Luanne L. Peters,1 Amy J. Lambert,1 Weidong Zhang,1 Gary A. Churchill,1 Carlo Brugnara,2. Orah S. Platt2. 1The Jackson Laboratory, 600 Main Street, Bar ...
Quantitative trait loci for baseline erythroid traits Luanne L. Peters,1 Amy J. Lambert,1 Weidong Zhang,1 Gary A. Churchill,1 Carlo Brugnara,2 Orah S. Platt2 1
The Jackson Laboratory, 600 Main Street, Bar Harbor, Maine 04609, USA Department of Laboratory Medicine, ChildrenÕs Hospital, Harvard Medical School, Boston, Massachusetts 02115, USA
2
Received: 22 October 2005 / Accepted: 7 December 2005
Introduction
Abstract
A substantial genetic contribution underlies variation in baseline peripheral blood counts. We performed quantitative trait locus/loci (QTL) analyses to identify chromosome (Chr) regions harboring genes influencing the baseline erythroid parameters in F2 intercrosses between NZW/LacJ, SM/J, and C57BLKS/J inbred mice. We identified multiple significant QTL for red blood cell (RBC) count, hemoglobin (Hgb) and hematocrit (Hct) levels, mean corpuscular volume (MCV), mean corpuscular hemoglobin (MCH), and mean cell hemoglobin concentration (CHCM). We identified four RBC count QTL: Rbcq1 (Chr 1, peak LOD score at 62 cM,), Rbcq2 (Chr 4, 60 cM), Rbcq3 (Chr 11, 34 cM), and Rbcq4 (Chr 10, 60 cM). Three MCV QTL were identified: Mcvq1 (Chr 7, 30 cM), Mvcq2 (Chr 11, 6 cM), and Mcvq3 (Chr 10, 60 cM). Single significant loci for Hgb (Hgbq1, Chr 16, 32 cM), Hct (Hctq1, Chr 3, 42 cM), and MCH (Mchq1, Chr 10, 60 cM) were identified. The data support the existence of a common RBC/MCH/MCV locus on Chr 10. Two QTL for CHCM (Chcmq1, Chr 2, 48 cM; Chcmq2, Chr 9, 44 cM) and an interaction between Chcmq2 with a locus on Chr 19 were identified. These analyses emphasize the genetic complexity underlying the regulation of erythroid peripheral blood traits in normal populations and suggest that genes not previously recognized as significantly impacting normal erythropoiesis exist.
Correspondence to: Luanne L. Peters; E-mail: Luanne.Peters@ jax.org
298
A significant genetic component underlies baseline peripheral blood parameters (Chen and Harrison 2002; Garner et al. 2000; Lin et al. 2005; Mahaney et al. 2005; Peters et al. 2004). In epidemiologic studies, baseline hematopoietic traits are significant, independent risk factors for human complex diseases with considerable morbidity and mortality. The baseline white blood cell (WBC) count correlates with early mortality, heart disease, and stroke in the general population. Increased hematocrit (Hct) level is a significant and independent risk factor for cardiovascular disease (coronary heart disease, myocardial infarction), cerebrovascular disease, peripheral vascular disease, and mortality (Gagnon et al. 1994; Lippi et al. 2002; Wannamethee et al. 1994). The mean platelet volume (MPV) is associated with heart disease and stroke (Bath et al. 2004; Martin et al. 1991, 1992). In sickle cell disease, increased baseline white blood cell (WBC) count is a strong predictor of acute chest syndrome with an impact comparable to hemoglobin F levels and is a significant predictor of infarcts and hemorrhagic stroke as well (Castro et al. 1994; de Labry et al. 1990; Kinney et al. 1999; Miller et al. 2000; Platt et al. 1994). Total hemoglobin (Hgb) is also a risk factor of sickle cell disease severity; high Hgb is associated with higher risk for pain crises and acute chest syndrome, while low Hgb levels are associated with higher risk of stroke (Castro et al. 1994; Ohene-Frempong et al. 1998; Platt et al. 1989). The number of F cells persisting in adults is an important modulator of thalassemia and sickle cell disease (Lal and Vichinsky 2004; Platt et al. 1994). We have shown that modifiers of red cell mean corpuscular volume (MCV) significantly impact the course of inherited hemolytic anemia (hereditary spherocytosis) in mice (Peters et al. 2004). From these examples it is clear that genes regulating peripheral blood traits profoundly affect
DOI: 10.1007/s00335-005-0147-3 Volume 17, 298 309 (2006) Springer Science+Business Media, Inc. 2006
L.L. PETERS
ET AL.:
QTL
FOR
ERYTHROID TRAITS
