QTL influencing baseline hematocrit in the C57BL/6J and DBA/2J lineage: age-related effects Frank Johannes,1,2 David A. Blizard,1 Arimantas Lionikas,1 Dena H. Lang,3 David J. Vandenbergh,1,2 Joseph T. Stout,1 James A. Strauss,4 Gerald E. McClearn,1,2 George P. Vogler1,2 1

Center for Developmental and Health Genetics, The Pennsylvania State University, University Park, Pennsylvania 16803, USA Department of Biobehavioral Health, The Pennsylvania State University, University Park, Pennsylvania 16803, USA 3 Center for Locomotion Studies, The Pennsylvania State University, University Park, Pennsylvania 16803, USA 4 Department of Biology, The Pennsylvania State University, University Park, Pennsylvania 16803, USA 2

Received: 19 January 2006 / Accepted: 28 February 2006

Abstract

Introduction

Baseline serum hematocrit varies substantially in the population. While additive genetic factors account for a large part of this variability, little is known about the genetic architecture underlying the trait. Because hematocrit levels vary with age, it is plausible that quantitative trait loci (QTL) that influence the phenotype also show an age-specific profile. To investigate this possibility, hematocrit was measured in three different age cohorts of mice (150, 450, and 750 days) of the C57BL/6J (B6) and the DBA2/J (D2) lineage. QTL were searched in the B6D2F2 intercross and the BXD recombinant inbred (RI) strains. The effects of these QTL were explored across the different age groups. On the phenotypic level, baseline serum hematocrit declines with age in a sex-specific manner. In the B6D2F2 intercross, suggestive QTL that influence the phenotype were located on Chromosomes (Chr) 1, 2, 7, 11, 13, and 16. With the exception of the QTL on Chr 2, all of these QTL exerted their largest effect at 750 days. The QTL on Chr 1, 2, 7, 11 and 16 were confirmed in the BXD RIs in a sex- and age-specific manner. Linkage analysis in the BXD RIs revealed an additional significant QTL on Chr 19. Baseline serum hematocrit is influenced by several QTL that appear to vary with the age and sex of the animal. These QTL primarily overlap with QTL that have been shown to regulate hematopoietic stem cell phenotypes.

Steady-state serum hematocrit is the result of complex physiologic processes that coordinate plasma volume and red blood cell mass (Donelly 2003). Normal maintenance of hematocrit is a crucial aspect of the general health of the organism. Despite its tight physiologic regulation, substantial variability in baseline hematocrit exists at the population level (Fulwood et al. 1980). Heritability studies in both animals (Stino and Washburn 1973; Weibust and Schlager 1968) and humans (Evans et al. 1999) have shown that additive genetic factors account for 20% 60% of this variability, suggesting that the phenotype is under polygenic control. This is supported by selection experiments for high and low hematocrit in mice, which have shown that several generations of selected breeding are required before the trait values of the extreme lines begin to plateau (Schlager and Weibust 1976; Stino and Washburn 1973). Attempts to characterize the genetic architecture underlying the trait using QTL methods so far have been limited to two studies. A study by Pravenec et al. (1997) demonstrated significant linkage of a QTL to the enolase 2 (Eno2) marker on Chromosome (Chr) 4 of the rat genome. A significant QTL for hematocrit was also mapped to the telomeric region of Chr 1 in the rabbit (Van Hearingen et al. 2002). One complication in the mapping of QTL for hematocrit relates to the fact that the phenotype undergoes changes throughout the life course of the organism (Heller et al. 1998; Strauss 1992), which raises the possibility that the genetic architecture underlying the trait also follows an age-dependent profile. This can result in mapping different sets of QTL depending on the age structure of the mapping

Correspondence to: Frank Johannes, Department of Biobehavioral Health, Center for Developmental and Health Genetics, The Pennsylvania State University, 101 Amy Gardner House, University Park, PA 16803, USA; E-mail: [email protected]

DOI: 10.1007/s00335-006-0009-7  Volume 17, 689 699 (2006)  Ó Springer Science+Business Media, Inc. 2006

