Quantitative trait loci for baseline erythroid traits

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

1The Jackson Laboratory, 600 Main Street, Bar Harbor, Maine 04609, USA2Department of Laboratory Medicine, Children�s Hospital, Harvard Medical School, Boston, Massachusetts 02115, USA

Received: 22 October 2005 / Accepted: 7 December 2005

Abstract

A substantial genetic contribution underlies varia-tion in baseline peripheral blood counts. We per-formed quantitative trait locus/loci (QTL) analysesto identify chromosome (Chr) regions harboringgenes influencing the baseline erythroid parametersin F2 intercrosses between NZW/LacJ, SM/J, andC57BLKS/J inbred mice. We identified multiple sig-nificant QTL for red blood cell (RBC) count, hemo-globin (Hgb) and hematocrit (Hct) levels, meancorpuscular volume (MCV), mean corpuscularhemoglobin (MCH), and mean cell hemoglobinconcentration (CHCM). We identified four RBCcount QTL: Rbcq1 (Chr 1, peak LOD score at62 cM,), Rbcq2 (Chr 4, 60 cM), Rbcq3 (Chr 11,34 cM), and Rbcq4 (Chr 10, 60 cM). Three MCVQTL were identified: Mcvq1 (Chr 7, 30 cM), Mvcq2(Chr 11, 6 cM), and Mcvq3 (Chr 10, 60 cM). Singlesignificant 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 exis-tence of a common RBC/MCH/MCV locus on Chr10. Two QTL for CHCM (Chcmq1, Chr 2, 48 cM;Chcmq2, Chr 9, 44 cM) and an interaction betweenChcmq2 with a locus on Chr 19 were identified.These analyses emphasize the genetic complexityunderlying the regulation of erythroid peripheralblood traits in normal populations and suggest thatgenes not previously recognized as significantlyimpacting normal erythropoiesis exist.

Introduction

A significant genetic component underlies baselineperipheral blood parameters (Chen and Harrison2002; Garner et al. 2000; Lin et al. 2005; Mahaneyet al. 2005; Peters et al. 2004). In epidemiologicstudies, baseline hematopoietic traits are significant,independent risk factors for human complex diseaseswith considerable morbidity and mortality. Thebaseline white blood cell (WBC) count correlateswith early mortality, heart disease, and stroke in thegeneral population. Increased hematocrit (Hct) levelis a significant and independent risk factor forcardiovascular disease (coronary heart disease, myo-cardial infarction), cerebrovascular disease, periph-eral vascular disease, and mortality (Gagnon et al.1994; Lippi et al. 2002; Wannamethee et al. 1994).The mean platelet volume (MPV) is associated withheart disease and stroke (Bath et al. 2004; Martin et al.1991, 1992). In sickle cell disease, increased baselinewhite blood cell (WBC) count is a strong predictor ofacute chest syndrome with an impact comparable tohemoglobin F levels and is a significant predictor ofinfarcts and hemorrhagic stroke as well (Castro et al.1994; de Labry et al. 1990; Kinney et al. 1999; Milleret al. 2000; Platt et al. 1994). Total hemoglobin (Hgb)is also a risk factor of sickle cell disease severity; highHgb is associated with higher risk for pain crises andacute chest syndrome, while low Hgb levels areassociated 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 animportant modulator of thalassemia and sickle celldisease (Lal and Vichinsky 2004; Platt et al. 1994). Wehave shown that modifiers of red cell mean corpus-cular volume (MCV) significantly impact the courseof inherited hemolytic anemia (hereditary spherocy-tosis) in mice (Peters et al. 2004).

From these examples it is clear that genes reg-ulating peripheral blood traits profoundly affect

Correspondence to: Luanne L. Peters; E-mail: [email protected]

298 DOI: 10.1007/s00335-005-0147-3 � Volume 17, 298�309 (2006) � � Springer Science+Business Media, Inc. 2006

disease progression. Identifying the primary geneticdeterminants of peripheral blood indices will notonly enhance our understanding of hematopoiesisbut also provide novel targets for risk assessment,diagnostics, and therapeutic intervention in hema-tologic and cardiovascular pathologies (Lal andVichinsky 2004; Platt et al. 1994). QTL analysis isan unbiased approach to the identification of novelgenes and their functions and can point towardpreviously unrecognized genetic regulatory path-ways. Concordance between QTL in rodents andhumans has been demonstrated for multiple com-plex traits and diseases. Thus, QTL analysis ininbred mice is highly significant and can identifygenes for subsequent meaningful, focused associa-tion studies in human populations (Paigen 2002;Sugiyama et al. 2001; Wang and Paigen 2005a,2005b).