disease progression. Identifying the primary genetic determinants of peripheral blood indices will not only enhance our understanding of hematopoiesis but also provide novel targets for risk assessment, diagnostics, and therapeutic intervention in hematologic and cardiovascular pathologies (Lal and Vichinsky 2004; Platt et al. 1994). QTL analysis is an unbiased approach to the identification of novel genes and their functions and can point toward previously unrecognized genetic regulatory pathways. Concordance between QTL in rodents and humans has been demonstrated for multiple complex traits and diseases. Thus, QTL analysis in inbred mice is highly significant and can identify genes for subsequent meaningful, focused association studies in human populations (Paigen 2002; Sugiyama et al. 2001; Wang and Paigen 2005a, 2005b). We are using inbred mouse strains to identify QTL for baseline hematologic parameters as a first step in elucidating the genes that regulate these traits in normal populations. In this article we focus on erythroid traits. In single-locus genomewide scans we have identified four significant red blood cell (RBC) QTL, three significant MCV QTL, two significant mean cell hemoglobin concentration (CHCM) QTL, and one QTL each for Hgb, Hct, and mean corpuscular hemoglobin (MCH). Moreover, multiple regression analysis reveals additional significant loci for these traits. Several suggestive QTL are also apparent. Our analyses confirm that multiple genes contribute to baseline erythropoietic traits and suggest that a network of genes not previously appreciated as significantly impacting hematopoiesis exist. The results reported here are a significant first step in identifying those genes. Materials and methods Animals. Mice were housed in humidity- and temperature-controlled rooms (12-h light cycle) with free access to acidified water and food (NIH 5K52). The Jackson Laboratory Animal Care and Use Committee approved all protocols. The Jackson Laboratory is fully accredited by the American Association for Accreditation of Laboratory Animal Care (AAALAC). Complete blood counts. Whole blood (275 ll) from 8-week-old adult mice was drawn from the retro-orbital sinus through EDTA-coated microhematocrit tubes directly into Eppendorf tubes containing 20 ll 20% EDTA in murine phosphatebuffered saline (PBS), as described previously (Peters
299
et al. 2004). Complete blood counts were determined using an Advia 120 multispecies whole-blood analyzer (Bayer Corporation, Tarrytown, NY). Statistical analysis. To identify QTL, two F2 intercrosses were established between F1 hybrids of NZW/LacJ · SM/J (NZSM cross) and C57BLKS/J · SM/J mice (KSSM cross). The parentals strains (NZW/LacJ, SM/J, C57BLKS/J) are considered priority strains by the Mouse Phenome Project based on their wide use in the research community and their genetic diversity. (For a complete listing of priority mouse strains, see the Mouse Phenome Database, www.jax.org/phenome.) Complete blood counts in 186 F2 progeny from each cross were performed as described above. Approximately equal numbers of F2 males and females were obtained; females accounted for 57% and 49% of the NZSM and KSSM F2 progeny, respectively. Simple sequence length polymorphic (SSLP) markers spaced at 10 30-cM intervals throughout the genome were typed by the polymerase chain reaction (PCR) using multiplexed fluorescent markers and ABI 3700 instrumentation (Applied Biosystems, Foster City, CA). A total of 88 markers were typed in each of the two crosses (marker information available upon request). Genome-wide scans were performed using log-transformed RBC count, Hgb and Hct values, MCV, MCH, and CHCM as quantitative traits. Results were analyzed in three stages, as described previously (Sen and Churchill 2001). First, genome-wide scans with sex and body weight as additive covariates were performed to detect single loci associated with each trait (main-effect QTL) using Pseudomarker software (www.jax.org/research/churchill) (Cordell et al. 1998). We also carried out single-locus scans that included sex as an interactive covariate to identify sex-specific QTL (Korstanje et al. 2004). Significance thresholds were determined by permutation testing (n = 1000 permutations) (Churchill and Doerge 1994). A LOD (logarithm of the odds ratio) score of the 95th percentile or greater (p < 0.05) of the permutation distribution was considered significant; a score meeting or exceeding the 37th percentile (p < 0.63) was considered suggestive (Lander and Kruglyak 1995). Confidence intervals (CI) were determined by computing the region of the posterior probability density curve containing 95% of the total area (Sen and Churchill 2001). The posterior probability density is proportional to 10LOD and gives results that are similar to the 1.5 ‘‘LOD support interval’’ but it is better justified on theoretical grounds (Sen and Churchill 2001). In the second stage of the analysis, a simultaneous search for pairs to detect epistatic interactions
300
L.L. PETERS
ET AL.:
QTL
FOR
ERYTHROID TRAITS
Fig. 1. Genome-wide scans for RBC count, Hgb, Hct, MCV, MCH, and CHCM QTL in the KSSM and NZSM crosses. Significant (p = 0.05) and suggestive (p = 0.63) LOD scores are indicated by the upper and lower dotted lines, respectively. Gene symbols are indicated above significant QTL. Chromosomes 1 through X are displayed on the ordinates with the relative spacing representative of the relative length of each chromosome. LOD scores are given on the y axis.
