Cocaine hydrochloride was provided by the National Institute on Drug Abuse. SKF-525A, cimetidine, α-tocopherol, 3(2)-tert-butyl-4-hydroxyanisole (BHA), and deferoxamine (DFO) were obtained from Sigma-Aldrich.

The AF5 neural progenitor cell line was maintained as previously described [19]. The AF5 cell line is homogenous and over long-term culture maintains growth stability [19], differentiation capacity [19,20], and an intact p53 function [19,21]. Notably, the total length of the cell cycle of AF5 cells (24 h) is close to that in the rodent ventricular zone (VZ) (18 h at E15 or later) [22]. 5 × 10³ cells/well were plated in 96-well plates for cell proliferation and cytotoxicity assays, and 4 × 10⁴ cells/well in 12-well plates for immunostaining and reactive oxygen species (ROS) measurement 24 h prior to use. Primary human fetal CNS cells (ScienCell Research Laboratories) were from ∼20-wk human fetal cerebral cortexes, obtained in accordance with principles embodied in the Declaration of Helsinki (Code of Ethics of the World Medical Association) and were cultured at 37 °C, 5% CO₂ using the recommended human cell media obtained from ScienCell Research Laboratories, except that human neural progenitor cells were maintained in DMEM/F12 (1:1, Invitrogen) supplemented with N2 supplement (R & D Systems), 20 ng/ml EGF (R & D Systems), 20 ng/ml bFGF (R & D Systems), 5 μg/ml heparin (Sigma-Aldrich), 100 U/ml penicillin G, 100 μg/ml streptomycin, and 50 μg/ml gentamicin (Sigma-Aldrich). Purity of respective types of CNS cells was evaluated by immunocytochemistry using antibodies against specific cell markers. For evaluation of neural progenitors, neurospheres were first dissociated into single cells using papain (Worthington Biochemical Corporation) for 10 min at 37 °C, and plated onto laminin/poly-L-ornithine-coated slides. Once attached, cells were fixed with 4% paraformaldehyde and evaluated by immunocytochemistry. A2B5 progenitor cells differ from neural progenitor cells in that they are committed to a glial cell lineage, eventually differentiating to type-2 astrocytes or oligodendrocytes, whereas neural progenitor cells can become both neurons and glial cells. The human neurons used in this study were MAP-2 positive, and immunoreactive for glutamate (44% ± 3%) and GABA (69% ± 5%), but not for tyrosine hydroxylase (TH), choline acetyltransferase (ChAT), or 5-hydroxytryptamine (5-HT) (unpublished data).

AF5 cells were treated with cocaine at various concentrations, and cell proliferation was measured using CyQUANT cell proliferation assay (Invitrogen). Cocaine-induced cytotoxicity was evaluated by lactate dehydrogenase (LDH) release from the cytosol into the medium after exposure of AF5 cells to various concentrations of cocaine for 24 h, according to the manufacturer's protocol (Roche Applied Science). For single-stranded DNA immunostaining, cells were fixed with methanol/PBS (6:1) for 24 h at −20 °C followed by incubation in formamide at 70 °C for 5 min. Fixed cells were immunostained with mouse anti-single-stranded DNA monoclonal antibodies (1:10, Chemicon) and fluorescein-conjugated anti-mouse IgM (1:200, Jackson Immunoresearch). Data were represented as: (number of single-stranded DNA-positive nuclei/number of DAPI-positive total nuclei) × 100%.

Cultures were synchronized by maintaining in serum-free medium for 24 h, followed by exposure to 0, 10, or 100 μM cocaine in serum-containing medium for 24 h. AF5 cells were analyzed by flow cytometry on a FACS Calibur flow cytometer (Becton Dickinson). Proportions of cells in the G1, S, and G2/M phases of the cell cycle were determined by using ModFit LT software (Verity Software House).

AF5 cultures were treated with 20 μM 5-bromo-2′-deoxyuridine (BrdU) (BD Biosciences) in the presence/absence of cocaine for 24 h, fixed with 95% ethanol, and permeabilized with 2 N HCl. Nonspecific staining was blocked with 5% normal goat serum and 0.1% Nonidet P-40 in PBS for 20 min at room temperature. Cells were double-labeled with monoclonal mouse anti-BrdU (1:100, BD Biosciences) and polyclonal rabbit anti-phospho-histone H3 (1:200, Upstate Biotechnology) overnight at 4 °C. After washing, secondary antibodies Alexa Flour 594 goat anti-mouse and Alexa Flour 488 goat anti-rabbit (1:500, Invitrogen) were applied, and nuclei were labeled with DAPI. Data were obtained by dividing numbers of nuclei positive for BrdU or phospho-histone H3 by total numbers of nuclei.

Total RNA was extracted from AF5 cultures using RNA STAT-60 (TEL-TEST). cDNA microarray analysis was performed using a mouse developmental cDNA microarray containing 15k clones derived from early Kargul libraries using procedures for processing as previously described [23]. z-Score transformation was employed to compare array data between different treatments [24]. z-Score transformation allows analysis of array data independent of the original pixel intensities and can be used in calculation of p-values for significance estimates. To calculate gene expression changes after cocaine treatment, z scores were converted to z ratios, which represent fold-like changes for each gene.

