The postgenome era has provided resources to link disease phenotypes to the genomic sequence, i.e., creating a disease “phenome.” Our detailed characterization of the sequenced BN rat strain (BN/NHsdMcwi) provides the first concerted effort in creating a direct link between a sequenced genome and its resulting biology. For the BN sequence to be of broad value to investigators, these measures need to be put into the context of the spectrum of the laboratory rats, so that their physiology can be benchmarked against the sequenced BN. As a major step in generating a comprehensive cardiovascular and pulmonary disease phenome, we measured 281 traits related to diseases of the heart, lung, and blood (http://pga.mcw.edu) in the sequenced BN. We compared these data with those of the same traits measured across multiple genetic backgrounds, both genders, and differing environments. We show that no single strain, inbred or outbred, can be considered a physiological control strain; what is normal depends on what trait is being measured and the strains' genome backgrounds. We find vast differences between the genders, also dependent on genome background. By combining the values across all strains studied, we generated a “population” mean and normal range of values for each of these traits, which are more genetically representative than the measured values in any single inbred or outbred strain. These data provide a baseline for physiological comparison of traits related to cardiovascular, lung, blood, and renal function in the sequenced BN rats relative to the major strains of rats studied in biomedical research.
- physiological genomics
- BN/NHsdMcwi rat strain
the rat has long been the model system of choice for biochemists, neuroscientists, pharmacologists, and physiologists because of the many models (>500) that capture key aspects of human disease (15). The finished genome sequence of the human, and the draft sequence of the mouse, rat, dog, cow, and other mammals and model organisms, has led to numerous meetings and articles on what tools and resources are going to be needed to annotate the human genomic sequence and increase our understanding of human disease (19), a major effort of the National Institutes of Health (NIH) Road Map (8). A common theme has been the need to link phenotypes to the genome. Indeed, there has even been a call for the development of mouse (6) and human (11) phenome projects. Given the need to annotate the human genome with function, linking the rat into this process is a logical and necessary requirement for accelerating improvements in health care, as virtually every drug is tested in the rat before human. Here we report the results from the first large cardiovascular, renal, and pulmonary phenome project for the sequenced BN rat and 10 additional commonly used laboratory rat strains, enabling us to make unparalleled comparisons at the physiological level.
Annotation of an organism's genome with physiological traits offers several significant challenges. First, a phenotype is context dependent (e.g., environment, experimental conditions, genome background). As a result, the physiological value, without reference to at least one other strain measured using the same protocol, is of little value to the research community. Second, genetic mapping is a powerful tool for “attaching” physiological traits to the genome; however, genetic mapping is also context dependent, with genomic background effects having a significant impact on the traits that can be mapped (20, 26, 28). To achieve a large-scale biological annotation, i.e., a comprehensive phenome “scan” of the sequenced BN rat for cardiovascular and pulmonary systems, 281 traits were measured in both males and females of the sequenced BN rat strain (BN/NHsdMcwi).
Because complex traits are often context dependent, environmental stressors can be useful in assessing an organism's physiology and are sometimes necessary to overcome homeostasis of well-regulated systems (12). To address how some environments affect the BN, we measured many of the phenotypes after a high-salt diet and/or chronic hypoxia, as well as after standard rat chow and normoxia. To compare the measures in BN rats with other strains, we performed the same phenotypic characterization of 10 commonly used strains of rats: CDF, FHH, GH, LE, LEW, SHR, SS, and WKY, as well as the CD®IGS and SD, both of which are outbred Sprague Dawley substrains from different commercial suppliers. [More detailed information for the Programs for Genomic Applications (PGA) strains is as follows: Brown Norway, BN/NHsdMcwi; Fischer CDF(F-344)CrlBR, Charles River Laboratories; fawn-hooded hypertensive, FHH/EurMcwi; genetically hypertensive, GH/Omr/Mcwi; Long Evans, LE/BluGill, Univ. of Illinois at Urbana-Champaign; Lewis, LEW/CrlBR, Charles River Laboratories; spontaneously hypertensive rat, SHR/NCrlBR, Charles River Laboratories; Hsd:Sprague Dawley, SD/Hsd; Sprague Dawley CD®IGS, Charles River Laboratories; Dahl salt sensitive, SS/JrHsdMcwi; WKY/NCrlBR, Charles River Laboratories.] Selection of the strains to be characterized was based on their high genetic diversity, their common use, and their known physiology and pathophysiology (http://rgd.mcw.edu/strains). The selected strains span nearly the entire known phylogenetic tree of the laboratory rat (8 of 10 phylogenetic taxa) based on the analysis of >4,200 simple sequence repeats (SSRs) (27), carrying an average of 62% of all known alleles for a given SSR (data not shown). Consequently, these 11 genetically diverse strains, referred to here as “parental” strains, are also likely to capture a significant range of measured cardiovascular, renal, and pulmonary traits in the laboratory rat. Included in these were measures of clinical blood chemistries, renal function, blood pressure, respiratory and lung function, cardiac function, and vascular reactivity. The goal was to measure each of the 281 traits in 10 males and 10 females of each strain, resulting in a maximum number of 562 traits measured for each strain. Following the tradition of the genome projects, all data are freely available. To enable investigators to mine the physiological and genomic data in silico, we have built tools and analysis packages into an interactive web site, a novel concept in physiological research.
