Alcohol trait and transcriptional genomic ... - Wiley Online Library

10 downloads 56584 Views 486KB Size Report
Center, Department of Pharmacology and Neuroscience, Texas. Tech. University Health .... To gain insight into the molecular mechanics of phenotypic differences ... ad libitum water and NIH Mouse/Auto 6F 5K52 diet consisting of not less than ...
Genes, Brain and Behavior (2008) 7: 677–689

# 2008 The Authors Journal compilation # 2008 Blackwell Publishing Ltd

Alcohol trait and transcriptional genomic analysis of C57BL/6 substrains M. K. Mulligan†, I. Ponomarev†, S. L. Boehm II‡, J. A. Owen§, P. S. Levin†, A. E. Berman¶,**, Y. A. Blednov†, J. C. Crabbe††,‡‡, R. W. Williams§§, M. F. Miles¶¶ and S. E. Bergeson§,* †

Waggoner Center for Alcohol and Addiction Research, University of Texas at Austin, Austin, TX, ‡ Behavioral Neuroscience Program, Department of Psychology, Binghamton University, Binghamton, NY, §South Plains Alcohol and Addiction Research Center, Department of Pharmacology and Neuroscience, Texas Tech University Health Science Center, Lubbock, TX, ¶ Department of Neurology, University of California San Francisco, **Neurology Service, Veterans Affairs Medical Center, San Francisco, CA, †† Portland Alcohol Research Center, Portland VA Medical Center, ‡‡ Department of Behavioral Neuroscience, Oregon Health & Science University, Portland, OR, §§Department of Anatomy and Neurobiology, University of Tennessee Health Science Center, Memphis, TN, and ¶ ¶ Center for Study of Biological Complexity, Departments of Pharmacology, Toxicology and Neurology, Virginia Commonwealth University, Richmond, VA, USA *Corresponding author: S. E. Bergeson, South Plains Alcohol and Addiction Research Consortium, Department of Pharmacology and Neuroscience, Texas Tech University Health Science Center, 3601 4th St., STOP 6592, Lubbock, TX 79430, USA. E-mail: [email protected]

C57BL/6 inbred mice have been widely used as research models; however, widespread demand has led to the creation of several B6 substrains with markedly different phenotypes. In this study, we report that two substrains of C57BL/6 mice, C57BL/6J (B6J) and C57BL/6NCrl (B6C), separated over 50 years ago at two different breeding facilities differ significantly in alcohol consumption and alcohol preference. The genomes of these two substrains are estimated to differ by only 1–2% of all gene loci, providing a unique opportunity to extract particular expression signatures between these substrains that are associated with quantifiable behavioral differences. Expression profiling of the cortex and striatum, hippocampus, cerebellum and the ventral brain region from alcohol-naı¨ve B6C and B6J mice showed intervals on three chromosomes that are enriched in clusters of coregulated transcripts significantly divergent between the substrains. Additional analysis identified two genomic regions containing putative copy number differences between the substrains. One such region on chromosome 14 contained an estimated 3n copy number in the B6J genome compared with B6C. Within this interval, a gene of unknown function, D14Ertd449e, was found to doi: 10.1111/j.1601-183X.2008.00405.x

be both associated with alcohol preference and vary in copy number across several inbred strain lineages. H2afz, Psen1, Wdfy1 and Clu were also identified as candidate genes that may be involved in influencing alcohol consumption. Keywords: Alcohol, B6, brain, C57BL/6J, C57BL/6NCrl, consumption, expression, microarray, preference, substrain Received 18 January 2008, revised 14 March 2008, accepted 22 March 2008