689

690

F. JOHANNES ET AL.: AGE-RELATED QTL INFLUENCING HEMATOCRIT

population at hand. The above-mentioned QTL analyses of hematocrit could not attend to this temporal problem because the studies were limited to a single age cohort of animals. However, evidence for important age-related genetic changes in the regulation of hematocrit was noted very early by Russell and McFarland (1966) in an analysis of several genetic mutations and their effects on hematocrit levels during postnatal development of the mouse. These analyses demonstrated that the effect of the mutant alleles was not constant over time but manifested itself only during the first few weeks of development and thereafter completely disappeared. More recently, Henckaerts et al. (2002a, 2002b) conducted a QTL analysis of hematopoietic stem cell number, frequency, and proliferation capacity in response to early-acting cytokines in the BXD recombinant inbred (RI) strains. They identified several QTL in 8-week-old mice but were unable to map the same QTL in 18-month-old mice (Henckaerts et al. 2002b). Henckaert et al. concluded that different sets of genes underlying the same phenotype must be active at different points during the life process of the organism. Similar age-related hematologic changes have been observed in two inbred mouse strains C57BL/6J and DBA/2J (Geiger et al. 2001; Henckaerts et al. 2002b, 2004; Van Zant et al. 1990). Geiger et al (2001), for example, demonstrated that the proliferation capacity of early hematopoietic progenitor cells (eHPCs) is generally higher in DBA/2J than in C57BL/6J at 2 months of age but substantially lower at 20 months of age. Because the genetic processes regulating the continued renewal and differentiation of hematopoietic stem cells into mature red blood cells comprise a crucial aspect of steady-state hematocrit maintenance, we expected QTL for hematocrit to partially overlap with those that influence hematopoietic stem cell parameters. However, we also anticipated additional QTL that influence other physiologic aspects of hematocrit regulation. We were particularly interested in exploring the agedependent patterns of the effects of these QTL. A detailed characterization of the genetic changes throughout the life course of the organism may contribute to a better understanding of the molecular regulation of the trait during aging.

Table 1. Mean ages (± SD) and sample sizes at each

Materials and methods Animals. This study used two populations of mice, a B6D2F2 intercross and 23 BXD RI strains, both derived from the C57BL/6J (B6) and the DBA/2J (D2) strains. Each population consisted of three age cohorts, which were part of a larger developmental

measurement of hematocrit Animals (cohort)

Age (days) at measurement

N

B6D2F2 B6D2F2 B6D2F2 BXD RI BXD RI BXD RI

148.11 453.04 741.31 159.63 462.24 740.88

375 397 364 22a [22; 9 24] 22a [18; 10 24] 21b [13; 6 30]

(150 days) (450 days) (750 days) (150 days) (450 days) (750 days)

± ± ± ± ± ±

9.92 8.46 11.71 5.04 4.84 26.97

a

Strain 13 was excluded because of premature mortality. Strains 13 and 33 were excluded because of premature mortality. The numbers in brackets denote [mean number of animals per age group; range of numbers of animals in age groups] available for genotyping. b

study investigating the physiologic, neurobiologic, and behavioral aspects of aging. The mean ages and sample sizes can be found in Table 1. Animal breeding and maintenance were performed in the barrier colony developed by the Center for Developmental and Health Genetics (CDHG), at The Pennsylvania State University (for details about animal housing and maintenance, see Lionikas et al. 2003). All procedures were approved by the Pennsylvania State University Institutional Animal Care and Use Committee. Hematocrit measures. Whole blood was collected directly from a tail nick in heparinized microcapillary tubes (50 ll). The volume of blood collected was approximately 30 45 ll. Both ends of the microcapillary tube were sealed with clay, and the tube was spun for 3 min in an International microcapillary centrifuge (Model MB). The hematocrit was read on a Clay Adams Micro-Hematocrit Reader. For the B6D2F2 population, hematocrit was assessed three times within each of the 150- and 450-day-old cohorts and was assessed twice in the 750-day-old cohort, with measurements separated by four-week intervals. The average of these measures was used for all subsequent phenotypic and genetic analyses involving the B6D2F2 animals. For the BXD RI population, hematocrit was assessed only once within each age cohort. Genotyping. The B6D2F2 animals were genotyped at 96 microsatellite markers distributed throughout the autosomes and the X chromosome at an average spacing of 15 20 cM. Genotyping was performed as described by Vandenbergh et al. (2003). For the BXD RI strains, the Wellcome-CTC Mouse Strain SNP Genotype data set (build 34) was used for mapping analyses (http://www.well.ox.

