Repeated cocaine administration alters the expression of genes in corticolimbic circuitry after a 3-week withdrawal: a DNA macroarray study


Address correspondence and reprint requests to Dr Shigenobu Toda, Department of Physiology and Neuroscience, Medical University of South Carolina, 173 Ashley Avenue, Suite 403, PO Box 250510, Charleston, SC 29425, USA. E-mail:


Addiction to psychostimulants elicits behavioral and biochemical changes that are assumed to be mediated by alterations of gene expression in the brain. The changes in gene expression after 3 weeks of withdrawal from chronic cocaine treatment were evaluated in the nucleus accumbens core and shell, dorsal prefrontal cortex and caudate using a complementary DNA (cDNA) array. The level of mRNA encoded by several genes was identified as being up- or down-regulated in repeated cocaine versus saline subjects. The results from the cDNA array were subsequently confirmed at the protein level with immunoblotting. Of particular interest, parallel up-regulation in protein and mRNA was found for the adenosine A1 receptor in the accumbens core, neuroglycan C in the accumbens shell, and the GluR5 glutamate receptor subtype in dorsal prefrontal cortex. However, there was an increase in TrkB protein in the nucleus accumbens core of cocaine-treated rats without a corresponding alteration in mRNA. These changes of gene expression in corticolimbic circuitry may contribute to the psychostimulant-induced behavioral changes associated with addiction.

Abbreviations used

adenosine A1 receptor


mitochondrial ATP synthetase subunit D


brain-derived neurotrophic factor


complementary DNA


cyclic AMP responsive element binding protein


dopamine D1 receptor


dorsal prefrontal cortex


gastric inhibitory polypeptide


glutamate receptor




leukocyte common-antigen related tyrosine phosphatase


mitogen-activated protein kinase


neuroglycan C


RET ligand2


nucleus accumbens core


nucleus accumbens shell


cyclic AMP dependent protein kinase A


sodium dodecyl sulfate


saline sodium citrate


ventral tegmental area.

The abuse of psychostimulants causes enduring cellular changes that are manifested as addictive behaviors, such as sensitization and craving (Nestler 2001). Once established, these behavioral modifications are long-lasting and can be elicited after extended periods of drug abstinence by environmental cues, stress or additional drug exposure (Dackis and O'Brien 2001). These enduring behavioral phenomena imply that drug-induced changes in brain function may involve long-lasting changes in gene expression.

Different temporal categories of cocaine-induced alterations of gene expression have been reported. One type of change endures for a few hours to a few weeks after discontinuing chronic cocaine administration. For example, the expression of AMPA receptor subunits and of tyrosine hydroxylase is altered in the ventral tegmental area (VTA), and these changes endure for only a few days after the last daily cocaine injection (Sorg et al. 1993; Fitzgerald et al. 1996; Churchill et al. 1999). Other examples in this category include changes typically found in the dorsal and/or ventral striatum that endure for hours to weeks, such as ΔFosB, preprodynorphin and Cdk5 (Daunais and McGinty 1996; Bibb et al. 2001; Nestler et al. 2001). These moderately enduring changes in gene expression are thought to mediate the synaptic reorganization necessary for developing the enduring psychostimulant-induced behavioral changes associated with addiction (Nestler 2001). A final temporal category of cocaine-induced changes in gene expression are those that are not altered immediately after discontinuing repeated drug administration but show a late-developing change beginning a week or more after the last injection. Again, these changes have been identified predominately in the nucleus accumbens and include Gi, Homer1bc, mGluR1 and NAC-1 (Striplin and Kalivas 1993; Cha et al. 1997; Swanson et al. 2001). This category of late-appearing alterations contains genes that are candidate mediators of the expression of addictive behaviors.

It is likely that there are many other genes with similar late-developing alterations in expression after withdrawal from repeated cocaine treatment. Corticostriatal circuitry has been implicated by neuroimaging and lesion studies in mediating the expression of addictive behaviors, and constitutes a substrate where drug-induced changes in gene expression may reside (Childress et al. 1999; McFarland and Kalivas 2001). In order to identify the changes in gene expression produced by repeated cocaine administration present a later withdrawal times, gene expression in the core and shell of the nucleus accumbens (NAcore and NAshell), dorsal prefrontal cortex (dPFC) and caudate of rat brain were examined using cDNA arrays. Significant differences shown to exist between chronic saline and cocaine treated subjects were further investigated using immunoblotting to determine if differences in mRNA were translated into alterations in protein.

