|Home | About | Journals | Submit | Contact Us | Français|
If the fitness effects of amino acid mutations are conditional on genetic background, then mutations can have different effects depending on the sequential order in which they occur during evolutionary transitions in protein function. A key question concerns the fraction of possible mutational pathways connecting alternative functional states that involve transient reductions in fitness. Here we examine the functional effects of multiple amino acid substitutions that contributed to an evolutionary transition in the oxygenation properties of avian hemoglobin (Hb). The set of causative changes included mutations at intradimer interfaces of the Hb tetramer. Replacements at such sites may be especially likely to have epistatic effects on Hb function since residues at intersubunit interfaces are enmeshed in networks of salt bridges and hydrogen bonds between like and unlike subunits; mutational reconfigurations of these atomic contacts can affect allosteric transitions in quaternary structure and the propensity for tetramer–dimer dissociation. We used ancestral protein resurrection in conjunction with a combinatorial protein engineering approach to synthesize genotypes representing the complete set of mutational intermediates in all possible forward pathways that connect functionally distinct ancestral and descendent genotypes. The experiments revealed that 1/2 of all possible forward pathways included mutational intermediates with aberrant functional properties because particular combinations of mutations promoted tetramer–dimer dissociation. The subset of mutational pathways with unstable intermediates may be selectively inaccessible, representing evolutionary roads not taken. The experimental results also demonstrate how epistasis for particular functional properties of proteins may be mediated indirectly by mutational effects on quaternary structural stability.
When an evolutionary transition in protein function involves multiple mutational steps, a number of important questions can be addressed by experimentally examining the full set of possible intermediate genotypes that connect the ancestral starting point and the evolved endpoint (Weinreich 2010; Weinreich, et al. 2013; de Visser and Krug 2014; Hartl 2014). For example, Are particular endpoints selectively accessible from many different ancestral starting points? What fraction of possible pathways connecting ancestral and descendant genotypes involve transient reductions in fitness? What fraction of possible pathways could never be realized—representing evolutionary “roads not taken”—because they include nonfunctional/inviable intermediate steps? (Maynard Smith 1970; DePristo et al. 2005; Poelwijk et al. 2007; Kondrashov and Kondrashov 2015). These are fundamentally questions about the form and prevalence of intramolecular epistasis (nonadditive interactions between mutant sites in the same protein) and its role in shaping evolutionary trajectories through protein sequence space (Weinreich et al. 2005; Wagner 2008; Bloom and Arnold 2009; Carneiro and Hartl 2010; Harms and Thornton 2013; Kondrashov and Kondrashov 2015; Miton and Tokuriki 2016; Starr and Thornton 2016; Storz 2016a).
If n mutations are associated with an evolved change in protein function, and if a binary combination of ancestral/derived amino acids are possible at each site, then the key question is whether all n mutations have either neutral or beneficial effects in each of the 2n possible multi-allelic combinations. If the sign of a mutation’s fitness effect depends on the allelic state at one or more other sites [“sign epistasis” (Weinreich et al. 2005)], then a subset of the n! mutational pathways connecting the ancestral and descendant genotypes will include intermediate steps with reduced fitness even if the fitness of the descendant genotype is greater than or equal to that of the ancestor (Malcolm et al. 1990; DePristo et al. 2005; Lunzer et al. 2005; Weinreich et al. 2005; Bridgham et al. 2006; Weinreich et al. 2006; Poelwijk et al. 2007; Bridgham et al. 2009; Lozovsky et al. 2009; Brown et al. 2010; da Silva et al. 2010; Weinreich 2010; Carroll et al. 2011; Kvitek and Sherlock 2011; Salverda et al. 2011; Dickinson et al. 2013; Gong et al. 2013; Schenk et al. 2013; de Visser and Krug 2014; Harms and Thornton 2014; Kondrashov and Kondrashov 2015; Palmer et al. 2015; Tufts et al. 2015; Bank et al. 2016; Wu et al. 2016).
There may be considerable scope for intramolecular epistasis during evolutionary shifts in the function of allosterically regulated, multimeric proteins like hemoglobin (Hb) because many residues are enmeshed in networks of site–site interactions that undergo discrete reconfigurations during ligation-dependent transitions in quaternary structure. The Hb tetramer is composed of two identical, semi-rigid α1β1 and α2β2 dimers that undergo a relative rotation of 15° during the oxygenation-linked transition between the deoxy (low affinity, T) conformation and the oxy (high affinity, R) conformation (Baldwin and Chothia 1979; Perutz 1989). Mutual rotation of the α1β1 and α2β2 dimers involves no appreciable change in the intradimer contact surfaces but substantial changes in intersubunit interactions at the α1β2 and α2β1 contacts (Pettigrew et al. 1982). The conformational equilibrium between the T- and R-states is central to the allosteric function of Hb as an O2-transport molecule, as it governs the cooperativity of O2-binding and the regulation of Hb-O2 affinity by allosteric effectors (nonheme ligands such as Cl− ions and organic phosphates). Allosteric effectors reduce Hb-O2 affinity by preferentially binding and stabilizing Hb in the deoxy T-state, thereby shifting the allosteric equilibrium in favor of this low-affinity conformation.