Primary human fetal CNS cells were fixed in 4% paraformaldehyde in PBS for 10 min and processed for immunostaining [20]. The following primary antibodies were used: rabbit anti-nestin (1:200, Chemicon); mouse anti-A2B5 (supernatant, clone 105, 1:3, ATCC); mouse anti-MAP2 (1:500, BD Biosciences); mouse anti-OX42 (1:200, BioLegend); and mouse anti-GFAP (1:200, Sigma-Aldrich). Cells were developed using Alexa Fluor 488 goat anti-rabbit or mouse secondary antibody (1:500, Invitrogen), and nuclei were labeled with DAPI.

Reverse transcription was performed as described previously [20]. To quantify the cyclin A2 transcript, quantitative real-time reverse transcription (RT)-PCR using the DNA Engine Opticon Fluorescence Detection System (MJ Research) was performed using SYBR green according to the manufacturer's protocol. The primer sequences and sizes of the PCR products for rat cyclin A2 were ATATGAAGAGGCAGCCAGACA (sense), AGGCAGCTCCAGCAATAAGCG (antisense), 483 bp; and for human cyclin A2 were GCAAACAGTAAACAGCCTGCG (sense), TCAACTAACCAGTCCACGAGG (antisense), 386 bp. The results were analyzed using Opticon software. Relative expression was determined by normalizing to 18S ribosomal RNA (Ambion) using 1.0 for the control.

Pregnant Sprague-Dawley rats (Charles River Laboratories) received cocaine at early (E13 and E14), middle (E15 and E16), or late periods (E17 and E18) of neocortical neurogenesis. Rats received 20 mg/kg cocaine (intraperitoneally [IP]) twice at an interval of 12 h followed by 50 mg/kg BrdU (Sigma-Aldrich), IP, 24 h after the last injection of cocaine. Rats were euthanized by CO₂ inhalation 2 h after BrdU. Control animals received physiological saline. All animal procedures were performed according to the “Guide for the Care and Use of Laboratory Animals,” according to an animal protocol approved by the Institutional Animal Care and Use Committee of the NIDA Intramural Research Program.

For measurements of cocaine concentrations, prefrontal cortex and peri-ventricular region were dissected from fetal rat brains at the early period of neurogenesis (E15). Tissues from all fetuses of each pregnant dam were pooled to become one individual sample. Detection and quantification of cocaine in fetal rat brain was accomplished utilizing a modification of a previously published method [25]. The method employed ultrasonic homogenization of brain tissue in pH 4.0 sodium acetate buffer followed by solid phase extraction. Extracts were derivatized with N,O-bis-[trimethylsilyl]trifluoroacetamide (BSTFA). Cocaine was separated by capillary gas chromatography and simultaneously quantified by electron impact mass spectrometry in selected ion mode. The calibration curve for cocaine was linear to 15,000 ng/g of brain, and the limit of quantification was 50 ng/g. This method provided sufficient analytical sensitivity to allow quantification of cocaine in small amounts of tissue.

For cyclin A expression studies, tissues (prefrontal cortex and peri-ventricular region) from fetal rat brain were dissected. Tissues from three fetal rats were pooled for each individual assay to obtain sufficient material. RNA and proteins were extracted with RNA STAT-60 (TEL-TEST) and lysis buffer, respectively. For BrdU labeling, coronal brain sections were labeled with monoclonal mouse anti-BrdU (1:200, BD Biosciences) and polyclonal rabbit anti-Ki67 (1:500, Novocastra Laboratories) overnight at 4 °C and visualized using Alexa Flour 594 goat anti-mouse and Alexa Flour 488 goat anti-rabbit antibodies (Invitrogen). BrdU labeling index [(number of BrdU-positive nuclei/number of Ki67-positive nuclei) × 100%] was calculated in the regions with 160 μm width (red rectangles in Figure S1) that locate on one-third and two-thirds of the cerebral cortex defined by arrowheads at the superior sagittal sinus (SSS) and the causal pole of the internal capsule in Figure S1. Six brain sections from both brain hemispheres were examined in each animal. The outer half of the VZ (the half farthest from the ventricle) contains a layer of closely packed cells positive for Ki67 and BrdU S-phase incorporation. The outer edge of this epithelium-like layer was used to define the boundary between VZ and subventricular zone (SVZ).

Western blotting was performed as previously described [20] using antibodies to CDK2 and pRb (BD Biosciences), α-tubulin (Sigma-Aldrich), phospho-CDK2 (Thr-160), phospho-eIF2α (Ser-51), and eIF2α (Cell Signaling), cyclin A, ATF1, ATF2, ATF3, ATF4, phospho-CREB (Ser-133), CREB, JunB, JunD, c-Jun, c-Fos, p21, and p27 (Santa Cruz Biotechnology). Immunoreactive bands were densitometrically quantitated using Kodak Image Station 440 CF.

AF5 cells were transfected with a plasmid encoding cyclin A (pRc/CMV-CycA; gift of P. Hinds, Harvard Medical School, Boston, Massachusetts, United States of America [26]) or a control plasmid (pcDNA3.1; Invitrogen). Briefly, one million AF5 cells in suspension (100 μl) were mixed with 2 μg of either plasmid, and electroporation was performed using nucleofection protocol T-20 (Amaxa Biosystems). Immediately after nucleofection, cells were plated in six-well plates for the Western blot analysis and 96-well plates for the CyQUANT cell proliferation assay.

