Isolation and Confirmation of a Calcium Excretion Quantitative Trait Locus on Chromosome 1 in Genetic Hypercalciuric Stone-Forming Congenic Rats
Richard R. Hoopes, Jr.*,
Frank A. Middleton,
Saunak Sen,
Paul A. Hueber*,
Robert Reid*,
David A. Bushinsky and
Steven J. Scheinman*
Departments of *Medicine; Neuroscience & Physiology, State University of New York Upstate Medical University, Syracuse, New York; Department of Epidemiology and Biostatistics, University of California San Francisco, San Francisco, California; and Department of Medicine and Physiology, University of Rochester, Rochester, New York
Address correspondence to: Dr. Richard R. Hoopes, Jr., Department of Medicine, SUNY Upstate Medical University, 750 E. Adams Street, Syracuse, NY 13210. Phone: 315-464-5290; Fax: 315-464-5797; hoopesr{at}mail.upstate.edu
Received for publication August 8, 2005.
Accepted for publication February 20, 2006.
Hypercalciuria is the most common risk factor for kidney stonesand has a substantial genetic component. The genetic hypercalciuricstone-forming (GHS) rat model displays complex changes in physiologyinvolving intestine, bone, and kidney and overexpression ofthe vitamin D receptor, thereby reproducing the human phenotypeof idiopathic hypercalciuria. Through quantitative trait locus(QTL) mapping of rats that were bred from GHS female rats andnormocalciuric Wistar Kyoto (WKY) male rats, loci that are linkedto hypercalciuria and account for a 6 to eight-fold phenotypicdifference between the GHS and WKY progenitors were mapped.GHS x WKY rats were backcrossed to breed for congenic rats withthe chromosome 1 QTL HC1 on a normocalciuric WKY background.Ten generations of backcrosses produced N10F1 rats, which wereintercrossed to produce rats that were homozygous for GHS lociin the HC1 region between markers D1Mit2 and D1Mit32. On a high-calciumdiet (1.2% calcium), significantly different levels of calciumexcretion were found between male congenic (1.67 ± 0.71mg/24 h) and male WKY control rats (0.78 ± 0.19 mg/24h) and between female congenic (3.11 ± 0.90 mg/24 h)and female WKY controls (2.11 ± 0.50 mg/24 h); the congenicspreserve the calcium excretion phenotype of the GHS parent strain.Microarray expression analyses of the congenic rats, comparedwith WKY rats, showed that of the top 100 most changed genes,twice as many as were statistically expected mapped to chromosome1. Of these, there is a clear bias in gene expression changefor genes in the region of the HC1. Of >1100 gene groupsanalyzed, one third of the 50 most differentially expressedgene groups have direct or secondary action on calcium metabolismor transport. This is the first QTL for hypercalciuria to beisolated in a congenic animal.
Through successive inbreeding of the most hypercalciuric normalSprague-Dawley rats over 64 generations, we have developed astrain of rats that has a marked, consistent increase in urinecalcium excretion (1–3). With successive generations ofinbreeding, urine calcium excretion rose in a linear mannerand seemed to plateau at approximately the 30th generation,consistent with polygenic inheritance of the phenotype. Urinecalcium excretion in these rats now exceeds that of the parentalstrain by eight- to 10-fold, and all of these rats develop kidneystones (4–6). The rats now are termed "GHS" for genetichypercalciuric stone-forming rats. Extensive characterizationof the underlying physiology of excess calcium excretion inthe GHS rats has revealed that they absorb an excessive amountof dietary calcium, fail to reabsorb sufficient filtered calcium,and have increased bone resorption in response to vitamin D,suggesting a systemic dysregulation of calcium homeostasis (7–10).Serum levels of calcium are normal, but there are excessivelevels of the vitamin D receptor in all target tissues (11–14)and of the calcium-sensing receptor in kidney (13). The pathophysiologyof the hypercalciuria in the GHS rats parallels that in humanswith idiopathic hypercalciuria, many of whom have been shownto have excessive intestinal calcium absorption, reduced renaltubular calcium reabsorption, and excessive bone demineralization(15–18). Humans with idiopathic hypercalciuria also havebeen found to have excessive numbers of vitamin D receptors(19). The GHS rat is the only spontaneous animal model for kidneystone formation.
We previously performed quantitative trait locus (QTL) mappingof F2 rats from a cross between GHS and normocalciuric WKY ratsas a first step in mapping the genes that contribute to hypercalciuriain the rat model (20). In that study, we identified a calciumexcretion QTL (hypercalciuria 1 [HC1]) with a logarithm of oddsscore of 2.91 on chromosome 1, centered near the marker D1Rat169.Several other possible QTL were identified on other chromosomesin this study, but these did not meet the same level of significance.The next step in identifying the gene(s) underlying the HC1QTL was to isolate this QTL onto a control, normocalciuric genomeby developing congenic strains. We used the approach of speedcongenics (21,22), selecting for progeny of GHS x WKY backcrossesthat contained the D1Rat32-D1Mit32 interval of the GHS chromosome1 over the course of 10 generations. The N10F1 generation ofthese rats was intercrossed, and the resulting progeny thatwere homozygous for the HC1 GHS region were phenotyped for calciumexcretion. We report here that the calcium excretion phenotypewas preserved in the GHS HC1 congenics, confirming unequivocallythat the HC1 QTL contributes significantly to hypercalciuriain the GHS rat model. We also compared gene expression betweenthe congenic and control strains using whole-genome microarrayexpression profiling of the kidney, small intestine, and femoralosteoclast cells (OC) and report that the HC1 QTL region containsa significantly greater number of genes with altered expressionthan expected by chance alone. Moreover, comprehensive functionalgene group analysis indicates that a considerable number ofgene groups that are involved in primary or secondary calciummetabolism are among the most altered. Taken together, thesedata strongly indicate a primary role for the HC1 QTL gene(s)in regulating the hypercalciuric phenotype in GHS rats.
Construction of GHS Rat Congenic Strains
Inbred GHS rats, developed by the selective breeding of themost hypercalciuric rats from a Sprague-Dawley colony, havebeen maintained at the University of Rochester for more than60 generations. GHS congenic strains were developed by breedingthe GHS donor strain to a WKY recipient strain (maintained asa closed colony at the University of Rochester; these are thesame WKY rats that were used to generate F2 rats for QTL mapping[20]) using a speed congenics approach (21,22). GHS x WKY F1rats were generated from GHS female x WKY male cross. Male progenythen were bred for 10 successive backcrosses to WKY female rats,each time selecting progeny that were heterozygous for the regionof chromosome 1 that contained the HC1 QTL that we had identifiedpreviously (20). The marker interval that we preserved throughto the N10F1 congenics, containing the GHS alleles, was fromD1Rat32 to D1Mit32, which comprised a genetic distance of approximately100 cM (Figure 1). N10F1 male and female rats then were intercrossed,and their progeny were genotyped. N10F3 progeny that had thechromosome 1 interval desired then were intercrossed to generatean N10F4, the WKY.GHS(RN01) rats, and these rats were phenotypedfor calcium excretion as described here.
