Evolutionary and Functional Properties of a Two-Locus β-Globin Polymorphism in Indian House Mice

Abstract

Electrophoretic surveys of hemoglobin (Hb) polymorphism in house mice from South Asia and the Middle East have revealed that two alternative β-globin haplotypes, Hbb^d and Hbb^p, are often present at intermediate frequencies in geographically disparate populations. Both haplotypes harbor two functionally distinct β-globin paralogs, HBB-T1 (which encodes the β-chain subunits of the major Hb isoform) and HBB-T2 (which encodes the β-chains of the minor Hb isoform). The Hbb^d and Hbb^p haplotypes share identical HBB-T1 alleles, but products of the alternative HBB-T2 alleles (d_minor and p_minor) are distinguished by two amino acid substitutions. To investigate the possible adaptive significance of the Hbb^d/Hbb^p polymorphism we conducted a population genetic analysis of the duplicated β-globin genes of Indian house mice (Mus castaneus) in conjunction with experimental studies of Hb function in inbred strains of mice that carry the alternative Hbb^d and Hbb^p haplotypes. The main objectives of this study were (i) to characterize patterns of nucleotide polymorphism and linkage disequilibrium in the duplicated β-globin genes of M. castaneus, (ii) to test the hypothesis that the Hbb^d and Hbb^p haplotypes are maintained as a balanced polymorphism, and (iii) to assess whether allelic differences in the alternative minor Hb isoforms (d_minor and p_minor) are associated with different O₂-binding properties. A multilocus analysis of polymorphism and divergence revealed that levels of diversity at the HBB-T2 gene exceeded neutral expectations, and reconstructed haplotype networks for both β-globin paralogs revealed extensive allele sharing with several other closely related species of Mus. However, despite this suggestive evidence for balancing selection, O₂-equilibrium curves revealed no discernible functional differences between red cell lysates containing the d_minor and p_minor Hb isoforms. If the d_minor and p_minor alleles are maintained as a balanced polymorphism, our results indicate that the associated fitness variance is not directly related to respiratory functions of Hb.

BALANCING selection at a particular locus is expected to produce elevated levels of nucleotide diversity and linkage disequilibrium (LD) due to the partitioning of sequence variation between unusually long-lived allele classes (Hudson and Kaplan 1988; Charlesworth et al. 2003; Charlesworth 2006). Although elevated levels of nucleotide polymorphism at a particular gene may provide suggestive evidence for a history of balancing selection, conclusive inferences regarding the selective maintenance of allelic polymorphism ultimately require experimental evidence that the alternative alleles are functionally distinct. The well-characterized β-globin polymorphism in house mice (genus Mus) represents a system where it is possible to integrate evolutionary and functional approaches to evaluate the role of balancing selection in maintaining protein polymorphism.

The two tandemly duplicated β-globin genes of house mice, HBB-T1 and HBB-T2, encode the β-chain subunits of adult hemoglobin (Hb) and are separated by ∼12–15 kb on chromosome 7 (Hoffmann et al. 2008; Figure 1). Five main classes of β-globin haplotypes have been characterized in natural populations of house mice: Hbb^d, Hbb^p, Hbb^s, Hbb^w1, and Hbb^w2 (Sato et al. 2008; Runck et al. 2009). Electrophoretic surveys of β-globin polymorphism in natural populations of Mus musculus and M. domesticus have revealed that the Hbb^d and Hbb^s haplotypes are consistently present at intermediate frequencies in population samples from across the species' range, a pattern that is not paralleled at other unlinked autosomal genes (Petras 1967; Selander et al. 1969a,b; Selander and Yang 1969; Berry and Murphy 1970; Wheeler and Selander 1972; Myers 1974; Berry and Jakobson 1975; Berry and Peters 1975, 1977, 1981; Berry et al. 1978; Sage 1981; Petras and Topping 1983; Sage et al. 1986). This striking uniformity of two-locus haplotype frequencies has led a number of authors to conclude that the polymorphism may be maintained by some form of balancing selection (reviewed by Berry 1978). More recently, surveys of nucleotide variation at the two β-globin paralogs have revealed patterns consistent with the idea that the Hbb^d and Hbb^s haplotypes are maintained as a balanced polymorphism. First, high levels of silent site polymorphism at the HBB-T1 and HBB-T2 genes of M. domesticus exceeded neutral expectations on the basis of levels of interspecific divergence (Storz et al. 2007). Second, the Hbb^d and Hbb^s haplotypes are shared among multiple Eurasian species in the subgenus Mus, indicating that the time to the most recent common ancestor of the two haplotypes may predate multiple speciation events (Runck et al. 2009). However, transpecific polymorphism appears to be relatively common in house mice due to incomplete lineage sorting and introgressive hybridization (e.g., Salcedo et al. 2007; Geraldes et al. 2008). Thus, central assumptions of standard neutrality tests may often be violated, making it especially important to use direct, experimental approaches to test hypotheses about balancing selection.

Figure 1.—

Genomic structure of the β-globin gene family of the house mouse (BALB/c strain) on chromosome 7.