Endogenous ROS were measured by incubating AF5 cells with 100 μM 2′, 7′-dichlorofluorescein diacetate (DCFH-DA) (Sigma-Aldrich) during the last 20 min of indicated treatments. The treated AF5 cells were washed, dissolved with 1% Triton X-100 in PBS, and fluorescence was measured at an excitation wavelength of 485 nm, and an emission wavelength of 530 nm using a fluorescence microplate reader.

All values were expressed as means ± standard error of the mean (SEM). Mean values were compared using the Student's t-test for two-sample comparisons, or analysis of variance (ANOVA) followed by Newman-Keuls tests for multiple comparisons as indicated in the legends for Figures 1–7, S2, and S6. The criterion for statistical significance was p < 0.05. It should be noted that this study includes a large number of statistical tests; thus, it is possible that some differences, even though statistically significant, are due to chance. For all differences found to be statistically significant by Student's t-tests, the differences were additionally assessed by nonparametric Mann-Whitney U tests, the results of which are listed in Table S1. p-Values obtained by both tests were similar. Also, as compared to the experiments performed in vitro, the variability between animals was relatively large; therefore, scattergram plots showing the data for individual animals are included as Figures S4 and S7.

Log-phase cultures of AF5 cells were treated with varying concentrations of cocaine (1–100 μM) for 24 h. The number of cells in each condition was then measured and compared to the number at the initiation of the treatment. As shown in Figure 1A, the number of AF5 cells in untreated cultures doubled in 24 h, while numbers of cells in cocaine-treated cultures were significantly reduced in a dose-dependent manner as compared to controls. When exposure of AF5 cells to 100 μM cocaine was extended to 6 d, the inhibitory effect was even more pronounced (Figure S2). Because cocaine substantially inhibited proliferation of AF5 cells after only 24 h exposure, we employed this exposure duration for subsequent experiments.

The effect of cocaine was not due to an increase in cell death, as we did not observe a change in either extracellular LDH activity (Figure 1B) or numbers of apoptotic nuclei immunoreactive for single-strand DNA (Figure 1C). We also did not observe any morphological changes indicative of necrosis or apoptosis in cocaine-treated AF5 cells. Therefore, as has been reported for several other cell types [11,27,28], cocaine inhibits proliferation of AF5 neural progenitor cells.

We next examined the cell cycle distribution of cocaine-treated cells by FACS (Figure 1D). Cocaine resulted in a dose-dependent increase in cells in G1 phase and a reduction in the number of cells in S phase, indicating that cocaine suppresses the G1-to-S phase transition (Figure 1D). To further clarify the transition suppression, cell populations going through S phase were monitored by BrdU incorporation over 24 h. Both 10 μM and 100 μM cocaine significantly decreased the percentage of BrdU-positive cells that had entered S phase, supporting the notion that cocaine interferes with the G1/S cell cycle transition (Figure 1E). The percentage of mitotic cells was low (<3%) as measured by phospho-histone H3 immunocytochemistry, and was not affected by cocaine (Figure 1E).

To identify molecules that could mediate the cocaine-induced G1/S transition impairment, we used a microarray that contains 93 cell cycle-related genes including 16 G1/S phase transition controllers (cyclin A2, C, D1, D2, D3, E1, E2, G1, CDK2, p12CDK2-AP1, CDK4, p27Kip1, p57Kip2, p18INK4c, PITSLRE/CDK11p58, and p53). Of these, cocaine (10 and 100 μM, 24 h) significantly down-regulated only cyclin A2 (Figure 2A). The decrease in cyclin A2 expression by 10 or 100 μM cocaine was confirmed by quantitative real-time RT-PCR (Figure 2B).

To characterize the time course of cocaine-induced down-regulation of cyclin A, we treated AF5 cells with 100 μM cocaine for 6 d and found that the cyclin A protein level, as measured by Western blotting, had started to decrease by day 1, continued to decline at day 3, and finally resulted in an undetectable amount of cyclin A protein by day 6 (Figure S3). These data show that cocaine-induced inhibition of AF5 cell proliferation is correlated with down-regulation of cyclin A.

To determine whether our findings in AF5 cells are relevant to primary cells, we measured cyclin A2 mRNA in several types of primary human fetal CNS cells obtained from ∼20-wk (second trimester) human fetal cerebral cortexes. As measured by quantitative real-time RT-PCR, in vitro cocaine exposure (100 μM, 24 h) significantly decreased cyclin A2 mRNA level in both human neural progenitor cells (>95% of cells are Nestin-positive and A2B5-negative, unpublished data) and A2B5+ progenitor cells (>90% of cells are A2B5-positive and Nestin-positive, unpublished data), whereas cyclin A2 mRNA was not altered in neurons or microglia (Figure 3A). In contrast, cocaine increased the cyclin A2 transcript in human astrocytes (Figure 3A).

The cyclin A protein level was also significantly decreased by cocaine (10 and 100 μM, 24 h) in both human neural and A2B5+ progenitor cells (Figure 3B and 3C). Cocaine caused a maximum down-regulation of cyclin A protein level at 10 μM in human neural progenitor cells (Figure 3B), but decreased expression of cyclin A in a dose-dependent manner in A2B5+ progenitor cells (Figure 3C). These data verify that cocaine down-regulates cyclin A in both types of progenitor cells derived from human fetal cerebral cortexes.

