INTRODUCTION

Molecular biology and molecular genetics provide complementary approaches to the interesting but complex problems posed by human addictions. There are large complex genetic influences on human substance abuse vulnerabilities (Kendler et al, 1994, 1999, 2000; Gynther et al, 1995; Tsuang et al, 1996, 1998, 1999; Uhl et al, 1997; Kendler and Prescott, 1998; Merikangas et al, 1998; Stallings et al, 1999; True et al, 1999a, 1999b; Uhl, 1999, 2004; Karkowski et al, 2000). Much of the genetic vulnerability to abuse of different legal and illegal addictive substances is shared. Linkage- and association-based genome scanning studies have identified a number of chromosomal regions likely to contain variants that confer vulnerabilities to dependence on illegal substances, ethanol and/or nicotine (Uhl et al, 2002; Uhl, 2004). The mid region of human chromosome 7 is an interesting candidate site for allelic variants that alter substance abuse vulnerability based on linkages to alcoholism and associations with polysubstance abuse (Reich, 1996; Long et al, 1998; Reich et al, 1998; Uhl et al, 2001, 2002).

Molecular biologic studies of drug-induced gene regulation provide another approach to understanding substance abuse. We and others have used differential display and microarray approaches to identify constellations of gene expression patterns that are altered by acute or chronic administration of morphine and other addictive substances (Douglas et al, 1995; Wang et al, 1997; Schafer et al, 2001; Toyooka et al, 2002; Savoie et al, 2003; Sokolov et al, 2003). Identifying convergent data from genome scanning studies and studies of gene regulation could help to focus on interesting aspects of the large data sets generated by each of these approaches.

We now report results of association-based fine mapping of the chromosome 7 regions that contains positive markers from prior linkage and association studies, WIAF1680, D7S1793, D7S1830, and D7S1809. We describe initial identification of NrCAM based on convergence between initial association findings in this region and results from studies of morphine-regulated gene expression. We report additional fine mapping results that provide NrCAM haplotypes that are associated with substance abuse vulnerability in several samples and with different levels of NrCAM mRNA expression in human post-mortem tissue samples. We document reduced reward from morphine, cocaine, and amphetamine, which is manifest in mice with reduced NrCAM expression. This converging evidence supports involvement of NrCAM and its regulation in mechanisms of reward and in addiction vulnerability. This converging evidence also supports contributions of common NrCAM allelic variants to human individual differences in addiction vulnerability.

MATERIALS AND METHODS

Subtracted Differential Display, Northern, and RT-PCR Experiments

Male Sprague–Dawley rats (250–300?g) and male wild-type C57/BJ mice (25–30?g) were housed under 12-h light/12-h dark conditions and given food and water ad libitum. Morphine treatments for rat subtracted differential display experiments were 20?mg/kg morphine or saline i.p. with killing 4?h after injection. For RT-PCR experiments, mice were treated with saline or with 20 or 100?mg/kg/i.p. morphine and killed 4?h after the last injection.

Brains were rapidly removed, regions dissected, and total and polyA+ RNAs were extracted. Subtracted differential display was performed as described (Wang et al, 1997), with sequence comparisons using BLAST (National Library of Medicine, Bethesda, MD). Northern analyses were carried out as described (Liu et al, 2002). RT-PCR experiments used TaqMan real-time PCR expression assays, an ABI 7900HT Sequence Detection System and TaqMan GAPDH control reagents (ABI, Foster City, CA, USA). NrCAM mRNA was assessed using oligonucleotide primers 5′-AGTGATGCAGAAGATC and 5′-GATGATGGGACGTTTGGAGAATAC and probe 5′-TCCTGTCT GAAGGTGTTCGACTT.

Identification of NrCAM cDNAs

Plaques (8 × 106) from a size-selected rat brain DNA library prepared in lambda ZAP II (Stratagene, La Jolla, CA) were screened using a 340?bp EcoRI fragment of the MD50 product radiolabeled by random priming with (32P)dCTP as described (Liu et al, 1993). DNAs from 17 positively hybridizing plaques obtained from three cycles of plaque purification were sequenced using dideoxynucleotide termination manual and automated methods. These sequences were assembled into 6783 and 7099?bp subclones termed NC1–4 subclone and NC2–5, respectively.

Human NrCAM Sequences and Variants and their Homology with Mouse and Rat Sequences and Variants

Human NrCAM sequences came from NCBI (http://www.ncbi.nlm.nih.gov) and Celera databases (http://www.celeradiscoverysystem.com/index.cfm). Human genomic and cDNA sequences included BAC and scaffold sequences AC005058, AC007567, AC005683, AF172277, hCG17111, hCG1654651, hCG1654652, hCG1806278, and GA_x54KREB2RL0. We assembled more than 400?kb of human NrCAM genomic sequence, with attention to correcting scrambled and/or misaligned sequences found in the original sources. Human sequences were also compared to mouse and rat cDNA and genomic sequences from NCBI and Celera databases that included mCT6280, mCT6282, mCG7335, mCG7337, GAx5J8B7TUNR9, GAx5J8B7W4G5V, and NW_047760. Variants were sought by assembly of sequences obtained from DNAs of different individuals and by resequencing amplicons obtained from DNA from different individuals. Haplotypes were sought in BAC sequence comparisons, with attention to comparisons between AC005683 sequences and those forming a consensus from other assembled sequences.

