Profound Prevention of Experimental Brain Metastases of Breast Cancer by Temozolomide in an MGMT-Dependent Manner

Breast cancer is the second most frequent cause of brain metastases, after lung cancer. Brain metastases often occur in patients with advanced HER2-positive breast cancer, including those with stable extracranial disease or while responding to systemic therapy (1, 2). For patients with triple-negative (estrogen receptor and progesterone receptor negative, HER2-normal) advanced breast cancer, a similar percentage develop brain metastases in a setting of progressive systemic disease (3). Current treatments for brain metastasis are palliative, including whole-brain radiotherapy (WBRT), stereotactic radiosurgery, neurosurgery, and steroids (4).

Chemotherapy or molecular therapies have played only a limited role in the treatment of brain metastases (5–11), as the brain is protected from most drugs by the blood–brain barrier (BBB). Pharmacokinetic and imaging studies in mouse models indicate that the extent of BBB opening following its disruption by the formation of a brain metastasis is limited and heterogeneous (12–14), a conclusion supported by clinical studies (reviewed in refs. 15–17). Using an experimental brain metastasis model system, we previously tested multiple drugs for the ability to prevent the formation of brain metastases or to shrink established brain metastases. Six drugs have shown partial efficacy in the prevention setting (reviewed in ref. 18), none were able to shrink established brain metastases.

Temozolomide is an oral, brain-permeable alkylating agent characterized by significant uptake in the central nervous system, and is used in the treatment of primary brain tumors (19–21). This compound induces a number of different DNA lesions and acts in an O⁶-methylguanine-DNA methyltransferase (MGMT)–dependent manner (review in ref. 22). Knowledge on the efficacy of temozolomide in advanced breast cancer is scarce (23), and the potential of this compound in prevention of brain metastases from breast cancer is unknown. We investigated the preventive effect of temozolomide using an experimental model of breast cancer brain metastasis, and determined the functional contribution of MGMT expression in this setting.

We hypothesized that temozolomide would significantly prevent the formation of brain metastases in patients with breast cancer whose tumors have low-to-no MGMT activity. The percentage of patients with breast cancer representing this category, as well as whether the primary breast tumor is a good indication of MGMT status in tumor cells metastatic to the brain was unknown. To address these questions, we determined MGMT expression immunohistochemically in patient-matched sets of primary breast cancers and resected brain metastases.

For in vitro and in vivo experiments, temozolomide was obtained from Sigma or the Drug Synthesis and Chemistry Branch, Developmental Therapeutics Program, NCI, respectively.

A brain metastatic derivative of the triple-negative breast cancer cell line MDA-MB-231 was transduced with EGFP (231-BR-EGFP), as previously reported (24). Cells were maintained in high glucose Dulbecco's Modified Eagle Medium (Life Technologies) supplemented with 10% FBS (Life Technologies) at 37°C in 5% CO₂.

Cell viability was assessed using MTT (Sigma) as previously described (25). Three separate experiments were performed, with n = 4 for each data point within an experiment. For clonogenic assays, cells were plated at a single-cell density and treated with vehicle or temozolomide 24 hours later and every third day for 10 days. Colonies were fixed and stained with crystal violet for quantification. Three separate experiments were performed with n = 3 for each data point within an experiment. Western blot analysis was preformed per standard procedures. Primary antibodies for MGMT (Cell Signaling Technologies) and α-tubulin (Oncogene Sciences) were used.

Derivation of the HER2-positive JIMT-1-BR3 cells is described below. JIMT-1-BR3 cells were maintained similarly to 231-BR-EGFP cells, but with the addition of 200 μmol/L L-glutamine (Life Technologies).

For short hairpin RNA (shRNA)–mediated MGMT knockdown in JIMT-1-BR3 cells, MISSION VSV-G pseudotyped lentiviral particles expressing shRNA-targeting MGMT (sequence #1 CCGGAGCCTGGCTGAATGCCTATTTCTCGAGAAATAGGCATTCAGCCAGGCTTTTTTTG or #2 CC GGTGAGCGACACACACGTGTAACCTCGAGGTTACACGTGTGTGTCGCTCATTTTTTG) and scrambled nontarget control (sequence CCTAAGGTTAAGTCGCCCTCGCTCTAGCGAG GGCGACTTAACCTTAGG; Sigma-Aldrich) were used to transduce JIMT-1-BR3 cells at multiplicity of infection of 10, according to the manufacturer's recommendations. Polyclonal populations were selected for 2 weeks in the presence of puromycin (Life Technologies).

