Multiple polymorphic loci determine basal hepatic and splenic iron status in mice

Authors


  • Potential conflict of interest: Nothing to report.

Abstract

Polymorphisms of genes linked to iron metabolism may account for individual variability in hemochromatosis and iron status connected with liver and cardiovascular diseases, cancers, toxicity, and infection. Mouse strains exhibit marked differences in levels of non-heme iron, with C57BL/6J and SWR showing low and high levels, respectively. The genetic basis for this variability was examined using quantitative trait loci (QTL) analysis together with expression profiling and chromosomal positions of known iron-related genes. Non-heme iron levels in liver and spleen of C57BL/6J × SWR F2 mice were poorly correlated, indicating independent regulation. Highly significant (P < .01) polymorphic loci were found on chromosomes 2 and 16 for liver and on chromosomes 8 and 9 for spleen. With sex as a covariate, additional significant or suggestive (P < 0.1) QTL were detected on chromosomes 7, 8, 11, and 19 for liver and on chromosome 2 for spleen. A gene array showed no clear association between most loci and differential iron-related gene expression. The gene for transferrin and a transferrin-like gene map close to the QTL on chromosome 9. Transferrin saturation was significantly lower in C57BL/6J mice than in SWR mice, but there was no significant difference in the serum level of transferrin, hepatic expression, or functional change in cDNA sequence. β2-Microglobulin, which, unlike other loci, was associated with C57BL/6J alleles, is a candidate for the chromosome 2 QTL for higher iron. In conclusion, the findings show the location of polymorphic genes that determine basal iron status in wild-type mice. Human equivalents may be pertinent in predisposition to hepatic and other disorders. (HEPATOLOGY 2006;44:174–185.)

Fundamental processes such as oxygen transport, respiration, cell cycle and signaling and metabolism of endogenous and exogenous chemicals are dependent on the redox potential of iron.1 At the same time redox activity occurring in tissues such as liver, kidney, brain, and heart may have pathological consequences in circumstances of abnormal iron deposition, malfunctions of iron homeostasis, or oxidative stress.2 Iron overload of hereditary hemochromatosis as a result of homozygosity for HFE mutations is associated with cirrhosis, diabetes, cardiomyopathy, arthritis, cancer, and sterility. However, penetrance of these mutations is low and probably depends on modifier genes.1, 3–5 More subtle differences of higher iron burden in populations have also been linked with liver and cardiovascular disease, diabetes, cancer, increased oxidative DNA damage, and neurodegenerative disorders.6–11 In addition, the iron status of a host may be a major factor in infections,12 and iron deficiencies associated with anemias are of wide concern.13

The absorption of iron and its subsequent use and turnover are highly regulated and likely to be influenced by individual variation.1, 3 Mice null for specific iron metabolism genes have greatly facilitated the understanding of iron homeostasis, but the severity of the hepatic phenotype of these genes may depend considerably on other genes that have yet to be identified.14–16 In particular, the C57BL/6J mouse background confers resistance to hepatic iron loading.14–17 However, the influence of mouse genes explored with null gene technology may not correspond to polymorphic susceptibility loci observed in non–genetically manipulated animals. Many important genetic traits in humans and experimentally, such as blood pressure and survival from infection, are recognized to be complex and quantitative, with polymorphism of modifier genes having profound effects on the dominance of a primary gene defect.18 So far most investigations of iron-overload diseases have focused on the locations of modifier genes for hepatic iron loading in Hfe-null mice.17, 19 Variation also occurs in the basal iron status of common inbred mouse strains.20–27 C57BL strains show much lower iron than some others, with SWR mice having much higher levels of hepatic and splenic iron, which may explain their different responses to toxins, infection, and age-related changes.22–24, 28

In this article, we show the location and contribution of polymorphic genes that determine basal iron status in apparently normal individuals using quantitative trait loci (QTL) analysis of an intercross between C57BL/6J and SWR mice.

Abbreviations

QTL, quantitative trait loci; EST, expressed sequence tag; RT-PCR, reverse-transcriptase PCR; LOD, log odds ratio.

Materials and Methods

Generation of F2 Mice.