691

F. JOHANNES ET AL.: AGE-RELATED QTL INFLUENCING HEMATOCRIT

ac.uk/mouse/INBREDS). The final single nucleotide polymorphism (SNP) map consisted of 5773 SNPs distributed throughout the BXD RI genome. Statistical analysis Phenotypic analyses. Descriptive and inferential statistics for hematocrit were obtained for the two parental strains B6 and D2, the BXD Ris, and the B6D2F2 intercross. Emphasis was placed on the differences in hematocrit levels as a function of age. QTL identification. In the B6D2F2 population (males and females combined), a one-dimensional genome-wide search was performed using interval mapping (Lander and Botstein 1989). Three different search models were initially used. Model 1 was a search model that included sex and age as additive and interacting covariates. Model 2 included only sex as an additive and interacting covariate. Model 3 included no additional covariates. Significant thresholds were derived empirically from 1000 permutations of the data (Churchill and Doerge 1994), resulting in a 5.97-LOD threshold for ‘‘significant’’ QTL (>95th percentile) and a 4.03-LOD threshold for ‘‘suggestive’’ QTL (>63rd percentile). All significant and/or suggestive QTL were then used in a multiple regression model to assess their independent effects on hematocrit and their interactions with sex and age. QTL confirmation. To confirm the QTL that were identified in the B6D2F2 intercross, the NCBI database (http://www.ncbi.nlm.nih.gov) (build 34.1) was accessed to identify genomic markers at either end of the ± 1-LOD confidence interval (CI) surrounding each of the B6D2F2 QTL peaks. The physical location of these markers was determined, and all of the BXD SNPs within this physical region were searched for linkage with the putative QTL via t tests. Because this confirmatory analysis in the BXD RIs relied on specific expectations of QTL positions based on the F2 mapping results, genomewide significant thresholds were no longer applicable. Similarly, standard procedures for repeated testing, such as Bonferoni adjustments, were also deemed too conservative because of the high density of the SNP marker map and the subsequent lack of independence between tests. To solve this problem, a ‘‘localized’’ version of the traditional genome-wide permutation approach was taken. This involved permuting the phenotypic data n times and performing, on each permuted data set, repeated t tests for all SNPs contained within the ± 1-LOD CI surrounding the B6D2F2 QTL peaks. Thus, similar to

Fig. 1. Mean hematocrit values for the parental strains (B6 and D2) in each age group; *p < 0.05, **p < 0.01.

the genome-wide permutation approach, this method generates an approximate null distribution of no linkage between any of the SNPs in the ± 1-LOD confidence window and the phenotype. The resulting significant thresholds are naturally lower than the genome-wide threshold because it is less likely to observe a significant t statistic by chance in only a small part of the genome as opposed to the whole genome. All analyses were performed with the statistical program R (R Development Core Team 2005) and its library package Rqtl (Broman et al. 2000). Results Parental strains. Mean hematocrit (± SEM) was compared between the B6 and D2 parental strains at 150, 450, and 750 days of age (Fig. 1). A highly significant strain-by-age interaction was noted (F2,121 = 5.65, I = 0.0045). At 150 days of age, the B6s had lower hematocrit compared with the D2s [42.83% (0.97) vs. 44.87% (0.3), t = 3.85, df = 45, p = 0.05]. This trend was reversed at 450 days of age at which point the B6s had significantly higher hematocrit than the D2s [44.83% (0.97) vs. 42.05% (0.57), t = 11.96, df = 41, p = 0.001]. No significant differences were found at 750 days (t = 2.93, df = 35, p = 0.09). B6D2F2 intercross. In the B6D2F2 intercross, mean (± SEM) hematocrit levels for males at 150, 450, and 750 days of age were 46.6% (0.06), 43.6% (0.08), and 43.4% (0.12), respectively. For females, they were 45.6% (0.06), 43.9% (0.07), and 44.2% (0.12), respectively. Results showed that hematocrit levels decline significantly with age for both sexes (F2,1120 = 155.6, p < 0.0001). In addition, this agerelated decline had significant interactions with sex (F1,1120 = 13.5, p < 0.0001) (Fig. 2). At 150 days of age, males had significantly higher hematocrit compared

692

F. JOHANNES ET AL.: AGE-RELATED QTL INFLUENCING HEMATOCRIT

Fig. 2. Decline in mean hematocrit values (± SEM) in the F2s and RIs with age. Males, filled symbols; females, opened symbols.

with females (t = )5.51, p < 0.001). This trend was reversed at 750 days, with females displaying higher hematocrit compared with males (t = 1.95, p = 0.05), (Fig. 2). BXD Ris. A similar, significant age-dependent decline was also noted in the BXD RIs (Fig. 2), with hematocrit declining from 43.95% (150 days) to 39.74% (750 days) in males (F2,121 = 7.13, p = 0.002), and from 44.12% (150 days) to 40.33% (750 days) in females (F2,121 = 12.23, p < 0.0001). However, no significant interactions between sex and age were detected in the BXD RIs (F2,121 = 0.05, p = 0.95). Heritability estimates. The significant mean differences in hematocrit between the two parental strains indicate that allelic variability between the strains accounts for variability in the trait. Additional evidence was obtained from the strain distribution patterns (SDP) of the BXD RIs (Fig. 3). Mean hematocrit was as low as 35% in strain BXD 2 at 750 days of age, and as high as 50% in strain BXD 30 at 150 days of age. A one-way analysis of variance (ANOVA) by strain (males and females combined) was used to estimate the heritability of hematocrit (Plomin and McClearn 1993). Heritability estimates were 0.38 at 150 days, 0.41 at 450 days, and 0.24 at 750 days of age. These results are generally consistent with previous research (Stino and Washburn

