A protocol for high-throughput phenotyping, suitable for ... - Springer Link

A protocol for high-throughput phenotyping, suitable for quantitative trait analysis in mice Leah C. Solberg,1 William Valdar,1 Dominique Gauguier,1 Graciela Nunez,1 Amy Taylor,1 Stephanie Burnett,1 Carmen Arboledas-Hita,1 Polinka Hernandez-Pliego,1 Stuart Davidson,1 Peter Burns,1 Shoumo Bhattacharya,1 Tertius Hough,2 Douglas Higgs,3 Paul Klenerman,4 William O. Cookson,1 Youming Zhang,1 Robert M. Deacon,5 J. Nicholas P. Rawlins,5 Richard Mott,1 Jonathan Flint1 1

Wellcome Trust Centre for Human Genetics, University of Oxford, Oxford Roosevelt Drive, Oxford OX3 7BN, UK Mammalian Genetics Unit and UK Mouse Genome Centre, Medical Research Council, Harwell OX11 ORD, UK 3 Medical Research Council Molecular Haematology Unit, Weatherall Institute of Molecular Medicine, John Radcliffe Hospital, Headington, Oxford OX3 9DS, UK 4 Nuffield Department of Medicine, University of Oxford, Peter Medawar Building for Pathogen Research, Oxford OX1 3SY, UK 5 Department of Experimental Psychology, University of Oxford, Oxford, OX1 3UD, UK 2

Received: 26 August 2005 / Accepted: 25 October 2005

Abstract

Introduction

Whole-genome genetic association studies in outbred mouse populations represent a novel approach to identifying the molecular basis of naturally occurring genetic variants, the major source of quantitative variation between inbred strains of mice. Measuring multiple phenotypes in parallel on each mouse would make the approach cost effective, but protocols for phenotyping on a large enough scale have not been developed. In this article we describe the development and deployment of a protocol to collect measures on three models of human disease (anxiety, type II diabetes, and asthma) as well as measures of mouse blood biochemistry, immunology, and hematology. We report that the protocol delivers highly significant differences among the eight inbred strains (A/J, AKR/J, BALBc/J, CBA/J, C3H/HeJ, C57BL/6J, DBA/2J, and LP/J), the progenitors of a genetically heterogeneous stock (HS) of mice. We report the successful collection of multiple phenotypes from 2000 outbred HS animals. The phenotypes measured in the protocol form the basis of a large-scale investigation into the genetic basis of complex traits in mice designed to examine interactions between genes and between genes and environment, as well as the main effects of genetic variants on phenotypes.

Systematic attempts to determine gene function in the mouse have hitherto relied on the production of novel, highly penetrant mutations that can either be targeted to a gene of interest or introduced randomly in the genome. While the use of mutagenesis has proved to be a powerful technique to determine the genetic basis of abnormal phenotypes (Brown and Hardisty 2003), many aspects of gene function become apparent only from the analysis of lowpenetrance mutations that are very hard to detect in a standard mutagenesis screen. In some cases fully penetrant effects are lethal so that the phenotypic consequences of a mutation can be obtained only by investigating more subtle alterations of gene structure; in other cases, discovering and exploring the often unexpected pleiotropic action of a gene requires the analysis of low-penetrance mutations (Greenspan 2004). Consequently, there is a need to develop screens for quantitative phenotypes. Naturally occurring genetic variants, the source of quantitative genetic variation, frequently have small phenotypic effects and would be a useful source of low-penetrance mutations were it not for the difficulty they present for mapping and cloning strategies. Developments in a number of areas are now making it feasible to uncover the molecular basis of quantitative variation in the mouse (Flint et al. 2005). One option is to carry out whole-genome genetic association studies in outbred mouse populations, using novel analytical approaches such as

Correspondence to: Jonathan Flint; E-mail: [email protected]

DOI: 10.1007/s00335-005-0112-1 Volume 17, 129 146 (2006) Springer Science+Business Media, Inc. 2006

129

130

ancestral probabilistic haplotype reconstruction (Mott et al. 2000). In this way, quantitative trait loci (QTLs) can be mapped in a single step to regions of less than 1 cM in heterogeneous stock (HS) mice, an outbred population derived from eight inbred strains such that the chromosomes of each mouse contain a genetic mosaic of the founding inbred animals (Flint et al. 2005). One major obstacle to the use of HS is the high cost of genotyping. Measuring many phenotypes in parallel on each mouse reduces the cost of genotyping relative to the information gained per phenotype to the point where it is economically attractive to genotype a cohort of outbred animals at high resolution and phenotype them for as many traits as possible. However, obtaining multiple phenotypes from each animal raises problems of interaction between phenotypes (particularly for behavioral assays) and potentially limits the amount of information that can be obtained. A similar problem confronts phenotypic analysis of mutagenized mice, where a protocol for comprehensive phenotypic collection (SHIRPA) has been developed (Rogers et al. 1997). SHIRPA and similar protocols are designed to detect abnormal phenotypes and therefore tend to favor the detection of outliers, since these are more likely to be a result of highly penetrant heritable mutations. These protocols are not suitable for QTL mapping experiments, which use information from all animals and where, in fact, an outlier is likely to be an unreliable measurement that should be discarded rather than a valuable data point as is the case with mutagenesis screens. In this article we describe the development of a protocol to collect multiple phenotypes from any population of mice. Because we have previously shown that high-resolution mapping of QTL is possible in HS mice (Mott et al. 2000; Talbot et al. 1999), we decided to apply the protocol to a large population of HS mice and to the progenitors of the HS [eight inbred strains: A/J, AKR/J, BALBc/J, CBA/J, C3H/HeJ, C57BL/6J, DBA/2J, and LP/J (Demarest et al. 2001)] to ensure that our phenotypic assays could detect significant variation between inbred strains. The QTL protocol was designed to yield information on phenotypes of importance to human health. We targeted three diseases: anxiety, type II diabetes, and asthma. In addition, we collected hematology and immunology profiles of each mouse. The phenotypes measured in the protocol form the basis of a large-scale investigation into the genetic basis of complex traits in mice designed to examine interactions between genes and between genes and environment, as well as the main effects of genetic variants on phenotypes. This article describes the

L.C. SOLBERG

ET AL.:

HIGH-THROUGHPUT PHENOTYPING

phenotyping protocol in detail and presents a statistical analysis of the phenotypes measured on the inbred strains. Results from mapping QTLs in the HS will be presented elsewhere. Materials and methods Experimental design. The protocol starts when mice are five weeks old, with a blood sample for immunology and the implantation of a microchip. Using only a local anesthetic, blood is drawn from the tail. At six weeks of age, behavioral tests are carried out, i.e., ethologic assays preceding tasks involving fear conditioning. At the end of the last of the unconditioned tasks (home-cage activity), the mice are given the first of two weekly intraperitoneal injections of ovalbumin, so that by week eight they have been fully sensitized for studies of respiratory physiology using a whole-body plethysmograph. This means that animals are immunized during the week in which fear conditioning is assessed. Table 1 gives the order of tests and the age at which they were carried out. Animals. Original Northport HS mice were obtained from Dr. Robert Hitzemann at the Oregon Health Sciences Unit (Portland, OR). At the time the animals arrived they had passed 40 generations of pseudorandom breeding (Demarest et al. 2001). A breeding colony was established at Oxford University to generate animals for phenotyping. Inbred strains were obtained from Harlan (United Kingdom) or the Jackson Laboratory (Bar Harbor, ME) at five weeks of age. For each of the eight inbred strains (A/J, AKR/J, BALB/cJ, C3H/HeJ, C57BL/6J, CBA/J, DBA/2J, and LP/J) we obtained six male and six female mice. Animals were housed at a maximum of six mice per cage (average of 4) and placed on a 12:12 light:dark cycle with food and water available ad libitum unless otherwise specified. The pH of the drinking water was not monitored. Mice were left undisturbed for two weeks before testing. Animals were housed in the same hallway as the experimental rooms to minimize the disruption of transporting animals. Unless otherwise specified, mice were brought to the experimental room a minimum of 5 min before testing. At five weeks of age, all HS animals were weighed and implanted with a microchip for identification. A 50-ll blood sample was taken from the tail vein for immunology (this occurs before allergen sensitization), and a 2mm hole was made in the center of the cartilaginous part of both ears using a metal ear punch (Fisher Scientific, Catalog No 01-337B) (see Table 1 for the order of testing).