Figure 1. Region of genetic hypercalciuric stone-forming (GHS) chromosome 1 introgressed into WKY rat to produce the GHS congenic rats, overlaid onto the quantitative trait locus (QTL) logarithm of odds score map for chromosome 1. The solid liner indicates the markers that are homozygous for the GHS alleles in the congenic rats. The 95% confidence interval for HC1 extends from D1Rat193 to D1Rat142.
Genotyping
Genomic DNA was prepared using the Puregene DNA Isolation Kit(Gentra Laboratories, Inc., Minneapolis, MN) from kidneys thathad been isolated immediately after death, frozen, and storedat –70°C. Primers for microsatellite markers wereobtained from Research Genetics, Carlsbad, CA; www.resgen.comproductsRtMPs.php3).PCR reactions were carried out as recommended and analyzed onDNA sequencing gels, on 4% agarose gels by standard methods,or on an ABI 3100 Genetic Analyzer. For the last, fluorescentlylabeled primers (Applied Biosystems, Foster City, CA) were used,with analysis of the products done using ABI GeneMapper 3.0software for allele calling. Amplifications were performed usingAmplitaq (Applied Biosystems), according to the manufacturer’srecommendations, on MJ Research PTC200 DNA Engine thermocyclers(23).
Phenotyping
All rats were maintained on normal rat diet from weaning until8 wk of age. At 8 wk, the rats were transferred to individualmetabolic cages, given free access to distilled water, and fed13 g/d of a defined calcium diet. The rats were phenotyped essentiallyas described previously, although no vitamin D was added totheir food (3–8,11,24). The high-calcium diet contained1.2% calcium and 0.65% phosphate (TD.90312.PWD; Harlan Teklad,Indianapolis, IN), whereas the low-calcium diet contained 0.02%calcium and 0.65% phosphate (TD.89379.PWD; Harlan Teklad). Ratswere placed on a particular calcium diet on the first day ofthe phenotyping protocol, and urine calcium excretion was measuredin 24-h urine collections on days 6 through 9. For this study,32 rats (16 congenic, eight of each gender, and 16 WKY, eightof each gender) were placed on a high-calcium diet for the first9-d period, and then the diet of these rats was switched toa low-calcium diet for the second 9-d period. Rats were killedat the end of this low-calcium period for tissue sampling forRNA analysis as described in the RNA Extraction and cRNA ProbeLabeling section. A second group of 16 rats (eight congenic,four of each gender, and eight WKY, four of each sex) were placedon a high-calcium diet for another 9-d period, duplicating thephenotyping on the high-calcium diet so that tissue samplescould be obtained after the high-calcium diet. The 24-h urinesamples were collected in 50-ml tubes that contained 0.25 mlof concentrated HCl. Food consumption was monitored for theentire experiment, and all calcium excretion values were excludedfrom the analysis, for any study period, for any rats that ate<10 g of the 13 g of food provided during each day of urinecollection.
Whole urine calcium levels were measured using o-CresolphthaleinComplexone reagent with a calcium calibration standard (C7508-400and C7503-STD; Pointe Scientific, Canton, MI), according tothe manufacturer’s protocol.
Statistical Analyses
Calcium excretion phenotypes are presented as mean ±SD. Comparisons of calcium excretion between the GHS congenicstrain and the normocalciuric WKY strain were performed by one-wayANOVA (GraphPad Prism 4; GraphPad, San Diego, CA), with Tukey-Kramercorrection for multiple comparisons, to assess significance.
Microarray Methods Tissue Samples Used for Microarray Analysis.
To screen for candidate genes that could underlie the hypercalciuricphenotype, we performed whole transcriptome analysis on selectedtissue samples from the GHS and WKY rats. The tissues includedwere from OC from the femurs, whole kidney, and small intestine(from rats that received both high-calcium and low-calcium diets),for a total of 12 tissue types (two genotypes x three tissuesx two diets). Each of these tissues was processed in a differentmanner to extract the RNA as described in the RNA Extractionand cRNA Probe Labeling section. Only tissue samples from maleGHS and male WKY rats were used in these studies.
RNA Extraction and cRNA Probe Labeling.
Intact kidneys and small intestine samples from each of fourrats in each group were pulverized in liquid nitrogen usinga nuclease-free mortar and pestle. The total RNA was extractedfrom approximately 30 mg of tissue powder using the RNeasy kit,with the QiaShredder columns, according to the manufacturer’sinstructions (Qiagen, Valencia, CA). The total RNA from theintestines then was subjected to polyA selection, to removebacterial RNA contamination, using the Oligotex kit (Qiagen).
To isolate RNA from OC in the femurs, we used the method ofDavid et al. (25), with modification. Briefly, femurs were scrapedcleaned of bone marrow and minced in ice-cold minimal essentialmedium (MEM)/HEPES buffer and centrifuged to collect the cells.These cells were incubated in charged tissue culture flasksthat contained MEM (37°C, 5% CO2) for 4 h to permit adhesionof OC and remaining stromal cells. The plates then were treatedwith trypsin/EDTA to remove any remaining stromal cells andwashed in fresh medium. These cells then were scraped from theplates using RLT Lysis Buffer in the RNeasy kit (Qiagen) andprocessed for RNA purification.
The purified total RNA or mRNA from each sample was eluted in40 µl of RNAase-free water and concentrated by vacuumcentrifugation to 11 µl. For assessment of the qualityand the concentration of the total RNA, 1 µl was analyzeddirectly on an Agilent Technologies Bioanalyzer RNA Pico Chip(Agilent Technologies, Palo Alto, CA) following the manufacturer’sinstruction. For all total RNA samples, the intensity of the28S rRNA band exceeded that of the 18S band by a ratio of atleast 1.5, and no obvious degradation was seen. For the mRNAsamples, no ribosomal bands were detected, and only a high-weightsmear of mRNA was visible, with no obvious degradation.