The Hbb^d haplotype harbors two distinct β-globin paralogs that are distinguished from one another by 9 amino acid substitutions. The more highly expressed HBB-T1 gene encodes the β-chain subunits of the major Hb isoform (isoHb), d_major, whereas HBB-T2 encodes the β-chains of the minor isoHb, d_minor (Hutton et al. 1962; Gilman 1972, 1974; Whitney 1977). In contrast to the two divergent β-globin paralogs on the Hbb^d haplotype, the Hbb^s haplotype harbors two β-globin paralogs that are identical in sequence due to a history of HBB-T1→HBB-T2 gene conversion (Erhart et al. 1985; Storz et al. 2007). Consequently, mice that are homozygous for the Hbb^s haplotype synthesize a single β-chain isoHb during postnatal life. The Hbb^d and Hbb^s haplotypes are distinguished by 3 amino acid substitutions at HBB-T1 and 11 amino acid substitutions at HBB-T2 (Erhart et al. 1985; Storz et al. 2007). One of the most salient differences between Hbb^d and Hbb^s is that β-chain products of HBB-T1 and HBB-T2 on the Hbb^d haplotype (d_major and d_minor, respectively) contain a derived cysteine residue at site 13, which appears to play an important role in intraerythrocyte glutathione metabolism and nitric oxide metabolism (Miranda 2000; Giustarini et al. 2006; Hempe et al. 2007).

A polymorphism involving the Hbb^d and Hbb^p haplotypes has been documented in house mice from South Asia and the Middle East (Minezawa et al. 1979; Ritte and Neufeld 1982; Miyashita et al. 1985; Kawashima et al. 1995), which exhibits some similarities to the Hbb^d/Hbb^s polymorphism in house mice from Europe and the Americas. Similar to the pattern described for the Hbb^d/Hbb^s polymorphism, the Hbb^d and Hbb^p haplotypes are present at intermediate frequencies in geographically disparate population samples from across the range of the Asian house mouse M. castaneus (Miyashita et al. 1985; Kawashima et al. 1995). As in the case with the Hbb^d haplotype, the HBB-T1 and HBB-T2 paralogs on the Hbb^p haplotype encode the β-chain subunits of the major and minor isoHbs (p_major and p_minor), respectively. The Hbb^d and Hbb^p haplotypes share identical HBB-T1 sequences (d_major = p_major), but they are distinguished by two amino acid substitutions at HBB-T2: 22Glu→Ala and 23Val→Ile (d_minor→p_minor, in both cases). The sequence identity between HBB-T1 alleles on the Hbb^d and Hbb^p haplotypes reflects the fact that Hbb^p is a recombinant chromosome that was produced by intergenic crossing over between Hbb^d and Hbb^w1 parental chromosomes (Ueda et al. 1999; Sato et al. 2008; Runck et al. 2009). Consequently, Hbb^p carries an HBB-T1 allele derived from Hbb^d and an HBB-T2 allele derived from Hbb^w1. In contrast to the structural differences between products of Hbb^d and Hbb^s in M. domesticus and M. musculus, the β-chain isoHbs produced by Hbb^d and Hbb^p do not differ in cysteine content. If there is some functionally significant difference between the β-chain isoHbs produced by the Hbb^d and Hbb^p haplotypes it must be related to the two amino acid differences that distinguish the minor isoHbs, d_minor and p_minor.

To investigate the possible adaptive significance of the Hbb^d/Hbb^p polymorphism we conducted a population genetic analysis of the HBB-T1 and HBB-T2 genes in a natural population of M. castaneus from northern India. In conjunction with this evolutionary analysis of sequence variation, we also conducted an experimental study of Hb function in inbred strains of mice that carry each of the alternative two-locus β-globin haplotypes. The main objectives of this study were (i) to characterize patterns of nucleotide polymorphism and LD in the duplicated β-globin genes of M. castaneus, (ii) to test the hypothesis that the Hbb^d and Hbb^p haplotypes are maintained as a balanced polymorphism, and (iii) to assess whether allelic differences in the alternative minor isoHbs (d_minor and p_minor) are associated with different O₂-binding properties.

MATERIALS AND METHODS

Samples for the population genetic analysis:

Our survey of nucleotide variation at the HBB-T1 and HBB-T2 genes was based on a total of 24 M. castaneus specimens that were collected by B. Harr in northern India [for details, see Baines and Harr (2007) and Geraldes et al. (2008)]. The mice used in our survey were collected from four closely situated localities within a 130 km radius that spanned the border between the Indian states of Himachal Pradesh and Uttarakhand. Mice from the Indian subcontinent have been referred to as M. bactrianus by some authors (see Boursot et al. 1993). Following Baines and Harr (2007) and Geraldes et al. (2008), we refer to the specimens from northern India as M. castaneus.