L.C. SOLBERG

ET AL.:

131


Table 1. Order of behavioral and physiologic testing in HS mice

Age (days)

Test

35 42 42 42 42

Microchip, blood sample for immunology Open field arena: distance in the perimeter, the center, and total distance in 5 min Elevated plus maze:distance traveled, time spent, and entries into close and open arms Food hyponeophagia: time taken to sample a novel foodstuff (overnight food deprivation) Activity measured in the home cage in 30 min; number of pellets removed from burrow in 1.5 h Potentiation of the startle response after fear conditioning Freezing to the context in which a tone is associated with a foot shock Freezing to a tone after association with a foot shock Animals sensitized by injection with ovalbumin inhale metacholine, and changes in lung function are measured by plethysmography (a model of asthma) Glucose and insulin values taken at 0, 15, 30, and 75 min after intraperitoneal glucose injection (a model of diabetes) Tissue harvest

49 49 49 56 63 65

Open field test. As an unconditioned test of anxiety we used the open field (OF), a brightly lit white circular arena 60 cm in diameter. This was arbitrarily divided into an inner (20 cm in diameter) and an outer circular area. At the time of testing, the mouse was removed from a cage and placed at the periphery of the OF. Mouse movements were monitored every second for 5 min by a video camera mounted on top of the OF. A Videotrack (vNT4.0) automated tracking system from Viewpoint (Champagne Au Mont D’Or, France) was used to analyze video data. Total activity (in distance), time spent in the inner area, and latency to enter the inner area were measured. At the end of the test, the number of fecal boli in the OF area was counted. Elevated plus maze. The second unconditioned test of anxiety, the elevated plus maze (EPM), is in the shape of a plus sign, made up of two opposite enclosed arms (30 cm · 6 cm · 21 cm), two opposite open arms (29.5 cm · 6 cm · 0.5 cm), and a central (junction) area (14.5 cm · 14 cm). The apparatus is elevated 73 cm from the floor on a stand. At the beginning of the test, the mouse was placed at the center of the plus maze facing toward the end of a closed wall. Mouse movements were monitored every second for 5 min by a video camera mounted above the EPM and the video was analyzed using the Videotrack system (Viewpoint). We collected the number of entries, time in seconds, and distance traveled in the open arms, closed arms, and junction areas. Food hyponeophagia. The third unconditioned test of anxiety is a measure of latency to eat a novel food [food hyponeophagia (FN)]. The apparatus is a white Perspex base (30 cm2) to which a food well (1.2 cm in diameter, 0.9 cm high) is glued. We used full-cream sweetened condensed milk diluted 50:50

with water as the novel food, placed in the food well. Mice were partly food-deprived overnight (all food was removed except for 1 g of food chow per mouse, placed inside each cage). Five minutes before the start of the test mice were put in individual cages. At the time of the test, the mouse was placed facing away from the food well under a 1.5-L measuring jug made of transparent plastic. The jug was placed over the Perspex base so that the food well was positioned in the spout of the jug. The latency from being placed in the apparatus to the time when the animal started drinking was measured. If a mouse did not drink within 2 min, it was removed from the apparatus, returned to the individual cage, and tested again in 5 min. The mouse was given three trials of 2 min each and was given a score of 360 sec if it did not drink within this time. New home-cage activity. Because the unconditioned tests of anxiety are assessed by changes in motor activity, we included a measure of activity thought to be relatively unaffected by emotional state. Baseline, or new home-cage, activity was measured using a photoactivity system from San Diego Instruments (San Diego, CA). Mice were individually placed in a plastic cage (46 cm · 15 cm · 21 cm) which has seven infrared photobeams that cross the width of the cage floor, are spaced 5.5 cm apart, and are 1.5 cm from the floor. A thin layer of clean wood chips along with the animal’s own bedding was placed in the cage. The number of beam breaks was measured in 5-min bins during a 30-min period. Species-typical behavior: burrowing. Gray plastic cylinders, 20 cm long, 6.8 cm in diameter, and sealed at one end, were filled with 200 g of normal diet food pellets and placed in individual mouse cages. Cages were lined with a thin layer of clean bedding and outfitted with a water bottle. Mice

132

were tested individually for 1.5 h in the afternoon (15:00 16:30). After the test was complete, the weight of food pellets taken out of the tube was measured. Fear-potentiated startle. We used a number of conditioned tests of anxiety, starting with fearpotentiated startle (FPS), which was measured in a startle response system from San Diego Instruments (San Diego, CA). On the first day of the test (day 1), startle response was measured in an accelerometer by ten presentations of a 30-sec (70 dB, 10 kHz) tone coterminating with a 100-dB noise burst (lasting 50 msec) and ten presentations of 50-msec 100-dB noise burst alone (not preceded by a tone). Trial types were presented randomly over 7 min with a fixed intertrial interval of 60 sec. On the following day (day 2), animals were trained to associate the tone with foot shock in training chambers. The training session consisted of ten 29.5-sec presentations of the tone (70 dB, 10 kHz, as before) that coterminated with a 500-msec, 0.3-mA foot shock. The pseudorandom intertrial interval ranged from 2 to 4 min and the entire training session lasted 47 min. White noise (60 dB) was present in the chambers during the entire testing period. Finally, on day 3, animals were again placed in the accelerometer and a startle response was measured using exactly the same protocol as on day 1, i.e., by ten presentations of the tone preceding a noise burst, and ten presentations of noise burst alone. Fear-potentiated startle was measured as the difference in startle response between the two trial types (tone + startle noise and startle noise alone) (Heldt et al. 2000). Context freezing. We measured freezing to the context, the context being a training chamber in which the animal had been trained to associate a tone with a mild foot shock. One day after completing FPS, mice underwent a training session in which the 29.5-sec tone (70 dB, 10 kHz, as before) was followed by a 0.5-sec, 0.3-mA shock. Three shock trials were randomly administered over a 12min period. Four hours later mice were placed into the shock chamber for 5 min and amount of time freezing was measured using Videotrack (Viewpoint) (no shock was administered during the 5-min test). Cue conditioning. We measured variation in freezing to a tone previously with a mild foot shock. The day after the assessment of context freezing, mice were placed in an OF that was lit with a dim red light instead of a bright white light. The 70-dB, 10-kHz tone was administered for 30 sec between

L.C. SOLBERG

ET AL.:


1:30 2:00 and 3:30 4:00 min into the 5-min test. Time spent freezing was measured using Videotrack (Viewpoint). Plethysmography. As a model of asthma, we measured airway resistance in unrestrained conscious mice after sensitization and airway challenge with ovalbumin. At 6 and 7 weeks of age, mice were immunized with an intraperitoneal injection of 100 lg ovalbumin (Sigma-Aldrich, Dorset, UK) once a week. At 8 weeks of age, over three consecutive days, mice were placed in a nebulizing chamber and exposed to aerosolized ovalbumin for 5 min. Twenty-four hours after the last ovalbumin challenge, unrestrained conscious mice were placed into a plethysmograph chamber (Buxco Europe Ltd., Hampshire, UK). Airway resistance was measured as Penh (enhanced pause). Figure 1 shows the components measured by the plethysmograph in the respiratory cycle. Pause is defined as (Te ) Ti)/Tr and Penh is calculated as pause multiplied by Pef/Pif, where Te is expiratory time, Ti is inspiratory time, Tr is relaxation time, Pef is expiratory flow, and Pif is peak inspiratory flow (Eum et al. 1995; Hamelmann et al. 1997). Baseline airway resistance was measured for 5 min, after which time an aerosol of 1.5 · 10)2 M methacholine (Sigma-Aldrich, Dorset, UK) was given and airway resistance was again measured for 5 min. The change in Penh was calculated between baseline and response values. Intraperitoneal glucose tolerance test (IPGTT). We used the response to an intraperitoneal injection of glucose as a model to assess both glucose tolerance and stimulated insulin secretion in vivo, which are key type 2 diabetes phenotypes. Following an overnight fast, mice were weighed and anesthetized with pentobarbitone (Sagatal, Merial Animal Health, Abingdon, UK). An initial sample of blood was collected. Mice were then injected with a solution of 2 g glucose/kg body weight. Blood samples were collected 15, 30, and 75 min after the glucose injection to evaluate accurately both the acute plasma glucose and insulin responses and the initial phase of glucose disposal by the body. Samples were centrifuged and plasma was collected and stored at )20C for subsequent determination of glucose and insulin concentrations. Plasma glucose concentrations were measured by a glucose oxidase method on a glucose analyzer 2 (Beckman, Fullerton, CA). Plasma immunoreactive insulin (IRI) was measured by a commercial insulin ELISA kit (Mercodia, Uppsala, Sweden) using mouse insulin as a standard. The total area under the curve and the area under

L.C. SOLBERG

ET AL.:


Fig. 1. Measures used in whole-body plethysmography. Ti = inspiratory time (sec), time from start of inspiration to end of inspiration; Te = expiratory time (sec), time from end of inspiration to start of next inspiration; Pif = peak inspiratory flow (ml/sec), maximal negative box pressure occurring in one breath; Pef = peak expiratory flow (ml/ sec), maximal positive box pressure occurring in one breath; respiratory rate; Tr = relaxation time (sec), time of the pressure decay to 36% of total box pressure during expiration. Pause is defined as (Te ) Tr )/Tr and enhanced pause (Penh) is pause multiplied by Pef/Pif.

the curve above baseline were calculated using trapezoidal analysis of plasma glucose (AUC-G and DG, respectively) and insulin (AUC-IRI and DIRI, respectively) values during the IPGTT (Fig. 2). The parameters provide estimates of glucose tolerance (AUC-G and DG) and overall insulin secretion (AUC-IRI, AUC-IRI/AUC-G, DIRI, and DIRDG) in response to glucose. Finally, the slope of the decay curve of plasma glucose during the IPGTT (K parameter), which provides an index of insulin sensitivity, was calculated from regression analysis (Fig. 2). Immunology. Immunophenotyping was performed using standard whole-blood methodology. A 50-ll blood sample was stained for 20 min at 4C, washed, and lysed using Cell Lysis solution (BD). Three monoclonal antibodies (mAbs) were used for antibody staining: (1) fluorescein isothiocynate (FITC) conjugated mAbs specific for CD3, (2) APC conjugated mAbs for CD8, and (3) peridinin chlorophyll A (perCP) for CD4 and phycoerthyrin (PE) conjugated mAbs for B220 (all from BD Pharmingen). Cell suspensions were analyzed with a Becton Dickinson FACS caliber flow cytometer using CellQuest software. Hematology. Blood was collected via heart puncture. Red and white cell indices were measured using a hematology analyzer (Coulter counter) designed for routine processing of human blood samples, allowing the independent measurement of cell volume, hemoglobin content of red blood cells, and platelet and white cell indices.

