Epistasis Among Adaptive Mutations in Deer Mouse Hemoglobin

See allHide authors and affiliations

Science  14 Jun 2013:
Vol. 340, Issue 6138, pp. 1324-1327
DOI: 10.1126/science.1236862

Holding Your Breath

Hemoglobin and myoglobin are widely responsible for oxygen transport and storage (see the Perspective by Rezende). The ability of diving mammals to obtain enough oxygen to support extended dives and foraging is largely dependent on muscle myoglobin (Mb) content. Mirceta et al. (p. 1303) found that in mammalian lineages with an aquatic or semiaquatic lifestyle, Mb net charge increases, which may represent an adaptation to inhibit self-association of Mb at high intracellular concentrations. Epistasis results from nonadditive genetic interactions and can affect phenotypic evolution. Natarajan et al. (p. 1324) found that epistatic interactions were able to explain the increased hemoglobin oxygen-binding affinity observed in deer mice populations at high altitude. In mammals, the offloading of oxygen from hemoglobin is facilitated by a reduction in the blood's pH, driven by metabolically produced CO2. However, in fish, a reduction in blood pH reduces oxygen carrying capacity of hemoglobin. Rummer et al. (p. 1327) implanted fiber optic oxygen sensors within the muscles of rainbow trout and found that elevated CO2 levels in the water led to acidosis and elevated oxygen tensions.


Epistatic interactions between mutant sites in the same protein can exert a strong influence on pathways of molecular evolution. We performed protein engineering experiments that revealed pervasive epistasis among segregating amino acid variants that contribute to adaptive functional variation in deer mouse hemoglobin (Hb). Amino acid mutations increased or decreased Hb-O2 affinity depending on the allelic state of other sites. Structural analysis revealed that epistasis for Hb-O2 affinity and allosteric regulatory control is attributable to indirect interactions between structurally remote sites. The prevalence of sign epistasis for fitness-related biochemical phenotypes has important implications for the evolutionary dynamics of protein polymorphism in natural populations.

Nonadditive interactions between mutations (epistasis) can exert a strong influence on the rate and direction of evolutionary change (1, 2). Insights into mechanisms of epistasis between beneficial mutations can reveal the causes of constraints on adaptive protein evolution (310). Mechanisms of epistasis are often best revealed through detailed examinations of interactions between amino acid mutations in the same protein that contribute to variation in a measurable biochemical phenotype (7, 915). Such studies are especially relevant to our understanding of evolutionary process when genetically based changes in the measured phenotype contribute to variation in fitness under natural conditions.

We investigated the nature of epistatic interactions between adaptive mutations in the hemoglobin (Hb) of deer mice (Peromyscus maniculatus). Deer mice that are native to high altitude have evolved an elevated Hb-O2 affinity relative to lowland conspecifics (1618), and this modification of protein function contributes to an adaptive enhancement of whole-animal physiological performance under hypoxia (19, 20). Comparisons between highland deer mice from the Rocky Mountains and lowland deer mice from the Great Plains revealed genetic differences in Hb-O2 affinity that are attributable to the independent or joint effects of 12 amino acid polymorphisms: 8 mutations in the α-chain subunits of the α2β2 Hb tetramer, and 4 mutations in the β-chain subunits. These 12 amino acid polymorphisms exhibit pronounced altitudinal shifts in allele frequency, and population genetic analyses of nucleotide variation in the α- and β-globin genes revealed evidence for divergent selection between deer mouse populations that are native to different elevations (18, 2123).

Structural variation in deer mouse Hb has a modular organization that reflects the linkage arrangement of the 12 amino acid polymorphisms. Within the α-chain subunit, five amino acid replacements are located in exon 2 of the underlying gene, and the remaining three replacements are located in exon 3. Polymorphic sites within the same exon are in nearly complete linkage disequilibrium (LD) with one another, but intragenic recombination has produced a partial uncoupling between the two exons (21, 23). The two most common α-globin allele classes are distinguished from each other by eight amino acid replacements at sites 50, 57, 60, 64, 71, 113, 115, and 116 (fig. S1A). The four amino acid polymorphisms in the β-globin gene are also in nearly complete LD with one another (18, 22). The two most common β-globin allele classes are distinguished from each other by four amino acid replacements at sites 62, 72, 128, and 135 (fig. S1B). Thus, in deer mouse populations, most of the naturally occurring variation in Hb structure is captured by combinatorial permutations of allelic variants at three loci: α-globin exon 2, α-globin exon 3, and β-globin.