Detailed phenotyping protocols are posted and downloadable (http://pga.mcw.edu/pga/) and are summarized in Fig. 1 and below. For each protocol, 10 males and 10 females for each strain were characterized in parallel. With each strain, we concurrently phenotyped two SS males to be used as sentinels for quality control purposes including technical and seasonal variation. The ages of the rats entering each protocol varied from 6 to 10 wk. Diet and environmental stressors also varied among the protocols (Fig. 1). It is important to note that, although these numbers of animals entered each protocol, the data collected may not be complete for all measurements within a protocol, because of loss of data from technical problems, loss of the animal, or subsequent quality control screens.
Animals were housed in standard rat cages with lights on from 6:00 AM to 6:00 PM in the Animal Care Facility, which is approved by the American Association for the Accreditation of Laboratory Animal Care. Chronic hypoxia was induced by placing animals in a hypobaric chamber at 12% O2 for 2 wk before study. All protocols were approved by the Medical College of Wisconsin's Animal Welfare Committee.
Surgery: Renal and Respiratory Studies
While animals were under deep anesthesia (ketamine, 28 mg/kg; xylazine, 1.6 mg/kg; acepromazine, 0.4 mg/kg im), catheters were implanted aseptically in the femoral artery and vein and exteriorized at the nape of the neck for the direct measurement of heart rate and blood pressure and for sampling of arterial blood in the renal and respiratory protocols. The rats were housed in stainless steel metabolic cages, and food and water were provided ad libitum.
At 10 wk of age, the food was changed from a 0.4% to a 4% NaCl diet. Three weeks later, catheters were implanted. After 3 days of postsurgical recovery, the blood pressure response to increasing doses of angiotensin II (5.0–50 ng·kg−1·min−1 iv) and norepinephrine (0.2–1.0 μg·kg−1·min−1 iv) was measured. One week postsurgery, 3 consecutive days of systolic, diastolic, and mean arterial pressures were recorded (9:00 AM to 12:00 PM), with rats housed in their home cages. An overnight (16 h) urine sample was collected to determine levels of protein, microalbumin, sodium, potassium, and creatinine. On the third and final day of pressure recording, whole blood was collected (0.8 ml) to determine plasma renin activity and sodium, potassium, and creatinine concentrations. After the third day of blood pressure recording, each rat received an intravenous injection of furosemide (10 mg/kg), and the diet was switched to a low-salt (0.4% NaCl) diet. The following night, an overnight urine sample was collected. Blood pressure measurements were made for 2 days after the salt depletion. A final arterial blood sample was collected to determine plasma renin activity and plasma creatinine concentration while on a low-salt diet. The kidneys were then collected and weighed.
The ventilatory responses to hypoxia and hypercapnia were determined in 10-wk-old rats, using standard plethysmographic techniques in a sealed 10-liter plethysmograph. After 10 min of acclimation, control data (eupnic ventilation, arterial blood pressure, and heart rate) were collected for 5 min for each rat, followed by 10 min of hypoxia [fractional concentration of O2 in inspired air (FiO2) = 0.12] or hypercapnia [fractional concentration of CO2 in inspired air (FiCO2) = 0.07]. Ventilation [total ventilation (V̇e), frequency, tidal volume (VT)], blood pressure, and heart rate were collected throughout the protocol. Arterial blood samples were drawn during the control period and during min 8–9 of the hypoxic exposure. Finally, on a separate day, the rats were exercised on a treadmill for 5 min of walking (0.8 m/m, 5% grade) and 5 min of running (1.8 m/m, 5% grade). Arterial blood samples were drawn after 5 min of rest on the treadmill, during a 5-min walk (0.8 m/m, 5% grade), and during a 5-min run (1.6 m/m, 5% grade).
