|Home | About | Journals | Submit | Contact Us | Français|
Brain metastases are highly resistant to chemotherapy. Metastatic tumor cells are known to exploit the host microenvironment for their growth and survival. We report here that melanoma brain metastases are surrounded and infiltrated by activated astrocytes, and we hypothesized that these astrocytes can play a role similar to their established ability to protect neurons from apoptosis. In coculture experiments, astrocytes, but not fibroblasts, reduced apoptosis in human melanoma cells treated with various chemotherapeutic drugs. This chemoprotective effect was dependent on physical contact and gap junctional communication between astrocytes and tumor cells. Moreover, the protective effect of astrocytes resulted from their sequestering calcium from the cytoplasm of tumor cells. These data suggest that brain tumors can, in principle, harness the neuroprotective effects of reactive astrocytes for their own survival and implicate a heretofore unrecognized mechanism for resistance in brain metastasis that might be of relevance in the clinic.
In the United States, up to 40% of cancer patients develop brain metastasis . The top five primary tumors that lead to brain metastasis are lung, breast, melanoma, renal, and colorectal tumors . The median survival duration of untreated patients is 1 to 2 months, and chemotherapy, in conjunction with surgery and radiation therapy, extends survival to only 4 to 6 months . Poor prognosis in these patients is primarily due to chemoresistance . Traditionally, resistance of brain metastasis to chemotherapy has been attributed to the blood-brain barrier (BBB), which prevents toxic substances from reaching the brain parenchyma [4,5]. However, recent studies have revealed that the BBB is not intact in brain metastasis because metastatic tumor cells release vascular permeability factor also known as vascular endothelial growth factor [4,6]. Moreover, more than 70% of brain metastasis cases exhibit enhanced lesions onmagnetic resonance imaging due to leakage of contrast agent from blood vessels, thus indicating BBB permeability .
The outcome of brain metastasis in patients depends on multiple interactions between tumor cells and the tumor microenvironment, which metastatic cells exploit to their advantage [8,9]. Astrocytes, the predominant cells in the brain maintain homeostasis of the brain microenvironment [10,11]. They transport various nutrients from the circulation to the neurons, participate in neural signal transduction, and buffer the ionic balance of the extracellular matrix [12–15]. Under pathological conditions, such as trauma, ischemia, and neurodegenerative disease, astrocytes became activated, as indicated by the up-regulation of glial fibrillary acidic protein (GFAP) . These reactive astrocytes have been shown to protect neurons from injury induced apoptosis [11,17,18]. These observations raise the intriguing possibility that reactive astrocytes can also protect melanoma cells in brain metastases from cytotoxicity induced by chemotherapeutic drugs. To test this hypothesis, we studied the sensitivity of various melanoma cells to chemotherapeutic agents when cocultured with mouse astrocytes or fibroblasts. We conclude that astrocytes (but not fibroblasts) significantly reduced apoptosis mediated by the chemotherapeutic agents and that the protective effects of astrocytes resulted from their sequestering calcium from the cytoplasm of the tumor cells.
Athymic Ncr-nu/nu male mice were purchased from National Cancer Institute-Frederick Cancer Research Facility (Frederick, MD). Animal protocols were approved and supervised by the University of Texas MD Anderson Cancer Center (Houston, TX) Institutional Animal Care and Use Committee.
Human metastatic melanoma cells were introduced into the brain parenchyma of mice through internal carotid artery injection under microscopy as described previously . After mice developed neurological symptoms (e.g., circling), mice were killed, and the brain tissue was harvested for analysis.