C57BL/6J/Ola and SWR/Ola were obtained from Harlan UK. They were fed an RM1 diet (Special Diet Services, Witham, UK) for at least 2 weeks before use and maintained at 21°C with a 12-hour light/dark cycle. (C57BL/6 × SWR) × (C57BL/6 × SWR) and (SWR × C57BL/6) × (SWR × C57BL/6) F2 mice were bred from F1 mice with either C57BL/6J/Ola or SWR/Ola as paternal grandmothers. A total of 284 male and female F2 mice were bred and culled at approximately 8 weeks of age. Livers and spleens were frozen at −80°C. Tissues from mice for histological and immunohistochemical analyses were fixed in buffered formalin.29 Antimouse DCYTB11-A peptide antibody was from Alpha Diagnostic (San Antonio, TX).

Non-Heme Iron.

Non-heme iron in liver and spleen was quantitated by the method of Torrance and Bothwell as previously reported22, 23 Levels are expressed as micrograms of Fe per gram of wet tissue. The non-heme iron level in the diet was 143 μg Fe/g. The manufacturer's estimate of total iron content was 159 μg Fe/g. The distribution of non-heme iron in tissues was shown by Perls' stain.

Analysis of Serum Transferrin.

Blood was collected by cardiac puncture under terminal anesthesia (Home Office license 40/2571). Serum iron, total iron-binding capacity (TIBC), and transferrin saturation were determined with a kit (Diagnostic Chemicals Ltd., Canada). Analysis of serum transferrin by electrophoresis was performed on 10% SDS polyacrylamide gels followed by immunoblotting using a rabbit polyclonal anti–human transferrin antibody (Stressgen Biotechnologies Corp., Canada) and mouse transferrin as a control (US Biological, MA) with detection by a chemiluminescence system (Amersham Biosciences, Amersham, UK).

Genotyping.

Genomic DNA was isolated from mouse liver by phenol-chloroform extraction. Individuals were genotyped using polymerase chain reaction (PCR) amplification of polymorphic microsatellite markers.30 Positions of markers were based on data of the Whitehead Institute (WI; http://www.broad.mit.edu/cgi-bin/mouse/index/). In many cases these were <2 cM different from those of Mouse Genome Informatics (MGI; http://www.informatics.jax.org/). Where positions differed markedly, this was taken into account when considering candidate genes. Information about SWR allele sizes was obtained as previously described, or primers were selected and tested for polymorphisms30 and can be obtained from the corresponding author. Pairs of fluorescently labeled primers were purchased from Applied Biosystems (Warrington, UK), and unlabeled primers were synthesized in the University of Leicester. Fluorescent PCR products were analyzed on an ABI 377 sequencer using GeneScan and Genotyper (Applied Biosystems) or were separated on 4% MetaPhor agarose (Cambrex) for unlabeled products. The extreme iron burden cohorts representing the most useful information for linkage analysis were genotyped for markers spread across the genome (2 or 3 per chromosome at 20-30 cM).30 Where the evidence was strongest for a QTL or a chromosomal region was of particular interest, further markers were included (if available) and the remainder of the mice genotyped.

Mapping and Statistical Analysis.

The R/qtl program (http://www.biostat.jhsph.edu/∼kbroman/qtl) was used for QTL mapping.31 As the distributions were skewed from normality, the phenotypes were analyzed as the log10 values. Untransformed data gave similar QTL results. Males and females were considered separately and also combined (in which case sex was included as an additive covariate). Genomewide thresholds of 10% and 1% significance were derived from 10,000 permutations of shuffled genotypes.32 Significant differences between groups were assessed by the Student t test or by one-way analysis of variance and correlation coefficients calculated using the Statistica 6 package.

Gene Expression Profiling.

The mouse cDNA microarray consisted of approximately 9,000 candidate EST clones obtained from Research Genetics (http://www.resgen.com) and the IMAGE collection and analyses performed as previously described.29, 33 mRNA was reverse-transcribed using aminoallyl dUTP. The resulting cDNA from C57BL/6 tissue was labeled with Alexa Fluor 555 Reactive Dye (Invitrogen, Paisley, UK) and cDNA from SWR mouse tissue labeled with Alexa Fluor 647. The two were then mixed and hybridized against the ESTs on the cDNA arrays. The procedure was repeated by reversing the dye labeling. Data from the arrays were normalized and averaged for each pair34 and can be found at ArrayExpress (E-MEXP-486; www.ebi.ac.uk/arrayexpress/).

Real-Time RT-PCR.

Reverse-transcriptase PCR (RT-PCR) was performed with 400 ng of mRNA, Superscript III Reverse Transcriptase (Invitrogen), and random hexamers (Amersham Biosciences). Primer Express software (Applied Biosystems) was used to design real-time RT-PCR primers (Table 1). Reactions were performed in triplicate using SYBR green PCR Master Mix (Applied Biosystems) on an ABI 7000 sequence detection system under standard conditions.35 Results were normalized to reference genes β- or γ-actin using the comparative CT method.