PCR, cloning, and sequencing:

Genomic DNA was extracted from ethanol-preserved liver tissue of each mouse using DNeasy kits (Qiagen, Valencia, CA). We cloned and sequenced both alleles of the HBB-T1 and HBB-T2 genes in the full sample of mice (48 experimentally phased sequences per gene, 96 sequences total). We therefore obtained complete diploid genotypes for both HBB paralogs in each mouse, and we were able to determine the exact haplotype phase of all heterozygous sites. Amplification of the complete coding regions of HBB-T1 and HBB-T2 followed established protocols (Runck et al. 2009). For HBB-T1, the sequenced fragment was 1809 bp in length, including 329 bp of 5′ flanking sequence and 276 bp of 3′ flanking sequence. For HBB-T2, the sequenced fragment was 1793 bp, including 338 bp of 5′ flanking sequence and 300 bp of 3′ flanking sequence. PCR products were cloned into pCR4-TOPO vector following the manufacturer's protocols (Invitrogen, Carlsbad, CA). For each species, we sequenced a total of 8–10 colonies per gene using the vector primers T3 and T7 (54° annealing). Automated DNA sequencing was performed on an ABI 3730 capillary sequencer using Big Dye chemistry (Applied Biosystems, Foster City, CA). Sequences were deposited in GenBank under the accession numbers GU057161–GU057256.

Analysis of DNA sequence variation:

HBB-T1 and HBB-T2 sequences were aligned and assembled into contigs using ClustalX (Thompson et al. 1997) and Sequencher (Gene Codes, Ann Arbor, MI). For the purpose of reconstructing the evolutionary history of the HBB-T1 and HBB-T2 sequences from Indian M. castaneus, and to test for evidence of transpecific polymorphism, we reconstructed haplotype networks that included previously published sequence data from the orthologous β-globin genes of four additional species of Mus (M. domesticus, M. macedonicus, M. musculus, and M. spicilegus) and an additional M. castaneus specimen from Thailand (Storz et al. 2007; Runck et al. 2009). M. castaneus, M. domesticus, M. macedonicus, M. musculus, and M. spicilegus are each known to carry the Hbb^d haplotype, and M. castaneus and M. musculus are also known to carry the Hbb^p haplotype (Runck et al. 2009). Haplotype networks were constructed using equally weighted characters in the program NETWORK v4.5.1 (www.fluxus-engineering.com). To ensure that a full median network was calculated, we set the weighted genetic distance measure (ɛ) to equal the maximum number of pairwise differences between haplotypes at each paralog (ɛ = 10 for HBB-T1 and ɛ = 11 for HBB-T2). We used the maximum parsimony option to remove unnecessary median vectors in the full median network (Polzin and Daneschmand 2003).

After binning HBB-T1 and HBB-T2 amino acid sequences into discrete allele classes, we calculated Weir and Cockerham's (1984) estimator of the inbreeding coefficient, F. To assess whether observed genotype frequencies deviated from Hardy–Weinberg equilibrium we used the exact test of Guo and Thompson (1992). We used a similar Markov-chain contingency table method to test the null hypothesis that genotypes at one β-globin gene were independent of genotypes at the other gene.

Summary statistics of nucleotide polymorphism and LD were computed with the programs SITES (Hey and Wakeley 1997) and DnaSP v5 (Librado and Rozas 2009). To detect intragenic recombination within each of the two β-globin paralogs, we used the four-gamete test of Hudson and Kaplan (1985) to estimate R_M, the minimum number of recombination events in the history of the sample. We computed two different measures of DNA sequence variation: nucleotide diversity, π (the average number of pairwise differences between sequences) and Watterson's (1975) estimator of the scaled mutation rate, θ_W (4Nu, where N is the effective population size and u is the mutation rate per nucleotide). We used variation at coding and noncoding sites to compute Hey and Wakeley's (1997) estimator of the population recombination rate, γ. We used the method of Betrán et al. (1997) to test for evidence of gene conversion between the two β-globin paralogs.

As a measure of LD between pairs of nucleotide polymorphisms, we used the squared allele-frequency correlation, r², and we used Fisher's exact test to determine the probability of obtaining estimates of LD that were more extreme than the observed values under the null hypothesis of linkage equilibrium. The analysis was based on biallelic nucleotide polymorphisms where the minor allele was present at least twice in the sample. We used nonlinear regression to model the decay of intragenic LD as a function of physical distance under a model of recombination–drift equilibrium that incorporated mutation (Hill and Weir 1988). Specifically, we used a nonlinear regression model based on the Gauss–Newton algorithm, as implemented in the nls function of the R statistical computing package (www.r-project.org). To summarize the effects of recombination on intragenic LD, we also computed the ZZ test statistic of Rozas et al. (2001), which measures the difference between the average r² between adjacent nucleotide polymorphisms and the average of pairwise r² values across the entire gene. To compute confidence intervals of the ZZ test statistic we conducted 10,000 coalescent simulations (with no recombination) that were conditioned on the observed number of segregating sites.

Using data on silent site polymorphism at both β-globin paralogs, we conducted neutrality tests on the basis of two different summary statistics: Tajima's (1989) D, which provides a measure of the site-frequency distribution, and Kelly's (1997) Z_nS, which provides a measure of intralocus LD. We obtained critical values for each test statistic by conducting coalescent simulations as described above.