1973; Weibust and Schlager 1968). Differences in these estimates across age cohorts were also found to be significant (F2,121 = 5.81, p < 0.0001). QTL identification. The first step in identifying QTL for baseline hematocrit involved the B6D2F2 intercross. The interval mapping LOD profiles for Models 1, 2, and 3 are shown in Fig. 4. Because age appeared to be an important factor in the mapping of QTL for hematocrit, the mapping results obtained with Model 1 were used for all subsequent analyses. Suggestive QTL were found on Chr 1, 2, 7, 11, 13, and 16 (Table 2). It is noteworthy that the parental alleles generally associated with increased hematocrit values were not consistently the same across age cohorts (Table 2). To assess the relative effects of the identified QTL and their interactions with sex and age, a multiple regression model was estimated that contains all QTL main effects and their interaction terms with sex and age. Then, a backward stepwise model selection approach was taken; it involved sequentially dropping nonsignificant covariate terms (at p > 0.05) from the model until the final model contained only significant terms (at p < 0.05). The multiple regression results can be viewed in Table 3. Independent main effects were found for QTL 1 (Chr 1), QTL 2 (Chr 2), QTL 4 (Chr 11), and QTL 5 (Chr 13). Although QTL 3 (Chr 7) and QTL 6 (Chr 16) had no detectable main effects on the phenotype,

Fig. 3. Strain distribution patterns (SDP) for hematocrit in each age cohort, triangles point to parental strains. Data for strain 13 (450 and 750 days), strain 33 (750 days), and strain 14 (750 days) are missing because of premature mortality.

693

F. JOHANNES ET AL.: AGE-RELATED QTL INFLUENCING HEMATOCRIT

Fig. 4. LOD profile for genome-wide

search for QTL in B6D2F2; Model 1: sex and age as additive and interacting covariates (dotted line); Model 2: sex as additive and interacting covariate (solid thick line); Model 3: no additional covariates (solid thin line); Horizontal line at LOD = 4 represents the empirically derived threshold for ‘‘suggestive’’ QTL based on Model 1.

they interacted significantly with age. QTL interactions with age were also found for QTL 5. The total regression model explained approximately 20% of the variance in baseline hematocrit. The total QTL effects (including main effects and their interaction effects with age) accounted for roughly 11% of the total variance, which amounts to 25% 40% of the observed genetic variability underlying the trait (Total QTL effect * 100/Estimated heritability). B6D2F2 QTL effects across age. The regression analysis revealed that approximately one third of the total QTL effects were accounted for by QTL interactions with age. Figure 5 illustrates the behavior of the B6D2F2 QTL across the different age cohorts. QTL 1, 3, 4, 5, and 6 appear to be primarily ‘‘lateacting’’ QTL, because their effects are largest in the 750-day age cohort. QTL 2 exerts its largest effect in the 450-day age cohort. Interestingly, the direction of the effect of the different marker classes with respect to each other is not constant over time. This is most apparent for QTL 3, where the D2 marker genotype decreases hematocrit in the 150- and 450-day-old animals but increases the trait at 750 days. Similar patterns were observed for QTL 1, 4, and 6. In addi-

tion, the heterozygote marker genotypes (B6D2) for QTL 1, 3, and 4 show clear overdominance in the 750-day-old age cohort, insofar that hematocrit values in this marker class are substantially lower compared with either of the homozygotes. QTL confirmation. The QTL that were identified in the B6D2F2 intercross reached only ‘‘suggestive’’ significance levels and therefore had to be treated with caution. To provide supporting evidence for these QTL, a confirmatory analysis was performed in a separate experiment using BXD RI strains. Figure 6 shows BXD RI whole-genome scans for each sex and age cohort separately. The vertical lines mark of the approximate 1-LOD CI surrounding the QTL identified in the B6D2F2 intercross. Several BXD RI QTL peaks fall within these CIs. Significant associations with SNPs were observed in the 1-LOD confidence regions surrounding QTL 1 (Chr 1), 2 (Chr 2), 3 (Chr 7), 4 (Chr 11), and 6 (Chr 16) (Table 4). QTL 1 is the only QTL that was consistently confirmed across sex and age cohorts in the BXD RIs. The most significant SNPs confirming QTL 1 in the females are slightly different across the three age

Table 2. Interval mapping search results based on Model 1

Label QTL QTL QTL QTL QTL QTL

Chr Markera

1 1 2 2 3 7 4 11 5 13 6 16

Increasing alleleb Approx. position (Mbp) Peak (cM) ± 1-LOD CI (cM) Peak LOD 150 days 450 days 750 days

D1MIT87 114 D2MIT343 168 D7MIT253 109 D11MIT227 17 D13MIT125 76 D16MIT71 41

CI = confidence interval. a Marker most closely linked the putative QTL. b Average allelic action across sex.