For clonogenic assays in JIMT-1-BR3 cells, 250 cells were plated at single-cell density and, after 24 hours, treated with either vehicle or temozolomide at 10, 50, or 100 μmol/L final concentrations. After 14 days, colonies were fixed and stained with crystal violet for quantification.

All animal experiments were conducted under an approved Animal Use Agreement with the NCI.

To assess the preventive role of temozolomide, 5- to 7-week-old female NRC nu/nu mice (Charles River Laboratories) were inoculated with 175,000 231-BR-EGFP cells in 0.1 mL PBS in the left ventricle of the heart (25–27). Three days after tumor cell inoculation, mice were randomized to temozolomide at a dose of 50 mg/kg delivered by oral gavage in saline, 5 days a week for 4 weeks, or vehicle (saline). Subsequent experiments used temozolomide doses of 25, 10, 5, 1, and 0.5 mg/kg. To evaluate the efficacy of temozolomide in treating established brain metastases, mice received temozolomide (50 mg/kg) beginning on either day 18 or day 24 after injection of 231-BR-EGFP cells, 5 days a week for 2 and 1 week, respectively. In all experiments, mice were euthanized under CO₂ asphyxiation 28 days after tumor cell injection, and the brains were removed at necropsy. Five hematoxylin and eosin (H&E)–stained serial sections (10-μm-thick), one every 600 μm in a sagittal plane through the right hemisphere of the brain were analyzed at ×50 magnification using an ocular grid. Every micro- or large (≤300 and >300 μm along the longest axis, respectively) metastasis in each section was tabulated. The left hemisphere of the brain was used for immunohistochemical analysis.

To investigate the impact of temozolomide on survival, mice injected with 231-BR-EGFP cells were randomized to vehicle, temozolomide on days 3 to 14, or temozolomide on days 17 to 28 after injection, per the schedule described above. Mice were maintained without further treatment up to 109 days and were sacrificed for signs of metastatic progression (loss of 20% body weight, seizures, and paralysis).

To determine the functional contribution of MGMT expression in the 231-BR-EGFP brain metastasis prevention model, expression of human MGMT or a control vector was induced in independent polyclonal populations using a lentiviral expression system. Briefly, the human MGMT cDNA was purchased from Origene Technologies and cloned into the pCDH-CMV-Hygro lentiviral vector (Systems BioScience) for lentivirus production and subsequent infection of cells per the manufacturer's recommended protocol. After infection, 231-BR-EGFP cells infected with MGMT or a vector control virus were selected in hygromycin for 2 weeks. Three independent populations of cells expressing MGMT or vector virus were harvested.

Two vector-expressing and two MGMT-expressing polyclonal populations were injected into the left cardiac ventricle of female nude mice, 16 mice per arm. On day 3 after injection, each group was randomly divided into vehicle or 50 mg/kg temozolomide by oral gavage arms, 5 days a week for 4 weeks.

A brain-tropic derivative of HER2-positive JIMT-1 breast cancer cells (28) was selected. In derivation experiments, 500,000 JIMT-1 cells (28) were injected in the left ventricle of 5- to 7-week-old female NRC nu/nu mice, and mice were housed until they showed signs of distress. Mice were euthanized if they lost greater than 20% of their starting body weight or they became paralyzed. At necropsy, brains were removed, manually dissociated, and placed in tissue culture. Cells that grew out were pooled as JIMT-1-BR1 cells, and the procedure was repeated two additional times to establish JIMT-1-BR3 cells that form brain metastases in 100% of mice injected over a 3- to 6-week period. For temozolomide experiments, 175,000 JIMT-1-BR3 cells were injected and randomized to vehicle or 50 mg/kg temozolomide identical to MGMT-negative 231-BR-EGFP experiments.

MGMT status was assessed in primary tumors and in corresponding brain metastases. Two sets of matched primary breast tumors and resected brain metastases were selected from the Polish brain metastases Consortium (n = 106) and the University of Kiel (Kiel, Germany; n = 14) databases, respectively. Formalin-fixed, paraffin-embedded (FFPE) samples were used for construction of a tissue microarray (TMA). MGMT immunohistochemical expression in the primary tumor and brain metastasis was compared for 62 matched sets; clinical parameters were established for 49 of those sets (excluded were 51 samples with no viable tumor tissue and six with insufficient clinical information).