We are using inbred mouse strains to identifyQTL for baseline hematologic parameters as a firststep in elucidating the genes that regulate thesetraits in normal populations. In this article we fo-cus on erythroid traits. In single-locus genome-wide scans we have identified four significant redblood cell (RBC) QTL, three significant MCV QTL,two significant mean cell hemoglobin concentra-tion (CHCM) QTL, and one QTL each for Hgb,Hct, and mean corpuscular hemoglobin (MCH).Moreover, multiple regression analysis revealsadditional significant loci for these traits. Severalsuggestive QTL are also apparent. Our analysesconfirm that multiple genes contribute to baselineerythropoietic traits and suggest that a network ofgenes not previously appreciated as significantlyimpacting hematopoiesis exist. The results reportedhere are a significant first step in identifying thosegenes.

Materials and methods

Animals. Mice were housed in humidity- and tem-perature-controlled rooms (12-h light cycle) with freeaccess to acidified water and food (NIH 5K52). TheJackson Laboratory Animal Care and Use Commit-tee approved all protocols. The Jackson Laboratoryis fully accredited by the American Associationfor Accreditation of Laboratory Animal Care(AAALAC).

Complete blood counts. Whole blood (275 ll)from 8-week-old adult mice was drawn from theretro-orbital sinus through EDTA-coated microhe-matocrit tubes directly into Eppendorf tubes con-taining 20 ll 20% EDTA in murine phosphate-buffered saline (PBS), as described previously (Peters

et al. 2004). Complete blood counts were determinedusing an Advia 120 multispecies whole-blood ana-lyzer (Bayer Corporation, Tarrytown, NY).

Statistical analysis. To identify QTL, two F2

intercrosses were established between F1 hybrids ofNZW/LacJ · SM/J (NZSM cross) and C57BLKS/J ·SM/J mice (KSSM cross). The parentals strains(NZW/LacJ, SM/J, C57BLKS/J) are considered prior-ity strains by the Mouse Phenome Project based ontheir wide use in the research community and theirgenetic diversity. (For a complete listing of prioritymouse strains, see the Mouse Phenome Database,www.jax.org/phenome.) Complete blood counts in186 F2 progeny from each cross were performed asdescribed above. Approximately equal numbers of F2

males and females were obtained; females accountedfor 57% and 49% of the NZSM and KSSM F2 progeny,respectively. Simple sequence length polymorphic(SSLP) markers spaced at 10�30-cM intervalsthroughout the genome were typed by the polymer-ase chain reaction (PCR) using multiplexed fluores-cent markers and ABI 3700 instrumentation(Applied Biosystems, Foster City, CA). A total of 88markers were typed in each of the two crosses(marker information available upon request). Gen-ome-wide scans were performed using log-trans-formed RBC count, Hgb and Hct values, MCV,MCH, and CHCM as quantitative traits. Resultswere analyzed in three stages, as described previ-ously (Sen and Churchill 2001). First, genome-widescans with sex and body weight as additive covari-ates were performed to detect single loci associatedwith each trait (main-effect QTL) using Pseudomar-ker software (www.jax.org/research/churchill)(Cordell et al. 1998). We also carried out single-locusscans that included sex as an interactive covariate toidentify sex-specific QTL (Korstanje et al. 2004).Significance thresholds were determined by permu-tation testing (n = 1000 permutations) (Churchilland Doerge 1994). A LOD (logarithm of the oddsratio) score of the 95th percentile or greater (p < 0.05)of the permutation distribution was considered sig-nificant; a score meeting or exceeding the 37th per-centile (p < 0.63) was considered suggestive (Landerand Kruglyak 1995). Confidence intervals (CI) weredetermined by computing the region of the posteriorprobability density curve containing 95% of the totalarea (Sen and Churchill 2001). The posterior proba-bility density is proportional to 10LOD and gives re-sults that are similar to the 1.5 ‘‘LOD supportinterval’’ but it is better justified on theoreticalgrounds (Sen and Churchill 2001).

