CRH Haplotype as a Factor Influencing Cerebrospinal Fluid Levels of Corticotropin-Releasing Hormone, Hypothalamic-Pituitary-Adrenal Axis Activity, Temperament, and Alcohol Consumption in Rhesus Macaques

Context  Both highly stress-reactive and novelty-seeking individuals are susceptible to alcohol use disorders. Variation in stress reactivity, exploration, and response to novelty have been attributed to differences in corticotropin-releasing hormone (CRH) system function. As such, CRH gene variation may influence risk for alcohol use and dependence.

Objective  To determine whether CRH variation influences relevant intermediate phenotypes, behavior, and alcohol consumption in rhesus macaques.

Design  We sequenced the rhesus macaque CRH locus (rhCRH) and performed cladistic clustering of haplotypes. In silico analysis, gel shift, and in vitro reporter assays were performed to identify functional variants. Cerebrospinal fluid (CSF) and blood samples were obtained, and levels of CRH and corticotropin (ACTH) were measured by radioimmunoassay. Behavioral data were collected from macaques during infancy. Among adolescent/adult animals, we recorded responses to an unfamiliar conspecific and measured levels of ethanol consumption.

Setting  National Institutes of Health Animal Center.

Participants  Rhesus macaques.

Main Outcome Measures  Animals were genotyped for a single-nucleotide polymorphism disrupting a glucocorticoid response element, rhCRH −2232 C>G, and the effects of this allele on CSF levels of CRH, plasma levels of ACTH, behavior, and ethanol consumption were assessed by analysis of variance.

Results  We show that −2232C>G alters DNA×protein interactions and confers decreased sensitivity of the CRH promoter to glucocorticoids in vitro. Consistent with the known effects of glucocorticoids on CRH expression in the brain, carriers of the G allele had lower CSF levels of CRH but higher levels of ACTH. Infants carrying the G allele were more exploratory and bold, and among adolescent and adult male macaques, the G allele was associated with exploratory/bold responding to an unfamiliar male. Adults with the C/G genotype also exhibited increased alcohol consumption in the social group, a model for high-risk alcohol-seeking behavior.

Conclusion  Haplotypes that differ in terms of corticosteroid sensitivity have been identified in humans. Our data may suggest that functionally similar CRH variants could influence risk for externalizing disorders in human subjects.

Corticotropin-releasing hormone (CRH) is critical to behavioral and neuroendocrine adaptation to stress and is the primary neuropeptide responsible for hypothalamic-pituitary-adrenal (HPA) axis activation. Studies in rodents have shown that various drugs of abuse, including alcohol, can influence CRH release and HPA axis output,^1-4 and alcohol-preferring strains have been shown to exhibit differences in CRH system function, sensitivity to stress, and HPA axis activity even prior to alcohol exposure.^5-8 In humans, perturbations of both the CRH system and the HPA axis are associated with various stress-related neuropsychiatric disorders, and a number of studies have shown that the CRH system and HPA axis are dysregulated in alcohol-dependent subjects.⁹ However, whether these observed differences antedate or result from prolonged alcohol use is unknown. It is also unknown whether genetic variation at the CRH locus increases the risk for developing stress-related or alcohol use disorders.

Studies using experimental manipulations of CRH system activity suggest that naturally occurring CRH gene variation may mediate individual variability in behavioral and physiological traits that are key to determining an individual's coping style. One of the most consistent behavioral correlates of CRH system activity is the way an organism approaches novelty and unfamiliar conspecifics.¹⁰ Individuals who readily seek out and investigate novel stimuli are considered “exploratory” or “bold”; those more likely to show fear or withdrawal when confronted with new objects or individuals are described as more “inhibited” or “shy.” Individuals show stable tendencies in their behavioral reactions, which are linked in mice, humans, and nonhuman primates, among others, to biological traits such as heart rate variability,¹¹ frontal brain electrical activity, and cortisol levels.^12,13 Because individuals with both anxious/inhibited and impulsive temperaments are more likely to regularly consume alcohol,^14,15 it is possible that those who have either very high or very low levels of CRH system function could be at increased risk for developing alcohol problems. Corticotropin-releasing hormone system genes are therefore good candidates for investigating genetic variation as it relates to vulnerability to stress-related disorders and alcohol dependence.

Genetic variants that are functionally similar to those that exist in humans have been identified in the rhesus macaque (eg, HTTLPR, MAOA-LPR, and OPRM1 C77G), offering the opportunity to examine how genetic variation may influence traits linked to alcohol dependence in subjects living in a controlled environment.^16-19 The aim of the present study was to screen the rhesus macaque CRH gene (rhCRH) and its transcriptional control region for variation. Because corticosteroids have a critical role in regulating CRH, and human CRH haplotypes that confer differences in corticosteroid sensitivity in vitro have been reported,²⁰ we were interested in identifying variants that might alter glucocorticoid receptor (GR)–mediated transcriptional control. We then wanted to determine whether rhCRH variation influenced central CRH levels, HPA axis function, temperament, and behavioral reactions to social challenge. Because both dysregulation of the CRH system and certain temperament traits can increase vulnerability to alcohol use disorders, we also tested whether rhCRH genotype influenced voluntary alcohol consumption.

