Medroxyprogesterone Acetate Elevation of Nm23-H1 Metastasis Suppressor Expression in Hormone Receptor Negative Breast Cancer

Transcription

1 ICLEARTICLES ART Medroxyprogesterone Acetate Elevation of Nm23-H1 Metastasis Suppressor Expression in Hormone Receptor Negative Breast Cancer Diane Palmieri, Douglas O. Halverson, Taoufi k Ouatas, Christine E. Horak, Massimiliano Salerno, Jennifer Johnson, W. Douglas Figg, Melinda Hollingshead, Stephen Hursting, David Berrigan, Seth M. Steinberg, Maria J. Merino, Patricia S. Steeg Background: Reestablishment of metastasis suppressor gene expression may constitute a therapeutic strategy for highrisk breast cancer patients. We previously showed that medroxyprogesterone acetate (MPA), a progestin that has been tested as treatment for advanced breast cancer, elevates expression of the Nm23-H1 metastasis suppressor gene in hormone receptor negative metastatic human breast carcinoma cell lines in vitro via a glucocorticoid receptor based mechanism. Here, we tested whether MPA treatment inhibits metastatic colonization of a hormone receptor negative breast cancer cell line in vivo. Methods: We tested the soft-agar colony-forming efficiency of untransfected MDA-MB-231T human breast carcinoma cells and MDA- MB-231T cells transfected with antisense Nm23-H1 in the presence and absence of MPA. Pharmacokinetic studies were used to establish dose and injection schedules that led to MPA serum levels in mice similar to those achievable in humans. For in vivo studies, nude mice were injected intravenously with MDA-MB-231T cells. After 4 weeks, mice were randomized to control or MPA arms. Endpoints included incidence, number, and size of gross pulmonary metastases; Nm23-H1 protein expression in gross metastases; and side effects. All statistical tests were two-sided. Results: MPA reduced colony formation of MDA-MB-231T cells by 40% 50% but had no effect on colony formation of Nm23- H1 antisense transfectants. Metastases developed in 100% (95% confidence interval [CI] = 78% to 100% and 77% to 100%, respectively) of control mice injected with MDA- MB-231T cells. In two independent experiments, only 73% (95% CI = 45% to 92%) and 64% (95% CI = 35% to 87%) of mice injected with 2 mg of MPA developed metastases. Mice injected with 2 mg of MPA showed reductions in the mean numbers, per mouse, of all metastases and of large (>3 mm) metastases ( P =.04 and.013, respectively). Nm23-H1 was expressed at high levels in 43% of pulmonary metastases in MPA-treated mice but only 13% of metastases in untreated mice. Mice receiving at least 1-mg doses of MPA gained more weight than control-treated mice but exhibited no bone density alterations or abnormal mammary fat pad histology. Conclusion: Our preclinical results show that MPA appears to elevate Nm23-H1 metastasis suppressor gene expression, thereby reducing metastatic colonization. The data suggest a new use for an old agent in a molecularly defined subset of breast cancer patients. [J Natl Cancer Inst 2005;97:632 42] Breast cancer patients at high risk of recurrence whose tumors are estrogen receptor (ER) negative and Her-2 unamplified have no molecularly targeted therapeutic options. One promising avenue of investigation involves targeting the metastatic process, specifically by increasing the expression of metastasis suppressor genes. These genes suppress in vivo metastatic capacity with no substantial effects on primary tumor size after reintroduction into metastatic tumor cell lines ( 1 3 ). Nm23 was the first metastasis suppressor gene identified (4,5 ). Subsequently, eight members of the Nm23 gene family (Nm23-H1 through -H8) have been reported. Although the eight genes share homology, the in vivo data show that Nm23-H1 is most likely to function as a metastasis suppressor gene ( 6 ). Eleven independent studies have confirmed that restoration of Nm23 expression reduces metastasis of breast, colon, prostate, oral squamous cell carcinoma, and melanoma tumor cell lines to the lymph nodes, lungs, and liver (5,7 16 ). In vitro correlates of elevated Nm23-H1 expression in breast carcinoma cell lines include reduced softagar colony formation, reduced cell motility, reduced invasion, and the induction of morphological and biochemical aspects of differentiation in three-dimensional culture (7,17 20 ). For both tumor suppressor genes and metastasis suppressor genes, when the gene for the wild-type protein is lost, a major Affi liations of authors: Women s Cancers Section, Laboratory of Pathology, Center for Cancer Research, National Cancer Institute, Bethesda, MD (DP, DOH, TO, CEH, MS, PSS); Laboratory Animal Sciences Program, SAIC, Frederick, MD (JJ); Cancer Therapeutics Branch, Center for Cancer Research, National Cancer Institute, Bethesda, MD (WDF); Developmental Therapeutics Program, National Cancer Institute, Bethesda, MD (MH); Laboratory of Biosystems and Cancer, Center for Cancer Research, National Cancer Institute, Bethesda, MD (SH, DB); Biostatistics and Data Management Section, Center for Cancer Research, National Cancer Institute, Bethesda, MD (SMS); Surgical Pathology Section, Laboratory of Pathology, Center for Cancer Research, National Cancer Institute, Bethesda, MD (MJM). Correspondence to: Patricia S.Steeg, PhD, National Institutes of Health, Bldg. 10, Rm. 2A33, Bethesda MD ( steegp@mail.nih.gov ). Present address: Taoufik Ouatas, Endotis Pharma, 70 rue du Docteur Yersin, Loos, France. See Notes following References. DOI: /jnci/dji111 Journal of the National Cancer Institute, Vol. 97, No. 9, Oxford University Press 2005, all rights reserved. 632 ARTICLES Journal of the National Cancer Institute, Vol. 97, No. 9, May 4, 2005