In the second stage of the analysis, a simulta-neous search for pairs to detect epistatic interactions

L.L. PETERS ET AL.: QTL FOR ERYTHROID TRAITS 299

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 therelative spacing representative of the relative length of each chromosome. LOD scores are given on the y axis.

300 L.L. PETERS ET AL.: QTL FOR ERYTHROID TRAITS

was performed using a two-way analysis of variance(ANOVA) model (Sen and Churchill 2001; Sugiyamaet al. 2001). Third, to determine the combined effectsof all QTL on each trait, multiple regression analyseswere performed that included all significant andsuggestive QTL and all possible QTL interactions forthe trait. Terms that failed to meet significancelevels were eliminated one-by-one until all remain-ing QTL were significant, resulting in the finalmodel. The percent variance reported for each QTLderives from this analysis.

To narrow the CI for MCV QTL on Chr 7, datafrom both crosses were combined and analyzed asdescribed previously (Li et al. 2005). Other statisticalanalyses were performed using JMP 5.1.2 software(SAS Institute Inc., Cary, NC). Between-group com-parisons were analyzed by Tukey HSD test todetermine statistical significance.

Results

Identification of QTL. Because SM females are notthe best breeders, we set up unidirectional (NZ · SMand KS · SM) crosses to generate F1 progeny andintercrossed the F1 s to obtain F2 progeny for analy-sis. Originally, we established these crosses toidentify QTL for WBC count, platelet (Plt) count, andmean platelet volume (MPV) (Peters et al. 2005).These traits differed substantially between theparental strains. In this study we have extended ouranalysis to RBC traits using these same crosses.Notably, the parental strains do not differ signifi-cantly for any of the erythroid traits analyzed. [For

complete blood count data for all priority inbredstrains of mice, see Mouse Phenome Database,www.jax.org/phenome; Project MPD:62, Peters1(Peters and Barker 2005).] Hence, one would predictthat if any QTL were identified, the parental strainswould necessarily be contributing both high and lowalleles to the trait. This proved to be the case for alltraits analyzed in which more than one QTL wasidentified (see below for allele effects).

Genome-wide scans for single QTL are shown inFig. 1. Centimorgan (cM) positions are from theMouse Genome Informatics database (www.infor-matics.jax.org). For each cross 186 F2 animals wereanalyzed using 88 simple sequence length polymor-phism (SSLP) markers. The data for all significantQTL detected in this study are summarized inTable 1.

QTL for baseline RBC count. Analysis for sin-gle-locus QTL in both crosses using sex and bodyweight as additive covariates revealed significant locifor RBC count on chromosomes (Chr) 1, 4, 10, and 11(Fig. 1). The peak LOD score and the posterior prob-ability density curves used to define the 95% CIs forall RBC QTL are shown in Fig. 2 and summarized inTable 1. No influence of sex was detected for any ofthe RBC QTL, as determined by repeating the anal-ysis with sex as an interacting covariate (no signifi-cant differences in LOD scores were seen between thetwo scans). A simultaneous search to detect inter-acting gene pairs (epistasis) was negative. By con-vention, these QTL were named Rbcq1�Rbcq4 (redblood cell quantitative loci 1�4) (Abiola et al. 2003).

Table 1. QTL for baseline erythroid traits

Name Trait ChrPeak, cM(Mb)

95% CI(cM)

Humanlocation High allele Inheritance Peak marker LOD

Rbcq1 RBC 1 62 (115) 34�74 18q21 SM/J Recessive D1Mit308 3.4Rbcq2 RBC 4 60 (131) 50�66 1p34-p36 SM/J Recessive D4Mit170 5.3Rbcq3 RBC 11 34 (60) 18�56 17p11, 1q41-q42 SM/J Dominant D11Mit30 3.8Rbcq4 RBC 10 60 (112) 40�60 12q21 C57BLKS/J Dominant D10Mit150 3.7Hgbq1 Hgb 16 32 (44) 20�42 3q12-q13 SM/J Dominant D16Mit60 3.7Hctq1 Hct 3 42 (89) 30�62 1q21 NZW/LacJ Recessive D3Mit98 3.9Mchq1 MCH 10 60 (112) 48�60 12q21 SM/J Recessive D10Mit150 4.4Mcvq1a MCV 7 30 (50) 26�38 15q14 SM/J Recessive D7Mit83 9.8Mcvq2 MCV 11 6 (12) 0�34 7p13-p11 NZW/LacJ