To assess whether relative levels of polymorphism and divergence deviated from neutral expectations, we conducted a multilocus Hudson–Kreitman–Aguadé (HKA) test (Hudson et al. 1987). The analysis included polymorphism data from the HBB-T1 and HBB-T2 genes in addition to previously published data from 8 loci that were sequenced in the same panel of Indian M. castaneus specimens (Geraldes et al. 2008). The set of effectively unlinked reference loci included four autosomal genes (Chrng, Med19, Prpf3, and Clcn6), two X-linked genes (G6pdx and Ocr1), one Y-linked gene (Jarid 1d), and the mtDNA control region. For each of the 10 loci we used orthologous sequences from M. caroli (Geraldes et al. 2008; Runck et al. 2009) to estimate locus-specific values of D_xy, the average pairwise sequence divergence between species (Nei and Kumar 2000). Since the 5′ end of the M. caroli HBB-T1 gene has experienced a history of gene conversion by HBB-T2 (Runck et al. 2009), we treated the 624-bp conversion tract as missing data. Thus, for the HBB-T1 gene, our estimate of net divergence between M. castaneus and M. caroli was based on 1369 bp of orthologous sequence that spanned intron 2, exon 3, and the 3′-UTR. For the mtDNA control region, the estimate of net sequence divergence between M. castaneus and M. caroli was corrected for multiple hits (see Geraldes et al. 2008). We jointly estimated neutral parameters from all 10 loci to obtain expected values for the HKA test, and we then used the resultant estimates to conduct coalescent simulations using the HKA program (http://genfaculty.rutgers.edu/hey/software).

Experimental analysis of Hb function:

To test for functional differences between the β-chain isoHbs produced by the Hbb^d haplotype (d_major and d_minor) and the Hbb^p haplotype (p_major and p_minor), we measured O₂-binding properties of hemolysates from the BALB/c and MSM/s inbred strains. Blood samples from BALB/c (which is homozygous for Hbb^d) and MSM/s (which is homozygous for Hbb^p) were obtained from the Jackson Lab (Bar Harbor, ME). Blood samples from the MSM/s strain were procured under a material transfer agreement with the National Institute of Genetics (Mishima, Japan).

Hemolysates were prepared according to standard methods and were stripped of organic phosphates and other ionic cofactors as described previously (Storz et al. 2009). The isoHb composition of hemolysates from each mouse strain was confirmed by using thin-layer isoelectric focusing (PhastSystem, GE Healthcare Biosciences, Piscataway, NJ). Using a modified diffusion chamber, O₂ equilibria of Hb solutions were measured in 10 mm HEPES buffer, pH 7.4, at constant temperature, 37°. The met-Hb enzymatic reducing system of Hayashi et al. (1973) was used to prevent oxidation of ferrous heme. Changes in the absorbance of Hb solutions were recorded in conjunction with stepwise changes in the partial pressure of O₂ [PO₂] inside the chamber (prepared using cascaded Wösthoff gas-mixing pumps; Weber 1981, 1992; Weber et al. 2004). Values of P₅₀ (the PO₂ at 50% oxygenation of the heme groups) and n₅₀ (Hill's cooperativity coefficient at P₅₀) were interpolated from linear Hill plots (log ([OxyHb]/[Hb]) vs. log PO₂). The P₅₀ values for the stripped (i.e., cofactor-free) hemolysates provide an inverse measure of the intrinsic Hb–O₂ binding affinities of the major and minor isoHbs occurring in their natural relative concentrations. To test for differences in cofactor sensitivity between the d- and p-type Hbs, we measured O₂-equilibrium curves for each sample in the absence of allosteric cofactors (stripped hemolysates), in the presence of 2,3-diphosphoglycerate (DPG), in the presence of Cl⁻ ions (added as potassium chloride, KCl), and in the presence of both cofactors [(Cl⁻), 0.10 m; (NaHEPES), 0.1 m; DPG/Hb tetramer ratio, 2.0; (Heme), 0.16 mm].

RESULTS

Genetic variation:

Our survey of nucleotide polymorphism in the HBB-T1 and HBB-T2 genes revealed that Indian house mice segregate two main β-globin haplotypes that appear to be referable to Hbb^d and Hbb^p. The survey of nucleotide polymorphism at HBB-T1 revealed the presence of three distinct allele classes that were present at roughly equal frequencies (Table 1). Allele classes T1^A, T1^B, and T1^C were defined by the following two-site amino acid combinations at residue positions 16 and 20: T1^A = Gly-Ser, T1^B = Gly-Ala, and T1^C = Ala-Ser. The amino acid sequence of the T1^A allele is identical to the canonical d_major and p_major β-globin sequences that were originally characterized in the BALB/c and AU/SsJ inbred strains, respectively (Gilman 1972, 1974; Erhart et al. 1985). The T1^B and T1^C alleles differ from the canonical d_major/p_major sequence by single amino acid substitutions. The survey of nucleotide variation at HBB-T2 revealed the presence of two distinct allele classes, referable to d_minor and p_minor, that were also present at roughly equal frequencies (Table 1). The HBB-T2 allele classes were distinguished by the following two-site amino acid combinations at residue positions 22 and 23: d_minor = Glu-Ile and p_minor = Ala-Val. Relative to the expected genotype frequencies at Hardy–Weinberg equilibrium, both β-globin genes exhibited a highly significant deficit of heterozygotes (HBB-T1: F = 0.573, P < 0.0001; HBB-T2: F = 0.676, P = 0.0011). A more detailed inspection of genotype frequencies revealed that the observed heterozygote deficits were mainly attributable to population structure. Although the d_minor and p_minor alleles were present at respective frequencies of 0.54 and 0.46 in the pooled sample, the three more northern samples from Himachal Pradesh were enriched for p_minor (frequency = 0.86, n = 11 mice) and the more southern sample from Uttarakhand was enriched for d_minor (frequency = 0.88, n = 13 mice).