Table 1. Primers Used for Real-Time RT PCR
GeneSymbolPrimer
  • *

    Primers span exons 9 and 10.

  • Primers span exons 11 and 12.

  • Tflg primers were designed to a unique region to avoid detection of Trf.

  • §

    Preliminary nomination.

β-ActinActbForward: GATTACTGCTCTGGCTCCTAGCA
  Reverse: GTGGACAGTGAGGCCAGGAT
γ-ActinActgForward: CCCTAGCACCTAGCACGATGA
  Reverse: GCCACCGATCCA ACTGAGTAC
Heme oxygenase 1Hmox1Forward: CACTTCGTCAGAGGCCTGCTA
  Reverse: GTCTGGGATGAGCTAGTGCTGAT
HepcidinHampForward: TTGCGATACCAATGCAGAAGAG
  Reverse: ACTGGGAATTGTTACAGCATTTACAG
β2-MicroglobulinB2mForward: CATACGCCTGCAGAGTTAAGCA
  Reverse: GATCACATGTCTCGATCCCAGTAG
Transferrin*TrfForward: TGAGGACTGCATTGAAAAGATTGT
  Reverse: GTAGGCATGTCCTCCATCCAA
Transferrin Forward: ACAACAAGGTATCTTTCCTAAAGGGT
  Reverse: GGTTGTTCCAGGTGATGCTAGTG
Transferrin-likeTflg§Forward: GGTCCCTCGAAAGATAGACATCA
  Reverse: GGGAGTCTTCCAGACCTCTTTTAA

Sequencing of cDNA.

Using the sequence for C57BL/6J (accession number AK085754), primers for the amplification of transferrin cDNA were forward 5′-CTGTACTCCCCGCTCCTCGC-3′ positioned 50 bp 5′ to the translation start site, and reverse 5′-GAGCATCTGTCTCCACCACAG-3′ approximately 40 bp 3′ of the stop codon. Amplicons were made by pooling eight PCRs for each mouse with three mice per strain. Sequencing reactions and analyses were performed using the above primers and a Big Dye terminator v3.1 cycle sequencing kit (Applied Biosystems). New primers were designed to extend the sequence to the center of the amplicons (not shown). The data were aligned and edited using BioEdit software (http://www.mbio.ncsu.edu/BioEdit/bioedit.html). A similar strategy was used for the transferrin-like gene (1300017J02Rik). Primers used to make the PCR amplicon were forward 5′-GAGAGCAAGCAGAACGCAGT-3′ and reverse 5′-ACATGGAGGCAAGCAGTTG-3′.

Results

Distribution of Iron in Parental Strains.

Non-heme iron in liver and spleen of C57BL mice is significantly lower than that in SWR mice (Table 2).22–24, 36 Figure 1 illustrates the distribution of non-heme iron by Perls' stain in the liver and spleen of fully adult mice. Whereas there was little staining of hepatocytes and only positive-staining of occasional Kupffer cells in C57BL/6J mouse liver, there was pronounced staining in SWR mouse liver, mainly of hepatocytes but also of some Kupffer cells (Fig. 1A-B). Iron in the spleen was particularly pronounced in the SWR mice, mainly in the macrophages of the red pulp (Fig. 1C-D). Perls' staining in the SWR mice was also observed in kidney proximal convoluted tubules at the apex of the columnar epithelial cells and their brush border lining the tubule. It was far less intense in kidneys from C57BL/6J mice. Although expression was variable, no consistent strain difference for duodenal ferric reductase (DCYTB; Cybrd1 gene) was observed by immunohistochemistry (Fig. 1E-F).

Table 2. Phenotypes in C57BL/6J and SWR Mice and Their F2 Crosses
StrainNSexLiver Iron (μg/g)Spleen iron (μg/g)
  • Values are means ± SDs from mice 6-8 weeks of age.

  • *

    Consisted of both (C57BL/6 × SWR) × (C57BL/6 × SWR) F2 and (SWR × C57BL/6) × (SWR × C57BL/6) F2 progeny.

  • Significantly less than females (P < .05, two-tailed Student t test).

  • SWR significantly greater than the C57BL/6J strain (P < .05). The C57BL/6J and SWR mice shown are not the parents of the F2 generation reported.