and SM/JHeterozygotes

lowbD11Mit2 4.7

Mcvq3 MCV 10 60 (112) 48�60 12q21 SM/J Recessive D10Mit15 4.7Chcmq1 CHCM 2 48 (80) 30�60 2q32 C57BLKS/J Recessive D2Mit300 4.7Chcmq2 CHCM 9 44 (80) 40�66 6p12, 6q13 SM/J Dominant D9Mit198 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/sear-ches/linkmap_form.shtml) and Ensemble Mouse Genome Browser (Build 34) (http://www.ensembl.org/Mus_musculus/).aData derived from combined cross analysis.bFor QTL for which heterozygotes are low, we hypothesize that there may be closely linked QTL with effects in the opposite direction.


Allele effects. Allele effects were determined bycalculating the phenotypic mean at the peak markerfor each of the three possible genotypes (Lyons et al.2003). We then determined which strain contributedthe allele that increased the trait and whether it actedin a dominant, additive (codominant), or recessivemanner. In the NZSM cross, recessive SM alleles atthe Rbcq1 and Rbcq2 loci increase RBC count, whilea dominant SM allele increases RBC count at Rbcq3(Fig. 2). In the KSSM cross, a dominant KS allele atRbcq4 increases RBC count (Fig. 2).

Combined effects of RBC QTL. Single-locusgenome scans as displayed in Fig. 1 do not considerother QTL. Multiple regression models, however,allow a QTL effect to be assessed in the context of allother QTL. We performed multiple regression anal-ysis integrating all suggestive and significant RBCQTL and all possible RBC QTL interactions todetermine the contribution of each QTL to the totalvariance. The regression analysis predicts that a Chr8 QTL, which was suggestive only in the main scan(Fig. 1), is significant in the final model for the KSSM

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. Resultsare based on single-locus scans of 186 F2 animals. (E�H) Effect plots for RBC QTL. Allelic contributions determined at thepeak marker for Rbcq1�4. The three possible genotypes are indicated on each ordinate. K, C57BLKS/J allele; N, NZW/LacJallele; S, SM/J allele. The y axis represents RBC counts. Error bars represent SEM. *p < 0.05, ns = not significant.


cross (Table 2). With Rbcq4, this QTL accounted for15.8% of the variance in RBC count in this cross. Inthe NZSM cross, Rbcq1�3 accounted for 27.1% of

the variance in RBC count. No influence of sex wasdetected for any of these RBC QTL, and a simulta-neous search to detect interacting pairs was nega-

Table 2. Multiple regression analysis of RBC QTL

df % Variance F value P(f)

KSSM crossSex 1 0.43 0.909 0.341619 (ns)Rbcq4 2 8.4 9.011 0.000186 (p < 0.001)Chr 8@26 cM 2 6.8 7.29 0.000908 (p < 0.001)Total 15.8

NZSM crossSex 1 1.3 3.09 0.076030 nsRbcq1 2 3.8 4.63 0.010987 (p < 0.05)Rbcq2 2 10.6 12.83 6.27 · 10)6 (p < 0.001)Rbcq3 2 8.8 10.68 4.20 · 10)5 (p < 0.001)Total 27.1

df = degrees of freedom; ns = not significant.aPercent variances are based on type III sums of squares (SS). The adjusted SS used to compute the % variance takes into account othermajor QTLs in the population and thus does not necessarily sum up to the total variance explained. A discussion can be found in Sokaland Rohlf (1981).

Fig. 3. (A�C) Genome-wide scans (solid lines) and posterior probability densities (dashed lines) for Hgbq1, Hctq1, andMchq1. 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 forHgbq1, 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.


tive. This was also true for all other traits analyzedin this study.

Identification of QTL for Hgb, Hct, andMCH. Genome-wide scans revealed a single signif-icant QTL for each of these traits: Hgbq1 (hemo-globin quantitative locus-1) on Chr 16, Hctq1(hematocrit quantitative locus-1) on Chr 3, andMchq1 (mean corpuscular hemoglobin quantitativelocus-1) on Chr 10 (Fig. 1, Table 1). A dominant SMallele at Hgbq1 increases hemoglobin levels; arecessive NZ allele increases the Hct at Hctq1, and arecessive SM allele increases the MCH at Mchq1(Fig. 3, Table 1). Regression analysis identified asecond significant hematocrit QTL on Chr 6(Table 3). Hgbq1 accounted for 10.7% of the traitvariance, Mchq1 10.3%, and Hctq1/Chr 6 15.5%(Table 3).