Genomic DNA was extracted from whole blood from rhesus macaques (Macaca mulatta) from the National Institutes of Health Animal Center (NIHAC), and direct sequencing was performed using samples from 20 unrelated animals (pairwise identity by descent, ≤0.0125) that were selected on the basis of variable HPA axis activity. We used primers designed from published human sequence and, subsequently, from rhesus macaque sequence generated in our laboratory (available on request). Cycle sequencing was performed using the BigDye Terminator Version 3.1 reaction in 96-well optical plates (Applied Biosystems, Inc, Foster City, California). Variants were detected by visualization of electropherograms generated by ABI Sequencing Analysis software (Applied Biosystems, Inc).

Haplotype frequencies and agreement with Hardy-Weinberg expectations were predicted using the default settings of PHASE, version 2.1 (http://www.stat.washington.edu/stephens/instruct2.1.pdf) (based on variation identified from the sequencing of 40 chromosomes). To identify putatively functional variants, we examined regions containing consensus sites for factors known to regulate CRH transcription and also used Web-based transcription factor binding site prediction algorithms (TfSitescan, http://www.ifti.org/cgi-bin/ifti/Tfsitescan.pl²¹ and Compel Pattern Search 1.0, http://compel.bionet.nsc.ru/FunSite/CompelPatternSearch.html²²).

Human haplotypes for Nigeria (YRI) and Utah (CEU) samples in the CRH gene region plus 5 kilobase upstream and 1 kilobase downstream were downloaded from HapMap project Release 21 (http://www.hapmap.org/). The following markers were used in our analyses (in 5′-3′ orientation: rs10098823, rs5030877, rs7839698, rs5030875, rs4501563, rs6472257, rs6999780, rs12721507, rs12721508, rs12721509, rs3176921, rs12721510, rs6158, rs12721511, rs7459924, rs6982394, rs11986876, and rs11997816). Based on Manhattan distances weighted by minor allele frequency and marker average linkage disequilibrium (LD), haplotypes for rhesus macaque (5 most common) and human were clustered hierarchically using R (http://www.r-project.org). The ancestral sequences at each single-nucleotide polymorphism (SNP) site were estimated as described by Pollard et al,²³ with multiple alignments of human (March 2006 assembly), chimp (March 2006 assembly), and rhesus macaque (January 2006 assembly) from UCSC Genome Browser (http://genome.ucsc.edu/).

DNA samples from 2 other rhesus macaque colonies (AlphaGenesis, n = 69 and University of Wisconsin, n = 94, the kind gift of M. Schneider, PhD, University of Wisconsin) were screened for alleles thought to cosegregate with −2232G, based on our initial sequencing. Focused resequencing was also performed on DNA from NIHAC animals of known −2232C/G genotype (C/C, n = 40; C/G, n = 60) to verify that rhCRH-H2 variants were in allelic identity.

Various other primate samples (New World monkey, Ateles geoffroyi; Old World monkey, Macaca nemestrina, Macaca fascicularis, Papio hamadryas, and Papio anubis; ape, Pan troglodytes) were sequenced for determination of sequence identity in the region surrounding the −2232C/G site. DNA samples from 10 wild baboons (P hamadryas) from the Awash National Park, Ethiopia, were the kind gift of J. Phillips-Conroy, PhD, and C. J. Jolly, PhD, New York University. Additional nonhuman primate samples were obtained from the Coriell Institute for Medical Research. Alignments of CRH and 5′ flanking region sequences were performed using ClustalW (http://align.genome.jp/) and/or DNAsis-Mac V.2.0 (http://www.bio.net/hypermail/methds-reagnts/1994-August/017671.html) software. Sequence identity percentages for interspecies comparisons were calculated using the Web-based pairwise alignment program EMBOSS (http://www.ebi.ac.uk/emboss/align/), using the default parameters.

Based on the identification of a variant located within a glucocorticoid response element (GRE) half-site (−2232C/G), double-stranded DNA oligonucleotides containing the consensus (TTAGCAGTGTGAGAAAGACAAATACA) and nonconsensus (TTAGCAGTGTGAGAAAGACAAATACA) sequences were generated for performance of gel shift assays using nuclear extract generated from an immortalized hypothalamic cell line (IVB cells, the kind gift of J. Kasckow, MD, PhD, University of Cincinnati). Assays were performed using the Gel Shift Assay System (Promega, Madison, Wisconsin) per the manufacturer's instruction. After annealing complementary oligonucleotides (95°C for 5 minutes, 25°C for 30 minutes), double-stranded probes were [³²P]-ATP labeled using T4 kinase (Promega) and purified using a Bio-Spin 30 chromatography column (Bio-Rad, Hercules, California). Incorporation of radiolabel was more than 1 × 10⁵ cpm per nanogram of DNA. Binding assays were performed using the Gel Shift Assay System per the manufacturer's instructions. Nuclear extract (5 μg) was incubated for 20 minutes with 100 000 cpm of each oligonucleotide probe. Competitor oligonucleotides were added at × 10 the concentration of the labeled probes. Samples were immediately separated by electrophoresis (250 V for 20 minutes) on a Novex 6% DNA retardation gel (Invitrogen, Carlsbad, California), after which gels were dried and bands visualized by autoradiography.