L.L. PETERS
ET AL.:
QTL
FOR
301
ERYTHROID TRAITS
Table 1. QTL for baseline erythroid traits
Name
Trait
Chr
Peak, cM (Mb)
95% CI (cM)
Human location
Rbcq1 Rbcq2 Rbcq3 Rbcq4 Hgbq1 Hctq1 Mchq1 Mcvq1a Mcvq2
RBC RBC RBC RBC Hgb Hct MCH MCV MCV
1 4 11 10 16 3 10 7 11
62 (115) 60 (131) 34 (60) 60 (112) 32 (44) 42 (89) 60 (112) 30 (50) 6 (12)
34 50 18 40 20 30 48 26 0
18q21 1p34-p36 17p11, 1q41-q42 12q21 3q12-q13 1q21 12q21 15q14 7p13-p11
Mcvq3 Chcmq1 Chcmq2
MCV CHCM CHCM
10 2 9
60 (112) 48 (80) 44 (80)
48 60 30 60 40 66
74 66 56 60 42 62 60 38 34
12q21 2q32 6p12, 6q13
High allele
Inheritance
Peak marker
LOD
SM/J SM/J SM/J C57BLKS/J SM/J NZW/LacJ SM/J SM/J NZW/LacJ and SM/J SM/J C57BLKS/J SM/J
Recessive Recessive Dominant Dominant Dominant Recessive Recessive Recessive Heterozygotes lowb Recessive Recessive Dominant
D1Mit308 D4Mit170 D11Mit30 D10Mit150 D16Mit60 D3Mit98 D10Mit150 D7Mit83 D11Mit2
3.4 5.3 3.8 3.7 3.7 3.9 4.4 9.8 4.7
D10Mit15 D2Mit300 D9Mit198
4.7 4.7 6.1
Gene symbols, by convention, are given in lowercase and italics for RBC (red blood cell count), Hgb (hemoglobin), Hct (hematocrit), MCH (mean corpuscular hemoglobin), MCV (mean corpuscular volume), CHCM (mean cell hemoglobin concentration). Estimated peak positions in Mb are determined using Mouse Genome Informatics linkage maps (http://www.informatics.jax.org/searches/linkmap_form.shtml) and Ensemble Mouse Genome Browser (Build 34) (http://www.ensembl.org/Mus_musculus/). a Data derived from combined cross analysis. b For QTL for which heterozygotes are low, we hypothesize that there may be closely linked QTL with effects in the opposite direction.
was performed using a two-way analysis of variance (ANOVA) model (Sen and Churchill 2001; Sugiyama et al. 2001). Third, to determine the combined effects of all QTL on each trait, multiple regression analyses were performed that included all significant and suggestive QTL and all possible QTL interactions for the trait. Terms that failed to meet significance levels were eliminated one-by-one until all remaining QTL were significant, resulting in the final model. The percent variance reported for each QTL derives from this analysis. To narrow the CI for MCV QTL on Chr 7, data from both crosses were combined and analyzed as described previously (Li et al. 2005). Other statistical analyses were performed using JMP 5.1.2 software (SAS Institute Inc., Cary, NC). Between-group comparisons were analyzed by Tukey HSD test to determine statistical significance. Results Identification of QTL. Because SM females are not the best breeders, we set up unidirectional (NZ · SM and KS · SM) crosses to generate F1 progeny and intercrossed the F1 s to obtain F2 progeny for analysis. Originally, we established these crosses to identify QTL for WBC count, platelet (Plt) count, and mean platelet volume (MPV) (Peters et al. 2005). These traits differed substantially between the parental strains. In this study we have extended our analysis to RBC traits using these same crosses. Notably, the parental strains do not differ significantly for any of the erythroid traits analyzed. [For
complete blood count data for all priority inbred strains of mice, see Mouse Phenome Database, www.jax.org/phenome; Project MPD:62, Peters1 (Peters and Barker 2005).] Hence, one would predict that if any QTL were identified, the parental strains would necessarily be contributing both high and low alleles to the trait. This proved to be the case for all traits analyzed in which more than one QTL was identified (see below for allele effects). Genome-wide scans for single QTL are shown in Fig. 1. Centimorgan (cM) positions are from the Mouse Genome Informatics database (www.informatics.jax.org). For each cross 186 F2 animals were analyzed using 88 simple sequence length polymorphism (SSLP) markers. The data for all significant QTL detected in this study are summarized in Table 1. QTL for baseline RBC count. Analysis for single-locus QTL in both crosses using sex and body weight as additive covariates revealed significant loci for RBC count on chromosomes (Chr) 1, 4, 10, and 11 (Fig. 1). The peak LOD score and the posterior probability density curves used to define the 95% CIs for all RBC QTL are shown in Fig. 2 and summarized in Table 1. No influence of sex was detected for any of the RBC QTL, as determined by repeating the analysis with sex as an interacting covariate (no significant differences in LOD scores were seen between the two scans). A simultaneous search to detect interacting gene pairs (epistasis) was negative. By convention, these QTL were named Rbcq1 Rbcq4 (red blood cell quantitative loci 1 4) (Abiola et al. 2003).