Materials and methods

Animal housing

Male Sprague–Dawley rats weighing 250–350 g (Harlan, Indianapolis, IN, USA) were housed in groups of two with food and water available ad libitum. A 12-h light/dark cycle was used with the lights on at 07.00 h. All saline or cocaine injections were performed during the light cycle. All experiments were conducted in accordance with approved National Institutes of Health guidelines.

Repeated cocaine or saline treatment

All rats were acclimatized to the housing facility for 1 week prior to beginning cocaine injections using a regimen shown previously to produce behavioral sensitization (Pierce et al. 1996). Rats were treated with either saline or cocaine daily for 7 days in their home cages (15 mg/kg on the first and last day, 30 mg/kg on the intervening 5 days, intraperitoneally). After a 3-week withdrawal period, the animals were decapitated, the dPFC, NAshell and NAcore, and caudate were dissected immediately, frozen on dry ice and stored at −80°C until use. Separate sets of animals were used for the measurement of mRNA (n = 9 in each group and tissue from three individuals was pooled) and protein (n = 5–9 in each group).

RNA extraction and cDNA array

All the array procedures were performed according to the manual of Atlas cDNA Expression Arrays (BD Bioscience Clontech, Palo Alto, CA, USA) with some modifications. All paired comparisons (chronic saline versus chronic cocaine) were repeated three times for each brain region using three different sets tissue pooled from three individual subjects. Briefly, total RNA was extracted from each tissue sample with TRIzol (Invitrogen, Carlsbad, CA, USA), and was treated with 1000 units of DNase I at 37°C for 30 min (Roche, Indianapolis, IN, USA). The quality of extracted total RNA was checked by spectroscopy and agarose gel electrophoresis. An equal amount of total RNA was created by pooling tissue from three different animals for each group. 32P-labeled cDNA probe was synthesized by reverse-transcription with 3–10 µg of total RNA from each pool using CDS primer mix (BD Bioscience Clontech), [α-32P]deoxyATP (10 µCi/µL; 3000 Ci/mmol, Amersham Bioscience, Piscataway, NJ, USA), and ThermoScript avian myeloblastosis virus reverse transcriptase (Invitrogen) at 50°C for 60 min Atlas Rat 1.2 Array nylon membranes (BD Bioscience Clontech) were prehybridized with salmon sperm DNA in UltraHyb buffer (BD Bioscience Clontech) at 68°C for 3 h. The nylon arrays were then hybridized at 68°C for 16 h with the 32P-labeled cDNA probe (5–10 × 106 cpm) purified with Atlas Nucleospin Extraction Kit (BD Bioscience Clontech) to remove unincorporated radioisotope. After hybridization, the membranes were washed three times with wash solution 1 [2 × saline sodium citrate (SSC; 0.3 m NaCl/0.03 m sodium citrate), 1% sodium dodecyl sulfate (SDS)] at 68°C, 1 time with wash solution 2 (0.1 × SSC, 0.5% SDS), and once with wash solution 2 at room temperature (20–25°C) for 5 min. The hybridized membranes were exposed to a phospho-imaging screen (Molecular Dynamics, Sunnyvale, CA, USA) for 1–2 days, and the radioactive signals were visualized with a Storm 860 Phospho-imager (Molecular Dynamics). Array images were analyzed with Atlas Image software (BD Bioscience Clontech).

Array data analysis

Signal intensities of each gene were calculated by subtracting the background value from the density measurement of each gene. Local background was used for the calculation except in case of genes with high signal intensity above 20 000 in both paired membranes when an external background measurement was applied. The genes whose adjusted signal intensities were below 0.5 × the mean background value were removed from analysis. A normalization factor was determined by averaging the signal intensities of two house-keeping genes that were determined not to change from membrane to membrane (e.g. hypoxanthine-guanine phosphoribosyltransferase and glyceraldehyde 3-phosphate dehydrogenase). The genes detected commonly in all assays were selected for further statistical analysis. A ratio between cocaine and saline paired arrays of > 1.4 was used as a criterion for further evaluation using a two-tailed unpaired Student's t-test and immunoblotting (see Results section for details on how the criterion was established).