We pooled equal quantities of RNA from each of the rats in eachof the groups into a single total RNA sample for microarrayanalysis. In the labeling reactions for these experiments, themRNA fraction from the total RNA was reverse-transcribed usingan oligo-dT primer coupled to a T7 RNA polymerase recognitionsequence (all reagents used in the labeling reactions were partof the Two-Cycle cDNA Synthesis Kit and the IVT Labeling Kit[both from Affymetrix, Santa Clara, CA]). After second-strandsynthesis in the presence of RNase H and subsequent DNA purification,the double-stranded cDNA template was used as a template forin vitro transcription (IVT). After IVT, the antisense RNA productwas reverse-transcribed, using random primers that filled inthe T7 ends. After second-strand synthesis and cDNA templatepurification, a second round of IVT was carried out using approximately200 ng of template. This time, during IVT, a fixed concentrationof biotinylated ribonucleotides was incorporated into the cRNAproducts. After 4 to 6 h, the IVT reaction was stopped, andDNAase 1 was added to the tube to eliminate the template.
Gene Chip Hybridization Procedures.
After purification and quantification, 15 µg of biotinylatedcRNA was hydrolyzed randomly to 35 to 200 nucleotides in a fragmentationbuffer solution (94°C, 35 min). It then was added to a hybridizationbuffer (100 mM 2-morpholinoethanesulfonic acid, 1 M [Na]+, 20mM EDTA, 0.01% Tween-20, 0.1 mg/ml herring sperm DNA, and 0.5mg/ml acetylated BSA) that contained known concentrations ofpositive control genes (50 pM Oligo B2; and 1.5, 5, 25, and100 pM of Escherichia coli bioB, bioC, bioD, and cre, respectively).The entire hybridization solution was heated (99°C, 5 min),equilibrated (45°C, 5 min), and centrifuged at maximum speedfor 5 min before being injected into an RAE230 2.0 GeneChip.The RAE230 2.0 GeneChip contains >31,000 probe sets, interrogating>28,000 well-substantiated rat genes, and represents themost complete analysis platform available for transcriptionalprofiling in this species. After sample loading, the GeneChipswere hybridized at 45°C for 16 h with constant rotation(60 rpm, then washed and stained on the Fluidics Station [Affymetrix])according to the EukGE-WS2v4 protocol. After washing and staining,fluorescence images were acquired using the GeneChip 3000 Scanner.After the GeneChips were scanned, the Affymetrix software (GeneChipOperating System, Santa Clara, CA) calculated the intensityof the signal from each perfect-match probe relative to thesignal for the mismatch probe and also determined whether thegene was present in the sample (and generated a probabilityvalue associated with this determination).
Analysis of Microarray Data.
To assess changes in gene expression in our animal groups, wefirst implemented the robust multichip analysis method of normalization(26) on the set of 12 microarrays using GeneTraffic (IobionInformatics, La Jolla, CA). Then, to determine the genes withthe most consistent (and therefore the most strain-dependentor diet-dependent) differences in expression across the tissuetypes, we analyzed pairwise differences in expression for thematched samples from each strain for each tissue type. Genesthat showed statistically significant changes in expressionby genotype were ranked by the magnitude of the fold change(Table 1), and the 100 largest changes in expression were mappedby position across the entire genome and across chromosome 1,using 40-megabase bins. To evaluate positional biases in thedistribution of highly changed genes, we used 2 statistics tocompare the distribution of all genes represented on the RAE230array with the distribution of the top 100 changed genes. Thisanalysis indicated a clear bias toward transcripts on chromosome1 and particularly in the 160- to 200-megabase region on 1q.These calculations were performed without correction for multipletesting. The top changed genes (in terms of the significanceof the effect for either diet or genotype) were used in a heatmap cluster analysis.
Table 1. Top 25 increased and decreased genes with significant changes in expression according to genotypea
To evaluate potential interactions of diet and genotype, weperformed a two-way ANOVA using genotype and calcium diet asindependent variables and tissue source as a blocking factor.The P values for this comparison were adjusted using the Benjamini-Hochbergfalse discovery rate correction. Genes with significant interactionsare listed in Table 2.
Table 2. All genes in the QTL with significant changes according to both diet and genotypea
To determine whether robust changes in expression of functionallyrelated transcripts were involved in calcium metabolism, weused custom-written software (PathStat [27]) to analyze theexpression of >1100 different gene groups represented bythe content on the RAE230 array. These groups represent theentire publicly curated content of the array, according to theGene Ontology and Enzyme Commission databases. In this methodof data analysis, a log-normalized ratio of each of the genesin the four pairwise comparisons is constructed and the meanand SD of each set of ratios for each functional gene groupis calculated, permitting gene group effects to be transformedinto z scores. The 25 most consistently increasing and decreasinggene groups, according to z scores, subsequently were identifiedand used in a "heat map" cluster analysis. This analysis revealedthe patterns and magnitudes of the most consistent gene groupeffects in our data, in a completely unbiased manner. Thesegene groups were inspected to determine their involvement incalcium metabolism or transport.
Real-Time Quantitative Reverse Transcription–PCR Methods
We used quantitative reverse transcription–PCR (qRT-PCR)to validate a selected number of results from the microarrayanalysis. Specifically, we chose to examine each of the high-calciumsample sets from the two genotypes (GHS congenic versus WKY)from the three tissues (kidney, osteoclast, and small intestine)and one of the low-calcium sample sets (kidney) from the twogenotypes (n = 8 sample sets total). Matching amounts of RNA(3 µg) from each of the four independent samples in eachset of eight sample types were used in a RT reaction (SuperscriptII; Invitrogen, Carlsbad, CA) with an oligo dT primer. The productsof the RT reaction then were subjected to RNase H treatmentand second-strand synthesis using DNA Polymerase I (Invitrogen).The genes that we selected for validation included two geneswith increased expression in the congenic tissue types ( hemoglobin[Hbb] and protein tyrosine phosphatase, receptor type, epsilonpolypeptide [Ptpre]), and two genes with decreased expressionin the congenic tissues (NAD synthetase 1 [NADsynth1] and regulatorof G protein signaling 10 [RGS10]). We also included a referencegene to control for levels of starting material in each sample(18S RNA). Primers for each of these genes were designed usingPrimer3 software to produce amplicons of 80 to 130 bp. Eachprimer pair was tested on a mixed cDNA template to ensure amplificationof a single specific band, using HotMaster Taq(Eppendorf) andstandard cycling conditions (hot start at 94°C for 2 minfollowed by 40 cycles of melt at 94°C for 30 s and annealingand extension at 59°C for 60 s). After the specificity ofeach of the amplicons was confirmed, real-time qRT-PCR was performedusing four replicates from each sample set for each gene (32PCR reactions total per gene) with 12 ng of cDNA as template.Reactions were performed in a 25-µl volume using the sameenzyme and cycling parameters as before (HotMasterTaq and buffer)except that 0.25 µl of SYBRGreenI dye (Molecular Probes,Eugene, OR) was added to each well before cycling. These reactionswere run on 96-well plates using an ABI 7000 Real-Time SequenceDetection instrument (Applied Biosystems). End point melt-curveanalysis confirmed the presence of single amplicons in eachreaction well, and amplification in the absence of templatefailed to produce any signal as a result of primer dimerizationand extension. Statistical analysis of the real-time data wasperformed using a one-tailed pairwise t test to compare thedifference in the mean number of cycles to threshold CT forthe transcript of interest in the two matching sample types(GHS congenic and WKY) from each of the four tissues. A foldchange was calculated according to the following formula: FoldChange = 2CT.
