L E T T E R S

The global health impact of malaria is enormous, with anestimated 300–500 million clinical cases and 1 million annualdeaths1. In humans, initial susceptibility to infection withPlasmodium species, disease severity and ultimate outcome ofmalaria (self-healing or lethal) are under complex geneticcontrol. Alleles associated with sickle cell anemia, β-thalassemia and deficiency in glucose-6-phosphatedehydrogenase have a protective effect against malaria and mayhave been retained by positive selection in areas of endemicmalaria2. Likewise, genetic variations in erythrocyte antigensand levels of host cytokines affect type and severity ofdisease3,4. A mouse model of infection with Plasmodiumchabaudi was used to study the genetic component of malariasusceptibility. Segregation analyses in informative F2 crossesderived from resistant C57BL/6J and susceptible A/J, C3H andSJL strains using extent of blood stage replication of the parasiteand survival as traits mapped three P. chabaudi resistance(Char) loci on chromosomes 9 (Char1), 8 (Char2) and 17(Char3, MHC-linked)5–7. Recombinant congenic strains AcB55and AcB61 are unusually resistant to malaria despite carryingsusceptibility alleles at Char1 and Char2. Malaria resistance inAcB55 and AcB61 is associated with splenomegaly andconstitutive reticulocytosis, is inherited in an autosomalrecessive fashion and is controlled by a locus on chromosome 3(Char4)8. Sequencing of candidate genes from the Char4 regionidentified a loss-of-function mutation (269T→A, resulting in theamino acid substitution I90N) in the pyruvate kinase gene (Pklr)that underlies the malaria resistance in AcB55 and AcB61.These results suggest that pyruvate kinase deficiency maysimilarly protect humans against malaria.

Recombinant congenic strains AcB55 and AcB61 derived ∼ 12.5% oftheir genome from C57BL/6 (B6) and ∼ 87.5% from A/J, fixed onchimeric chromosomes. Despite carrying A/J susceptibility alleles atChar1 and Char2, AcB55 and AcB61 are resistant to malaria, as shownby low levels of parasitemia at the peak of infection compared withlevels observed in C57BL/6 and A/J controls, early clearance of theparasite and no mortality (Fig. 1a,b). Linkage analysis in (AcB55 ×A/J) F2 mice using a partial genome scan for B6 chromosomal seg-ments fixed in AcB55 located a B6-derived locus (Char4, lod score =6.57, ∼ 14% of variance) on chromosome 3 that controls blood para-

sitemia, with additional suggestive linkage detected on chromosome10 (lod score = 2.35)9. A similar analysis of an (AcB61 × A/J) F2 popu-lation did not detect an effect of B6 chromosome segments in AcB61on the level of parasitemia8.

Macroscopic examination at the time of necropsy showed signifi-cant splenomegaly in AcB55 and AcB61 strains (spleen index =∼ 1.2) compared to A/J and B6 (spleen index = ∼ 0.5–0.6) controls.Because the erythroid compartment is a key site of Plasmodiumreplication in vivo, we used transcriptional profiling with spleenRNA to investigate possible gene expression profiles specific toAcB55 and AcB61 that could indicate cell type or biochemical path-ways underlying splenomegaly in these two strains (Fig. 2). Weobserved changes in transcript abundance in both strains relative toA/J (Fig. 2 and Supplementary Table 1 online). Two overlappinggroups of transcripts had greater expression in AcB55 and AcB61:(i) erythroid-specific genes or proteins involved in iron metabolismand (ii) genes involved in cell cycle, DNA replication and proteinsynthesis. Hematological analyses of peripheral blood showed asmaller number of erythrocytes in AcB55 (8.2 × 109 per ml) andAcB61 (7.2 × 109 per ml) strains compared with B6 and A/J controls(9.8–10 × 109 per ml) but also showed markedly higher reticulocytenumbers in AcB55 and AcB61 strains (∼ 21% versus <4% in con-trols). Thus, hematological analyses and transcriptional profilingstudies suggest increased erythropoietic activity as a common phe-notypic feature of AcB55 and AcB61 mice.

Segregation analyses in (AcB55 × A/J) and (AcB61 × A/J) F1 and F2crosses (Fig. 1c–e) showed that, unlike peak parasitemia after P. chabaudi infection, which segregates as a complex trait8, reticulo-cytosis segregates as a simple trait. Indeed, reticulocytosis is inher-ited as a recessive trait, which is controlled by a single autosomal andfully penetrant locus in both strains, with no detectable sex effect.(AcB55 × AcB61) F1 mice doubly heterozygous with respect to thelocus controlling reticulocytosis had elevated reticulocyte counts(7–16%; Fig. 1e), suggesting that both strains carry mutations eitherin the same gene or in the same biochemical pathway.