Identification of QTL for MCV. Three signifi-cant 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 tookadvantage of a statistical method, combined crossanalysis, to significantly narrow the 95% CI (Fig. 4)(Li et al. 2005). Recessive SM alleles at Mcvq1 andMcvq3 increase red cell volume, while a recessiveNZ allele does so at Mcvq2 (Fig. 4). Multipleregression did not reveal any additional significantMCV QTL. The MCV QTL identified accounted forapproximately 25% of the total variance of the trait(Table 4).

Identification of QTL for CHCM. Like theMCV, CHCM is directly measured on a cell-by-cellbasis on the Advia 120 analyzer, i.e., it is not a

calculated parameter. CHCM is analogous to thecalculated parameter MCHC (mean corpuscularhemoglobin content). We identified two significantCHCM QTL in our single-locus genome-wide scans:Chcmq1 on Chr 2 and Chcmq2 on Chr 9 (Figs. 1 and5). A recessive KS allele increases CHCM atChcmq1, and a dominant SM allele at Chcmq2(Fig. 5) increases CHCM. Multiple regression analy-sis revealed a third significant CHCM QTL on Chr19 (Table 5). Interestingly, the Chr 19 locus inter-acted with Chcmq2; homozygosity for the Chr 19 KSallele significantly increases CHCM when Chcmq2is also homozygous for the KS allele, but otherwise ithas no effect (Fig. 5). In total, the CHCM QTL ac-count for 44% of the variance (Table 5).

Discussion

QTL analysis in mice is a powerful, unbiased ap-proach to the identification of novel genes andpathways influencing complex phenotypes. We re-cently reported QTL for WBC count, Plt count, andMPV (Peters et al. 2005), and another recent studyidentified loci associated with baseline Plt counts inmice (Cheung et al. 2004), but identification of QTLfor baseline blood counts in mice has otherwise re-ceived scant attention. No studies in mice to localizeQTL for steady-state erythroid traits, important riskfactors for heart disease and stroke, have been re-ported to our knowledge. Likewise, analysis of ery-throid traits as a complex phenotype in humans isnewly emerging. Recent QTL analysis using theFramingham Heart cohort has identified a signifi-cant QTL for Hct on 6q23-24 (Lin et al. 2005). Asuggestive Hct QTL was also found on human Chr 1(Lin et al. 2005). The positions of the loci identifiedin humans thus far, with the exception of the sug-gestive locus on Chr 1, do not correspond to the

Table 3. Multiple regression analysis of Hgb, Hct, and MCH QTL


KSSM crossSex 1 4.1 8.32 0.004389 (p < 0.01)Hgbq1 2 7.4 7.90 0.000515 (p < 0.001)Total 10.7

Sex 1 1.3 2.76 0.098210 (ns)Mchq1 2 10.3 10.55 4.62 · 10)5 (p < 0.001)Total 11.4

NZSM crossSex 1 0.52 1.1 0.295779 (ns)Hctq1 2 9.09 9.51 0.000112 (p < 0.001)Chr 6@60 cM 2 6.20 6.54 0.001824 (p < 0.01)Total 15.5

ns = not significant.


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), indicatedby the gray bars. Position along the chromosome is shown on the ordinates. The y axis shows LOD scores. Results arebased on single-locus scans of 186 F2 animals. (E) Genome wide scan (solid line) and posterior probability density (dashedline) 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.


human syntenic positions of our loci (Table 3),confirming our suspicion that multiple QTL forbaseline peripheral blood traits exist and many re-main to be identified.

In the current study using two crosses betweeninbred strains of mice, we identified multiple QTLinfluencing baseline erythroid traits—RBC count,Hgb, Hct, MCV, MCH, and CHCM (summarized inTable 1). A common QTL for RBC count, MCH, andMCV was identified on Chr 10, suggesting that theserelated parameters are under common geneticinfluences. Independent loci influencing each ofthese traits were also identified. These data reinforcethe concept that steady-state peripheral blood valuesare truly complex traits. Given that these parametersare significantly associated with common diseases inman, elucidating the underlying genetic networksinfluencing these traits is critical.