133

Fig. 2. Calculated measures derived from a glucose tolerance test. The K parameter is the slope of decline of plasma glucose values from the highest point. AUC-G = total area under the curve. DG = under the curve above baseline. Identical measures were taken after measuring blood insulin and the equivalent acronyms are AUC-IRI and DIRI, respectively.

Biochemistry. Blood samples for biochemistry were obtained from cardiac puncture after overnight fasting. Samples in lithium heparin anticoagulant were separated by centrifugation and stored at )70C. We used assays developed for a biochemical screen of ENU-induced mutant mice (Hough et al. 2002). The method uses an automated clinical chemistry analyzer with each analysis optimized for small samples. We analyzed kidney, bone, and liver profiles as previously described (Hough et al. 2002). Tissue harvest and wound healing. At the end of the protocol all animals were killed by injection of phenobarbitone; after taking a blood sample by cardiac puncture, internal organs were removed and snap frozen in liquid nitrogen. Both ears were collected and the holes measured for wound closure using a grid etched reticle (Clark et al. 1998). Data management. To track the progress of each animal through the phenotyping schedule and to ensure that each phenotype was associated with the correct animal and that no phenotypes were lost, we wrote software to upload phenotypes directly into the IGS relational database (Integrated Genotyping System, www.well.ox.ac.uk/bioinformatics/ project_lims.shtml). Each mouse was identified by a microchip that was implanted subcutaneously before the start of the test. By using portable microchip readers that communicated directly with computers running the tests, we were able to ensure that all data collected related to the correct animal. Data from just two tests were collected manually: Food hyponeophagia and the intraperitoneal glucose tolerance test were administered by the experimenter with no automation. Results from these two tests

134

were entered manually into a spreadsheet. At the end of each week, data for each phenotype were checked for inconsistencies and formatted for upload using a suite of PERL scripts that interrogated the relevant computer or spreadsheet. Covariates were collated with the experimental data at the same time, including the time of each test, the experimenter, the cage number of the animal, and, where appropriate, weight of the mouse and an identifier for the measuring instrument used. Statistical analysis. Results were analyzed using the R statistical package (v2.0.0) (www.r-project.org) (R Development Core Team 2004). For the data on the inbred strains, the effect of strain was assessed by a two-factor analysis of variance: gender (two levels) and strain [the number of levels depending on the number of strains in the analysis (either eight or two)]. Latency measures (EPM, FN, OFT) were analyzed as survival data using the R package ‘‘survival.’’ Results We began our study by piloting assays for each phenotype using at least two inbred strains to ensure that we could detect strain differences. Because we assumed that behavioral tests would be the least robust, our choice of strains for the pilot project was guided by prior reports. As most data are available for the two strains C57BL/6J and DBA/ 2J, we used these strains for our pilot. Tests for which we obtained statistically significant evidence of a strain effect were then combined into a continuous battery of phenotyping assays and used to phenotype all eight progenitor strains of the HS. Finally, all tests were administered to a large cohort of HS animals. We did not explore the effects of altering the order of tests. The list of tests that we used and the order in which they were administered are given in Table 1. We first describe the results for each phenotype in the inbred strains and then the results obtained in the HS animals. All results reported here, unless otherwise specified, are for mice that completed the entire battery of tests. The number of mice used in each test is indicated by the degrees of freedom of the relevant F-statistic (in all cases this figure was more than 95% of the 96 animals used). We calculated the power (using the power.anova.test function in R) to detect a strain effect for a quantitative trait using 12 mice of each strain (6 male, 6 female). Analyzing the data ignoring gender (comparing 8 groups), the within-group variance (the residual variance) must be less than 5.4 times the

L.C. SOLBERG

ET AL.:


between-group variance (the variance due to the strain differences) to obtain 80% power to detect an effect at a p value of less than 0.05. Analyzing the data taking into account gender (as 16 groups of 6 animals), the within-group variance must be less than 4.1 times the between-group variance to deliver the same power. Behavioral measures. We were unable to find any significant effect of gender on any of the behavioral measures and consequently report results for a data set of males and females combined. We found very significant strain effects in unconditioned tests of anxiety in the elevated plus maze [open arm entries (F7,96 = 14.1, p < 10)11), time in open arms (F7,96 = 47.3, p < 10)29), close arm entries (F7,96 = 17.5, p < 10)14), time in closed arms (F7,96 = 17.3, p < 10)14)] and open-field arena [total activity (F7,92 = 31.7, p < 10)21), latency to enter centre area (F7,92 = 3.9, p < 0.001), center time (F7,92 = 13.7, p < 10)11)]. There was a similarly large effect of strain on the measure of home-cage activity (30 min of activity in a cage similar to the home cage) (F7,93 = 29.7, p < 10)20) and strain effect on food hyponeophagia (v27 = 95.8, p < 10)17). We also collected one species-typical behavior, burrowing, which has been shown to be affected by hippocampal lesions yet does not reflect emotional reactivity (Deacon and Rawlins 2005). We observed a highly significant effect of strain (F7,96 = 17.1, p < 10)13). Table 2 gives the mean and standard deviations for each strain. We were unable to develop a protocol to detect a robust potentiation of the startle response. Protocols for inducing fear-potentiated startle in rodents typically begin by measuring startle response to a loud noise (100-dB startle stimulus), then pairing an aversive unconditioned stimulus (foot shock) with a light [conditioned stimulus (CS)], and finally looking for an increase in the startle response when the CS is followed by the startle stimulus. For the pilot experiments (using C57BL/6J and DBA/2J strains), we tested the effect of using a visual cue (a light applied for 29.5 sec) but obtained comparable results to those described below (results not shown). We had to take into account that two HS progenitor strains (C3H/HeJ and CBA/J) carry retinal degeneration mutations (Pde6brd1) and would be difficult to train using a light CS. Therefore, we decided to use an auditory tone as a cue, a pure tone of 10 Hz to contrast with the white noise startle (100 dB). We used a 29.5-sec tone (70 dB, 10 kHz) as the CS and compared C57BL/6J and DBA/2J (two strains examined by other investigators). After measuring the startle response to a 100-dB noise preceded by a

139.6 ± 53.1 137.2 ± 54.4 123.6 ± 58.4 Data are expressed as mean ± standard deviation.

45.1 ± 57.1 24.9 ± 29.9 18.25 ± 18.4

33.3 ± 23.6

57.7 ± 36.4

405 ± 130 547 ± 123 467 ± 134 691 ± 145 248 ± 147

408 ± 104

1052 ± 205

76 ± 98 86 ± 111 118 ± 134 15 ± 8 72 ± 98 194 ± 137 280 ± 118

± ± ± ± ± 869 79 148 5 82 132 70 21 10 66 ± ± ± ± ± 1121 111 186 16 47 108 229 14 223 276 ± ± ± ± ± 648 859 82 432 473 126 115 17 215 115 ± ± ± ± ± 1195 283 170 404 282 174 206 36 3 181 ± ± ± ± ± 402 697 69 8 342 281 209 9 22 8 ± ± ± ± ± 534 188 101 24 161 208 67 22 110 94 ± ± ± ± ± 1336 261 146 83 219 ± ± ± ± ± 859 115 182 20 175

339 219 79 36 150

± ± ± ± 1519 3.4 140 2.3 407 4.6 85 1.6 ± ± ± ± 2448 7.6 113 4.6 512 6.1 69 2.2 ± ± ± ± 2337 10.8 106 5.7 888 7.8 61 2.6 ± ± ± ± 4289 14.3 67 2.4 559 7.9 104 1.8 ± ± ± ± 1774 9.8 95 2.5 709 ± 258 0 300 4.9 ± 3.1 820 3.6 78 1.9 ± ± ± ± 743 0.89 85 1.4 ± ± ± ±

Open field arena Total activity Center time (s) Latency (sec) Boli Elevated plus maze Closed arm distance Open arm distance Closed arm time (sec) Open arm time (sec) Open arm latency (sec) Food hyponeophobia FN latency (sec) New home-cage activity Total beam breaks Species-typical behavior Pellets burrowed (g)

1440 0.33 274 2.2

365 ± 230

the 10-kHz tone, six males of each strain were subjected to a fear-conditioning paradigm consisting of ten trials in which the CS and US presentations were randomized (nonpaired group), and six males of each strain were subjected to ten trials in which CS and US were paired (as described in the Methods section). We then repeated the measurement of the startle response paired to the tone. The results, shown in Fig. 3, demonstrate a very large effect of strain on the response to an unpaired stimulus. The results also show a potentiation in the startle response in the paired groups compared with the unpaired groups, but this difference did not reach statistical significance, nor was there a significant effect of strain. On the assumption that the effect might exist but be relatively small, we included FPS in the protocol. We assessed fear conditioning by the amount of time that mice froze when, after a training session in which an aversive unconditioned stimulus (foot shock) was paired with a cue (a tone), they were reexposed to the context in which they were trained, or to the cue, in the absence of shock. In contrast to the nonsignificant results obtained with fear-potentiated startle, we found robust strain differences in the amount of time freezing to the cue (F7,96 = 14.4, p < 10)12) and context (F7,96 =13.9, p < 10)12). Figure 4 shows the proportion of time spent freezing when exposed to the cue for two strains (C57BL/6J and DBA/2J). The means and standard deviation of the measures from each behavioral test are given in Table 3.