To understand the relationship between reticulocytosis (simpletrait) and resistance to malaria (complex trait controlled in part byChar4) in AcB55 mice, we mapped the reticulocytosis locus. We firstexamined the chromosome 3 Char4 region in (AcB55 × A/J) F2 mice,including two distal B6-derived congenic segments defined byD3Mit109 (61.8 cM) and D3Mit44 (78.5 cM; Fig. 3a). We readily

1Department of Biochemistry, McGill University, Montreal, QC H3G 1Y6, Canada. 2Research Institute of the McGill University Health Center, Montreal GeneralHospital, 1650 Cedar Ave, Montreal, QC H3G 1A4, Canada. 3Biotechnology Research Institute, National Research Council, 6100 Royalmount, Montreal, QC H4P 2R2, Canada. Correspondence should be addressed to P.G. (philippe.gros@staff.mcgill.ca)

Published online 2 November 2003; doi:10.1038/ng1260

Pyruvate kinase deficiency in mice protects against malariaGundula Min-Oo1, Anny Fortin1, Mi-Fong Tam2, André Nantel3, Mary M Stevenson2 & Philippe Gros1

NATURE GENETICS VOLUME 35 | NUMBER 4 | DECEMBER 2003 357

©20

03 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

reg

enet

ics

L E T T E R S

detected linkage of reticulocytosis to the chromosome 3 Char4region in (AcB55 × A/J) F2 mice, but the noted 18% and 28% recom-bination between reticulocytosis and D3Mit109 and D3Mit44,respectively (Fig. 3b), indicates that the locus is in fact proximal toD3Mit109. The locus either maps to a small, B6 congenic segmentthat escaped detection in the initial genotyping of AcB55 (ref. 10) orcorresponds to a new mutation on the A/J portion of chromosome 3.High-resolution genotyping of AcB55 with additional chromosome3 informative markers did not detect B6-derived sequences proximalto D3Mit109 (data not shown). Additional linkage studies in a fullyinformative (AcB55 × DBA/2) F2 cross subsequently mapped thereticulocytosis locus between D3Mit75 and D3Mit22, tightly linked(position 41 cM) to D3Mit175 (Fig. 3c,d). Reticulocytosis is caused

by homozygosity with respect to AcB55-derived A/J alleles atD3Mit175, which are inherited in a fully recessive fashion in theAcB55 × DBA/2 cross, suggesting a new mutation in AcB55 probablyoccurring during its derivation.

A search of public sequence databases for erythroid-specific genesnear D3Mit175 identified the liver- and red cell–specific pyruvatekinase (Pklr, 0.5 cM from D3Mit175) as a strong positional candi-date. Pklr, which catalyzes the transformation of phosphoenolpyru-vate into pyruvate, is essential for ATP production in erythrocytes,and mutations at PKLR cause hemolytic anemia in humans11.Sequence analysis (Fig. 4a) identified an isoleucine-to-asparaginesubstitution at residue 90 of the Pklr protein (caused by an T→Asubstitution at position 269 in Pklr) in AcB55 mice. We found the

same mutation in AcB61 mice (Fig. 4b). ThePKLR I90N mutation has also beendescribed in a human case of pyruvate kinasedeficiency12. Additional studies with SNPs(data not shown) confirmed that the

358 VOLUME 35 | NUMBER 4 | DECEMBER 2003 NATURE GENETICS

Figure 1 Blood-stage replication of P. chabaudiin AcB55 and AcB61 mice and segregationanalysis of reticulocytosis. Parasitemia values at the peak of infection (peak parasitemia,expressed as a percentage of pRBCs) are shown for strains AcB55 and AcB61 comparedwith parental strains for males (a) and females(b) after intraperitoneal infection with 106

erythrocytes infected with P. chabaudi. Eachcircle represents one mouse, and filled circlesrepresent mice that succumbed to infection.Averages for each group are indicated byhorizontal bars. F1 and F2 mice derived fromAcB55 × A/J (c) or from AcB61 × A/J (d) crosseswere generated, and the number of reticulocytesin individual mice (percentage of total circulatingred cells) was determined. Each dot represents amouse. Mice from an AcB55 × AcB61 F1 crosswere similarly tested (e).

