INTRODUCTION

Although drug abuse and addiction has been studied extensively for decades, the underlying molecular mechanisms are still not well understood. The contributions of animal models to our understanding of drug abuse are undeniable but, at the same time, many features of human drug addiction (eg spontaneous drug self-administration, and the pattern and length of drug use) are difficult, if not impossible, to recapitulate in experimental models. Fortunately, the human post-mortem brain is amenable to neurochemical analysis, providing a complementary strategy for examining the effects of drug abuse at a molecular level in human drug addicts. The advent of high-throughput microarray technology enables investigators to simultaneously examine changes in gene expression across hundreds of genes or even entire genomes. In the last several years, microarray analysis of post-mortem brain has been successfully applied to the study of gene expression in the human drug abuser (Albertson et al, 2004; Bannon et al, 2005; Lehrmann et al, 2003; Lewohl et al, 2000; Mayfield et al, 2002; Tang et al, 2003). Such studies are likely to increase our understanding of the consequences and complexities of addiction.

In a previous microarray study investigating gene expression changes in the nucleus accumbens of human cocaine abusers, we identified a striking downregulation of numerous myelin-related genes (Albertson et al, 2004). The decreased myelin basic protein (MBP), proteolipid protein (PLP), and myelin-associated oligodendrocyte basic protein (MOBP) gene expression seen in cocaine abusers (Albertson et al, 2004; Bannon et al, 2005; Lehrmann et al, 2003; Tang et al, 2003) may be a molecular correlate of white matter changes identified through neuroimaging studies (Bartzokis et al, 2002; Bartzokis et al, 1999b; Bartzokis et al, 1999a; Chang et al, 1997; Chang et al, 1999; Lim et al, 2002). We have posited (Albertson et al, 2004) that the excessive extracellular levels of the neurotransmitter dopamine induced by cocaine abuse may retard the conversion of immature oligodendrocytes into mature myelin-producing cells, a process thought to be dopamine-sensitive (Bongarzone et al, 1998; Howard et al, 1998). If this hypothesis is correct, then similar changes in myelin gene expression should be seen in the nucleus accumbens of abusers of other illicit drugs known to increase extracellular dopamine levels. Heroin, through the activation of μ opioid receptors located on inhibitory GABAergic interneurons in the midbrain, disinhibits mesolimbic dopamine neurons and may elicit increased levels of extracellular dopamine in the nucleus accumbens (Wise et al, 1995; Xi et al, 1998). The recent resurgence of heroin abuse and its effects on nucleus accumbens dopamine make heroin-abusing subjects a potentially important cohort for post-mortem analysis of gene expression and comparison to cocaine abusers.

In the present study, we employed high-throughput microarray technology to identify gene expression changes in the nucleus accumbens of chronic heroin abusers. To our knowledge, this is the first such analysis reported. Data were analyzed independently and in relation to our previously reported study on human cocaine abusers in an effort to determine which alterations in gene expression are drug specific and which may be common to many or all drug addictions.

MATERIALS AND METHODS

Cocaine Study

A portion of the cocaine-related microarray data set has been previously described (Albertson et al, 2004; Bannon et al, 2005). Briefly, RNA from the nucleus accumbens of 10 cocaine abusers and 10 matched controls (listed in Table 1) were hybridized to Affymetrix u133A and u133B microarrays and data were analyzed in pairs using a Wilcoxon signed rank test (p⩽0.05). For a transcript to be considered significantly different between groups, it had to be present in all samples and changed in the same direction in the majority of pairs (⩾6 out of 10 pairs).

Table 1 Study Subject Characteristics
Full size table

Tissue Acquisition and Subject Characterization

For this study, brain specimens were collected as part of the routine autopsy process under a protocol approved by Wayne State University's Human Investigation Committee, as previously described (Albertson et al, 2004; Bannon et al, 2005). Heroin users (n=7) exhibited a positive blood toxicology for heroin and/or its metabolites (eg 6-monoacetyl morphine, morphine, morphine glucuronide). Control subjects (n=7) were matched pairwise with heroin users for gender, race, age, and brain pH (Table 1). Of the 14 subjects, one control and two heroin subjects tested positive for moderate levels of alcohol (⩽0.11 g/dl), but did not exhibit common signs of chronic alcohol abuse. All heroin and control subjects tested negative for all other common drugs of abuse including cocaine, barbiturates, benzodiazepines, and phencyclidine.