C57BL/6J5Male40 ± 5100 ± 9
 5Female57 ± 8157 ± 75
SWR5Male234 ± 40337 ± 128
 5Female254 ± 15342 ± 65
C57BL/6 × SWR F110Male53 ± 14102 ± 25
 13Female84 ± 16192 ± 75
C57BL/6 × SWR F2*145Male78 ± 34317 ± 172
 139Female112 ± 35427 ± 201
Figure 1.

Iron staining and immunohistochemistry of tissue from 12-week-old female C57BL/6J and SWR mice. (A) and (B) Hepatic non-heme iron by Perls' stain in C57BL6/J and SWR tissue, respectively. All hepatocytes of SWR tissue showed blue staining, whereas little was seen in C57BL/6J tissue. There was no consistent lobular distribution of iron in SWR tissue. In some sections, iron staining of Kupffer cells was also observed. Original magnification: 200×. (C) and (D) Splenic iron by Perls' stain in C57BL/6J and SWR tissue, respectively (wp, white pulp; rp, red pulp; arrows show marginal zone). Original magnification ×100. (E) and (F) Immunohistochemistry of duodenum for ferric reductase (DCYTB) in C57BL6/J and SWR tissue, respectively. Sections with immunoglobulin G as control showed no brown apical staining. Original magnification ×200. Bar size is 100 μ.

Non-Heme Iron Levels in Genetic Crosses.

To investigate the genetic basis of the strain variation in levels of liver and spleen iron, an F2 intercross between the parent C57BL/6J and SWR strains was examined at 6-8 weeks (Table 2), when divergence in iron status was still developing.23 In addition to a marked difference between the parental strains, there was a propensity for female C57BL/6J F1 and F2 mice to have greater iron contents than males. This was not detected at this age in the SWR parental animals. In other studies a greater distinction between the sexes was observed, perhaps reflecting contributions from age and diet. Figure 2A-B shows the distribution of iron levels for liver and spleen. The low correlation (r < 0.3) between liver and spleen iron (Fig. 2C) implies that genetic and environmental factors determining levels in the two tissues were largely, but not totally, independent.

Figure 2.

Distribution plots of tissue non-heme iron levels in 284 6- to 8-week-old male and female C57BL/6J × SWR F2 mice: (A) liver iron; (B) spleen iron (shading, males; no shading, females). (C) Correlation analysis (showing 95% confidence limits) of log10 iron levels in liver and spleen between males and females. Results of analyses by Pearson correlations and Bonferroni probabilities of difference were: females, r = 0.270, P = .027; males, r = 0.259, P = .035.

Quantitative Trait Locus Analysis.

F2 mice representing the extreme 17th percentiles for male and female liver and spleen phenotypes (24 per sex) were genotyped using microsatellite markers polymorphic between the parent strains and distributed across the genome. These represented a total of 55% of all F2 progeny. Further markers and mice were added to strengthen potential quantitative trait locus (QTL) positions. Combined data from the sexes were analyzed in a genomewide scan using R/qtl (Fig. 3). Permutation analyses of the two phenotypes were set to define genomewide LOD thresholds of 10% and 1% (termed here suggestive and highly significant, respectively).32 QTL locations, microsatellite markers nearest the calculated peak of the QTL, and LOD scores and confidence intervals of the QTL are shown in Table 3.

Figure 3.

Genome scans using R/qtl for the iron status of mice of both sexes combined: (A) liver; (B) spleen. Horizontal lines represent genomewide significance thresholds of 10% (suggestive) and 1% (highly significant), as determined by permutation tests. The broad bands under each chromosome represent the relative width between the first and last markers.

Table 3. Significant QTL for Iron Status Observed in C57BL/6J × SWR F2 Mice by Sex and Overall*
OrganSexChrLocation (cM)Peak MarkerLODPConfidence Interval (cM)
  • *

    Genomewide significance was based on permutation analysis (see Materials and Methods section) giving 10% (suggestive) and 1% (highly significant) LOD score thresholds for liver of 3.07 and 4.30, respectively; the thresholds for spleen were 3.01, and 4.20 respectively. The locations of QTL are given at calculated maximum LODs with markers closest to these. Confidence intervals for the QTL were calculated as the chromosomal region with LOD score within 1.5 units of its maximum for the chromosome using WI, but ranges and peak marker positions may differ to some extent depending on use of WI or MGI databases, most notably D2Mit17 and D7Mit71.