a b

c d e

10.2 0.3 0.4 0.5 0.6 0.8 2 3 4 5 6

1

0.2

0.3

0.4

0.6

0.8

2

3

4

56

AcB61

AcB55

AcB55 AcB61

567 289162

a b

Figure 2 Differentially expressed transcripts in thespleens of AcB55 and AcB61 mice. Total cellularRNA from the spleens of malaria-resistant AcB55or AcB61 strains was compared with transcriptsfrom malaria-susceptible A/J controls using 15kspotted cDNA microarrays. (a) The scatter plotrepresents the average fluorescence ratiosobtained from 6–8 independent pair-wisehybridizations. Transcripts that are significantlymodulated (by a factor of 1.5, FDR < 0.05%) inAcB55 are colored in red, and AcB61-modulatedgenes are colored in green. Genes that weremodulated in both strains are colored in yellow,and the 14,232 genes that did not pass thestatistical cutoff are colored in gray. The numberof genes differentially expressed between the threemouse strains is also shown in a Venn diagram (b).

©20

03 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

reg

enet

ics

L E T T E R S

Pklr269A allele corresponds to a single mutation on the A/J back-ground that became fixed in AcB55 and AcB61. In AcB55 and AcB61,splenomegaly, constitutive erythropoiesis and reticulocytosis arecompensatory reactions to hemolytic anemia caused by pyruvatekinase deficiency.

We investigated the relationship between susceptibility to malariaand reticulocytosis caused by Pklr269A (Fig. 5 and Table 1). First, weanalyzed the association between reticulocyte numbers and peak par-asitemia levels after infection in (AcB55 × A/J) F2 mice (Fig. 5a). Micewith the lowest parasitemia (≤34% parasitized red blood cells(pRBCs), n = 16) all had high reticulocyte counts (>10%), whereasmice with the highest parasitemia (>58% pRBCs, n = 35) showedreticulocyte counts <5%. Second, using a SfaNI polymorphismcaused by the I90N mutation (Fig. 5b), we observed that homozygos-ity with respect to Pklr269A was associated with reduced peak para-sitemia in (AcB55 × A/J) F2 mice, from 51.6% (in homozygouswild-type and heterozygous mice) to 38.4% of pRBCs (t = 7.77,degrees of freedom (d.f.) = 169, P < 0.0001). We noted a similar pro-tective effect of Pklr269A homozygosity in (AcB61 × A/J) F2 mice(46.6% versus 56.9% pRBCs in homozygous wild-type and heterozy-

gous mice; t = 5.98, d.f. = 193, P < 0.0001; Fig. 5c). Third, homozy-gosity with respect to Pklr269A was associated with significantly lessmortality after P. chabaudi infection, from 35% to 8.5% in the AcB55× A/J cross (χ2 = 11.9, d.f. = 1, P < 0.001) and from 65% to 15.5% inthe AcB61 × A/J cross (χ2 = 31.7, d.f. = 1, P < 0.001; Fig. 5d). Theseresults strongly suggest that loss of function at Pklr (Pklr269A) in miceprotects against malaria.

A genetic approach in the mouse to characterize naturally occur-ring interstrain differences in disease susceptibility has proven tobe a valuable tool in identifying genes that function in defensesagainst bacterial13–15, viral16,17 and parasitic infections18,19.Although formal proof that Char4 and Pklr are allelic awaits cre-ation of gain- or loss-of-function mutations at Pklr in transgenicmice, several lines of evidence suggest that this is indeed the case:(i) Char4 and Pklr269A map to chromosome 3 and are derived fromthe AcB55 chromosome; (ii) Char4-controlled resistance to malariaand Pklr269A-induced reticulocytosis have the same mode of inher-itance; (iii) Pklr269A affects biochemical activity and life span of acell type that is the primary site for Plasmodium replication in vivo;(iv) homozygosity with respect to Pklr269A causes a significant

NATURE GENETICS VOLUME 35 | NUMBER 4 | DECEMBER 2003 359

Table 1 Pklr269A associated with malaria resistance

Number Percentage of Statistical StatisticalCross Genotypea of mice reticulocytesb Mortalityc significance Peak parasitemiab,d significance

(AcB55 × A/J) F2 +/+ and +/– 124 1.61 ± 0.13 34% χ2 = 11.9, d.f. = 1, 51.58 ± 0.86 t = 7.77, d.f. = 169,P < 0.001 P < 0.0001

–/– 47 17.78 ± 0.65 8.5% 38.39 ± 1.56

(AcB61 × A/J) F2 +/+ and +/– 150 1.53 ± 0.11 64.8% χ2 = 31.7, d.f. = 1, 56.87 ± 0.79 t = 5.98, d.f. = 193,P < 0.001 P < 0.0001

–/– 45 17.54 ± 0.46 15.5% 46.57 ± 1.70

a+/+, homozygous wild-type mice; +/–, heterozygous mice; –/–, homozygous Pklr269A mutant mice. Homozygous wild-type mice and heterozygous mice were grouped together in theanalysis. bValues given are mean ± s.e.m. cMortality is given as the percentage of mice that succumbed to infection. dPeak parasitemia is expressed as the percentage of pRBCs.

