Compartment-Specific Roles of ATP-Binding Cassette Transporters Define Differential Topotecan Distribution in Brain Parenchyma and Cerebrospinal Fluid Free

Malignant central nervous system (CNS) tumors carry a poor prognosis and are the most common solid tumors in children (1). Surgical resection and radiotherapy have been the cornerstones of treatment for most brain tumors, but chemotherapy is gaining an equally important role. In children ages <3 years, chemotherapy is preferentially given to avoid or delay irradiation of the developing brain (2). Because drugs approved for adults, such as temozolomide, have not shown efficacy in pediatric trials (3), new agents are needed to treat childhood brain tumors.

Camptothecin analogues (topotecan and irinotecan) exerted potent antitumor activity in xenograft models of human primary CNS tumors, including glioma, medulloblastoma, and ependymoma (4–6). Subsequent research in nonhuman primates showed that topotecan readily penetrates into the cerebrospinal fluid (CSF; ref. 7), and clinical trials showed the activity of topotecan against pediatric CNS tumors that are disseminated through the subarachnoidal space (8). However, topotecan given as a single agent has shown no activity against high-grade gliomas in children (9, 10). These disparate results may be explained in part by different drug penetration of the blood-brain barrier (BBB) and the blood-CSF barrier (BCB; ref. 11).

Among the functional components of the BBB and the BCB (for reviews, see refs. 12–14), proteins of the ATP-binding cassette (ABC) transporter family appear to play a significant role in transporting topotecan (15, 16) and are likely to affect its distribution in the brain parenchyma and CSF compartments (17–19). The ABC transporters are expressed in the brain vascular endothelial cells in the BBB and in the monolayer of ependymal cells of the choroid plexus in the BCB (20). They can also be present on the surface of tumor cells, and several in vitro studies have shown that tumors overexpressing the ABC transporters human breast cancer resistance protein (BCRP/ABCG2), P-glycoprotein (P-gp/MDR1/ABCB1), or multidrug resistance protein 4 (MRP4/ABCC4) are resistant to topotecan (21–23).

Transgenic rodent models in which specific transporters are deleted provide a powerful tool for examining the roles of individual transporters in the brain distribution of drugs. Our previous studies in a Mrp4-deficient mouse model suggested that Mrp4 reduces topotecan penetration into the ventricular CSF (vCSF; ref. 18). However, we observed very high vCSF accumulation of topotecan in wild-type mice (17), strongly suggesting that other drug transporters actively move topotecan from the brain parenchyma or blood into the vCSF. As topotecan is also a substrate of Bcrp1 and P-gp (16, 23, 24), these two transporters are likely to be responsible for the robust movement of topotecan across the BCB. Using immunohistochemistry, we observed the expression of Bcrp1 on the apical side of the ependymal cells of the choroid plexus and the uniform presence of P-gp in the ependymal cell cytoplasm. We also confirmed that gefitinib, a tyrosine kinase inhibitor, impedes topotecan transport into the CSF probably by inhibiting the ABC transporters at the BCB (17).

In the present study, we examined the individual roles of Bcrp1 and P-gp at the BBB and BCB by using knockout mouse models lacking Bcrp1, P-gp, or both proteins, gefitinib as a potent inhibitor of the ABC transporters, and an intracerebral microdialysis sampling technique. We found that the presence of these proteins at the BBB and BCB explains the transport of topotecan into the vCSF and the efflux of topotecan from the brain parenchyma. These findings improve our understanding of active mechanisms regulating drug penetration into the CNS and enhance our ability to develop more effective brain tumor chemotherapy.

Drugs and chemicals. Topotecan hydrochloride (Hycamtin; GlaxoSmithKline) was prepared in sterile water (1 mg/mL). Gefitinib tablets (Iressa; AstraZeneca; 250 mg) were pulverized, and the powder was reconstituted with 0.5% Tween 20 (20%, v/v) and suspended in carboxymethylcellulose (0.25%, w/v) to a final concentration of 40 mg/mL. All other solvents and chemicals used were analytical grade or better.

Animals. FVB wild-type, Bcrp1^−/−, Mdr1a/b^−/−, and Mdr1a/b^−/−Bcrp1^−/− female mice weighing 20 to 25 g were purchased from Taconic. All gene knockouts were created on the FVB genetic background. All procedures were approved by the St. Jude Institutional Animal Care and Use Committee and met the guidelines of the Association for Assessment and Accreditation of Laboratory Animal Care.