We next examined cyclin A expression in fetal rat brains exposed to cocaine in utero. Neocortical neurogenesis occurs within two proliferative strata of the embryonic cerebral wall, which is adjacent to the ventricle. Neocortical neurogenesis starts at E12 and ends at E19 in the rat [29]. During this process, the VZ appears first and is followed by the SVZ. To investigate the time-frame within which neocortical development is susceptible to prenatal cocaine exposure, we examined three different time periods: the early neurogenesis period from E13 to E15, the middle period of neurogenesis from E15 to E17, and the late neurogenesis period from E17 to E19. Pregnant animals received cocaine according to the injection schedule shown in Figure 4A, and the frontal cortex of developing fetuses was dissected as shown in Figure 4B.

To examine cocaine concentrations in the fetal neocortex under our injection schedule and for comparison to our in vitro studies, tissue concentrations of cocaine were analyzed at the early period of neurogenesis after the second cocaine administration. Cocaine concentrations in fetal neocortex reached at least 30 μM (9,812 ng/g) 0.5 h after injection, dropped to ∼2 μM (median value of 594 ng/g) at 1 h, and gradually declined to ∼0.2 μM (median value of 63 ng/g) at 6 h (Figure 4C). Thus, the injection schedule used produced exposure of fetal brains to maximum concentrations of cocaine within the range of doses that we found to inhibit proliferation of progenitor cells in vitro.

Down-regulation of cyclin A2 mRNA was seen in prefrontal cortex of the developing fetuses when cocaine was injected at either early or middle periods of neurogenesis (Figures 4D and S4A). Cyclin A2 mRNA was not, however, changed by injections during the late period of neurogenesis (Figures 4D and S4A). Similarly, cyclin A protein was also significantly decreased by cocaine in both early and middle neurogenesis periods, but not during the late period of neurogenesis (Figures 4E and S4B). This result suggests that cocaine down-regulates cyclin A expression only during earlier periods of neocortical neurogenesis that involve active proliferation of neural progenitor cells in the VZ.

Since cocaine causes down-regulation of cyclin A in fetal brains, we also examined cell cycle progression of neural progenitors in the VZ and SVZ in fetal brains in utero exposed to cocaine. Pulse labeling with BrdU was used to quantify cortical progenitor cells that had entered S phase during a period of 2 h, whereas Ki67 immunocytochemistry was used to monitor the total fraction of progenitors that are in any phase of the cell cycle except for G₀ [30]. Cocaine decreased the number of BrdU-labeled progenitor cells in the VZ during both the early and middle neurogenesis periods (Figures 4F, 4G, S4C, and S5), suggesting that cocaine inhibits the proliferation of progenitor cells in the VZ.

In contrast to the findings in the VZ, BrdU positive progenitor cells in the SVZ were not changed by cocaine (Figures 4F, 4H, S4D, and S5). Also, since cortical germinal zones in the late period of neurogenesis comprise SVZ only, cocaine did not change BrdU labeling during this period (Figures 4H, S4D, and S5). Taken together, these data suggest that cocaine promotes cyclin A down-regulation and cell proliferation inhibition both in vitro and in vivo.

Western blotting confirmed that cocaine (100 μM, 24 h) decreases cyclin A2 protein in AF5 cells (Figure 5A). Cyclin A induces a conformational change in CDK2, and thereby permits activation of CDK2 through phosphorylation of Thr-160 catalyzed by CDK-activating kinase (CAK) [31]. Activated CDK2 phosphorylates the downstream molecule pRb, which in turn promotes the expression of genes required for progression from G1 to S. Western blotting showed that cocaine significantly decreased the phosphorylated forms of CDK2 and pRb (Figure 5A), indicating that cocaine-induced down-regulation of cyclin A results in a hypoactive state of the downstream cell cycle signaling pathway.

To demonstrate a causal relationship between cyclin A down-regulation and cocaine-induced inhibition of AF5 cell proliferation, we attempted to compensate for cyclin A down-regulation by gene transfer using an expression vector encoding cyclin A (pRc/CMV-CycA).

To examine the appropriate timing of transfection, we first examined the time course of cyclin A expression after either cocaine treatment or vector transfection. Quantitative real-time RT-PCR showed that cyclin A2 mRNA was decreased as early as 6 h after 10 or 100 μM cocaine, and the down-regulation lasted up to 24 h (Figure 5B). Cyclin A protein was significantly decreased 12 h after 10 or 100 μM cocaine with nearly maximum effects (64%) at 12 h for 10 μM cocaine, and a 48% decrease at 24 h for 100 μM cocaine (Figure 5C). Since pRc/CMV-CycA transfection (with ∼80% transfection efficiency) also showed the maximum level of overexpression at 24 h post-transfection (unpublished data), we exposed cells to 100 μM cocaine immediately after electroporation of the CMV-Cyc A vector, and measured cyclin A protein and cell proliferation 24 h later. As shown in Figure 5D, cyclin A transfection counteracted the cocaine-induced down-regulation of cyclin A, as well as the inhibition of proliferation caused by cocaine. These data suggest that cocaine-induced down-regulation of cyclin A contributes to the proliferation inhibition seen in cocaine-exposed AF5 cells.