All procedures were performed according to the manufacturer's instructions (Imgenex, Corp. and LifeSpan, BioSciences, Inc.). The antigen–antibody complex was visualized using the Novolink Polymer Detection System. TMA sections were deparaffinized in xylene and rehydrated through graded alcohol concentrations (100%, 96%, 80%, and 70%). For antigen retrieval, slides were pretreated with a low pH target retrieval solution (Dako). Endogenous biotin was blocked with an appropriate kit. Sections were incubated for 1 hour with an antibody against the human MGMT (monoclonal antibody, mAb; clone MT3.1; dilution, 1:50; Imgenex, Corp. and mAb from LifeSpan, BioSciences, Inc.; dilution, 1:50). Only a nuclear staining was considered positive. Tonsil tissue served as a positive control. The immunoreactivity was scored semiquantitatively as follows: 0, <5% positive tumor cells; 1+, 5% to 75% positive tumor cells; 2+, 75% to 95% positive tumor cells; 3+, >95% positive tumor cells (29); data were subsequently combined into a MGMT-negative (≤5% positive) and MGMT-positive (>5% positive) categories (Supplementary Table S1). The testing laboratory was blinded to patient characteristics.

The Wilcoxon rank sum test compared distributions from two-samples. In the rare case the data were normally distributed, a one-way or factorial ANOVA was performed on the data. The Holm's method was used to adjust for multiple comparisons. Analysis of MGMT immunohistochemical data used the following tests as appropriate: The Fisher exact test for 2 × 2 tables, the Cochran–Armitage trend test for 2 × C ordered tables, and the Jonckheere–Terpstra trend test for doubly ordered R × C tables. Actuarial analyses used the Kaplan–Meier method. Overall survival was measured from the date of first primary tumor surgery to date of death or last follow-up. Survival times were censored if the subject was alive as of the last follow-up. The log-rank test was used to test for differences between strata. All reported P values are two-sided. Considering the large number of tests performed, only P values of <0.005 were deemed statistically significant, whereas 0.005 < P < 0.05 were deemed a strong trend.

A previously characterized model system using a brain-tropic subline of the triple-negative human MDA-MB-231 breast carcinoma cell line (231-BR-EGFP) was used to generate experimental brain metastases. Tumor cells were injected into the left cardiac ventricle; on day 3 after injection, mice were randomized to receive either vehicle or temozolomide. At necropsy on day 28 after injection, experimental brain metastases were quantified in step sections through one brain hemisphere as micrometastases or large metastases (based on a cutoff of 300 μm along the longest axis, comparable with a several mm lesion in a human brain). Temozolomide, administered at a dose of 50 mg/kg, 5 days a week for 4 weeks, prevented the formation of all large- and micro-brain metastases over two replicate experiments (Table 1); even single tumor cells could not be detected in sections of the brains. Vehicle-treated mice developed brain metastases at normal rates. This dose was reported to be consistent with clinically achievable doses of temozolomide (30, 31). Doses of 25, 10, and 5 mg/kg on the same schedule yielded the same results, complete prevention of brain metastases formation in vivo (Table 1). At a dose of 1 mg/kg, 50-fold lower than the starting dose, medians of 0.9 and 1.2 large metastases per section were present in treated brains, as compared with 2.3 and 4.8 large metastases per section in vehicle-treated mice, respectively, corresponding to 61% and 75% reductions (P = 0.80 and 0.059, respectively) in the two experiments conducted. Similar inhibition of micrometastatic lesions was observed. At a dose of 0.5 mg/kg, two logs lower than a widely reported preclinical regimen, temozolomide did not significantly prevent the formation of large- or micrometastases. The data identify a potent preventive effect of temozolomide on the formation of 231-BR-EGFP experimental brain metastases of breast cancer over a wide dose range.

