|Home | About | Journals | Submit | Contact Us | Français|
A key oncogenic force in acute promyelocytic leukemia (APL) is the ability of the promyelocytic leukemia–retinoic acid receptor α (PML-RARA) oncoprotein to recruit transcriptional repressors and DNA methyltransferases at retinoic acid–responsive elements. Pharmacological doses of retinoic acid relieve transcriptional repression inducing terminal differentiation/apoptosis of the leukemic blasts. APL blasts often harbor additional recurrent chromosomal abnormalities, and significantly, APL prevalence is increased in Latino populations. These observations suggest that multiple genetic and environmental/dietary factors are likely implicated in APL. We tested whether dietary or targeted chemopreventive strategies relieving PML-RARA transcriptional repression would be effective in a transgenic mouse model. Surprisingly, we found that 1) treatment with a demethylating agent, 5-azacytidine, results in a striking acceleration of APL; 2) a high fat, low folate/choline–containing diet resulted in a substantial but nonsignificant APL acceleration; and 3) all-trans retinoic acid (ATRA) is ineffective in preventing leukemia and results in ATRA-resistant APL. Our findings have important clinical implications because ATRA is a drug of choice for APL treatment and indicate that global demethylation, whether through dietary manipulations or through the use of a pharmacologic agent such as 5-azacytidine, may have unintended and detrimental consequences in chemopreventive regimens.
Cancer is driven by an initiating event that leads to the progressive accumulation of genetic mutations and of epigenetic changes that deregulate critical cancer genes.1,2 The retinoic acid receptor α (RARA) gene fuses to the promyelocytic leukemia (PML) gene in the vast majority of acute promyelocytic leukemia (APL) cases.3 The PML-RARA fusion protein initiates APL when expressed in the early myeloid compartment.4-6 The PML-RARA fusion protein not only forms aberrant complexes with histone deacetylases (HDAC) and DNA methyltransferases (DNMT) interfering with RARA transcription at multiple levels but also acts as a PML dominant-negative mutant.7-11 In addition, APL blasts often harbor recurrent chromosomal abnormalities, strongly suggesting that multiple genetic events cooperate over time to give rise to overt leukemia.12,13
Mouse models of APL faithfully recapitulate this chain of genetic and epigenetic events. For instance, PML-RARA transgenic mice succumb to APL after a long latency period, suggesting that in addition to PML-RARA, several genetic events including recurrent karyological abnormalities are required for full-blown APL.14,15
Epidemiological data point to a high rate of APL among Latino populations, which may suggest a genetic predisposition to and/or exposure to distinct dietary/environmental factor(s).16 In this regard, it is noteworthy that numerous epidemiological and animal studies have demonstrated that a high-fat, low-folate/choline diet (i.e., “Western-style diet”) is a cancer risk factor.17,18 However, to date, little is known on the direct molecular mechanisms by which environmental, dietary, or chemopreventive strategies may favor or antagonize leukemogenesis.
The experimental evidence indicating that PML-RARA transforms hematopoietic progenitors through the transcriptional repression of RARA target genes because of the ability of PML-RARA to recruit corepressors and HDAC onto RARA target genes led to the hypothesis that targeted agents that specifically relieve PML-RARA transcriptional repression would effectively chemoprevent APL. We reasoned that this strategy would be devoid of the toxic long-term effects of cytotoxic chemotherapy. Arsenic trioxide has striking anti-APL properties because of its ability to promote PML-RARA degradation. However, arsenic trioxide targets multiple cellular networks, and its long-term use poses significant hazards.3 Therefore, we tested the chemopreventive effects of targeted treatment aimed at relieving PML-RARA transcriptional repression in a well-established APL mouse model.6 We used all-trans retinoic acid (ATRA), the first drug of choice for APL both in induction and maintenance protocols. This drug is effective in APL because of the ability to target the RARA moiety of PML-RARA relieving transcriptional repression and converting PML-RARA into a transcriptional activator.3 We also used 5-azacytidine (5-AC), a DNA methylation inhibitor used for the treatment of myelodysplastic syndromes19 expected to interfere with aberrant DNA methylation induced by PML-RARA.
ATRA has been used with varying degrees of success in chemoprevention.20 However, the long-term effects of ATRA and 5-AC on APL pathogenesis are currently unknown. In fact, point mutations within the RARA ligand-binding domain of PML-RARA occur in APL cases with acquired resistance to ATRA,21,22 while long-term treatment with 5-AC might result in genomic instability and lymphomagenesis in mice and humans.23-25 Moreover, to begin addressing the contribution of diet to APL leukemogenesis, we tested the long-term effects of the Western diet (WD), a high fat, low folate/choline–containing diet that is known to induce hypomethylation.26 The key endpoint biomarker in these experiments was the occurrence of leukemia.