We next examined the levels of several transcription factors involved in regulation of the cyclin A2 promoter activity. Western blot analysis showed that cocaine (100 μM, 3 h) significantly up-regulated only ATF4 among a total of nine candidate transcription factors tested, including ATF1–4, CREB, JunB, JunD, c-Jun, and c-Fos (Figure 6A). The phosphorylation status of CREB, which is involved in activation of cyclin A transcription at G1/S [32], was not affected by 100 μM cocaine at 3 h (Figure 6A). In addition, cocaine did not change levels of p21 and p27, CDK inhibitors that decrease transcription of cyclin A2 (Figure 6A) [33,34].

An increase in ATF4 protein occurred as early as 1 h after 100 μM cocaine exposure, and reached the maximal level at 3 h (Figure 6B). Cocaine at concentrations higher than 1 μM resulted in dose-dependent induction of ATF4 (Figure 6B). The phosphorylated form of the alpha subunit of translation initiation factor 2 (eIF2α) is known to promote ATF4 induction [35]. Indeed, we found that 10 μM cocaine significantly increased the phosphorylated form of eIF2α 0.5 h after exposure to cocaine, preceding induction of ATF4 (Figure 6C).

The eIF2α-ATF4 pathway is activated by PERK, an endoplasmic reticulum (ER) stress sensor protein [35,36]. The ER stress pathway has been shown to be activated by a variety of chemicals or pathological stress including oxidative ER stress [35]. Notably it has been demonstrated that the cytochrome P450 dependent N-oxidative pathway is responsible for generation of ROS and glutathione (GSH) depletion during cocaine biotransformation in the liver [37]. We therefore hypothesized that ROS generation caused by the N-oxidative metabolism of cocaine may trigger activation of the eIF2α-ATF4 pathway.

Accordingly, we first examined whether cocaine induces ROS production in progenitor cells. As shown in Figure 6D, 10 μM cocaine caused a significant increase in ROS 30 min after treatment. Cocaine at 100 μM caused ROS generation even earlier, with a significant increase 15 min after the exposure.

Because the rise of cocaine concentrations over 10 μM lasts for less than 1 h in fetal brains after cocaine injection (Figure 4C), AF5 cells were next treated with 30 μM cocaine for 30 min followed by incubation in cocaine-free medium for 2.5 h to examine ATF4 protein levels. We found that ATF4 was significantly up-regulated by this transient 30 min exposure to cocaine (Figure S6). Thus, exposure to cocaine for a period of only 30 min can generate ROS sufficient to initiate ER stress, and result in up-regulation of ATF4 2.5 h later.

Pretreatment with the cytochrome P450 inhibitors SKF-525A or cimetidine, drugs that have been shown to potently block N-oxidative metabolism of cocaine, completely blocked cocaine-induced ROS formation (Figure 6E), confirming that N-oxidative metabolism of cocaine is involved in ROS formation in AF5 cells. Further, ROS generation appears to be the source of cocaine-induced ER stress, as both SKF-525A and cimetidine also completely inhibited cocaine-induced ATF4 up-regulation and cyclin A down-regulation (Figure 6F and 6G).

Finally, we tested effects of SKF-525A and cimetidine on the inhibition of cell proliferation by cocaine. Both drugs significantly diminished cocaine-induced proliferation inhibition (100 μM for 24 h) (Figure 6H). On the other hand, the lipophilic free radical scavengers α-tocopherol and BHA and the iron chelator DFO (each 50 μM) did not prevent cocaine-induced proliferation inhibition (unpublished data). These data indicate that cytochrome P450-dependent ROS formation occurring during cocaine metabolism is responsible for both cocaine-induced proliferation inhibition and cyclin A down-regulation.

To determine whether P450 inhibitors can block cocaine-induced proliferation inhibition in neural progenitor cells in the developing neocortex, pregnant rats at the early period of neurogenesis (E13–E15) were pretreated with 100 mg/kg cimetidine IP 1 h before each cocaine administration. Cimetidine is known to cross the placenta [38–40]. Cimetidine itself did not affect neural progenitor cell proliferation, survival, density, or fetal mortality (Figures 7A, 7B, S7A, S7B, and S8A–S8E). As shown in Figures 7A and S7A, pretreatment of pregnant rats with cimetidine resulted in recovery of the cocaine-induced decease in BrdU-positive progenitor cells in the VZ. As expected, no differences in BrdU-positive progenitor cells were observed in the SVZ (Figures 7B and S7B).

To determine whether the protection afforded by cimetidine was due to the recovery of cocaine-induced down-regulation of cyclin A and mediation by ER stress, the effects of cimetidine on expression of ATF4 and cyclin A were also measured in prefrontal cortex of cocaine-treated fetuses. Pretreatment of pregnant rats with cimetidine significantly inhibited the cocaine-induced up-regulation of ATF4 and the down-regulation of cyclin A (Figures 7C, 7D, S7C, S7D). These results suggest that blockade of cocaine N-oxidative metabolism by the P450 inhibitor cimetidine reverses cocaine-induced proliferation inhibition of neural progenitor cells in the VZ through normalizing cocaine-induced oxidative ER stress and consequent cyclin A down-regulation.

Using the AF5 neural progenitor cell line, we determined that cocaine treatment for 24 h causes proliferation inhibition at concentrations higher than 1 μM. Similar results were reported by Hu et al. [11] showing that cocaine ranging 1–100 μM (7 d) suppresses growth of human neural precursor cells, as measured by thymidine incorporation. Nevertheless, this group also did not detect proliferation inhibition by 1 μM cocaine for 24 h, although a significant effect was produced by treatment with 1 μM cocaine for 5 d [11]. The findings of Hu et al. [11] are therefore consistent with our data.