Total protein (10–80 µg) was determined from DC assay (Bio-Rad, Hercules, CA, USA) of homogenates from each brain region for each animal, and was subjected to 7.5–15% denaturing polyacrylamide gel electrophoresis and transferred to polyvinylidene difluoride membranes (Bio-Rad). In the case of neuroglycan C (NGC), 10 µg of homogenized tissue was treated with 0.05 U protease-free chondroitinase ABC (EC4.2.2.4; Calbiochem, San Diego, CA, USA) at 37°C for 1 h before applying to SDS–polyacrylamide gel electrophoresis to remove chondroitin sulfate side chains (Oohira et al. 1991). TrkB immunoreactive protein was detected using a rabbit polyclonal antibody (diluted 1: 200, Santa Cruz, Santa Cruz, CA, USA), Adenosine A1 receptor immunoreactive protein was detected using a rabbit polyclonal antibody (diluted 1 : 2000, Sigma-Aldrich, St Louis, MO, USA), glutamate receptor 5 (GluR5) immunoreactive protein was detected using a goat polyclonal antibody (diluted 1 : 200, Santa-Cruz), and NGC immunoreactive protein was detected using a mouse monoclonal antibody (diluted 1 : 2000, BD Pharmingen, San Diego, CA, USA). Immunolabeling was detected using horseradish peroxidase-conjugated IgG (diluted 1 : 10 000) followed by enhanced chemiluminescence (Amersham Bioscience). Levels of immunoreactivity were quantified by measuring the density of each band using NIH image 1.62. Density was normalized to the content of actin, and the chronic cocaine groups were compared with chronic saline groups using two-tailed unpaired Student's t-test.


Determination of the threshold for assay with cDNA array

Before starting assays using Atlas Rat 1.2 arrays, we determined the threshold for detection of differences in gene expression. The ratio of hybridization signal intensities using probes derived from the same pooled caudate RNA sample was determined among three pairs of membranes. Figure 1(a) shows the results from one pair of arrays, whereas Fig. 1(b) illustrates the mean densities for all three arrays pairs. Note the decrease of variance by averaging three pairs of arrays, especially in the region of high optical density. The average of signal intensity ratios of three comparisons ± two standard deviations was 1.00 ± 0.38. Therefore, a ratio of 1.4-fold (e.g. for cocaine/saline this would be 40% induction and 29% reduction) was used as a criterion in this study, to indicate a cocaine-induced change in gene expression. Moreover, a statistical significance for each gene was required in order to warrant further examination. Thus, while the mean ratio for all three pairs may have achieved 1.4 (e.g. see Fig. 3), this was not sufficient to proceed with immunoblotting.

Figure 1.

Determination of the threshold for assays using cDNA arrays. The same 32P-labeled cDNA probe was hybridized to three pairs of membranes. Forty per cent induction and 29% reduction cut-offs are drawn as dashed and solid lines, respectively. (a) represents one of three results. (b) shows the averaged signal intensities of three array pairs including the data in (a).

Figure 3.

Hybridization array scatterplots for specific brain nuclei. Three sample pairs of arrays were evaluated. Each array contained pooled tissue from three animals (i.e. a total of nine subjects was used for each treatment group). Array data are plotted as a function of the mean chronic cocaine signal (y-axis) versus the mean chronic saline signal (x-axis) in diamonds or circles. Forty per cent induction and 29% reduction cut-offs are drawn as dashed and solid lines, respectively, and the values in parenthesis represent the mean ratio of three different assays. Arrows indicate background cut off criteria. The genes achieving the criterion (> 1.4) for a difference between chronic saline and chronic cocaine are indicated as numbered circles, whereas the genes not achieving criterion are indicated as diamonds. Inset is a magnified representation of signals with low intensity in NAcore.

Messenger RNA expression altered by chronic cocaine administration

The Atlas Rat 1.2 array contains cDNAs encoding 1176 genes. Of these 1176 genes, 280 in NAcore, 410 in NAshell, 428 in dPFC, and 400 in caudate showed a discernible signal in all three comparisons. Figure 2 shows a representative comparison of array images for NAcore. Array assays identified a number of potential cocaine-responsive genes in each brain region. Figure 3 shows that the changes at the level of mRNA of 5 genes in NAcore, 2 genes in NAshell, 1gene each in dPFC and caudate achieved the 1.4 ratio criterion (see Methods).