Among the most commonly studied and widely utilized inbred strains is C57BL/6 (B6). The B6 strain was isolated from C57BL/10 prior to 1937 and was used to found the C57BL/6J (B6J) colony at The Jackson Laboratory in 1948. The B6J colony was subsequently used to found the C57BL/6N (B6N) colony at the National Institutes of Health (NIH) in 1951. In 1974, Charles River Laboratory started their own subcolony, C57BL/6NCrl (B6C), from the 32nd generation of the B6N colony. Thus, the B6J and B6C substrains have arisen through the propagation of B6 colonies by different breeding facilities and have been inbred separately for over 5 decades and 200 generations. Colony separation leads to genetic divergence over time, and genetic drift can occur within 20 generations after separation (Bailey 1977, 1982). If a subcolony is founded from an inbred strain prior to the 40th generation of inbreeding, genetic drift can result from the remaining heterozygosity within the original inbred strain. Indeed, single nucleotide polymorphism (SNP) analysis of B6 substrains supports the existence of residual heterozygosity despite extensive inbreeding at the time the substrains were separated (Petkov et al. 2004). Genetic drift can also occur if spontaneous gametal mutations become fixed within the subcolony, and several examples have been identified between B6J and B6C. For instance, a spontaneous deletion occurring in the B6J strain resulted in the loss of several exons of the nicotinamide nucleotide transhydrogenase (Nnt) gene (Huang et al. 2006). Additionally, of 867 microsatellite markers, 13 were found to be polymorphic between B6J and B6C (Hovland et al. 2000). Phenotypically, the two B6 substrains differ in a wide range of traits including cardiac function during anesthesia (Roth et al. 2002), body weight (Green et al. 2007) and aspects of the fear response (Radulovic et al. 1998; Siegmund et al. 2005). B6J mice have been used to study alcohol-related traits since the discovery that B6J mice have a high alcohol preference (HAP) phenotype (McClearn & Rodgers 1959). B6J mice also consume more alcohol than other commercially available inbred mouse strains (Belknap et al. 1993). The phenotype of HAP in the B6J substrain has remained

677

Mulligan et al.

remarkably stable over time, likely because alcohol preference is a complex trait controlled by many genes of small to moderate effect size (Wahlsten et al. 2006). Unfortunately, most of the genes involved in driving initial alcohol preference have yet to be elucidated. In 1982, researchers discovered a difference in alcohol preference between B6J and a now extinct B6 substrain derived from B6N (Blum et al. 1982). Recently, it was discovered that B6J also differed from B6C for several alcohol-related phenotypes, including sensitivity to fetal alcohol syndrome (Green et al. 2007), ethanol deprivation effect (Khisti et al. 2006) and brain dopamine release after exposure to alcohol (Ramachandra et al. 2007). To gain insight into the molecular mechanics of phenotypic differences observed between these two substrains, we further characterized several alcohol-related behaviors and completed brain region transcriptome analyses on alcoholnaı¨ve B6J and B6C mice.

Materials and methods Animal handling at the Jackson Laboratory C57BL/6J mice colonies from the Jackson Bar Harbor facility were fed ad libitum water and NIH Mouse/Auto 6F 5K52 diet consisting of not less than 18% crude protein, 6% crude fat, 5% crude fiber, 8% ash and 3% added minerals. Animals were maintained on a 12-h light/dark cycle, and all other environmental controls were standard. The breeding protocol at The Jackson Laboratory used a trio of two females and one male, with offspring weaned at 28 days. All mice purchased directly from The Jackson Laboratory were shipped at approximately 56 days and held with same-sex animals for 2 weeks or more prior to killing or testing at the University of Texas at Austin.

Animal handling at the Charles River Laboratories Breeding at the Charles River facility also occurred using a breeding trio, and C57BL/6NCrl offspring were weaned at 21 days. Mice were fed a diet of Purina 5L79 that consists of 18% crude protein, 5.2% crude fat, 5.2% crude fiber, 5.7% ash and 3.3% minerals (Cano et al. 2001). All mice from Charles River Laboratories were shipped at approximately 56 days and held with same-sex animals for 2 weeks or more prior to killing or testing at the University of Texas at Austin. Animals were maintained on a 12-h light/dark cycle, and all other environmental controls were standard.