353 ± 28

135

271 136 53 8 125


758 2.8 118 1.3

ET AL.:

2126 2.2 254 2.3

C3H/HeJ BALB/cJ AKR/J A/J

Table 2. Unconditioned tests of anxiety for eight inbred strains (12 mice of each strain)

C57BL/6J

CBA/J

DBA/2J

LP/J

L.C. SOLBERG

Glucose homeostasis and insulin secretion. Table 4 shows the increase in blood glucose concentration in the eight strains of mice after an intraperitoneal injection of glucose. All strains showed a significant increase in glycemia in response to the glucose stimulus and there were significant differences between strains, most marked for the measurement taken 15 min after the injection (F7,96 = 18.9, p < 10)15 compared with F7,96 = 9.9, p < 10)8 for fasting blood glucose, F7,96 = 8.7, p < 10)7 at 30 min and F7,96 = 9.9, p < 10)8 at 75 min postinjection). We observed a significant difference between males and females in fasting glucose concentrations (p = 0.02), but not for the other time points. All three calculated variables of glucose regulation showed significant strain effects (AUC-G: F7,96 = 7.1, p < 10)6; DG: F7,96 = 6.8, p < 10)6; K: F7,96 = 10.3, p < 10)8). BALBc/J and DBA2/J mice showed the lowest values of fasting glycemia and A/J and AKR/J mice showed the highest values of this variable (Table 4). The BALBc/J strain exhib-

136

L.C. SOLBERG

ET AL.:


Fig. 3. Potentiation of the startle response in two strains (C57BL/6J and DBA/2J). Mean amplitude startle response, before training and after training, for animals where the cue (tone) was either paired or unpaired with foot shock in the training session.

Fig. 4. Conditioned freezing to a tone in two strains (C57BL/6J and DBA/2J). The horizontal axis is time (min) and the vertical axis is the degree of freezing. The black carets (^) are markers placed every 10 sec; carets at the top of the graph show the time at which the two tones were delivered.

124.1 86.3 80.7 45.7


137

403.6 376.7 294.0 290.5

± ± ± ±

483.4 361.7 210.7 223.4

346.9 284.5 341.6 313.8

± ± ± ±

215.3 125.4 172.6 152.1

164.4 108.0 155.8 127.8

± ± ± ±

121.8 88.6 234.8 146.4

74.8 99.8 114.1 94.3

± ± ± ±

64.6 100.3 107.0 108.9

191.8 133.2 416.3 180.4

± ± ± ±

235.9 185.1 276.9 127.3

167.4 155.8 294.7 217.0

± ± ± ±

168.7 162.3 228.0 150.6

67.9 49.5 66.6 61.1

± ± ± ±

ited the highest range of glucose tolerance, whereas A/J mice showed relative glucose intolerance when compared with the other strains (Table 4). We carried out identical analyses using the plasma insulin measures. We found that all blood glucose and immunoreactive insulin (IRI) levels were significantly correlated (e.g., fasting levels: p = 0.001, correlation = 0.39) but this was not true for the calculated measures (AUC and D). As for glucose, we observed significant strain effects on all measures (fasting insulin: F7,92 = 5.3, p < 0.001; plasma IRI at 15 min: F7,92 = 10.2, p < 0.001; plasma IRI at 30 min: F7,92 = 6.9, p < 10)5; Plasma IRI at 75 min: F7,92 = 5.8, p < 10)4; AUC_IRI: F7,92 = 6.5, p < 10)5; DIRI: F7,92 = 4.8, P = 0.0001). A/J and C57BL/6J mice showed relatively low fasting insulin values and poor insulin secretory response to glucose in vivo when compared with the other strains tested (Table 4). In contrast, BALB/cJ and DBA/2J mice showed the highest insulin secretory response to the glucose challenge when compared with the other strains.

Data are expressed as mean ± standard deviation.

41.1 83.1 26.8 18.2 ± ± ± ± 125.4 113.8 109.7 116.1

0.52 ± 0.29 0.35 ± 0.17 0.35 ± 0.17 0.19 ± 0.12 0.50 ± 0.19 0.23 ± 0.15 0.20 ± 0.25

0.15 ± 0.03

0.7 ± 0.6 1.2 ± 1.0 2.8 ± 1.8 3.3 ± 0.4

1.5 ± 1.6

C57BL/6J A/J

AKR/J

BALB/cJ

C3H/HeJ

0.5 ± 0.4

1.3 ± 0.4

2.0 ± 0.5

ET AL.:

Context freezing Minutes freezing (min) Cue conditioning Time spent freezing at cue (min) FPS Pretrain tone + startle Pretrain startle Test tone + startle Test startle

Table 3. Conditioned tests of anxiety for eight inbred strains (12 mice of each strain)

CBA/J

DBA/2J

LP/J

L.C. SOLBERG

Physiologic tests: allergen-mediated alteration of airway responsiveness. To confirm the effectiveness of the protocol for allergen-induced airway responsiveness, we sensitized two groups of 12 mice (C57BL/6J) with ovalbumin; one group was challenged with phosphate buffered saline (PBS) and one group with nebulized ovalbumin. There were two further controls: one group of 12 mice was left unsensitized but was challenged with ovalbumin and one group of 12 mice was untreated. All four groups were then given nebulized metacholine. We used the empirically derived parameter Penh as a measure of changes in the respiratory system caused by bronchoconstriction (Hamelmann et al. 1997). Penh and its derivation are explained in Fig. 1. Sensitization with ovalbumin followed by metacholine challenge significantly increased airway responsiveness, as indicated by the change in Penh (p < 0.001). The control groups showed similar increases in Penh in response to aerosolized metacholine compared to the Penh values after receiving nebulized PBS. Mice that were sensitized and challenged with allergen had a significantly greater increase in Penh (in response to nebulized metacholine) than controls (p < 0.01). When animals from all eight strains were sensitized and challenged with nebulized metacholine, significant effects of strain were found for changes in Penh (F7,91 = 35.5, p < 10)22). C3H/HeJ and C57BL/6J strains showed the smallest change in Penh, indicating that these strains have the lowest airway response to allergen challenge. In addition,

± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

17866 16860 134 387 436 312 0.28 82.21 78.71 0.30 1.12 1.57 1.31 0.06 0.43 0.42

15151 13989 155 386 419 271 0.33 22.96 20.48 0.12 0.44 0.71 0.81 0.01 0.59 0.56

3105 3223 21 31 39 44 0.06 51.96 31.27 0.40 1.08 1.20 0.65 0.03 0.28 0.25

1775 1794 19 32 30 37 0.07 16.45 77.90 0.07 0.31 0.42 0.46 0.05 0.11 0.10 12778 11720 141 326 362 244 0.26 23.15 22.33 0.03 0.43 0.53 0.35 0.02 0.07 0.07

8263 7177 145 274 305 164 0.33 38.77 33.40 0.22 0.54 0.96 0.27 0.02 0.16 0.14 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± 2792 2774 16 54 54 60 0.15 75.30 54.29 0.28 1.07 1.35 1.59 0.02 0.04 0.03

718 780 10 12 15 7 0.04 77.90 116.45 0.51 2.30 2.95 1.08 0.04 0.06 0.05

AKR/J

11108 10190 122 328 337 153 0.38 88.14 67.64 0.42 0.55 1.80 0.85 0.02 0.20 0.19

10108 9481 84 282 260 137 0.26 35.68 35.30 0.08 0.57 1.24 0.63 0.03 0.20 0.19 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± 3040 3053 8 53 58 28 0.11 148.79 209.70 0.81 3.29 5.02 2.60 0.07 0.09 0.08

1912 1973 10 26 32 36 0.10 59.27 140.51 0.71 2.82 3.67 1.63 0.05 0.08 0.07

BALB/cJ

14616 13585 138 415 386 225 0.34 42.33 18.94 0.38 0.73 0.33 0.09 0.00 0.10 0.10

14754 13897 114 398 369 204 0.32 38.31 24.33 0.17 0.59 0.92 0.41 0.01 0.10 0.10 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± 1702 1724 12 42 30 44 0.11 26.00 80.61 0.78 2.73 2.28 0.79 0.04 0.06 0.06

3369 3427 15 51 53 45 0.09 83.20 33.20 0.77 2.88 2.32 0.73 0.04 0.06 0.06

C3H/HeJ

14728 13811 122 313 369 279 0.20 358.53 300.78 1.56 4.11 7.78 6.67 0.08 0.50 0.48

12365 11574 105 275 320 222 0.22 474.11 340.71 1.81 6.82 10.33 7.59 0.03 0.63 0.60 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± 3693 3828 32 26 48 62 0.05 454.82 337.22 1.57 5.57 9.23 8.37 0.03 0.27 0.24