68.0 82.0 41.4 13.0 30.0 70.7

54.0 50.0 34.4 4.00 24.0 66.3

79.0 99.0 49.4 43.0 38.0 77.7

5.2 4.5 4.9 4.1 4.5 4.0

D2 B6 B6 D2 D2 D2

D2 B6 B6 D2 D2 = B6 B6

D2 B6 D2 D2 B6 B6

694

F. JOHANNES ET AL.: AGE-RELATED QTL INFLUENCING HEMATOCRIT

Table 3. Multiple regression results using the B6D2F2 QTL, age, and sex as covariates

Variable Age and sex effects age sex age*sex Variance accounted QTL main effects QTL 1 (Chr 1) QTL 2 (Chr 2) QTL 3 (Chr 7) QTL 4 (Chr 11) QTL 5 (Chr 13) QTL 6 (Chr 16) Variance accounted QTL interaction effects QTL 3 (Chr 7)*age QTL 5 (Chr 13)*age QTL 6 (Chr 16)*age Variance accounted

Effect

SE

t

p

Variance %

)0.003 1.174 )0.003

0.0004 0.283 0.0006

)6.98 4.15 )4.89

<0.000 <0.000 <0.000

5.99 1.24 1.70 8.93

)0.386 0.134 0.346 )0.152 )0.268 0.543 0.239 )0.216 )0.405 )0.118 0.226 )0.138

0.090 0.181 0.111 0.169 0.235 0.324 0.098 0.144 0.207 0.284 0.195 0.284

)4.11 0.74 3.11 )0.89 )1.14 1.67 2.45 )1.49 )1.95 )0.42 1.14 )0.48

<0.000 NS 0.002 NS NS NS 0.014 NS 0.05 NS NS NS

1.77

0.0006 )0.002 0.0011 )0.0002 )0.0012 0.0003

0.0006 0.0006 0.0004 0.0004 0.0004 0.0006

1.03 )2.37 2.75 )0.39 )3.04 0.54

NS 0.018 0.006 NS 0.002 NS

1.08 1.45 0.82 1.32 1.60 8.04 1.21 0.87 0.76 2.84

NS = nonsignificant. The final model was obtained through a backward stepwise approach, where terms with p > 0.05 were excluded. Additive effects (bold) and dominant effects (nonbold) are given with their standard errors. The total phenotypic variance explained by the QTL locus is shown in the right-hand column.

Fig. 5. For each B6D2F2 age group, hematocrit values (y axis) are shown based on the genotypes of markers linked to the QTL; Filled circle, solid line = B6B6; triangle, dashed line = B6D2; open circle, solid line = D2D2.

F. JOHANNES ET AL.: AGE-RELATED QTL INFLUENCING HEMATOCRIT

695

Fig. 6. Whole-genome scans of the BXD RI are shown (males: top; females: bottom) to confirm the QTL previously identified in the B6D2F2. Each set of parallel vertical lines marks off the ± 1-LOD confidence interval surrounding the putative B6D2F2 QTL. Because of the dense BXD RI map, a single-marker mapping approach was taken, which tested for mean hematocrit differences at each analysis point (SNP) via t tests. The resulting t-test statistic is shown on the y axis. The solid horizontal line indicates the suggestive genome-wide significance threshold.

groups. This discrepancy is likely a result of sampling error, which is particularly large with small sample sizes. Other QTL were either age, sex, or age

and sex specific (QTL 2: 450-day females; QTL 3: 150-day males; QTL 4: 450- and 750-day males and 450-day females; QTL 6: 750-day males). In general,

696

the same parental alleles were associated with increased hematocrit values in both the B6D2F2 and the BXD RI strains for each of the QTL. The exception is QTL 4, for which the B6 parental allele increased trait values in the BXD RI males but decreased trait values in the BXD RI females. t-O = observed t value; t-R = required t value based on 1000 permutations of the data with respect to the SNPs within the B6D2F2 confidence interval.