Human Subjects for Association Genome Scanning

DNAs came from NIDA research volunteers of self-reported European- and African-American descent balanced for gender whose drug use and dependence were characterized as described (Smith et al, 1992; Persico et al, 1996; Uhl et al, 2001). Unrelated European-American ‘abusers’ averaged 35 years of age, scored the maximal 3+ score on a total drug use scale of use of addictive substances, and/or demonstrated DSMIII-R or DSMIV dependence on at least one illegal abused substance. ‘Control’ European-Americans averaged age 32 reported no significant lifetime histories of use of any addictive substance and 0 or 1+ DUS total drug use scores indicating that this group mixed individuals with little or no lifetime experience with legal or illegal addictive substances. ‘Abusers’ of self-reported African-American descent averaged 34 years of age while 181 controls averaged 36 years of age. We excluded subjects with 2+ DUS scores who did not reach DSM dependence criteria for an illegal substance. Other DNAs came from 177 unrelated individuals selected from COGA (wave I) pedigrees generously supplied from NIAAA with the assistance of COGA investigators. The substance abusing proband was selected from each pedigree based on (a) diagnosis of alcohol dependence and (b) manifestation of the highest level of abuse or dependence on illegal substances of any individual within that pedigree. Controls with no histories of substance abuse or dependence were selected from individuals who married into these pedigrees. A fourth sample contains 288 alcohol-dependent Japanese subjects and 472 gender-balanced Japanese control research volunteers who reported no alcoholism or dependence on other illegal substances, as previously described (Ishiguro et al, 1998). Alcohol-dependent individuals averaged 52 years of age when tested and 37 years of age at the onset of their alcohol dependence. Controls averaged 49 years of age were selected from healthy volunteers, who displayed no psychiatric diagnoses and reported levels of alcohol use less than two drinks on fewer than two occasions per week during any year.

Genotyping

Alleles for 54 SNPs lying in the mid chromosome 7 region of interest were assessed in DNA pools using SNP microarrays as described (Uhl et al, 2001). Alleles of 37 additional simple sequence length polymorphism (SSLP) markers were assessed in pooled DNA samples based on peak heights in ABI 3100 chromatograms. A total of 11 SNPs were genotyped by PCR-restriction fragment length polymorphism (RFLP) methods using 3% agarose/1.5% NuSieve gels, ethidium bromide staining, and ultraviolet transillumination. Additional SNPs at the NrCAM locus were genotyped with 5′ nuclease assays using TaqMan/ABI 7900HT sequence detection system assays and SnapShot/ABI 3100 genetic analyzer assays. Nominal significance values are reported for results from samples that nominated regions for further analyses. Significance values are reported for hypothesis-directed studies that replicate associations at these markers. Nominal significance values are reported for associations using haplotypes generated in studies of mRNA expression in brains of heterozygous individuals (see below).

Allele-Specific Expression in Human Brain RNAs

Brains were surgically removed from epileptic patients and/or obtained post mortem (Brain Tissue Bank at University of Maryland). Samples heterozygous for selected NrCAM markers were identified. Total and polyA+ RNAs and cDNAs were prepared from these specimens as described above. Allele-specific expression was assessed using: (a) real-time PCR, SNP-specific genotyping primers and probes and an ABI7900HT sequence detection system or (b) cDNAs amplified by PCR so that the allele-specific primer extension products produced peaks whose heights could be determined using an ABI 3100 (SnapShot, ABI, Foster City, CA) analyzer, compared to results obtained from genomic DNAs and normalized by comparison to standards constructed using known samples mixed in differing proportions. To specify the haplotypes conferring differential expression, we focused on alleles whose cDNA expression differences were larger than twice the standard deviation of the techniques used, as assessed in data from heterozygote genomic DNAs studied in parallel with the cDNA samples.

In Situ Hybridization and Immunohistochemistry

Adult rats and C57BL/6 mice were anesthetized, perfused, brains dissected, postfixed, equilibrated, and cut at 20–40?μm as described (Moriwaki et al, 1996). In situ hybridization (Uhl and Kitayama, 1993; Morales and Bloom, 1997) used (a) antisense T3 and control sense T7 transcripts that corresponded to the 0.2?kb 3′-UTR of clone NC9 sequences radiolabelled with [35S]UTP and (b) a 45 base antisense oligonucleotide complementary to NrCAM 3′-untranslated region sequences AACGAAGTACCC AACCACGTCGCA ATCACATACACAG CTAGTCCT that was internally radio-labeled with [35S]dCTP. Immunofluoresence (McNaught et al, 2002) used goat anti-NrCAM (Santa Cruz), rabbit antidopamine transporter (Chemicon), Alexa fluor 594 (red) anti-goat and Alexa fluor 488 (green) anti-rabbit secondary antibodies (Molecular Probes), and confocal microscopy.

Morphine- and Cocaine-Conditioned Place Preference

NrCAM knockout mice (25–30?g) were produced as described, maintained on mixed C57/129 genetic backgrounds under conditions noted above, and bred by heterozygote-heterozygote breeding strategies so that the average genetic backgrounds of homozygous NrCAM knockout, heterozygous NrCAM knockout, and matching wild-type mice were identical (Sakurai et al, 1997). Neither homozygous nor heterozygous NrCAM knockout mice displayed abnormal body size, locomotor activity, or growth rates, as previously noted (Sakurai et al, 1997). Conditioned place preferences were assessed as described (Sora et al, 2001). There were 4-day conditioning sessions using morphine sulfate (20?mg/kg i.p.; 30?min exposures), cocaine (20?mg/kg s.c; 20?min exposures), or amphetamine (2.0?mg/kg i.p.; 20?min exposures) with horizontal locomotor activity monitored during these trials as described (Hall et al, 2004). Conditioned place preference was calculated as the difference in the duration of time spent in the drug-paired compartment before and after drug conditioning. A positive score thus represents a preference for the drug-paired environment.