3548 3581 12 33 59 61 0.07 499.07 434.81 1.46 6.90 9.35 9.45 0.03 0.31 0.29

C57BL/6J

14802 13831 129 390 392 220 0.38 42.37 35.73 0.12 1.17 0.80 0.29 0.02 0.62 0.60

15312 14553 101 351 392 193 0.44 37.96 36.18 0.10 0.77 0.69 0.56 0.01 0.57 0.55 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± 2992 3065 12 54 46 25 0.08 66.69 101.49 0.46 3.33 2.06 0.78 0.05 0.28 0.25

1912 1940 10 28 39 24 0.08 49.64 81.24 0.42 2.05 1.92 0.80 0.03 0.30 0.28

CBA/J

11997 11098 120 278 327 231 0.21 143.22 138.81 0.20 0.76 2.92 1.82 0.01 0.18 0.17

13981 13386 79 287 323 211 0.25 248.06 202.59 0.61 8.20 3.90 1.97 0.14 0.18 0.17 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± 1877 1843 12 33 38 33 0.05 170.55 201.13 0.41 3.56 3.58 2.89 0.02 0.03 0.03

1904 1937 0.6 1.5 2.3 4.4 0.07 226.65 210.41 0.43 15.43 1.04 1.35 0.26 0.05 0.06

DBA/2J

18227 17422 107 367 402 316 0.18 31.35 16.78 0.22 0.33 0.19 0.49 0.02 0.06 0.06

13079 12296 104 300 325 227 0.22 34.72 32.18 0.51 0.54 0.58 0.58 0.01 0.07 0.07 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

2564 2536 9 28 46 53 0.06 63.17 98.97 0.48 1.61 1.72 2.30 0.04 0.02 0.02

1736 1718 10 25 33 32 0.03 76.94 22.78 0.98 1.97 2.05 1.97 0.02 0.03 0.02

LP/J

L.C. SOLBERG


Females AUC-G DG Glucose 0 (mg/dl) Glucose 15 (mg/dl) Glucose 30 (mg/dl) Glucose 75 (mg/dl) Glucose K AUC-IRI DIRI Insulin 0 (ng/ml) Insulin 15 (ng/ml) Insulin 30 (ng/ml) Insulin 75 (ng/ml) Insulin K DIRI/DG AUC-IRI/AUC-G Males AUC-G DG Glucose 0 (mg/dl) Glucose 15 (mg/dl) Glucose 30 (mg/dl) Glucose 75 (mg/dl) Glucose K AUC-IRI DIRI Insulin 0 (ng/ml) Insulin 15 (ng/ml) Insulin 30 (ng/ml) Insulin 75 (ng/ml) Insulin K DIRI/DG AUC-IRI/AUC-G

A/J

Table 4. Intraperitoneal glucose tolerance test for eight inbred strains (6 males and 6 females of each strain)

138 ET AL.:


L.C. SOLBERG

ET AL.:

139


Table 5. Plethysmograph phenotypes for eight inbred strains (12 mice of each strain)

A/J Penh difference Respiratory rate (sec) Tidal volume (ml) Inspiratory time (sec) Expiratory time (sec)

1.2 414 0.27 0.07 0.10

± ± ± ± ±

0.5 53 0.07 0.01 0.01

AKR/J 4.8 348 0.29 0.07 0.15

± ± ± ± ±

1.7 73 0.04 0.01 0.04

BALB/cJ 2.4 518 0.25 0.04 0.08

± ± ± ± ±

0.6 38 0.03 0.01 0.01

C3H/HeJ 1.1 518 0.26 0.04 0.09

± ± ± ± ±

0.6 50 0.03 0.01 0.01

C57BL/6J 1.0 549 0.32 0.04 0.08

± ± ± ± ±

0.5 29 0.04 0.01 0.01

CBA/J 1.2 518 0.26 0.05 0.08

± ± ± ± ±

0.4 46 0.05 0.01 0.01

DBA/2J 1.4 532 0.32 0.05 0.08

± ± ± ± ±

0.4 23 0.04 0.01 0.01

LP/J 2.3 390 0.28 0.06 0.12

± ± ± ± ±

0.7 58 0.03 0.01 0.02


the whole-body plethysmograph provided a number of other measures of lung function (as illustrated in Fig. 4) and significant strain effects were observed for a number of these: respiratory rate (F7,91 = 30.4, p < 10)20), tidal volume (F7,91 = 8.9, p < 10)7), inspiratory time (F7,91 = 21.5, p < 10)16), and expiratory time (F7,91 = 26.5, p < 10)18). We found no significant effect of gender on these measures. Table 5 shows the mean and standard deviations of these phenotypes. Immunology and hematology. We obtained a full-blood count and antibody stains for CD4, CD3, CD8 and B220 on all eight strains. There were significant effects of gender, so we report means and standard deviations for both sexes in Table 6. Significant strain effects were found for all hematologic phenotypes, varying from a remarkably large effect on mean cellular volume (F7,95 = 212, p < 10)54) to a smaller effect on white blood cell count (F7,95 = 4.1, p = 0.0006), with other measures lying between these extremes: hemoglobin concentration (F7,95 = 7.9, p < 10)6), hematocrit (F7,95 = 11.4, p < 10)9), platelet count (F7,95 = 5.3, p < 10)4), and red blood cell count (F7,95 = 6.9, p < 10)5). Large effects were also found for the immunologic parameters: percent B220+ cells (lymphocytes) (F7,95 = 17.6, p < 10)14), percent CD4+/CD3+ cells (T helper cells) (F7,95 = 25.1, p < 10)18), and percent CD8+/CD3+ cells (cytotoxic T cells) (F7,95 = 24.3, p < 10)17) Biochemistry. We found significant gender effects on high-density lipoprotein, triglyceride, and sodium and chloride concentrations. Two measures showed no significant strain effect: a liver enzyme (aspartate transaminase) and creatinine. Strain effects on triglycerides were marginally significant (p = 0.02). All other measures were significant (at p < 10)5), with notably huge effects on urea (F7,92 = 26.1, p < 10)19), total cholesterol (F7,92 = 16.9, p < 10)14), and high-density lipoprotein concentration (F7,92 = 21.2, p < 10)16). Means and standard deviations are shown in Table 7. It should be noted that our measurements were taken after 18 h of food deprivation and are not comparable

with values obtained while animals were on high fat diets or allowed free access to food (Hough et al. 2002). Heterogeneous stock. Phenotype data were collected for 16 tests from 2491 HS mice (1220 female, 1271 male) between the ages of 37 and 72 days. The entire experiment took 620 days (89 weeks). Testing took place in six testing rooms located off one corridor. Ninety-five percent of all animals were phenotyped by six experimenters; an additional six experimenters carried out the remaining 5% of the experiments. The only phenotype we discarded was the potentiation of startle because we were not able to find the expected increase after fear conditioning. Figure 5 is a box-and-whisker plot for the three startle measures (cue, startle, and cue + startle) taken before (labeled ‘‘Pre’’ in Fig. 5) and after training (‘‘Test’’). The figure shows that animals exposed to a startle stimulus preceded by a cue showed an increase in the startle response compared with the startle response alone, but that this difference was present before training. Furthermore, the potentiation of the startle response was less after fear conditioning, falling from a mean of 32.8 to 17.6. One possible explanation was that there had been a failure in fear conditioning, but we were able to exclude that explanation because all animals showed an increase in freezing response when presented with the cue. We were unable to explain the absence of startle potentiation and decided to discard the potentiation phenotype from further analyses. Tables 8 and 9 give the number of usable phenotypes collected from each test and their means and standard deviations. Discussion We have developed a protocol suitable for highthroughput phenotyping of a large number of mice and suitable for the genetic analysis of both inbred and heterogeneous stock mice. We modeled anxiety using both unconditioned measures (open-field activity, elevated plus maze, and food neophobia) and conditioned measures (fear potentiated startle, context freezing, cue conditioning). Fear potentiated

± ± ± ± ± ±

1.3 0.5 0.8 97.7 0.2 0.4

± ± ± ± ± ±

2.2 0.8 0.5 115.0 0.5 0.5

17.5 ± 4.5 78.4 ± 1.2 20.6 ± 1.1

32.7 11.3 44.7 761.3 7.3 1.1

19.4 ± 3.3 79.3 ± 1.3 19.7 ± 1.2

32.6 11.3 45.3 584.3 7.2 0.9

± ± ± ± ± ±

6.6 2.6 0.5 120.0 1.6 1.1

± ± ± ± ± ±

2.2 0.9 0.5 50.7 0.5 0.4

16.6 ± 2.6 78.9 ± 2.7 20.1 ± 2.4

36.1 13.5 42.4 640.5 8.6 0.8

15.5 ± 2.7 82.5 ± 1.0 16.5 ± 0.9

34.8 12.8 42.5 592.2 8.2 3.8

AKR/J


Female Hematology HCT (%) HGB (g/dl) MCV (fl) PLT (n/l) RBC (n/ll) WBC(n/ll) Immunology %B220± %CD4/CD3 %CD8/CD3 Male Hematology HCT (%) HGB (g/dl) MCV (fl) PLT (n/ll) RBC (n/ll) WBC (n/ll) Immunology %B220+ %CD4+/CD3+ %CD8+/CD3+