For these studies, we used a murine transgenic mouse model in which PML-RARA is expressed in myeloid cells under the MRP8 promoter. PML-RARA transgenic mice develop impaired myeloid differentiation early in life and APL, with a penetrance of 64% and a median latency of 8.5 months. APL that develops in these transgenic mice fully recapitulates the features of the human disease, including achieving remissions when treated with retinoic acid, which is the drug of choice in this disease. Moreover, these features are retained when preleukemic bone marrows are transplanted in recipient mice.6,15 Thus, this mouse model of APL has attractive features to test preclinical chemopreventive and therapeutic strategies.
We generated cohorts of mice transplanted with bone marrow cells obtained from preleukemic APL transgenic mice to perform our chemoprevention studies. Results are summarized in Figure 1. Body weights and blood counts were comparable and within the normal range for all treatment groups until leukemia occurred (data not shown). Nonleukemic deaths, unless stated otherwise, were due to acute events secondary to manipulations that are required for venipuncture. We noted no significant drug-related toxic events.
Significantly, we observed 9 APL cases in the 5-AC treatment group. The mean onset of 5.9 months posttransplantation was strikingly accelerated as compared to the 10 months of the PBS injection group (Fig. 1B). There were 7 nonleukemic deaths in this group, 2 of which were due to sarcoma. We have also observed a substantial but nonsignificant APL acceleration in the WD group when compared to the control diet group (P = 0.08) (Fig. 1C).
Surprisingly, we observed no difference in leukemia-free survival between the ATRA and the control diet groups (Fig. 1D). In addition, 2 of the leukemias that arose in this group were proven to be insensitive to ATRA treatment (data not shown). We found no significant morphological and phenotypic differences between the APL that arose in the various treatment groups (Fig. 1F--H;H; note increased percentage of doubly positive Gr1 and Mac1 cells in F-H as compared to E, and data not shown).
To identify at the molecular level the determinants of the APL acceleration in 5-AC–treated mice, we examined the RNA expression profiles of the APL cases that arose in 5-AC–treated mice and in their control group. Among the top 100 most differentially expressed genes, we found several cancer genes down-regulated in the leukemias that arose during treatment with 5-AC such as DMTF1 (a cyclin D–binding protein that exerts a growth-suppressive function), FOXO1A (a tumor suppressor gene), Bach2 (a growth suppressor), DUSP1 (a down-regulator of multiple MAPK signaling pathways and protein kinases that play a role in cell cycle regulation), and sestrin 1 (a transcriptional target of the tumor suppressor p53). Among the genes up-regulated in the leukemias that arose during 5-AC treatment were Birc1e (a IAP inhibitor that antagonizes apoptosis), Sept9 (a proposed growth suppressor), and angiopoietin-like 4 (a gene with prometastatic functions) (Online Mendelian Inheritance in Man: http://www.ncbi.nlm.nih.gov/entrez) (Table 1 and Suppl. Table S1). When analyzed with the DAVID Bioinformatics Resource and Ingenuity Systems Pathway Analysis tools, these genes cluster in functional classes involved in the regulation of transcription, adaptive immune responses, and apoptosis. Therefore, it is tempting to speculate that these processes are involved with the mechanism(s) causing APL acceleration in 5-AC–treated mice. However, the relevance of these gene alterations versus the effects of 5-AC on APL is yet to be established.
Our findings allow us to draw the following conclusions: 1) Chronic 5-AC treatment resulted in a striking acceleration of APL leukemogenesis. This result is surprising given the fact that the PML-RARA oncoprotein induces hypermethylation and transcriptional silencing at its target promoter sites and that this activity contributes to its leukemogenic potential.11 DNA hypomethylation has been associated with chromosomal instability, reactivation of transposable elements, loss of imprinting, and activation of proto-oncogenes.25 We speculate that 5-AC may cause genome-wide genetic and epigenetic effects that promote leukemogenesis. For example, recent reports show that genomic hypomethylation causes genomic instability and lymphomagenesis in mice.23,24 2) Chemoprevention with ATRA did not prevent APL leukemogenesis and instead results in ATRA-resistant APL, a known complication of ATRA therapy.21,22 3) WD induced a trend toward APL acceleration that, although did not reach statistical significance (P = 0.08), suggests that WD may represent a risk factor for APL leukemogenesis. Further, because the effects of WD appear to parallel those by 5-AC, we may infer that these effects are due, in large measure, to its enhanced hypomethylating potential.