While animals were under deep anesthesia (sodium pentobarbital, 50 mg/kg ip), hearts were rapidly removed for retrograde perfusion via the aorta using a Langendorff apparatus and perfused with modified Krebs-Henseleit bicarbonate buffer. Left ventricular function was measured continuously throughout the experiment. End diastolic pressure was set to 5 mmHg, and developed pressure was measured during steady-state conditions. Preischemic leakage of lactate dehydrogenase (LDH) and coronary flow rate were measured at the end of the equilibration period. A 25-min period of global ischemia was created by stopping the flow of perfusate to the heart. After ischemia, the hearts were reperfused for 180 min. Eight-week-old rats were studied under normoxic conditions (21% O2) or after 14 days of chronic hypoxia (12% O2).
Under anesthesia (ketamine, 30 mg/kg im, and Inactin, 0.0375 g/kg ip), 10 wk-old rats were instrumented with a tracheal cannula, a carotid catheter, a saline-filled esophageal catheter, and an intraperitoneal catheter. A blood sample (1 ml) was taken via the carotid catheter and put into EDTA (0.4 ml) and serum separator (0.6 ml) tubes for biochemical analysis. The rats were placed in a whole body plethysmograph. Ventilation, air flow, and esophageal pressure measurements during increasing doses of methacholine (0.2–22 mg/kg ip) were used to calculate changes in pulmonary conductance. After the methacholine challenge, the lungs were removed, and tracheal tube and pulmonary arterial cannula were used to ventilate and perfuse the lungs, enabling measurements of pulmonary arterial pressure (as the flow rate was increased to 40 ml/min), calculation of pulmonary vascular resistance, and assessment of pulmonary endothelial cell activity [angiotensin-converting enzyme (ACE) activity and redox state]. Rats were studied under normoxic conditions or after 14 days of exposure to chronic hypoxia (12% O2).
While animals were under deep anesthesia (sodium pentobarbital, 60 mg/kg ip), a 3- to 5-cm segment of the thoracic aorta was removed and divided into 3-mm-wide rings. Two aortic rings from each rat were mounted. Passive tension was adjusted to 1.5 g and equilibrated for 30 min. One aortic ring was used for contractile responses to increasing doses of phenylephrine followed by hypoxic relaxation, and the other ring was used to measure the relaxation responses to increasing doses of acetylcholine and sodium nitroprusside. Twelve-week-old rats were studied on a low-salt diet (0.4% NaCl) or after 3 wk on a high-salt diet (4.0% NaCl).
A blood sample was collected from the carotid artery from rats used in the lung protocol, and a small animal clinical blood panel (ANP-15) and a complete differential blood count were performed by Marshfield Laboratories (Marshfield, WI).
In each group of animals phenotyped, two SS rats were studied in parallel as sentinels. These sentinel animals provided a means to assess continuity in the phenotypes across time. To address potential seasonal effects on the measured phenotypes, data from male SS sentinel rats under normal O2 from four consecutive years were grouped according to seasons (quarters) as follows: quarter 1 (Q1) consists of males born from January 1 through March 31, Q2 consists of males born from April 1 through June 30, Q3 consists of males born from July 1 through September 30, and Q4 consists of males born from October 1 through December 31. Phenotypic data from 88 different traits spanning the four seasons were evaluated for normality and tested for significant differences between the four groups by either conventional ANOVA (normal distribution) or by nonparametric ANOVA (if data were not normally distributed). Traits found to have a significant difference between the groups were subject to a post hoc pairwise test to determine which season(s) significantly differed. Seven percent of the traits (6 of 88 tested) showed significant differences between one season and another (data not shown). Of these, body weight was determined to significantly differ in two protocols (cardiac and vascular); however, the particular seasons showing variation were not consistent. Therefore, for the purpose of this study, we assumed the seasonal variation did not exceed normal variation and did not consider it further in the analyses of the parental data.
As part of the quality control procedure, data outliers were determined before posting of data on the web site. Mean values and standard deviations (SD) of all pooled data were determined, and measures exceeding 3 standard deviations were eliminated from the released data set.
Because all commercial strains are caesarean derived at the distributor, strain authenticity is verified by the genotyping the breeders used to establish each colony.
Web Site Data and Visualization
All phenotype and genotype data are freely available on the PhysGen web site (http://pga.mcw.edu/pga/jsp/data/index.jsp) and released on a quarterly basis. To date, phenotypic measures have been completed and released for all nine inbred and two outbred parental strains. In addition, PhysGen has developed two complete panels of consomic rats, whereby each BN chromosome was introgressed into both the SS and FHH strains (15). Currently, available data include 390,910 mean values and all corresponding raw data. To mine the PhysGen data, we have created a visualization tool to dynamically display strain distribution patterns for each trait (http://pga.mcw.edu/pga/data_status.html), partitioned by protocol (see Supplemental Presentation 1; available at the Physiological Genomics web site).1 To visualize a trait, an investigator can select “Data” on the PhysGen home page and then choose “Phenotype Data Download and Analysis.” On this page, one can choose the protocol of interest, e.g., biochemistry, and select “Visualization and Statistics.” The following web page offers a variety of options to dynamically view strain phenotype distribution patterns. For instance, an investigator could choose to visualize the “Parental Strains,” under “Hypoxic (12% O2)” preconditioning, and “Males,” for the phenotype “plasma cholesterol” and “Do Analysis.” The result is presented as a histogram showing the mean value for each strain, automatically sorted in ascending order. Below the histogram is a table listing additional detail (e.g., mean + SE, N, environment, diet) as well as options for dynamic statistical analysis (described below). Each bar has mouse-over capability that displays the strain name, the environmental and diet conditions, gender, and the mean value. Clicking on the histogram bar returns a pop-up window displaying the distribution of the raw trait values.