A/J

± ± ± ± ± ±

5.3 1.5 1.4 64.3 1.1 0.7

± ± ± ± ± ±

2.3 0.8 0.6 48.0 0.5 0.3

23.2 ± 7.0 77.8 ± 2.0 21.4 ± 1.9

37.0 13.2 44.0 497.3 8.4 1.6

24.8 ± 5.3 78.3 ± 1.1 20.8 ± 1.2

32.7 12.1 43.0 434.5 7.6 1.3

BALB/cJ

± ± ± ± ± ±

3.2 1.1 1.0 118.7 0.7 1.0

± ± ± ± ± ±

2.9 0.9 0.8 55.1 0.6 0.6

13.5 ± 4.6 81.4 ± 1.1 17.4 ± 1.3

36.3 12.7 48.3 631.5 7.5 2.6

15.8 ± 4.2 78.7 ± 3.0 20.3 ± 2.8

30.6 11.1 47.7 539.7 6.4 2.1

C3H/HeJ

± ± ± ± ± ±

8.8 2.7 1.5 138.0 1.6 0.9

± ± ± ± ± ±

2.0 0.6 1.3 172.8 0.3 0.7

33.4 ± 5.3 72.4 ± 3.1 26.0 ± 2.8

36.2 11.1 49.8 749.2 7.3 2.3

37.9 ± 6.2 71.0 ± 2.4 27.1 ± 2.3

33.6 10.6 50.4 457.8 6.6 2.4

C57BL/6J

Table 6. Hematology and immunology for eight inbred strains (6 males and 6 females of each strain)

± ± ± ± ± ±

3.6 1.4 0.5 91.1 0.9 0.3

± ± ± ± ± ±

1.3 0.5 0.8 56.1 0.4 0.4 21.0 ± 5.5 76.5 ± 1.7 22.2 ± 1.8

29.5 11.8 39.8 598.5 7.4 1.6

26.9 ± 9.7 71.4 ± 2.3 27.1 ± 2.4

29.3 11.6 39.7 476.5 7.4 1.1

CBA/J

± ± ± ± ± ±

2.9 1.0 0.5 21.4 0.6 0.8

± ± ± ± ± ±

2.5 0.9 0.0 45.0 0.6 0.8 27.8 ± 4.1 82.2 ± 2.1 16.6 ± 1.9

38.0 12.6 42.0 682.6 9.1 2.0

30.1 ± 5.9 80.1 ± 1.4 18.5 ± 1.4

32.2 10.8 41.5 602.8 7.7 1.5

DBA/2J

± ± ± ± ± ±

1.7 0.6 0.7 39.1 0.3 1.4

± ± ± ± ± ±

2.8 0.8 0.5 35.1 0.5 0.5 22.1 ± 4.1 76.0 ± 1.1 22.6 ± 1.1

43.2 14.5 48.6 496.4 8.9 1.1

27.1 ± 5.1 74.2 ± 1.8 24.4 ± 1.6

39.4 13.4 49.0 449.0 8.1 2.4

LP/J

140 L.C. SOLBERG ET AL.:


Female Sodium (mmol) Chloride (mmol) Urea (lmol) Creatinine (lmol) Calcium (mmol) Phosphorus (mmol) ALP (U/L) ALT (U/L) AST (U/L) Total protein (g/L) Albumin (g/L) Total cholesterol (mmol) HDL (mmol) LDL (mmol) Triglycerides (mmol) Male Sodium (mmol) Chloride (mmol) Urea (lmol) Creatinine (lmol) Calcium (mmol) Phosphorus (mmol) ALP (U/L) ALT (U/L) AST (U/L) Total protein (g/L) Albumin (g/L) Total cholesterol (mmol) HDL (mmol) LDL (mmol) Triglycerides (mmol) ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

129.4 93.8 5.96 17.8 2.00 2.38 95.6 42.4 155.8 48.0 30.3 2.29 1.03 0.22 0.26

136.2 100.7 6.93 21.8 2.06 1.92 94.8 43.0 139.5 48.5 30.3 2.39 1.27 0.27 0.92

A/J

136.8 100.7 7.27 19.5 2.06 2.30 46.3 22.5 93.2 48.7 28.7 2.02 0.85 0.14 0.30

135.3 95.5 6.00 18.5 2.14 1.86 47.2 17.8 103.2 49.4 31.0 2.15 0.96 0.19 0.64 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

± ± ± ± ± ± ± ± ± ± ± ± ± ± ± 7.3 5.1 1.53 2.4 0.06 0.38 6.9 3.8 18.1 1.4 1.4 0.30 0.09 0.03 0.12

4.7 4.0 0.66 1.3 0.07 0.28 7.9 1.7 17.1 2.5 2.0 0.25 0.07 0.02 0.36 123.7 90.8 7.62 19.0 1.83 2.60 81.7 24.0 121.0 50.4 28.1 3.40 1.81 0.32 0.29

123.0 91.3 6.66 19.8 1.78 2.44 66.2 24.0 128.5 45.5 28.7 2.81 1.45 0.23 0.29 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

± ± ± ± ± ± ± ± ± ± ± ± ± ± ± 6.1 5.2 1.45 1.3 0.09 0.20 14.2 7.4 44.6 3.0 1.6 0.53 0.15 0.07 0.12

11.4 8.7 1.31 1.0 0.35 0.25 26.9 5.6 24.7 7.4 4.3 0.41 0.22 0.04 0.17

BALB/cJ

131.5 98.2 5.95 18.3 1.97 2.18 86.0 32.5 151.5 48.2 27.6 3.22 2.01 0.32 0.60

126.5 95.5 5.67 17.3 1.86 2.10 85.3 22.0 164.2 43.7 26.9 2.70 1.61 0.30 0.37 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

± ± ± ± ± ± ± ± ± ± ± ± ± ± ± 3.4 3.3 0.57 1.0 0.07 0.25 15.8 14.4 81.2 2.6 2.1 0.28 0.19 0.05 0.11

3.8 3.7 1.00 0.6 0.12 0.14 9.1 12.8 18.0 3.6 2.2 0.36 0.25 0.06 0.09

C3H/HeJ

126.7 91.2 9.97 18.6 1.95 2.86 65.8 26.8 111.3 48.2 28.1 2.67 1.03 0.35 0.28

121.2 85.8 10.82 18.3 1.83 2.62 103.0 20.2 102.0 45.4 28.4 2.38 0.82 0.29 0.17 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

± ± ± ± ± ± ± ± ± ± ± ± ± ± ± 10.7 7.6 0.87 2.1 0.16 0.66 12.0 20.1 59.4 3.5 2.7 0.26 0.27 0.04 0.15

4.2 2.9 1.38 1.6 0.09 0.08 23.1 4.9 19.0 2.2 2.0 0.58 0.17 0.07 0.04

C57BL/6J

128.8 97.7 7.55 20.3 1.99 2.63 73.8 37.2 147.8 46.9 26.3 2.87 1.59 0.33 0.70

129.5 95.8 6.12 18.8 1.90 2.95 95.0 35.2 184.3 43.4 25.6 2.04 1.07 0.23 0.44 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

± ± ± ± ± ± ± ± ± ± ± ± ± ± ± 6.9 4.9 1.16 1.0 0.09 0.49 16.0 13.7 66.5 2.4 1.9 0.25 0.21 0.13 0.31

5.5 4.5 0.50 1.9 0.14 0.56 15.6 19.4 41.7 2.6 4.4 0.59 0.27 0.06 0.21

CBA/J

138.8 104.2 6.46 20.0 2.07 2.49 103.0 17.2 121.6 49.5 27.3 3.08 1.83 0.39 1.11

131.7 96.7 6.28 21.3 2.01 2.57 109.5 20.2 142.8 47.9 28.0 2.49 1.42 0.36 0.43 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

± ± ± ± ± ± ± ± ± ± ± ± ± ± ± 5.4 3.3 0.77 1.4 0.15 0.41 13.7 6.1 12.5 2.3 1.3 0.15 0.15 0.04 0.28

5.9 5.2 0.45 2.1 0.08 0.43 10.5 10.5 64.5 3.0 2.2 0.31 0.10 0.07 0.08

DBA/2J

131.2 99.0 10.28 20.0 2.08 2.40 82.8 18.8 95.8 52.2 29.7 3.44 1.98 0.29 0.73

124.5 91.5 8.98 17.7 1.90 2.44 82.9 17.6 151.0 53.8 34.1 3.70 1.57 0.39 0.32

± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

± ± ± ± ± ± ± ± ± ± ± ± ± ± ±

3.1 3.0 1.67 2.5 0.16 0.28 10.1 3.9 12.4 1.4 1.2 0.24 0.10 0.04 0.16

5.7 4.3 1.04 1.2 0.13 0.31 10.2 3.3 26.1 2.5 1.7 0.47 0.29 0.08 0.07

LP/J

ET AL.:

6.7 4.9 0.84 2.6 0.17 0.25 12.9 7.2 40 3.2 2.0 0.32 0.16 0.05 0.34

4.6 5.0 0.78 1.3 0.07 0.34 11.1 13.4 9.2 1.6 0.9 0.28 0.15 0.02 0.06

AKR/J

Table 7. Biochemistry phenotypes for eight inbred strains (6 males and 6 females of each strain)

L.C. SOLBERG HIGH-THROUGHPUT PHENOTYPING

141

142

L.C. SOLBERG

ET AL.:


Fig. 5. Box-and-whisker plot of fear potentiated startle in 2000 HS mice. Data for the three startle measures are shown: cue, startle, and cue + startle. The measures were taken before (labeled ‘‘Pre’’) and after training (labeled ‘‘Test’’). The vertical scale shows the startle response. Means of the three groups are shown in shaded boxes above the box-and-whisker plots.

startle (Davis et al. 1993) was of particular interest because it can be used to assess fear conditioning in humans, providing there is a homologous measure of anxiety in rodents (Grillon et al. 1991), and because it appears to share a common neuroanatomical basis in different species (Davis and Whalen 2001). Fear potentiation can be reliably induced in rats but there are only two publications that describe its measurement in mice (Falls et al. 1997; McCaughran et al. 2000). We also measured conditioned freezing, a widely used paradigm for investigating the neurobiology of fearfulness. As a model of type II diabetes, we used an intraperitoneal glucose tolerance test that measured both blood glucose and insulin after glucose infusion. We used an allergen-mediated alteration of airway responsiveness as a model of asthma. Barometric whole-body plethysmography was used to measure airway resistance in unrestrained conscious mice after sensitization and airway challenge with ovalabumin (Eum et al. 1995; Hamelmann et al. 1997).