Topotecan plasma protein binding. Topotecan lactone plasma protein binding was determined as reported previously (17). The unbound fraction was calculated as the proportion of unbound plasma topotecan lactone to total plasma topotecan lactone.

Cannula implantation for microdialysis. The implantation of a guide cannula in anesthetized mice for microdialysis study of vCSF (25) or extracellular fluid (ECF; refs. 17, 26) has been reported. Briefly, for vCSF microdialysis, a MD-2255 guide cannula (Bioanalytical Systems) was inserted into the lateral ventricle at an angle of 20° posterior (1 mm lateral, 0.8 mm posterior, and 2 mm ventral to the bregma point). For ECF microdialysis, the guide cannula was inserted vertically into the striatum (1.8 mm lateral, 0.6 mm anterior, and 2 mm ventral to the bregma point). Mice were then allowed 3 to 5 days to recover.

Microdialysis studies. The microdialysis procedure and probe calibration have been described in detail (17, 26). Briefly, on the day of an experiment, the microdialysis probe (MD-2211; Bioanalytical Systems) was primed and flushed with artificial CSF (27). The probe was inserted through the cannula into the brain lateral ventricle or striatum, and the artificial CSF perfusion rate was set at 0.5 μL/min. The probe was allowed to equilibrate in vivo for at least 1 h. Topotecan (4 mg/kg) was then administered by intravenous bolus via the lateral tail vein. Gefitinib-pretreated animals received a 200 mg/kg gefitinib dose by oral gavage 1 h before topotecan. After the injection of topotecan, the dialysate samples were directly loaded onto a sample loop (2 μL) and topotecan lactone and carboxylate were simultaneously analyzed through an online microbore high-performance liquid chromatography system until topotecan concentration was undetectable (26). Topotecan carboxylate and lactone concentrations in ECF or vCSF were first corrected for the hydrolysis of topotecan lactone in pH 7.4 artificial CSF using previously described method (26). Probe recovery was determined by the in vivo retrodialysis method after each experiment (26). After the microdialysis experiment, the mice were euthanized and their brains were fixed in 10% neutral buffered formalin for 24 h and embedded in paraffin. H&E-stained sections (4 μm) were examined microscopically to confirm the location of the microdialysis probe track. Three to six animals were used for each condition studied.

Topotecan plasma pharmacokinetics. As in our previous murine microdialysis studies (17), we assessed plasma population pharmacokinetics in a preliminary group of mice. These results were then used as a prior distribution when assessing individual mouse plasma pharmacokinetics with limited sampling in the microdialysis studies. We estimated the plasma population pharmacokinetic parameters in each genotype and treatment group (topotecan 4 mg/kg intravenously alone versus after pretreatment with oral gefitinib 200 mg/kg) as described previously (17). Briefly, each preliminary study included six to seven mice, and blood was sampled from each mouse randomly at four of five time points (0.25, 0.5, 1.5, 3, and 6 h after drug administration chosen based on our previous experience with topotecan murine pharmacokinetic studies) either by retro-orbital bleed or (at the final time point) by cardiac puncture. Sample processing and topotecan lactone and carboxylate bioanalysis have been reported in detail. A two-compartment model was used to describe the topotecan lactone data and a third compartment was included to account for the topotecan carboxylate, and the population parameters were estimated by nonlinear mixed-effects modeling (NONMEM version VI; refs. 28, 29).

For each microdialysis experiment, three plasma samples were collected from the retro-orbital plexus (0.25, 1, and 3 h after topotecan administration; ref. 17) for plasma pharmacokinetic analysis. The above-described plasma pharmacokinetic model for topotecan was fit to the plasma topotecan lactone and carboxylate concentration-time data using maximum a posteriori probability Bayesian estimation (with the prior parameter distribution determined above) as implemented in ADAPT II (30). The plasma area under the curve (AUC) from zero time to infinity (AUC_0→∞) for unbound topotecan lactone and carboxylate was calculated by integration of the simulated concentration-time data from the model estimates. These simulated plasma AUCs were used to calculate the plasma lactone-to-carboxylate ratio of unbound topotecan.

Three-compartment analysis of combined plasma and brain topotecan lactone pharmacokinetic data. A three-compartment pharmacokinetic model (two compartments to describe the plasma topotecan lactone and one for the microdialysis obtained brain unbound topotecan lactone concentrations) was fit to the vCSF or ECF data. Specifically, topotecan lactone plasma pharmacokinetic parameters including volume of central compartment (V_c), systemic clearance (CL_t), volume of peripheral compartment (V_p), and intercompartmental clearance (CL_p) for each mouse were fixed at the values obtained by the methods described in the above section. Then, the model parameters describing the brain disposition (CL_in, CL_out, and V_ECF or V_CSF) were estimated by maximum likelihood estimation via ADAPT II (17, 31). The vCSF or ECF AUC_0→∞ for unbound topotecan lactone was calculated by integration of the simulated concentration-time data from the model estimates.