Figure 3 Linkage mapping of the monogenic traitcontrolling reticulocytosis in the AcB55 strain.(a) Schematic representation of chromosome 3from AcB55, including the position ofinformative microsatellite markers used forgenetic linkage studies. The position of eachmarker (in cM) with respect to the centromere(identified by a dot) is indicated to the left of thediagram. Areas shaded in gray correspond to thetwo C57BL/6J-derived congenic segments fixedin AcB55 on the A/J background (white areas).(b) Segregation of reticulocyte counts inperipheral blood of (AcB55 × A/J) F2 mice withmarkers D3Mit109 (defining the Char4 QTL) and D3Mit44 derived from the two B6-congenicsegments of chromosome 3. Mice (shown asindividual dots or triangles) were furtherseparated according to homozygosity orheterozygosity with respect to A/J (a) andC57BL/6J (b) alleles for both markers. Therecombinant fraction is shown for each haplotype(in parenthesis), along with the overall deducedrecombination frequency. (c) Haplotype map of(AcB55 × DBA/2) F2 mice for the central portionof chromosome 3. Each column represents achromosomal haplotype, with the total number ofmice bearing this haplotype indicated at the bottom (n). For genetic markers D3Mit228, D3Mit22, D3Mit175, D3Mit75 and D3Mit56, homozygosity withrespect to either parental DBA/2J (solid boxes) or A/J (empty boxes; AcB55-derived) or heterozygosity (half-filled boxes) is shown. The presence (emptyboxes) or absence (half-hatched boxes) of reticulocytosis (Retic; >5% circulating red cells) is shown. (d) Segregation of reticulocyte counts and D3Mit175alleles in (AcB55 × DBA/2) F2 mice; A/J (AcB55-derived; a) and DBA/2J (d) alleles are identified.

a

c d

b

©20

03 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

reg

enet

ics

L E T T E R S

reduction in peak parasitemia and mortality after acute infectionwith P. chabaudi (Fig. 5c,d). Notably, the Char4 quantitative traitlocus (QTL) mapped in (AcB55 × A/J) F2 mice was not detected in(AcB61 × A/J) F2 mice similarly infected with P. chabaudi, initiallysuggesting independent genetic control of malaria resistance inAcB55 and AcB61 (refs. 8,9). The current study establishes that thisis not the case. Linkage to Char4 in the AcB61 cross was notdetected because chromosome 3 of AcB61 does not bear B6 seg-ments and thus was not informative.

In the present studies, a trait that seems to be complex can beexplained by the segregation of a monogenic trait regulating a distinctbut related phenotype. The distribution of peak parasitemia in(AcB55 × A/J) F2 mice seems complex, with the Char4 locus showingsignificant linkage (χ2 = 23.49; lod = 5.1; P = 0.000008) but account-ing only for a fraction of the variance (∼ 14%)9. On the other hand,reticulocytosis in (AcB55 × A/J) F2 mice has a bimodal distributionand segregates as a monogenic trait, with homozygosity at Pklr269A

showing complete linkage (χ2 > 50) and explaining 100% of the vari-ance. For peak parasitemia in these mice, Pklr alleles show higher sta-tistical association than Char4 alleles (χ2 = 44.3; lod = 9.6) but stillexplain only a fraction of the variance. This may be due to variability

of phenotypic ascertainment when using peak parasitemia, incom-plete penetrance or further modulation by other genetic traits, suchas gender, or autosomal contributions, such as from chromosome 10previously detected in AcB55 (ref. 9). Our results suggest that a com-plete loss-of-function mutation behaving as a monogenic trait forone phenotype (hemolytic anemia) affects another phenotype inwhich apparent complexity may be due to variability in ascertain-ment or additional genetic components.