Preservation of Haplotype
We produced the GHS congenic strain WKY.GHS(RN01) by introgressingthe rat chromosome 1 region that contained the HC1 QTL fromthe GHS donor into the WKY recipient. Ten backcrosses were performedto produce essentially complete homozygosity for the WKY genomein the region outside HC1 (all 90 markers tested on chromosomes2 through X and on chromosome 1 beyond the HC1 region were homozygousfor WKY). N10F4 intercrosses were generated and genotyped todemonstrate preservation of the HC1 QTL region of chromosome1. The congenic intercrosses were homozygous for the GHS allelesof chromosome 1 markers D1Rat32, D1Rat200, D1Rat193, D1Mgh11,D1Rat169, D1Rat76, D1Rat119, D1Rat142, and D1Mit32 (Figure 1).This region of the chromosome extends from approximately 64to 153 cM on the genetic map, encompassing more than one halfof chromosome 1. It fully encompasses the 95% confidence intervalfor HC1, which extends from D1Rat193 to D1Rat142 (20).
Phenotypes Calcium Excretion.
The HC1 congenic rats were phenotyped on high- and low-calciumdiets for 9 d, with 24-h urine collections made on days 6 through9. Urine calcium levels were measured, and averages of the fourcollections were compared between the congenic and the controlWKY rats. Urine calcium excretion for congenic male rats thatwere on the high-calcium diet averaged 1.67 ± 0.71 comparedwith 0.78 ± 0.19 mg/24 h for the WKY male controls (Figure 2).For the female rats that were on the high-calcium diet, thecongenics excreted 3.11 ± 0.90 compared with 2.11 ±0.50 mg/24 h for the WKY rats (Figure 3). On the high-calciumdiets, excretion differences between the male congenic and maleWKY rats were significant, as were the differences between femalecongenic and the female WKY rats (for the one-way ANOVA, F =39.88 with 7 df). On the low-calcium diets, calcium excretionlevels were much lower, and the differences in calcium excretionswere much smaller for the same comparisons (Figures 2 and 3).
Figure 2. Calcium excretion phenotype for male congenic and control WKY rats on high- and low-calcium diets. Each bar represents the mean excretions values for four 24-h measurements for rats on the high- or the low-calcium diet. Numbers of rats in each group are indicated in parentheses.
Figure 3. Calcium excretion phenotype for female congenic and control WKY rats on high- and low-calcium diets. Each bar represents the mean excretions values for four 24-h measurements for rats on the high- or the low-calcium diet. Numbers of rats in each group are indicated in parentheses.
Microarray Gene Expression.
The results of four pairwise comparisons between male congenicand WKY tissues strongly supported the localization of the HC1QTL to chromosome 1q and suggested a number of genes and biologicpathways that might underlie the phenotype (Tables 1 and 2;Figures 4 through 6). In this study, we first focus on describingthe genes that were most affected and then discuss the biologicpathways that were most altered. Finally, we present the resultsof a formal analysis of localization biases in the data set,as we attempt to define attractive candidate genes for furtherstudy.
Figure 4. Hierarchical cluster analysis of genes with significant changes in expression according to genotype or diet reveals alteration of chromosome 1 QTL region genes and calcium regulatory pathways. (A) Top genes according to changes in congenic rats compared with WKY rats. (B) Top genes according to changes in high-calcium diet versus low-calcium diet in both strains. Genes are clustered by their Pearson intercorrelations, so those that are increased in the six comparative analyses are clearly separated from those that are decreased. Genes with high levels of expression are colored in red, whereas those with reduced expression are colored in green.
Figure 5. Hierarchical cluster analysis of functional gene groups with significant changes in expression according to genotype or diet reveals alteration of chromosome 1 calcium regulatory pathways. Top 50 functional gene groups in the microarray data comparisons between matching congenic and WKY tissue samples (A) or high and low calcium diet animals (B). Only the top 50 gene groups are shown (of >1100 groups that were analyzed). Each of these gene groups contains at least three probes that were expressed (average five genes per group were expressed). Groups of genes are clustered by their Pearson intercorrelations, so those that are increased in the six comparative analyses are clearly separated from those that are decreased. Approximately one third of these gene groups have primary or secondary action on calcium metabolism or transport (indicated by * or **, respectively). Gene groups with high levels of expression are colored red, whereas those with reduced expression are colored green.
Figure 6. Selective enrichment of significantly changed genes in the chromosome 1 QTL region. These graphs show the distribution of the top 100 changed genes (according to the magnitude of the fold change, where P < 0.05 in the four pairwise comparisons). (A) The localization of significantly changed transcripts is significantly different from the expected distribution according to 2 analysis (P = 0.02006). (B) The chromosome 1 bias is strongest in the 160- to 200-megabase bin, and this difference also is significantly different than expected by 2 analysis (P = 0.000001). Taken together, the array data strongly support the localization of HC1 to this chromosomal region.
To begin to examine the single-gene data, we first looked foreffects of genotype in a pairwise manner. Genes with the mostrobust changes among all those that were changed significantlyaccording to genotype (n = 661) are listed in Table 1. We pointout that of these 50 genes, 16 were localized to the QTL regionon chromosome 1q (boldface entries). We also examined the geneson the basis of their P values and performed a heat map clusteranalysis of the top genes for both genotype and diet effects(Figure 4). In terms of the genes with the most statisticallyrobust genotype effect, we point out again that one third ofthese genes are located in the chromosome 1 HCI QTL region (boldfaceentries; the figure shows 31 of the top 50 genes). In additionto the single genes with the largest or most robust effect ofgenotype, we identified an even greater number of genes withsignificant effects of diet (n = 1579) and a small number ofgenes (n = 74) with significant changes in expression accordingto both diet and genotype. Eight of these genes were locatedin the chromosome 1 QTL region (Table 2). Examination of thepatterns of change in these genes indicated that some transcriptsmight display an interaction of diet and genotype on their expressionpattern. A separate two-way ANOVA confirmed that a small numberof genes (10 genes) did indeed display such significant interactions,even after corrections for multiple comparisons. Only one ofthese genes, protocadherin 16, was located in the chromosome1 QTL region.