302
L.L. PETERS
ET AL.:
QTL
FOR
ERYTHROID TRAITS
Fig. 2. (A D) Genome-wide scans (solid lines) and posterior probability densities (dashed lines) for significant RBC QTL, Rbcq1 4. Posterior probability density is the likelihood statistic that is used to compute the 95% confidence interval, indicated by the gray bars. Position along the chromosome is shown on the ordinates. The y axis shows LOD scores. Results are based on single-locus scans of 186 F2 animals. (E H) Effect plots for RBC QTL. Allelic contributions determined at the peak marker for Rbcq1 4. The three possible genotypes are indicated on each ordinate. K, C57BLKS/J allele; N, NZW/LacJ allele; S, SM/J allele. The y axis represents RBC counts. Error bars represent SEM. *p < 0.05, ns = not significant.
Allele effects. Allele effects were determined by calculating the phenotypic mean at the peak marker for each of the three possible genotypes (Lyons et al. 2003). We then determined which strain contributed the allele that increased the trait and whether it acted in a dominant, additive (codominant), or recessive manner. In the NZSM cross, recessive SM alleles at the Rbcq1 and Rbcq2 loci increase RBC count, while a dominant SM allele increases RBC count at Rbcq3 (Fig. 2). In the KSSM cross, a dominant KS allele at Rbcq4 increases RBC count (Fig. 2).
Combined effects of RBC QTL. Single-locus genome scans as displayed in Fig. 1 do not consider other QTL. Multiple regression models, however, allow a QTL effect to be assessed in the context of all other QTL. We performed multiple regression analysis integrating all suggestive and significant RBC QTL and all possible RBC QTL interactions to determine the contribution of each QTL to the total variance. The regression analysis predicts that a Chr 8 QTL, which was suggestive only in the main scan (Fig. 1), is significant in the final model for the KSSM
L.L. PETERS
ET AL.:
QTL
FOR
303
ERYTHROID TRAITS
Table 2. Multiple regression analysis of RBC QTL
KSSM cross Sex Rbcq4 Chr 8@26 cM Total NZSM cross Sex Rbcq1 Rbcq2 Rbcq3 Total
df
% Variance
1 2 2
0.43 8.4 6.8 15.8
1 2 2 2
1.3 3.8 10.6 8.8 27.1
F value 0.909 9.011 7.29 3.09 4.63 12.83 10.68
P(f) 0.341619 (ns) 0.000186 (p < 0.001) 0.000908 (p < 0.001) 0.076030 ns 0.010987 (p < 0.05) 6.27 · 10)6 (p < 0.001) 4.20 · 10)5 (p < 0.001)
df = degrees of freedom; ns = not significant. a Percent variances are based on type III sums of squares (SS). The adjusted SS used to compute the % variance takes into account other major QTLs in the population and thus does not necessarily sum up to the total variance explained. A discussion can be found in Sokal and Rohlf (1981).
cross (Table 2). With Rbcq4, this QTL accounted for 15.8% of the variance in RBC count in this cross. In the NZSM cross, Rbcq1 3 accounted for 27.1% of
the variance in RBC count. No influence of sex was detected for any of these RBC QTL, and a simultaneous search to detect interacting pairs was nega-
Fig. 3. (A C) Genome-wide scans (solid lines) and posterior probability densities (dashed lines) for Hgbq1, Hctq1, and Mchq1. Posterior probability density is the likelihood statistic that is used to compute the 95% confidence interval, indicated by the gray bars. Position along the chromosome is shown on the ordinates. The y axis shows LOD scores. Results are based on single-locus scans of 186 F2 animals. (D F) Allelic contributions determined at the peak markers for Hgbq1, Hctq1, and Mchq1. The three possible genotypes are indicated on each ordinate. K, C57BLKS/J allele; N, NZW/ LacJ allele; S, SM/J allele. The y axis represents the trait value. Error bars represent SEM. *p < 0.05, ns = not significant.
304
L.L. PETERS
ET AL.:
QTL
FOR
ERYTHROID TRAITS
Table 3. Multiple regression analysis of Hgb, Hct, and MCH QTL
KSSM cross Sex Hgbq1 Total Sex Mchq1 Total NZSM cross Sex Hctq1 Chr 6@60 cM Total
df
% Variance
F value
1 2
4.1 7.4 10.7
8.32 7.90
1 2
1.3 10.3 11.4
2.76 10.55
1 2 2
0.52 9.09 6.20 15.5
1.1 9.51 6.54
P(f) 0.004389 (p < 0.01) 0.000515 (p < 0.001) 0.098210 (ns) 4.62 · 10)5 (p < 0.001) 0.295779 (ns) 0.000112 (p < 0.001) 0.001824 (p < 0.01)
ns = not significant.