The mechanisms by which pyruvate kinase deficiency affects theability of Plasmodium to replicate inside erythrocytes are unknownbut may involve the following possibilities: (i) greater membranerigidity affecting parasite invasion; (ii) altered membrane propertiesresulting in shortened half-lives of noninfected and infected erythro-cytes through increased phagocytosis; (iii) greater abundance ofmetabolic intermediates, such as 2,3-bis-phosphoglycerate, and ofoxidative species, resulting in a less hospitable intracellular environ-

360 VOLUME 35 | NUMBER 4 | DECEMBER 2003 NATURE GENETICS

a

b

Figure 4 A mutation in liver- and erythrocyte-specific Pklr causes reticulocytosis in AcB55and AcB61 mice. A partial transcript map ofthe chromosome 3 region surroundingD3Mit175 is shown (a), indicating some of the genes and markers in the region, with their distance (in cM) from the centromere(identified by a filled circle) shownimmediately below. The exon-intronorganization of Pklr is shown, with exonsnumbered E1 to E11. The sequence of eachexon was determined in A/J, B6, AcB55 andAcB61 strains. The amino acid sequenceencoded by exon 2 (E2) of the four strains is shown, highlighting the I90N mutationfound in AcB55/AcB61. The correspondingsequencing gels are shown (b).

Figure 5 Association of Pklr-I90N with protection against malaria. (a) Correlation between the number of reticulocytes in peripheral blood(expressed as a percentage of red cells) of 200 individual mice of an(AcB55 × A/J) F2 cross and the extent of blood-stage replication of P.chabaudi AS parasite at the peak of infection (peak parasitemia,expressed as pRBCs). Each dot represents a mouse. The horizontal linesdelineate the extremities of the distribution of peak parasitemia valueswhere a complete correlation is observed between presence or absence of reticulocytosis and low or high peak parasitemia, respectively (χ2 = 51,d.f. = 1, P < 0.001). (b) The 269T→A mutation in Pklr (resulting inI90N) eliminates a SfaNI restriction site in genomic DNA. Diagnosticdigestion products corresponding to B6 and A/J (93 bp + 243 bp;Pklr269T), AcB55/AcB61 (336 bp, Pklr269A) and to homozygous andheterozygous combinations seen in segregating F2 mice are shown afteragarose gel electrophoresis. The effect of homozygosity or heterozygositywith respect to either wild-type (+; Pklr269T) or mutated (–; Pklr269A) Pklralleles on peak parasitemia and survival to infection with P. chabaudi AS(106 pRBCs) is shown for informative (AcB61 × A/J) (c) and (AcB55 ×A/J) (d) F2 mice. Mice were grouped according to their Pklr haplotypes,and the horizontal bar indicates the average peak parasitemia values ineach group. The effect of Pklr alleles on mortality is shown as standardsurvival plots.

a

c d

b

©20

03 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

reg

enet

ics

L E T T E R S

ment20; (iv) altered ratio of reticulocytes to mature erythrocytes incirculating blood affecting replication of Plasmodium species prefer-ring mature red cells as a host; (v) impairment of intraerythrocyticparasite glucose metabolism. Pyruvate kinase activity is substantiallygreater in infected erythrocytes, and although this increase seemslargely due to parasite-encoded pyruvate kinase21, one cannotexclude the possibility that host pyruvate kinase has a role in theearly stages of infection.

Strong evidence suggests that several forms of thalassemia, certainhemoglobin variants (HbS, HbC and HbE), deficiency in glucose-6-phosphate dehydrogenase, the Duffy blood group and southeastAsian ovalocytosis (deletion of ‘band 3’ NHE1 gene) may have a pro-tective effect against malaria in humans (reviewed in refs. 3,4).Heterozygosity at severe alleles, such as in HbS and β-thalassemia22,and homozygosity at milder alleles (such as in HbC, α-tha-lassemia23,24) both have a protective effect against malaria. Ourresults in mice suggest that homozygosity with respect to pyruvatekinase deficiency confers malaria protection, as in HbC23. Inhumans, homozygosity or compound heterozygosity with respect topartial or complete loss-of-function alleles at PKLR cause pyruvatekinase deficiency and are the most common cause of hereditary non-spherocytic hemolytic anemia25. The possibility that pyruvate kinasedeficiency may also affect susceptibility to malaria in humans isbeing investigated.

METHODSMice. We purchased inbred C57BL/6J, DBA/2J and A/J mice from the JacksonLaboratory. The AcB55 and AcB61 strains are part of a set of 37 AcB/BcArecombinant congenic strains derived from A/J and C57BL/6J parents andproduced by E. Skamene at the Montreal General Hospital Research Institute,according to a breeding scheme and genotyping protocol previouslydescribed10. We generated (AcB55 × A/J), (AcB61 × A/J) and (AcB55 × DBA/2)F1 and F2 crosses by standard brother-sister mating. All AcB and BcA strainsunderwent a complete analysis of hematological parameters in peripheralblood, including erythrocyte and leukocyte counts, hemoglobin, hematocrit,platelets, mean corpuscular volume and mean concentration of hematocrit.All mice were handled according to guidelines and regulations of the CanadianCouncil on Animal Care.