Our findings have in turn important clinical implications and suggest that further investigations should be considered to address the long-term effects of ATRA and 5-AC in APL and myelodysplastic syndromes. The study of the function of the genes differentially expressed between the 5-AC–treated and control mice may provide insights into the mechanisms underlying 5-AC–induced leukemogenesis.
At 8 weeks of age, FVB/N mice underwent allogenic bone marrow transplant with 107 viable preleukemic hMRP8-PML-RARA marrow cells obtained from 8-week-old donors.6 After hematopoietic reconstitution (about 8 weeks), mice were randomly assigned to treatment with standard AIN-76A diet (control diet) alone, control diet with ATRA (30 µg/g of diet), or WD alone. The control diet standard provides 3.6 kcal/g, and its composition is as follows: fat (corn oil), 5%; calcium, 5 mg/g; vitamin D3, 1 IU/g; phosphorus, 4 mg/g; fiber, 5%; folic acid, 2 µg/g; methionine, 0.3%; choline bitartrate, 0.2%. The WD provides 4.5 kcal/g, and its composition is as follows: fat (corn oil), 20%; calcium, 0.5 mg/g; vitamin D3, 0.11 IU/g; phosphorus, 3.6 mg/g; fiber, 2%; folic acid, 0.23 µg/g; cysteine, 0.3%; choline bitartrate, 0.12%. Research Diet Inc. (Newark, NJ) formulated all diets. 5-AC (4-amino-1-β-D-ribofuranosyl-1,3,5-triazin-2(1H)-1), provided by the Chemopreventive Agent Development Research Group of the National Cancer Institute, was administered by weekly intraperitoneal injections at a concentration of 5 µg/g of body weight with PBS as a control. Each treatment arm consisted of 16 or 17 mice (Fig. 1A). Mice were evaluated clinically weekly and weighted biweekly. Hematopoiesis was monitored by tail vein venipuncture biweekly. Treatment continued until leukemia was diagnosed by venipunture, by signs of distress, or until mice turned 18 months old. Postmortem analysis included necroscopy, detailed pathological examination, and flow cytometry analysis. Animal handling was carried out according to the guidelines of the Memorial Sloan-Kettering Cancer Center institutional animal care committees.
Mice were inspected weekly and weighted biweekly. Treatment continued until leukemia was diagnosed, signs of distress occurred, or mice turned 18 months old. To diagnose leukemia, mice were bled twice a month from the tail to determine white blood cells, hemoglobin, and platelet counts. APL was diagnosed on the basis of the following concomitant criteria: 1) presence of blasts/promyelocytes (>1%) in peripheral blood; 2) leukocytosis (white blood cells >30 × 103/µL); and 3) anemia (hemoglobin <10 g/dL) and/or thrombocytopenia (platelets <500 × 103/µL).27 Diagnosis was confirmed by morphological and flow cytometric analyses of bone marrow cells with Mac-1 (CD11b), Gr-1, c-kit (CD117), Sca-1, B220, CD3, and Ter119 antibodies (BD Biosciences, Franklin Lakes, NJ).28 Animal handling was carried out according to the guidelines of the Memorial Sloan-Kettering Cancer Center institutional animal care committees.
Affymetrix U74Av2 arrays (Santa Clara, CA) were used to analyze leukemic samples obtained from the bone marrow or spleen of 4 mice treated with 5-AC and 4 that arose from vehicle control–treated mice, following procedures already described.29 The data were analyzed using the R statistical program and the Bioconductor microarray packages (Bioconductor Open Source Software for Bioinformatics). The raw data were quantitated using the GC-RMA algorithm to determine signal levels and the MAS5 algorithm to determine the absent/present flags. Genes that scored absent in more than 75% of both groups were filtered from further analysis. We used a variant of the standard t test (the LIMMA method) to determine the genes, which had the lowest P values for the test and had a fold change difference of at least 1.5. Pathway analysis was performed with the DAVID Bioinformatics Resource (http://david.abcc.ncifcrf.gov/home.jsp) and Ingenuity Systems pathway analysis.
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
This work was supported by the National Cancer Institute [grant number N01-CN-25122 (P.P.P.)]; the National Institutes of Health (NIH) [grant numbers R01 CA-71692 (P.P.P.), CA 112325-01 (P.P.S.), R01 CA 095274 (S.C.K.)]; the ASCO Young Investigator Award (P.P.S.); the Cancer and Leukemia Group B Oncology Fellows Award (P.P.S.); the Charles A. Dana Foundation (P.P.S.); the Michael and Ethel L. Cohen Foundation (P.P.S.); and the Steps for Breath Foundation (P.P.S.).