tive. This was also true for all other traits analyzed in this study. Identification of QTL for Hgb, Hct, and MCH. Genome-wide scans revealed a single significant QTL for each of these traits: Hgbq1 (hemoglobin quantitative locus-1) on Chr 16, Hctq1 (hematocrit quantitative locus-1) on Chr 3, and Mchq1 (mean corpuscular hemoglobin quantitative locus-1) on Chr 10 (Fig. 1, Table 1). A dominant SM allele at Hgbq1 increases hemoglobin levels; a recessive NZ allele increases the Hct at Hctq1, and a recessive SM allele increases the MCH at Mchq1 (Fig. 3, Table 1). Regression analysis identified a second significant hematocrit QTL on Chr 6 (Table 3). Hgbq1 accounted for 10.7% of the trait variance, Mchq1 10.3%, and Hctq1/Chr 6 15.5% (Table 3). Identification of QTL for MCV. Three significant MCV QTL were detected, Mcvq1 on Chr 7, Mcvq2 on Chr 11, and Mcvq3 on Chr 10 (Fig. 1). Mcvq1 was detected in both crosses. Hence we took advantage of a statistical method, combined cross analysis, to significantly narrow the 95% CI (Fig. 4) (Li et al. 2005). Recessive SM alleles at Mcvq1 and Mcvq3 increase red cell volume, while a recessive NZ allele does so at Mcvq2 (Fig. 4). Multiple regression did not reveal any additional significant MCV QTL. The MCV QTL identified accounted for approximately 25% of the total variance of the trait (Table 4). Identification of QTL for CHCM. Like the MCV, CHCM is directly measured on a cell-by-cell basis on the Advia 120 analyzer, i.e., it is not a
calculated parameter. CHCM is analogous to the calculated parameter MCHC (mean corpuscular hemoglobin content). We identified two significant CHCM QTL in our single-locus genome-wide scans: Chcmq1 on Chr 2 and Chcmq2 on Chr 9 (Figs. 1 and 5). A recessive KS allele increases CHCM at Chcmq1, and a dominant SM allele at Chcmq2 (Fig. 5) increases CHCM. Multiple regression analysis revealed a third significant CHCM QTL on Chr 19 (Table 5). Interestingly, the Chr 19 locus interacted with Chcmq2; homozygosity for the Chr 19 KS allele significantly increases CHCM when Chcmq2 is also homozygous for the KS allele, but otherwise it has no effect (Fig. 5). In total, the CHCM QTL account for 44% of the variance (Table 5). Discussion QTL analysis in mice is a powerful, unbiased approach to the identification of novel genes and pathways influencing complex phenotypes. We recently reported QTL for WBC count, Plt count, and MPV (Peters et al. 2005), and another recent study identified loci associated with baseline Plt counts in mice (Cheung et al. 2004), but identification of QTL for baseline blood counts in mice has otherwise received scant attention. No studies in mice to localize QTL for steady-state erythroid traits, important risk factors for heart disease and stroke, have been reported to our knowledge. Likewise, analysis of erythroid traits as a complex phenotype in humans is newly emerging. Recent QTL analysis using the Framingham Heart cohort has identified a significant QTL for Hct on 6q23-24 (Lin et al. 2005). A suggestive Hct QTL was also found on human Chr 1 (Lin et al. 2005). The positions of the loci identified in humans thus far, with the exception of the suggestive locus on Chr 1, do not correspond to the
L.L. PETERS
ET AL.:
QTL
FOR
ERYTHROID TRAITS
305
Fig. 4. (A D) Genome-wide scans (solid lines) and posterior probability densities (dashed lines) for significant MCV QTL. Posterior probability density is the likelihood statistic that is used to compute the 95% confidence interval (CI), indicated by the gray bars. Position along the chromosome is shown on the ordinates. The y axis shows LOD scores. Results are based on single-locus scans of 186 F2 animals. (E) Genome wide scan (solid line) and posterior probability density (dashed line) for Mcvq1 following combined analysis of KSSM (A) and NZSM (B) data. Note the decrease in the 95% CI. (F H) Allelic contributions determined at the peak markers. The three possible genotypes are indicated on each ordinate. K, C57BLKS/J allele; N, NZW/LacJ allele; S, SM/J allele. The y axis represents the MCV. Error bars represent SEM. *p < 0.05, ns = not significant.