Genomic DNA from animals confirmed to be homozygous for the rhCRH-H1 haplotype was extracted for polymerase chain reaction (PCR)–based cloning. Primers were modified from those previously reported for human CRH reporter constructs.²⁴ The forward primer was 5′ GCG GAA TTC GGC TCA TAA CTC CTT TAT GTG CTT GC 3′ (containing an EcoRI site) and the reverse primer was 5′ AAA GGA TCC GAG GGA CGT CTC CGG GGC 3′ (containing a BamHI site). These primers produced a 783–base pair (bp) amplification product (from rhCRH −660 to +123) in which none of the rhCRH-H2 cosegregating loci were present. Reaction mixtures (50 μL) contained 100 ng of DNA, 0.1mM deoxyribonucleotide triphosphate, 0.5μM each primer, and 2.5 U of PfuUltra High Fidelity DNA Polymerase and PfuUltra Buffer (Stratagene, La Jolla, California). Amplifications were performed using a thermocycler (9700; PerkinElmer, Waltham, Massachusetts) with 1 cycle at 95°C; 30 cycles of 95°C for 30 seconds, 61.5°C for 30 seconds, and 72°C for 5 minutes; and a final 10-minute extension at 72°C. Following cleavage using EcoRI and BamHI, PCR product was separated by electrophoresis and isolated using a QIAquick gel extraction kit (Qiagen, Valencia, California). Cleaved products were then ligated into EcoRI/BamHI–digested pDsRed-2.1 (BD Biosciences, Mountainview, California) using standard molecular cloning techniques.

Attempts to test functionality of the −2232C/G site using rhCRH-H1 and rhCRH-H2 reporter constructs (−3458 to +974) generated inconsistent results because they exhibited both extremely low levels of transfection efficiency and increased cell mortality, which we attributed to the size of these constructs. This was true among experiments performed in a variety of cell lines (COS-7, IVB, HT22, and JEG). Glucocorticoid response element half-sites are known to interact across long distances,²⁵ and there are several GREs and GRE half-sites in the proximal promoter of the CRH gene. Since we specifically wanted to test the function of the −2232C/G polymorphism, located in a more distal GRE half-site, we designed double-stranded DNA cassettes (−2269>−2200) containing either the −2232C or −2232G allele using oligonucleotides obtained from a commercial vendor (Operon, Huntsville, Alabama). These cassettes were digested with BglII and HindIII and ligated into the pDsRed multiple cloning site in a position 5′ to the rhCRH −660 to +123 insert (which contained the proximal promoter). Fidelity of resultant constructs was verified by sequencing.

We used mouse HT22 hippocampal cells (the kind gift of David Schubert, PhD, The Salk Institute). This cell line²⁶ was selected because of its high concentration of GRs²⁷ and because the hippocampus expresses relatively high levels of CRH, especially after stimulation.²⁸ HT22 cells were propagated in Dulbecco modified Eagle medium (Gibco; Invitrogen, Carlsbad, California) supplemented with 10% fetal bovine serum and maintained at 37°C in a humidified incubator (5% carbon dioxide). Cells were seeded at a density of 5×10⁴ per well in a 12-well tissue culture plate. When 60% to 70% confluent, either the −2232C or −2232G rhCRH pDsRed reporter construct (1 μg) was cotransfected with a green fluorescent protein reporter (0.5 μg of pGlow-TOPO [Invitrogen]) using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. Cells were treated with 30μM forskolin (Sigma, St Louis, Missouri) or 30μM forskolin + cortisol (Sigma)^20,27,29 12 hours following transfection. We selected a low cortisol dose (1nM) for this study based on reports that this dose produced type I GR activation, without type II GR activation or type I:type II dimerization, in the HT22 cell line.³⁰ This is important because high cortisol doses (sufficient to activate type II receptors) induce cell death in the hippocampus.^31,32 Expression of DsRed was monitored 48 hours following treatment using a microscope (IX70; Olympus America, Melville, New York) interfaced with a digital camera (ORCA-ER; Hamamatsu, Tokyo, Japan). Region-of-interest analysis was performed using OpenLab software (http://www.improvision.com/products/openlab/). A CRH promoter expression value was obtained by dividing the DsRed intensity by that for green fluorescent protein. Experiments were performed 4 times, with treatments in some instances performed in duplicate. Measurements were made by 2 observers (M.G. and C.S.B.) and averaged. Data were analyzed by analysis of variance (ANOVA), with treatment (no treatment, forskolin, forskolin + cortisol) and genotype (−2232C and −2232G) included as independent variables.