Temozolomide was ineffective in the treatment (i.e., shrinking) of metastatic breast cancer (23). In experiments testing the ability of temozolomide to treat established brain metastases, mice were randomized to receive temozolomide beginning on either day 18 or day 24 after injection. Temozolomide administered on day 18 after injection of 231-BR-EGFP cells resulted in a median of 1.4 large metastases per section, compared with 4.7 for the vehicle controls, a 70% reduction (P = 0.0002; Fig. 1A). A 59% reduction was observed in micrometastases (P = 0.0001; Fig. 1B). Day 18 after injection represents a time point where multiple micrometastases and occasional large metastases are present. When treatment was further delayed to day 24 after injection, a time when greater numbers of large metastases had formed, as demonstrated in imaging studies (32), the efficacy of temozolomide was lost (a median of 5.8 large metastases per section compared with 4.7 in vehicle-treated mice, and similar trends were observed for micrometastases; Fig. 1B). Prolongation of the experiment was impossible, as mice administered vehicle or temozolomide on day 24 required euthanasia for paralysis and other indications. The data indicate that the inhibitory effect of temozolomide was reduced by late administration, possibly due to decreased delivery throughout larger tumor masses and/or shorter exposure to the drug.

To investigate the impact of preventing experimental brain metastases on mouse survival, mice were injected with 231-BR-EGFP cells in the left cardiac ventricle, and then randomized to three arms: Vehicle from days 3 to 28; temozolomide (50 mg/kg, 5 days a week) from days 3 to 14 or temozolomide from days 17 to 28. All mice were then left untreated and monitored for signs of paralysis, weight loss, or seizures requiring euthanasia. All vehicle-treated mice required euthanasia by day 45 after injection, with a median survival of 5 weeks after injection compared with 10.9 weeks for the delayed treatment (Fig. 2; median survival was not reached in the early treatment). The two schedules of 2-week temozolomide treatment significantly increased survival (P = 0.0003 by log-rank test). Earlier (days 3–14) and delayed (days 17–28) administration of temozolomide resulted in long-term survival of 60% (6/10 mice) and 18% (2/11 mice), respectively; a significant difference compared with vehicle (P < 0.0001 and 0.0008, respectively). Taken together, the data indicate that temozolomide administration prolonged mouse survival and was associated with cures. This is consistent with data from previous experiments showing no evidence of disease from histopathologic counts. In this experiment the early versus late treatment arms received the same cumulative dose of temozolomide, and the data favored early treatment.

Whereas temozolomide induces a number of DNA lesions, evidence suggests that the formation of the O⁶-methylguanine DNA lesions is associated with temozolomide activity in primary brain tumors (22). This lesion is repaired by the enzyme MGMT. In in vitro experiments, temozolomide completely abolished the ability of single 231-BR-EGFP cells to form colonies (Fig. 3A, P = 0.003 at 10 μmol/L and P = 0.0004 at 50 μmol/L) but decreased proliferation by only 24% (P = 0.0002; Supplementary Fig. S1). In agreement with the sensitivity of 231-BR-EGFP cells to temozolomide, the cells were MGMT negative by Western blot analysis (Fig. 3B) and quantitative real-time PCR (data not shown). To determine the functional contribution of MGMT expression in our brain metastasis preventive model, expression of a MGMT cDNA was induced using a lentiviral expression system. Independent polyclonal populations of vector or MGMT-expressing 231-BR-EGFP cells were isolated (Fig. 3B). When 231-BR-EGFP cells were forced to express MGMT, in vitro colonization was similar to vehicle-treated cells at both 10 (P = 1.0) and 50 μmol/L (P = 0.92) concentrations of temozolomide; proliferation in vitro was unaffected.

The impact of MGMT expression on temozolomide prevention of 231-BR-EGFP brain metastases was tested in experimental brain metastases assays. Two vector– and two MGMT–positive polyclonal populations were injected into mice as described, and randomized to vehicle or 50 mg/kg temozolomide on the same schedule as previously described. Data from the two vector and two MGMT-expressing polyclonal cell populations were combined as no significant difference was detected between the populations. A representative experiment of two conducted is shown (Fig. 3C and D). For the two experiments conducted, vehicle-treated mice injected with 231-BR-EGFP-vector cells developed medians of 1.4 and 0.7 large brain metastases per section, whereas temozolomide again abrogated all large and micrometastasis development. Mice that received 231-BR-EGFP cells expressing MGMT, and were treated with temozolomide, developed a median of 1.7 and 2.8 large brain metastases per section, respectively. These data were similar to that of untreated mice injected with 231-BR-EGFP and represented a strong trend of distinction from the temozolomide-treated mice injected with 231-BR-EGFP (P = 0.066 and P = 0.0003, respectively). The data suggest that MGMT expression may modulate brain metastatic activity to a limited extent (Fig. 3C). In agreement with previous data, temozolomide completely inhibited the formation of brain metastasis by 231-BR-EGFP-vector cells (P < 0.001 for both experiments), bringing to four the number of independent experiments with complete preventive activity at 50 mg/kg (see Table 1). For the MGMT-expressing 231-BR-EGFP cells, temozolomide administration resulted in a median of 0.6 and 2.8 large brain metastases per section in the two experiments conducted, as compared with 1.7 and 2.8 large brain metastases per section in vehicle-treated mice, respectively (P = 0.010 and P = 0.48). Similar trends were observed for micrometastases (Fig. 3D). Expression of MGMT in the tumor cells that formed metastases in the brain was heterogeneous at the end of the experiment; however, the majority of lesions maintained some level of expression of the transduced gene (Fig. 3E). Thus, temozolomide prevention of brain metastases formation in this model system was MGMT dependent. We have been unable to identify a second brain metastasis model system that is low in MGMT expression for validation purposes.

