, 30) and a TUNEL assay was used for assessment of apoptosis. Either RA, arsenic, or both treatments sharply induced CD11b expression in the infiltrated liver at day 4 (Fig. 4), as shown previously for APL blasts in patients (29). In the liver, a basal level of TUNEL positivity was noted in the leukemic cells of untreated mice (Fig. 5 A), consistent with high rates of spontaneous apoptosis of tumor cells in vivo. Arsenic sharply enhanced TUNEL positivity, particularly in the small leukemic aggregates in the liver sinusoids (Fig. 5 A). With RA treatment, intense TUNEL positivity was found in the red pulp of spleen, whereas liver was completely negative, suggesting that RA triggered the migration of differentiated leukemic cells to the spleen where they underwent apoptosis. Double RA/arsenic therapy led to an even more dramatic enhancement of TUNEL positivity in the spleen (Fig. 5 B), suggestive of accelerated differentiation and migration to this site.

To see whether RA and arsenic also influenced survival, 20 mice were transplanted, allowed to engraft for 12 d, and were then left untreated or were treated with arsenic, RA, or both until the first mouse in the control group died (40 d). Hence, mice were treated for 28 d, and survival was monitored. After arsenic therapy, all animals eventually died within a narrow time range (80 d; Fig. 6 A), as reported above with a shorter implantation time before treatment. In the case of RA therapy, relapses were more scattered but all animals died between 78 and 220 d after transplantation. In striking contrast, all double-treated animals were alive 9 mo after transplantation. The log–rank test demonstrates that differences between the survival of these four groups are highly statistically significant (P = 0.0001). Moreover, dual RA and arsenic therapy was significantly better than RA alone (P = 0.002). These observations are consistent with the synergistic effects of RA and arsenic on tumor regression.

To know whether the double treatment had actually eradicated the leukemia, surviving animals were killed at day 280 after transplantation. Microscopic examination of the bone marrow and spleen showed no leukemic infiltrate (not shown). The presence of leukemic cells was molecularly assessed by PCR amplification of the leukemia- specific PML/RARα fusion gene. In splenic DNA from all four mice tested, no amplification products were found with a nested PCR assay that detects 1 leukemic cell in 1,000–10,000 cells (32; data not shown), whereas the mouse p13 gene was amplified in all four cases. Thus, after dual RA and arsenic therapy, leukemic cells have become undetectable.

This report presents evidence that two drugs that specifically target the PML/RARα fusion protein in APL cooperate in vivo to induce tumor regression and dramatically prolong survival. This model offers the advantage that it closely mimics the APL situation: a population of malignant cells is present in an immunocompetent organism, and only this population is PML/RARα positive, in contrast to transgenic animals where all myeloid cells express the fusion protein. The behavior of the leukemic cells versus the nontransformed hematopoiesis is much better assessed in this setting, and immune response against the leukemia can occur.

Despite previous claims (28), it seems logical that these two drugs which target an oncogene for degradation through distinct pathways cooperate rather than antagonize, confirming our previous findings in vitro (30). A double dominant-negative model was proposed to explain APL pathogenesis, whereby PML/RARα blocks the functions of the normal RARα (differentiation) and the normal PML (apoptosis) proteins (20). Apart from inducing PML/RARα degradation, RA transcriptionally activates RAR, promoting differentiation. In addition, RA induces RARα degradation (30; our unpublished observations). Similarly, arsenic induces PML/RARα degradation. Arsenic also targets PML onto NBs, enhancing its proapoptotic properties (7) and subsequently promoting PML degradation (25). Hence, in this double dominant-negative model, PML/RARα degradation by one agent should favor the action of the other and vice versa (Fig. 6 B). Our results, both in vitro and in vivo showing enhanced differentiation and apoptosis with dual treatments, are consistent with this model. Nevertheless, it is also possible that arsenic modifies the function of RARα, as it enhances RARα phosphorylation (25) (which was recently shown to modify its function [33, 34]) and induces RARα catabolism (30). Together with PML/RARα degradation, arsenic's effects on RARα could account for the moderate differentiation induced by this agent. Moreover, the most striking synergy in the double treatments concerns differentiation, suggesting that arsenic enhances RA's effects more than the reverse.

Some toxicity occurred, but under our conditions it was acceptable and never led to deaths. Arsenic alone was hepatotoxic as assessed by moderate edema and steatosis, whereas dual treatment induced some hepatocyte apoptosis suggested by dense rims of nuclear heterochromatin and nuclear condensation on electron micrographs (not shown; see also arrows, Fig. 2). Some endothelial toxicity was also noted with dual treatment. However, the absence of major toxicity in a pilot case of dual treatment in a relapse APL patient (Dombret, H., and L. Degos, personal communication) suggests that toxicity is unlikely to limit the association of these two drugs.

In our experimental model, mice relapse quickly after single treatment discontinuation. One obvious possibility is that our treatments were too short. Alternatively, the therapeutic route (subcutaneous for RA, intraperitoneal for arsenic), different from that used in patients (oral for RA, intravenous for arsenic), may not have been optimal. Nevertheless, in human APL, resistance to RA or arsenic as single agents is quite rapid (31, 35, 36). In addition, rate of spontaneous resistance to RA or arsenic of APL cell lines is also high (30, 37). Such high intrinsic resistance of APL cells to these agents could account for the high incidence of relapses with single agent therapy. Here, the apparent eradication of the leukemic clone may reflect the direct differentiating/ proapoptotic properties of these two agents. Alternatively, the small number of cells resistant to both RA and arsenic may be eradicated by NK cell activity or by an immune response against the graft. In that sense, the necrotic-like death of arsenic-treated APL cells (Fig. 3) could induce an antileukemia immune response, as proposed in another setting (38).

To our knowledge, these studies represent the first example of clinical trials in a mouse model derived from a transgenic system of a human leukemia. Current protocols use induction therapies based on the simultaneous or sequential use of RA and chemotherapy (39). To date, arsenic is used as a single agent, principally in relapse APL patients (31, 35). The dramatic synergy between these two agents has obvious therapeutic indications: eradication of the leukemic clone in dual-treated animals clearly favors the use of arsenic as a first line drug, suggesting that combined therapies should be assessed in APL patients.

APL

acute promyelocytic leukemia

arsenic

arsenic trioxide

NB

nuclear body

PML

promyelocytic leukemia

RA

retinoic acid

TUNEL

terminal deoxynucleotidyltransferase– mediated dUTP nick end labeling