2 challenge lies in how to restore its suppressive function to a metastatic cancer cell. If the suppressive function of the protein can be identified and mapped to a single, specific domain of the protein, then screening methods may be able to identify a compound whose activity can replace that of the suppressor gene (21 ). However, many suppressor proteins have multiple biochemical functions and functional domains, all of which may contribute to the overall phenotype. Nm23-H1 is a case in point: This protein interacts with several different proteins, exerts DNA regulatory functions, and exhibits a histidine kinase activity, all of which may contribute to its suppression of metastasis. For suppressor proteins of this type, multiple specific replacement compounds would need to be identified. Although allelic deletion of Nm23-H1 does occur in human tumors, most poor-prognosis tumors are associated with reduced Nm23-H1 expression rather than with its allelic deletion or mutation, making reexpression of Nm23 in such tumors an attractive therapeutic option ( 22 ). We and others have identified compounds that stimulate metastatic tumor cells to reexpress Nm23-H1 in vitro ( ). In addition, we have found that the Nm23-H1 promoter contains a cassette of mammary-specific transcription factor binding sites regulated by glucocorticoid response elements that contribute to Nm23-H1 expression ( 28 ). Indeed, dexamethasone, prednisolone, and other glucocorticoids elevated Nm23-H1 expression in metastatic MDA-MB- 435 and -231 breast carcinoma cell lines in vitro in the presence of charcoal-stripped medium ( 29 ), which contains no endogenous corticosteroids. High-dose inhibition of Nm23-H1 was observed, suggesting that these compounds are relevant to physiologic but not pharmacologic elevation of Nm23-H1 expression ( 29 ). Another glucocorticoid, medroxyprogesterone acetate (MPA), has also been shown to elevate Nm23-H1 expression (29 ). MPA has a long clinical history. It is a progestin that is found at low doses in the contraceptive Depo-Provera (30 ) and is combined with estrogen in hormone replacement therapy. MPA has been tested at high doses as a single agent or in combination regimens as a hormonal treatment for advanced breast and uterine cancers [ reviewed in (31) ]. Although some patient responses were observed in clinical trials, an optimum dose and schedule were not determined. Two of the MPA clinical trials (32,40 ) have suggested that a longer course of MPA treatment would increase its clinical benefit in postmenopausal women. These two trials randomly assigned a total of 950 patients to chemotherapy (cytoxan, methotrexate, and 5-fluoruracil [CMF] in one trial and cytoxan, adriamycin, and 5-fluoruracil [CAF] in the other), with or without MPA. MPA was given for a 6-month time course after either CMF or CAF, compared with other trials, which limited MPA treatment to several weeks (31 ). Administration of MPA was divided into two phases, an induction phase of 500 mg daily for 28 days followed by a maintenance phase of 500 mg twice weekly for 6 months. After 12 and 13 years of follow-up, respectively, the MPA-treated postmenopausal subsets of patients had improved disease- and metastasis-free survival compared with patients who did not receive MPA (CAF, P =.01; CMF, P =.06 for node-negative and P =.002 for node-positive patients) and, for patients in the CAF ± MPA trial, longer overall survival (P =.02) (32,40 ). Paradoxically, patient responses were not well correlated with progesterone receptor (PR) expression (32 34 ). This lack of association made better sense years after the trials commenced with reports that MPA can interact not only with PR but also with the androgen receptor (35) and, as a glucocorticoid, with the glucocorticoid receptor (GR) the latter in both traditional (36) and nontraditional, i.e., DNA binding independent (37 39), pathways. MPA elevates Nm23-H1 expression of the MDA-MB- 435 and MDA-MB 231 metastatic breast carcinoma cell lines two- to fourfold in vitro in culture medium supplemented with 10% fetal bovine serum (29). Both cell lines are ER and PR negative, and the effect of MPA on Nm23-H1 expression involves a nontraditional, GR-dependent molecular pathway that is regulated posttranslationally (29). In this article, we characterized the effect of MPA on metastatic colonization of an ER-negative, PR-negative, GR-positive MDA-MB-231 breast carcinoma subline both in vitro (using soft-agar colony-forming assays) and in vivo (in a mouse model). One goal of the in vivo studies was to use MPA doses that are achievable in humans. To this end, we conducted a nine-arm pharmacokinetic study to determine an MPA dose and treatment schedule that would yield serum MPA levels in mice comparable to those attained in women in the clinical trials. For the in vivo metastasis experiments we adapted an in vivo experimental metastasis assay to focus on metastatic colonization the outgrowth of tumor cells from micrometastases to life-threatening lesions. Metastatic colonization is the last step in the metastatic process and has been hypothesized to represent the optimal therapeutic window of opportunity ( 1 ). For breast cancer patients with aggressive, lymph node positive disease, tumor cells may have already invaded past the primary tumor at time of diagnosis and surgery. Therefore, the outgrowth of tumor cells to a detectable size in a distant organ i.e., metastatic colonization and angiogenesis represent the only parts of the metastatic process that are known to be incomplete in lymph node positive patients. We compared Nm23-H1 protein expression in pulmonary metastases and the incidence, number, and size of pulmonary metastases between MPAtreated and untreated mice. M ATERIALS AND METHODS Materials MPA was purchased from Sigma (St. Louis, MO) and resuspended in 20% chloroform 80% ethanol at 100 mm and subsequently diluted in 100% ethanol to 1 mm for use in in vitro experiments. Clinical grade MPA (Depo-Provera; Pharmacia & Upjohn, Kalamazoo, MI) was diluted to 20 mg/ml in 8.6 mm polyethylene glycol (PEG 3350; Sigma) and 0.15 M NaCl (vehicle) for use in in vivo experiments. Cell Culture A subline of human MDA-MB-231 cells, which were designated MDA-MB-231T cells (gift of Dr. Zach Howard, Laboratory of Immunoregulation, NCI), was chosen for its reliable in vivo experimental metastatic potential. MDA-MB-231T cells and MDA-MB-231 cells (American Type Culture Collection, Manassas, VA) were genotyped by the Core Genotyping Facility, NCI. The Applied Biosystems Profiler Plus kit was used to polymerase chain reaction (PCR) amplify tetranucleotide short tandem repeat loci and the amelogenin locus. The amelogenin locus was used to identify sex. Sixteen loci were Journal of the National Cancer Institute, Vol. 97, No. 9, May 4, 2005 ARTICLES 633