We tested the brain metastases preventive ability of temozolomide in a second experimental brain metastasis model system. A brain-tropic subline of the HER2-positive breast cancer cell line JIMT-1 (28), which is MGMT-positive (Fig. 3B), was derived by three rounds of intracardiac injection, formation of experimental brain metastases, sterile harvest, and ex vivo culture. JIMT-1 cells are reported to be lapatinib-resistant in vitro and may, therefore, be representative of advanced disease (33). In keeping with its MGMT-positive status, there was no significant prevention of brain metastases formation by temozolomide when mice were injected with JIMT-1-BR3 cells and randomized to vehicle or 50 mg/kg temozolomide beginning on day 3 after injection (5 d/wk; Supplementary Fig. S2A and S2B). Vehicle-treated mice developed a median of 8.6 large metastases and 64.6 micrometastases per section, compared with 8.5 large metastases and 57.7 micrometastases for mice treated with 50 mg/kg temozolomide (P = 0.36 and P = 0.63, respectively). JIMT-1-BR3 cells were then transduced with shRNAs, using a scrambled control and two independent MGMT-targeting constructs. MGMT expression of the resulting polyclonal populations is shown on Supplementary Fig. S2C. Colonization assays, which were a faithful indicator of in vivo experimental brain metastasis for 231-BR-EGFP cells (Fig. 3A), were performed in the presence and absence of temozolomide. Knockdown of MGMT had no significant effect on the colonization of JIMT-1-BR3 cells in the absence of temozolomide; both shRNA constructs strongly sensitized the cells to temozolomide, at 50 and 100 μmol/L concentrations (P < 0.0001 for comparisons with scramble control; Supplementary Fig. S2D). The data are consistent with a requirement for low MGMT expression for temozolomide prevention of experimental brain metastasis.

We hypothesize, based on the preclinical data presented, that temozolomide may be effective in preventing the formation of brain metastases from single tumor cells or micrometastases in patients with tumors that have low-to-no MGMT activity. What percentage of patients with breast cancer this represents, as well as whether the primary breast tumor is an accurate predictor of MGMT status in brain metastases, remains a question. MGMT has been screened by multiple methodologies, with the methylation status of its promoter used most often. We reasoned that many molecular mechanisms can downregulate MGMT, besides DNA methylation, and elected to use an immunohistochemical assay for overall protein levels. Two TMAs were stained for MGMT and the percentage of MGMT-positive tumor cells scored by a pathologist. A total of 62 matched sets were stained, each consisting of a primary breast tumor and resected brain metastasis from the same patient, a cohort of rare specimens. Initially a previously reported set of cutoffs was used, including >5%, 6% to 75%, 76% to 95% and >95% positive tumor cells (29). On the basis of the paucity of samples in each of the latter three categories, the data were dichotomized to low (<5% positive tumor cells) and high (≥5%) staining categories. All four potential patterns of MGMT staining in the primary tumor and matched resected brain metastasis were observed (low MGMT expression in both primary tumor and brain metastasis; low expression in the primary tumor, high expression in the brain metastasis; high expression in the primary tumor, low expression in the brain metastasis; low expression in both the primary tumor and brain metastasis; Fig. 4). Overall, the concordance of primary tumor and brain metastasis MGMT expression was weak: in only 60% of cases dichotomized MGMT status in primary tumor and brain metastasis was concordant. Of note, 44% (n = 27) contained concordant low MGMT staining in primary tumors and brain metastases and in another 9 cases (15%) high MGMT expression in the primary tumor corresponded with low expression in the brain metastasis. Taken together, a majority of brain metastases (36 of the 62 matched sets, 58%) had low MGMT–expressing brain metastases. The data indicate that a majority of brain metastases are low in MGMT expression and, therefore, potentially preventable by temozolomide.