CNS penetration estimation. The CNS penetration of topotecan lactone was measured as the ratio of the unbound topotecan lactone AUC in brain ECF (AUC_u,ECF) or vCSF (AUC_u,vCSF) to that in plasma (AUC_u,plasma; ref. 31). The CL_in-to-CL_out ratio is also considered to express the extent of CNS penetration of topotecan lactone in the three-compartment analysis (32). Thus, we compared the topotecan penetration estimates calculated by both approaches based on the relationship AUC_brain/AUC_plasma = CL_in/CL_out (32).

Immunohistochemistry. The brains of FVB wild-type, Bcrp1^−/−, and Mdr1a/b^−/− mice were fixed overnight in 10% neutral buffered formalin and embedded in paraffin. The presence of Bcrp1 was immunohistochemically assessed in the sectioned tissues as reported previously (17). The area and intensity of staining in digital images was quantified using NIH ImageJ software. After background subtraction, the stained area was assessed by setting the threshold of the red channel of each image at 4 SDs below the mean pixel intensity and calculating the integrated intensity of the selected area. Bcrp1 staining area and intensity values were normalized to the mean values of negative control slides.

In vitro study of topotecan lactone and carboxylate transport by BCRP. To assess BCRP transport of topotecan lactone and carboxylate, BCRP-expressing Saos2 cells (Saos-BCRP; ref. 23) and Saos2-pcDNA control cells were incubated with topotecan at different pH values to shift the distribution of the lactone (pH 6) and carboxylate (pH 8) forms. Cells (1 × 10⁶) were incubated in DMEM (10% fetal bovine serum, 1% l-glutamine) in 10 cm² dishes for 16 h at 37°C. The medium was carefully aspirated and the cells were washed twice with PBS (pH 6 or 8). For topotecan treatment at pH 6, a topotecan stock solution in DMSO was diluted to 500 ng/mL in pH 6 PBS (37°C). For topotecan treatment at pH 8, a topotecan stock solution in 0.4 N borax solution was diluted to 500 ng/mL in pH 8 PBS (37°C). The prepared topotecan treatment solutions (8 mL) were added to each 10 cm² dish (n = 3). After incubation for 1 or 30 min at 37°C without CO₂, the treatment medium was removed, the plate was washed once with 5 mL cold PBS at pH 6 or 8, respectively, 1 mL PBS at pH 6 or 8 was added to the dishes, and the cells were detached by using a cell scraper. The cells were then ultrasonically lysed (10 s × 3). The lysate (200 μL) was added to 800 μL cold methanol for topotecan extraction to determine the intracellular drug concentration. Proteins in the remaining lysate were quantified by colorimetric assay (Dc Protein Assay; Bio-Rad Laboratories).

Statistics. In cases where two groups were compared, a Student's t test was done. For the comparison of more than two groups, a one-way ANOVA was done followed by a post hoc t test with Bonferroni correction.

Topotecan lactone plasma protein binding. The mean ± SD unbound fraction of plasma topotecan lactone was 30.1 ± 1.0% in wild-type, 29.2 ± 2.7% in Mdr1a/b^−/−, 30.9 ± 2.2% in Bcrp1^−/−, and 27.5 ± 1.9% in Mdr1a/b^−/−Bcrp1^−/− mice. One-way ANOVA showed no significant difference among these results (P = 0.10).

Bcrp1 and P-gp enhance topotecan lactone penetration into the vCSF and decrease penetration into the brain ECF. For these studies, a microdialysis probe was placed in the lateral ventricle or the brain parenchymal tissue of each mouse, and proper placement was verified by postmortem histologic examination (Supplementary Fig. S1). After a single 4 mg/kg topotecan intravenous bolus injection, the mean ± SD vCSF-to-plasma ratio of unbound topotecan lactone (AUC_u,vCSF/ AUC_u,plasma) was 3.07 ± 0.09 in wild-type, 2.57 ± 0.17 in Mdr1a/b^−/−, 1.63 ± 0.12 in Bcrp1^−/−, and 0.86 ± 0.05 in Mdr1a/b^−/−Bcrp1^−/− mice (Fig. 1). The mean ± SD in vivo probe recovery in these groups was 8.1 ± 2.4%, 7.1 ± 1.1%, 7.5 ± 1.5%, and 6.0 ± 0.6%, respectively. The CL_in-to-CL_out ratio at the BCB was consistent with the AUC ratio (Table 1).