Statistical Analyses

Genotype frequencies were fit to Hardy–Weinberg equilibrium values, maximum-likelihood haplotype frequencies were estimated using expectation-maximization (EM) algorithms and Arlequin ver 1.1 (http://lgb.unige.ch/arlequin/) and standardized linkage disequilibrium estimates (D′=D/Dmax) were computed from estimated haplotype frequencies. For SSLP genotyping with pooled DNA analysis, estimated frequencies of each allele were multiplied by the total number of alleles in each pool. Differences in observed allele, genotype frequencies, and haplotype frequencies between groups were tested for nominal significance using Fisher's exact tests on 2 × 2 or 2 × (number of genotypes) contingency tables. For comparisons of initial screening estimates of allele and haplotype frequencies, p-values <0.05 were considered nominally significant. Nominally significant results from initial association studies were then tested in additional samples. Behavioral results were assessed using one-way ANOVA and post hoc t-tests. RNA expression data were assessed using t-tests.

RESULTS

Identification of NrCAM as a Morphine-Regulated Gene

Subtracted differential display results

Subtracted differential display PCR (SDD) identified MD50 as one of more than 40 regulated subcloned SDD cDNAs that corresponded to mRNAs whose expression was altered in striata of rats killed 4?h after treatment with 20?mg/kg morphine. MD50 displayed a 340?bp sequence with strong nucleotide similarities to human sequences 3860–4134?bp of GenBank entry U55258 previously reported as 3′ untranslated sequences of NrCAM.

A radiolabeled 340?bp MD50 fragment identified full-length cDNAs in a rat cerebral cortex λZapII library (Clontech, Palo Alto, CA). Sequences obtained from these clones defined the full-length NrCAM cDNA and several apparent NrCAM isoforms.

Full-length NrCAM cDNAs encode a single-transmembrane-domain protein with an N-terminal signal peptide, six immunoglobulin (Ig) domains, four to five fibronectin III repeats, a transmembrane domain, and a C-terminal cytoplasmic domain with tyrosine kinase phosphoacceptor sites (FIGQY) (Figure 1). Multiple rat NrCAM isoforms were also identified as the products of differential RNA splicing that produced NrCAM translation products with short inserted peptide sequences (Figure 1).

Figure 1
figure 1

(Top row) Cartoon of NrCAM's structure with fibronectin, immunoglobulin, transmembrane, and intracellular (C-terminal) domains indicated. Arrows indicate sites where splice variants encode additional inserted additional amino acids. These NrCAM structures are conserved in human, mouse, and rat. (Second row) Cartoon of the distribution of NrCAM exons, including those used in most transcripts (open boxes) and those that are alternatively spliced (filled boxes). Most exons are conserved in human, mouse and rat. Exons 1b and 33 have not been identified in humans. The candidate alternatively transcribed exons produce the shorter rat and human isoforms are shown in the bottom two rows. (Rows 3–8) NrCAM clones (Nc)1, 3, 6, 7, 14, 17 that represent splice variant isoforms identified in rat brain cDNAs. (Rows 9–10) Shorter NrCAM gene product expressed isoforms that are identified from ESTs in rat and human databases.

NrCAM cDNA and Genomic Sequences: Human, Rat, Mouse

Intron/exon structures of NrCAM genes from several species

Assembly of human NrCAM cDNA and genomic sequences reveals 35 exons encompassing 300?kb of chromosome 7q31.1–31.2. At least a dozen NrCAM isoforms can be produced from these exonic sequences by alternative splicing and transcriptional start site usage (Wang et al, 1998). Mouse and rat NrCAM cDNA and genomic sequences reveal 36 NrCAM exons that encompass about 300?kb of mouse chromosome 12A3 and rat chromosome 6q16.

Sequences that flank each NrCAM exon contain consensus splicing acceptor (AG) and donor (GT) sites. 5′-untranslated regions (UTRs) of both human and rodent NrCAM mRNAs are encoded by exons 1–3. These exons are separated by 60, 70–80, and 70–80?kb introns, respectively. Other NrCAM exons lie within 70–90?kb regions of both human and rodent genomic sequences. Most of these NrCAM exons encode functional domains. Two exons encode each Ig or fibronectin domain, one exon encodes the signal peptide, one encodes the transmembrane domain, and two encode the intracellular domain and 3′UTR (Figure 1). NrCAM exon–intron structures from human, mouse, and rat are well conserved. NrCAM peptides from these species share ⩾93% identity.

Alternative NrCAM promoters, splice variants and isoforms

Analyses of 17 rat NrCAM cDNA clones reveal both novel and previously identified splicing isoforms. The six amino acid insertion encoded by exon 5 is located between the signal peptide and the first Ig domain. The 19 amino acid insertion encoded by exon 10 falls between Ig domains II and III. The 10 amino acid insertion encoded by exon 19 is located between the Ig domain VI and the first fibronectin III (FnIII) domain. The C-terminus of the fourth fibronectin domain displays substantial variability that is conferred by isoforms that include a three amino acid insertion encoded by exon 27B, a 15 amino acid insertion encoded by exons 27B and 28A and 12- and four amino acid insertions encoded by exons 28A and B, respectively. Exons 28, 29, and 30 encode 93 and 105 amino acid insertions that include a fifth fibronectin domain (Figure 1). A novel isoform found in both mouse and rat NrCAM genes uses an additional NrCAM exon that we term rodent exon 1b. This exon encodes an alternative transcriptional start site that could not be identified in human EST or genomic sequences. Transcription beginning at the 5′ end of rodent exon 1b produces the 4.5?kb mRNA identified in rodent Northern studies (see below). This exon's products are also identified in transcripts mCT6282 and AU08595 that contain only NrCAM's fifth fibronectin, transmembrane, and intracellular domains. An additional novel NrCAM isoform observed in both human and rodent genomic and EST (BG001715) sequences uses exons 1c, the unique exon 33 and then the common exon 34 to produce an NrCAM isoform that contains only the intracellular domain of other NrCAM isoforms (Figure 1). Transcription of the human NrCAM is thus likely to start at two promoters. Transcription of the rodent NrCAM is likely to utilize three promoters to create NrCAM isoforms with different structures.