and a sipper tube containing tap water. Sipper tubes were constructed from sterile BD Falcon 50-ml conical tubes (BD Biosciences, San Jose, CA, USA), rubber stoppers and dual ball bearing steel sipper spouts. In this two-bottle choice paradigm, animals were allowed free access to food, alcohol solutions and water. Starting at a 3% alcohol and water solution (w/v) for 4 days and increasing in increments of 3% up to 21% alcohol and water (w/v), animals were allowed access to each additional concentration for 5 days before being switched to the next highest concentration. The physical location (right or left) of the water and alcohol solutions was switched daily to reduce the effects of potential side preference. Sipper tubes were weighed daily to the nearest 0.01 g to measure consumption of alcohol and water. To avoid including aberrant data from leaky tubes or from mice that played with their bottles, an outlier test was conducted on the daily total fluid intake, and outliers were excluded on a per day basis. Consumption was determined as grams of alcohol consumed per kilogram body weight per day. Preference was measured as the amount of alcohol consumed over the total amount of fluid consumed per day, with >50% designated as a preference for alcohol. The statistics software GRAPHPAD PRISM 4 (Jandel Scientific, Costa Madre, CA, USA) was used to complete two-way mixed design ANOVAs for alcohol consumption and alcohol preference. Strain was a betweensubjects factor and alcohol concentration was a within-subjects factor.

Mice bred in-house Adult mice from each substrain and both sexes were tested using the same two-bottle choice paradigm described above with the exception that mice were tested for 10 days using a 20% alcohol and water solution (n ¼ 12 females per group and n ¼ 6 and 7, respectively, for B6JUT and B6CUT male mice). Differences in consumption and preference between the strains were tested for significance by a ttest within each sex.

Taste discrimination Vendor-purchased mice Taste preference was measured using a saccharine and quinine twobottle choice paradigm similar to that described above. Female adult mice from each strain were housed individually in single cages and provided with two sipper tubes containing either a 0.033% saccharine solution (sweet) and tap water or a 0.03 mM quinine solution (bitter) and tap water. Consumption was measured for 4 days, and bottles were alternated each day to avoid side preference effects.

Loss of righting reflex Vendor-purchased mice

C57BL/6J (B6J) and C57BL/NCrl (B6C) mice were bred in-house from adult male and female mice (1:1) supplied from both Jackson and Charles River Laboratories, respectively. The abbreviations B6JUT and B6CUT shall be used to designate the offspring of mice purchased from each vendor and subsequently bred in-house to differentiate them from the vendor-purchased animals. B6JUT and B6CUT mice were weaned at 21 days and fed a diet of PROLAB RMH 1800 that contains a minimum of 18% crude protein, 5% minimum crude fat and 5% minimum crude fiber. Weaned animals were housed with up to five same-sex animals on a 12-h light/dark cycle until experimentation began after they were at least 9 weeks of age.

Sensitivity to the hypnotic effects of alcohol was assessed by testing the duration of the loss of righting reflex (LORR) after exposure to alcohol. Ten alcohol-naı¨ve B6J and 10 B6C adult female animals were injected with 3.8 g/kg ethanol intraperitoneally (i.p.). Ethanol-induced LORR was measured by placing mice on their backs in a 908 plastic trough. Loss of righting reflex was described as the inability of the mouse to right itself within 30 seconds. Return of the righting reflex was operationally defined as the ability of the mouse to right itself twice in a 1-min period. The duration of LORR was measured as the time between the LORR and the return of the righting reflex. Upon recovery of the righting reflex, retro-orbital sinus blood was taken for gas chromatographic determination of blood alcohol concentration (BAC). The statistical software GRAPHPAD PRISM 4 was used to complete t-tests to evaluate the effect of strain on the duration of LORR and the BAC at the time of recovery.