TABLE 1.

Amino acid variation among HBB-T1 and HBB-T2 alleles in Indian house mice

Gene	Allele class	Frequency	Amino acid
			16	20	22	23
HBB-T1	T1A	0.27	Gly	Ser	Glu	Val
	T1B	0.35	Gly	Ala	Glu	Val
	T1C	0.38	Ala	Ser	Glu	Val
HBB-T2	dminor	0.54	Gly	Ser	Glu	Ile
	pminor	0.46	Gly	Ser	Ala	Val

Both HBB-T1 and HBB-T2 exhibited high levels of nucleotide diversity at silent sites (π = 0.0051 and 0.0110, respectively; Table 2). The estimates of θ_W for the HBB-T1 and HBB-T2 genes (θ_W = 0.0071 and 0.0148, respectively; Table 2) exceeded the upper range of values for the seven unlinked nuclear loci that were sequenced in the same panel of mice (θ_W = 0.0014–0.0062, where values for the X- and Y-linked loci were multiplied by 4/3 and 4, respectively, to account for differences in effective population size; Geraldes et al. 2008). The multilocus HKA test was highly significant due to a higher-than-expected level of polymorphism at the HBB-T2 gene (χ_d.f.=9² = 34.661, P = 0.00007; Table 3). The test also remained highly significant when the analysis was restricted to the six autosomal loci (χ_d.f.=5² = 22.644, P = 0.00030). The HBB-T1 and HBB-T2 genes both exhibited an excess of low-frequency polymorphisms, as indicated by negative values for Tajima's D, but the skews in the site-frequency distributions were not statistically significant in either case (Table 2).

TABLE 2.

Summary of nucleotide polymorphism and linkage disequilibrium at the HBB-T1 and HBB-T2 genes of Indian house mice

Gene	Allele class	Length (bp)	N	S	H	HD	π(silent)	θw(silent)	Tajima's D	k	Rm	γ/bp	ZnS	ZZ
HBB-T1	All	1815	48	61	24	0.946	0.00505	0.00708	−0.887	10.17	5	0.00975	0.0936	0.1587*
	T1A	1816	13	9	5	0.628	0.00112	0.00160	−1.188	2.03
	T1B	1816	17	10	7	0.816	0.00156	0.00163	−0.148	2.83
	T1C	1814	18	44	12	0.954	0.00485	0.00673	−1.141	8.98
HBB-T2	All	1802	48	125	35	0.986	0.01101	0.01475	−0.8377	21.38	8	0.01052	0.0739	0.1093*
	dminor	1804	26	78	18	0.966	0.00692	0.01060	−1.348	12.53
	pminor	1803	22	50	17	0.978	0.00440	0.00761	−1.664	7.76

*P < 0.05. N, sampled chromosomes; S, total number of segregating sites; H, number of haplotypes; H_D, haplotype diversity; and k, average number of nucleotide differences between sequences.

TABLE 3.

Multilocus HKA test involving the HBB-T1 and HBB-T2 genes and eight reference loci

			Polymorphism				Divergence
Locus	Chromosome	I	θ (2.5–97.5%)	S/ES	Var	Dev	D/ED	Var	Dev
HBB-T1	7	1.00	0.0056 (0.0038–0.0084)	57/45.46	215.82	0.618	44/55.54	126.82	1.051
HBB-T2	7	1.00	0.0082 (0.0056–0.0119)	118/65.39	417.90	6.624	44/96.61	312.25	8.865
Chrng	1	1.00	0.0059 (0.0040–0.0093)	61/55.44	308.90	0.100	42/47.56	99.80	0.309
Med19	2	1.00	0.0057 (0.0039–0.0082)	15/44.05	193.69	4.358	92/62.95	154.48	5.464
Prpf3	3	1.00	0.0032 (0.0021–0.0046)	31/35.25	132.57	0.136	55/50.75	110.25	0.164
Clcn6	4	1.00	0.0058 (0.0039–0.0082)	54/52.68	270.00	0.006	75/76.32	210.89	0.008
G6pdx	X	0.75	0.0037 (0.0025–0.0053)	19/26.35	95.43	0.567	63/55.65	99.13	0.545
Orc1	X	0.75	0.0056 (0.0038–0.0081)	32/31.89	141.81	0.000	68/68.11	133.26	0.000
Jarid 1d	Y	0.25	0.0072 (0.0049–0.0105)	4/13.56	40.99	2.232	68/58.44	64.72	1.413
Control region	mtDNA	0.25	0.0598 (0.0438–0.0803)	33/53.93	338.90	1.292	329/308.07	482.64	0.907