LiverCombined257D2Mit175.2.00251-68
  747D7Mit713.5.04823-54
  1549D15Mit1593.1.09631-49
  1629D16Mit308.1< .00125-34
 Male256D2Mit173.4.04038-73
  747D7Mit714.4.00434-54
  1629D16Mit305.6< .00117-35
 Female838D8Mit1953.1.08427-46
  1150D11Mit363.4.04727-66
  1938D19Mit373.3.06216-38
SpleenCombined814D8Mit45.2.0016-23
  956D9Mit18214.3< .00149-61
 Male255D2Mit173.3.06145-69
  956D9Mit18210.8< .00149-61
 Female957D9Mit1824.4.01241-61

Highly significant liver QTL for the combined sexes were found on chromosome 2 and, particularly, on chromosome 16, and suggestive QTL were found on chromosomes 7 and 15 (Fig. 3A, Table 3). When the sexes were considered separately, highly significant QTL were detected on chromosomes 7 and 16 (males) and suggestive QTL on chromosomes 8, 11, and 19 (females) (Table 3). For the combined spleen data two highly significant QTL were detected on chromosome 8 and, particularly, on chromosome 9 (Fig. 3B). The spleen chromosome 8 QTL was at a different position than that for the liver (Table 3). A QTL on chromosome 2 only reached the 90% confidence threshold for males but appeared to be identical to that observed for the liver (Table 3). Sex explained 15.4% of total variance for liver and 11% for spleen. There was no evidence of linkage on the X chromosome for either tissue. Although most of the time the same QTL was found for males and females, relative strength varied considerably by sex (Figs. 4 and 5). Only on chromosome 2 for spleen (D2Mit304) was the sex difference statistically significant (P = .009).

Figure 4.

Chromosome scans and allele effects for major liver QTL: (A-D) chromosome scans; (E-H) effect of genotype on iron levels (μg/g liver) at markers closest to peak LOD scores. Continuous line shows combined data. The dotted line represents females, and the dashed line shows males. Horizontal lines correspond to genome-wide LOD thresholds of 10% and 1%. Shaded bands show the confidence intervals for the combined data. Triangles show marker positions. SS, SB, and BB represent homozygous and heterozygous alleles of SWR and C57BL/6J for markers closest to the peak LOD scores. QTL on chromosomes 2 and 16 explained 6.7% and 7.5%, respectively, of total variance.

Figure 5.

Chromosome scans and allele effects for spleen QTL: (A-C) chromosome scans; (D-F) effect of genotype on iron levels (μg/g liver) at markers closest to peak LOD scores. Continuous line shows combined data. The dotted line represents females, and the dashed line shows males. Horizontal lines correspond to genomewide LOD thresholds of 10% and 1%. Shaded bands show the confidence intervals for the combined data. Triangles show marker positions. SS, SB, and BB represent homozygous and heterozygous alleles of SWR and C57BL/6J for markers closest to the peak LOD scores. QTL on chromosomes 2, 8, and 9 explained 3.4%, 5.3%, and 14.5%, respectively, of the total variance.

Plots of iron levels versus allele genotypes at markers nearest to peak LOD scores showed approximately additive effects (Figs. 4 and 5). Two-dimensional, two-QTL scans showed no evidence of epistasis, in which the effect of the alleles at one locus depend on the alleles at other loci. For most of the QTL the SWR allele was associated with higher iron levels. In contrast, the higher iron phenotype QTL on chromosome 2, for both tissues, was associated with the C57BL/6J allele (Figs. 4E and 5D).

Iron-Related Gene Profiling.

We used our cDNA arrays29 to compare the expression of many known genes associated with iron metabolism in the liver, spleen, and duodenum of males of both mouse strains (Table 4). These cDNA arrays broadly contain those genes linked to iron and heme metabolism and were first reported by Muckenthaler.37 No significant differences in the expression of most genes according to major QTL were detected. These included hepatic L ferritin (Ftl1) and IRP1 (Aco1), genes that previously showed strain variation,23 which may have been a result of posttranscriptional responses. A notable exception for all three tissues was the β2-microglobulin (B2m) gene (chromosome 2), which apparently was expressed to a greater degree in the C57BL/6J mice than in the SWR mice. A small but significant elevation of the heme oxygenase 1 (Hmox1) gene in liver and spleen was observed for the SWR mice, which was confirmed by RT-PCR (Table 5).

Table 4. Comparison Between SWR and C57BL/6J Mice of Expression of Some Known Iron-Related Genes
GeneSymbolChromosomePosition (cM or Band)LiverSpleenDuodenum
  • The result for the EST of each gene used is expressed as a mean of the ratio of the response from SWR mice relative to that of paired C57BL/6J mice with initial and reverse labeling.