To examine the functional gene groups with the most robust alterationsin expression, we performed a comprehensive analysis of >1100biologic pathways in the microarray data across the six-pairedcomparisons (these functional gene group lists are availableas supplementary data). Groups that showed the most consistentalterations were ranked by their mean z score and subjectedto hierarchical cluster analysis (Figure 5). This analysis revealedthat very few functional gene groups showed consistent alterations.However, among those that did show large effects of genotype,nearly one third had primary or secondary roles in regulatingcalcium metabolism or transport. For example, the inositol-polyphosphate-5-phosphatasegene group contains genes that are involved directly in theintracellular transport of calcium, and this group showed veryconsistent increases in expression. Other gene groups with knownroles in calcium regulation and with increased gene expressionincluded the cholesterol metabolism group, the actin filamentorganization and striated muscle contraction groups, and thephospholipase C activation and voltage-gated calcium channelactivity groups. Gene groups with known roles in calcium regulationthat showed decreased gene expression included the phospholipaseD group, the long-chain fatty acid transport and steroid metabolismgroups, and the antiporter activity groups, among others. Collectively,these observations suggest a number of ways in which the HC1QTL region that has been established in the GHS rats could contributeto the hypercalciuric phenotype.
Our pathway analysis also revealed a number of biologic genegroups that were affected prominently by the intake of a high-calciumdiet compared with low-calcium diet (Figure 5B). Some of thesegroups were related to the gene groups that differed most accordingto genotype (Figure 5A) and represent well-described targetsof intracellular calcium, including the ligand-dependent nuclearreceptor activity group, the cytoskeleton group, the inositoltriphosphate kinase activity group, the G1 group, and the bonemorphogenic protein signaling pathway group, among others.
To determine whether there was a true chromosome 1q bias inthe distribution of genes with altered expression, we performeda genome-wide examination of the top 100 changed genes accordingto their chromosomal position and compared this distributionwith that of all of the genes on the RAE230 GeneChip (Figure 6A).This analysis revealed that chromosome 1 had approximately twiceas many of the top 100 significantly changed genes as wouldbe expected by chance (Figure 6A). This differential bias wassignificantly different according to 2 analysis (P = 0.0006)only when chromosome 1 was included and disappeared entirelyin the absence of chromosome 1. We next analyzed the localizationof the top 100 changed genes within chromosome 1 (n = 28), using40-megabase binning of the data. Once again, we observed a significantbias in the distribution of these genes (P = 0.000001), witha clear enrichment in the 160- to 200-megabase region (Figure 6B).This difference was eliminated when the 160- to 200-megabasebin was eliminated.
Array Validation: Real-Time qRT-PCR
We chose to validate four genes of interest from the list ofgenes with significant changes in expression according to genotypeacross the three tissue types. For these experiments, only RNAfrom the four high-calcium samples of the three tissue types(kidney, osteoclast, and small intestine) for each genotypeand one low-calcium tissue type (kidney) for each genotype wasused. The four genes that we chose to examine revealed a modestbut significant correlation in the differences reported by thearray and real-time quantitative PCR assays (R = 0.58, P <0.02, 14 df; Table 3). For some genes (e.g., RGS10), the contributionto this correlation was strong, compared with other genes (e.g.,Hbb). Nonetheless, the data from the two techniques did notindicate any false discoveries in this sampling of genes andtissue sets (Figure 7). In addition to the overall correlationin the changes reported with the two techniques, the significanteffect for three of the four genes that we chose to validate(Ptpre, NADsynth1, and RGS10) was confirmed. The replicatedchange observed for Hbb, however, was only marginally significant(P = 0.06).
Figure 7. Relationship of log2 differences by array and quantitative reverse transcription–-PCR. Note that none of the points in this plot is found in the false-positive quadrants (top left, bottom right).
In addition to validating the changes in expression, we comparedthe levels of expression for the WKY and GHS sample sets usingthe absolute levels of expression determined from the arrayand PCR data. Two genes (Ptpre and NADsynth1) had significantcorrelations (R = 0.61 and 0.74, respectively; P < 0.02)in the absolute levels reported across the eight sample sets,whereas the other two genes did not. The most likely explanationfor this is that the Hbb and RGS10 array probes and PCR ampliconsdiffered in their overall GC content and labeling efficiency(with biotin or SYBRGreen).
In contrast to the four selected genes of interest, we observedno difference in the levels of expression of a very commonlyused reference gene, the 18S ribosomal RNA transcript (overallfold change 1.25 in GHS versus WKY; P = 0.70). Therefore, overall,we find fairly good correspondence between the two approachesand believe that the majority of the differences that we havereported in the array data also will be independently verifiable,using real-time PCR.
The goal of this study was to isolate the HC1 QTL region ofthe GHS rat on a normocalciuric WKY background to assess thecontribution of the QTL to hypercalciuria. We previously estimatedthat HC1 contributes 7% of the variation in calcium excretionin the GHS rat, in the context of several other suggestive QTLon other chromosomes, likely to be contributing to the hypercalciuria(20). We now have used a high-calcium diet to push successfullythe phenotype of the congenic rats to enhance the calcium excretiondifference between the congenic and control strains. We demonstrateunequivocally that the HC1 QTL contributes significantly tothe hypercalciuric phenotype of our WKY.GHS(RN01) congenic strain.We see a 1.5-fold increase in calcium excretion in the femalecongenic rats and a nearly 2.5-fold increase in the male congenicrats on a high-calcium diet, compared with the control WKY strains.Although we still observe parallel differences in calcium excretionbetween the congenic and control rats on a low-calcium diet,the calcium excretion levels are much lower and the differencesare much smaller. The larger difference in calcium excretionfor the rats on the high-calcium diet is consistent with thephysiology of the GHS rat in that the major component of hypercalciuriain the model results from increased dietary calcium absorption,with observed overexpression of the vitamin D receptor in theserats (11–13). The persistence of calcium excretion differencesfor the rats on a low-calcium diet also is consistent with theGHS rat, in which two additional contributors to the hypercalciuricphenotype are increased bone resorption and the failure to reabsorbsufficient filtered calcium (7–9,28). The higher levelsof calcium excretion in female compared with male rats is consistentwith what was observed with the parental GHS and WKY strains,in which the GHS female rats excreted 1.4 times as much calciumas the male rats and in which the WKY female rats excreted threetimes as much calcium as the male rats on normal-calcium diets(0.6% calcium) (20). This is the first isolation of a calciumexcretion QTL in any congenic animal.