306
L.L. PETERS
ET AL.:
QTL
FOR
ERYTHROID TRAITS
Table 4. Multiple regression analysis of MCV QTL
KSSM cross Sex BW Mcvq1 Mcvq3 Total NZSM cross Sex BW Mcvq1 Mcvq2 Total
df
% Variance
F value
P(f)
1 1 2 2
1.25 1.94 6.77 8.23 24.5
2.97 4.59 8.03 9.75
0.86709 (ns) 0.033555 (p < 0.05) 0.000460 (p < 0.001) 9.57 · 10)5 (p < 0.001)
1 2 2 2
0.08 0.76 9.50 8.75 24.6
0.18 4.14 11.15 10.26
0.670353 (ns) 0.043413 (p < 0.05) 2.76 · 10)5 (p < 0.001) 6.08 ·10)5 (p < 0.001)
ns = not significant.
human syntenic positions of our loci (Table 3), confirming our suspicion that multiple QTL for baseline peripheral blood traits exist and many remain to be identified. In the current study using two crosses between inbred strains of mice, we identified multiple QTL influencing baseline erythroid traits—RBC count, Hgb, Hct, MCV, MCH, and CHCM (summarized in Table 1). A common QTL for RBC count, MCH, and MCV was identified on Chr 10, suggesting that these related parameters are under common genetic influences. Independent loci influencing each of these traits were also identified. These data reinforce the concept that steady-state peripheral blood values are truly complex traits. Given that these parameters are significantly associated with common diseases in man, elucidating the underlying genetic networks influencing these traits is critical. The CIs for our QTL are quite large, not unexpected in a ‘‘first pass’’ analysis. The smallest interval, Mcvq1, owes its reduced size to combined cross analysis, a statistical method developed by Li et al. (2005). As its name implies, combined cross analysis considers the phenotypic information from multiple crosses in a single QTL analysis to provide increased recombination and, thus, statistical power, allowing one to detect QTL with small effects, resolve closely linked QTL, and narrow QTL confidence intervals compared with individual QTL crosses. The more crosses combined, the more power achieved. Clearly, a need exists to establish additional mouse crosses in order to effectively apply new statistical and bioinformatic methods (e.g., haplotype analysis) (Wang et al. 2004) to the study of hematopoiesis as a complex trait and begin unraveling the regulatory genetic networks. Given the current 95% CIs, examination of potential candidate genes is currently unwieldy at best. For all QTL intervals (Table 1), we examined the
databases for potential candidate genes. Because most the 95% CIs are quite large, we confined our analysis to 10 Mb of 5¢ and 3¢ sequence flanking the QTL peak (20 Mb total) using the Ensembl mouse server (http://www.ensembl.org/Mus_musculus/index.html). Interestingly, the genes not present within our intervals are the most notable. For example, EPO, EPOR, and other growth factors and transcription factors (e.g., GATA, EKLF, SCL) known to influence erythroid expansion, lineage commitment, or differentiation (Cantor and Orkin 2001, 2002, 2005; Orkin 2000; Orkin and Zon 2002; Orkin et al. 1999) are not present. In some cases an intriguing candidate was observed but could be eliminated based on expression data and/or knockout phenotyping data. For example, Csf2 [colony stimulating factor 2 (granulocyte-macrophage)], located near the Rbcq3 peak, is not expressed beyond the median in erythroid cells or tissues according to the Novartis microarray databases (http://symatlas.gnf.org/SymAtlas/), and the knockout phenotype is one of lung pathology with no evidence of erythroid involvement [Mouse Genome Informatics (MGI), http://www.informatics.jax.org/). For Rbcq3, the gene encoding the erythroid membrane skeleton component protein 4.1 (Epb4.1) is located near the peak. Protein 4.1 is critical to red cell stability (Peters and Barker 2001) and thus represents a viable candidate gene at this locus. Similar extensive in silico ‘‘mining’’ of our other QTL intervals failed to identify obvious, strong candidates, suggesting that a complex regulatory network consisting of multiple genes not previously recognized as significantly impacting normal erythropoiesis indeed exists. We are establishing additional crosses to further explore QTL for baseline hematologic parameters in mice. The parental strains used in the current analysis are members of the 40 ‘‘priority’’ strains of the
L.L. PETERS
ET AL.:
QTL
FOR
307
ERYTHROID TRAITS