NrCAM Expression and Regulation

Northern and RT-PCR analyses

Northern analyses reveal NrCAM mRNA in rat brain but not in spleen, muscle, lung, liver, kidney, or heart (Figure 2). These analyses reveal 7.8, 6.5, 6.0, and 4.5?kb brain mRNA species likely to represent several NrCAM alternative splice and transcriptional start site variants. NrCAM mRNA displays high expression in cerebral cortex and hippocampus and moderate levels in striatum and thalamus.

Figure 2
figure 2

NrCAM mRNA size, distribution, and regulation. Northern analyses of NrCAM distribution and its regulation by saline (S) and morphine (M) treatment in cortex (CTX), striatum (STR), and hippocampus (HIP). There is substantial expression of several NrCAM transcripts in brain (B) but not in heart (H), lung (L), kidney (K), liver (L), skeletal muscle (M), or spleen (S).

Rats killed after single treatments with 20?mg/kg morphine reveal NrCAM mRNA upregulation to 183±20, and 141±10% of control values in Northern analyses of mRNA from striatum, and cortex, respectively. RT-PCR analyses of NrCAM expression in brains of wild-type mice killed after the same morphine doses reveal NrCAM levels that are 107±4, 110±7, 115±7% of control values in striatum, midbrain, and hippocampus, respectively.

In situ hybridization

In situ hybridization using NrCAM cRNA and oligonucleotide cDNA hybridization probes complementary to 3′ untranslated region sequences common to all NrCAM isoforms produce largely similar patterns for brain distribution that are not found when mRNA-sense hybridization probe is applied in control experiments (Figure 3) using film and emulsion autoradiography. The majority of hybridization signal is associated with neurons. White matter tracts including the corpus callosum, anterior commissure, and cerebellar peduncles display only background autoradiographic grain densities.

Figure 3
figure 3

In situ hybridization autoradiographic and immunohistochemical documentation of the distributions of NrCAM mRNA and protein in the brain. (a–f) Hybridization to radiolabeled antisense probes complementary to common regions of the NrCAM 3′ untranslated regions. Film autoradiograms of brain sections containing: (a) forebrain, (b) midthalamus, (c) midbrain, (d) brainstem, and (e) cerebellum. (f) Higher magnification view of in situ hybridization emulsion autoradiogram of a Nissl-stained midbrain section hybridized with the radiolabeled 3′-UTR NrCAM probe. (g) NrCAM immunohistochemistry in midbrain sections detected with Alexa Fluor 594 (red) in midbrain sections. (h) Immunohistochemistry in the same midbrain section detecting immunoreactivity for the dopamine transporter (DAT) using Alexa Fluor 488 (green) in midbrain sections. (i) Overlay of G and H documenting NrCAM expression in many dopaminergic neurons. In repeated experiments, most (80%), but not all, DAT immunopositive neurons display NrCAM staining and visa versa.

NrCAM mRNA expression is high in hippocampal pyramidal cells. Scattered neurons in other hippocampal layers also display significant hybridization densities. Higher densities of hybridization in mid- to deep layers of the cerebral cortex noted on film autoradiograms correspond to especially high levels of hybridization to the largest neurons in cerebral cortical laminae V and VI. Cerebellar purkinge cells display high densities of hybridization. In the striatum, hybridization densities are highest over large neurons consistent with interneurons. Hypothalamic zones contained spotty hybridization densities, including moderate hybridization signals over neurons of the supraoptic and paraventricular nuclei. Densities are noted in the habenula and basal medulla, including inferior olive and nucleus raphé magnus. Cells of the substantia nigra and ventral tegmental area (VTA) express hybridization densities.

Densities over individual cells in the nucleus paranigralis of the VTA are similar to those of hippocampal pyramidal neurons. NrCAM immunoreactivity is also found in ventral midbrain neurons (Figure 3). Many, but not all, of the NrCAM immunopositive neurons in the VAT also express dopamine transporter (DAT) immunoreactivity, a marker for dopaminergic neurons (Figure 3g–i). However, minor NrCAM immunopositive neuronal subpopulations do not clearly express DAT and minor subgroups of DAT immunopositive neurons fail to clearly express NrCAM (data not shown).

Association Genome Scanning of the Chromosome 7 Region of Interest

Hypothesis generation: NrCAM variants might contribute to individual differences in human addiction vulnerability

In total, 11 of 37 tested mid-chromosome 7 simple sequence polymorphisms display allele frequency differences between abusers and controls in NIDA European- and/or African-American samples that reach nominal significance (eg uncorrected for multiple comparisons). Three display nominally significant differences in both populations, while fewer than one would have been expected by chance. The nominally significant results thus encompass WIAF1680, D7S2420, and other markers located between 98 and 103?Mb of chromosome 7, a region that contains NrCAM. Since we had just identified this gene as drug-regulated, we used these nominally significant results and this expression data to generate the testable hypothesis that allelic variants in NrCAM might contribute to individual differences in human addiction vulnerability.

We also tested alternative hypotheses by studying markers in other nearby genes that include intracellular membrane-associated calcium-independent phospholipase A2γ (iPLA2)(hCG17116, XM_027224), a ribosomal protein L7 homologue (hCG17114, XM_027229), the laminin-β precursor gene (LAMB4), and laminin β1 (LAMB1) gene (Figure 4).