  • *

    Significant at P < .05 with four or five separate mouse comparisons per group.

  • Some EST clones for transferrin that are not shown had a greater response with C57BL/6J.

ATP binding cassette, B7Abcb7X391.170.991.01
Aconitase 1 (IRP1 IREBP)Aco14211.190.981.00
Aconitase 2, mitochondrialAco215130.921.010.96
Aminolevulinate synthase 1 (H)Alas19F10.850.991.02
Aminolevulinate synthase 2 (E)Alas2X631.061.32*1.02
β2 -MicroglobulinB2m2690.69*0.60*0.61
CeruloplasminCp3550.800.941.02
Cytochrome b reductase 1 (DCYTB)Cybrd12C31.110.951.08
Cytochrome b-561Cyb56111651.151.011.00
Erythropoietin receptorEpor950.991.011.0
Ferredoxin 1Fdx19B0.990.991.01
FerrochelataseFech18391.001.180.99
Ferritin heavy chainFth11920.881.161.03
Ferritin light chainFtl17250.881.120.99
FrataxinFxn19C11.101.021.02
HepcidinHamp7111.641.000.99
HephaestinHephX361.120.960.95
HemochromatosisHfe13150.960.920.98
Hemochromatosis (juvenile)Hfe23F20.990.941.04
Heme oxygenase 1Hmox18351.11*1.43*0.99
Heme oxygenase 2Hmox21631.030.941.00
HaptoglobinHp8550.681.020.98
HemopexinHpxn7F10.570.921.08
IRP2Ireb29291.171.000.99
Lipocalin 2Lcn22271.10*1.16*1.02
LactotransferrinLtf9700.841.20*0.94
Metallothionein 1Mt18451.461.161.19
NRAMPSlc11a11391.070.95*1.01
NRAMP2 (DMT2)Slc11a215601.16*0.961.15
IREG1 (MTP, ferroportin)Slc40a11B1.211.181.10
TransferrinTrf9560.931.021.0
Transferrin receptorTfrc16211.121.111.02
Transferrin receptor 2Trfr25G21.040.961.07
Table 5. Expression of Selected Genes as Estimated by Real-Time RT-PCR
GeneLiverSpleen
SWR:C57BL/6JPSWR:C57BL/6JP
  • Expression of target genes was calculated relative to γ-actin except for the β-2 microglobulin and heme oxygenase 1 genes, which were calculated relative to β-actin. We used both β- and γ-actin for normalization because the reference gene must have approximately the same amplification efficiency as the target gene for the comparative CT (threshold cycle) calculation to be valid.35 Results are means ± SDs of 5 male mice per group, and ratios are of SWR mice compared with C57BL/6J mice.

  • Primers span exons 9 and 10.

  • Primers span exons 11 and 12. The expression of transferrin and especially hepcidin was far lower in spleen than in liver.

β2-microglobulin0.97 ± 0.14.7410.67 ± 0.08< .001
Hepcidin2.65 ± 0.16.0531.59 ± 0.35.008
Heme oxygenase 11.78 ± 0.36.0011.79 ± 0.06< .001
Transferrin1.3 ± 0.24.0671.04 ± 0.13.636
Transferrin1.14 ± 0.26.2791.1 ± 0.1.133
Transferrin-like1.51 ± 4.9.0592.31 ± 1.0.047

Candidate Genes for QTL.

Particular consideration was given to candidate genes for highly significant QTL. There was a trend toward higher hepatic expression of hepcidin (Hamp) genes in SWR mice compared with that in C57BL/6J mice, although not significantly so, as judged by the array (Table 4) and RT-PCR (Table 5). A similar finding (data not shown) was made for Hamp2, whose role in iron metabolism is questionable.38–40 However, the location of the Hamp genes (11 cM) is markedly different from that of the chromosome 7 QTL (Table 3). Although transferrin receptor (21 cM) is a candidate gene for the chromosome 16 QTL, it lies on the edge of the confidence interval, some distance from the peak marker D16Mit30 (Whitehead, 30.6 cM; MGI, 36.5 cM). A hypothetical gene much closer, E330017A01Rik, has a ferritin domain, but we failed to amplify cDNA of this gene from tissue samples.