Our microarray gene expression analyses demonstrated that roughlyone third of the individual genes with the greatest mean foldchange in expression, comparing the congenic rats with the controls,are encoded on chromosome 1. These genes show a clear bias inlocalization to the region of the chromosome where HC1 maps.Many of these are uncharacterized expressed sequence tags thatmay prove to have a role in calcium metabolism or transport,whereas a few of these genes are transcription factors thatmay regulate the expression of genes that are involved in calciumpathways. The comprehensive pathway analysis of the gene expressiondata revealed that relatively few of the >1100 pathways thatwere examined in the array data showed consistent changes inany group of genes between the control and congenic rats. Someof the gene groups, however, did show increased or decreasedexpression for all the tissues examined: small intestine, kidney,and femoral OC. Approximately one third of the consistentlyaffected groups clearly are involved in either primary or secondarycalcium metabolism.
Using real-time qRT-PCR, selected changes in gene expressionaccording to genotype that were identified by the microarraystudy (Ptpre, NADsynth1, and RGS10; P < 0.05) were confirmed,with another gene showing a strong trend for validation (Hbb;P = 0.06). In contrast, a widely used reference gene (18S RNA)showed no change between the various samples according to genotype.Moreover, the relative expression values that were determinedby real-time PCR showed a significant correlation with thosethat were obtained by the microarray study. Together, thesefacts strengthen our confidence that most of the single-genefindings that we have identified as changing according to genotypeor diet (or the interaction of the two) also will be independentlyverifiable.
The bias in the individual gene expression data for localizationto the HC1 region perhaps is not surprising given that the congenicrats that we analyzed differed from the WKY rats only in chromosome1. However, combined with the bias in differential gene groupexpression for gene groups involved in calcium metabolism ortransport, these observations strongly support the utility ofour approach. The introgression of the HC1 region of the GHSrat onto the WKY rat background to generate our WKY.GHS(RN01)congenic rats transfers the hypercalciuric phenotype to theWKY strain, significantly altering the gene expression profileof this region relative to the WKY strain. The altered geneexpression demonstrates a strong bias in the alteration of genegroups involved in the direct or indirect metabolism or transportof calcium. The HC1 QTL region contains at least one gene thatcontributes either directly or indirectly to hypercalciuriain the GHS rat model. If that gene exerts a direct effect, thenit is likely to be at the level of calcium metabolism or transport,and if its effect is indirect, then it may be a transcriptionfactor or protein processing factor that affects the level ofgene expression of proteins that are involved in calcium homeostasis.A number of potential candidate genes in the region are uncharacterizedexpressed sequence tags and would require further work shouldany become strong candidates for the QTL.
The preservation of a significant phenotype in the congenicrats is a critical step toward the identification of a geneor genes in the HC1 QTL region that contribute to hypercalciuria.The microarray data for the HC1 congenic rats clearly demonstratethe perturbation of calcium metabolism and transport pathwaysin these rats. The next step in our analysis of the GHS congenicrats will be the fine structure mapping of HC1, through theconstruction and characterization of subcongenic strains. Wethereby will localize more closely the HC1 QTL and reduce thenumber of candidate genes to be tested. This approach is beingused to study many rat models for polygenic human diseases,including insulin-dependent diabetes (29), hypertension (30,31),alcohol withdrawal (32), liver fibrosis (33), radiation sensitivity(34), arthritis (35), and obesity (36). Genes underlying morethan eight QTL in mouse and rat models have been identified(37), including a gene underlying a type 1 diabetes QTL in theKDP rat (38), a gene for BP (39), and a gene that modifies liverfibrosis in mice and humans (33). Gene expression profilingis being integrated with QTL mapping as one approach to acceleratingthe identification of genes underlying QTL (40–43). Thisintegrated approach has been used to identify several genesfor QTL, including the Cd36 gene for defective fatty acid metabolism(44), the Alox15 gene for bone mineral density (45), and thegene encoding complement factor 5 for allergic asthma (46).
Acknowledgments
This study was supported by National Institutes of Health grantsDK57716 and DK56788 and by a grant from the Hendricks Foundation.
We thank Dr. Wesley Beamer and Dr. Michael Lane for helpfuldiscussions and Jason A. Horton and Karen Gentile for experttechnical assistance.
Footnotes
Published online ahead of print. Publication date availableat www.jasn.org.
Bushinsky DA: Genetic hypercalciuric stone forming rats.
Semin Nephrol 16
: 448
–457, 1996[Medline]
Bushinsky DA, Favus MJ: Mechanism of hypercalciuria in genetic hypercalciuric rats: Inherited defect in intestinal calcium transport.
J Clin Invest 82
: 1585
–1591, 1988[Medline]
Bushinsky DA, Asplin JR: Thiazides reduce brushite, but not calcium oxalate, supersaturation, and stone formation in genetic hypercalciuric stone-forming rats.
J Am Soc Nephrol 16
: 417
–424, 2005[Abstract/Free Full Text]
Bushinsky DA, Grynpas MD, Nilsson EL, Nakagawa Y, Coe FL: Stone formation in genetic hypercalciuric rats.
Kidney Int 48
: 1705
–1713, 1995[Medline]
Bushinsky DA, Asplin JR, Grynpas MD, Evan AP, Parker WR, Alexander KM, Coe FL: Calcium oxalate stone formation in genetic hypercalciuric stone-forming rats.
Kidney Int 61
: 975
–987, 2002[CrossRef][Medline]
Evan AP, Bledsoe SB, Smith SB, Bushinsky DA: Calcium oxalate crystal localization and osteopontin immunostaining in genetic hypercalciuric stone-forming rats.
Kidney Int 65
: 154
–161, 2004[CrossRef][Medline]
Kim M, Sessler NE, Tembe V, Favus MJ, Bushinsky DA: Response of genetic hypercalciuric rats to a low calcium diet.
Kidney Int 43
: 189
–196, 1993[Medline]
Tsuruoka S, Bushinsky DA, Schwartz GJ: Defective renal calcium reabsorption in genetic hypercalciuric rats.
Kidney Int 51
: 1540
–1547, 1997[Medline]
Krieger NS, Stathopoulos VM, Bushinsky DA: Increased sensitivity to 1,25(OH)2D3 in bone from genetic hypercalciuric rats.