5.2 4.2 48

4.8 4.2 149.6 151.3

124.4 142.8 rs3090765 rs13476148 7.2 4.3 125.6 142.8 rs3680116 rs6263578 Cel_2_135876979 4.4 4.4 135.6 rs13481015 rs6189020 rs13476148 24.3 4

rs13481021 rs3664190 rs6389547 6.3 4.4 76.7 rs13481111

8.4 4.2 50.8 4.5 4.5 92.4 92.8

rs6185344-Cel_1_152747565 5.7 4.4 146.1 152.6 9.3 4.7 155.6 rs6194543 rs3667720 12.9 4.6 120.6 rs13476936 7.77 4.3 178.1 rs13479363 rs3663343 5.52 4.2 74.9 75.3

Males 1 2 7 11 16 Females 1 2 11

SNP Position t-O t-R (Mbp) SNP Chr

150 days

Table 4. Confirmation of QTL in the BXD RI

450 days

750 days

Position t-O t-R (Mbp) SNP Position t-O t-R (Mbp)

F. JOHANNES ET AL.: AGE-RELATED QTL INFLUENCING HEMATOCRIT

BXD RI SNP effects across age. To explore the age- and sex-related effects of the confirmed QTL, the most significant BXD RI SNPs within the confirmed regions of QTL 1, 2, 3, 4, and 6 were selected, and the mean hematocrit values for each of the SNP marker classes were plotted across age (Fig. 7). In the BXD RIs, QTL 1 displays a fairly constant effect across age in both sexes (Fig. 7). QTL 2 has a significant effect only in the 150-day males and can therefore be viewed as an ‘‘early-acting’’ QTL (Fig. 7b). In the females, on the other hand, QTL 2 appears to have no effect at 150 days of age but shows an effect at 450 and 750 days (Fig. 7g). In both sexes, the influence of QTL 4 becomes more pronounced in the later-age cohorts, although there appears to be a tendency toward an ‘‘effect reversal’’ in the females insofar that the D2 genotype for this SNP displays a trend toward decreasing hematocrit at 150 days but increases the trait at 450 and 750 days (Fig. 7d, h). In males, QTL 3 appears to be active mostly at 150 days (Fig. 7c), whereas QTL 6 has its largest effect at 750 days (Fig. 7e). Other BXD RI QTL. Although the BXD RI genome scan revealed several peaks that fell within the B6D2F2 ± 1-LOD CIs, several other QTL peaks outside of these intervals were noted that exceeded suggestive or significant genome-wide threshold values (Table 5). Again, these QTL were fairly sexand age specific. Most notably, a significant QTL was located on Chr 19 (150-day males). In addition, suggestive QTL were observed on Chr 4 and 6 (750day males) and on Chr 8 (150-day males). In females, a suggestive QTL was found on Chr 9 (150 days). Finally, a suggestive QTL on Chr 14 was noted; it appeared to be fairly consistent across sex and age groups (450-day males; 150- and 450-day females). Interestingly, the D2 allele for this latter QTL appears to increase hematocrit values in males (450 days), whereas the B6 allele increases the trait in females (150 and 450 days) (Table 5). Discussion On the phenotypic level, a significant age-related and sex-specific decline in hematocrit was found, an observation that is generally in line with previous studies of the age-related changes in hematocrit. The

F. JOHANNES ET AL.: AGE-RELATED QTL INFLUENCING HEMATOCRIT

697

Fig. 7. The most significant BXD RI SNPs confirming the QTL identified in the B6D2F2 were selected and hematocrit values (y axis) are plotted for each of the SNP genotypes across age (x axis). Filled circle = B6B6; open circle = D2D2; M = males, F = females.

B6D2F2 intercross generally had higher hematocrit values compared with the BXD RIs, which may be attributable to heterosis. Also, the inbred strains may be more susceptible to the development of bleeding lesions with advancing age, which could possibly lower their hematocrit more rapidly than in the F2 animals. In the parental strains we found that hematocrit is higher in the D2 than in the B6 at 150 days of age, but that this trend was reversed at 450 and 750 days of age, with the B6 displaying higher hematocrit than the D2. A remarkably similar pattern was observed by Geiger et al. (2001) in a strain comparison of hematopoietic stem cell proliferation capacity. They demonstrated that the D2 had significantly higher proliferation capacity at 2 months (60 days) of age compared with the B6 strain. At 20 months (670 days) of age, however, the B6 strain Table 5. BXD RI QTL that did not coincide with the

B6D2F2 ± 1-LOD confidence interval but exceeded suggestive or significant genome-wide thresholds based on 1000 permutations of the data Age-cohort Males 150 days 450 days 750 days Females: 150 days 450 days 750 days a

Increasing allele

Locus

8a 19b 14a 4a 6a

rs6373291 rs4136257 rs4230383 rs13477546 rs3713705

93.6 14.5 48.8 7.6 89.1

B6 B6 DBA B6 B6

9a 14a 14a —

rs13461459 rs13482195 rs13482195 —

109 50.8 50.8

DBA B6 B6 —

Suggestive QTL. Significant QTL.