Alcohol consumption and preference measurement

Initial sensitivity to alcohol

Vendor-purchased mice

Vendor-purchased mice

Ten individually housed adult female mice from each vendor were provided with a sipper tube containing an ethanol and water solution

Nine alcohol-naı¨ve adult female mice of each substrain were used. Initial sensitivity to the hypnotic effects of ethanol was assessed

Animal handling at the University of Texas at Austin

678

Genes, Brain and Behavior (2008) 7: 677–689

B6 substrain trait and genomic analysis using a modified version of LORR (Ponomarev & Crabbe 2002, 2004). Briefly, an animal was injected with 3 g/kg ethanol i.p. (20% v/v in saline) and then immediately placed in a cylindrical restrainer. The restrainer was then gently rotated 908 every 2–3 seconds. Within 2 min of the injection, the mice are heavily intoxicated and remain on their back after two successive 908 turns. Thus, the mouse is considered to have lost its righting reflex when it is no longer able to right itself within 5 seconds from a supine position. Latency to LORR in seconds was used as a measure of initial sensitivity.

Custom complementary DNA microarrays Complementary DNA (cDNA) microarrays were printed in-house at the University of Texas, exactly as previously published (Mulligan et al. 2006). The spotted clones on the microarrays represented approximately 13 000 unique known genes (annotations current as of 29 June 2007 using SOURCE at http://source.stanford.edu/cgi-bin/source/ sourceSearch). Specific details for custom-spotted cDNA microarray printing has been described elsewhere (Schena et al. 1995).

and for each experiment, Ch1S ¼ Ch1 þ (Ch2med  Ch2), where Ch1S represents the standardized Ch1 intensity for an experimental value of a gene, Ch1 and Ch2 are the original green and red intensities, respectively, that represent the experimental value for that gene and Ch2med is the median value of the red channel intensities for that gene across all experiments. Outliers were removed by gene within each experimental group from the resulting normalized and standardized Ch1 intensity data. Two filters were applied to the microarray dataset. First, a filter was applied that removed genes that did not meet a minimum criteria of 6 (out of 8) Ch1 intensity values for each experimental group. Second, genes that did not exceed a minimum average Ch1 intensity threshold of 4.64 (log2 scale) across all experimental groups tested were discarded on the basis that the signal was below the threshold for biological relevance to be determined. Statistical significance was determined for each gene between experimental groups by t-test and false discovery rate estimation in the form of Storey’s Q value (Storey & Tibshirani 2003) or additionally as described below. Both P value and q value (q) were calculated using the free statistical software program R (Ihaka & Gentleman 1996). Highly significant transcripts were selected for each brain region based on q < 0.05.

Brain tissue collection and RNA isolation Alcohol-naı¨ve intact female animals aged 84 days from each supplier were killed by cervical dislocation, and the cortex (including striatum), hippocampus, cerebellum and ventral brain region (remaining brain region) were dissected. The olfactory bulb was excluded from this study. Dissected tissue was immediately frozen in liquid nitrogen and stored at 808C. RNA was extracted according to the manufacturer’s protocol using RNA STAT-60ä (Tel-Test, Inc., Friendswood, TX, USA). RNA concentration and integrity were determined using the NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA) and Agilent 2100 Bioanalyzer and RNA 6000 Nano LabChip kit (Agilent Technologies, Palo Alto, CA, USA), respectively.

Microarray hybridization and normalization Total RNA for each brain region was used in subsequent microarray experiments. Microarray hybridizations were performed according to the manufacturer’s protocol using the Genisphere (Hatfield, PA, USA) Array 350 kit using 3 mg per sample. All experimental samples were end labeled with Cy-3 (green channel 1) and hybridized against a common reference sample (total RNA from 100 adult male C57BL/6J mice) that was end labeled with Cy-5 (red channel 2). The B6 reference sample was selected for use over other potential reference samples because it is the strain most commonly used in alcohol studies, was the reference for the mouse sequence in the Genome Project and has been used for all other microarray experiments in our laboratory. The use of this particular reference allows for uncomplicated comparisons with other datasets generated at our microarray facility. Use of a common reference resulted in a total of 64 microarrays completed; 8 arrays for each B6 substrain in each of the four brain regions. Hybridized microarrays were immediately scanned using Axon GenePix 4000B dual channel laser scanners (Molecular Devices, Union City, CA, USA). Microarrays were gridded using the software package GENEPIX version 5.0 (Molecular Devices). Lowess within-print-tip normalization was performed for each array using the R package, Statistics for Microarray Analysis (Yang et al. 2002). All normalized intensity information from the arrays used in this study was uploaded and stored in the Longhorn Array Database (LAD) (Killion et al. 2003).