Sample Preparation

Coronal sections measuring 2–3 cm were taken throughout the rostrocaudal extent of the basal ganglia. The nucleus accumbens (core and shell) was dissected as previously described (Bannon et al, 1992), flash frozen in isopentane cooled in liquid nitrogen, and stored at −80°C. Determination of the pH of an adjacent piece of cortex was used as an initial indicator of sample integrity, with all post-mortem samples exhibiting pH values within the desired range (pH⩾6.1; Kingsbury et al, 1995) (Table 1). RNA was extracted and characterized as previously described (Albertson et al, 2004). Briefly, frozen tissue was rapidly homogenized by Polytron in 10 × w/v Tri Reagent (Sigma, St Louis, MO), RNA was chloroform/isopropanol extracted, precipitated, and reconstituted in sodium citrate (Ambion, Austin, TX). Contaminating DNA was eliminated using a Qiagen RNeasy Mini Kit (Valencia, CA). The Agilent 2100 Bioanalyzer (RNA Nano LabChip Kit, Agilent Technologies, Palo Alto, CA) was used to verify RNA abundance and sample quality (as indicated by a 2 : 1 ratio of 28S to 18S rRNA and the absence of DNA and degraded RNA species).

Microarray Experiments and Data Analysis

Affymetrix oligonucleotide arrays (Santa Clara, CA) were used in all studies. Initial examination of the glyceraldehyde-3-phosphate dehydrogenase (GAPDH) 3′/5′ ratios generated by cRNA hybridization to test chips provided an additional measure of RNA quality as well as a measure of the efficiency of the RT-PCR and in vitro transcription reactions. According to Affymetrix quality control parameters, this ratio should be <3.0 (see Table 1). For subsequent full-scale analysis, samples were hybridized to both human u133A and u133B arrays, representing over 39 000 transcripts. All sample labeling, hybridization, and scanning followed the Affymetrix GeneChip® Expression Analysis Technical Manual (www.affymetrix.com). The complete absence in all subject samples of dopamine transporter and tyrosine hydroxylase transcripts (data not shown) supports the contention that nucleus accumbens mRNA was derived predominantly, if not exclusively, from cells intrinsic to this nucleus.

Data were analyzed with the Affymetrix Microarray Suite 5.0 software package as previously described (Albertson et al, 2004; Bannon et al, 2005). Images were scaled for signal intensity to account for any differences between hybridization efficiencies. Subjects were analyzed in pairs, comparing each heroin sample with its matched control. Significant differences between subject pairs were calculated using the Wilcoxon signed rank test (p⩽0.05); marginal calls were considered nonsignificant. For purposes of the present study, transcripts increased or decreased in the majority (⩾4 of 7) of pairs were considered differentially expressed. Functional groups were created using annotation information provided by Affymetrix. A post hoc group-wise analysis of all genes identified on the microarray as residing within the category of synaptic transmission was performed with the non-parametric Mann–Whitney U-test (p⩽0.05).

RT-PCR

RNA from all subjects was used for verification of the microarray data. Reverse transcription (RT) was performed (Sensiscript RT Kit, Qiagen) with random hexamer primers, whereas subsequent PCR used sequence-specific primers (amplicons: synaptogyrin 3 (SYNGR3) 1651–1849, synaptotagmin 1a (SYT1) 2098–2171, synapsin 2b (SYN2b) 3258–3453, synaptic vesicle protein 2A (SV2a) 2339–2410, prodynorphin (PDYN) 2265–2342, and β-actin 1396–1578 or 2366–2631). PCR was performed in the LightCycler version 3.3 with the Qiagen SYBR Green PCR Kit (Roche, Indianapolis, IN) as described previously (Albertson et al, 2004). Equivalent amounts of RNA from each subject were pooled to create standard curves (input RNA 1–16 ng) that were assayed in parallel with replicate samples (5 ng RNA) from individual subjects. A linear standard curve was generated and all samples fell within the range of the curve. For sample normalization, individual transcript values were divided by the subject's β-actin values determined using the same RT reaction. β-Actin transcript levels did not differ between heroin abusers and control subjects, as determined by either RT-PCR (p=0.85) or microarray (p=0.10). Differences in transcript abundance between heroin abusers and matched controls were assessed by Wilcoxon signed rank tests.