Am J Physiol 271
: C130
–C135, 1996
Li XQ, Tembe V, Horwitz GM, Bushinsky DA, Favus MJ: Increased intestinal vitamin D receptor in genetic hypercalciuric rats: A cause of intestinal calcium hyperabsorption.
J Clin Invest 91
: 661
–667, 1993[Medline]
Yao J, Kathpalia P, Bushinsky DA, Favus MJ: Hyperresponsiveness of vitamin D receptor gene expression to 1,25-dihydroxyvitamin D3. A new characteristic of genetic hypercalciuric stone-forming rats.
J Clin Invest 101
: 2223
–2232, 1998[Medline]
Yao JJ, Bai S, Karnauskas AJ, Bushinsky DA, Favus MJ: Regulation of renal calcium receptor gene expression by 1,25-dihydroxyvitamin D3 in genetic hypercalciuric stone-forming rats.
J Am Soc Nephrol 16
: 1300
–1308, 2005[Abstract/Free Full Text]
Karnauskas AJ, van Leeuwen JP, van den Bemd GJ, Kathpalia PP, DeLuca HF, Bushinsky DA, Favus MJ: Mechanism and function of high vitamin D receptor levels in genetic hypercalciuric stone-forming rats.
J Bone Miner Res 20
: 447
–454, 2005[CrossRef][Medline]
Scott P, Ouimet D, Valiquette L, Guay G, Proulx Y, Trouve M-L, Gagnon B, Bonnardeaux A: Suggestive evidence for a susceptibility gene near the vitamin D receptor locus in idiopathic calcium stone formation.
J Am Soc Nephrol 10
: 1007
–1013, 1999[Abstract/Free Full Text]
Reed BY, Gitomer WL, Heller HJ, Hsu MC, Lemke M, Padalino P, Pak CY: Identification and characterization of a gene with base substitutions associated with the absorptive hypercalciuria phenotype and low spinal bone density.
J Clin Endocrinol Metab 87
: 1476
–1485, 2002[Abstract/Free Full Text]
Levy FL, Adams-Huet B, Pak CYC: Ambulatory evaluation of nephrolithiasis: An update of a 1980 protocol.
Am J Med 98
: 50
–59, 1995[CrossRef][Medline]
Pietschmann F, Breslau NA, Pak CY: Reduced vertebral bone density in hypercalciuric nephrolithiasis.
J Bone Miner Res 7
: 1383
–1388, 1992[Medline]
Favus MJ, Karnauskas AJ, Parks JH, Coe FL: Peripheral blood monocyte vitamin D receptor levels are elevated in patients with idiopathic hypercalciuria.
J Clin Endocrinol Metab 89
: 4937
–4943, 2004[Abstract/Free Full Text]
Hoopes RR Jr, Reid R, Sen S, Szpirer C, Dixon P, Pannett AA, Thakker RV, Bushinsky DA, Scheinman SJ: Quantitative trait loci for hypercalciuria in a rat model of kidney stone disease.
J Am Soc Nephrol 14
: 1844
–1850, 2003[Abstract/Free Full Text]
Weil MM, Brown BW, Serachitopol DM: Genotype selection to rapidly breed congenic strains.
Genetics 146
: 1061
–1069, 1997[Abstract]
Markel P, Shu P, Ebeling C, Carlson GA, Nagle DL, Smutko JS, Moore KJ: Theoretical and empirical issues for marker-assisted breeding of congenic mouse strains.
Nat Genet 17
: 280
–284, 1997[CrossRef][Medline]
Hoopes RR Jr, Raja KM, Koich A, Hueber P, Reid R, Knohl SJ, Scheinman SJ: Evidence for genetic heterogeneity in Dent’s disease.
Kidney Int 65
: 1615
–1620, 2004[CrossRef][Medline]
Bushinsky DA, Kim M, Sessler N, Nakagawa Y, Coe F: Increased urinary saturation and kidney calcium content in genetic hypercalciuric rats.
Kidney Int 45
: 58
–65, 1994[Medline]
David JP, Neff L, Chen Y, Rincon M, Horne WC, Baron R: A new method to isolate large numbers of rabbit osteoclasts and osteoclast-like cells: Application to the characterization of serum response element binding proteins during osteoclast differentiation.
J Bone Miner Res 13
: 1730
–1738, 1998[CrossRef][Medline]
Irizarry RA, Hobbs B, Collin F, Beazer-Barclay YD, Antonellis KJ, Scherf U, Speed TP: Exploration, normalization, and summaries of high density oligonucleotide array probe level data.
Biostatistics 4
: 249
–264, 2003[Abstract]
Middleton FA, Ramos EJ, Xu Y, Diab H, Zhao X, Das UN, Meguid M: Application of genomic technologies: DNA microarrays and metabolic profiling of obesity in the hypothalamus and in subcutaneous fat.
Nutrition 20
: 14
–25, 2004[Medline]
Bushinsky DA, Neumann KJ, Asplin J, Krieger NS: Alendronate decreases urine calcium and supersaturation in genetic hypercalciuric rats.
Kidney Int 55
: 234
–243, 1999[CrossRef][Medline]
Kose H, Moralejo DH, Ogino T, Mizuno A, Yamada T, Matsumoto K: Examination of OLETF-derived non-insulin-dependent diabetes mellitus QTL by construction of a series of congenic rats.
Mamm Genome 13
: 558
–562, 2002[CrossRef][Medline]
Rapp JP, Garrett MR, Dene H, Meng H, Hoebee B, Lathrop GM: Linkage analysis and construction of a congenic strain for a blood pressure QTL on rat chromosome 9.
Genomics 51
: 191
–196, 1998[CrossRef][Medline]
Garrett MR, Meng H, Rapp JP, Joe B: Locating a blood pressure quantitative trait locus within 117 kb on the rat genome: Substitution mapping and renal expression analysis.
Hypertension 45
: 451
–459, 2005[Abstract/Free Full Text]
Fehr C, Rademacher BL, Buck KJ: Evaluation of the glutamate decarboxylase genes Gad1 and Gad2 as candidate genes for acute ethanol withdrawal severity in mice.
Prog Neuropsychopharmacol Biol Psychiatry 27
: 467
–472, 2003[Medline]
Hillebrandt S, Wasmuth HE, Weiskirchen R, Hellerbrand C, Keppeler H, Werth A, Schirin-Sokhan R, Wilkens G, Geier A, Lorenzen J, Kohl J, Gressner AM, Matern S, Lammert F: Complement factor 5 is a quantitative trait gene that modifies liver fibrogenesis in mice and humans.