χ_d.f.=9² = 34.661, P = 0.00007. I, inheritance scalar (1.00 for autosomal loci, 0.75 for X-linked loci and 0.25 for the Y-linked locus and mtDNA); S, observed number of segregating sites; ES, expected number of segregating sites; Var, variance under the model; Dev, deviation of each observed value from the model-based expectation; D, observed divergence to M. caroli; and ED, expected divergence.

Levels and patterns of LD:

In addition to the departure from Hardy–Weinberg genotype frequencies, we also observed a highly nonrandom association between genotypes at each of the two β-globin genes (χ_d.f.=2² = 19.807, P < 0.0001). This high level of intergenic LD is partly attributable to population structure, as the difference in allele frequencies between the northern and southern collection localities augments the covariance in allele frequencies between the two β-globin genes. At the HBB-T1 gene, Fisher's exact test revealed significant LD between 189 of 703 pairwise comparisons (56 of which remained statistically significant after Bonferroni correction), and at the HBB-T2 gene, significant LD was observed between 563 of 2485 such comparisons (304 of which remained significant after Bonferroni correction). Neither of the two β-globin genes exhibited an excess of intragenic LD, as indicated by nonsignificant Z_nS values (Table 2). HBB-T1 and HBB-T2 both showed evidence for a history of intragenic recombination (R_m = 5 and 8, respectively), and we detected no evidence for interparalog gene conversion.

To assess whether the elevated polymorphism at HBB-T1 and HBB-T2 could plausibly be explained by the effects of balancing selection at a linked locus, we measured rates of decay of intralocus LD at each of the two β-globin genes. The rationale is that if LD decays to near background levels within each individual gene, then the observed allelic dimorphism at the HBB-T1 and HBB-T2 genes cannot be ascribed to the effects of selection at a linked locus. At both β-globin genes, mean r² declined to <0.1 within 200 bp (Figure 2). Although a relatively small number of nonrandom associations persisted for site pairs separated by up to 1.5 kb, it is clear that LD does not extend very far into flanking chromosomal regions. Consistent with these results, estimated values of the ZZ test statistic were significantly positive for both HBB-T1 and HBB-T2 (Table 2), indicating that intragenic recombination has played an important role in randomizing pairwise associations between polymorphisms within each gene.

Figure 2.—

Rates of decay of gametic linkage disequilibrium within the HBB-T1 and HBB-T2 genes. The red lines show nonlinear regressions of r² against physical distance using a mutation–recombination–drift model (see text for details). Open symbols denote r² estimates for pairs of polymorphic sites in the HBB-T1 gene, and solid symbols denote estimates for pairs of polymorphic sites in the HBB-T2 gene.

Evolutionary relationships among house mouse β-globin sequences:

The minimum spanning network of HBB-T1 coding sequences revealed that many of the HBB-T1 alleles from M. castaneus were more similar to alleles from M. domesticus than they were to other HBB-T1 alleles from M. castaneus (Figure 3A). The Hbb^w1 and Hbb^w2 alleles of M. musculus (Kawashima et al. 1991; Sato et al. 2008) were the most distantly related sequences in the network of HBB-T1 alleles. Relative to HBB-T1, we observed an even greater degree of transpecific polymorphism at the HBB-T2 gene. The d_minor alleles from M. castaneus were far more closely related to d_minor alleles from M. domesticus, M. macedonicus, M. musculus, and M. spicilegus than they were to p_minor alleles from M. castaneus (Figure 3B). Even within the d_minor allele class, many d_minor sequences from M. castaneus were more similar to those of M. domesticus and M. musculus than they were to other conspecific d_minor sequences.

Figure 3.—

Median joining network showing relationships among β-globin coding sequences from M. castaneus and four other species of Mus that are known to segregate the Hbb^d and Hbb^p haplotypes (M. domesticus, M. macedonicus, M. musculus, and M. spicilegus). Haplotype networks for the HBB-T1 and HBB-T2 paralogs are shown in A and B, respectively. The size of each circle is proportional to the corresponding haplotype frequency. Inferred intermediate haplotypes are shown as black circles on branches connecting observed haplotypes. Branches between haplotypes indicate one mutational step, and tick marks denote additional steps. Amino acid changes in HBB-T1 sequences are shown in relation to the canonical d_major/p_major sequence from the BALB/c and MSM/s inbred strains, and amino acid changes in HBB-T2 are shown in relation to the canonical d_minor sequence from BALB/c. The three allele classes at HBB-T1 are indicated in A. The black line in B separates haplotypes that are referable to the d_minor and p_minor allele classes.