Using standard extraction methods, DNA was isolated from whole blood, collected from the femoral vein under ketamine anesthesia (15 mg/kg, intramuscular). To genotype for −2232C>G, a functional SNP that represented one of the major haplotype clades, a portion of the 5′ flanking region (−2730>−2204) was amplified from 25 ng of genomic DNA with flanking oligonucleotide primers (forward: 5′GGT TCT CAT TTA AAC CGA GTG ATC 3′; reverse: 5′AAG TGG CTC CAA CTA GGG AGT AAG 3′) in 20-μL reactions using AmpliTaq Gold (Applied Biosystems) and 4mM magnesium chloride (Invitrogen) according to the manufacturer's instructions. Amplifications were performed on a thermocycler (9700; PerkinElmer) with 1 cycle at 96°C followed by 30 cycles of 94°C for 15 seconds, 55°C for 15 seconds, and 72°C for 30 seconds and a final 3-minute extension at 72°C. Because results of 5′ exonuclease assay genotyping for −2232C/G were unreliable, we performed restriction digest by EarI (1.5 μL; New England Biolabs, Beverly, Massachusetts) using 10 μL of PCR product in a total volume of 20 μL for 4 hours at 37°C. Samples were separated by electrophoresis on 10% polyacrylamide gels, and the C and G alleles were identified by direct visualization following ethidium bromide staining. Three hundred seven animals from the NIHAC colony were genotyped using this method.

Rhesus macaque (M mulatta) infants at NIHAC were randomly selected to be reared with their mothers or in a nursery by human caregivers (peer reared). Mother-reared animals were reared in social groups composed of 8 to 14 females (about half of whom had same-aged infants) and 2 adult males. Peer-reared animals were separated from their mothers at birth and hand reared in a neonatal nursery for the first 37 days of life. For the first 14 days, they were kept in an incubator and hand-fed. From day 15 until day 37, they were placed alone in a nursery cage and provided a blanket and a terry-cloth–covered, rocking surrogate. A bottle from which the infants would feed was fixed to the surrogate. At 37 days of age, they were placed in a cage with 3 other age-mates with whom they had continuous contact. At approximately 8 months of age, animals (both mother reared and peer reared) were placed into age-matched social groups and housed in large indoor/outdoor runs through late adolescence/adulthood (3.5-5 years of age), at which point the cohorts were divided into sex-limited groups. All procedures described were approved by the National Institute on Alcohol Abuse and Alcoholism and National Institute of Child Health and Human Development Animal Care and Use committees.

Blood and cerebrospinal fluid (CSF) samples were taken at 3.5 to 5 years of age (n = 72 and n = 139, respectively) under ketamine anesthesia (10 mg/kg, intramuscular) for measurement of levels of CRH, corticotropin (ACTH), and cortisol. Samples were obtained within 5 minutes of capture. Cisternal CSF samples were immediately aliquoted in polypropylene tubes and frozen in liquid nitrogen. Blood samples were placed on wet ice and centrifuged at 4°C for 20 minutes, after which plasma was aliquoted and immediately frozen in liquid nitrogen. The CSF and plasma samples were stored at −70°C until assayed.

The CSF was assayed for CRH by radioimmunoassay coupled with C18 Sep-pak extraction using our own antibodies as previously described.³³ Extracted samples were reconstituted in a buffer at a concentration 3-fold higher than the original samples. The sensitivity of the assays was 15 pg/mL for CRH. Plasma ACTH and cortisol radioimmunoassays used commercially available kits (ICN and DPC, respectively, Los Angeles, California) and were run according to the instructions of the manufacturers. All assays were run in duplicate and the interassay and intraassay coefficients of variation were all less than 12%.

Mother-reared rhesus macaque infants (n = 95) were scored for 10 minutes twice a week in the social group for the first 24 weeks of life. Behavior definitions are listed in the eTable. Interobserver reliability was established at greater than r = 0.85 for all behaviors.

Behavioral responses to an unfamiliar intruder were recorded in adolescent/adult rhesus macaques (n = 113). “Intruder” animals were selected based on the age and sex of the test subjects, such that adult animals (mean [SD], 7.5 [1.6] years of age) were exposed to an unfamiliar adult of the same sex and subadults (mean [SD], 3.3 [1] year of age) were exposed to subadult animals of the same sex. Intruders were also selected to match the size of the test subjects as closely as possible. All intruder animals were completely unfamiliar to the test subjects. Prior to the intruder challenge test, the intruder animal was placed into an individual transfer cage, measuring 0.76 m wide × 0.63 m deep × 0.91 m high, for a 30-minute acclimation period. All subjects were tested 3 at a time in the home run. For the test, 3 randomly grouped animals were locked into the outdoor portion of their home run, an enclosure measuring 2.64 m wide × 3 m long × 2.44 m high. After 10 minutes, the intruder animal's cage was placed directly at the front of the enclosure and behavioral scoring of the test subjects was initiated. One observer was assigned to each test subject, which was observed for 30 minutes using focal animal continuous recording. Behaviors recorded are listed in the eTable. Behaviors were generally scored in seconds in duration. Vocalization, aggression, and approach intruder were scored in frequency. Interobserver reliability was established at greater than r = 0.85 for all behaviors.