The mean ± SD ECF-to-plasma AUC ratios (AUC_u,ECF/AUC_u,plasma) of unbound topotecan lactone were 0.36 ± 0.06 in wild-type, 0.42 ± 0.06 in Bcrp1^−/−, and 0.89 ± 0.07 in Mdr1a/b^−/−/Bcrp1^−/− mice (Fig. 2). The topotecan lactone ECF concentrations in Mdr1a/b^−/− mice were below the lower limit of detection of the assay. The in vivo probe recovery (mean ± SD) in these groups was 8.5 ± 2.4%, 5.4 ± 0.6%, and 4.3 ± 0.2%, respectively. The CL_in-to-CL_out ratio at the BBB was consistent with the AUC ratio (Table 1).

Effect of gefitinib on topotecan penetration of vCSF and ECF. After gefitinib pretreatment, no significant difference in topotecan lactone vCSF penetration (mean ± SD AUC_u,vCSF/AUC_u,plasma) was noted between wild-type and Mdr1a/b^−/−Bcrp1^−/− mice (Fig. 3; 1.29 ± 0.09 versus 1.13 ± 0.13; P = 0.147, Student's t test), suggesting that gefitinib fully inhibited Bcrp1 and P-gp function at the BCB. For the BBB, mean ± SD AUC_u,ECF/AUC_u,plasma ratios differed significantly between wild-type and Mdr1a/b^−/−Bcrp1^−/− mice (Fig. 3; 0.74 ± 0.14 versus 1.07 ± 0.03; P < 0.05, Student's t test).

Preferential transport of topotecan lactone into the vCSF. We investigated whether unbound topotecan lactone would be preferentially transported in vivo. The mean ± SD vCSF lactone-to-carboxylate AUC ratio (AUC_lactone,vCSF/AUC_{carboxylate,vCSF}) was 5.7 ± 0.8 in wild-type, 3.9 ± 0.6 in Mdr1a/b^−/−, 3.6 ± 0.5 in Bcrp1^−/−, and 0.8 ± 0.2 in Mdr1a/b^−/−Bcrp1^−/− mice. The ratios of the knockout groups were significantly different compared with the wild-type group (P = 0.003 for Mdr1a/b^−/− and Bcrp1^−/− mice and P < 0.001 for Mdr1a/b^−/−Bcrp1^−/− mice, ANOVA with post hoc t test with Bonferroni correction). The mean ± SD plasma lactone-to-carboxylate AUC ratio (AUC_{lactone,plasma}/AUC_{carboxylate,plasma}) was not significantly different between wild-type and Mdr1a/b^−/−Bcrp1^−/− mice (1.24 ± 0.38 versus 0.95 ± 0.28; P = 0.11, Student's t test). Different AUC_lactone/AUC_carboxylate ratios in vCSF but similar ones in plasma suggest the preferential active transport of topotecan lactone across the BCB by these ABC transporters.

Saos2-BCRP cells transport topotecan lactone. To determine whether BCRP is capable of specifically transporting topotecan lactone, we evaluated topotecan lactone efflux in cells overexpressing BCRP. After a 1-min incubation with topotecan lactone (stable at pH 6), the mean ± SD intracellular topotecan lactone concentration was significantly higher in Saos2-pcDNA cells than in Saos2-BCRP cells (0.063 ± 0.01 versus 0.043 ± 0.005 ng/μg protein; P = 0.05, Student's t test). A similar difference was observed after a 30-min incubation at the same pH (0.039 ± 0.003 ng/μg in Saos2-pcDNA versus 0.026 ± 0.004 ng/μg in Saos2-BCRP cells; P = 0.05). Topotecan carboxylate was not detected at pH 6. We observed no difference in intracellular topotecan carboxylate accumulation between Saos2-pcDNA and Saos2-BCRP cells at pH 8 (data not shown).

Mdr1a/b^−/− mice overexpress Bcrp1 protein at the BBB. Because topotecan lactone ECF concentrations in the Mdr1a/b^−/− mice were below the lower limit of detection, we hypothesized that Bcrp1 protein might be overexpressed in these animals, thereby increasing the overall drug efflux into the brain microvessels. A semiquantitative immunohistochemical analysis of Bcrp1 expression in the microvascular endothelium in wild-type and Mdr1a/b^−/− mice showed greater expression in the Mdr1a/b^−/− mice, with no apparent differences in vessel density (Fig. 4).