3 genotyped to confirm the relationship between the different strains of the MDA-MB-231cell line. Four of these loci differed between MDA-MB-231T and MDA-MB-231 cells ( Table 1, Supplementary Data). Vials of MDA-MB-231T cells that were confirmed negative for murine antigen protein were stored in the NCI animal facility repository. Cells were cultured in Dulbecco s minimal essential medium containing 10% fetal bovine serum. Transfection with Antisense Nm23-H1 The cdna for human Nm23-H1 was inserted in reverse orientation into pcdna3.1 (Invitrogen, Carlsbad, CA) at the Bam H1 restriction site in the multiple cloning region. MDA- MB-231T cell clones stably expressing antisense Nm23-H1 were generated by transfection with Effectene (Qiagen, Valencia, CA) and selection in 1 mg/ml Geneticin (Invitrogen). Control cells were transfected with empty vector. Two individual stable clones of control and antisense Nm23-H1, respectively, were used for the experiments described herein. Western Blot Analysis Cells ( ) were plated in 100-mm 2 tissue culture dishes and incubated overnight. Cells were then incubated in MPA or ethanol (vehicle) for 72 hours. Cell lysates were obtained from cells treated with MPA or vehicle using cold RIPA buffer (20 mm Tris HCl [ph 8], 100 mm NaCl, 10% glycerol, 1% NP-40, 0.5% sodium deoxycholate, and 0.1% sodium dodecyl sulfate) containing complete mini EDTA-free protease inhibitor cocktail (Roche Molecular Biochemicals, Indianapolis, IN) and phosphatase inhibitor cocktails 1 and 2 (Sigma). Cellular lysates (30 mg total protein) were resolved by sodium dodecyl sulfate polyacrylamide gel electrophoresis and transferred to nitrocellulose membranes. Western blot analysis was performed for anti Nm23-H1 (1 : 500; Cymbus Chemicon, Temecula, CA), anti α-tubulin (1 : 5000; Ab-1, Oncogene, Cambridge, MA), anti- ER (1 : 300; clone EMR02, NovoCastra, Newcastle upon Tyne, UK), anti-pr (1 : 500; Upstate, Chicago, IL) and anti-gr (1 : 100; Oncogene) overnight at 4 C. Appropriate horseradish peroxidase conjugated secondary antibodies were used and detected by autoradiography with LumiGlo Reagents (Cell Signaling, Beverly, MA). Anchorage-independent Proliferation Assays Anchorage-independent colony-forming assays in soft agar were performed according to the previously published protocol (29 ). In brief, MDA-MB-231T cells were plated in a layer of 0.3% noble agar containing the indicated concentrations of MPA. This layer was overlaid on a layer of 0.7% noble agar in 1.88-cm 2 wells of a 24-well tissue culture plate. Cultures were grown at 37 C in 5% CO2 for 3 weeks. Total numbers of colonies (>50 cells) per well were counted using a light microscope; triplicate wells were done for each condition. Dose and Schedule Study Mice (n = 180) were randomized to groups of 20 that received MPA in nine different combinations of dose, schedule, and route of administration (i.e., intramuscular or subcutaneous). In all cases, mice received 4 weeks of induction MPA and 8 weeks of maintenance MPA. Induction dosing consisted of either biweekly or daily MPA injections for 4 weeks, and maintenance dosing consisted of one injection every fourth week thereafter for 8 weeks. Terminal bleeds were collected from five mice of each group on days 7, 21, 49, and 56 after receipt of the first MPA dose. Blood from each group of five mice was pooled and centrifuged to separate the serum. Pooled serum samples were stored at 70 C. Mammary fat pads were removed at necropsy from mice killed on days 7 and 56, preserved in Bouin s solution, fixed in formalin, and embedded in paraffin for immunohistochemistry. All in vivo experiments were performed in accordance with an approved animal use contract. Determination of Serum MPA Levels Fifty microliters of 0.1 mg/ml nortestosterone (internal standard) was added to 300 μ L of each serum sample. Hexane (3 ml) was then added, and the samples were vortexed and centrifuged for 5 minutes. The organic layer was collected and evaporated under liquid nitrogen. The residue was reconstituted in 200 μl of mobile phase, and 100 μ L was injected into a high-pressure liquid chromatography (HPLC) system. A C18 column (Phenomenex, Torrance, California) was used with a 75 : 25 : 0.03 methanol 0.1 M sodium phosphate (ph 3.5) triethylamine mobile phase at a flow rate of 1.2 ml/minute with a run time of Table 1. Pharmacokinetic analysis of medroxyprogesterone acetate (MPA) dose and schedule analysis * Experimental group Treatment Induction (4 wks) Dose (mg) Schedule 2W 2W 2W 2W 2W 2W QD 5 QD 5 QD 5 Route IM IM IM SC SC SC SC SC SC Maintenance (8 wks) Dose Schedule M M M M M M M M M Route IM IM IM SC SC SC SC SC SC Serum MPA (ng/ml) Day Day Day Day * Mice were divided into nine groups of 20 mice per group. 2W = twice weekly; QD 5 = 5 days per week; M = monthly; IM = intramuscular; SC = subcutaneous. Mean serum concentrations for n = 5 mice at each time point (from duplicate determinations). 634 ARTICLES Journal of the National Cancer Institute, Vol. 97, No. 9, May 4, 2005

4 12 minutes. The retention time of MPA was approximately 9.1 seconds. The intraday variation of the standard curve was 5.4%, and the interday variation was 2.9%. Metastasis Assay Two independent in vivo experiments were conducted. In Experiment 1, 6-week-old Nu/Nu mice (n = 100) were acclimated for 2 weeks and were then injected intravenously in the lateral tail vein with MDA-MB-231T cells. Four weeks later, 90 mice were randomized to one of three treatment groups (30 mice per group). Each group received subcutaneous injections of vehicle (see Materials and Methods) or 2 mg or 4 mg MPA twice weekly for 4 weeks (induction phase), followed by a maintenance phase of the same dose 4 weeks later. In Experiment 2, 6-weekold Nu/Nu mice (n = 160) were acclimated for 2 weeks and were then injected intravenously with MDA-MB-231T cells. Four weeks later, 150 mice were randomized to one of five treatment groups (30 mice per group). Four groups received subcutaneous injections of vehicle or MPA (0.5-mg, 1-mg, or 2-mg doses) twice weekly for 4 weeks (induction phase), followed by a