We used site-directed mutagenesis to engineer all eight combinations of the α- and β-chain variants in recombinant Hb (rHb), and we measured O2-binding properties of the purified proteins (24). In addition to the chimeric multipoint mutants, we also engineered 10 additional single- and double-mutant rHbs to measure the functional effects of specific point mutations individually and in pairwise combination. We synthesized rHbs representing the two most common variants from high- and low-altitude populations, designated “HH-H” and “LL-L,” respectively (the first two letters denote the separate α-chain subdomains encoded by exons 2 and 3, and the third letter denotes the β-chain subunit). To test for epistasis, we also synthesized rHbs representing the remaining six combinations of H- and L-type alleles at each of the three loci (Fig. 1A). To examine variation in the allosteric regulation of Hb-O2 affinity, we measured O2-binding properties of each rHb mutant in the presence and absence of the two principal allosteric effectors present in mammalian red blood cells: Cl ions and 2,3-diphosphoglycerate (DPG). These effectors reduce Hb-O2 affinity by preferentially binding and stabilizing deoxyHb, thereby shifting the allosteric equilibrium in favor of the low-affinity T-state quaternary structure. By using standardized concentrations of Cl and DPG in the physiological range, we ensured that in vitro measurements were relevant to in vivo conditions (24).

Fig. 1 Structural and functional variation among recombinant deer mouse Hbs (rHbs).

(A) rHbs representing all combinatorial permutations of allelic variants at α-globin exon 2, α-globin exon 3, and β-globin. Shaded regions represent the products of low-altitude L-type alleles (red), and unshaded regions represent products of high-altitude H-type alleles (blue). (B) Variation in the allosteric regulation of Hb-O2 affinity by DPG. Sensitivity to DPG is indexed by the difference in log-transformed P50 values between stripped Hb in the presence and absence of DPG.

Our experiments revealed substantial variation in intrinsic Hb-O2 affinity, as P50 values (the O2 tension at 50% heme saturation) for stripped, cofactor-free rHbs ranged from 4.55 to 7.09 torr (Table 1). The high-altitude HH-H variant exhibited a 26% lower stripped P50 (i.e., higher intrinsic O2 affinity) relative to the low-altitude LL-L variant (Table 1). Hb-O2 affinity was reduced in the presence of Cl ions (added as 0.1 M KCl), in the presence of DPG at a twofold molar excess over tetrameric Hb, and in the simultaneous presence of both effectors (Table 1). All rHbs exhibited cooperative O2 binding, and Hill coefficients ranged from 1.36 to 2.28 in the presence of Cl and DPG.

Table 1 O2 affinities (P50, torr; mean ± SEM) and allosteric properties of purified rHbs.

Sensitivities to allosteric effectors are measured as the difference in log-transformed P50 values in the presence and absence of each effector, individually and in combination. VA and VE are estimated components of additive and epistatic variance, respectively (24).

View this table:

Contrary to the expectations of an additive null model, the phenotypic effects of allelic substitutions (L→H and H→L) at α-globin exon 2, α-globin exon 3, and β-globin were highly dependent on genetic background, as pairwise epistasis accounted for 40% of the variance in P50 values in the absence of allosteric effectors and 90% in the simultaneous presence of Cl and DPG (Table 1). In the presence of both allosteric effectors, the HH α-globin allele conferred an increased affinity on the βL background and a decreased affinity on the βH background. Similarly, the H-type β-globin allele conferred an increased affinity on the αLL background and a decreased affinity on the αHH background. These are examples of sign epistasis (2), where the sign of the phenotypic effect of an allele is conditional on genetic background.

Because mammalian Hb is a heterotetramer (α2β2), epistatic interactions could involve closely linked sites in the same gene or sites in unlinked genes that encode different subunits of the protein. Intragenic (within-subunit) epistasis could stem from localized modifications of secondary or tertiary structure, whereas intergenic (between-subunit) epistasis could stem from allosteric transitions in quaternary structure between different oxygenation states of the Hb tetramer. Epistasis for Hb-O2 affinity is mainly attributable to the suppressed DPG sensitivity of chimeric rHb variants that incorporate the products of α- and β-globin alleles of unlike type (αHH combined with βL, and vice versa; Table 1 and Fig. 1B). Although DPG sensitivity was suppressed in HH-L and LL-H, allosteric regulatory capacities of the chimeric rHbs were partially restored by reciprocally converting either of the two α-chain subdomains to the type that matched the associated β-chain subunit: DPG sensitivity of the chimeric LL-H was partially restored by L→H substitutions at α-globin exon 2 or exon 3, and reciprocally, DPG sensitivity of the chimeric HH-L was partially restored by H→L substitutions at these same loci (Fig. 1B). In principle, a suppressed DPG sensitivity (and hence, increased Hb-O2 affinity) could be produced by charge-changing amino acid replacements that eliminate phosphate-binding sites in the β-chain subunits. Because the positively charged phosphate-binding sites are invariant in deer mouse β chains (17, 18), allelic variation in DPG sensitivity must stem from indirect, second-order perturbations.