No candidate for the spleen chromosome 8 QTL was identified. The most significant QTL detected was that on chromosome 9 for the spleen with a maximum LOD score within 1 cM of the transferrin gene (Trf) at 56 cM (Table 3). Consistent differences in transferrin expression in liver or spleen were not detected by gene array or by real time RT-PCR (Tables 4 and 5). Transferrin levels in serum estimated as TIBC (Table 6) and by immunoblotting (Fig. 6) were not significantly different according to mouse strain. In contrast, serum iron level and saturation of transferrin were significantly lower in C57BL/6J mice compared with SWR mice, similar to the findings for AKR.20, 26 The cDNA for transferrin showed polymorphisms between C57BL/6J and SWR mice in codons for amino acids 203 (nt 609, A→G) and 508 (nt 1524, C→T), respectively, but these are synonymous. A gene highly homologous to transferrin (1300017J02Rik) lies approximately 20 kb downstream from Trf.41 This transferrin-like gene, which we termed Tflg, shares 67.9% of its identity with Trf and 55.3% of its amino acid identity with the protein Trf encodes, as determined by EMBL-EBI sequence alignment. RT-PCR showed greater expression of Tflg in SWR mice than in C57BL/6J mice, but this difference was statistically significant only for the spleen (Table 5). A cDNA polymorphism between the strains (nt 107 C→T) corresponded to a change in amino acid 36 from threonine (C57BL/6J) to isoleucine (SWR).

Table 6. Serum Transferrin Saturation in Parent Strains
StrainSexSerum Iron (μg/dL)TIBC (μg/dL)TRF Saturation (%)
  • Values represent means ± SD of 4 or 5 mice (8 weeks old) per group.

  • *

    Significantly different from females (P < .05, two-tailed Student t test).

  • Significantly different from C57BL/6J strain.

  • Spleen and liver non-heme iron contents of AKR mice (data not shown) were not significantly different from the SWR strain shown in Table 2.

C57BL/6JMale133 ± 37*343 ± 1738 ± 9*
 Female178 ± 19347 ± 1751 ± 7
SWRMale272 ± 17376 ± 5073 ± 11
 Female347 ± 61394 ± 6389 ± 16
AKRMale291 ± 4391 ± 474 ± 4
 Female306 ± 20444 ± 10771 ± 15
Figure 6.

Levels of transferrin in serum and liver of C57BL/6J and SWR mice by immunoblotting. Lanes contained equal amounts of serum or liver protein. Repeated blots with different samples showed no reproducible differences between strains.

The β2-microglobulin (B2m) is a candidate for the chromosome 2 QTL. B2m showed significantly greater expression in the spleen of C57BL/6J male mice than in that of SWR mice according to gene array and RT-PCR (Tables 4 and 5). The B2m gene (69 cM) does not appear to be close to the two peak markers, D2Mit17 and D2Mit304, 56.8 and 59 cM, respectively (Table 3, Figs. 3 and 4), in the WI database but is much closer to MGI (69 and 73 cM, respectively).

Discussion

In humans, individual differences in apparently normal iron status may confer important predispositions for some diseases and responses to toxins.6, 12 In the present study we used an intercross of SWR and C57BL/6J mice to show basal levels of hepatic and splenic non-heme iron in non–genetically manipulated mice are determined by complex genetic traits that are tissue- and sex-dependent. SWR mice are resistant to the hepatic protoporphyria caused by some iron-chelating drugs,22 but differences in iron metabolism between SWR and C57BL/6J mice may be a factor in their susceptibility to uroporphyria, a model of the human iron-dependent hepatic disorder sporadic porphyria cutanea tarda, and in aging and oxidative stress.23, 24, 36, 42 We identified highly significant QTL on chromosomes 2 and 16 for the liver and chromosomes 8 and 9 for the spleen. In addition, other QTL were observed, in particular, on chromosomes 7 and 11 for the liver and chromosome 2 for the spleen, that were either significant only for both sexes or were significant for one sex. The QTL we have observed potentially represent some of the same genes as those of the QTL on chromosomes 1, 3, 7, 8, and 11, reported by Bensaid et al. to be modifiers of the Hfe null gene.17 On the other hand, because loci may be on the same chromosome does not mean they necessarily are associated with identical genes. However, they provide important information about the mechanisms for individual variation.

Although we found no evidence for an X chromosome–linked gene, there was a tendency for greater levels of iron in females, which was especially noted in the C57BL/6J parent strain and the F1 and F2 offspring. This is probably related to sexual dimorphism in the penetrance of some QTL, for example, liver chromosome 7. Sex-modified expression of an autosomal locus can occur by imprinting during development and in response to hormonal differences.