RESULTS

Overview of Sample Characterization and Microarray Data

The integrity of the human post-mortem brain samples used in this study was evident from appropriate tissue pH values (Table 1), and the quality of RNA extracted from these samples was confirmed using spectrophotometric and electropherographic analyses (data not shown). Furthermore, the appropriately low 3′/5′ GAPDH ratios obtained by microarray analysis (Table 1) reflected both the quality of input RNA and the efficiency of the RT-PCR and in vitro transcription reactions used to generate cRNAs for microarray hybridizations (for more details, see Materials and methods).

Following sample hybridization to human u133A and u133B microarrays (representing over 39 000 transcripts) and data analysis as previously described (Albertson et al, 2004; Bannon et al, 2005), it was determined that 49.7% of the transcripts represented on the arrays were expressed (‘present’) in the nucleus accumbens of all subjects. Global visualization of transcript abundances, as determined in the present study and our previous cocaine study (Albertson et al, 2004), revealed that most transcripts did not differ significantly in abundance between heroin abusers (Figure 1a) or cocaine abusers (Figure 1b) and their matched drug-free controls (represented as drug/control signal log ratio). Pairwise analyses between heroin abusers and their matched controls revealed that a total of 1050 transcripts were differentially expressed across the majority of subject pairs (for a complete list of these transcripts, see Supplementary Table 1). Those transcripts that are annotated encode proteins thought to be involved in a wide variety of cellular processes including synaptic transmission, structural organization, metabolism, energy pathways, apoptotic processes, stress and immune responses, transcriptional regulation, and other processes. Within these broad functional categories, however, few patterns of gene expression changes were evident.

Figure 1
figure 1

Altered transcript abundance in the nucleus accumbens of human cocaine and heroin abusers. The signal log ratio (log base 2) distribution of all transcripts detected in the majority of subject pairs was plotted after binning averages into groups of 0.05 (a:n=18 055, b:n=17 688). Values falling to the left of zero indicate a downregulation of transcript abundance in drug abusers, whereas those to the right of zero are indicative of increases. The mean signal log ratio of myelin-related transcripts (MBP, MOBP, and PLP) for each study was plotted in relation to the normal distribution of all present transcripts, and is represented by black arrows (a:mean=0.13; b:mean=−0.81). The mean signal log ratio for all presynaptic machinery transcripts (listed in Figure 2a) is represented by gray arrows (a:mean=−0.39; b:mean=−0.001). Note that the majority of transcripts are unchanged in either heroin abusers (a) or cocaine abusers (b), and that changes in myelin-related and presynaptic machinery transcripts are drug specific.

Full size image

Myelin-Related Transcripts

In our previous analysis of chronic cocaine abusers (Albertson et al, 2004), we had identified a significant downregulation of multiple transcripts derived from the myelin-related genes MBP, MOBP, and PLP (see Figure 1b), a finding confirmed by RT-PCR and immunohistochemical experiments. In the present study, examination of human heroin abusers failed to reveal any significant change in the expression of this group of genes (Figure 1a). There was no directional trend toward a change in the abundance of any of these myelin-related transcripts individually (data not shown), suggesting that the effects seen on myelin-related gene expression were selective for cocaine abusers in comparison to heroin abusers.