The active transport mechanisms that control topotecan penetration and pharmacokinetics in different CNS compartments have not been well described previously. By using our modified surgical procedure (25), microdialysis sampling, transporter-deficient mice, and pharmacokinetic modeling, we elucidated the roles of Bcrp1 and P-gp in the differential penetration of the CNS barriers by topotecan. Although these transporters enhanced the entry of topotecan into the CSF through the BCB, they restricted exposure of parenchymal brain to topotecan by acting as efflux transporters at the BBB. Our results may be clinically relevant in several ways. First, they may explain why topotecan is more active against CNS tumors that spread via leptomeningeal route than those located in the brain hemispheres. Second, they indicate that the use of CSF drug levels to predict CNS penetration warrants reconsideration. Finally, our results may help to predict the effects of ABC transporter inhibitors on the efficacy of chemotherapy for CNS tumors.

Our results suggest that Bcrp1 plays a major role, whereas P-gp plays a lesser role in transporting topotecan into the vCSF. Both proteins are present in the ependymal cells of the choroid plexus (17, 33, 34), and the role of P-gp in the active efflux of potentially toxic substances into the CSF has been described (33, 35). However, ours is the first study to show that Bcrp1 plays a significant role in this BCB-mediated protective mechanism. There appeared to be no functional redundancy between Bcrp1 and P-gp at the BCB in our mouse models, as animals deficient in either protein showed a correspondingly reduced topotecan vCSF penetration. Interestingly, the vCSF-to-plasma AUC ratio in Mdr1a/b^−/−Bcrp1^−/− mice was <1.0, which suggests active efflux from the vCSF compartment (31). This result may be explained by our previous finding that Mrp4 mediates efflux of topotecan from vCSF, a mechanism that opposes Bcrp1 and P-gp activity at the BCB (18).

In studies of the BBB penetration of topotecan, the loss of either transporter appeared to be compensated or even surpassed by overexpression of the other (Fig. 2). The nondeleted transporter may have been up-regulated at the BBB in these mice. In the Mdr1a^−/− model, for example, Bcrp1 mRNA is reportedly up-regulated 3-fold in the cerebral microvessels (36), which is consistent with the stronger Bcrp1 staining we observed in Mdr1a/b^−/− mice compared with wild-type mice (Fig. 4). Although Bcrp1 protein in whole-brain tissue homogenates was recently reported not to be overexpressed in the Mdr1a/b^−/− murine model compared with wild-type mice (19), the change in protein expression in the microvasculature is likely to have been masked in that study's Western blots of whole-brain homogenates. In the Bcrp1^−/− model, the ECF-to-plasma AUC ratio was comparable with that in the wild-type mice. To our knowledge, these are the first findings to suggest compensatory mechanisms for the loss of Bcrp1 function. Our results agree with those of previous studies using the Bcrp1^−/− mouse model, which found no increased BBB penetration in vivo by drugs that are Bcrp1 substrates in vitro (19, 37). Compensatory up-regulation of the nondeleted transporter would mean that neither the Mdr1a/b^−/− nor the Bcrp1^−/− model can reliably be used to distinguish the roles of P-gp or Bcrp1 in BBB transport in vivo. Therefore, results obtained using animal models with single-transporter deletions should be interpreted cautiously. However, the significantly higher ECF-to-plasma AUC ratio of unbound topotecan lactone in the Mdr1a/b^−/−Bcrp1^−/− mice compared with wild-type mice in our study (Table 1) reveals that P-gp and Bcrp1 work together in restricting topotecan penetration through the BBB.

The markedly different ECF and vCSF topotecan penetration we observed highlights the limitations of using CSF sampling or whole-brain homogenates to study CNS drug penetration. Studies using total homogenates may misinterpret drug penetration of brain parenchyma, as they include both protein-bound and unbound drug fractions from brain vessels, brain ECF, brain CSF, and brain intracellular fluid. A recently published study used the total (lactone + carboxylate) topotecan concentration values in whole-brain homogenates from knockout animals to define the roles of P-gp and Bcrp1 in brain penetration, concluding that both transporters act together at the BBB and inhibit topotecan penetration of the brain (19). However, this method was unable to define intercompartmental drug kinetics. Several studies have shown differential drug distribution in CSF compared with brain ECF. For example, Venkatakrishnan and colleagues found that CP-615,003, a GABA_A receptor agonist confirmed to be a P-gp substrate, accumulates in rat CSF 7-fold higher than in ECF (38). Similar reports about other drugs are available (39, 40). It was recently postulated that the CSF-to-plasma ratio of free drug can be used to estimate the brain parenchymal penetration of highly permeant drugs that are not efflux transporter substrates (41).