Microarray analysis Red channel 2 (Ch2) and green channel 1 (Ch1), net median intensities, were downloaded from LAD, and the Ch1 intensities for each gene were adjusted (standardized) by the reference Ch2 intensities for each gene as follows. All intensity values less than 0 were given a value of 1. All linear intensity information was then converted to log2 intensity values. For every gene on the microarray Genes, Brain and Behavior (2008) 7: 677–689

Detection of similarly expressed genes in all four brain regions Transcripts that passed the intensity and missing data filter were evaluated for consistency of divergent expression between B6J and B6C across all four brain regions. The Student’s two-tailed t-test was used to compare B6J and B6C in each brain region. The resulting tvalue (t) was used to estimate the effect size for each brain region as follows: Cohen’s d: d ¼ 2t/Odf, where d is the effect size measure and df is the degrees of freedom. A two-tailed Z-test (z) was performed on the effect sizes to assess significance. For a transcript to be considered to be highly significantly divergently expressed between B6J and B6C in all four brain regions, we defined the following criteria: P(z) < 0.01, average jdj  0.8 and average fold change j1.5j. Significant genes were clustered and displayed using CLUSTER 3.0 (de Hoon et al. 2004) and JAVA TREEVIEW 1.1.1 (Saldanha 2004) freely available software.

Validation of expression data and estimation of gene copy number on chromosome 14 Quantitative real-time polymerase chain reaction to confirm expression levels of D14Ertd449e and Plac9 Probes for D14Ertd449e and Plac9 were purchased from Applied Biosystems (Foster City, CA, USA). About 1 mg of total RNA was taken from each brain region of four B6J and four B6C mice and pooled together to create n ¼ 4 whole brain RNA pools for each substrain. Validation was performed as described in Ponomarev et al. (2006) with the exception that 18 S ribosomal RNA was used as the endogenous control.

Comparison of expression between inbred strains Whole brain expression data corresponding to loci within the chromosome 14 region containing putative changes in gene copy number were collected for several inbred strains from the UCHSC BXD Whole Brain M430 2.0 (Nov06) RMA database available through the GeneNetwork resource at http://www.genenetwork.org. The number of replicates, n, per strain as well as the mean probe expression values and the corresponding standard error of the mean (SEM) were retrieved for B6J, DBA/2J, FVB/NJ and SJL/J and used to calculate t-values and test for significant differences in gene expression between strains using a two-tailed t-test. Affymetrix probes 1452590_a_at, 1428738_a_at, 1445238_at and 1447939_a_at were used to retrieve expression data for Plac9, D14ertd449e, Ccl21b and 4933409K07Rik, respectively. Average whole brain expression in B6C was estimated from our microarray data by averaging the intensity value for each gene across all four brain regions.

679

Mulligan et al.