We demonstrated significant strain differences between eight progenitor strains of the heterogeneous stock using data from our protocol and then collected phenotypes for 2500 HS mice. With the exception of the fear-potentiated startle, we acquired usable data on all phenotypes, ranging from 1900 hematologic and immunologic phenotypes to 2500 results for open-field activity. Over 100 phenotypes were collected on 2000 HS animals. This is the first time that such an extensive set of phenotypes has been collected on such a large cohort of animals, making it possible to investigate relationships between a range of physiologic and behavioral parameters. With a few exceptions (Stoll et al. 2001), segregating populations are phenotyped for a limited number of measures so that the genetic architecture of each complex trait or disease model has been studied in isolation. Our protocol makes it possible to apply a systems biology approach to complex traits in the mouse. An important question is the extent to which results are comparable to those collected by other

L.C. SOLBERG

ET AL.:

143


Table 8. Heterogeneous stock mice: means and standard deviations of phenotypes

Test

Number

Open field arena Total activity Center time Latency Boli Elevated plus maze Closed arm distance Open arm distance Closed arm time Open arm time Open arm latency Food hyponeophagia FN latency New home-cage activity Total beam breaks Species-typical behavior Pellets burrowed

2504

Mean

SD

2233.3 7 192.5 4.5

817.4 8.1 98.4 2.5

963.4 238.2 122.8 37.2 63.8

277.1 229 38.9 32.6 98.5

114.9

134.5

698.8

216

2452

2474 2445

Test

Number

Context freezing Minutes freezing Cue conditioning Minutes freezing at cue Fear-potentiated startle Pretrain tone + startle Pretrain startle Test tone + startle Test startle Plethysmography Penh difference Respiratory rate Tidal volume Inspiratory time Expiratory time

2070

Mean

SD

1.38

0.99

3.59

2.67

2110 2005 439.1 406.3 505.2 487.6

323.9 310.4 317.2 321.3

2304 1.305 505.2 0.298 0.048 0.089

1.006 69.1 0.062 0.008 0.016

2455 88.9

65.3

SD = standard deviation.

investigators, because there is evidence that some phenotypes may be specific to an environment and that results may not be generalized from one laboratory to another (Crabbe et al. 1999). However, because other investigators have collected typically only a small number of phenotypes on any one animal, it is not clear how reliable the comparisons with published data sets will be. We compared our results with those reported in the Mouse Phenome Database, a community resource for mouse strain characterization data (http:// www.jax.org/phenome) (Grubb et al. 2004). Currently, the database does not include relevant data for all the strains we have analyzed, nor does it include every phenotype. Table 10 summarizes comparable data for 12 phenotypes. Absolute values differed considerably between the data sets, presumably because of differences in measurement (e.g., the Coulter counter used in our study for hematologic assays had been calibrated to measure human samples, which is not true for the data reported to the Jackson Laboratory). Consequently, we looked for correlations between the phenotypic means in each data set and calculated the correlation coefficients (Pearson’s product-moment correlation). The coefficients were greater than 0.5 for all measures except lymphocyte counts (0.26) and mean cellular volume (0.46). A behavioral measure, homecage activity, was highly correlated between the data sets (0.75). Our results are also consistent with published strain comparisons, although there have been no reports that include all eight HS progenitor strains. So, for example, our findings that C57BL/6 mice are relatively hyperglycemic and glucose intolerant

have been reported elsewhere (Goren et al. 2004; Rossmeisl et al. 2003); the strain distribution pattern of airway response [in which C3H/HeJ and C57BL/6J strains are lowest (Table 5)] replicates previous studies (Ewart et al. 2000; Konno et al. 1993; Levitt and Mitzner 1988; Longphre and Kleeberger 1995). Our behavioral analyses also concur with earlier reports that C57BL/6J is among the least, while A/J and C3H/HeJ are among the most anxious strains [as measured by time spent in the open arms of the elevated plus maze (Table 2)] (Crabbe et al. 1999; Crawley et al. 1997). We were also able to replicate the finding that C57BL/6J freezes less than DBA/2 to context conditioning (Table 3) (Ammassari-Teule et al. 2000; Paylor et al. 1994). However, there are a number of discrepancies; for example, we find that BALB/cJ is a relatively inactive strain in the open field, while others have found that it is second to C57BL/6 (Table 2) (Crabbe et al. 1999). In fact, given that we have tested only six animals of each gender for these phenotypes, the degree of agreement with other behavioral studies is surprising because differences in the equipment and protocols and other environmental variables will affect test results. We will be able to determine some of the sources of environmental variation in an analysis of covariates collected during our study of 2500 HS animals. Finally, we note that a number of our assays are included in the EUMORPHIA consortium (http:// www.eumorphia.org), which describes a series of standardized operating procedures (SOP) for mouse phenotyping (Green et al. 2005; The Eumorphia Consortium 2005). The biochemistry screen was performed using a EUMORPHIA SOP, although our

144

L.C. SOLBERG

ET AL.:


Table 9. Heterogeneous stock mice: means and standard deviations of phenotypes for each sex

Test

No. Males

IPGTT AUC-G DG Glucose 0 (mg/dl) Glucose 15 (mg/dl) Glucose 30 (mg/dl) Glucose 75 (mg/dl) Glucose K AUC-IRI DIRI Insulin 0 (ng/ml) Insulin 15(ng/ml) Insulin 30 (ng/ml) Insulin 75 (ng/ml) Insulin K DIRI/DG AUC-IRI/AUC-G Immunology %B220+ %CD4+CD3+ %CD8+/CD3+ Hematology HCT (%) HGB (g/dl) MCV (fl) PLT (n/ll) RBC (n/ll) WBC (n/ll) Biochemistry ALP (U/L) ALT (U/L) AST (U/L) Albumin (g/L) Calcium (mmol) Chloride (mmol) Creatinine (mmol) Glucose (mmol) HDL (mmol) LDL (mmol) Phosphorus (mmol) Potassium (mmol) Sodium (mmol) Total cholesterol (mmol) Total protein (g/l) Triglycerides (mmol) Urea (mmol)

1188

Mean

SD

11,860 12,952 146 327 361 254 2.29 93.9 98.4 0.608 1.831 1.999 2.067 0.023 0.824 0.872

3,201 3,159 32 63 64 73 1.05 57.4 58.2 0.500 1.269 1.168 1.359 0.024 0.541 0.603

30.2 70.6 27.3

9.7 6.4 6.2

32.6 12.3 45.6 623.5 7.8 1.53

2.7 1.5 1.7 129.3 1 0.77

120.20 32.06 125.76 28.45 2.06 97.15 20.69 9.31 1.838 0.444 2.553 7.564 132.40 3.355 50.52 0.598 7.892

35.53 20.34 81.81 2.73 0.19 7.06 2.66 2.45 0.446 0.143 0.383 1.200 8.22 0.592 4.43 0.352 1.789

No. Females

Mean

SD

12,352 13,221 116 208 343 324 2.64 84.3 88.0 0.498 1.739 1.933 1.488 0.026 0.733 0.761

3,497 3,479 22 64 72 85 1.16 54.1 54.5 0.470 1.258 1.102 1.139 0.022 0.509 0.556

26.8 71.3 26.2

9.4 6.6 6.4

34.4 11.9 45.4 559.2 7.56 1.47

4.3 1.4 1.6 108.7 0.94 0.87

130.62 29.34 164.57 29.80 2.02 94.72 19.86 8.57 1.354 0.394 2.531 7.443 128.63 2.807 48.82 0.338 7.687

37.52 20.85 112.54 2.91 0.21 7.57 2.55 2.59 0.390 0.117 0.423 1.108 8.89 0.567 4.59 0.185 1.692

1145

966

905

971

920

1057

1029

blood samples did not undergo the same quality control checks. Consequently, there may have been more hemolysis than allowed by EUMORPHIA SOPs, which may explain why our sodium, chloride, and triglyceride measurements differ from those reported by EUMORPHIA. Differences may also have arisen because our mice have been through a particular battery of assays rather than assayed for each measure individually. Our protocol for FACs analysis is also comparable to that used by EUMORPHIA.