The CIs for our QTL are quite large, not unex-pected in a ‘‘first pass’’ analysis. The smallestinterval, Mcvq1, owes its reduced size to combinedcross analysis, a statistical method developed by Liet al. (2005). As its name implies, combined crossanalysis considers the phenotypic information frommultiple crosses in a single QTL analysis to provideincreased recombination and, thus, statistical power,allowing one to detect QTL with small effects, re-solve closely linked QTL, and narrow QTL confi-dence intervals compared with individual QTLcrosses. The more crosses combined, the more powerachieved. Clearly, a need exists to establish addi-tional mouse crosses in order to effectively applynew statistical and bioinformatic methods (e.g.,haplotype analysis) (Wang et al. 2004) to the study ofhematopoiesis as a complex trait and begin unrav-eling the regulatory genetic networks.

Given the current 95% CIs, examination of po-tential candidate genes is currently unwieldy at best.For all QTL intervals (Table 1), we examined the

databases for potential candidate genes. Becausemost the 95% CIs are quite large, we confined ouranalysis to 10 Mb of 5¢ and 3¢ sequence flanking theQTL peak (20 Mb total) using the Ensembl mouseserver (http://www.ensembl.org/Mus_musculus/in-dex.html). Interestingly, the genes not presentwithin our intervals are the most notable. Forexample, EPO, EPOR, and other growth factors andtranscription factors (e.g., GATA, EKLF, SCL) knownto influence erythroid expansion, lineage commit-ment, or differentiation (Cantor and Orkin 2001,2002, 2005; Orkin 2000; Orkin and Zon 2002; Orkinet al. 1999) are not present. In some cases anintriguing candidate was observed but could beeliminated based on expression data and/or knock-out phenotyping data. For example, Csf2 [colonystimulating factor 2 (granulocyte-macrophage)], lo-cated near the Rbcq3 peak, is not expressed beyondthe median in erythroid cells or tissues according tothe Novartis microarray databases (http://symat-las.gnf.org/SymAtlas/), and the knockout phenotypeis one of lung pathology with no evidence of ery-throid involvement [Mouse Genome Informatics(MGI), http://www.informatics.jax.org/). For Rbcq3,the gene encoding the erythroid membrane skeletoncomponent protein 4.1 (Epb4.1) is located near thepeak. Protein 4.1 is critical to red cell stability (Pe-ters and Barker 2001) and thus represents a viablecandidate gene at this locus.

Similar extensive in silico ‘‘mining’’ of our otherQTL intervals failed to identify obvious, strongcandidates, suggesting that a complex regulatorynetwork consisting of multiple genes not previouslyrecognized as significantly impacting normal eryth-ropoiesis indeed exists.

We are establishing additional crosses to furtherexplore QTL for baseline hematologic parameters inmice. The parental strains used in the current anal-ysis are members of the 40 ‘‘priority’’ strains of the

Table 4. Multiple regression analysis of MCV QTL


KSSM crossSex 1 1.25 2.97 0.86709 (ns)BW 1 1.94 4.59 0.033555 (p < 0.05)Mcvq1 2 6.77 8.03 0.000460 (p < 0.001)Mcvq3 2 8.23 9.75 9.57 · 10)5 (p < 0.001)Total 24.5

NZSM crossSex 1 0.08 0.18 0.670353 (ns)BW 2 0.76 4.14 0.043413 (p < 0.05)Mcvq1 2 9.50 11.15 2.76 · 10)5 (p < 0.001)Mcvq2 2 8.75 10.26 6.08 ·10)5 (p < 0.001)Total 24.6

ns = not significant.