Fig. 5. Genome-wide scans (solid lines), posterior probability densities (dashed lines), and effect plots for significant CHCM QTL. Posterior probability density is the likelihood statistic that is used to compute the 95% confidence interval, indicated by the gray bars for (A) Chcmq1, (B) Chcmq2. Position along the chromosome is shown on the ordinates. The y axis shows LOD scores. Results are based on single-locus scans of 186 F2 animals. (C, D) Effect plots. The three possible genotypes are indicated on each ordinate. K, C57BLKS/J allele; S, SM/J allele. The Y axis represents CHCM counts. Error bars represent SEM. *p < 0.05, ns = not significant. (E) Interaction between Chcmq2 and a locus on Chr 19. The effects of gene interaction predicted by multiple regression analysis are shown. The three possible genotypes for Chcmq2 are indicated on the ordinate. Genotypes for the interacting locus on Chr 19 are indicated by the circles, boxes, and diamonds. K, C57BLKS/J allele; S, SM/J allele. The y axis represents CHCM value. Error bars represent SEM. Table 5. Multiple regression analysis of CHCM QTL
KSSM cross Sex BW Chcmq1 Chcmq2 Chr 19@32 Chr 9@44:Chr 19@32 Total variance
df
% Variance
F value
P(f)
1 1 2 6 6 4
1.429 1.618 7.661 16.812 8.207 5.302 43.982
4.415 4.996 11.829 8.653 4.224 4.093
0.037082 (p < 0.05) 0.026679 (p < 0.05) 1.53e-005 (p < 0.001) 3.08e-008 (p < 0.001) 0.000540 (p < 0.001) 0.003398 (p < 0.01)
308
Mouse Phenome Project, chosen on the basis of their wide use and genetic diversity, and share ancestry with other widely used mouse strains. Thus, analysis of haplotypes using publicly available high-density SNP data can be used to identify other strains that are likely to share alleles. Our ultimate goal is to identify overlapping QTL in multiple crosses to provide the framework for the application of new, powerful statistical and bioinformatic techniques to significantly narrow QTL intervals and, ultimately, identify the underlying genes (Li et al. 2005; Wang et al. 2004). Such studies will provide critical focal points for clinical interventions. Acknowledgments The authors thank Drs. Beverly Paigen and David Harrison for critical review of the manuscript. This work was supported by National Institutes of Health grants HL68922 (OSP) and HL64885 and HL66611 (LLP), and The National Cancer Institute CA34196 (The Jackson Laboratory). References 1. Abiola O, Angel JM, Avner P, Bachmanov AA, Belknap JK, et al. (2003) The nature and identification of quantitative trait loci: a communityÕs view. Nat Rev Genet 4, 911 916 2. Bath P, Algert C, Chapman N, Neal B (2004) Association of mean platelet volume with risk of stroke among 3134 individuals with history of cerebrovascular disease. Stroke 35, 622 626 3. Cantor AB, Orkin SH (2001) Hematopoietic development: a balancing act. Curr Opin Genet Dev 11, 513 519 4. Cantor AB, Orkin SH (2002) Transcriptional regulation of erythropoiesis: an affair involving multiple partners. Oncogene 21, 3368 3376 5. Cantor AB, Orkin SH (2005) Coregulation of GATA factors by the Friend of GATA (FOG) family of multitype zinc finger proteins. Semin Cell Dev Biol 16, 117 128 6. Castro O, Brambilla DJ, Thorington B, Reindorf CA, Scott RB, et al. (1994) The acute chest syndrome in sickle cell disease: incidence and risk factors. The Cooperative Study of Sickle Cell Disease. Blood 84, 643 649 7. Chen J, Harrison DE (2002) Quantitative trait loci regulating relative lymphocyte proportions in mouse peripheral blood. Blood 99, 561 566 8. Cheung CC, Martin IC, Zenger KR, Donald JA, Thomson PC, et al. (2004) Quantitative trait loci for steady-state platelet count in mice. Mamm Genome 15, 784 797
L.L. PETERS
ET AL.:
QTL
FOR
ERYTHROID TRAITS
9. Churchill GA, Doerge RW (1994) Empirical threshold values for quantitative trait mapping. Genetics 138, 963 971 10. Cordell HJ, Todd JA, Lathrop GM (1998) Mapping multiple linked quantitative trait loci in non-obese diabetic mice using a stepwise regression strategy. Genet Res 71, 51 64 11. de Labry LO, Campion EW, Glynn RJ, Vokonas PS (1990) White blood cell count as a predictor of mortality: results over 18 years from the Normative Aging Study. J Clin Epidemiol 43, 153 157 12. Gagnon DR, Zhang TJ, Brand FN, Kannel WB (1994) Hematocrit and the risk of cardiovascular disease—the Framingham study: a 34-year follow-up. Am Heart J 127, 674 682 13. Garner C, Tatu T, Reittie JE, Littlewood T, Darley J, et al. (2000) Genetic influences on F cells and other hematologic variables: a twin heritability study. Blood 95, 342 346 14. Kinney TR, Sleeper LA, Wang WC, Zimmerman RA, Pegelow CH, et al. (1999) Silent cerebral infarcts in sickle cell anemia: a risk factor analysis. The Cooperative Study of Sickle Cell Disease. Pediatrics 103, 640 645 15. Korstanje R, Li R, Howard T, Kelmenson P, Marshall J, et al. (2004) Influence