Presynaptic Machinery Genes

Perhaps the most striking finding arising from the microarray analysis of the nucleus accumbens of the heroin cohort was the decreased abundance of a number of transcripts encoding presynaptic machinery proteins (Figure 1a). Genes encoding proteins involved in multiple aspects of neurotransmitter release, including vesicle storage, release, and recycling, were impacted, as illustrated in detail in Figure 2a. No corresponding downregulation of gene expression was seen in cocaine abusers (Figures 1b and 2a), supporting the specificity of this microarray finding.

Figure 2
figure 2

Decreased expression of presynaptic machinery genes in human heroin abusers. (a) Specific presynaptic machinery genes, their putative functions, and Affymetrix probe ID numbers are listed. Median and individual signal log ratios for the seven heroin–control pairs are shown (as are median SLR from the cocaine study as a comparison). (b) Quantitative real-time RT-PCR was used to determine the abundance of four representative transcripts (synapsin 2b, synaptotagmin 1a, synaptic vesicle protein 2a, and synaptogyrin 3). All transcript levels were divided by β-actin levels, which did not differ between groups (see Materials and methods). Significant heroin-related decreases in transcript levels were validated in the case of synapsin 2b (p=0.046), synaptic vesicle protein 2a (p=0.028), synaptogyrin 3 (p=0.027), but not synaptotagmin 1a (p=0.128). Median percent decrease across all subject pairs is also shown for each of the four transcripts.

Full size image

In subsequent experiments, we employed quantitative real-time PCR to determine the abundance of four representative transcripts (namely synapsin 2b, synaptotagmin 1a, synaptic vesicle protein 2a, and synaptogyrin 3; Figure 2b). When examined as a group, the abundance of these four transcripts was significantly decreased in heroin subjects as compared to matched controls (Z=−3.758, p=0.001, Wilcoxon signed rank test), and the log 2 ratios for each subject pair as determined by microarray and RT-PCR were significantly correlated (p=0.049, Pearson's correlation). When considering each transcript separately, however, a significant correlation between pairwise microarray and PCR data was seen only in the case of synaptogyrin 3 (p=0.040), with correlations for the other transcripts failing to reach significance. Nevertheless, in spite of the limited power afforded by the modest sample size, PCR analysis did validate the central microarray finding of heroin-related decreases in transcripts encoding synaptogyrin 3 (Z=−2.213, p=0.027), synapsin 2b (Z=−1.859, p=0.046), and synaptic vesicle protein 2a (Z=−2.197, p=0.028) (Figure 2b). In the case of synaptotagmin 1a, the decreased transcript abundance seen in five of seven subject pairs did not achieve statistical significance (Z=−1.532, p=0.128).

General Comparison of Cocaine and Heroin Data Sets

In contrast to the relatively small list of transcripts significantly altered in the nucleus accumbens of cocaine abusers (Albertson et al, 2004), 1050 transcripts were significantly changed in the majority of heroin cohort subject pairs (Supplementary Table 1). When the lists of changed transcripts from both drug-abusing cohorts were compared directly, only 25 out of the approximately 39 000 possible transcripts were differentially expressed in both data sets, with 10 of these changed in opposite directions (Table 2). As is common with microarray studies, many of the transcripts within this overlap list are insufficiently annotated, although a few genes have been previously associated in some manner with drug abuse (see Discussion). To date, we have not further investigated the gene expression changes reported in Table 2, with the exception of the opioid peptide-encoding prodynorphin transcript, which was found to be increased in cocaine abusers and decreased in heroin abusers by RT-PCR (median cocaine expression=149% of control; median heroin expression=66% of control), with a significant correlation between the PCR and microarray data (r=0.769; p=0.001).

Table 2 Genes Significantly Altered in Both the Heroin and Cocaine Abusers
Full size table

DISCUSSION

High-throughput microarray technology makes it possible to simultaneously interrogate tens of thousands of transcripts in a discovery-driven process that should provide novel insights into disease processes. In this study, we have profiled the changes in gene expression in the nucleus accumbens of human heroin abusers relative to matched controls, and found nearly 10 times the number of genes changed compared to a cocaine cohort we have analyzed in an identical manner (Albertson et al, 2004). We have directly compared data from the two groups of chronic drug abusers. Transcript changes specific to either the heroin- or cocaine-abusing group could be due to the unique sites and mechanisms of action of these drugs. Transcript changes common to both cohorts of drug abusers could reflect excessive dopamine neurotransmission in the nucleus accumbens, as this is thought to be a common mechanism of all drugs of abuse.