Young adult macaques (aged 3.5-5 years; n = 74) were allowed to freely consume an aspartame-sweetened 8.4% (volume to volume ratio) alcohol solution for 1 hour per day, 5 days a week in their home run. This method consisted of 3 phases, which have previously been reported¹⁷: (1) spout training; (2) initial alcohol exposure; and (3) experimental period. During the 6-week experimental phase, alcohol and vehicle were dispensed 5 days a week (Monday-Friday) from 1 PM to 2 PM while the animals were in their home-cage environment. Animals were fitted with a collar implanted with an identifier chip so that volumes consumed could be recorded for each individual.

Using data collected from mother-reared infants in the home cage, we performed factor analysis to generate behavioral dimensions from behaviors averaged over the 18th through the 24th weeks of life, a developmental stage during which infants are being weaned by their mothers and begin to spend much of their time interacting with other members of the social group and exploring. Factor analysis was also performed on behaviors collected during the intruder challenge test. In each instance, principal components analysis, followed by standard orthogonal (varimax) normalized rotation, was performed, and factor scores were generated for each individual. Behavioral dimensions were labeled by investigators with expertise in primate behavior and were consistent with factors that have been generated by other primatologists studying temperamental differences³⁴ or responses to an intruder.³⁵ The behavioral dimensions generated were then used as dependent variables in ANOVA to test the hypothesis that animals carrying the −2232G allele (carried on the rhCRH-H2 haplotype) would show different scores on one or more behavioral dimensions. Because age and sex are known to influence responses to an intruder, age (adult vs subadult) and sex (male vs female) were included as independent measures for these analyses.

We analyzed baseline levels of CRH, ACTH, and cortisol using ANOVA. Analyses were performed with rearing condition as a coindependent variable, but since there were no interactions with genotype and because the inclusion of this variable did not reduce the residual variance, it was removed from the analyses. Alcohol consumption data were also analyzed by ANOVA. To control for potential differences in drinking patterns among test cohorts, we also generated z scores controlling for cohort. Prior reports from this laboratory have shown that alcohol consumption is higher among peer-reared macaques and that functional genetic variants can interact with early rearing history to influence levels of ethanol self-administration.^17,36 Because of these established differences and the fact that males generally consume more ethanol than females, both sex and rearing condition were included as coindependent variables in the analysis.

Although this is an outbred colony of macaques,³⁷ to verify that our effects were attributable to CRH variation and not to general heritability of stress responsivity, we repeated our analyses using a set of 6 randomly selected biallelic genetic markers used for genotyping in our colony. There were no effects of the other markers tested (data not shown), suggesting our results to be attributable to effects of rhCRH variation on our phenotypes of interest. −2232C/G allele frequencies were in agreement with Hardy-Weinberg. There were 2 animals with the G/G genotype among these data sets, which were collapsed with G allele carriers for the purpose of this analysis. To account for nonhomogeneity of variances, dependent variables were rank-transformed when appropriate. Analyses were performed using Statview 5.01 statistical software (SAS Inc, Cary, North Carolina). Criterion for significance was set at P ≤ .05.

We screened the rhCRH gene and 3′ and 5′ flanking regions and identified 40 polymorphic sites (Figure 1A). Variants were assigned positions relative to the rhCRH transcription start site, which was predicted at 68 804 514 of chromosome 8 (January 2006 assembly, rheMac2, http://genome.ucsc.edu/). The 5 most common haplotypes are shown in Figure 1B. rhCRH-H1 carried the common allele at almost every site. The second most common (rhCRH-H2) was composed of the minor alleles for a 21-marker haplotype that extends across the 5740-bp region (Figure 1 and Figure 2). Genotyping across populations (NIHAC, University of Wisconsin, Morgan Island) demonstrated this haplotype to be present at similar frequencies (minor allele frequency = 8%, data not shown), and focused resequencing showed they were in perfect linkage disequilibrium (LD) (D′) (Figure 1B).

Cladistic clustering of the 5 most common rhesus macaque haplotypes demonstrated 2 major haplotype clades (Figure 2, cluster 1 = H1, H3-H5 and cluster 2 = H2). Using data from European (CEU), Nigerian (YRI), and Asian (ASN) human populations for comparison, we also generated cladograms for the human CRH locus (Figure 2). The most common haplotypes (rhCRH-H1 in rhesus macaque and C1 haplotype cluster in human) share most alleles with the ancestral sequence (data not shown). The Nigerian population (YRI) showed more haplotype complexity at the CRH locus than did the European (CEU) population (Figure 2), whereas the entire Asian population was clustered in C1. The cladograms also demonstrate the existence of alternative (yin-yang) haplotype clades in both species.

Glucocorticoids are critical to the regulation of the CRH system, and among the SNPs that differed between the 2 major haplotype clades was one that was present in a consensus GRE half-site (−2232C/G, boxed in Figure 1B).^38,39 Rhesus macaques exhibited only 92% sequence similarity with humans for the 173 bp surrounding this site (−2341>−2168), but among all primate species screened, the GRE half-site within this sequence was 100% conserved (data not shown). In silico analysis indicated that −2232C>G would result in complete disruption of this GRE.