Figure 4
figure 4

Diagram of the mid-chromosome 7 region subjected to initial hypothesis-generating and follow-up hypothesis-testing association studies. Positions of genes and markers are noted. Markers named here display nominally significant associations with addiction in either European- or African-American samples from NIDA. Other markers studied include: (1) WIAF-514, (2) WIAF-678, (3) WIAF-902, (4) WIAF-2308, (5) WIAF-1810, (6) WIAF-2750, (7) WIAF-2752, (8) WIAF-4576, (9) WIAF-1960, (10) WIAF-2601, (11) WIAF-3321, (12) WIAF-350, (13) WIAF-806, (14) WIAF-961, (15) WIAF-4581, (16) WIAF-4573, (17) WIAF-4537, (18) WIAF-4521, (19) WIAF-1772, (20) WIAF-3603, (21) WIAF-2264, (22) WIAF-550, (23) WIAF-570, (24) WIAF-1781, (25) WIAF-1477, (26) WIAF-3755, (27) WIAF-672, (28) WIAF-671, (29) WIAF-2086, (30) WIAF-807, (31) WIAF-983, (32) WIAF-2152, (33) WIAF-3292, (34) WIAF-3585, (35) WIAF-3285, (36) WIAF-3650, (37) WIAF-3309, (38) D7S2213, (40) UT6195, (41) D7S817, (42) D7S1818, (43) SSLP in AC004920, (44) SSLP in AC006971, (45) GGAT3D12, (46) D7S2472, 47: SSLP in AC006001, (48) D7S1839, (49) GTAT1A10, (50) UT6234, 51: GATA8G02, (52) GATA21D08, (53) D7S820, (54) D7S821, (55) ATA21D06, (56) D7S530, (57) D7S501, (58) D7S2459, (59) UT899, (60) D7S1817, (61) D7S2847, and (62) D7S2203. Enlargement shows the region subjected to follow-up studies including NrCAM and surrounding genes, with the positions of key markers that help to define the 5′ (left) and 3′ (right) NrCAM haplotypes indicated.

NrCAM Region Markers and Haplotypes

Elucidation and validation of SNPs

NCBI and Celera databases, sequence alignments, and NrCAM genomic and cDNA resequencing allowed identification of the NrCAM missense SNP rs6958498/hCV25644546 that produces an Ala520Pro coding variant. We also identified four additional common and one rare coding region SNPs, an SNP in NrCAM's 5′ untranslated region sequences and two SNPs in its 3′ untranslated region sequences. We could not identify the previously reported missense SNP rs1043895/hCV8856898 in any of 200 European-Americans sampled (data not shown).

Definition of NrCAM 3′ and 5′ blocks of restricted haplotype diversity

Identifying these SNP variants allowed us to define 3′ and 5′ blocks of restricted haplotype diversity in European-American individuals. The 3′ haplotype block includes the rs1269634/hCV2686185/Ala520Pro variant; rs401433/hCV2298076/Ala710Ala variant and rs439587/hCV2298025 variant that encodes 1343c>a in the 3′UTR, and displays moderate linkage disequilibrium (Supplementary materials and data not shown). The more 5′ block of restricted haplotype diversity includes SNPs rs3763462/hCV2631385, rs1859769/hCV11607927, and rs2284284/hCV2686200. Haplotypes involving these markers provided information about at least four additional 5′ NrCAM SNPs that lie in introns, four that lie in 5′ exons, and five that lie in 5′ flanking sequences.

In African-American samples, more complex patterns of linkage disequlibrium produce smaller and more complex haplotype blocks especially noted in the 3′ haplotype block (Supplementary materials and data not shown). A region of historically higher recombination thus separates the 3′ and 5′ haplotype blocks in both ethnic samples, while additional recombinations also appear to provide small haplotype blocks most notable in 3′ NrCAM aspects in African-American samples (data not shown). These patterns are generally in accord with those documented more recently in HapMap data (Supplementary material).

NrCAM Markers and Haplotype Associations with Substance Abuse Vulnerabilities

Testing the hypothesis that variants contained in 3′ NrCAM haplotypes contribute to individual differences in human addiction vulnerability: identification of associations with opposite phases in European- and African-American samples

To test this hypothesis, we first sought association between addiction and allelic frequencies at markers in NrCAM's 3′ block of restricted haplotype diversity. The 3′ haplotype formed by rs6958498/hCV25644546/Ala520Pro, rs401433/hCV2298076/Ala710Ala, and rs439587/hCV2298025/3′UTR+1343c>a was about twice as abundant in NIDA European-American controls as in abusers (nominal p=0.0006) (Table 1). To seek generalization and extension of these initial results, we studied unrelated European-American individuals selected from COGA pedigrees. The 520Pro/Ala710c/+1343a 3′UTR haplotype was again more frequent in individuals who did not meet any alcoholism diagnosis by DSM criteria than in individuals with alcohol dependence (p=0.003). To further test whether the 3′ 520Pro/Ala710c/+1343a 3′UTR haplotype was also present in different frequencies in addicts and controls, we tested NIDA African-American drug abusers vs controls. These individuals again displayed significant differences between 520Pro/Ala710c/+1343a 3′UTR haplotype frequencies in abusers and controls (nominal p=0.0006). However, these samples displayed the opposite phase of association from that displayed by the two European-American samples. Frequencies of the addiction-associated 3′ NrCAM haplotype were thus higher in African-American abusers than in controls but lower in two samples of European-American substance abusers than in the corresponding control samples (Table 1).

Table 1 Frequencies of 3′ NrCAM Haplotypes Formed by A520P, 716A g>c, and +1343a>c 3′UTR in Substance Abusing and Control Samples

Testing the hypothesis that variants contained in 5′ NrCAM haplotypes contribute to individual differences in human addiction vulnerability: identification of associations with the same phase in European-American, African-American and Japanese samples

Since the phase of addiction vulnerability association of the 3′ NrCAM 520Pro/Ala710c/+1343a 3′UTR haplotype was similar in two European-American substance abuser/control comparisons but opposite in African-American samples (Table 1), we then asked whether more 5′ haplotypes might display association with the same phase in individuals of both ethnicities.