Nat Genet 37
: 835
–843, 2005[CrossRef][Medline]
Tsuji AB, Sugyo A, Ogiu T, Sagara M, Kimura T, Ishikawa A, Sudo H, Ohtsuki M, Aburatani H, Imai T, Harada YN: Fine mapping of radiation susceptibility and gene expression analysis of LEC congenic rat lines.
Genomics 86
: 271
–279, 2005[Medline]
Joe B, Cannon GW, Griffiths MM, Dobbins DE, Gulko PS, Wilder RL, Remmers EF: Evaluation of quantitative trait loci regulating severity of mycobacterial adjuvant-induced arthritis in monocongenic and polycongenic rats: Identification of a new regulatory locus on rat chromosome 10 and evidence of overlap with rheumatoid arthritis susceptibility loci.
Arthritis Rheum 46
: 1075
–1085, 2002[CrossRef][Medline]
Stylianou IM, Clinton M, Keightley PD, Pritchard C, Tymowska-Lalanne Z, Bunger L, Horvat S: Microarray gene expression analysis of the Fob3b obesity QTL identifies positional candidate gene Sqle and perturbed cholesterol and glycolysis pathways.
Physiol Genomics 20
: 224
–232, 2005[Abstract/Free Full Text]
Glazier AM, Nadeau JH, Aitman TJ: Finding genes that underlie complex traits.
Science 298
: 2345
–2349, 2002[Abstract/Free Full Text]
Yokoi N, Komeda K, Wang HY, Yano H, Kitada K, Saitoh Y, Seino Y, Yasuda K, Serikawa T, Seino S: Cblb is a major susceptibility gene for rat type 1 diabetes mellitus.
Nat Genet 31
: 391
–394, 2002[Medline]
Cicila GT, Garrett MR, Lee SJ, Liu J, Dene H, Rapp JP: High-resolution mapping of the blood pressure QTL on chromosome 7 using Dahl rat congenic strains.
Genomics 72
: 51
–60, 2001[CrossRef][Medline]
Flint J, Valdar W, Shifman S, Mott R: Strategies for mapping and cloning quantitative trait genes in rodents.
Nat Rev Genet 6
: 271
–286, 2005[CrossRef][Medline]
Schadt EE, Lamb J, Yang X, Zhu J, Edwards S, Guhathakurta D, Sieberts SK, Monks S, Reitman M, Zhang C, Lum PY, Leonardson A, Thieringer R, Metzger JM, Yang L, Castle J, Zhu H, Kash SF, Drake TA, Sachs A, Lusis AJ: An integrative genomics approach to infer causal associations between gene expression and disease.
Nat Genet 37
: 710
–717, 2005[CrossRef][Medline]
Chesler EJ, Lu L, Shou S, Qu Y, Gu J, Wang J, Hsu HC, Mountz JD, Baldwin NE, Langston MA, Threadgill DW, Manly KF, Williams RW: Complex trait analysis of gene expression uncovers polygenic and pleiotropic networks that modulate nervous system function.
Nat Genet 37
: 233
–242, 2005[CrossRef][Medline]
Hubner N, Wallace CA, Zimdahl H, Petretto E, Schulz H, Maciver F, Mueller M, Hummel O, Monti J, Zidek V, Musilova A, Kren V, Causton H, Game L, Born G, Schmidt S, Muller A, Cook SA, Kurtz TW, Whittaker J, Pravenec M, Aitman TJ: Integrated transcriptional profiling and linkage analysis for identification of genes underlying disease.
Nat Genet 37
: 243
–253, 2005[CrossRef][Medline]
Glazier AM, Scott J, Aitman TJ: Molecular basis of the Cd36 chromosomal deletion underlying SHR defects in insulin action and fatty acid metabolism.
Mamm Genome 13
: 108
–113, 2002[CrossRef][Medline]
Klein RF, Allard J, Avnur Z, Nikolcheva T, Rotstein D, Carlos AS, Shea M, Waters RV, Belknap JK, Peltz G, Orwoll ES: Regulation of bone mass in mice by the lipoxygenase gene Alox15.
Science 303
: 229
–232, 2004[Abstract/Free Full Text]
Karp CL, Grupe A, Schadt E, Ewart SL, Keane-Moore M, Cuomo PJ, Kohl J, Wahl L, Kuperman D, Germer S, Aud D, Peltz G, Wills-Karp M: Identification of complement factor 5 as a susceptibility locus for experimental allergic asthma.
Nat Immunol 1
: 221
–226, 2000[CrossRef][Medline]
Top
Abstract
Introduction
2 statistics to compare the distribution of all genes represented on the RAE230 array with the distribution of the top 100 changed genes. This analysis indicated a clear bias toward transcripts on chromosome 1 and particularly in the 160- to 200-megabase region on 1q. These calculations were performed without correction for multiple testing. The top changed genes (in terms of the significance of the effect for either diet or genotype) were used in a heat map cluster analysis. 
hemoglobin [Hbb] and protein tyrosine phosphatase, receptor type, epsilon polypeptide [Ptpre]), and two genes with decreased expression in the congenic tissues (NAD synthetase 1 [NADsynth1] and regulator of G protein signaling 10 [RGS10]). We also included a reference gene to control for levels of starting material in each sample (18S RNA). Primers for each of these genes were designed using Primer3 software to produce amplicons of 80 to 130 bp. Each primer pair was tested on a mixed cDNA template to ensure amplification of a single specific band, using HotMaster Taq(Eppendorf) and standard cycling conditions (hot start at 94°C for 2 min followed by 40 cycles of melt at 94°C for 30 s and annealing and extension at 59°C for 60 s). After the specificity of each of the amplicons was confirmed, real-time qRT-PCR was performed using four replicates from each sample set for each gene (32 PCR reactions total per gene) with 12 ng of cDNA as template. Reactions were performed in a 25-µl volume using the same enzyme and cycling parameters as before (HotMasterTaq and buffer) except that 0.25 µl of SYBRGreenI dye (Molecular Probes, Eugene, OR) was added to each well before cycling. These reactions were run on 96-well plates using an ABI 7000 Real-Time Sequence Detection instrument (Applied Biosystems). End point melt-curve analysis confirmed the presence of single amplicons in each reaction well, and amplification in the absence of template failed to produce any signal as a result of primer dimerization and extension. Statistical analysis of the real-time data was performed using a one-tailed pairwise t test to compare the difference in the mean number of cycles to threshold 
6). In this study, we first focus on describing the genes that were most affected and then discuss the biologic pathways that were most altered. Finally, we present the results of a formal analysis of localization biases in the data set, as we attempt to define attractive candidate genes for further study. 