Fig. 5. Genome-wide scans (solid lines), posterior probability densities (dashed lines), and effect plots for significantCHCM 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 yaxis shows LOD scores. Results are based on single-locus scans of 186 F2 animals. (C, D) Effect plots. The three possiblegenotypes are indicated on each ordinate. K, C57BLKS/J allele; S, SM/J allele. The Y axis represents CHCM counts. Errorbars represent SEM. *p < 0.05, ns = not significant. (E) Interaction between Chcmq2 and a locus on Chr 19. The effects ofgene interaction predicted by multiple regression analysis are shown. The three possible genotypes for Chcmq2 areindicated 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 crossSex 1 1.429 4.415 0.037082 (p < 0.05)BW 1 1.618 4.996 0.026679 (p < 0.05)Chcmq1 2 7.661 11.829 1.53e-005 (p < 0.001)Chcmq2 6 16.812 8.653 3.08e-008 (p < 0.001)Chr 19@32 6 8.207 4.224 0.000540 (p < 0.001)Chr 9@44:Chr 19@32 4 5.302 4.093 0.003398 (p < 0.01)Total variance 43.982


Mouse Phenome Project, chosen on the basis of theirwide use and genetic diversity, and share ancestrywith other widely used mouse strains. Thus, analy-sis of haplotypes using publicly available high-den-sity SNP data can be used to identify other strainsthat are likely to share alleles. Our ultimate goal isto identify overlapping QTL in multiple crosses toprovide the framework for the application of new,powerful statistical and bioinformatic techniques tosignificantly narrow QTL intervals and, ultimately,identify the underlying genes (Li et al. 2005; Wang etal. 2004). Such studies will provide critical focalpoints for clinical interventions.

Acknowledgments

The authors thank Drs. Beverly Paigen and DavidHarrison for critical review of the manuscript. Thiswork was supported by National Institutes of Healthgrants 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, BelknapJK, et al. (2003) The nature and identification ofquantitative trait loci: a community�s view. Nat RevGenet 4, 911�916

2. Bath P, Algert C, Chapman N, Neal B (2004) Associa-tion of mean platelet volume with risk of strokeamong 3134 individuals with history of cerebrovascu-lar disease. Stroke 35, 622�626

3. Cantor AB, Orkin SH (2001) Hematopoietic develop-ment: a balancing act. Curr Opin Genet Dev 11,513�519

4. Cantor AB, Orkin SH (2002) Transcriptional regulationof erythropoiesis: an affair involving multiple partners.Oncogene 21, 3368�3376

5. Cantor AB, Orkin SH (2005) Coregulation of GATAfactors by the Friend of GATA (FOG) family of mul-titype 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 insickle cell disease: incidence and risk factors. TheCooperative Study of Sickle Cell Disease. Blood 84,643�649

7. Chen J, Harrison DE (2002) Quantitative trait lociregulating relative lymphocyte proportions in mouseperipheral blood. Blood 99, 561�566

8. Cheung CC, Martin IC, Zenger KR, Donald JA,Thomson PC, et al. (2004) Quantitative trait loci forsteady-state platelet count in mice. Mamm Genome15, 784�797

9. Churchill GA, Doerge RW (1994) Empirical thresholdvalues for quantitative trait mapping. Genetics 138,963�971

10. Cordell HJ, Todd JA, Lathrop GM (1998) Mappingmultiple linked quantitative trait loci in non-obesediabetic 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 ofmortality: results over 18 years from the NormativeAging Study. J Clin Epidemiol 43, 153�157

12. Gagnon DR, Zhang TJ, Brand FN, Kannel WB (1994)Hematocrit and the risk of cardiovascular disease—theFramingham study: a 34-year follow-up. Am Heart J127, 674�682

13. Garner C, Tatu T, Reittie JE, Littlewood T, Darley J,et al. (2000) Genetic influences on F cells and otherhematologic variables: a twin heritability study. Blood95, 342�346

14. Kinney TR, Sleeper LA, Wang WC, Zimmerman RA,Pegelow CH, et al. (1999) Silent cerebral infarcts insickle cell anemia: a risk factor analysis. The Coop-erative 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 quantitativetrait loci for HDL cholesterol levels in an SM/J byNZB/BlNJ intercross population. J Lipid Res 45,881�888

16. Lal A, Vichinsky E (2004) The role of fetal hemoglobin-enhancing agents in thalassemia. Semin Hematol 41,17�22

17. Lander E, Kruglyak L (1995) Genetic dissection ofcomplex traits: guidelines for interpreting and report-ing linkage results. Nat Genet 11, 241�247

18. Li R, Lyons MA, Wittenburg H, Paigen B, ChurchillGA (2005) Combining data from multiple inbred linecrosses improves the power and resolution of quanti-tative trait loci mapping. Genetics 169, 1699�1709

19. Lin JP, O�Donnell CJ, Levy D, Cupples LA (2005)Evidence for a gene influencing haematocrit on chro-mosome 6q23�24: genomewide scan in the Framing-ham Heart Study. J Med Genet 42, 75�79