Poon et al. [10] found that 1 μM cocaine treatment for 30 min induced late-onset cell death (72 h later) in differentiated human neuronal progenitor cells. We did not observe cell death in cocaine (1–100 μM)-treated AF5 cells (Figure 1B and 1C) or in primary human fetal neural and A2B5+ progenitor cells (unpublished data) 24 h after treatment, although we did not examine cell death at later time points such as 72 h. In agreement with our findings, cocaine treatment (1 pM–100 μM) for 7 d did not induce cell death in undifferentiated human neural precursor cells [11]. Under certain conditions, cocaine can selectively induce cell death in neurons without affecting viability of other types of CNS cells [41]. The differentiated neuronal progenitor cells employed by Poon et al. [10] may therefore be relatively more sensitive to an effect of cocaine in promoting cell death as compared to undifferentiated neural progenitor cells.

Although cocaine has been shown to inhibit DNA synthesis as measured by thymidine incorporation in human neural precursor cells [11], the mode of action of cocaine on cell cycle progression has not previously been determined. We found that cocaine causes cell proliferation inhibition by interfering with the G1-to-S transition. Although one report suggested a disturbance in chromosome segregation induced by 2,500 μM cocaine treatment (14 or 24 h) of mouse oocytes [13], our flow cytometric and immunocytochemical analyses show that cocaine (10–100 μM) does not have any effect on the mitotic phase at least in our system (Figure 1D and 1E). A cDNA microarray identified cyclin A2 as a candidate molecule related to cocaine-mediated G1/S phase arrest. Cocaine causes specific down-regulation of cyclin A2 in AF5 cells, primary neural and A2B5+ progenitor cells, and fetal brains exposed to cocaine in utero. Further, compensating for cocaine-induced cyclin A down-regulation by gene transfer counteracted the growth-suppressing action of cocaine in AF5 cells (Figure 5D).

Reversal of cocaine-induced proliferation inhibition by cyclin A transfection is not, in itself, proof that the effect of cocaine on neural progenitor cell proliferation is caused by down-regulation of cyclin A. Nevertheless, among known modulators of the G1-to-S transition, microarray analysis identified only cyclin A2 as being down-regulated by cocaine. The cyclin A transfection results are consistent with the hypothesis that cocaine decreases proliferation via down-regulation of cyclin A. Moreover, the fact that the ROS-induced ER stress consequently promoted transcriptional down-regulation of cyclin A is further evidence that reductions in cyclin A signaling are responsible for decreased proliferation. These findings, taken together, suggest that cyclin A down-regulation constitutes at least one molecular mechanism by which cocaine causes dysfunction of neural progenitor cells.

Cocaine-induced cyclin A2 down-regulation was observed in human neural progenitor and A2B5+ progenitor cells, but not in human neurons and microglia. In contrast to progenitor cells, the cyclin A2 transcript was increased by cocaine in human astrocytes. Although we do not know the mechanism underlying the proliferative effect of cocaine on astrocytes, it is worth noting that cellular stress inhibits proliferation in CNS progenitors, whereas it activates proliferation in astrocytes, leading to gliosis in the brain [42]. Cocaine thus has at least two pharmacological actions on cellular proliferation in a cell type-specific manner. Moreover, in utero cocaine exposure down-regulated cyclin A expression in progenitor cells in the VZ, but not in the progenitor cells in the SVZ. Although the reason for this specific mode of action is unknown, variations in expression and metabolic capacities of cytochrome P450s, plasma membrane permeability of cocaine, and variations in cellular stress response mechanisms could be involved. Future studies exploring the activity of cytochrome p450s and oxidative stress in different types of CNS cells will address this issue.

Screening of molecules that can regulate the promoter activity of cyclin A2 demonstrated that ATF4 is specifically up-regulated by cocaine exposure; this effect of cocaine occurs 3 h prior to cyclin A2 down-regulation. Upon ER stress such as ER oxidation, the ER sensor protein PERK phosphorylates eIF2α, subsequently promoting translational activation of ATF4 [35,36]. ATF4 promotes not only up-regulation of genes involved in redox control (e.g., glutathione biosynthesis) [35], but also promotes down-regulation of cyclin A during G1-to-S progression, which can lead to cell cycle arrest [43]. Thus, ER oxidation is capable of causing cell cycle arrest via the eIF2α-ATF4 pathway.

In addition to promoting dopamine auto-oxidation, cocaine can itself get through the cell membrane in its nonprotonated form, where it produces ROS via N-oxidative metabolism catalyzed by cytochrome P450 at the ER [37]. Cocaine is first N-demethylated to norcocaine [44,45], followed by oxidation to N-hydroxynorcocaine [44–47] by cytochrome P450 and flavin-containing monooxygenases. N-hydroxynorcocaine is further converted to norcocaine nitroxide by one-electron oxidation, also through cytochrome P450 activity [44,48]. Norcocaine nitroxide can be rapidly reduced to N-hydroxynorcocaine again by flavoproteins such as cytochrome P450 reductase and FAD-monooxygenase [44,48]. Redox cycling between norcocaine nitroxide and N-hydroxynorcocaine is responsible for generation of ROS such as superoxide anion and hydrogen peroxide [37,44]. In addition, depletion of NADPH, an essential cofactor for maintenance of reduced glutathione, during the futile redox cycling of N-hydroxynorcocaine/norcocaine nitroxide is able to disrupt the homeostatic effect of cellular glutathione [48], possibly also contributing to cellular oxidative stress.