Thus far, QTL are not obviously explicable by differences in total duodenal, hepatic, or splenic tissue expression of many known iron-related genes. The QTL may represent polymorphisms and differences in expression of unknown genes as well as those currently characterized. They should not represent posttranscriptional regulation unless there is a genetic component. Loci for one organ also may not necessarily reflect differences detectable by expression in that tissue. In previous work no marked difference between SWR and C57BL/6J mice was found in iron absorption from the intestine, and the major hepatic difference seemed to be in clearance of iron from the circulation.23 The demonstration that iron accumulation in the spleen occurs in the red pulp in particular is consistent with polymorphisms associated with degradation of red cells and recirculation of iron from this pool by macrophages.

A polymorphism of the hepcidin (Hamp) gene looks like a promising candidate to be a modifier of the Hfe-knockout mouse phenotype.17 As expected, we found suggestions of greater expression in the SWR mice, perhaps to readdress the balance of higher iron in tissues as a result of processes other than iron absorption. However, in the present study the chromosome interval for the chromosome 7 QTL (Table 3, Fig. 4B) did not encompass Hamp or the Hamp regulator Cebpa (12 cM).43Hmox1 seems to be a potential candidate for the suggestive liver chromosome 8 QTL, as judged by position and expression (Tables 3 and 5). The transferrin receptor (Tfrc) gene (21 cM) may be a candidate for the major hepatic QTL on chromosome 16. In QTL analyses of the hepatic porphyria induced by 2,3,7,8-tetrachlorodibenzo-p-dioxin in combination with iron overload, we found a highly significant locus on chromosome 11.30 QTL on chromosome 11 for hepatic iron level have been reported in searches for modifier genes in Hfe-null mice17 and in iron levels in chromosome substitution strains.26 Only a weak hepatic locus was detected in the present study (Fig. 4C), and it remains to be compared more closely with previous findings.

That the QTL observed for spleen were not identical to those for liver was in agreement with the observation that iron level in these tissues, although strain dependent, is predominantly associated with independent factors (see Fig. 3), as was previously observed for transferrin saturation.20 Most of the difference between strains in the spleen was associated with the highly significant QTL on chromosome 9, which is close to the transferrin gene (56 cM). Differences between mouse strains in transferrins detected by electrophoresis have been reported,44 and C57BL/6J mice have lower saturation of serum transferrin and greater resistance to iron loading than do other strains (Table 5).20 We did not find polymorphisms in the transferrin cDNA that would predict amino acid changes that might lead to a difference between C57BL/6J and SWR mice in iron-binding properties. Strain-dependent saturation of transferrin may simply reflect iron status, but polymorphisms of transferrin were linked to iron status in a black population.45 We did find a polymorphism in a transferrin-like gene positioned nearby that may act as a plasma inhibitor of carbonic anhydrase II.41

For the QTL on chromosome 2, higher iron level was linked with the C57BL/6J genotype. One candidate is Cybrd1, which encodes a Fe3+ reductase proposed to be associated with duodenal iron absorption.46 We did not find consistent strain differences in duodenal expression by immunohistochemistry or gene array or in other studies (not reported here) by immunoblotting and RT-PCR. A more promising candidate for this locus is the B2m gene.14, 47, 48 Although knockout of the gene leads to iron overload47, 48 we observed indications of greater expression in C57BL/6J mice than in SWR mice. β2-Microglobulin was first detected as a surface antigen in the liver and spleen of C57BL mouse strains but was not observed in SWR mice.49 Interaction between β2-microglobulin and the HFE protein reportedly modulates iron uptake and recycling of the transferrin receptor and iron uptake by cells.50 It is interesting to speculate whether the differences between strains in iron status have a basis in immunity28 in light of many strains originally being selected for immunological investigations.

In summary, we have shown that variability in basal iron status in apparently normal mice is the consequence of complex genetic factors that can be tissue specific. Identification and penetrance of the polymorphic genes, especially those on chromosomes 2, 7, 8, 9, and 16, in conjunction with cellular studies and comparison with syntenic regions in humans should lead to further appreciation of complex iron homeostasis in normal and pathological conditions.

Acknowledgements

We thank the members of the Biomedical Services staff for their expertise and Dr. Tatyana Chernova, Dr. David Dinsdale, Professor Yuri Dubrova, Jenny Edwards, Dr. Timothy Gant, Sam Hoffman, Dr. Katie Ridd, Linda Wilkinson, and Dr. Shu-Dong Zhang for their advice and assistance.

Ancillary