Quantitative real-time polymerase chain reaction to confirm gene copy number Genomic DNA was obtained for B6J, B6C, FVB/NJ, NZB/BlNJ, BUB/ BnJ, SJL/J and DBA/2J. Reactions were performed in a total volume of 25 ml with 12.4 ml SYBRâ green master mix (Applied Biosystems), 0.5 ml of each primer (20 mM) and 50 ng genomic DNA. Polymerase chain reaction (PCR) amplification and detection were performed using the ABI 7300 Real-Time PCR System (Applied Biosystems) for 40 cycles each of 908C for 30 seconds, 568C for 30 seconds and 728C for 45 seconds, with an initial denaturation for 10 min at 908C. Primer sequence (Table 1) was chosen based on National Center for Biotechnology Information (NCBI) build 37.1, and custom primers were designed and purchased from Integrated DNA Technologies (Coralville, IA, USA). Peptidylprolyl isomerase F (Ppif) exists as a single copy and intergenic primer pairs 1 and 2 align to three repeated sequences on chromosome 14 (Table 1). Relative genomic copy number was calculated by 3  2DDCT using the comparative threshold cycle (CT) method described in the Applied Biosystems thermocycler manual to calculate the amount of the target normalized to an endogenous control and relative to a calibrator. Briefly, CT value was determined for intergenic primer sets 1 and 2 and Ppif in each inbred strain using the thermocycler software, and the average CT value for each primer set was calculated based on four replicates. DCT is calculated for each strain by target  endogenous control, where Ppif is the endogenous control and the targets are represented by intergenic primer sets 1 and 2. DDCT is the fold difference in gene expression and is given by DCT target  DCT calibrator, where the B6J substrain was used as the calibrator for primer sets 1 and 2. The 2DDCT values for intergenic primer sets 1 and 2 for each substrain were multiplied by 3 to estimate copy number because B6J is thought to have three copies of the genes within the chromosome 14 region based on sequence information. Complete sequence information within this region was not available for the other inbred strains.

Overrepresentation analyses Overrepresentation for pathways, transcription factor binding sites (TFBSs) and chromosomal location were completed using WebGestalt, oPOSSUM and a macro within Microsoft Excel, respectively. Genes with similar patterns of expression in each substrain for each brain region were analyzed for overrepresented pathways using GOTree in WebGestalt (Zhang et al. 2004, 2005) and for TFBS overrepresentation using oPOSSUM (Ho Sui et al. 2005). A P < 0.01 significance cutoff was applied in both cases unless otherwise stated. To investigate the effects of genotype on chromosome locationspecific gene expression, we compared chromosomal frequencies of genes differentially expressed between B6J and B6C with a random distribution. First, all transcripts that passed the expression detection threshold were sorted by their gene symbols and the level of statistical significance. Then, transcripts with unknown gene symbols and all duplicated gene copies with lower statistical significance were removed. The remaining genes with unique genetic identifiers were then resorted according to their chromosomal location. The numbers of tightly linked genes significantly different between the two strains and regulated in the same direction [P(z) < 0.05 and average jdj

 0.5] were then calculated within 15-gene overlapping bins along the whole genome. To calculate P values associated with these frequencies, the frequencies were then compared with a distribution of about 1.2 million frequencies calculated the same way and based on the same number of significantly regulated genes but distributed randomly along the whole genome. This distribution was generated using 100 genome-wise permutations and a random number generator in Excel. This random distribution was virtually identical to a binomial distribution of the number of successes in a sequence of n ¼ 15 and the probability of success equal to the number of significantly regulated genes in the same direction divided by the total number of unique genes that passed the filters (0.1). Therefore, we calculated P values based on the binomial distribution (‘binomdist’ function in Excel). The frequency of 7 closely located similarly regulated genes (of 15 genes in a bin) was considered to pass the significance threshold with P < 0.0003 and false discovery rate 0.8 and average fold change >j1.5j. Table S2: Brain region-specific differences between substrains that correspond to a 5% FDR. A small number of genes are highly statistically significantly divergent between B6 substrains. B6J indicates expression that is higher in B6J and B6C shading indicates expression that is lower in B6J relative to B6C, respectively. C, cerebellum; Chr, chromo-

Genes, Brain and Behavior (2008) 7: 677–689

some; CLID, clone ID; Entrez, Entrez ID; GBID, GenBank accession number; H, hippocampus; M, ventral brain region; X, cortex and striatum. Table S3: Functional group overrepresentation analysis for genes highly significantly divergent in one or more brain regions. Significant overrepresentation determined using the GOTree function in WebGestalt. Bold lettering and normal lettering indicate genes with higher expression and lower expression, respectively, in B6J compared with B6C. Please note: Blackwell Publishing is not responsible for the content or functionality of any supplementary materials supplied by the authors.

689