Figure 2.

A representative comparison of gene expression patterns between chronic saline- and chronic cocaine-treated rats in NAcore. Equal amounts of total RNA from three cocaine- and three saline- treated animals were pooled, labeled with 32P and hybridized to separate arrays. The genes highlighted in diamonds are up-regulated in cocaine treated subjects, whereas the gene in the square is down-regulated according to the criteria described in Methods. Higher magnifications of the boxed areas are shown in the right panels. 1, ATP5H; 2, RETL2; 3, LAR; 4, A1; 5, GIP.

Protein expression altered by chronic cocaine administration

Of the genes reaching the 1.4 ratio criterion, antibodies were available for adenosine A1 receptor (A1), leukocyte common-antigen related phosphatase (LAR), gastric inhibitory polypeptide (GIP), RET ligand2 (RETL2), TrkB, id3, and neuroglycan C (NGC); therefore these genes were selected for subsequent immunoblot analyses. In addition, the increase in GluR5 mRNA in the dPFC of cocaine treated subjects nearly reached criterion (i.e. two of three pairs had a ratio > 1.4) and was also evaluated with immunoblotting. In consideration of the possibility of presynaptic and/or postsynaptic localization of candidate proteins, we examined the levels of candidate proteins in not only the brain regions where alteration of mRNAs were observed but also other brain regions where receive the afferent projections from the former regions. Four of these proteins had altered levels as a consequence of repeated cocaine, including A1 receptor, NGC, GluR5 and the p95 and p145 isoforms of TrkB (Fig. 4). A1 protein was significantly up-regulated in NAcore (Table 1), but not altered in NAshell or ventral tegmental area (data not shown). Chondroitin sulfate side chain-free NGC protein (p120) was also significantly up-regulated in NAshell (Table 1). In spite of the reduction in mRNA, both the p95 and p145 TrkB isoforms did not show a significant reduction in dPFC (Table 1). However, both isoforms of TrkB were significantly up-regulated in the NAcore where mRNA was unchanged (Table 1). There was no difference in either TrkB isoform between cocaine and saline treatment groups in the NAshell (data not shown). In the case of GluR5, the protein was significantly up-regulated in dPFC in parallel with a non-significant change in mRNA (Table 1). Sufficient amounts of GluR5 could not be detected in the NAcore to quantify. In contrast with mRNA levels, RETL2 and LAR protein contents were not significantly different between the saline and cocaine treatment groups. Although abundant mRNA was measured, immunoreactive signals corresponding to GIP or id3 could not be detected in the NAcore or NAshell, respectively.

Figure 4.

Representative immunoblots as post-hoc confirmation of the changes in mRNA level detected with cDNA array. S, chronic saline-treated animals; C, chronic cocaine-treated animals. (a) A1 in NAcore, (b) p145 TrkB in NAcore, (c) p95 TrkB in NAcore, (d) GluR5 in dPFC, (e) p120 NGC in NAshell.

Table 1.  Summary of percent differences in mRNA and protein content between animals treated 3 weeks earlier with one week of daily cocaine (CC) or saline (CS) injections
CC/CS (n = 3)CS (n = 5–9)CC (n = 5–9)
  1. ND, not detectable; NP, no antibody available. *p < 0.05 comparing saline with cocaine treatment groups using a two-tailed paired (mRNA) or unpaired (protein) Student's t-test. **p < 0.005.

Na core
 ATP5H171 ± 21*NPNP
 RETL2p52159 ± 5**100 ± 17 94 ± 19
 p37 100 ± 17 93 ± 11
 GIP 62 ± 1**NDND
 TrkB p145106 ± 8100 ± 16204 ± 24*
 p95 100 ± 16170 ± 22*
 A1140 ± 10*100 ± 17155 ± 12*
 LAR p150155 ± 12*100 ± 22 94 ± 19
 p85 NPNP
 NGC145 ± 7**100 ± 11161 ± 21*
 ID3 58 ± 4**NDND
 TrkB p145 71 ± 7*100 ± 16 91 ± 14
 p95 100 ± 19 82 ± 16
 GluR5159 ± 26100 ± 13155 ± 20*
 nur77 71 ± 4**NDND