Dynamic online tools.
Several statistical tests exist on the web site. These dynamic analysis tools were used for the strain distribution patterns, and can be used for any phenotypic data selected for visualization. Each test is automatically performed on a hierarchical basis depending on a series of decision trees. The first test is for homoscedasticity of the data, using Levene's test. If Levene's test passed, a parametric ANOVA or t-test is used, depending on the number of strains selected for statistical analysis. If a t-test is selected, a parametric test is automatically selected based on the results of Levene's test; if an ANOVA is chosen, results of a conventional ANOVA are provided. For the ANOVA, two post hoc analysis methods can be selected, a pairwise analysis using Tukey's test or comparison with a control strain using Dunnett's test. Correction for multiple testing is accounted for in both tests. If the Levine's test for homoscedasticity fails, a nonparametric analysis is used. The t-test analyses are performed using a Mann-Whitney test; nonparametric ANOVA analyses are performed using the Kruskal-Wallis test. Post hoc analyses for the nonparametric ANOVA include Dunn's test for all pairwise comparisons and a nonparametric Dunnett's test for pairwise comparisons with a selected control strain. Again, each test includes a correction for multiple testing errors.
Parental strain vs. other combined parental strains.
To determine the degree to which an individual strain differs from the other parental rat strains, calculations were carried out for each trait for the 10 parental inbred and 2 outbred strains. For each trait, a nonparametric Mann-Whitney U-test was used to compare the mean value of a single strain with the mean value of all other parental strains combined as a single group. A Bonferroni correction for multiple testing was performed within each protocol (the same group of animals did not progress through multiple protocols). These data can be obtained online by strain by selecting the “Strain Profile” option on the PhysGen home page. A parental strain, e.g., BN, can be selected on the following page to return the profile of the chosen strain. An investigator can select a particular protocol(s) in which the trait(s) was measured and view or download a “Phenotype Report” with significant results from the t-test comparison. Included are the mean values ± SE for the selected strain, mean values ± SE for the other 10 strains combined, and both raw and Bonferroni-adjusted P values. All t-test results can be found in Supplemental Tables B and C (males and females, respectively).
To generate a reference set for phenotypes measured by PhysGen that may be used by investigators working in rat heart, lung, and blood physiology, an extrapolated “population” value for each measured trait was generated by determining the mean and standard deviation of the mean values for all 11 parental strains. This reference is found in Supplemental Table D.
Direct gender comparisons within the clinical chemistry panel were made for each parental strain housed under normal O2 conditions by a Mann-Whitney U-test, followed by a Bonferroni correction for multiple testing.
RESULTS AND DISCUSSION
Characterization of the BN
To characterize the BN that was sequenced with respect to the successfully measured phenotypes, the mean values in the BN strain were compared with the mean trait values of the other 10 strains combined. Table 1 shows the number of traits whose mean value for the BN was significantly different from the 10-strain mean value, and Supplemental Table A contains the details for each of those traits. Combined, male and female BN rats have measures for 529 traits, 258 in male and 271 in females; of those, 39% significantly differ from the 10-strain mean. The breakdown of significant traits by protocol and gender is found in Table 1. Forty-six percent (90/198) of the traits that were measured in the respiratory and lung protocols were significantly different in the BN, which is consistent with the view that the BN is considered to be a sensitive strain for the study of asthma and other inflammatory lung diseases (2, 9, 14). BN males appear more prone to lung disease than female BN rats, as 54% of the significant traits in males involve respiratory and lung function compared with 38% in females. In contrast to the respiratory and lung protocol, only 15% of the traits measured in the renal protocol (Renal_A) significantly differed in the BN vs. the 10-strain mean values. Therefore, while the BN could be considered a good disease model for respiratory and lung disease, it could be considered as an appropriate control strain for studying renal disease. In other words, the role of a specific strain, such as the BN, in studying disease highly depends on its baseline physiology relative to other strains and whether the model is more or less sensitive to the disease process than the range of responses seen across all other strains.