Functional properties of house mouse Hbs:

To test for functional differences between the minor β-chain isoHbs produced by the Hbb^d haplotype (d_minor) and the Hbb^p haplotype (p_minor), we compared O₂-binding properties of hemolysates from the BALB/c and MSM/s inbred strains. Hb–O₂ affinity is modulated by various allosteric cofactors, particularly protons (responsible for the Bohr effect) and organic and inorganic anions, such as DPG and Cl⁻ ions (Weber and Fago 2004; Storz and Moriyama 2008). In the red blood cells of most mammalian species, the binding of these effectors stabilize the low-affinity, deoxygenated “tense state” conformation of the Hb tetramer relative to the oxygenated “relaxed state,” thereby decreasing Hb–O₂ affinity. Thus, even if the d_minor and p_minor Hbs exhibit similar intrinsic O₂-binding affinities, differences in their cofactor sensitivities could still account for significant differences in blood–O₂ affinity.

The comparison between hemolysates containing the d_minor and p_minor isoHbs revealed no significant difference in O₂-binding properties (Figure 4). Across all treatments, the O₂-affinities and cooperativity coefficients were nearly identical (Table 4). In summary, the O₂-binding measurements revealed no discernible differences in the respiratory functions of d_minor and p_minor isoHbs under physiological conditions.

Figure 4.—

O₂-equilibrium curves of stripped house mouse Hbs at pH 7.40 and 37° in the presence and absence of allosteric cofactors [(Cl⁻), 0.10 m; (NaHEPES), 0.1 m; DPG/Hb tetramer ratio, 2.0; (Heme), 0.16 mm] and in the presence of the met-Hb reductase system (Hayashi et al. 1973). Representative curves for the Hbs from two inbred strains, BALB/c (which is homozygous for Hbb^d) and MSM/s (which is homozygous for Hbb^p), are shown in A and B, respectively. C shows mean values (± SEM) of P₅₀ (n = 3) measured under the above conditions and indicates that the β-chain isoHbs produced by the Hbb^d haplotype and the Hbb^p haplotype do not differ in intrinsic O₂ affinity (as revealed by the comparison of stripped Hbs) or in sensitivity of Hb–O₂ to the presence of allosteric cofactors such as DPG or Cl⁻ ions.

TABLE 4.

O2 binding properties (mean ± SEM, n = 3) of stripped hemolysates from the BALB/c and MSM/s inbred strains (which are homozygous for the Hbbd and Hbbp β-globin haplotypes, respectively)

Strain (β-haplotype)	Conditions	P50 (torr)	n50
BALB/c (Hbbd)	Stripped	8.44 ± 0.09	2.46 ± 0.05
	DPG	11.88 ± 0.17	2.65 ± 0.09
	KCl	16.27 ± 0.30	2.68 ± 0.12
	DPG + KCl	17.53 ± 0.27	2.78 ± 0.13
MSM/s (Hbbp)	Stripped	8.76 ± 0.04	2.40 ± 0.02
	DPG	11.64 ± 0.07	2.50 ± 0.04
	KCl	17.74 ± 0.06	2.67 ± 0.02
	DPG + KCl	17.03 ± 0.27	2.56 ± 0.09

P₅₀ and n₅₀ values indicate the O₂ tensions and cooperativity coefficients at half-saturation, respectively. O₂ equilibria were measured in 0.1 m HEPES buffer at pH 7.40, 37°, in the absence of allosteric cofactors (stripped hemolysates), in the presence of DPG alone, in the presence of KCl alone, and in the presence of both cofactors [(Cl⁻), 0.10 m; DPG/Hb tetramer ratio, 2.0; (Heme), 0.16 mm].

DISCUSSION

Consistent with the results of electrophoretic surveys of house mice from other localities in South Asia (Miyashita et al. 1985; Kawashima et al. 1995), our survey of nucleotide polymorphism in the HBB-T1 and HBB-T2 genes revealed that Indian M. castaneus segregate two main β-globin haplotypes that are referable to Hbb^d and Hbb^p. Remarkably, many of the HBB-T1 and HBB-T2 alleles that we recovered in our sample of Indian M. castaneus were shared by other species of Mus. The transpecific polymorphism was especially pervasive at the HBB-T2 gene, as identical d_minor and p_minor alleles were shared among representatives of M. castaneus, M. domesticus, M. macedonicus, and M. spicilegus. A lesser degree of shared polymorphism among M. castaneus and M. domesticus was also documented at several of the other unlinked reference loci (Geraldes et al. 2008), which appears to reflect the combined effects of unsorted ancestral polymorphism and introgressive hybridization. Hybridization between M. castaneus and M. musculus has been documented in northern China and in Japan (Bonhomme et al. 1989; Boursot et al. 1993), and hybridization between M. castaneus and M. domesticus has been documented in California (Orth et al. 1998). At least in the case of M. castaneus, M. domesticus, and M. musculus, which are thought to have diverged from one another 350,000–900,000 years ago (She et al. 1990; Boursot et al. 1996; Suzuki et al. 2004; Salcedo et al. 2007; Tucker 2007; Geraldes et al. 2008), retained ancestral polymorphism and introgressive hybridization appear to be sufficiently common that shared polymorphism can often be explained without invoking balancing selection.