Gene array and immunoblotting were used to demonstrate that 3 weeks following chronic cocaine administration, multiple alterations in mRNAs and their encoding proteins were present in rat corticolimbic system. The genes detected with the cDNA array that were altered to criterion in the cocaine treated subjects can be grouped in terms of their functions as follows: (i) receptors of neurotrophic factors, TrkB, RETL2; (ii) neurotransmitter receptors, GluR5, A1; (iii) proteoglycans, NGC, LAR; (iv) transcription factors, Nur77, id3; and (v) a xenobiotic metabolic enzyme, ATP5H. Consistent with involvement in cocaine-induced neuroadaptations, many of these genes have identified roles in neural plasticity (see Discussion below; Bäckman et al. 2001; Chen et al. 2001; Van Lieshout et al. 2001). Many of the alterations in gene expression as a consequence of repeated cocaine administration were confirmed in the NA. This is consistent with a distinctive involvement of the NA in chronic cocaine-induced neuroadaptations (Pierce et al. 1996; Vanderschuren and Kalivas 2000; Di Ciano and Everitt 2001; Thomas et al. 2001). Whether these alterations are late-developing and long-lasting or not should be examined in the future; however, it is highly likely that at least some of these alterations are late-developing, since TrkB mRNA is not altered in frontal cortex 18 h after repeated cocaine treatment (Nibuya et al. 1995). Also, it needs to be clarified whether these alterations are generalized in the animals sensitized in a different environment such as in an experimental box (Ostrander et al. 1998) or in animals sensitized using a different treatment paradigm such as cocaine self-administration.

Adenosine A1 receptor in nucleus accumbens core

Both A1 mRNA and protein were up-regulated in NAcore, but not in NAshell of cocaine-treated subjects. Previously reported interactions between A1 receptors and psychostimulant-induced neuroadaptions indicate that the up-regulation may be important. A1 receptor agonists inhibit the expression, but not the development of methamphetamine-induced behavioral sensitization in rats (Shimazoe et al. 2000). Also, caffeine is an A1 antagonist and prevents the extinction of cocaine-seeking behavior (Kuzmin et al. 1999).

A1 is known to localize at pre- and postsynaptic sites in rat forebrain (Ochiishi et al. 1999) and to modulate signaling through dopamine D1 receptors (Mayfield et al. 1999). The stimulation of A1 inhibits adenylyl cyclase via Gi/o. This contravenes D1 signaling via Gs to stimulate adenylyl cyclase. In addition, the A1 and D1 receptors form a heterodimer that favors the low-affinity binding conformation of D1 receptors, probably due to guanosine 5′-triphosphate (GTP)-independent uncoupling of Gs (Ferréet al. 1996; Ferréet al. 1998; Ginés et al. 2000). Given the up-regulation of A1 receptors, it is paradoxical that repeated cocaine elicits an increase in the electrophysiological response to D1 agonist administration in the nucleus accumbens (Henry and White 1995) and that the adenylyl cyclase and cyclic AMP protein kinase A (PKA) signaling cascade is up-regulated (Nestler 2001). While the up-regulation in PKA signaling may not endure beyond a week or two after discontinuing repeated treatment, the increase in D1 responsiveness parallels the time course of behavioral sensitization (Henry and White 1995). However, previous studies also found no significant alteration of D1 binding after repeated cocaine treatment (Mayfield et al. 1992; Unterwald et al. 1994). Thus, the extant data indicate that the up-regulated electrophysiological response to D1 receptor stimulation in the nucleus accumbens of cocaine treated subjects may arise via post-translational modification of proteins in the D1 signaling cascade, such as altered phosphorylation state. Another possibility is that the reported late-developing decrease in protein content for Gi may remove inhibitory tone on adenylyl cyclase, including tone provided by stimulating A1 receptors (Striplin and Kalivas 1993).

Given the opposite effects of A1 and D1 receptor stimulation, the up-regulation of A1 receptors in NAcore may be a compensatory phenomenon in response to the reduced Gi and increased D1 signaling (Striplin and Kalivas 1993; Henry and White 1995). Alternatively, following 1 week of withdrawal from chronic cocaine, the potency of adenosine to presynaptically inhibit glutamate release via A1 was reduced due to an augmented uptake of adenosine (Manzoni et al. 1998). Thus, the up-regulation of A1 receptors in NAcore could be a compensatory response to reduced extracellular adenosine.