b

Position (Mbp)

Chr

showed significantly higher proliferation capacity. These converging findings suggest that the proliferation capacity of hematopoietic stem cells may be a strong contributor to baseline hematocrit maintenance throughout the life course of the organism. Several suggestive QTL that influence baseline serum hematocrit were located in the B6D2F2 intercross on Chr 1, 2, 7, 11, 13, and 16. The QTL on Chr 7, 13, and 16 interacted significantly with the age of the animal. The QTL on Chr 1, 2, 7, 11, and 16 were confirmed in a separate experiment using 23 BXD RI strains with a highly dense SNP map. However, confirmation was highly sex and age specific. This may be partially attributable to the small sample size (23 RI strains), which could have prevented the detection of QTL because of a lack of statistical power. A majority of the QTL identified and confirmed in this study overlap with loci that influence hematopoietic stem cell phenotypes that have been identified in mapping populations derived from the D2 and B6 parental strains. In particular, the QTL on Chr 1, 2, 7, and 11 are near, or identical with, QTL that are involved in controlling the frequency, cycling activity, and proliferation response of hematopoietic stem cells (deHaan and Van Zant 1999; deHaan et al. 2002; Geiger et al. 2001; Henckaerts et al. 2002a, 2002b; Mueller-Sieburg and Riblet 1996). With respect to these latter phenotypes, similar age-related changes have been found for loci on Chr 2 and 7 (Geiger et al. 2001; Henckaerts et al. 2002b). The significance of the putative QTL on Chr 2, 13, and 16 is less clear. Only the QTL on Chr 2 and 16 were confirmed in the 750-day-old BXD RI. One candidate gene near the confirmed QTL on Chr 2 is

698

F. JOHANNES ET AL.: AGE-RELATED QTL INFLUENCING HEMATOCRIT

VAV2, which is involved in lymphocyte development and activation (Fujikawa et al. 2003). One strong candidate gene near the QTL on Chr 16 is Runx1, which is implicated in hematopoiesis (Bystrykh L et al. 2005; Wang 1996). Together, these results suggest that the variation in steady-state hematocrit is primarily due to variability in genetic factors directly underlying hematopoiesis. This is surprising given the physiologic complexity of hematocrit regulation. Future studies should explore if the same QTL are active in situations where the homeostatic control of hematocrit is experimentally challenged, such as after a blood draw. This would make it possible to delineate more effectively the genetic mechanisms responsible for regulating this phenotype. The partial age dependency of QTL influencing hematocrit offers new insights into the biological regulation of this phenotype throughout the life course of the organism. However, it also complicates our understanding of the underlying genetic events because the actions of specific genes must be seen in the context of the temporal organization of the genome. Large-scale microarray expression experiments, taken at different time points in the life course of the organism, may be necessary to delineate crucial genome-wide changes. Relating these changes to the effects of QTL may provide additional insights into the regulation of baseline hematocrit.

6. Donelly S (2003) Why is erythropoietin made in the kidney? The kidney functions as a ÔcritmeterÕ to regulate hematocrit. Adv Exp Med Biol 543, 73 87 7. Evans DM, Frazer IH, Martin NG (1999) Genetic and environmental causes of variation in basal levels of blood cells. Twin Res 2, 250 257 8. Fujikawa K, Miletic AV, Alt FW, Faccio R, Brown T, et al. (2003) Vav1/2/3-null mice define an essential role for Vav family proteins in lymphocyte development and activation but a differential requirement in MAPK signaling in T and B cells. J Exp Med 198, 1595 1608 9. Fulwood R, Johnson CL, Bryer JD, Gunter EW, McGrath CR (1980) Hematological and nutritional biochemistry reference data for persons 6 months-74 years of age: United States 1976 1980. Vital and Health Statistics Series 11-No.232, DHHS Publ. No. (PHS) 83-1682 (Washington, DC: Public Health Service) 10. Geiger J, True JM, deHaan G, Van Zant G (2001) Ageand stage-specific regulation patterns in the hematopoietic stem cell hierarchy. Blood 98, 2966 2972 11. Heller DA, Ahern FM, Stout JT, McClearn GE (1998) Mortality and biomarkers of aging in heterogeneous stock (HS) mice. J Gerontol 53, B217 230 12. Henckaerts E, Geiger H, Langer CJ, Rebollo P, Van Zant G, et al. (2002a) Genetically determined variation in the number of phenotypically defined hematopoietic progenitor and stem cells and in their response to early acting cytokines. Blood 99, 3947 3954 13. Henckaerts E, Langer JC, Snoeck H-W (2002b) Quantitative genetic variation in the hematopoietic stem cell and progenitor cell compartment and in lifespan are closely linked at multiple loci in BXD recombinant inbred mice. Blood 104, 374 379 14. Henckaerts E, Langer JC, Orenstein J, Snoeck H-W (2004) The positive regulatory effect of TGT-b2 on primitive murine hematopoietic stem and progenitor cells is dependent on age, genetic background, and serum factors. J Immunol 173, 2486 2493 15. Lander ES, Kruglyak L (1995) Genetic dissection of complex traits: Guidelines for interpreting and reporting linkage results. Nate Genet 11, 241 247 16. Lander LS, Botstein D (1989) Mapping mendelian factors underlying quantitative traits using RFLP linkage maps. Genetics 121, 185 199 17. Lionikas A, Blizard DA, Vandenbergh DJ, Glover MG, Stout JT, et al. (2003) Genetic architecture of fast- and slow-twitch skeletal muscle weight in 200-day-old mice of the C57BL/6J and DBA/2J lineage. Physiol Genomics 16, 141 152 18. Mueller-Sieburg CE, Riblet R (1996) Genetic control of the frequency of hematopoietic stem cells in mice: Mapping of a candidate locus to chromosome 1. J Exp Med 183, 1141 1150 19. Plomin R, McClearn GE (1993) Quantitative trait loci (QTL) analyses and alcohol-related behaviors. Behav Genet 126, 277 284 20. Pravenec M, Zidek V, Zdobinska M, Kren V, Krenova D, et al. (1997) Mapping genes controlling hematocrit in the spontaneously hypertensive rat. Mamm Genome 8, 387 389