The NrCAM 5′ flanking region haplotype that includes rs3763462/hCV2631385, rs1859769/hCV11607927, and rs2284284/hCV2686200 displays significant differences between abusers and controls in each of the three samples tested. Further, this haplotype's association displays the same phase in comparisons of NIDA European-American polysubstance abusers vs controls (p=0.0002), NIDA African-American polysubstance abusers vs controls (p=0.02), and COGA European-American alcoholics vs controls (p=0.06; Table 2). We were able to replicate the same finding in another addiction vulnerability sample of different ethnicity: in samples of Japanese alcoholics vs controls (p=0.01). These cumulative positive data strongly support our hypothesis that NrCAM variants contribute to human interindividual differences in addiction vulnerability.

Table 2 Frequencies of 5′ NrCAM Haplotypes Formed by rs3763462/hCV2631385, rs1859769/hCV1160727, and rs2284284/hCV2686200 in Substance Abusing and Control Samples

Contrasting results for markers at nearby genes

These positive data for NrCAM contrast with results obtained at nearby genes. Neither LAMB1 missense hcv2193705/Arg1022Gln, LAMB1 rs7561/hcv8856543/3′ untranslated region +30C>A, LAMB4 missense SNP rs1627354/hCV8856678/His1417Tyr, nor LAMB4 intron SNP rs2286251/hCV16179224 display allelic frequency differences between abusers and controls (data not shown). Studies of four common exonic and five 5′ flanking region or intronic iPLA2 variations reveal that none of these markers produces associations that displayed the same phase in comparisons of abuser vs control groups of European- and African-American samples, although several of the markers display nominally significant associations with addiction vulnerability in single samples (data not shown).

Haplotype-Specific Expression of NrCAM mRNA in Human Brain

NrCAM haplotype association with reduced mRNA expression

We hypothesized that functional NrCAM haplotypes might produce their effects on human addiction vulnerability, at least in part, by conferring different levels of NrCAM expression and/or regulation in ways that might contribute to individual differences in NrCAM mRNA levels in post-mortem human brains. To seek possible functional significance of the NrCAM locus haplotypes, we identified cerebral cortical, hippocampal, and midbrain samples from European-American individuals who were heterozygous for SNPs in NrCAM exons. We examined the ratios between the mRNAs marked by one allele, as defined using individual SNPs, vs the mRNAs marked by the other allele. These expression studies of allelic variants at eight SNPs within NrCAM revealed expression patterns skewed toward lower expression of one allele than the other (Figure 5a). The maximal differences in haplotype-specific mRNA expression correspond to the ‘gcgccPcg’ haplotype that is marked by the alleles: rs1269634/hCV2686185=g, rs2072546/hCV2686182/N121N=c, rs1269621/hCV2686158/N319N=g, rs381318/hCV2298103/N404N=c, rs404287/hcv2298094/A509A=c, rs6958498/hCV25644546/A520P=P, rs401433/hCV2298076/A710A=c, and rs449077/hCV2298026/+666 3′UTR=g.

Figure 5
figure 5

NrCAM mRNA expression levels, markers, and polysubstance abuse vulnerability. (a) Skew distribution of allele-specific NrCAM mRNA expression in brains from heterozygous European-American individuals. Ratios of expression levels for NrCAM mRNAs containing alternative alleles for eight exonic NrCAM SNPs are shown with circles=rs1269634 (5UTR) g/a, triangles=rs2072546 (N121N) c/t, squares=rs1269621 (N319N) g/a, stars=rs381318 (V404V) c/a, hearts=rs40428 (A509A) c/t, quarter moons=rs6958498 (A520P) P/A, diamonds=rs6958498 (A710A) c/g, lightning bolts=rs449077 (3utr) g/a. Subjects with nearly equal levels of expression of mRNAs corresponding to both allelic variants are indicated in black, and those with >2 standard deviations lower expression of one variant than the other indicated in open symbols. The heterozygote gcgccPcg haplotype that correlated with the reduced levels of expression was constructed from the allelic variants that displayed skewed distributions toward lower levels of expression. (b) NrCAM mRNA expression in cerebral cortical (circle), midbrain (square), and hippocampal (triangle) samples from individuals heterozygous for the gcgccPcg haplotype (white, open symbols) and individuals who lack the gcgccPcg haplotype (black, filled symbols). (c) Frequencies of the gcgccPcg NrCAM haplotype in European-American NIDA controls (n=198; lighter bar) and abusers (n=316; darker bar). *p<0.003.

Comparisons of haplotype-specific NrCAM mRNA expression levels in brains of heterozygous individuals

mRNA corresponding to the gcgccPcg haplotype is expressed at an average of 26% of the levels of expression of mRNAs encoded by the alternative haplotypes in these samples from heterozygous individuals (Figure 5a).

Comparisons between levels of NrCAM mRNA expression in brains of heterozygous individuals vs those with other hapoltypes

Supporting results were obtained when we compared NrCAM mRNA levels in cerebral cortex, midbrain, and hippocampal samples from individuals who were heterozygotes for the gcgccPcg haplotype to samples from individuals who lacked a gcgccPcg haplotype (Figure 5b). Expression is about 40% lower in the brains of individuals who displayed a gcgccPcg haplotype. Much of this difference can be captured by a core two-SNP ‘Pc’ haplotype (data not shown).

Contrasting results for expression of nearby genes

Data for NrCAM allele specific expression contrast with the results of attempts to identify possible patterns of haplotype-specific differential expression of adjacent genes. None of these genes revealed evidence for haplotype-specific differential expression. iPLA2 expression from each haplotype was similar in heterozygotes for the rs42200/hCV2631461 5′UTR and novel 3′UTR SNPs. There were no allele-specific differences in expression of the hypothetical gene hCG2014062 (data not shown).