However, EUMORPHIA behavioral assays do not overlap with the phenotypes described here. Our data set of multiple phenotypes collected on over 2000 heterogeneous stock animals provides a starting point for genetic analysis of quantitative variation in a large outbred population. By mapping multiple QTLs simultaneously, it will be possible to detect genetic effects that are invisible to single gene-based approaches, such as mutagenesis and gene targeting. Not only can gene by gene (epistatic)

14.0 16.4 41.6 48.8 82.4 13.5 0.293 0.319 453.0 532.3 0.055 0.079 0.093 0.079 22.4 21.8 22.6 22.1

14.9 16.2 44.0 47.8 77.4 6.5 11.6 13.9 34.8 41.7 81.4 11.7 15.7 29.4 39.8 91.4 18.2 14.4 15.9 42.1 47.2 81.7 9.4

Haemoglobin concentration 14.2 11.3 Mean cell hemoglobin 15.2 15.6 Hematocrit 43.6 32.6 Mean red cell volume 44.9 45.0 % lymphocytes 85.2 87.8 Home-cage activity 10.4 8.3 Mean tidal volume 0.269 0.275 Mean respiratory rate 375.0 414.2 Mean inspiratory time 0.061 0.101 Mean expiratory time 0.130 0.101 Body weight, 6 weeks of age 20.1 21.1

14.8 15.9 42.7 45.8 71.2 11.8

13.1 15.7 35.4 42.5 82.8 23.0

14.8 12.7 16.3 16.0 43.6 34.8 47.0 43.5 78.0 84.6 14.1 13.6 0.290 0.246 498.0 518.4 0.049 0.088 0.085 0.088 28.6 23.6 22.8 17.5

14.9 16.9 42.7 49.1 74.0 8.0

11.9 17.1 33.5 48.0 87.8 15.6

14.6 11.4 15.2 15.6 45.8 36.6 46.1 50.1 87.1 84.1 20.1 35.1 0.321 0.322 472.0 549.6 0.048 0.083 0.100 0.083 22.1 19.1 20.6 19.4

WT WT JAX WT JAX WT JAX WT JAX WT JAX Phenotype

145


14.1 13.7 44.8 41.5 79.5

WT JAX WT

ET AL.:

JAX

JAX

DBA/2J CBA/J C57BL/6J C3H/He BALB/cJ AKR/J A/J

Table 10. Comparison between mean values of phenotypes acquired in this study (WT) and the phenome project at the Jackson laboratory (JAX).

LP/J

L.C. SOLBERG

interactions be detected, but when multiple phenotypes are analyzed, pleiotropic gene action can be identified. Furthermore, it is possible to search for interactions between genetic variants and environmental variables, assuming these can be measured. We are currently investigating whether our data can be used to study the genetic architecture of complex traits in mice. Acknowledgments This work was supported by the Wellcome Trust. References 1. Ammassari Teule M, Passino E, Restivo L, de Marsanich B (2000) Fear conditioning in C57/BL/6 and DBA/2 mice: variability in nucleus accumbens function according to the strain predisposition to show contextual- or cue-based responding. Eur J Neurosci 12, 4467 4474 2. Brown SD, Hardisty RE (2003) Mutagenesis strategies for identifying novel loci associated with disease phenotypes. Semin Cell Dev Biol 14, 19 24 3. Clark LD, Clark RK, Heber Katz E (1998) A new murine model for mammalian wound repair and regeneration. Clin Immunol Immunopathol 88, 35 45 4. Crabbe JC, Wahlsten D, Dudek BC (1999) Genetics of mouse behavior: interactions with laboratory environment. Science 284, 1670 1672 5. Crawley JN, Belknap JK, Collins A, Crabbe JC, Frankel W, et al. (1997) Behavioral phenotypes of inbred mouse strains: implications and recommendations for molecular studies. Psychopharmacology (Berl) 132, 107 124 6. Davis M, Whalen PJ (2001) The amygdala: vigilance and emotion. Mol Psychiatry 6, 13 34 7. Davis M, Falls WA, Campeau S, Kim M (1993) Fearpotentiated startle: a neural and pharmacological analysis. Behav Brain Res 58, 175 198 8. Deacon RM, Rawlins JN (2005) Hippocampal lesions, species-typical behaviours and anxiety in mice. Behav Brain Res 156, 241 249 9. Demarest K, Koyner J, McCaughran J Jr, Cipp L, Hitzemann R (2001) Further characterization and highresolution mapping of quantitative trait loci for ethanol-induced locomotor activity. Behav Genet 31, 79 91 10. Eum SY, Haile S, Lefort J, Huerre M, Vargaftig BB (1995) Eosinophil recruitment into the respiratory epithelium following antigenic challenge in hyper-IgE mice is accompanied by interleukin 5-dependent bronchial hyperresponsiveness. Proc Natl Acad Sci USA 92, 12290 12294 11. Ewart SL, Kuperman D, Schadt E, Tankersley C, Grupe A, et al. (2000) Quantitative trait loci controlling allergen-induced airway hyperresponsiveness in inbred mice. Am J Respir Cell Mol Biol 23, 537 545

146

12. Falls WA, Carlson S, Turner JG, Willott JF (1997) Fearpotentiated startle in two strains of inbred mice. Behav Neurosci 111, 855 861 13. Flint J, Valdar W, Shifman S, Mott R (2005) Strategies for mapping and cloning quantitative trait genes in rodents. Nat Rev Genet 6, 271 286 14. Goren HJ, Kulkarni RN, Kahn CR (2004) Glucose homeostasis and tissue transcript content of insulin signaling intermediates in four inbred strains of mice: C57BL/6, C57BLKS/6, DBA/2, and 129X1. Endocrinology 145, 3307 3323 15. Green EC, Gkoutos GV, Lad HV, Blake A, Weekes J, et al. (2005) EMPReSS: European mouse phenotyping resource for standardized screens. Bioinformatics 21, 2930 2931 16. Greenspan RJ (2004) E pluribus unum, ex uno plura: quantitative and single-gene perspectives on the study of behavior. Annu Rev Neurosci 27, 79 105 17. Grillon C, Ameli R, Woods SW, Merikangas K, Davis M (1991) Fear-potentiated startle in humans: effects of anticipatory anxiety on the acoustic blink reflex. Psychophysiology 28, 588 595 18. Grubb SC, Churchill GA, Bogue MA (2004) A collaborative database of inbred mouse strain characteristics. Bioinformatics 20, 2857 2859 19. Hamelmann E, Schwarze J, Takeda K, Oshiba A, Larsen GL, et al. (1997) Noninvasive measurement of airway responsiveness in allergic mice using barometric plethysmography. Am J Respir Crit Care Med 156, 766 775 20. Heldt S, Sundin V, Willott JF, Falls WA (2000) Posttraining lesions of the amygdala interfere with fearpotentiated startle to both visual and auditory conditioned stimuli in C57BL/6J mice. Behav Neurosci 114, 749 759 21. Hough TA, Nolan PM, Tsipouri V, Toye AA, Gray IC, et al. (2002) Novel phenotypes identified by plasma biochemical screening in the mouse. Mamm Genome 13, 595 602 22. Konno S, Adachi M, Matsuura T, Sunouchi K, Hoshino H, et al. (1993) [Bronchial reactivity to methacholine and serotonin in six inbred mouse strains.]. Arerugi 42, 42 47

L.C. SOLBERG

ET AL.:


23. Levitt RC, Mitzner W (1988) Expression of airway hyperreactivity to acetylcholine as a simple autosomal recessive trait in mice. FASEB J 2, 2605 2608 24. Longphre M, Kleeberger SR (1995) Susceptibility to platelet-activating factor-induced airway hyperreactivity and hyperpermeability: interstrain variation and genetic control. Am J Respir Cell Mol Biol 13, 586 594 25. McCaughran JA Jr, Bell J 3rd, Hitzemann RJ (2000) Fear-potentiated startle response in mice: genetic analysis of the C57BL/6J and DBA/2J intercross. Pharmacol Biochem Behav 65, 301 312 26. Mott R, Talbot CJ, Turri MG, Collins AC, Flint J (2000) A method for fine mapping quantitative trait loci in outbred animal stocks. Proc Natl Acad Sci USA 97, 12649 12654 27. Paylor R, Tracy R, Wehner J, Rudy JW (1994) Dba/2 and C57bl/6 mice differ in contextual fear but not auditory fear conditioning. Behav Neurosci 108, 810 817 28. R Development Core Team (2004) A language and environment for statistical computing (Vienna: R Foundation for Statistical Computing) 29. Rogers DC, Fisher EM, Brown SD, Peters J, Hunter AJ, et al. (1997) Behavioral and functional analysis of mouse phenotype: SHIRPA, a proposed protocol for comprehensive phenotype assessment. Mamm Genome 8, 711 713 30. Rossmeisl M, Rim JS, Koza RA, Kozak LP (2003) Variation in type 2 diabetes-related traits in mouse strains susceptible to diet-induced obesity. Diabetes 52, 1958 1966 31. Stoll M, Cowley AW Jr, Tonellato PJ, Greene AS, Kaldunski ML, et al. (2001) A genomic-systems biology map for cardiovascular function. Science 294, 1723 1726 32. Talbot CJ, Nicod A, Cherny SS, Fulker DW, Collins AC, et al. (1999) High-resolution mapping of quantitative trait loci in outbred mice. Nat Genet 21, 305 308 33. The Eumorphia Consortium (2005) EMPReSS: standardized phenotypic screens for functional annotation of the mouse genome. Nat Genet 37, 1 2