Indeed, we found that cocaine induces endogenous ROS accumulation as early as 15 min after treatment of AF5 cells (Figure 6D). Further, the P450 inhibitors SKF-525A and cimetidine, which have been shown to inhibit N-oxidative metabolism of cocaine, blocked cocaine-induced endogenous ROS generation, translational activation of ATF4, down-regulation of cyclin A, and proliferation inhibition in AF5 cells (Figure 6E–6H). In contrast, lipophilic free radical scavengers and the iron chelator DFO did not block cocaine-induced inhibition of proliferation, suggesting that lipid peroxidation and iron-mediated ROS production (Fenton reaction) in mitochondria are not primarily involved in cocaine-induced ROS production in AF5 cells. We therefore conclude that cocaine biotransformation by microsomal cytochrome P450 is the source of cocaine-induced ROS generation, which promotes cell cycle arrest via ER stress-induced cyclin A down-regulation (Figure 6E–6H).

In cultured human neural precursor cells, cocaine-induced (1–100 μM for 7 d) cell proliferation inhibition has been suggested to be related to increased expression of p21, a major transcriptional target of p53, as well as down-regulation of nuclear antigen (PCNA), an essential DNA replication factor [11]. Activation of the p53-p21 pathway has been implicated in hyperoxia-induced inhibition of G1/S progression, through suppression of CDK activity and subsequent repression of cyclin A2 transcription and inactivation of PCNA [34,49]. Increased expression of p21 was not observed 24 h after exposure to cocaine in our study; however, we did detect an increase of p21 48 h after exposure to cocaine (unpublished data). These data suggest that the p53-p21 pathway may be involved in longer-term effects of cocaine, but not in the short-term action of cocaine that promotes down-regulation of cyclin A2 within 24 h.

Cocaine has been shown to inhibit protein synthesis in rat fibroblasts, associated with cocaine-induced inhibition of cell proliferation [28]. One of the cellular adaptive responses to oxidative ER stress is inhibition of translation, which may explain the inhibition of protein synthesis observed in cocaine-treated fibroblasts. Also, cocaine acts as a local anesthetic by inhibiting sodium influx into cells. An in vitro study, however, has shown that cocaine-induced proliferation inhibition in lymphocytes was not related to sodium channel blockade [50].

Neural progenitor cells in ventricular proliferating zones are multipotent, having the potential to differentiate to neurons destined for different layers of the cortex. Alterations in cell cycle dynamics of neural progenitor cells have been linked to the production of various sizes of cortical areas and to the thickness of cortical layers [51,52]. In addition, disruption of neural progenitor proliferation can result in microcephaly [53], which has been reported in infants born to cocaine-abusing mothers [3]. We demonstrated here that oxidative ER stress-mediated down-regulation of cyclin A contributes to cocaine-induced inhibition of neural progenitor cell proliferation in the developing neocortex (Figure 8). Because cell cycle duration is regulated mainly via G1 phase and/or the G1-to-S transition during embryonic neurogenesis [54,55], cocaine-induced G1/S cell cycle arrest mediated by down-regulation of cyclin A may have a substantial impact on proliferation of progenitor cells, as well as on ensuing migration and differentiation (Figure 8).

Cocaine inhibited proliferation of AF5 cells in concentrations as low as 10 μM for 24 h of exposure. As shown in Figure 4C, the concentrations of cocaine in the developing neocortex following the second IP administrations of 20 mg/kg cocaine peaked higher than 30 μM within 30 min. A short duration of exposure of the fetal brain to cocaine may in fact be deleterious since, as shown in Figure S6, exposure of AF5 cells to 30 μM cocaine for 30 min significantly up-regulated the expression of ATF4 at 3 h. Thus, exposure to 30 μM cocaine for 30 min is sufficient to increase endogenous ROS to a degree that activates the ATF4 signaling pathway. Therefore, it is conceivable that exposure to cocaine for fairly short durations could initiate the molecular mechanisms underlying cocaine-induced inhibition of neural progenitor cell proliferation.

Maternal cocaine exposure has been associated with uterine vasoconstriction [56], which could cause low oxygenation of the developing fetal brain. A number of studies indicate, however, that cocaine-induced vasoconstriction is unlikely to mediate impairments in brain development [57–59] as measured by brain structure and DNA synthesis. Also, Lidow and Rakic [60] have shown that neurotransmitter receptors expressed in neural progenitor cells, including dopaminergic, serotonergic, and adrenergic receptors, are related to cell proliferation stimulation or inhibition in the developing primate occipital lobe. Therefore, it is possible that cocaine-induced accumulation of monoamine neurotransmitters and activation of corresponding receptors also contribute to the inhibition of cell cycle progression by cocaine, especially in vivo. Moreover, in addition to ROS induced by N-oxidation of cocaine, auto-oxidation of monoamines could contribute to ATF4-mediated cyclin A down-regulation in fetal brains. Nevertheless, our data demonstrated that cytochrome P450-dependent N-oxidation of cocaine promotes cyclin A2 down-regulation by causing ER stress, which constitutes at least one mechanism underlying cocaine-induced inhibition of ventricular progenitor cell proliferation in the developing neocortex.