To address the effect of stressors on the measured traits in the BN, we determined which phenotypes showed a significant difference from the 10-strain mean under varying levels of environmental O2 (12 vs. 21% O2 exposure for 14 days before measurement) and dietary salt (0.4 vs. 4.0% exposure for 3 wk before measurement). Of the 38 significant traits having atmospheric O2 as a variable in male BN, 12 (32%) were significant in both conditions, indicating that atmospheric O2 did not affect these phenotypes (Table 2). However, 15 traits (39%) were significant only under 12% O2 (hypoxia), while 11 traits (29%) were significant only under 21% O2 (normoxia). Interestingly, after 2 wk of hypoxic preconditioning in the BN, hematocrit and hemoglobin levels do not increase to the same level as seen in other strains, indicating that the BN may suffer an impaired ability to increase erythropoiesis in response to chronic hypoxia, the typical response (7) to prevent anemia and provide sufficient O2 to tissues under oxygen stress. Alternatively, it is possible that the BN strain shows better ventilatory compensation and is not subjected to as much hypoxic stress as other strains, or has a greater decrease in metabolic rate under hypoxic conditions, thus reducing the need for additional O2.
Of the 19 significant traits in the vascular protocol having dietary salt as a variable, 11 (58%) were significantly different on both diets, meaning dietary salt has no effect on these phenotypes. Four traits (21%) in the vascular protocol were significant only on a low-salt or a high-salt diet, respectively (data not shown).
Variability in the phenotypes between the BN and the other parental strains likely indicates the genetic variability for each trait, as the experimental conditions for each strain are essentially constant. To more easily visualize the phenotypic data and identify their genetic variability, we have developed an online dynamic visualization tool (http://pga.mcw.edu/pga/data_status.html) so that each trait can be visualized across all 11 strains (i.e., display a strain distribution for each trait). We also developed corresponding dynamic statistical analysis tools to determine significant differences between the strains. For example, one protocol (cardiac) involves an ex vivo measurement of heart tissue damage after 25 min of global ischemia followed by 180 min of reperfusion in a Langendorff heart preparation. Figure 2A displays the online visualization of the mean trait value distribution and statistical analysis for the area of the left ventricle damaged after ischemia/reperfusion in male rats across all 11 strains, under normal environmental conditions. The wide range of ischemic injury among the strains strongly suggests a genetic component in susceptibility or resistance to ischemia/reperfusion injury. For example, while only 3% of the BN left ventricle is damaged, >27% of the SS heart is damaged, as has been previously reported (3). Three additional strains (SHR, LEW, and SD) also show significantly more ischemic injury than the BN. One could speculate that these strains share a common genetic factor(s) predisposing them to susceptibility to ischemia/reperfusion.
The genetic differences between BN and SS for infarct size, as well as any other traits that significantly differ between these strains, can also be mapped using the consomic rat panel (4, 15, 23) generated by our group, in which each BN chromosome was systematically introgressed onto the SS genome. The consomic rat panel is scheduled for phenotype completion and release later this year. All phenotypic data from the characterized consomic strains are also released on the PhysGen web site. Because each consomic strain is characterized for the same phenotypes as the BN and SS parental strains, one can map the measured traits in silico. Furthermore, the ability to download the complete phenotypic data set allows for correlation studies between the measured traits, to identify primary and intermediate phenotypes.
To identify chromosomes that play a role in a trait that differs between BN and SS using the PhysGen consomic panel, one can visualize and perform statistical analyses in same manner as described for the parental strains (Fig. 2B). For example, in the case of the protection against ischemia/reperfusion injury, evaluation of the available chromosome substitution strains (all but SS-1BN) indicates that, in male rats raised in normal O2 and on a low-salt diet, substitution of some SS chromosomes with corresponding chromosomes of the BN have no effect on the percentage of the heart damaged by global ischemia. However, nearly complete protection against the ischemia/reperfusion injury was conferred by either BN chromosome 2 or 6 when introgressed on the SS genome background. Interestingly, there are no significant differences in mean arterial pressure between SS-2BN, SS-6BN, and their parental derivatives on a low-salt diet and normal O2 (from the respiratory protocol), suggesting that the resistance to ischemia/reperfusion damage is independent of blood pressure in these consomic strains. These results cannot rule out the possibility that the resistance to ischemia/reperfusion injury is secondary to another related trait that also maps to chromosomes 2 and/or 6, or that a BN allele on these chromosomes modifies the phenotype through interaction with a gene(s) on the SS background. However, use of the consomic strains to generate multiple subcongenic strains can address these issues by mapping the phenotype to a chromosomal region small enough for traditional positional cloning experiments. Furthermore, the large number of related phenotypes should allow for the determination of intermediate phenotypes.