Consistent with the prediction that −2232C>G would lead to loss of a GRE half-site, gel shift assays performed using hypothalamic nuclear extract (generated from IVB cells, which express GRs and in which corticosteroid effects on CRH expression have been extensively studied) demonstrated an attenuation in protein binding to −2232G probes (Figure 1C). Similar results were obtained using a nuclear extract enriched for GR (data not shown). To test whether corticosteroid sensitivity of the CRH regulatory region was altered as a consequence of this SNP, reporter assays were performed in HT22 cells transfected with −2232C or −2232G pDsRed constructs. There were main effects of treatment (F_2,27 = 11.8; P < .001) and genotype (F_1,27 = 4.7; P = .04). A significant decrease in reporter activity following cortisol treatment was observed in forskolin-stimulated cells expressing the −2232C allele constructs (Figure 1D) (Tukey-Kramer, P < .05), but this effect was not observed in those transfected with G allele constructs (Figure 1D, P=.60) (no treatment: C, n = 5, G, n = 4; forskolin, C, n = 6, G, n = 5; and forskolin + cortisol, C, n = 6, G, n = 7).

No other variants in the H2 haplotype were predicted to disrupt transcription factor binding sites for factors known to be involved in CRH transcriptional control. While another site (−2336A>T) that was part of this haplotype predicted the creation of a YY1 site, gel shift assays indicated that DNA×protein interactions were not altered as a consequence of this variant (data not shown). There were neither nonsynonymous SNPs nor variants predicted to alter exon/intron splicing or messenger RNA stability.

Among adolescent/young adult animals, those with the rhCRH −2232C/G genotype (H2) had lower CSF levels of CRH (Figure 3A) (F_1,70 = 4.7; P < .04) and higher plasma levels of ACTH (Figure 3B) (F_1,138 = 7.96; P = .005), with genotype accounting for 6% to 7% of the variance. Cortisol levels did not differ according to genotype (P=.70, data not shown). These findings were the same after controlling for differences in early stress exposure (ie, rearing condition).

Principal component extraction, followed by standard varimax normalized rotation, identified 3 orthogonal factors with eigenvalues more than 1, which, together, explained 73.0% of the variance (Table 1). These were labeled as “anxious,” “exploratory/bold,” and “attached.”³⁴ Animals with the C/G genotype rated higher on the “exploratory/bold” dimension (Figure 4A) (F_1,93 = 6.6; P = .01) than did those homozygous for the C allele, with genotype accounting for 7% of the total variance of this trait. In contrast, the other behavioral dimensions were not influenced by genotype (anxious, P=.70; attached, P=.90).

When we performed factor analysis of the behavioral responses to an unfamiliar intruder, 5 orthogonal factors with eigenvalues greater than 1 were generated: “agonistic/highly aggressive,” “harm avoidant,” “curious/bold,” “threatening,” and “reactive.” These factors explained 68% of the total variance (Table 2). Prior reports from other laboratories have demonstrated there to be both sex and age differences in responses to an unfamiliar intruder.³⁵ In agreement with this, we found that during social intrusion subadults were more harm avoidant than adults (F_1,105 = 12.19; P = .007) and that males rated higher on both the curious/bold (F_1,105 = 6.7; P = .01) and reactive factors (F_1,105 = 5.6, P = .02) (data not shown).

While there were no main effects of rhCRH genotype on the curious/bold response to an unfamiliar intruder (F_1,105 = 1.13; P = .29), −2232C/G genotype strongly interacted with sex (F_1,105 = 10.4; P = .002) to influence this type of response, accounting for 8% of the total variance. Males that were carriers of the G allele exhibited significantly increased levels of curious/bold responding when presented with an unfamiliar male compared with C homozygotes (Figure 4B) (Tukey-Kramer, P < .05), while no such differences were observed among females (not significant). This remained true following rank-transformation of the data (F_1,105 = 8.3; P < .005). There were no main effects of genotype nor were there interactions among genotype, age, and sex on the other reactivity dimensions.

C/G animals consumed more alcohol in the social group than did those with the C/C genotype (Figure 5) (F_1,72 = 8.5; P = .005). Genotype accounted for 12% of the total variance in ethanol consumption, and a main effect of genotype remained after controlling for rearing condition (peer reared vs mother reared) and sex (male vs female) (F_1,66 = 5.04; P ≤ .03; C/C vs G, Tukey-Kramer, P < .05). It also remained following both statistical correction for testing cohort (F_1,66 = 4.5; P ≤ .04) and rank-transformation of the data (F_1,66 = 7.9; P < .007). In each of these analyses, G allele carriers exhibited higher levels of alcohol self-administration (Tukey-Kramer, P < .05).

The neurobiological systems that influence addiction vulnerability may do so by acting on reward pathways, behavioral dyscontrol, and vulnerability to stress and anxiety. The CRH system is one that is known to be dysregulated in alcohol-dependent subjects and in which variation may influence both stress responding and novelty seeking. We screened the rhCRH gene and regulatory region and identified a SNP that disrupted a GRE half-site in the CRH regulatory region. We demonstrated that this functional variant predicted endocrine and behavioral differences in addition to levels of alcohol self-administration in the rhesus macaque.