P. chabaudi AS infection. We phenotyped mice 8 to 12 weeks of age withrespect to susceptibility to malaria, as previously described26. We infectedmice intraperitoneally with 106 pRBCs and determined the percentage ofpRBCs daily on duplicate thin blood smears stained with Diff-Quik (DadeBehring) on days 4–28 after infection. We recorded death daily and killedmoribund mice. We determined the percentage of reticulocytes by differen-tial counts based on 400 cells on thin blood smears stained with Diff-Quik(Dade Behring)26.

Transcription profiling studies. Mice were killed and organs were rapidlyfrozen in liquid nitrogen and then stored at –80 °C. We homogenized tissuesmechanically and extracted total cellular RNA (TRIzol; Life Technologies). Weisolated the poly(A)+ fraction of the RNA by chromatography with oligo-d(T)cellulose27. For cDNA labeling, we converted either 25 µg of total RNA or 2.5µg of poly(A)+ RNA into cDNA using reverse transcriptase (Super Script II,Invitrogen) and either Cy5- or Cy3-labeled dCTP (1 mM, Perkin Elmer-Cetus/NEN) in a reaction mixture containing 1.5 µl oligo(dT) (100 pmolµl–1), 3 µl dNTP-dCTP (6.67 mM each), 1 µl dCTP (2 mM), 4 µl dithiothreitol(100 mM) and 8 µl 5× RT Buffer (Invitrogen). The reactions were carried outat 42 °C for 3 h, and the RNA was then degraded by the addition of 0.5 µlRNase A (1 µg µl–1) and 1.5 µl RNaseH (5 units µl–1). We separated labeledcDNA from unincorporated nucleotides and further concentrated it by evapo-ration under vacuum. We then used labeled cDNA to hybridize Mouse 15k v.1cDNA spotted arrays purchased from the UHN Microarray Facility (J.Woodgett, Toronto, Ontario, Canada) containing 15,250 expressed-sequencetags (NIA clone set) spotted in duplicate. Briefly, the arrays were prehybridized

for 1–2 h with DIGEasy hybridization buffer (Roche) containing 10 µg ml–1

denatured salmon sperm DNA and 10 µg ml–1 yeast tRNA. We combined theCy5- and Cy3-labeled cDNAs, hybridized them in the same medium and incu-bated them with the arrays for 16–18 h at 37 °C. Finally, we washed the arraysthree times for 10 min each in 0.1× saline sodium citrate (SSC; 20× SSC is 3 Msodium chloride, 0.3 M sodium citrate, pH 7.0) and 0.1% SDS at 50 °C andfour times for 3 min each in 0.1× SSC at room temperature and then driedthem by centrifugation. We scanned the slides, acquired digitized images usinga ScanArray 5000 instrument (Perkin Elmer) and quantified the intensity ofindividual spots from 16-bit TIFF images using the QuantArray softwarepackage (Perkin Elmer).

Microarray data analysis. We normalized the raw data generated byQuantArray (Perkin Elmer) using the GeneSpring software package (SiliconGenetics) by the Lowess scatter smoothing algorithm. We analyzed eighthybridizations consisting of dye-swap hybridizations of four biologicalreplicates for AcB61 versus A/J and analyzed six hybridizations (dye swapsof three biological replicates) for AcB55 versus A/J. We used tissues fromindividual mice for each biological replicate. We identified genes withreproducible changes in transcript abundance with the ‘one class’ algorithmin the SAM (Significance Analysis of Microarrays) application28. SAMassigns a score to each gene on the basis of the change in expression relativeto the standard deviation of repeated measurements for that gene. We chosesignificant genes with a false discovery rate below 0.05% and further culledthem by keeping only those with a relative variation >1.5. We visualized theresults in GeneSpring.

Genotyping. Before infection, we obtained tail biopsy samples from all miceand isolated genomic DNA by a standard procedure involving proteinase Ktreatment10. We purchased primer pairs defining dinucleotide repeat markersinformative for the A/J, C57BL/6J and DBA/2J mouse strains from ResearchGenetics. We obtained linkage map positions from the Jackson LaboratoryMouse Genome Informatics website. We genotyped sequence length polymor-phisms by a standard PCR-based method using [α-32P] dATP labeling andseparation on denaturing 6% polyacrylamide gels10. We used the followingchromosome 3 markers: D3Mit22, D3Mit175, D3Mit75, D3Mit56, D3Mit228,D3Mit109 and D3Mit44.