maintenance phase of half the original dose every other week thereafter for 6 weeks. The fifth group was injected with 2 mg of MPA twice weekly for the entire 10-week treatment period. Throughout both experiments, mice were weighed weekly and monitored for signs of ill health and labored breathing. Mice were killed if pathologic conditions unrelated to the study (e.g., breathing difficulties) developed. Mice from each group were killed at various time points for collection of serum (four mice per group) and flash-frozen tissue (six mice per group). Additional mice not assigned to a treatment group were killed at various time points to determine whether gross pulmonary metastases had developed (10 mice). All mice alive at the end of each experiment were killed and necropsied to quantitate gross metastases. Lungs, skin, mammary fat pads, and liver were preserved in Bouin s solution. An investigator who was blinded to the experimental arm counted gross pulmonary metastases and measured them with a ruler while viewing through a magnifying glass. Immunohistochemistry Lung tissue obtained at necropsy from five mice per treatment group in Experiment 2 that had multiple metastases was fixed in formalin, embedded in paraffin, and sectioned (6- μm sections). One randomly chosen section from each of the five mice per treatment group was stained for Nm23 protein using an affinitypurified polyclonal rabbit anti Nm23-H1/H2 antibody (Cymbus). An isotype-matched control antibody (goat anti rabbit immunoglobulin G; Santa Cruz Biotechnology, Santa Cruz, CA) was used on a second, independent section to control for nonspecific binding. In brief, sections were deparaffinized in xylene, dehydrated in 100% ethanol, rehydrated in 95% ethanol and phosphate-buffered saline (PBS), and blocked sequentially with hydrogen peroxide and goat serum. Antigen retrieval was accomplished by incubating the slides in 10 m M citrate buffer, ph 3.0, for 30 minutes at 37 C. Slides were incubated with primary antibody (diluted 1 : 5 in PBS with 1% goat serum) in a humidified chamber overnight at room temperature. Staining was visualized using the Vectastain ABC kit and the DAB Substrate kit (Vector Laboratories, Burlingame, CA). Stained sections were examined under a microscope, and every visible metastasis in the tissue section was counted. Staining intensities were categorized as homogeneous low (i.e., staining of pulmonary metastases was equivalent in intensity to staining of the surrounding lung parenchyma); heterogeneous high (i.e., the pulmonary metastases contained some areas of staining higher in intensity than that in the parenchyma); and homogeneous high (i.e., staining in all areas of the pulmonary metastases was higher in intensity than that in the parenchyma). Body Content Analysis Fat weight, lean weight, bone mineral density (BMD), and bone mineral content (BMC) were determined for 10 mice per treatment arm from Experiment 2 using dual-energy X-ray absorptiometry (DXA) (GE Lunar Piximus II, Madison, Wisconsin). Necropsied carcasses were weighed and then placed on the specimen tray for scanning. Lean weight was calculated by subtracting fat weight from carcass weight. [Validation studies for estimates of body composition and bone characteristics using the GE Lunar Piximus Dual Energy X-ray Absorptiometer on mouse carcasses suggest that DXA is as effective for mice as it is for humans (41 ).] Histopathologic Analysis of the Mammary Fat Pad The fourth mammary fat pad of each animal in Experiment 2 was preserved at necropsy. A hematoxylin and eosin stained, formalin-fixed, paraffin-embedded section was examined by a pathologist who was blinded to experimental arm. Fat pads were examined for the presence of any pathologic conditions, i.e., hyperplasia, dysplasia, or carcinoma. Statistical Methods We used the Wilcoxon rank sum test to compare the number of total metastases and the number of metastases 3 mm or larger in any dimension in control mice with those values in mice treated with each dose of MPA. The 3-mm cutoff for large metastases was chosen arbitrarily after necropsy as a second measure of metastatic colonization. In addition, we used the Jonckheere Terpstra trend test to evaluate whether the total number of metastases or the number of metastases 3 mm or larger increased with increasing MPA dose. We used the Wilcoxon rank sum test to compare the presence of any metastases, the total number of metastases, and the number of large metastases between control mice and mice treated with any dose of MPA. The fraction of all mice with any metastases was also evaluated across all dose levels within each experiment using the exact Cochran Armitage test for trend as implemented in SAS Version 8 (SAS Institute, Cary, NC). The fraction of mice with large metastases, both among all mice and among mice with any metastases, was also evaluated using the exact Cochran Armitage trend test. Because both experiments contained a vehicle-treated control group as well as a group that received a 2-mg dose of MPA, the results for the 2-mg treatment groups from both experiments were compared within each of those two dose groups to determine if they were similar (P >.05), which would allow the data to be pooled to assess differences in metastatic colonization by dose. In the case of the number of metastases and number of large metastases, the Wilcoxon rank sum test was used. The fraction of any metastases that were 3 mm or larger was compared between experiments using the likelihood ratio test derived from logistic Journal of the National Cancer Institute, Vol. 97, No. 9, May 4, 2005 ARTICLES 635