Functional properties of BALB/c and MSM/s hemoglobins:

We detected no differences in the O₂ equilibria of hemolysates from BALB/c (containing the d_minor isoHb) and MSM/s (containing the p_minor isoHb) under physiological conditions. This indicates that structural differences in the products of d_minor and p_minor do not have any significant effects on blood–O₂ transport. As shown in Figure 4 and Table 4, Hbs from BALB/c and MSM/s showed no physiologically significant differences in O₂ affinity (P₅₀) or cooperativity (n₅₀) under identical buffer conditions. As previously observed for Hbs of the deer mouse, Peromyscus maniculatus (Storz et al. 2009), Hb–O₂ affinity was not markedly reduced in the presence of DPG alone (ΔlogP_{50(DPG-stripped)} = 0.15 and 0.13 in BALB/c and MSM/s, respectively; Table 4). The magnitude of the DPG effect in the house mouse Hbs was similar to that observed in lowland deer mice (ΔlogP_{50(DPG-stripped)} = 0.09–0.14), and is indicative of weak DPG binding to the tense-state deoxyHb structure. In contrast to the relatively weak DPG effect, the effect of Cl⁻ ions was much more pronounced (ΔlogP_{50(KCl-stripped)} = 0.29 and 0.31 in BALB/c and MSM/s, respectively; Table 4), which may be attributable to the presence of one or more distinct Cl⁻ binding sites in the tense-state quaternary structure.

Reconciling the results of neutrality tests and functional tests:

A number of studies have documented patterns of DNA sequence variation that provide suggestive evidence for the maintenance of protein polymorphism by balancing selection (e.g., Filatov and Charlesworth 1999; Cork and Purugganan 2002; Tian et al. 2002; Baysal et al. 2007; Ferguson et al. 2008; Fumagalli et al. 2009). However, there are very few case studies where indirect, statistical evidence for balancing selection has been combined with documented functional differences between alternative alleles (for one notable exception involving house mice, see Johnsen et al. 2009). Our combined evolutionary and functional analysis of the two-locus β-globin polymorphism in Indian house mice revealed seemingly contradictory results. The population genetic analysis revealed excess levels of polymorphism at HBB-T2 that could not be reconciled with the expectations of a neutral equilibrium model, and yet the analysis of protein function revealed no discernible physiological differences between d_minor and p_minor isoHbs. There is clearly no basis for fitness variation among genotypes if the products of alternative alleles are functionally identical.

One possible explanation for the discrepancy is that results of the HKA test reflect a violation of model assumptions related to demographic history or population structure. Since the ratio of intraspecific polymorphism to between-species divergence can be inflated by population structure (Wakeley 2000), results of the HKA test may be biased if the test is applied to a set of loci that are characterized by different levels of population structure (Ingvarsson 2004). This may be a common problem in studies of house mice, since natural populations often show evidence of historical admixture and levels of introgression are highly variable among different genomic regions (e.g., Payseur et al. 2004; Geraldes et al. 2008; Teeter et al. 2008, 2010). Another possibility is that the elevated polymorphism that we observe at the β-globin genes reflects the effects of diversity-enhancing selection at one or more linked loci. We cannot rule out this possibility, but given the observed decay of intragenic LD at each of the two β-globin genes (Figure 3), it does not seem plausible that levels of diversity could be affected by associative overdominance or some other form of diversity-enhancing selection at a linked locus. If the unusual patterns of variation at the β-globin genes of Indian house mice can be explained by admixture or other complexities of population structure, then we should expect that surveys of additional autosomal genes will uncover similar examples of allelic dimorphism and allele sharing between M. castaneus and other closely related species.

Finally, it is also possible that the β-globin polymorphism in M. castaneus is maintained by balancing selection, but that the adaptive variation in protein function is not directly related to blood–O₂ transport. In contrast to the allelic differences in cysteine content that distinguish products of Hbb^d and Hbb^s in M. domesticus and M. musculus, amino acid differences between the d_minor and p_minor isoHbs of M. castaneus have no effect on the metabolism of thiol reactants or intraerythrocytic redox balance (which can affect aspects of the host immune response to pathogenic infection). Nonetheless, we cannot rule out the possibility that the β-globin polymorphism in M. castaneus affects some unknown biochemical function, especially in light of recent discoveries concerning the expression of globin chain monomers in nonerythroid cells (e.g., Liu et al. 1999; Mansergh et al. 2008; Nishi et al. 2008; Biagioli et al. 2009; Richter et al. 2009; Schelshorn et al. 2009). In light of our experimental results, we conclude that amino acid differences between products of the d_minor and p_minor alleles are functionally inconsequential with regard to the respiratory physiology of house mice. If the d_minor and p_minor alleles are maintained as a balanced polymorphism in M. castaneus, either directly or indirectly, the associated variance in fitness appears to be unrelated to respiratory functions of Hb.