Clinical parameters, including hormone receptor status, grade, and subtype of primary tumor, as well as treatment, type of first metastatic progression, and dominant site of metastatic disease, were compared for low versus high MGMT-staining tumors, either the primary tumor or the matched resected brain metastasis (Table 2). A strong trend was observed in brain metastases for an association of HER2 overexpression and MGMT negativity (P = 0.089), suggesting the eligibility of this subset for potential clinical trials. Nine patients (33%) with low MGMT brain metastases (n = 27) experienced brain as the first site of metastatic progression, as compared with 16 patients (76%) with high MGMT–expressing brain metastases (P = 0.004). Brain metastasis as a first site of progression was associated with overall survival (log-rank P = 0.008), whereas MGMT staining, either in the primary tumor or resected brain metastasis was not (Supplementary Table S2; Supplementary Fig. S3).

Temozolomide shows efficacy in primary brain tumors, but is considered inactive in metastatic breast cancer (23). This compound has also been tested in patients with established brain metastases from a variety of cancer types with limited responses, either as monotherapy (34–37), in combination with other cytotoxic agents (refs. 38–40, as examples), or with radiotherapy (refs. 41–43, as examples). Many of these studies enrolled patients with multiple cancer histologies, were focused on responses rather than on initial development of disease, and did not investigate molecular correlates.

We report the preclinical testing of temozolomide in the 231-BR-EGFP experimental brain metastasis model system. The 231-BR-EGFP model system was previously reported to be representative of breast cancer craniotomy specimens in terms of proliferation and apoptosis rates, and a neuroinflammatory response (44). The experimental metastasis model used histologic counts as opposed to imaging, because imaging signals can be variably diminished by their depth within the brain. This model was extensively tested for the prevention of brain metastases by 10 drugs (18). Partial prevention of brain metastases was noted for several drugs (18), but none was completely effective. In contrast, four experiments at 50 mg/kg, two experiments at 25 mg/kg, and one experiment each at 10 and 5 mg/kg dosing schedules completely prevented the formation of 231-BR-EGFP brain metastases. We attempted dual fluorescent staining of brain sections with antibodies specific to human mitochondria and endoplasmic reticulum to identify residual tumor cells that may have been unseen on H&E staining, but remain uncertain of any positive staining. The complete abrogation of brain metastases demonstrated in this study is unique. The extensive dose–response of temozolomide prevention of 231-BR-EGFP brain metastases, over a log of doses, suggests that lower doses of drug may be used in a prevention trial. Indeed, temozolomide efficacy has been studied in lower metronomic regimens with good results (45–47).

Our data also suggest that the inhibitory effect of temozolomide was reduced by late administration. Using histologic counts, a 2-week administration of temozolomide starting at day 17 after injection was superior to a 1-week regimen starting on day 24 after injection. This experiment contained two variables: The size of the brain lesions at the initiation of treatment and the cumulative dose of temozolomide delivered. However, in a survival analysis, equal cumulative doses of temozolomide were administered. Earlier administration of temozolomide (days 3–14 after injection) produced long-term survival in 60% of mice, superior to later administration (days 17–28 after injection). It remains possible that even greater numbers of mice were “cured” of brain metastases in the survival experiment, as some mice also develop bone metastases and require euthanasia for similar symptoms, such as paralysis.