Sequencing. We obtained regional transcript maps of chromosome 3 from theMouse Genome Assembly at the University of California Santa CruzBioinformatics website. We amplified individual exons or pairs of small exonsfrom Pklr genomic DNA of parental A/J and C57BL/6J strains as well as fromAcB61 and AcB55 strains. Oligonucleotides used for PCR amplification areavailable on request. We visualized the PCR products by agarose gel elec-trophoresis to confirm size and cloned them directly into a TOPO plasmidvector (Invitrogen). We recovered recombinant plasmids from independentEscherichia coli DH10β transformants and sequenced three independentclones for each exon using vector-derived primers flanking the cloning site. Wecarried out cycle sequencing in the presence of fluorescently labeled BigDyeTerminator (Applied Biosystems) and analyzed the products using an auto-mated ABI 3700 instrument (Applied Biosystems). We confirmed the presenceof sequence alterations in additional clones using standard dideoxynucleotidesequencing using [α-33P] dATPs. In the AcB55 and AcB61 strains, a singlenucleotide substitution at position 269 converts the wild-type isoleucineresidue at position 90 to asparagine (I90N) and also causes the disappearanceof a SfaNI recognition site (5′-GCATC (N)5 ^-3′). We assessed this polymor-phism by PCR amplification on genomic DNA using oligos for exon 2 as listedabove, followed by SfaNI digestion and agarose gel (1%) electrophoresis.

Statistical analysis. We used an unpaired, two-tailed t-test to analyze the effectof Pklr haplotypes on mean peak parasitemia in groups of (AcB55 × A/J) F2

and (AcB61 × A/J) F2 mice infected with P. chabaudi AS. In the same groups ofmice, we compared mortality rates using a standard χ2 test to detect significantdifferences. We carried out linkage analyses between the chromosome 3 mark-ers, including Pklr polymorphisms, and either reticulocyte counts or peak par-asitemia after P. chabaudi infection using MapManager QT29. We computed χ2

values using the expectation/maximization algorithm in Mapmaker/QTL.

NATURE GENETICS VOLUME 35 | NUMBER 4 | DECEMBER 2003 361

©20

03 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

reg

enet

ics

L E T T E R S

URLs. The Jackson Laboratory Mouse Genome Informatics website is athttp://www.jax.org. The University of California Santa Cruz Bioinformaticswebsite is at http://genome.ucsc.edu.

Accession numbers. The GenBank accession number for the Pklr cDNAsequence is NM01301. The GEO accession number of the data is GPL519.

Note: Supplementary information is available on the Nature Genetics website.

ACKNOWLEDGMENTSThe authors are indebted to E. Skamene for providing AcB55 and AcB61recombinant congenic strains and to J. Woodgett for providing spotted cDNAarrays. This work was supported by research grants to P.G. and M.M.S. from theCanadian Institutes of Health Research and from the Canadian Genetic DiseasesNetwork. Expertise in transcriptional profiling was made possible by the GenomeHealth Initiative of the National Research Council of Canada. P.G. is supported bya salary award from the CIHR (Distinguished Scientist) and by a James McGillProfessorship. G.M. is supported by the Maysie MacSporran Award from theStrategic Training Centre in Infectious Diseases and Autoimmunity of the Center for the Study of Host Resistance Group of McGill University.

COMPETING INTERESTS STATEMENTThe authors declare that they have no competing financial interests.

Received 9 July; accepted 3 October 2003Published online at http://www.nature.com/naturegenetics/

1. Sachs, J. & Malaney, P. The economic and social burden of malaria. Nature 415,680–685 (2002).

2. Cooke, G.S. & Hill, A.V. Genetics of susceptibility to human infectious disease. Nat.Rev. Genet. 2, 967–977 (2001).

3. Weatherall, D.J. & Clegg, J.B. Genetic variability in response to infection: malariaand after. Genes Immun. 3, 331–337 (2002).

4. Fortin, A., Stevenson, M.M. & Gros, P. Susceptibility to malaria as a complex trait:big pressure from a tiny creature. Hum. Mol. Genet. 11, 2469–2478 (2002).

5. Foote, S.J. et al. Mouse loci for malaria-induced mortality and the control of para-sitaemia. Nat. Genet. 17, 380–381 (1997).

6. Fortin, A. et al. Genetic control of blood parasitaemia in mouse malaria maps tochromosome 8. Nat. Genet. 17, 382–383 (1997).

7. Burt, R.A., Baldwin, T.M., Marshall, V.M. & Foote, S.J. Temporal expression of anH2-linked locus in host response to mouse malaria. Immunogenetics 50, 278–285(1999).