We have shown that pretreatment with the P450 inhibitor cimetidine effectively abolished both the cocaine-induced inhibition of neural progenitor cell proliferation in the VZ of the developing rat brain, and cyclin A down-regulation (Figures 7A, 7D, S7A, and S7D). This effect of cimetidine appeared to be mediated by prevention of N-oxidative metabolism of cocaine and consequent ER stress, as cimetidine also prevented the cocaine-induced increase in ATF4 (Figures 7C and S7C). These effects of cimetidine in the developing rat brain were consistent with the efficacy of cimetidine in blocking effects of cocaine on the AF5 cell line in vitro.

Cimetidine belongs to a class of drugs called histamine H2 receptor antagonists, and is often prescribed for the treatment of gastroesophageal reflux diseases and peptic ulcer diseases. Cimetidine crosses the placenta by passive diffusion [38] and has been allocated to pregnancy category B by the FDA, which means that cimetidine is not expected to be harmful to the fetus. There are conflicting reports on anti-androgenic effects of cimetidine in animals exposed in utero [61–64]; however, no anti-androgenic effects of cimetidine have so far been reported in human pregnancy.

We demonstrated that giving cimetidine (100 mg/kg IP twice at an interval of 12 h) to pregnant rats at an early period of neurogenesis did not affect proliferation (Figures 7A, 7B, S7A, and S7B), survival (Figure S8A and S8B), or density (Figure S8C and S8D) of neural progenitor cells in the VZ or SVZ. We also did not observe an increase in perinatal mortality in cimetidine-treated animals (Figure S8E). The expression of histamine H₂ receptors in the cerebral cortex is very low [65,66]; therefore, H₂ receptor activity is not likely to be involved in the ability of cimetidine to decrease cocaine-induced changes in the fetal brain.

The question naturally arises as to whether cimetidine or a similar drug could be employed to prevent the adverse effects of cocaine on brain development. Presumably, the manner in which this would be accomplished would be that women of child-bearing age with a history of cocaine abuse, and who are at risk for subsequent cocaine abuse, would be asked to take cimetidine as a prophylactic measure. Several issues would have to be addressed before this could be done. Most importantly, the possibility that cimetidine increases the systemic toxicity of cocaine, perhaps by interfering with cocaine metabolism and lengthening the half-life of cocaine, would have to be considered. Another question to be addressed is the half-life of the preventive drug used that would be necessary to achieve a satisfactory level of patient compliance; it is possible that a drug with a very long half-life, e.g., a depot preparation, would be needed. Also, although the mechanism that is described here appears to at least contribute to the developmental toxicity of cocaine, it is not necessarily the entire cause of cocaine's adverse effects on development. Thus, the long-term efficacy of cimetidine (or alternative drugs) on brain development and function, and efficacy in a second larger species, will also need to be examined.

During neocorticogenesis, the two germinal compartments VZ and SVZ are composed of heterogeneous populations of neural progenitor cells. It is difficult to isolate, define, and examine specific sub-types of neural progenitor cells derived from primary fetal cortex. It is also problematic to examine regional cellular and molecular parameters in vivo in tissue sections derived from embryonic rat brains at different ages following intrauterine cocaine administration. The AF5 neural progenitor cell line used in this study may not have properties identical to those of cortical VZ progenitors; however, the AF5 cell line was able to reveal the molecular mechanisms (e.g., free radical-induced ER stress), which are involved in cocaine-induced proliferation inhibition of neural progenitor cells. The AF5 cell model not only allowed us to identify these molecular mechanisms, but also was remarkably predictive of the in vivo findings. Nevertheless, the expression and metabolic capacities of cytochrome P450 in neural progenitor cells of developing rodent and human brains are not identical, which to same extent limits extrapolation of our findings to cocaine-exposed human fetal brains.

A second issue involves the dose of cimetidine that is required to prevent the effects of cocaine. The dose of cimetidine used to block the histamine H2 receptor in rodents is ∼2.5–10 mg/kg [67]; however, doses of cimetidine up to 100 mg/kg are used to inhibit the activity of cytochrome P450 in rodents [68,69]. We therefore used 100 mg/kg cimetidine for the present study. On a mg/kg basis, this dose is about ten times higher than the human therapeutic dose. Thus, it is not clear whether cimetidine, or any similar drug, would be effective in preventing the adverse effects of cocaine on neural progenitor cells in human cocaine abuse. It will therefore be essential to determine whether clinically relevant doses of cimetidine have the capacity to inhibit the adverse effects of exposure of the human fetus to cocaine, as well as whether there are other P450 inhibitors that are more effective.

Understanding the molecular mechanisms by which in utero cocaine exposure causes proliferation inhibition of neural progenitor cells is important for developing prevention and therapeutic strategies against long-lasting neurological and behavioral dysfunction caused by exposure of the developing fetus to cocaine. Future research might be focused on exploring the molecular and biochemical mechanisms involved in cellular functions such as differentiation and migration changed by cocaine in neural progenitor cells, determining the expression and metabolic capacity of cytochrome P450s in various subtypes of cortical and subcortical progenitor cells, and investigating whether oxidative ER stress is involved in other disorders that have been shown to be related to P450-dependent oxidative metabolism of cocaine, such as immunosuppression [70–72], hepatocyte injury [37,69,73,74], and cardiotoxicity [75].