In addition to visualizing the distribution of mean trait values across all parental strains using the dynamic online visualization and analysis tools, we also generated a strain report that enables an investigator to identify traits that are significantly different from the population mean in the BN strain, as well as the other 10 parental strains, thus enabling the determination of additional correlated cardiovascular or pulmonary traits. For each strain, a list of significantly different traits can be obtained online in a Strain Profile at http://pga.mcw.edu/pga-bin/strain_profile.cgi, with links to graphical visualization and statistical analyses for each trait. These data can also be directly downloaded as mean and raw data values.
Characterization of Other Commonly Studied Strains
In addition to the analyses of BN described above, we performed statistical analyses of each of the other parental strain traits, comparing their trait values with their corresponding 10-strain trait values. Table 3 summarizes the traits for each parental strain whose values are significantly different from the 10-strain mean, with detailed trait and phenotype information found in Supplemental Tables B and C (males and females, respectively). On average, ∼20% of the traits differ significantly in any one strain. The BN strain has the highest number of differences among all the strains tested. This may not be surprising, given that the BN strain is the most genetically distinct of all laboratory rat strains (27). What is surprising is that the number of traits significantly different in the outbred strains, considered a “normal” rat, is not dramatically different from the inbred strains. In male SD and CD®IGS, 18.2 and 16.1% of measured traits were significantly different, respectively. In comparison, only 13.5% (35/259) of measured traits differed in the inbred CDF, suggesting that outbred strains may not better represent the range of normal values than inbred strains. Moreover, characterization of a value as normal in an outbred rat depends on the outbred strain and the trait being measured. For example, in the distribution of the 29 blood biochemistry values measured in males under a normal O2 environment, the SD ranks as the lowest or highest of all the parental strains for 17 of the 29 traits (59%) (data not shown) and differs significantly from the 10-strain mean for 11 of the 29 blood traits (38%) (Supplemental Table B), suggesting that this strain is not representative of a normal rat for many of the blood phenotypes. Conversely, in the male CD®IGS, another outbred source of the “same” strain, only plasma segmented neutrophil levels significantly differ (under normal O2), indicating the presence of dramatic differences in two strains that are often interchangeably used as controls for physiological and pharmacological studies. Because these substrains (from 2 different suppliers) are outbred, we cannot distinguish whether the lack of overlap is due to substrain differences within the Sprague Dawley strain or because the two groups capture a different collection of genotypes. However, we consider the former to be more likely. Collectively, these data suggest that it is unlikely that any one strain can be classified as a normal rat for all traits, even with the somewhat increased genetic diversity in outbred strains. Therefore, use of a single inbred or outbred strain for physiological and pharmacological studies is akin to determining risk for common human diseases without considering the epidemiology of different populations or ethnic groups. Because such studies represent a limited sample of diversity in rats, it is not surprising that the results of such studies may have little or no predictive value in determining the range of responses that might be seen in human population studies.
The PhysGen Program also allowed us to compare physiological measures with respect to differences in gender and environmental O2 across the 11 parental strains, again comparing each strain with the remaining 10 strains combined. For example, it is rather surprising that 40% of the significantly different traits are significant in only one gender in BN (Table 2). The intragender strain distributions are even more dramatic in FHH, WKY, and CDF, where nearly 72% of the traits are significantly different in only one gender. On average, nearly 60% of the traits that significantly differ are gender specific, given an identical genetic background and experimental conditions. Because some of the protocols involved a 2-wk chronic preconditioning period in both genders under either normoxia or hypoxia, we were able to evaluate not only gender effects but also the physiological responses to differing O2 across the strains. Several traits were found to have a strain-dependent response here as well. For instance, plasma cholesterol levels were measured in both genders and after both normoxic and hypoxic preconditioning. The strain distribution in males subject to normal O2 is shown in Fig. 3. The strains that significantly differ from the BN male (dark gray bar) are highlighted in pink. When comparing cholesterol levels between the genders, across different strains raised in a normoxic environment, one can also determine that females of several strains (CDF, CD®IGS, LEW, SS, WKY, LE/BluGill) have significantly higher plasma cholesterol levels compared with males (Fig. 3). Furthermore, the gender differences are strain dependent. For instance, while cholesterol levels do not differ between male BN and CDF (P = 0.1716), they do significantly differ between female BN and CDF (P < 0.0001). When comparing males from the same strains after 2 wk of chronic hypoxia (12% O2), some strains show a significant decrease in plasma cholesterol levels in response to hypoxic preconditioning, whereas others do not. Furthermore, the extent of this decrease varies between the strains. For example, while BN shows virtually no response, there is nearly a 30-mg/dl decrease in cholesterol levels in the WKY in response to chronic hypoxia. These data suggest that two different “environments” (gender and environmental oxygen content) have strain-dependent effects on lipid metabolism. Therefore, before making global interpretations of physiological measurements, one must always be mindful that, not only can the genome background significantly affect baseline measurements, but different responses to gender and environmental stimuli can also depend on the different genome backgrounds.