The most robust and consistent finding from our previous cocaine study was a significant downregulation of myelin-related genes, as well as an apparent loss of MBP-immunoreactive oligodendrocytes (Albertson et al, 2004). At that time, we put forth the hypothesis that these cocaine-related expression changes could be due to an effect of excess dopamine in the nucleus accumbens, given the inhibitory effect of dopamine on the maturation of oligodendrocyte progenitors (Bongarzone et al, 1998; Howard et al, 1998). Because drugs of abuse are generally believed to facilitate nucleus accumbens dopamine release, according to this hypothesis myelin-related gene expression should also be decreased in our cohort of heroin abusers. On the contrary, we found no decrease (nor even a trend toward a decrease) in any myelin-related transcript in this group (Figure 1a). On the surface, these data seem to argue against our original ‘excess dopamine’ hypothesis as a plausible explanation for the cocaine study myelin findings. In fact, although some investigators have reported that heroin self-administration increases dopamine release in the nucleus accumbens (Wise et al, 1995; Xi et al, 1998), other data strongly indicate that, in contrast to cocaine, heroin does not influence dopamine outflow (Hemby et al, 1995, 1999; Smith et al, 2006). In any case, there is very clear evidence for a powerful dopamine-independent reward mechanism in the nucleus accumbens that mediates heroin self-administration (Ettenberg et al, 1982; Pettit et al, 1984). Thus, differences in the dopamine-releasing and other neurochemical effects of heroin could mediate differences we found in the profile of gene expression in the nucleus accumbens, including differential effects on myelin-related and presynaptic machinery genes.

In our heroin cohort, we found significant decreases in the expression of numerous genes encoding presynaptic machinery proteins involved in processes related to neurotransmitter release. Synaptotagmin 1 is located on synaptic vesicles and is thought to act as a calcium sensor for fast neurotransmitter release (Sudhof, 2004). SNAP25 participates in the SNARE complex mediating vesicle exocytosis (Sudhof, 2004). Munc 18-1 controls synaptic fusion by binding syntaxin, preventing SNARE formation until its release (Sudhof, 2004). Synapsin 2a and 2b are vesicular proteins required to maintain normal amounts of vesicles. Amphiphysin is a component of the presynaptic clathrin-coated intermediate formed during vesicular recycling (Slepnev and De Camilli, 2000). Pallidin, which is located on early endosomes and is known to bind syntaxin 13 on the vesicle surface, facilitates docking and fusion during recycling (Huang et al, 1999). Amysin (for which two transcripts were downregulated) regulates the assembly of the SNARE composite by forming stable complexes with the synaptic t-SNAREs syntaxin 1 and SNAP25, and inhibits exocytosis by interfering with synaptobrevin (Scales et al, 2002). Synaptogyrin 3 is a recently discovered brain-specific vesicular protein of unknown function (Belizaire et al, 2004). After identification of the many changes in this functional group by pairwise analysis, we conducted a post hoc group-wise analysis of other release-related genes represented on the microarray. Significant decreases in transcript abundance of both Vti1b, a vesicular SNARE protein functioning in endosomal recycling (Jahn et al, 2003), and SV2a, an essential protein implicated in the transport of calcium into vesicles (Sudhof, 2004), were identified in this manner. Subsequent quantitative PCR experiments validated the microarray findings for three of four transcripts tested. The downregulation of so many genes encoding known SNARE proteins, vesicular proteins, and recycling proteins implies a synaptic dysfunction within the brains of human heroin abusers.

Until relatively recently, gene expression studies have been limited to individual candidate genes and, to our knowledge, effects of heroin on presynaptic release proteins have not been reported. In a rat model utilizing morphine pellet implantation, striatal expression of synapsin 2a was found to be increased after several days of continuous, passive drug exposure. It is conceivable that the downregulation of presynaptic protein gene transcripts seen in the nucleus accumbens of our chronic heroin abusers is a compensatory response to an acute opiate-mediated upregulation. Differences in the method, pattern, and length of drug exposure (not to mention potential species differences) highlight the importance of pursuing complementary studies in both human material and animal or cell culture models.