20. Lippi G, Manzato F, Franchini M, Guidi G (2002)Relationship between hematocrit, primary hemosta-sis, and cardiovascular disease in athletes. Am Heart J144, E14

21. Lyons MA, Wittenburg H, Li R, Walsh KA, LeonardMR, et al. (2003) New quantitative trait loci thatcontribute to cholesterol gallstone formation detectedin an intercross of CAST/Ei and 129S1/SvImJ inbredmice. Physiol Genomics 14, 225�239

22. Mahaney MC, Brugnara C, Lease LR, Platt OS (2005)Genetic influences on peripheral blood cell counts: astudy in baboons. Blood 106, 1210�1214

23. Martin JF, Bath PM, Burr ML (1991) Influence ofplatelet size on outcome after myocardial infarction.Lancet 338, 1409�1411


24. Martin JF, Bath PM, Burr ML (1992) Mean plateletvolume and myocardial infarction. Lancet 339,1000�1001

25. Miller ST, Sleeper LA, Pegelow CH, Enos LE, WangWC, et al. (2000) Prediction of adverse outcomes inchildren 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 insickle cell disease: Rates and risk factors. Blood 91,288�294

27. Orkin SH (2000) Diversification of haematopoieticstem cells to specific lineages. Nat Rev Genet 1,57�64

28. Orkin SH, Zon LI (2002) Hematopoiesis and stemcells: 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 developmentand leukemia genes. Cancer Res 59, 1784S�1787S

30. Paigen K (2002) Understanding the human condition:experimental strategies in mammalian genetics. ILARJ 43, 123�135

31. Peters LL, Barker JE (2001) Spontaneous and targetedmutations in erythrocyte membrane skeleton genes:mouse models of hereditary spherocytosis. In: Hema-topoiesis A Developmental Approach, LI Zon (ed).(New York: Oxford University Press), pp 582�608

32. Peters LL, Barker JE (2005) Hematology, clotting, andthrombosis. MPD:62, Mouse Phenome Database website at http://www.jax.org/phenome (The JacksonLaboratory, Bar Harbor, ME)

33. Peters LL, Swearingen RA, Andersen SG, Gwynn B,Lambert AJ, et al. (2004) Identification of quantitative

trait loci that modify the severity of hereditary sphe-rocytosis in wan, a new mouse model of band-3 defi-ciency. Blood 103, 3233�3240

34. Peters LL, Zhang W, Lambert AJ, Brugnara C, Chur-chill GA, et al. (2005) Quantitative trait loci for base-line white blood cell count, platelet count, and meanplatelet volume. Mamm Genome 16, 749�763

35. Platt OS, Falcone JF, Lux SE (1989) A 12 kD trypticpeptide of ankyrin retains spectrin binding capacity.Blood 74, 104a

36. Platt OS, Brambilla DJ, Rosse WF, Milner PF, CastroO, et al. (1994) Mortality in sickle cell disease. Lifeexpectancy and risk factors for early death. New Engl JMed 330, 1639�1644

37. Sen S, Churchill GA (2001) A statistical framework forquantitative trait mapping. Genetics 159, 371�387

38. Sokal RR, Rohlf FJ (1981) Biometry, 2nd edn. (SanFrancisco: WH Freeman)

39. Sugiyama F, Churchill GA, Higgins DC, Johns C,Makaritsis KP, et al. (2001) Concordance of murinequantitative trait loci for salt-induced hypertensionwith rat and human loci. Genomics 71, 70�77

40. Wang X, Paigen B (2005a) Genetics of variation in HDLcholesterol in humans and mice. Circ Res 96, 27�42

41. Wang X, Paigen B (2005b) Genome-wide search fornew genes controlling plasma lipid concentrations inmice and humans. Curr Opin Lipidol 16, 127�137

42. Wang X, Korstanje R, Higgins D, Paigen B (2004)Haplotype analysis in multiple crosses to identify aQTL gene. Genome Res 14, 1767�1772

43. Wannamethee G, Perry IJ, Shaper AG (1994) Haemat-ocrit, hypertension and risk of stroke. J Intern Med235, 163�168


Quantitative trait loci for baseline erythroid traits

Documents

Transcript of Quantitative trait loci for baseline erythroid traits