A direct gender comparison (i.e., comparing mean values from males and females of the same strain) was performed for a panel of blood chemistries measured in each strain to determine whether different strains displayed different gender differences. Interestingly, the number of traits that displayed significant gender differences varied greatly, depending on the strain being evaluated (Fig. 4). Although not representative of a normal rat for the blood traits, none of the mean values of those traits differed significantly between males and females in the SD. At the other end of the spectrum, in the SS (SS/NHsdMcwi), 64% of these traits differed between the genders. This suggests that gender differences are affected by genome backgrounds that vary among the strains. It should be noted that, while the outbred SD had no differences between the genders, >24% of the blood traits differed between the genders in the other outbred strain, the CD®IGS. Gender differences have been determined in humans, as well, for many of the same blood chemistry measurements such as plasma alanine aminotransferase (ALT) and aspartate aminotransferase (AST) (21). These results suggest that different human populations may also require different normal reference measures for males and females (22).
A Reference Standard for the Laboratory Rat
An unintended, but powerful, outcome of our study is a comprehensive collection of cardiovascular and pulmonary measurements related to the laboratory rat. We used the physiological values across all 11 strains to extrapolate population values for >281 traits in male and female rats (Supplemental Table D). We believe this derived population mean from these strains is more representative of a range of normal values for the rat than values obtained in any single inbred or outbred strain, as it captures a wider spectrum of both phenotypic and genotypic diversity (10). The measures in Supplemental Table D provide data needed to benchmark other strains relative to both the population and the sequenced strain. These data facilitate the selection of strains for physiological and pharmacological studies related to the heart, lung, and blood traits studied here. Most importantly, this data set is of tremendous value for investigators who wish to integrate the “sequenced” rat into their experimental protocols, as it provides detailed baseline measures for the sequenced BN/NHsdMcwi rat compared with 10 commonly used strains representing the majority of the known genetic diversity of the laboratory rat.
As inbred strains are developed, multiple genes conferring disease may be concurrently fixed, resulting in multiple disease models within a single inbred strain, albeit some of these traits may remain unidentified. Because of this, there is a need to better characterize strains, at both the phenotypic and the genomic levels, i.e., generate a comprehensive rat phenome resource. Major efforts are focusing on generating a rat phenome. Our PhysGen Program has characterized 11 different strains (9 inbred and 2 outbred) for >280 different traits in the two genders (corresponding to >560 total traits), as reported here. Furthermore, we have generated and characterized two chromosome substitution panels (44 strains derived from the sequenced BN and the FHH and SS hypertensive strains) (15). These two panels of consomic strains can be used for genetic mapping of phenotypes from the sequenced BN in silico. Each consomic strain is characterized using the same phenotypes and experimental conditions. The data and analytical tools at the PhysGen web site provide the research community at large with the opportunity to mine rat physiology and genomics for many common heart, lung, and blood traits with respect to genetic background, gender, and environmental stressors, and to map many of these traits to a chromosome resolution in silico using the chromosome substitution panels, all in the context of the sequenced rat. In addition, the National Bio Resource Project (NBRP) has characterized 109 traits in 54 inbred rat strains (16) for some of the same biochemical and hematological traits as well as baseline blood pressures, and for additional traits relating to behavior, urology, and anatomy (16). The overlap allows these strains to also be directly compared with the sequenced BN rat and extends the publicly available resources for physiological measurements of the rat.
To complement the detailed phenotypes generated by these efforts, alleles of 48 common inbred strains have been determined for 4,328 simple sequence length polymorphisms (SSLPs) spanning the rat genome, as part of the United States Rat Genome Project (25). Furthermore, the NBRP has determined 357 SSLP genotypes in 98 strains, including the 54 strains from their Rat Phenome Project (16). These data allow the construction of haplotypes across all major rat strains using publicly available tools such as the ACP Haplotyper (http://rgd.mcw.edu/ACPHAPLOTYPER/) to identify common haplotypes within models with similar diseases. From these data, one can determine the “evolutionary” relatedness of the various inbred strains of rats (27). These allele data are now being greatly supplemented by the addition of >45,000 single nucleotide polymorphisms (SNPs) identified across multiple rat strains (http://www.ncbi.nlm.nih.gov/SNP/snp_summary.cgi) (13, 29). Furthermore, a growing number of rat strains are or will be resequenced for SNP discovery (1, 5, 24, 29) by two major SNP discovery projects underway in Europe (the Functional Genomics Group in the Netherlands and the Max-Delbruck-Center for Molecular Medicine in Germany) and by Baylor College of Medicine (Houston, TX) via a recent National Human Genome Research Institute white paper to identify SNPs in an additional eight strains. In all, 11 strains, including 7 of the parental strains (or substrains thereof) we present here (F344, FHH, LEW, SD, SS, SHR, and WKY), are being screened or proposed for SNP discovery, with the expected identification of >485,000 SNPs (1). Of note, the FHH and SS strains will be from the same source as the phenotypes presented here, further expanding the rat phenome with phenotypes directly attached to genome sequence. This genomic resource, combined with the large BN phenotypic data presented here, will provide a powerful data set for new types of analyses to identify genotype-phenotype relationships and environmental effects on clinical outcomes.