of sex and diet on quantitative trait loci for HDL cholesterol levels in an SM/J by NZB/BlNJ intercross population. J Lipid Res 45, 881 888 16. Lal A, Vichinsky E (2004) The role of fetal hemoglobinenhancing agents in thalassemia. Semin Hematol 41, 17 22 17. Lander E, Kruglyak L (1995) Genetic dissection of complex traits: guidelines for interpreting and reporting linkage results. Nat Genet 11, 241 247 18. Li R, Lyons MA, Wittenburg H, Paigen B, Churchill GA (2005) Combining data from multiple inbred line crosses improves the power and resolution of quantitative trait loci mapping. Genetics 169, 1699 1709 19. Lin JP, OÕDonnell CJ, Levy D, Cupples LA (2005) Evidence for a gene influencing haematocrit on chromosome 6q23 24: genomewide scan in the Framingham Heart Study. J Med Genet 42, 75 79 20. Lippi G, Manzato F, Franchini M, Guidi G (2002) Relationship between hematocrit, primary hemostasis, and cardiovascular disease in athletes. Am Heart J 144, E14 21. Lyons MA, Wittenburg H, Li R, Walsh KA, Leonard MR, et al. (2003) New quantitative trait loci that contribute to cholesterol gallstone formation detected in an intercross of CAST/Ei and 129S1/SvImJ inbred mice. Physiol Genomics 14, 225 239 22. Mahaney MC, Brugnara C, Lease LR, Platt OS (2005) Genetic influences on peripheral blood cell counts: a study in baboons. Blood 106, 1210 1214 23. Martin JF, Bath PM, Burr ML (1991) Influence of platelet size on outcome after myocardial infarction. Lancet 338, 1409 1411
L.L. PETERS
ET AL.:
QTL
FOR
309
ERYTHROID TRAITS
24. Martin JF, Bath PM, Burr ML (1992) Mean platelet volume and myocardial infarction. Lancet 339, 1000 1001 25. Miller ST, Sleeper LA, Pegelow CH, Enos LE, Wang WC, et al. (2000) Prediction of adverse outcomes in children with sickle cell disease. New Engl J Med 342, 83 89 26. Ohene-Frempong K, Weiner SJ, Sleeper LA, Miller ST, Embury S, et al. (1998) Cerebrovascular accidents in sickle cell disease: Rates and risk factors. Blood 91, 288 294 27. Orkin SH (2000) Diversification of haematopoietic stem cells to specific lineages. Nat Rev Genet 1, 57 64 28. Orkin SH, Zon LI (2002) Hematopoiesis and stem cells: plasticity versus developmental heterogeneity. Nat Immunol 3, 323 328 29. Orkin SH, Porcher C, Fujiwara Y, Visvader J, Wang LC (1999) Intersections between blood cell development and leukemia genes. Cancer Res 59, 1784S 1787S 30. Paigen K (2002) Understanding the human condition: experimental strategies in mammalian genetics. ILAR J 43, 123 135 31. Peters LL, Barker JE (2001) Spontaneous and targeted mutations in erythrocyte membrane skeleton genes: mouse models of hereditary spherocytosis. In: Hematopoiesis A Developmental Approach, LI Zon (ed). (New York: Oxford University Press), pp 582 608 32. Peters LL, Barker JE (2005) Hematology, clotting, and thrombosis. MPD:62, Mouse Phenome Database web site at http://www.jax.org/phenome (The Jackson Laboratory, Bar Harbor, ME) 33. Peters LL, Swearingen RA, Andersen SG, Gwynn B, Lambert AJ, et al. (2004) Identification of quantitative
34.
35.
36.
37. 38. 39.
40.
41.
42.
43.
trait loci that modify the severity of hereditary spherocytosis in wan, a new mouse model of band-3 deficiency. Blood 103, 3233 3240 Peters LL, Zhang W, Lambert AJ, Brugnara C, Churchill GA, et al. (2005) Quantitative trait loci for baseline white blood cell count, platelet count, and mean platelet volume. Mamm Genome 16, 749 763 Platt OS, Falcone JF, Lux SE (1989) A 12 kD tryptic peptide of ankyrin retains spectrin binding capacity. Blood 74, 104a Platt OS, Brambilla DJ, Rosse WF, Milner PF, Castro O, et al. (1994) Mortality in sickle cell disease. Life expectancy and risk factors for early death. New Engl J Med 330, 1639 1644 Sen S, Churchill GA (2001) A statistical framework for quantitative trait mapping. Genetics 159, 371 387 Sokal RR, Rohlf FJ (1981) Biometry, 2nd edn. (San Francisco: WH Freeman) Sugiyama F, Churchill GA, Higgins DC, Johns C, Makaritsis KP, et al. (2001) Concordance of murine quantitative trait loci for salt-induced hypertension with rat and human loci. Genomics 71, 70 77 Wang X, Paigen B (2005a) Genetics of variation in HDL cholesterol in humans and mice. Circ Res 96, 27 42 Wang X, Paigen B (2005b) Genome-wide search for new genes controlling plasma lipid concentrations in mice and humans. Curr Opin Lipidol 16, 127 137 Wang X, Korstanje R, Higgins D, Paigen B (2004) Haplotype analysis in multiple crosses to identify a QTL gene. Genome Res 14, 1767 1772 Wannamethee G, Perry IJ, Shaper AG (1994) Haematocrit, hypertension and risk of stroke. J Intern Med 235, 163 168