It has been suggested that expression of the vesicular protein synapsin is particularly sensitive to disorders of cognition and mood (Vawter et al, 2002), as evidenced by decreased abundances in human post-mortem brain studies investigating bipolar disorder, schizophrenia, and Alzheimer's disease (Ho et al, 2001; Mirnics et al, 2000; Vawter et al, 2002). Although our study focuses solely on the nucleus accumbens and therefore cannot speak to more global reductions in synapsin 2 expression, the cognitive deficits documented in human heroin abusers (Ornstein et al, 2000; Pau et al, 2002), in combination with our synapsin data, could lend support to this notion.

Interestingly, when we directly compared the data from heroin abusers with our previous study on gene expression changes in the nucleus accumbens of human cocaine abusers, few transcripts were seen in common. To our knowledge, only three of these transcripts are known to be regulated by, or involved in the signaling of, heroin or cocaine. None fall within the group of transcripts downregulated in both drug-abusing cohorts. Within the group of transcripts increased in both cocaine and heroin abusers was tumor necrosis factor receptor superfamily member 6, encoding the proapoptotic receptor FAS (Table 2). Chronic heroin and morphine administration elicits a similar effect in rodent brain to what we report in heroin abusers (Boronat et al, 2001; Garcia-Fuster et al, 2003; Yin et al, 1999). Although this is the first report of FAS mRNA upregulation by cocaine, this pathway is upregulated by methamphetamine (Jayanthi et al, 2005).

Nearly one-half of the transcripts affected in both drug-abusing cohorts were actually changed in opposite directions. As an example, the transcript encoding the endogenous opioid peptide prodynorphin (PDYN) was differentially regulated between groups. Both microarray analysis and PCR revealed that PDYN mRNA levels were increased in cocaine subjects but decreased in heroin abusers. Cocaine-induced increases in PDYN levels are well documented (Hurd et al, 1999). Several days of morphine treatment also increases PDYN (Gerfen et al, 1990; Turchan et al, 1997), although the effects of longer exposure are not known. Dynorphin attenuates cocaine-induced increases in dopamine levels and has been hypothesized to act via κ opioid receptors to modulate dopamine system responses to stimulant administration (Kreek et al, 2005). Dynorphin peptides are also likely involved in adaptations to opiate abuse, as they affect μ opioid receptor function and indirectly dopamine neurotransmission (Kreek et al, 2005). It is possible that the local effects of heroin on the accumbens opioid receptors supercede the common effects of dopamine to elicit the differences in PDYN gene expression seen in cocaine and heroin abusers. Another example of differential regulation is the neurotrophin receptor NTRK2 (also known as TrkB), which was identified by microarray analysis as significantly upregulated in heroin abusers but downregulated in cocaine-abusing subjects (Table 2). In rodents, cocaine's effects on NTRK2 expression differ markedly depending upon the length of drug administration and withdrawal (Freeman et al, 2003; Lucas et al, 2003; Toda et al, 2002). Although heroin effects of NTRK2 expression have not been reported, it is interesting to note that morphine increases autophosphorylation of this receptor, and its activation plays a critical role in opiate-induced analgesia (Freeman et al, 2003; Lucas et al, 2003).

In conclusion, we report herein significant and distinctive changes in the profile of gene expression in the human nucleus accumbens associated with cocaine and heroin abuse. Documentation of the molecular correlates of addiction may ultimately suggest novel therapeutic strategies for the treatment of drug abuse. To date, the focus of thinking about drugs of abuse has been on the commonality of mesolimbic reward pathway activation. These data suggest that the molecular consequences of heroin and cocaine abuse are largely unique despite some common effects on dopamine in the nucleus accumbens, and that a re-examination of our current assumptions of the importance of common mechanisms in all addictive substances is warranted.