Physiological characterization of the parental rat strains has indicated that genome backgrounds, gender, and environment can have a significant impact on phenotypic outcome. Although the gene environment effects on disease are certainly not a new concept, PhysGen has provided one of the largest sets of rat data to test these interactions in cardiovascular and pulmonary disease. Additional concerted efforts for strain characterization and data integration by the NBRP (17) and the Mouse Phenome Database (6) are powerfully controlled data sets, with large-scale systematic characterization of inbred organisms including some common cardiovascular traits. Importantly, these strain characterization resources are being linked with the SNP resources in mouse and rat. The dramatic differences due to genome backgrounds and environmental conditions are an important discovery from these phenome projects that can be extrapolated to the clinical situation in complex disease. Given the phenotypic results in different inbred strains, the level of heterogeneity in human clinical populations for complex disease (both at the genotypic and etiological levels) is not surprising. Therefore, careful population stratification may be critical to dissecting the genetic and etiological components of multifactorial disease. The animal phenome projects should lead to new approaches toward advancing personalized medicine, by better understanding how genetic variation can affect disease outcome and drug therapy and, perhaps, even lending predictions to pathways that control complex biological systems.
This project was supported by The Programs for Genomic Applications, National Heart, Lung, and Blood Institute Grant HL-066579.
H. J. Jacob and R. J. Roman are co-founders and board members of PhysioGenix, Incorporated, which has a commercial license for the consomic rat strains. PhysioGenix and the Medical College of Wisconsin receive royalties for sales from Charles River Laboratories to for-profit organizations.
We acknowledge members of the PhysGen Program, as follows, by their role in the program. Project Coordinator: Julia Jursinic. Genomics Component: Sheri Jene, Lisa Groth, Jody Klingkammer, Mary Kaldunsky, Roberta Rogge, Rebecca Majewski, Kathleen Kennedy, Mike Tschannen, Jaime Wendt-Andre, Mary Siegrist, Angela Lemke, Nadia Barreto, and Kristin Fritsch. Phenotyping Component: Renal Protocol, Michael Bregantini, Jess Powlas, Angelo Piro, Erica Liss, and Sarah Kaplan; Cardiac Protocol, Robert Beauvais, Kristin Berg, Janelle Curran, William Hutchins, and Jessica Laessig; Vascular Protocol, Kathryn Privett, Alison Kriegel, Julie Antczak, Janelle Yarina, Julie Holding, Jennifer Labecki, and Ashley Merritt; Lung Protocol, Cynthia Maas, Michael Kelm, Hart Moss, Nong Xiang, Tracy Enslow, and Bernadette Cabigas; Respiratory Protocol, Genevieve Hogan, Andrea Trevett, Lisa Gottschalk, Jaime Petersen, and Amy Rider; Biochemistry/Histology, Candace Jones, Anne Ansley, Zdravko Peric, Becky Bralich, Sherry Hahn, Austin Brill, and Glen Slocum; Research Services and Data Management, Greg McQuestion, John Govin, Dave Eick, and Mike Kloehn. Bioinformatics Component: Zhanchi Wang, Yan Wang, Yu Han, Jeff Nie, Dawei Li, Nan Jiang, Hao Jiang, Zhitao Wang, and Mike Chen. Education and Outreach: Sandra Grieger.
↵1 The Supplemental Material for this article (Supplemental Tables A–D and Supplemental Presentation 1) is available online at http://physiolgenomics.physiology.org/cgi/content/full/00288.2005/DC1.
Article published online before print. See web site for date of publication (http://physiolgenomics.physiology.org).
Address for reprint requests and other correspondence: A. E. Kwitek, Human and Molecular Genetics Center, Medical College of Wisconsin, 8701 Watertown Plank Rd., Milwaukee, WI 53226 (e-mail:).
- Copyright © 2006 the American Physiological Society