The human metastatic melanoma cell lines A375P, DM-4 (from lymph node metastasis specimens) and TXM-13 (from brain metastasis specimens) were all established at the MD Anderson Cancer Center. All melanoma cell lines were recovered from frozen stocks and cultured as previously described . Immortal mouse astrocytes were established in our laboratory . Calcein-AM, Fura-2AM, Fluo-3AM, Hochest 33342, BAPTA-AM, and connexin 43 antibodies were purchased from Invitrogen (Carlsbad, CA). CBX and propidium iodide were purchased from Sigma (St. Louis, MO). GFAP antibodies were purchased from Biocare Medical (Concord, CA). Paclitaxel was purchased from Bristol-Myers Squibb (New York, NY), cisplatin was from SICOR Pharmaceuticals (Irvine, CA), and 5-fluorouracil (5-FU) was from American Pharmaceutical Partners (Schaumburg, IL).
A375P human melanoma cells and murine astrocytes (1:1 ratio) were plated on sterilized glass coverslips in 24-well plates at a density of 2.4 x 104 cells/well. After 48 hours, the samples were processed as described previously .
Tumor cells were cultured alone or with green fluorescent protein (GFP)-expressing astrocytes at a ratio of 1:1. After being incubated for 72 hours with chemotherapeutic drugs, floating and adherent cells were harvested and fixed with 70% ethanol. The propidium iodide staining procedure was performed as described previously . The fluorescein isothiocyanate (FITC)-propidium iodide protocol was used to separate the GFP-expressing, FITC-positive astrocyte population from FITC-negative tumor cells (Figure 1). This was followed by an analysis of apoptosis in control untreated tumor cells, followed by tumor cells and tumor-astrocyte cocultured samples incubated with chemotherapeutic agents. Apoptotic tumor cells were assessed as the fraction of cells in the sub-G0 region. Untreated and apoptotic tumor cells were cultured alone and used as controls. Apoptosis in response to chemotherapeutic drugs in astrocytes cultured alone or cocultured with tumor cells was performed as described above for tumor cells. In this case, apoptosis was assessed as the fraction of GFP-positive cells in the sub-G0 region (Figure 1).
Transwell Boyden Chamber inserts (0.4-µm pore size; Costar, Corning, NY) were seeded with astrocytes (105 cells). Tumor cells (105 cells) were seeded in the bottom well. Tumor cells were treated with chemotherapeutic drugs and analyzed as described above.
The expression of GFAP and connexin 43 in specimens was analyzed by immunostaining using GFAP (1:50) and Connexin 43 antibodies (1:50) in conjunction with peroxidase (human specimens) or fluorescent (mouse specimens and cell staining) antimouse secondary antibody (1:500). Nuclei were stained using Hochest 33342.
The dye transfer assay was performed as described previously . In brief, astrocytes (donor cells) were labeled with calcein-AM (5 µM) for 30 minutes at 37°C. Tumor cells (recipient cells) were labeled with DiI (5 µM) for 30 minutes at 37°C. Astrocytes and tumor cells were plated (1:50) in six-well plates. After 6 hours, the cells were harvested and analyzed by fluorescence-activated cell sorting (FACS) to determine the fraction of DiI-positive tumor cells containing calcein.
Connexin 43 short hairpin RNA (shRNA; target sequence: shRNA1, -GAAGTTCAAGTACGGGATT- and shRNA 2, -CCATCTTCATCATCTTCAT-) and a nontargeting shRNA (target sequence: TTCTCCGAACGTGTCACGT) were used with the lentiviral system developed and kindly provided by Didier Trono (Ecole Polytechnique Fédérale de Lausanne, Lausanne, Switzerland) as described previously .
Total cell extracts (sonification) were separated by gradient (4%–12%) Nu-polyacrylamide gel electrophoresis and transferred to membranes using an iBlot dry blot analysis (Invitrogen). Connexin 43 was detected with antibodies (1:500), using β-actin antibodies (1:3000) as the loading control.
Reverse transcription was performed using the high-capacity RNA-to-cDNA kit (Applied Biosystems, Carlsbad, CA). The gene expression assay for connexin 43 was from Applied Biosystems (assay ID: Mm00439105_m1), with human 18S as the endogenous control (Applied Biosystems). Real-time polymerase chain reaction was performed with the ABI Prism 3000 spectrometer (Applied Biosystems). The relative CT was used for comparison.