Glucocorticoids are known to be critical to CRH transcriptional regulation via binding to GREs. Glucocorticoid response elements are hormone-inducible enhancer elements⁴⁰ that can activate or repress transcription through cooperative action with additional GRE half-sites or other DNA-bound proteins.^41,42 The nucleotide sequence in which we identified variation (−2232C/G) is a consensus GRE half-site. Glucocorticoid response elements exert effects in regions distant from the core promoter,⁴³ and half-sites can interact with one another over distances of hundreds of nucleotides.²⁵ It is also known that cooperative binding among receptors bound to GRE half-sites leads to high occupancy at low concentrations and nonlinearity of binding.⁴⁴ We used GR-containing nuclear extracts, generated from a hypothalamic cell line, to show that −2232G results in decreased levels of DNA×protein interactions. These results support the notion that the AGAACA (−2232C) site is functional.

Using an in vitro reporter assay, we also demonstrated there to be attenuated sensitivity of the rhCRH promoter to corticosteroids in cells expressing the G allele construct. Even with the introduction of the G allele, some degree of corticosteroid sensitivity was maintained. This is not surprising, given that other GR binding sites are present in the rhCRH promoter. However, it is possible that the GRE half-site in which we identified variation could be functionally important through another mechanism. Binding of a hormone-receptor complex to a consensus GRE half-site can aid in recruitment of GRs to a particular genomic region and facilitate binding of other receptor monomers to surrounding nonconsensus sites.⁴⁵ This site may aid in GR binding to the nonconsensus sites in the rhCRH promoter that have been demonstrated to be important to regulation of CRH transcriptional activity, an effect that would be unlikely to be appreciated in a reporter system, which is enriched for the genomic region of interest. Regardless of the exact mechanism, we do indeed find effects of the −2232G allele on CSF CRH levels and HPA axis output, and these effects are consistent with decreased sensitivity of the CRH promoter to glucocorticoids.

In the paraventricular nucleus of the hypothalamus, corticosteroids reduce CRH expression, but in some extrahypothalamic sites, such as the amygdala, they produce opposite effects.^29,46-49 We find that plasma levels of ACTH are increased, whereas, in CSF, levels of CRH are decreased. The lower CSF CRH levels can be accounted for based on the likelihood of a low or nonexistent contribution of paraventricular nucleus–median eminence CRH to the CSF pool relative to other sources, such as the amygdala.³³ Our finding is consistent with what has been previously reported in human and nonhuman primates, which indicates opposing rhythms of CSF levels of CRH and HPA axis output.^33,43 In rodents, some exploratory/novelty-seeking strains have higher levels of Crh expression in the hypothalamus but lower levels in the amygdala,⁵⁰ and studies have shown that bold/aggressive rat strains have higher baseline levels of ACTH.¹⁰ We examined behavioral dimensions during infancy and found that carriers of the −2232G allele were more “exploratory/bold,” exhibiting relatively high levels of environmental exploration and aggression. We also used a challenge that has been developed to test behavioral responses to social intrusion in adolescent or adult primates. This test is not only an excellent tool for looking at coping responses during exposures to social stress but is especially useful for assessment of individual differences in behavioral inhibition, as approaching an unfamiliar animal is considered to be a risky and impulsive behavior.³⁵ Consistent with the fact that G allele carriers rated higher on the exploratory/bold dimension during infancy, we found that male G allele carriers exhibited a more bold and active response to an unfamiliar intruder. An individual that readily approaches novel objects or conspecifics may do well in certain social situations but may face higher risk of predation or attack than a more cautious, harm-avoidant individual. Such behaviors might, therefore, be predicted to confer selective advantage at particular developmental or life-history stages and in certain environmental contexts.

Twenty mutations at the rhCRH locus were in LD (D′) with −2232C>G, and roughly 15% of individuals in the National Institutes of Health population were heterozygous at these sites (minor allele frequency = 8%). Sequencing of the regions containing each polymorphic site in this haplotype demonstrated that these 21 markers were in perfect LD in multiple macaque populations, in which they also occurred at similar frequencies. Neutral sites in LD with an allele maintained by selection have a better chance of remaining in the population.^51,52 The persistence of the 2 divergent rhCRH haplotypes over time may suggest that they have been subject to selection such that at least one of the alleles on each background is being selected—possibly in a particular environmental context—while the rest are hitchhiking. The effects of the H2 haplotype on our intermediate phenotypes (ie, decreased CSF CRH levels but increased plasma ACTH levels) are consistent with our in vitro findings, which demonstrated the loss of a GRE half-site and decreased sensitivity to corticosteroids. However, as we did not sequence beyond the boundaries of this haplotype, we do not know whether there is another functional marker in LD with −2232G, nor do we know about the block size, which might provide further evidence for selection. That being said, several studies in humans^53,54 have shown there to be evidence for selection at the CRH locus, in which, similar to the rhesus macaque, we have observed alternative, yin-yang haplotype clades (Figure 2). As in the rhesus macaque, the major human CRH haplotypes have been shown to vary in terms of their in vitro promoter activity, and among the observed differences are those pertaining to glucocorticoid sensitivity.²⁰