8. Fortin, A., Stevenson, M.M. & Gros, P. Complex genetic control of susceptibility tomalaria in mice. Genes Immun. 3, 177–186 (2002).

9. Fortin, A. et al. Identification of a new malaria susceptibility locus (Char4) in recombi-nant congenic strains of mice. Proc. Natl. Acad. Sci. USA 98, 10793–10798 (2001).

10. Fortin, A. et al. Recombinant congenic strains derived from A/J and C57BL/6J: atool for genetic dissection of complex traits. Genomics 74, 21–35 (2001).

11. Jacobasch, G. & Rapoport, S.M. Hemolytic anemias due to erythrocyte enzyme defi-ciencies. Mol. Aspects Med. 17, 143–170 (1996).

12. van Solinge, W.W., Kraaijenhagen, R.J., Rijksen, G. & Nielsen, F.C. Novel mutationsin the human red cell type pyruvate kinase gene: two promoter mutations in cis, asplice site mutation, a nonsense- and three missense mutations. Blood 90Supplement, 1197 (1997).

13. Vidal, S.M., Malo, D., Vogan, K., Skamene, E. & Gros, P. Natural resistance to infec-tion with intracellular parasites: isolation of a candidate for Bcg. Cell 73, 469–485(1993).

14. Diez, E. et al. Birc1e is the gene within the Lgn1 locus associated with resistance toLegionella pneumophila. Nat. Genet. 33, 55–60 (2003).

15. Poltorak, A. et al. Defective LPS signaling in C3H/HeJ and C57BL/10ScCr mice:mutations in Tlr4 gene. Science 282, 2085–2088 (1998).

16. Lee, S.H. et al. Susceptibility to mouse cytomegalovirus is associated with deletionof an activating natural killer cell receptor of the C-type lectin superfamily. Nat.Genet. 28, 42–45 (2001).

17. Brown, M.G. et al. Vital involvement of a natural killer cell activation receptor inresistance to viral infection. Science 292, 934–937 (2001).

18. Vladimirov, V. et al. Different genetic control of cutaneous and visceral disease afterLeishmania major infection in mice. Infect. Immun. 71, 2041–2046 (2003).

19. Kemp, S.J., Iraqi, F., Darvasi, A., Soller, M. & Teale, A.J. Localization of genes con-trolling resistance to trypanosomiasis in mice. Nat. Genet. 16, 194–196 (1997).

20. Roth, E. Jr. Plasmodium falciparum carbohydrate metabolism: a connectionbetween host cell and parasite. Blood Cells 16, 453–460 (1990).

21. Oelshlegel, F.J., Jr., Sander, B.J. & Brewer, G.J. Pyruvate kinase in malaria host-par-asite interaction. Nature 255, 345–347 (1975).

22. Willcox, M. et al. A case-control study in northern Liberia of Plasmodium falciparummalaria in haemoglobin S and beta-thalassaemia traits. Ann. Trop. Med. Parasitol.77, 239–246 (1983).

23. Modiano, D. et al. Haemoglobin C protects against clinical Plasmodium falciparummalaria. Nature 414, 305–308 (2001).

24. Allen, S.J. et al. α-Thalassemia protects children against disease caused by otherinfections as well as malaria. Proc. Natl. Acad. Sci. USA 94, 14736–14741 (1997).

25. Zanella, A. & Bianchi, P. Red cell pyruvate kinase deficiency: from genetics to clini-cal manifestations. Baillieres Best Pract. Res. Clin. Haematol. 13, 57–81 (2000).

26. Stevenson, M.M., Lyanga, J.J. & Skamene, E. Murine malaria: genetic control ofresistance to Plasmodium chabaudi. Infect. Immun. 38, 80–88 (1982).

27. Sambrook, J.R. & David, W. Extraction, purification, and analysis of mRNA fromeukaryotic cells. in Molecular Cloning: A Laboratory Manual vol. 1 (Cold SpringHarbor Laboratory Press, Cold Spring Harbor, New York, 2001).

28. Tusher, V.G., Tibshirani, R. & Chu, G. Significance analysis of microarrays applied tothe ionizing radiation response. Proc. Natl. Acad. Sci. USA 98, 5116–5121 (2001).

29. Manly, K.F. & Olson, J.M. Overview of QTL mapping software and introduction tomap manager QT. Mamm. Genome 10, 327–334 (1999).

362 VOLUME 35 | NUMBER 4 | DECEMBER 2003 NATURE GENETICS

©20

03 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

reg

enet

ics