Neuroglycan C in nucleus accumbens shell

NGC is a transmembrane chondroitin sulfate proteoglycan originally isolated from the developing rat brain. Its expression is restricted to the central nervous system, and in mature rat the protein level is reduced by almost half of the peak level in the developing brain (Watanabe et al. 1995). Although the function of NGC is still unknown, it has an EGF-like motif on its extracellular domain, therefore; NGC is likely a signaling molecule (Oohira et al. 2000). It is reported that NGC immunoreactivity is rich in the soma and the thick stems of the Purkinje cell dendrites, where Purkinje cells form synapses with climbing fibers in the postnatal developing cerebellum. These data suggest that NGC may play a role in selective synaptogenesis as a receptor for an unidentified ligand-like molecule (Aono et al. 2000). It has been suggested that the NAshell may be particularly important in mediating the acute rewarding affects of psychostimulants (Di Chiara et al. 1999), posing the possibility that the neuroadaptation in NGC may be involved in the altered reward thresholds associated with chronic cocaine abuse (Koob 1999).

Glutamate receptor 5 in dorsal prefrontal cortex

We also found that both mRNA and protein levels of GluR5 were up-regulated in dPFC after repeated cocaine administration, although only the latter alteration was statistically significant. The expression of GluR5 mRNA is regulated in an activity-dependent manner in developing brain and the content of GluR5 is greatly reduced in adult rat (Wüllner et al. 1997; Kidd and Isaac 1999). Since we could not detect a protein signal for GluR5 in NAcore even using a large amount of protein (80 µg), GluR5 protein in dPFC is likely to be localized in the soma and dendrites of glutamatergic pyramidal cells or GABAergic interneurons rather than being transported to axon terminal fields such as the NAcore. This interpretation is coincident with previous reports showing dense labeling for the GluR5/6/7 subunits of kainate (KA) receptors in both the soma and dendritic regions of pyramidal neurons in neocortex, but little labeling in axon terminals (Huntley et al. 1993). GluR5 forms a heteromeric complex with other KA receptor subunits and different subunit compositions have been shown in vitro to affect ligand binding and channel conductance (Cui and Mayer 1999; Muñoz et al. 1999; Paternain et al. 2000). For example, GluR5-KA2 heteromer exhibits a reduced affinity for the ligand, while an edited GluR5 [GluR5(R)]-KA2 heteromer generates channels with unitary conductance larger than for the corresponding homomeric channels (Swanson et al. 1996; Lerma et al. 2001). Of all the subunits capable of forming a KA receptor/ion channel (e.g. GluR5, 6, 7 and KA1, 2), GluR6 mRNA is expressed in highest density in cortex (Muñoz et al. 1999) and is a candidate heteromeric partner with GluR5. Unfortunately, GluR6 and KA1, 2 were not spotted on the membrane used in this study and GluR7 was spotted but not detectable. Although GluR5 is assumed to play a role in adult neocortex since RNA editing of GluR5 is enhanced in the temporal cortex of patients with epileptic episodes (Kortenbruck et al. 2001), there is no direct information regarding the functional properties of GluR5–GluR6 heteromers or how up-regulation of GluR5 by cocaine may affect these properties.

TrkB in nucleus accumbens core

TrkB proteins are receptors for brain-derived neurotrophic factor (BDNF) and neurotrophin-4/5, and one isoform (145 kDa) has an intracellular kinase domain while two truncated isoforms (95 kDa) lacking the catalytic domain have been identified (Middlemas et al. 1991). The isoform of TrkB containing the catalytic domain makes a homodimer in the presence of ligand which causes self-phosphorylation and subsequent phosphorylation of downstream molecules, notably mitogen-activated protein kinases (MAPK) (Barbacid 1995). In contrast, the truncated forms are most often proposed to inhibit signaling through the extended TrkB isoform. For example, the presence of p95 TrkB in glia, especially following neuronal damage, indicates that it may function as a ligand-trapping system to regulate extracellular BDNF concentration (Beck et al. 1993; Frísen et al. 1993). Also, truncated TrkBs may function as dominant negative receptors by dimerizing with p145 TrkB expressed in the same cells and inhibiting auto-phosphorylation of tyrosine (Eide et al. 1996; Gonzalez et al. 1999). However, truncated TrkBs have also been proposed to induce neurite outgrowth and promote elongation of distal dendrites (Haapasalo et al. 1999; Yacoubian and Lo 2000).