5 regression modeling, with adjustment for multiple comparisons by the Hochberg method as needed (42 ). Finally, the pooled results (number of metastases and number of large metastases) from the 2-mg dose from both experiments were compared with the pooled results from the control mice from both experiments by the methods described above. All P values are two-tailed and, except as stated above, are presented with no formal adjustment for multiple comparison. In view of the number of comparisons performed, P values between.01 and.05 were interpreted as being of borderline statistical significance, and P <.01 was considered statistically significant. For the immunohistochemical analysis of Nm23 expression, the distribution of the number of metastases exhibiting homogeneous low, heterogeneous high, and homogeneous high Nm23 expression was analyzed in two ways. First, we compared the absolute number of metastases at each expression level between all mice in the control and each of the MPA dose groups using a Wilcoxon rank sum test. In addition, we created a severity index for each mouse within a group that was equal to: 1 (number of metastases with homogeneous low expression) + 2 (number of metastases with heterogeneous high expression) + 3 (number of metastases with homogeneous high expression). The distribution of the severity indices was compared within each group, and a Wilcoxon rank sum test was used to compare severity indices between groups of mice. The P values are two-tailed and presented without adjustment for multiple comparisons. However, given the number of comparisons performed, P values between.01 and.05 are interpreted as being of borderline statistical significance, whereas P <.01 was considered statistically significant. Finally, analysis of variance (ANOVA) and analysis of covariance were used to assess the effects of treatment on body composition and bone characteristics. ANOVA was used to determine body weight independent effects of treatments. Analysis of covariance allowed the inclusion of weight as a covariate in the analysis (43 ). R ESULTS Effect of MPA-stimulated Elevation of Nm23-H1 Expression on Anchorage-independent Colony Formation MPA was previously reported to elevate Nm23-H1 expression in MDA-MB-231 breast carcinoma cells by two- to fourfold in vitro and to inhibit soft-agar colony formation by approximately 50% (29 ). We obtained similar results using the MDA-MB-231T subline ( Fig. 1 ), which was chosen for its reliable in vivo experimental metastatic potential. The MDA-MB-231T cells were confirmed to be ER negative, PR negative, and GR positive Fig. 1. Effect of medroxyprogesterone acetate (MPA) elevation of Nm23-H1 expression on colony formation of MDA-MB-231T human breast carcinoma cells. (A ) Western blot analysis of Nm23-H1 expression in MPA- or vehicle-treated MDA-MB-231T cells. Cells were treated with the indicated concentration of MPA for 72 hours. Densitometric analysis of fold elevation shown below. (B ) Softagar colony formation of MDA-MB-231T cells cultured with MPA for 3 weeks, expressed as percentage of control colony formation. Mean + 95% confidence interval for a representative experiment of n = 3. **P =.0026 compared to vehicletreated control by analysis of variance (ANOVA). (C ) Western blot analysis of control (pcdna3) and Nm23-H1 antisense (AS)-expressing MDA-MB-231T clones. (D ) Western blot analysis of Nm23-H1 expression in MPA-treated (100 nm ) control (pcdna3) and Nm23-H1 AS-expressing MDA-MB-231T cells. (E ) Soft-agar colony formation of control and Nm23-H1 antisense expressing clones treated with MPA at the indicated concentrations and expressed as percentage of control colony formation. Open bars, pcdna3.1 clone (c.) 5; gray bars, pcdna3.1 c.105; solid bars, nm23-h1 AS c.15; hatched bars, Nm23-H1 AS c.223. Mean + 95% confidence interval for a representative experiment of n = 3. **P =.0131 compared with vehicle-treated control by ANOVA. 636 ARTICLES Journal of the National Cancer Institute, Vol. 97, No. 9, May 4, 2005

6 by western blot analysis, and treatment with MPA did not alter this phenotype (data not shown). Nm23-H1 expression in MDA- MB-231T cells increased by approximately threefold after 72 hours of culture in 100 nm MPA ( Fig. 1, A ). Additionally, soft agar colony formation was reduced by 40% 50% for MDA- MB-231T cells treated with MPA compared with untreated controls ( Fig. 1, B ). To examine the extent to which the increased Nm23-H1 expression accounts for the ability of MPA to suppress colonization, MDA-MB-231T cells were stably transfected with an antisense Nm23-H1 construct or an empty-vector control before treatment with MPA. Nm23-H1 expression was decreased three- to fourfold in the antisense transfectants as compared with control transfectants ( Fig. 1, C ). The antisense transfectants also did not undergo an increase in Nm23-H1 expression when treated with MPA ( Fig. 1, D ). Moreover, soft agar colony formation of antisense-expressing clones was not inhibited by MPA ( Fig. 1, E ). Thus, MPA appeared to affect anchorage-independent colony formation via its effect on Nm23-H1 expression. Establishment of an In Vivo MPA Regimen A goal of the metastasis studies reported here was to use MPA doses that result in serum MPA levels in mice similar to those achievable in humans. Pharmacokinetic data exist from the clinical trials using MPA as hormonal therapy for breast cancer. In one trial, plasma concentrations of MPA were determined by HPLC in 79 patients with advanced or recurrent breast cancer who were treated orally with mg of MPA per day (44 ). Mean serum MPA concentrations were 60 ng/ml for complete responders and partial responders, 34 ng/ml for patients with stable disease, and 23 ng/ml for patients with progressive disease (P <.001). Similarly, among 129 patients treated orally with mg of MPA per day, mean serum concentrations ranged from 46 ng/ml in patients with progressive disease to ng/ml in patients with complete or partial responses or stable disease (45 ). We carried out an extensive dose and schedule study to achieve a similar range of MPA plasma concentrations in mice. Although most clinical trials of MPA for breast cancer used oral or intramuscular dosing, we did not consider the oral route because of health concerns related to repeated oral gavage. We also did not supply the MPA in the mice s water because of concerns that analysis of water consumed would be open to artifacts from dripping bottles. The dose schedule study therefore compared intramuscular and subcutaneous injection routes. The induction phase maintenance phase regimen reported in two clinical trials that showed improved disease-free survival in postmenopausal women with breast cancer was adapted for use in these experiments (32,40 ). Fig. 