Acknowledgments This work was supported by the National Institute on Aging (grant AG14731) of the National Institutes of Health.

References 1. Broman KW, Wu H, Sen S, Churchill GA (2000) R/qtl: QTL mapping in experimental crosses. Bioinformatics 19, 889 890 2. Bystrykh L, Weersing E, Dontje B, Sutton S, Pletcher MT, et al. (2005) Uncovering regulatory pathways that affect hematopoietic stem cell function using ‘‘genetical genomics’’. Nat Genet 37, 225 232 3. Churchill GA, Doerge RW (1994) Empirical threshold values for quantitative trait mapping. Genetics 138, 963 971 4. deHaan G, Van Zant G (1999) Genetic analysis of hemopoietic cell cycling in mice suggests its involvement in organismal life span. FASEB J 13, 707 713 5. deHaan G, Bystrykh LV, Weersing E, Dontje B, Geiger H, et al. (2002) A genetic and genomic analysis identifies a cluster of genes associated with hematopoietic cell turnover. Blood 100, 2056 2062

F. JOHANNES ET AL.: AGE-RELATED QTL INFLUENCING HEMATOCRIT

699

21. R Development Core Team (2002) R: A language and environment for statistical computing (Vienna, Austria: R Foundation for Statistical Computing), ISBN 3-900051-07-0. Available at http://www.R-project.org 22. Russell ES, McFarland EC (1966) Analysis of pleiotropic effects of W and f genic substitutions in the mouse. Genetics 53, 949 959 23. Schlager G, Weibust RS (1976) Selection for hematocrit percent in the house mouse. J Hered 67, 295 299 24. Stino FKR, Washburn KW (1973) Divergent selection under two nutritional environments for packed erythrocyte volume in Japenese quail. Genetics 74, 363 369 25. Strauss JA (1992) Physiological biomarkers of aging, using heterogeneous stock mice as a modeling system. PhD Dissertation, The Pennsylvannia State University 26. Van Hearingen WA, den Bieman MG, Lankhorst AE, van Lith HA, van Zutphen LFM (2002) Application of

AFLP markers for QTL mapping in the rabbit. Genome 45, 914 921 Van Zant G, Holland BP, Eldridge PW, Chen J-J (1990) Genotype-restricted growth and aging patterns in hematopoietic stem cell populations of allophonic mice. J Exp Med 171, 1547 1565 Vandenbergh DJ, Heron K, Peterson R, Shpargel KB, Woodroofe A, et al. (2003) Simple tests to detect errors in high-throughput genotype data in the molecular laboratory. J Biomol Tech 14, 9 16 Wang Q, Stacy T, Binder M, Marin-Padilla M, Sharpe AH, et al. (1996) Disruption of the Cbfa2 gene causes necrosis and hemorrhaging in the central nervous system and blocks definitive hematopoiesis. Proc Natl Acad Sci USA 93, 3444 3449 Weibust RS, Schlager G (1968) A genetic study of blood pressure, hematocrit and plasma cholesterol in aged mice. Life Sci 7, 1111 1119

27.

28.

29.

30.