In the 231-BR-EGFP model system, the preventive efficacy of temozolomide was MGMT dependent. Expression of MGMT in the MGMT-null 231-BR-EGFP subline abrogated the brain metastasis preventive activity of temozolomide. In addition, we developed a new model system for experimental brain metastasis of breast cancer based on the JIMT-1 breast cancer cell line (28). Temozolomide was ineffective at preventing brain metastases in the MGMT-positive JIMT-1-BR3 model. This observation raises the question of whether MGMT expression needs to be an enrollment criterion for a temozolomide brain metastasis preventive trial. In earlier studies MGMT expression was detected by enzymatic activity, promoter methylation, mRNA level, and immunohistochemistry (48, 49). We reasoned that many events, not just promoter methylation, can downregulate MGMT gene expression and enzymatic activity; therefore, we used an immunohistochemical assay. We used two TMAs containing rare matched sets of primary breast tumors and the resected brain metastasis from the same patient. In this series, 59% of patients had low MGMT–expressing brain metastases. However, only 60% of resected brain metastasis retained concordant MGMT staining with the primary tumor. Thus, primary tumor MGMT status is an unfaithful predictor of the brain metastasis status, and, therefore, should not be used as trial enrollment criterion. In patients undergoing brain metastases excision, MGMT status may be determined in the first metastasis, but its consistency in subsequent metastases remains to be established. Current knowledge indicates that all patients should be enrolled, and primary tumor MGMT expression should be quantified retrospectively. A higher than 40% MGMT-positivity rate of brain metastases might be factored into statistical calculations of a trial size.

Pharmacologic prevention of brain metastases remains an important goal that is rarely addressed in clinical trials. In small-cell lung cancer prophylactic WBRT has been prospectively demonstrated to reduce brain metastases incidence (50). Cognitive decline from WBRT occurs in a proportion of patients and is irreversible, leading to hesitation in using this modality in breast cancer where survival times can be relatively long. Thus, the identification of new brain metastasis preventive strategies remains an important goal. We advocate for randomized phase II secondary brain metastasis prevention trials to provide initial evidence of a preventive effect (51). Patients with limited numbers of brain metastases, treated with surgery or stereotactic radiosurgery are at high risk for the development of subsequent brain lesions. Such patients could be randomized to placebo or the proposed preventive, in addition to standard systemic therapy. The primary endpoint would be freedom from a new brain metastasis, distant from radiosurgical or surgical beds, and other endpoints should include time to WBRT and quality of life, rather than shrinkage of an established lesion. The finding that temozolomide more effectively prevents the outgrowth of a few tumor cells, as opposed to shrinking a lesion containing millions of tumor cells, makes intuitive sense. The number of tumor cells that must be impacted varies. Micrometastases may have a more normal vasculature and peritumoral hydrostatic pressure, both of which facilitate drug delivery. A cytostatic agent may be sufficient to prevent brain colonization, whereas a cytotoxic agent would be required to shrink an established lesion. There are some hints of clinical brain metastasis preventive activity of temozolomide. In advanced melanoma patients, 2 of 20 patients treated with temozolomide developed brain metastases, as compared with 9 of 21 treated with dacarbazine (P = 0.03; ref. 35). Extensive stable disease was reported in early trials of capecitabine/temozolomide and lapatinib/temozolomide for patients with breast cancer and brain metastases (7, 40).

In conclusion, our preclinical study suggests that temozolomide may effectively prevent the outgrowth of brain metastases in patients with high-risk advanced breast cancer. These data provide a compelling rationale for brain metastasis prevention trials using temozolomide in patients with high-risk advanced breast cancer.

P.S. Steeg reports receiving a commercial research grant from GlaxoSmithKline and Sanofi. No potential conflicts of interest were disclosed by the other authors.

Conception and design: D. Palmieri, R. Duchnowska, S. Woditschka, B. Gril, J. Jassem, P.S Steeg

Development of methodology: D. Palmieri, B. Gril, S. Hewitt, J. Jassem, P.S Steeg

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D. Palmieri, R. Duchnowska, S. Woditschka, Y. Qian, W. Biernat, K. Sosińska-Mielcarek, B. Gril, A. Stark, S. Hewitt

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D. Palmieri, R. Duchnowska, D.J Liewehr, S.M Steinberg, P.S Steeg

Writing, review, and/or revision of the manuscript: D. Palmieri, R. Duchnowska, S. Woditschka, W. Biernat, K. Sosińska-Mielcarek, B. Gril, A. Stark, S. Hewitt, D.J Liewehr, S.M Steinberg, J. Jassem, P.S Steeg

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R. Duchnowska, S. Woditschka, E. Hua, Y. Qian, A. Stark

Study supervision: R. Duchnowska, J. Jassem, P.S Steeg

This work was supported by the Intramural program of the National Cancer Institute and U.S. Department of Defense Breast Cancer Research Program, grant number: W81 XWH-062-0033, and an Intramural grant of the Medical University of Gdańsk (Gdańsk, Poland), grant number ST-51.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.