Analysis of the crystal structure of deer mouse Hb at 1.8 Å resolution (24, 25) revealed that each of the eight rHb mutants is characterized by a unique constellation of hydrogen bonds within and between subunits (Table 2 and fig. S2). Additional hydrogen bonds between subunits of the same αβ dimer are formed in the presence of β128Ser (an L-type residue; fig. S2), which contributes to the observed epistasis between allelic α- and β-chain variants. Structural analysis also revealed that in Hbs with L-type α-globin, the imidazole ring of α50His forms a hydrogen bond with α30Glu in the same subunit. The replacement of α50His with Pro (the H-type residue) eliminates this hydrogen bond and causes a subtle reorientation of the E helix and CD loop (Fig. 2), an effect that propagates to the α1β2 intersubunit contact and shifts the allosteric equilibrium in favor of the high-affinity oxyHb (R-state) quaternary structure.

Table 2 Allelic variation in the network of atomic contacts within and between subunits of deer mouse Hb.

Plus signs denote the presence of hydrogen bonds within subunits (α50His-α30Glu and α113His-α24Tyr) or between subunits of unlike type (α34Cys-β128Ser). Polymorphic sites are shown in bold.

View this table:
Fig. 2 Difference in the network of hydrogen bonds between high- and low-altitude Hb variants, HH-H and LL-L, respectively.

The α1 and β1 subunits of HH-H (light blue) and LL-L (light red) are superimposed, and van der Waals radii are shown for α-chain residues that are in atomic contact with β-chain residues of the opposing α2β2 dimer.

To test the effects of charge-changing α-chain mutations in the CD loop (α50His/Pro) and the adjacent E helix (α64Asp/Gly), we synthesized each of the alternative single and double mutants on both HH-H and LL-L backgrounds. The experiments revealed that, on the LL-L background, substitutions of H-type residues (α50His→Pro and α64Asp→Gly) did not produce a significant increase in Hb-O2 affinity individually or in combination; however, on the HH-H background, single-step reversions to L-type residues at both sites produced significant reductions in Hb-O2 affinity in the presence of allosteric effectors (fig. S3). We also measured the individual effects of all four amino acid mutations in the β-chain subunit. On the LL-L background, the substitution β128Ser→Ala, which removes an α1β1 hydrogen bond (fig. S2), produced an increased anion sensitivity (and hence, a decreased Hb-O2 affinity in the presence of Cl and DPG; fig. S4). However, on this same background, introducing all four H-type β-chain mutations in combination produced a highly significant increase in Hb-O2 affinity in the presence of allosteric effectors (fig. S4).

In summary, results of our mutagenesis experiments revealed pervasive epistasis among segregating amino acid variants in deer mouse Hb (Table 1). The individual and joint effects of α- and β-chain point mutations contribute to the elevated Hb-O2 affinity of highland deer mice, but the effects of these mutations are highly dependent on the allelic state of other residue positions.

Directed mutagenesis studies have unveiled “cryptic” epistasis between amino acid substitutions that distinguish deeply diverged orthologous proteins (7, 11, 13, 14). Similarly, experimental studies of microbial systems have revealed intragenic epistasis between sites that underwent successive allelic substitutions but that were never simultaneously polymorphic (6, 10). By contrast, the interacting mutations in deer mouse Hb are segregating in natural populations and, given the extensive intragenic and intergenic LD, the epistasis contributes to additive genetic variance in Hb function, providing an explanation for the previously documented variation in anion sensitivity of deer mouse Hbs (17, 18). Given the evidence for spatially varying selection on Hb polymorphism in relation to altitude, the pervasiveness of sign epistasis for Hb-O2 affinity suggests that the selection coefficient for a given allele will often be highly dependent on the allelic composition of the local population. Thus, sign epistasis among segregating amino acid variants may exert a strong influence on allele frequency dynamics and mutational pathways of protein evolution.

Supplementary Materials

Materials and Methods

Figs. S1 to S5

Table S1

References (2641)

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Acknowledgments: Funded by grants from the NIH–National Heart, Lung, and Blood Institute (R01 HL087216 and HL087216-S1), the NSF (DEB-0614342 and IOS-0949931), and the Faculty of Science and Technology, Aarhus University. We thank S. Kachman for statistical advice, A. Bang for assistance in the lab, and M. Harms, S. Smith, and two reviewers for helpful comments. All experimental data are tabulated in the main text and in the supplementary materials.
View Abstract

Stay Connected to Science

Editor's Blog

Navigate This Article