2. Immunohistochemical analysis of Nm23-H1 expression in the mammary fat pad of MPA-treated mice. Representative image from 2-mg treatment group on day 7 after MPA treatment is shown. Original magnification, 400. Fig. 3. Experimental design for the inhibition of metastatic colonization. (A ) Outline of experimental protocol for in vivo studies. MDA-MB-231T breast carcinoma cells were injected into the lateral tail vein of mice, and pulmonary micrometastases were allowed to develop for 4 weeks. Mice were then randomized (in groups of 30) to a vehicle control or doses of medroxyprogesterone acetate (MPA), administered in a 4-week induction, 4 6 week maintenance regimen. In Experiment 1, the MPA doses were 2 mg and 4 mg; in Experiment 2, the MPA doses were 0.5 mg, 1 mg, and 2 mg. In Experiment 1, the maintenance dose was the same as the induction dose, given once every fourth week after the end of the induction phase. In Experiment 2, the maintenance phase was half the dose given during the induction phase, given every other week for 6 weeks. (B ) Hematoxylin and eosin stained sections of mouse lungs containing micrometastatic lesions at week 4 immediately before randomization and induction phase of treatment. Micrometastases are visible as collections of tumor cells in the lung parenchyma. Original magnification, 200. MPA plasma concentrations similar to those achieved in the human clinical trials were obtained by both intramuscular and subcutaneous dosing throughout the study period ( Table 1 ). Serum MPA concentrations in mice that received subcutaneous injections 5 days per week were generally higher than those achieved in human clinical trials, whereas the serum MPA concentrations obtained with the twice-weekly dosing were more reflective of those achieved in the clinical trials. Therefore, for Experiment 1 we used subcutaneous injections of 2 and 4 mg MPA twice weekly throughout the 4-week induction phase and every 4 weeks throughout the 8-week maintenance phase. With these two doses, differences in plasma MPA concentration were seen mainly during the maintenance period (i.e., on days 49 and 56 after the initial MPA injection) ( Table 1 ). To determine whether these serum levels of MPA were biologically active, we investigated Nm23-H1 expression in normal mammary epithelial cells of mice treated twice weekly with the 2-mg and 4-mg doses of subcutaneous MPA. Immunohistochemical analysis of mammary fat pads collected at necropsy on day 7 showed that Nm23-H1 was localized to the luminal side of the mammary epithelial cells in untreated mice. In both treatment groups, Nm23-H1 expression was increased in both the Journal of the National Cancer Institute, Vol. 97, No. 9, May 4, 2005 ARTICLES 637

7 mammary epithelial cells and the stroma relative to expression in untreated mice ( Fig. 2 and data not shown). An isotypematched control antibody showed no staining (data not shown). These data suggest that MPA stimulates expression of Nm23-H1 in normal mouse mammary epithelial cells in vivo at doses that achieved serum MPA levels comparable to those reported in clinical trials. Effect of MPA on Metastatic Colonization of the Lung We conducted two independent in vivo metastasis experiments that differed slightly in the maintenance phase ( Fig. 3, A ). MDA- MB-231T cells ( ) were injected intravenously into nude mice, which were then housed for 4 weeks to permit formation of micrometastases. A pathologist s examination of hematoxylin and eosin stained sections of lungs obtained from 4-week postinjection mice revealed micrometastases (i.e., clusters of tumor cells within the lung parenchyma) ( Fig. 3, B ). In Experiment 1, an induction dose of vehicle or 2 mg or 4 mg of MPA was then given twice weekly for the next 4 weeks; in Experiment 2 the 4-mg dose was dropped due to excessive weight gain, and mice were given MPA doses of 0.5, 1, and 2 mg. In Experiment 1, the 4 weeks of induction treatment was followed by a single maintenance injection of the induction dose given 4 weeks later. Large metastases developed in the mice within the maintenance period, and all mice in the experiment were killed 1 day after the maintenance dose was administered. In Experiment 2 the maintenance dose was modified to half the induction dose (0.25, 0.5, and 1 mg, respectively) every other week for 6 weeks to maintain serum MPA levels throughout the maintenance phase. In Experiment 1, the percentage of mice with metastases was 100% in the vehicle control group and 73% (95% confidence interval [CI] = 45% to 92%) in the 2-mg MPA group. In Experiment 2 the respective percentages were 100% in the vehicle control group and 64% (95% CI = 35% to 87%) in the 2-mg MPA group ( Table 2 ). A comparison of the pooled results for the vehicle control versus 2-mg dose from both experiments showed a borderline statistically significant reduction in the percentage of mice with gross metastases with MPA treatment (P =.04). The mean numbers of gross metastases per mouse in the vehicle control arms were similar for the two experiments (32.3 [95% CI = 9.9 to 54.7] in Experiment 1 and 33.4 [95% CI = 11.8 to 55] in Experiment 2), although the number of metastases in individual mice varied widely. The mean numbers of gross metastases per mouse were 21.7 (95% CI = 7.7 to 35.7) (2-mg dose) and 15.6 (95% CI = 0.2 to 31) (4-mg dose) in Experiment 1 (P =.086) and 23.8 (95% CI = 10.7 to 36.9) (0.5-mg dose), 12.6 (0 to 25.6) (1-mg dose), and 14.5 (3.5 to 25.5) (2-mg dose) in Experiment 2 (P =.033). In general, mice treated with MPA had smaller metastases than control mice treated with vehicle; the fraction of mice with large (i.e., 3 mm) metastases and the number of large metastases per mouse were lower in MPA-treated mice than in vehicletreated mice ( Table 2 ). Data pooled from both experiments for the 2-mg dose showed that MPA-treated mice had a smaller proportion of large metastases per mouse than control mice (P =.013). In addition, in Experiment 1, a likelihood ratio test across doses (0 mg to 2 mg to 4 mg) demonstrated that there was a statistically significant trend toward fewer metastases being large as dose level increased (P =.002). In this experiment, each of the two active doses