Association Studies Using the Differentially Expressed NrCAM Haplotype

Definition of the European-American gcgccPcg haplotype that is associated with differential levels of expression of NrCAM mRNA and the core Pc haplotype allowed us to seek associations between these differentially expressed haplotypes and substance abuse. The extended gcgccPcg SNP haplotype displays significant frequency differences between NIDA European-American controls and polysubstance abusers (p⩽0.01) (Figure 5c). The core Pc haplotype based on rs6958498/hCV25644546/A526P P and rs401433/hCV2298076/A716A c displays significant associations in the NIDA European-American (p=0.0003), NIDA African-American (p=0.02), and COGA European-American (p=0.05) samples (data not shown).

Reduced Morphine and Stimulant Reward in Heterozygous and Homozygous NrCAM Knockout Mice

We hypothesized that if individual differences in levels of NrCAM expression contributed to individual differences in addiction vulnerability in humans, then mice with altered levels of NrCAM expression might display differences in murine models of substance reward and dependence. Conditioned place preference provides a relatively robust test for alterations in the reward and reward-memories induced by abused substances and displays good correlations with human addiction liabilities for many classes of abused substances. Heterozygous NrCAM knockout mice display 50–60% of the NrCAM expression of wild-type mice (data not shown), while homozygous knockouts display no NrCAM mRNA. Further, NrCAM knockout mice fail to display any robust deficit in mnemonic processes or locomotor abilities that would, a priori, invalidate conditioned place preference testing (MG and others, data not shown).

Both heterozygous and homozygous NrCAM knockout mice display striking reductions in their preferences for the places where they received 20?mg/kg morphine in comparison to wild-type control mice (Figure 6). Place preferences conditioned by 20?mg/kg of cocaine were also reduced in both heterozygous and homozygous NrCAM knockouts (Figure 6). The place preference conditioned by 2?mg/kg amphetamine was also attenuated, although only in homozygotes (Figure 6). There was no evidence for sedation found in locomotor data obtained from these mice during conditioning trials. Mice of all genotypes displayed similar sensitivities to the acute locomotor stimulant properties of all three drugs during conditioning trials (data not shown).

Figure 6
figure 6

Morphine (black bars), cocaine (striped bars), and amphetamine (white bars)-conditioned place preferences in wild type (+/+), heterozygous (+/−), and homozygous (−/−) NrCAM knockout mice. Postexposure−pre exposure differences in the time spent on the previously drug paired side of the apparatus are noted in seconds. Morphine (20?mg/kg/ip) cocaine (20?mg/kg/s.c.) and amphetamine (2.0?mg/kg/i.p.) both cause wild-type mice to spend more time on the side paired with drug than they did prior to drug pairings. Morphine data derive from19 wild type, 31 heterozygous, and 16 homozygous NrCAM knockout mice, cocaine data from 17 wild type, 36 heterozygous, and 22 homozygous NrCAM knockout mice, and amphetamine data from nine wild type, 25 heterozygous, and 5 homozygous NrCAM knockout mice. Both heterozygous and homozygous NrCAM knockouts reduce this measure of morphine and cocaine reward, while results were significant for amphetamine only in homoaygotes One-way ANOVA p<0.05. *Post hoc test: p<0.05 compared with wild-type mice.

DISCUSSION

The present report documents convergent results from searches for drug-regulated genes and from association-based genome scans for drug abuse vulnerability alleles in a chromosome 7 region implicated in prior linkage studies. These data are supported by results of work on the NrCAM gene, its haplotypes, the brain distribution of NrCAM, the apparent impact of NrCAM haplotypes on levels of brain expression, and the impact of NrCAM deletion on drug reward in mouse models. Each of these approaches suggests that NrCAM could play important roles in neuronal connections and functions important for drug reward and drug addiction. Taken together, the data provide strong support for the idea that NrCAM haplotypes contribute to human individual differences in addiction vulnerability.

Subtractive hybridization/differential display, microarray, and related approaches have proven to be powerful tools, identifying hundreds of novel and known genes that are regulated after acute or chronic treatments with abused substances (Douglas et al, 1995; Schafer et al, 2001; Toyooka et al, 2002; Savoie et al, 2003; Sokolov et al, 2003; Wang et al, 1997). One of these genes is NrCAM, a single transmembrane domain protein member of a family of ‘cell adhesion molecule’ genes. The products of these ‘cell adhesion molecule’ genes can form homomeric or heteromeric recognition complexes that are important for normal brain development and for many of the ways in which adult brains alter functional connections.

The substantial levels of expression of NrCAM protein found in adult brain, its localization in specific neuronal populations, observations of substantial synaptic plasticity in circuits that express NrCAM, the likelihood that NrCAM function is regulated by phosphorylation (Grumet, 1997), and the likelihood that NrCAM can regulate important phosphorylation cascades (Schaefer et al, 1999; Kolkova et al, 2000) each make this cell adhesion molecule a strong candidate to influence cellular regulatory pathways important for short- and long-term actions of abused substances.

NrCAM haplotypes, their expression and association with drug abuse vulnerability

We have identified NrCAM polymorphisms, 3′ haplotypes, and 5′ haplotypes that are found at different frequencies in individuals without any significant lifetime use of any addictive substance when compared to individuals dependent on addictive substances. These haplotypes and the racial or ethnic differences in haplotyes and allele frequencies identified here are supported by recent HapMap data as well. The different phases of association for the 3′ NrCAM haplotype block that we have identified in different ethnic groups are consistent with the idea that this block is close to but does not contain the functionally important allelic variations. Finding that the association for the 5′ haplotype block is in the same phase in several ethnic groups is consistent with the idea that this block contains the key functional allelic variants. These observations are also in accord with observations that NrCAM haplotypes correlate with levels of NrCAM expression in post-mortem brain samples and that drugs regulate NrCAM expression. Variations in NrCAM regulation are thus the strongest candidate mechanisms to explain the effects of NrCAM haplotypes on addiction vulnerability.