Because of differences in their behavioral and physiologic responses to stress, the types of stress-induced abnormalities to which bold, proactive individuals and harm-avoidant, reactive individuals are vulnerable are distinct. Whereas the latter are at risk for internalizing disorders, such as depression and anxiety, the former are more likely to develop externalizing conditions, primarily characterized by impaired impulse control.¹⁰ In humans, anxiety is a risk factor for developing alcohol problems, and stress exposures can lead to craving and relapse.^14,15,55 It is also known that impulsivity or behavioral dyscontrol can predispose individuals to early and uncontrolled alcohol intake.^14,15 We found that carriers of the −2232G allele exhibited higher levels of alcohol consumption when tested in the social group, a model for high-risk, impulsivity-related alcohol consumption. They also exhibited lower levels of the serotonin metabolite 5-hydroxyindoleatic acid (data not shown), a neurochemical endophenotype observed both in macaques exposed to early life stress and among individuals with early-onset, type II alcoholism.^37,56 It may be that, in humans, genetic variation that altered CRH system function could influence multiple behavioral dimensions (ie, both neuroticism and extraversion) and that variants that placed an individual at the extremes of these spectra (ie, inhibited and anxious/stress reactive vs bold/impulsive and novelty seeking) could increase the risk for developing alcohol use disorders.

The present investigation includes some of the largest sample sizes that are currently available for a relatively homogeneous outbred macaque colony. Nevertheless, the numbers of animals represented in the data sets used for this study are quite small relative to those used for performing association studies in human populations. For this reason, we neither have the power to perform refined haplotype analysis nor can we examine complex interactive effects with other functional genetic variants (ie, HTTLPR, MAOA-LPR, OPRM1 C77G) that may be contributing to the variance in our phenotypes of interest. We have elected to use an approach that relies on cladistic clustering of haplotypes and functional characterization of variants to determine whether genetic variation is associated with our phenotypes of interest, using factor analysis to generate behavioral clusters so that the number of statistical comparisons is minimized. Experimental studies in humans as well as nonhuman primates^16,37,57,58 have repeatedly demonstrated the gain in power that comes from designs that genotype a relatively small number of subjects for functional genetic variation; measure fine-grained, quantitative phenotypic traits; and apply the power of general linear models to test the a priori hypothesis. In this study, we demonstrate that a functional rhCRH promoter variant is associated with relevant intermediate phenotypes in addition to behavioral traits that were assessed at multiple developmental points and have been, in both primate and nonprimate species, repeatedly and reliably linked to differences in CRH system activity.

In a social environment and because of niche-specific and frequency-dependent selection, divergent physiologic and behavioral stress responses will be maintained in a population because the benefits and risks of opposing behavioral strategies will balance one another in different contexts. Two basic behavioral phenotypes have been observed and are thought to be maintained by balancing selection in a wide variety of animal species, including humans. While one is characterized as being aggressive and bold, adopting a more proactive coping style and fight-or-flight responses to stress, the other is more harm avoidant and exhibits reactive, anxietylike behaviors following stress. A conceptual framework has been put forth that suggests differences in these behavioral strategies have implications not only in terms of selection but in terms of vulnerability to psychopathology.¹⁰ The CRH system is proposed to critically contribute to such differences. There is evidence that variation at the CRH locus predicts levels of behavioral inhibition in children,⁵⁹ and genetic variation at the CRHR1 locus is suggested to be a risk factor for alcohol problems. Functional CRH haplotypes have been associated with various nonpsychiatric disorders such as rheumatoid arthritis,⁶⁰ but few studies examining the role of CRH variation in the vulnerability to psychiatric disorders have been reported. Our data suggest common selective mechanisms in humans and macaques and may indicate that corticosteroid-insensitive CRH haplotypes could increase risk for externalizing disorders in human subjects.

Correspondence: Christina S. Barr, VMD, PhD, Laboratory of Clinical Studies, Primate Section, National Institute on Alcohol Abuse Alcoholism Division of Intramural Clinical and Biological Research, TR 112, PO Box 529, Poolesville, MD 20837 (cbarr@mail.nih.gov).

Submitted for Publication: December 17, 2007; final revision received February 7, 2008; accepted March 12, 2008.

Author Contributions: Drs Barr, Dvoskin, Higley, Heilig, Suomi, and Goldman contributed equally to this work.

Financial Disclosure: None reported.

Funding/Support: This work was funded by NARSAD and the National Institute on Alcohol Abuse and Alcoholism and National Institute of Child Health and Human Development intramural programs.

Additional Contributions: Karen Smith, MLS, assisted in the preparation of the manuscript and the research and animal care staff at the National Institute on Alcohol Abuse and Alcoholism and National Institutes of Health Animal Center assisted in data collection.