led to a statistically significantly lower Table 2. Medroxyprogesterone acetate (MPA) effects on pulmonary metastatic outgrowth of MDA-MB-231T breast carcinoma cells * Expt 1 Expt 2 Dose Control 2 mg 4 mg Control 0.5 mg 1 mg 2 mg Mice per group N = 15 N = 15 N = 10 N = 14 N = 15 N = 14 N = 14 Mice with metastases Number 15/15 11/15 6/10 14/14 11/15 10/14 9/14 Percent 100% 73% 60% 100% 73% 71% 64% 95% Confidence Interval 78% to 100% 45% to 92% 26% to 88% 77% to 100% 45% to 92% 42% to 92% 35% to 87% Trend P =.021 P =.034 Metastases per mouse Mean % Confidence Interval 9.9 to to to to to to to 25.5 Range % Reduction Trend P =.086 P =.033 Mice with at least one large metastasis (>3 mm) Number 11/15 6/15 3/10 9/14 3/15 4/14 5/14 Percent 73% 40% 30% 64% 20% 29% 36% 95% Confidence Interval 45% to 92% 16% to 68% 7% to 65% 35% to 87% 4% to 48% 8% to 58% 13% to 65% Trend P =.043 P =.22 Large metastases per mouse Mean % Confidence Interval 0.7 to to to to to to to 1.8 % Reduction Trend P =.015 P =.16 *Mice were injected with MDA-MB-231T cells intravenously on day 0 and given 4 weeks to develop pulmonary micrometastases. Mice were then randomized to receive vehicle control or MPA at the indicated doses by subcutaneous injection twice weekly for 4 weeks. In Experiment 1, mice then received an injection every fourth week during the maintenance period of 4 weeks, and in Experiment 2 mice received half the induction dose every other week for a maintenance period of 6 weeks. Gross pulmonary metastases were quantitated at necropsy by visual inspection. Exact Cochran-Armitage test for trend. Jonckheere Terpstra test for trend. 638 ARTICLES Journal of the National Cancer Institute, Vol. 97, No. 9, May 4, 2005

8 fraction of mice with large metastases among mice with any metastases (P =.003 and P =.03, respectively, for comparison of the 2-mg dose versus control and the 4-mg dose versus control, after Hochberg adjustment for multiple comparisons). In Experiment 2, no such trend was identified overall (P =.46), although a likelihood ratio test comparing control with a pooled group of mice receiving all three dose levels identified larger metastases among mice that did not receive MPA (P =.064). We tested the impact of increasing the MPA dose during the maintenance period in an experimental arm of Experiment 2 (data not shown). Mice were given 2 mg of MPA twice weekly throughout the study rather than twice weekly for 4 weeks and then once every other week, as was done in the induction/maintenance regimen. The mean numbers (range) of metastases were 15.8 (0 65) and 14.5 (0 55) for mice given 2 mg of MPA twice weekly and mice given 2 mg of MPA in the standard induction/ maintenance regimen, respectively, and 33 (1 110) in the vehicle control arm. There were no statistically significant differences in either the percentage of mice without metastases or the number of large metastases between the two treatment groups. Thus, increasing the frequency of MPA injections during the maintenance period did not further reduce metastatic colonization. Analysis of Potential Side Effects of MPA Treatment The reported human side effects of MPA include weight gain (46 48 ) and decreased bone density (49 ). MPA induces mammary tumors in rodents, but only at doses 10-fold higher than those used in our experiments (50,51 ). In our study, all mice treated with MPA gained weight relative to vehicle-treated mice ( Fig. 4 ). In Experiment 1, the 4-mg dose caused up to a 21% weight gain during the treatment period, compared with up to 30% for the 2-mg dose relative to the control ( Fig. 4, A ). In Experiment 2, mice were treated for a longer period. The 1- and 2-mg doses exerted graded effects on weight gain during the treatment period (7% and 16%, respectively, relative to 13% for the controls) ( Fig. 4, B ). Further examination of the weight gain was conducted by measuring the lean (muscle) and fat weight of 10 mice per treatment group from Experiment 2, using a dual energy x-ray absorptiometer ( Table 3 ). The weight gain was balanced between lean and fat weight, and the distribution of weight gain (i.e., the balance of fat and lean) did not differ between MPA-treated and control mice. In addition to causing weight gain, steroids are known to affect bone density. The bones from 10 mice per treatment group in Experiment 2 were examined for bone mineral density and bone mineral content ( Table 3 ). No statistically significant differences were observed between the control and the MPA treatment groups. Finally, we conducted an analysis of mammary fat pad histology to examine whether MPA had any effects on breast tumor carcinogenesis. An axillary mammary fat pad was harvested from each mouse in Experiment 2 at necropsy. Histologic analysis of hematoxylin and eosin stained sections of formalin-fixed, paraffin-embedded mammary fat pads showed that the pads appeared normal in both control and MPA-treated mice. All pads were surrounded by mature and/or brown fat (data not shown). Nm23 Expression in Lung Metastases To examine the effect of MPA treatment on Nm23-H1 expression in lung metastases, we sectioned the lungs of five A Weight (g) B Weight (g) metastasis-containing mice from each treatment group and subjected one randomly chosen section from each mouse to immunohistochemical staining for Nm23 ( Fig. 5 ). Lung metastases were categorized as exhibiting homogeneous or heterogeneous high Nm23 expression or homogeneous low Nm23 expression based on the relative staining of metastases compared with that of surrounding lung parenchyma. All metastases in the stained section were scored. The metastases from the control mice exhibited a range of Nm23 expression intensities: 54% of lesions had homogeneous low expression, 13% had homogeneous high expression, and 33% had heterogeneous high expression. By contrast, mice treated with 1 or 2 mg MPA displayed homogeneous low Nm23 staining in only 9% and 6% of lesions, respectively, and homogeneous high Nm23 staining in 41% and 43% of lesions, respectively. In addition, the total percentages of metastases with homogeneous low expression were higher in control mice than in the 1 mg ( P =.024) or 2 mg ( P =.048) group. MPA at 0.5 mg had