Regardless of the exact function of the truncated TrkBs, protein levels of both the p95 and p145 isoforms were significantly up-regulated in the NAcore without a parallel increase in mRNA. This indicates that the increase in TrkB in the NAcore may be located predominantly on afferent terminals that arise from nuclei with projections to the NAcore such as the VTA, ventral pallidum, amygdala or hippocampus. However, it is unlikely that the elevated TrkB would be on afferents from the dPFC given the measured reductions in mRNA in this cortical region. Consistent with a view of cocaine-induced neuroadaptations in TrkB signaling in the nucleus accumbens and its afferent projections, previous studies have revealed the importance of BDNF-stimulated signaling cascades in the VTA and NAcore in the development of behavioral sensitization to cocaine (Berhow et al. 1996; Horger et al. 1999; Pierce and Bari 2001). However, this is the first demonstration of a change in the BDNF signaling cascade that occurs beyond the discontinuation of repeated drug administration. Thus, while the increase in TrkB is consistent overall with a persistent impact on BDNF signaling by repeated cocaine administration, it remains to be determined what other components of the signaling cascade may also be altered.

Methodological concerns

Previous studies using cDNA arrays have applied arbitrary cut-off ratios that are typically above 1.5 for screening altered gene expression (Cadet et al. 2001; Freeman et al. 2001). In the present study, the ratio 1.4 was employed since 0.38 was two standard deviations from unity as determined from three replicated pairs of arrays hybridized to the same striatal RNA sample (see Methods). However, even with a relatively stringent criterion, cDNA arrays warrant additional post-hoc assays to verify the relevance of any observed differences, such as RT-PCR or in situ hybridization at the level of mRNA or immunoblotting of the translated protein (Ang et al. 2001; Hata et al. 2001). In the present study, it was found that the alterations in mRNA detected with the cDNA array are not necessarily coincident with differences in protein between saline and cocaine treatment groups (see Table 1). Similarly, it has been reported that even in yeast, the correlation between mRNA and protein levels is about 0.6, probably due to post-transcriptional modifications (Ideker et al. 2001). Given the higher order of complexity in the brain, it is perhaps not surprising that concordance between protein and mRNA levels does not always occur. For example, a change in mRNA may alter protein content in the axon terminals in a nucleus receiving afferents from the neurons in the nucleus where the changes in mRNA were identified.

Another potential concern is that the number of genes identified in the present study as altered by repeated cocaine is surprisingly small. After 3 weeks of withdrawal from repeated cocaine administration, the number of altered genes would be expected to be less than the number detected after acute treatment (Cadet et al. 2001; Freeman et al. 2001). However, given the profound enduring behavioral effects of repeated cocaine administration, such as sensitization and relapse to drug-taking behavior, more changes in gene expression might have been expected. Explanations for this possible discrepancy include: (i) A number of genes that may have been affected by repeated cocaine administration were below the detection limit or were not included in the arrays. (ii) Changes in gene expression may have occurred in brain nuclei not examined in the present study. While this is undoubtedly true, based upon the extant clinical and preclinical literature, the regions examined, especially the prefrontal cortex and nucleus accumbens, are thought to undergo neuroadaptations critical to the development and expression of addictive behaviors (Childress et al. 1999; McFarland and Kalivas 2001; Nestler 2001). A final concern is that it is possible to detect false-positive or false-negative gene expression in the array. However, confirmation by replication makes this less of a concern. Likewise, the possibility of cross-hybridization cannot be ruled out. In this study, GIP might be an example of a false-positive or a cross-hybridization artifact because we could not detect immunoreactivity corresponding to GIP in the NAcore in spite of a very strong signal on the cDNA array.


Using the cDNA array technique, significant alterations in gene expression were observed 3 weeks after discontinuing chronic cocaine administration, especially in the nucleus accumbens. Subsequent immunoblotting verified the increase in the A1 receptor in the NAcore, NGC in the NAshell and GluR5 in the dPFC. In addition, TrkB protein in NAcore was found to be elevated in the absence of a parallel change in mRNA. These alterations may contribute to the persistent cocaine-induced behavioral neuroadaptations associated with addiction.


This work was supported in part by MH-40817 and DA-03960 (PWK). The authors thank Dr M. Valeria Gonzalez-Nicolini and Lan Jin for technical advice and discussions.