At different time intervals, melanoma cells treated with apoptotic agents were labeled with Fura-2AM (1 µM, 30 minutes at 37°C). The cells were then washed and relative changes to cytosolic Ca2+ were expressed as the ratio of fluorescence intensities at 340 and 380 nm (λem = 510 nm) using untreated cells as the control .
A375P melanoma cells were cultured alone or with red fluorescent protein (RFP)-expressing astrocytes. Paclitaxel (10 ng/ml) was added to the culture medium with or without the gap junctional communication (GJC) inhibitor CBX (100 µM). After a 24-hour incubation, the cells were labeled with the Ca2+ indicator, Fluo-3AM(2 µM, 30 minutes, 37°C), and washed twice with phosphate-buffered saline. Cytoplasmic Ca2+ levels in the harvested cells were determined by flow cytometry using the FITC-phycoerythrin (PE) protocol, which allowed the exclusion of RFP-positive (red) astrocytes from the analysis (Figure 2). Cytoplasmic Ca2+ in tumor cells (PE-negative population) was assessed as the mean FITC fluorescence. Data were expressed as the relative change in cytoplasmic calcium compared with control tumor cells that had not been treated with chemotherapeutic agents.
Data are presented as the mean ± SEM. A statistical comparison of treatment sets was performed using the Student's t test.
In the first set of experiments, we injected human A375 parental cells  into the internal carotid artery of nude mice . Mice exhibiting neurological symptoms were killed, and their brain tissue was processed for immunohistological analysis, which revealed that, similar to clinical specimen of human melanoma brain metastases (Figure 3, A–C), the experimental melanoma brain metastases (Figure 3, D–F) were surrounded and infiltrated by GFAP-positive reactive astrocytes.
In the next set of in vitro studies, we used an immortalized astrocyte cell line derived from the brain of H-2Kb-tsA58 mice  and co-cultured these astrocytes with melanoma cells. A scanning electron microscopic examination revealed that the astrocytes formed direct contacts with tumor cells through multiple podia. In some areas, astrocytes and tumor cells formed a seamless structure, which resembled GJC (Figure 3G).
Next, we evaluated chemotherapy-induced apoptosis in tumor cells in the absence and presence of astrocytes. Coculture with astrocytes dramatically reduced 5-FU-induced apoptosis in the human melanoma cells A375P, TXM13, and DM-4 by 78.6%, 62.8%, and 69.7%, respectively (Figure 4A). This protection was not specific to the chemotherapeutic agent because astrocytes demonstrated similar reductions in cytotoxicity against A375P, TXM13, and DM-4 cells with cisplatin (79.7%, 77.2%, and 59.2%, respectively; Figure 4B) and paclitaxel (43.5%, 56.8%, and 82%, respectively; Figure 4C). Similar results were also obtained using the MTT cell viability assay  (data not shown). Astrocytes cultured alone or cocultured with melanoma cells did not undergo apoptosis when incubated with the chemotherapeutic agents under similar conditions (Figure 4D). To determine whether protection by astrocytes required secreted factors or direct physical contact, we repeated the experiments described above but separated the astrocytes from tumor cells by a transwell membrane (0.4-µm pore size). Under these conditions, the astrocytes failed to protect the tumor cells from chemotherapy-induced apoptosis (Figure 4C). Substituting murine NIH3T3 fibroblasts for astrocytes in the coculture experiments failed to provide protection (Figure 4E). Moreover, because astrocytes protect melanoma cells against both P-glycoprotein-sensitive (paclitaxel) and insensitive (5-FU and cisplatin) drugs , the above observations suggest that astrocyte-tumor interactions, rather than P-glycoprotein expression, are a foremost indicator of chemoresistance in melanoma brain metastases . The chemoprotective nature of astrocytes was not restricted to melanoma cells. The protective effect was also demonstrated for human breast cancer cells, human lung cancer cells, and even 3T3 fibroblasts .