Interestingly, we observed preferential active transport in vivo of the active topotecan lactone over the inactive carboxylate. Transport of topotecan lactone by human BCRP was also confirmed in vitro. In contrast, topotecan carboxylate transport by BCRP was not detected, although at pH 8 (required to stabilize the carboxylate in our in vitro experiment), the activity of BCRP could be altered (42). These findings highlight the relevance of pharmacologic strategies that target topotecan transporters to modulate the concentration of active drug in the treated tissues. The tyrosine kinase inhibitor gefitinib provides an example. In our previous studies, gefitinib increased oral absorption and decreased systemic clearance of topotecan and irinotecan (23, 43). In the present study, we hypothesized that if gefitinib fully inhibits Bcrp1 and P-gp function at the BBB or BCB, similar topotecan AUC ratios would be observed in wild-type and Mdr1a/b^−/−Bcrp1^−/− mice treated with gefitinib. The wild-type and Mdr1a/b^−/−Bcrp1^−/− mice did have similar AUC ratios in the BCB penetration study, but in the BBB comparisons the brain-to-plasma ratio was significantly greater in Mdr1a/b^−/−Bcrp1^−/− mice. Assuming that other unknown transport mechanisms at the BBB are equal in both mouse models, this finding suggests that gefitinib can fully inhibit Bcrp1 and P-gp at the BCB but not at the BBB (see Supplementary Text for analysis). Alternatively, transport mechanisms other than P-gp and Bcrp1 may act differently either at the BCB, the BBB, or both barriers of the wild-type and Mdr1a/b^−/−Bcrp1^−/− mice. Such putative transport mechanisms appear to be active at the BCB of the Mdr1a/b^−/−Bcrp1^−/− mice, because a significant increase in topotecan vCSF penetration was noted when these animals received gefitinib (see representative plots for Mdr1a/b^−/−Bcrp1^−/− mice in Figs. 1 and 3). This finding may reflect either gefitinib inhibition of ABC transporters, the action of which opposes that of P-gp and Bcrp1 at the BCB (e.g., Mrp4; ref. 18), or the widespread presence of the organic anion transporter and organic anion-transporting peptides at the BBB and BCB in rodents (20, 44, 45). OAT3 is present in the renal basolateral membrane, where it enhances renal tubular secretion of topotecan carboxylate in the rat (46). In vitro studies have found that human OAT3 transports topotecan carboxylate (46). Because it is localized on the apical border of the choroid plexus at the BCB and on endothelial cells at the BBB in rats and mice (47), OAT3 could oppose the transport functions of Bcrp1 and P-gp but cooperate with that of Mrp4 to restrict topotecan penetration into CSF. Similarly, human OATP1B1 can mediate uptake of SN-38 (the active metabolite of irinotecan) in the liver (48). Although human OATP1B1 is not known to be expressed in the CNS, rat OATP2 (its counterpart) is found on both apical and basolateral sides of brain capillary endothelial cells and on the basolateral side of choroid plexus epithelial cells. Therefore, it may enhance topotecan penetration of the CSF, as do Bcrp1 and P-gp (49).

To summarize, topotecan influx into the vCSF and efflux from the brain parenchyma via the ECF are driven by the transporters P-gp and Bcrp1 in a manner consistent with their cellular orientation at the BCB and BBB. The blockade of P-gp and Bcrp1 activity by treatment with tyrosine kinase inhibitors reverses these effects, causing topotecan depletion in the vCSF and accumulation in the ECF. Although this study focused on a specific drug, we will use the same methods and models for investigations aimed at improving other chemotherapies for CNS tumors located in specific CNS compartments. Increasing evidence shows that histologically identical brain tumors can differ molecularly according to the site of origin of their precursors in the brain and thus may differ in their susceptibility to molecularly targeted chemotherapy agents (50). We propose that the site-specific penetration of candidate agents should be considered when selecting optimal chemotherapy treatments.

No potential conflicts of interest were disclosed.

Grant support: USPHS grants CA23099, CA21765, and GM071321-04 and American Lebanese Syrian Associated Charities.

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.

We thank Sharon Naron for editorial assistance.