None of these lines of evidence rules out the possibility that NrCAM variants that do not alter its regulated levels of its expression could also contribute to addiction vulnerability, however. Alanine/proline missense substitutions in extracellular motifs might alter NrCAM's binding properties. The rich patterns of differential splicing and exon use within the 3′ NrCAM haplotype block that we have identified might add to the diversity of NrCAM binding domains and allow NrCAM variants in these regions to subtly modify the cell-type specific recognition information that they appear to confer, while each of these polymorphisms did not always show association.

Reduced drug reward in NrCAM knockout mice

Haplotypes that correlate with reduced NrCAM expression in human post-mortem brains are found less frequently in human substance abusers than in human controls. These lines of evidence provide strong correlates with data obtained from heterozygous knockout mice. Mice with reduced levels of NrCAM expression manifest less preference for the places associated with injections of morphine or cocaine. Mice with no NrCAM expression demonstrate mutually no conditioned place preferences for morphine, cocaine, or amphetamine. These striking and convergent results should be considered in the context of possible confounding influences. It is conceivable that some of the effects observed in these single-dose CPP studies could be due to left- or rightward shifts in the dose–effect relationships of these drugs. While more detailed dose–effect studies will be necessary to identify such effects, it seems that potency shifts alone would be highly unlikely to confer the convergent results obtained here. Conceivably, alterations in drugs' sedative or stimulant properties might contribute to the CPP results obtained here. However, locomotor data obtained during conditioning sessions provide no support for such effects. We cannot totally exclude modest contributions of NrCAM knockout effects on general memory systems in the observations made here. However, there is a lack of large effects of NrCAM knockout on most memory tests that has been observed in preliminary studies of these animals (Grumet and Matzel unpublished observations), with only very modest influences in heterozygotes. The contrasting large, and consistent effects of NrCAM knockouts on morphine-, amphetamine-, and cocaine-conditioned place preferences suggests that most of the altered drug effects identified in these assays are unlikely to come from mnemonic effects alone.

Contribution of NrCAM variation to COGAwave I chromosome 7 linkage peak

The current data provide support for association-based genome scanning and point to its ability to identify specific vulnerability-associated genes and haplotypes. Our identification of NrCAM haplotype associations in COGA samples is consistent with contributions of the identified haplotype to the linkage signal reported by Reich et al in 1997. However, NrCAM variants appear unlikely to confer all of these linkage signals. Modest association signals found in other parts of this broad chromosome 7 region (HI and GRU, unpublished observations, 2004) could be consistent with contributions of variants at other loci that fall under the relatively large chromosome 7 linkage peak to the observed COGA Wave I linkage results. It is also interesting to note that the large linkage signals identified in this region in COGA Wave I samples do not appear as intensely in studies of smoking and alcohol phenotypes in COGA Wave II samples (Bierut et al, 2004). While there is no generally agreed way to estimate the power of association genome scanning approaches, further studies in additional and population-based samples will help to accurately estimate the addiction risks attributable to NrCAM haplotypes and other gene variants that lie in mid-chromosome 7.

Roles of neuronal cell adhesion molecule variants in addiction and other brain disorders

The current data are in accord with observations from recently reported studies of autism that describe contributions of 5′ NrCAM haplotypes to autism (Bonora et al, 2005). These additional observations support the idea that variants found in 5′ NrCAM haplotypes exert functional impact on the gene and its expression. NrCAM haplotypes thus appear likely to provide plieotrophic contributions to brain and behavior and contributions to vulnerability to addiction and autism that are likely to depend on which other genetic and environmental influences are present.

Potential roles in multiple brain functions and brain disease states are supported by in situ hybridization and immunohistochemical results. These data document NrCAM expression in regions such as hippocampus that are important for mnemonic functions and in midbrain dopaminergic neurons important for reward. NrCAM could thus influence the establishment and maintenance of cellular regulatory pathways important for drug reward and drug memory circuits. Drug regulation of NrCAM expression in some of these brain regions provides other levels of complexity. NrCAM expression, regulation, and individual differences are thus all well-positioned to play roles in responses to addictive drugs and in the development of addictions.

NrCAM is unlikely to play such roles alone, however. NrCAM is a member of a family of genes that can form homophilic or heterophilic recognition complexes that are important for normal brain development and can play key roles in adult brain connectivities. Data from higher density genome scanning SNP arrays and assessments of genes for other family members both suggest that human individual differences in other cell adhesion molecule genes could contribute to addiction vulnerability (Q-R Liu, GRU et al, in preparation; HI and GRU, in preparation). Molecules with Ig/fibronectin/single transmembrane domain/tyrosine kinase motifs such as those displayed by NrCAM, molecules with EGF-like motif/single transmembrane domain/tyrosine kinase motifs, and nontransmembrane cell adhesion molecules that contain glycoprotein membrane anchors are each implicated by these studies and/or by studies that document regulation by abused substances (Grigorenko et al, 2002; Miller and Luo, 2002; Dunckley and Lukas, 2003) (HI and GRU, in preparation). Regulatory events that involve NrCAM and other cell adhesion molecules appear to play significant roles in adaptive process that follow drug administration, including those relevant for addiction.

Convergence between positional cloning and disease-regulated gene results

The current observations concerning NrCAM appear to represent a striking example of convergence between genome scanning and gene expression results for addiction, a phenomenon that has been termed convergent functional genomics in studies of depression (Niculescu and Kelsoe, 2001). Gene expression data helped to focus our attention on an appropriate locus in genome scanning efforts. Regulatory differences are likely to contribute to the effects of many functional human allelic variants. Disease-regulated genes located on chromosomal regions in which markers are linked to or associated with the same disease should be considered strong pathogenic candidates and examined with care in fine mapping studies. As better and better validated animal models for disease expression studies and improved human gene expression profiles are available for more and more disorders, we anticipate more examples of the power of this approach.