GJC channels, which directly connect the cytoplasm between adjacent cells, have been shown to be involved in the transmission of survival and apoptotic signals between neighboring cells [30,31]. Both astrocytes and melanoma cells expressed connexin 43, the major GJC at the cell surface (Figure 5A). The functionality of GJC was evaluated using the dye transfer assay (Figure 5B), which measures the transfer of fluorescent calcein from the astrocytes to the tumor cells; this process can occur only by direct cell-to-cell transfer [30–32]. To determine whether astrocyte-mediated protection of tumor cells occurred through GJC, cocultures were assessed for tumor cell apoptosis in the absence or presence of carbenoxolone (CBX), a specific inhibitor of GJC channels [33,34]. Under these conditions, CBX completely reversed the protection (Figure 5C). To confirm the participation of GJC in tumor cell protection, we knocked down expression of connexin 43 in astrocytes . Two shRNA sequences specific against connexin 43  were inserted into lentiviral vectors and transduced into astrocytes. Stable cell lines were established, and the knockdown was confirmed at the protein (Figure 5D) and transcription (Figure 5E) levels. Analysis of tumor cell apoptosis in coculture experiments with connexin 43-deficient astrocytes (Ast-shRNA-1) revealed increased sensitivity of these tumors to chemotherapy when compared with nontargeted astrocytes (Ast-NT) (Figure 5F). Whereas the small interfering RNA approach precludes complete reversal of protection, these data, taken together with the results of the inhibitor experiments (Figure 5, B and C), underscore the importance of GJC to astrocytes' protection of tumor cells from chemotherapy.
Calcium, one of the most important second messengers transmitted through GJC channels , has been shown to play a causal role in cell death . Temporal analysis of melanoma cells incubated with chemotherapeutic agents revealed an increase in cytoplasmic calcium that is followed by the appearance of fragmented DNA, one of the hallmarks of apoptosis (Figure 6A). The importance of cytoplasmic calcium to melanoma apoptosis is demonstrated by our finding that its chelation with BAPTA inhibited paclitaxel-induced cell death (Figure 6B). Analysis of cytoplasmic Ca2+ in paclitaxel-treated tumor cells by flow cytometry revealed that coculture with astrocytes significantly reduced cytoplasmic calcium levels when compared with tumor cells incubated with paclitaxel in the absence of astrocytes (Figure 6C). Furthermore, this effect was reversed by the inclusion of the GJC channel blocker CBX in the culture medium (Figure 6C). Although we cannot rule out the participation of other molecules by these experiments, these data implicate Ca2+ sequestration through GJC channels as a key mechanism by which astrocytes protect the tumor cells from chemotherapy (Figure 6D).
Resistance to chemotherapy is one of the major causes of mortality in patients with brain metastasis. Until now, this resistance has been primarily attributed to the impermeable nature of the BBB and the expression of P-glycoprotein by metastatic cells, but we provide here a novel alternate mechanism that requires direct physical connections between activated astrocytes and tumor cells in the brain tumor microenvironment. Taken together, the data presented here underscore the ability of metastatic brain tumors to harness and adopt the neuroprotective properties of resident astrocytes for their own survival. Thus, successful treatment of brain metastasis will require chemotherapy in combination with agents that selectively interfere with the GJC channels between these tumor cells and tumor-associated astrocytes. This could potentially be achieved by targeted therapies that are delivered to tumor-associated astrocytes using GFAP antibodies or short nucleotide sequences complementary to GFAP messenger RNA .
The authors thank Ann Sutton for critical editorial comments and Lola López for expert assistance with the preparation of this article.
1This work was supported in part by Cancer Center Support Core Grant CA16672 and grant 1U54-CA143837 (I.J.F.) from the National Cancer Institute, National Institutes of Health. Q.L. was supported by a predoctoral fellowship from The Rosalie B. Hite Foundation at the University of Texas MD Anderson Cancer Center.