little or no stimulatory effect on Nm23 expression in metastases compared with vehicle alone. D ISCUSSION Weeks Postinjection Weeks Postinjection Fig. 4. Body weight in medroxyprogesterone acetate (MPA) treated mice over the course of the experiments. Means ± 95% confidence intervals are shown. (A ) Experiment 1: open diamond, control; filled square, 4 mg of MPA; filled triangle, 2 mg of MPA. A statistically significant difference (P <.001 by analysis of variance [ANOVA]) in weight between control and treatment groups was noted at experimental end. (B ) Experiment 2: open diamond, control; filled triangle, 2 mg of MPA; filled square, 1 mg of MPA; and filled circle, 0.5 mg of MPA. A statistically significant difference (P =.0046 by ANOVA) in weight between control and the 2-mg and 1-mg treatment groups was noted at the end of the experiment. We have used an in vivo experimental metastasis model to assay for the effects of MPA on metastatic colonization by permitting MDA-MB-231T breast carcinoma cells to arrest, extravasate, Journal of the National Cancer Institute, Vol. 97, No. 9, May 4, 2005 ARTICLES 639

9 Table 3. Effect of medroxyprogesterone acetate (MPA) treatment on body composition MPA treatment group * Body composition parameter Control 0.5 mg 1 mg 2 mg P Weight, g 23.4 (21.8 to 24.9) 24.6 (23.0 to 26.1) 25.0 (23.4 to 26.5) 25.8 (24.2 to 27.3) 0.26 % Fat 14.1 (13.1 to 15.1) 14.7 (13.6 to 15.7) 13.6 (12.6 to 14.6) 14.2 (13.1 to 15.2) 0.46 Lean Weight, g 20.1 (18.7 to 21.5) 20.9 (19.5 to 22.3) 21.6 (20.2 to 23.0) 22.2 (20.7 to 23.6) 0.27 Fat Weight, g 3.3 (3.0 to 3.6) 3.6 (3.3 to 3.9) 3.4 (3.1 to 3.7) 3.7 (3.3 to 4.0) 0.32 Bone mineral density, g/cm (0.047 to 0.052) (0.050 to 0.054) (0.050 to 0.054) (0.049 to 0.053) 0.50 Bone mineral content, g 0.53 (0.51 to 0.58) 0.57 (0.54 to 0.61) 0.57 (0.54 to 0.61) 0.57 (0.53 to 0.60) 0.32 * Results shown are from analysis of mice in Experiment 2. Mice were randomized to receive vehicle control or MPA at the indicated doses by subcutaneous injection twice weekly for 4 weeks (induction phase) followed by half the induction dose every other week for a maintenance period of 6 weeks. All values are from 10 mice per group and are presented with 95% confidence intervals. By analysis of variance (ANOVA). Weight after necropsy, means from one-way ANOVA. Percent fat reported from GE Lunar Piximus Dual-Energy x-ray Absorptiometer output. and proliferate in the lung for 4 weeks before beginning MPA treatment. Our goal was to test the effects of MPA in a model system that might provide insights into its clinical efficacy. Indeed, micrometastases were detectable histologically before initiation of treatment. In this model system, MPA inhibited several aspects of pulmonary metastatic colonization. The number of gross metastatic lesions declined substantially, the percentage of mice without metastases increased, and the number of large (i.e., 3 mm) metastases per mouse decreased. It is impossible to know whether the reduced metastasis would be associated with improved survival, because this endpoint is not attainable under current animal care regulations. Our finding that MPA elevated Nm23-H1 metastasis suppressor gene expression in MDA- MB- 231T cells in vitro and in pulmonary metastases in vivo validates a molecular mechanism of action. This study represents a first attempt to bring the rapidly developing metastasis suppressor field toward clinical application. Several metastasis suppressor genes contribute to the control of metastatic colonization (52,53 ), and such genes may represent ideal therapeutic targets because they are rarely mutated and simply need to be turned back on. Our preclinical modeling effort also included an extensive pharmacokinetic evaluation designed to allow us to identify MPA doses and schedules that would lead to serum concentrations that are achievable in humans. Elevation of mammary fat pad Nm23- H1 expression in the ductal epithelial cells was used as a surrogate biologic endpoint for the pharmacokinetic study. However, factors such as serum binding proteins may differ between mice and humans, leading to differences in bioavailability even when serum concentrations appear to be similar. The extent to which elevation of Nm23-H1 expression contributes to the phenotypic effects of MPA was estimated in colony formation assays. Antisense Nm23-H1 transfectants, which did not undergo the MPA-induced increase in Nm23-H1 expression, also exhibited no inhibition of colony formation. Although this result suggests that MPA exerts its suppressive effects on metastasis via its effect on Nm23-H1 expression, it is possible that MPA has additional suppressive activities in this model system. The in vitro and in vivo suppressive effects of MPA on breast neoplastic progression have been ascribed to other molecular events, mostly through the progesterone receptor (54 64 ). In addition, MPA has been noted to exert stimulatory effects on breast neoplastic progression (65 ), particularly in combination with estrogen (66,67 ), and the risk benefit ratio will determine the acceptability of this agent. Indeed, it was to avoid the effects of MPA on PR that we focused our preclinical experiments on ER /PR disease. Experiments are under way to define the Fig. 5. Medroxyprogesterone acetate induced elevation of Nm23-H1 expression in pulmonary metastases. Sections of lung metastases from Experiment 2 were stained immunohistochemically for Nm23-H1. Metastases were scored as either homogeneous high Nm23-H1 expressing, heterogeneous high expressing (stronger than the surrounding lung parenchyma), or homogeneous low Nm23 expression (equivalent in intensity to the lung parenchyma). Photomicrographs of pulmonary metastatic lesions of each category are shown. The fraction of metastases from each staining category from five mice from each arm (control, 0.5-, 1-, and 2-mg doses) of Experiment 2 is shown. All metastases within a section were analyzed. Original magnification of homogeneous low expressing section was 100. Original magnification of heterogeneous high expressing and homogeneous high Nm23-H1 expressing sections was ARTICLES Journal of the National Cancer Institute, Vol. 97, No. 9, May 4, 2005