Here we report an experimental analysis of multiple amino acid substitutions that occurred during an evolutionary transition in Hb function in South American nightjars (nocturnal birds in the family Caprimulgidae). We focus on one particular species, Tschudi’s nightjar (Hydropsalis decussata), a strictly lowland bird that has evolved a reduced Hb-O2 affinity relative to its highland sister species, the band-winged nightjar (Hydropsalis longirostris), which occurs at elevations ranging from ~2000 m to 4400 m above sea level in the Andes (Schulenberg et al. 2007; Benham et al. 2011). We chose to dissect the molecular basis of this particular transition in protein function because ancestral sequence reconstruction indicated that the evolved reduction in Hb-O2 affinity in the H. decussata lineage involved the independent or joint effects of four amino acid substitutions, two of which involve intradimeric (α1β1 and α2β2) interfaces. Amino acid replacements at intradimeric contacts can affect the allosteric transition in quaternary structure and the propensity for tetramer–dimer dissociation (Fronticelli et al. 1994; Tsuruga et al. 1998; Vasquez et al. 1999; Chang et al. 2002; Yasuda et al. 2002; Shikama and Matsuoka 2003; Bellelli et al. 2006; Natarajan et al. 2013). Such replacements may be especially likely to have epistatic effects on Hb function since residues at intersubunit interfaces are involved in networks of salt bridges and hydrogen bonds with multiple residues in like and unlike subunits.
By reconstructing and resurrecting the Hb of the H. decussata/H. longirostris ancestor, we deduced that H. decussata evolved a derived reduction in Hb-O2 affinity. We then used a combinatorial protein engineering approach to synthesize genotypes representing all possible mutational intermediates in each of 4!=24 forward pathways connecting the high-affinity ancestor to the low-affinity quadruple-mutant genotype of H. decussata. For each possible pathway, we experimentally characterized the trajectory of change in Hb-O2 affinity and a number of other properties (rates of autoxidation and measures of structural stability) that potentially trade-off with Hb-O2 affinity. The goal was to assess how mutational pleiotropy and epistasis influence evolutionary trajectories through protein sequence space. Specifically, we assessed the fraction of mutational pathways connecting functionally distinct ancestral and descendant genotypes that include potentially deleterious intermediate steps.
We start by presenting data on the functional and structural properties of native Hbs from the lowland H. decussata and the highland H. longirostris. We then explain how these results motivated additional protein-engineering experiments using site-directed mutagenesis.
Experiments based on a combination of isoelectric focusing (IEF) and tandem mass spectroscopy (MS/MS) revealed that H. decussata and H. longirostris both express two structurally distinct Hb isoforms in adult red blood cells (supplementary fig. S1, Supplementary Material online), consistent with data for most birds (Grispo et al. 2012; Opazo et al. 2015). The major and minor Hb isoforms (HbA and HbD, respectively) share identical β-type subunits, but the α-type subunits of HbA are encoded by the αA-globin gene and those of HbD are encoded by the closely linked αD-globin gene (Hoffmann and Storz 2007; Grispo et al. 2012; Opazo et al. 2015). There was no significant difference in the relative abundance of the two isoforms in the red blood cells of the two nightjar species; the percentage of total adult Hb accounted for by the minor HbD isoform (mean±1 SD) was 18.4%±4.1 for H. decussata (n=5 individual specimens) and 21.0%±5.3 for H. longirostris (n=6).
After isolating and purifying HbA and HbD from multiple specimens of each species, we measured oxygenation properties of the purified Hb solutions under standardized experimental conditions (Imai 1982; Mairbäurl and Weber 2012). Genetically based differences in Hb-O2 affinity can stem from changes in intrinsic O2-affinity and/or changes in the responsiveness to allosteric effectors like Cl− ions and organic phosphates. To reveal the functional mechanisms that are responsible for observed differences in Hb-O2 affinity, we measured O2 equilibria of purified Hbs under four standardized treatments: (i) in the absence of allosteric effectors (stripped), (ii) in the presence of Cl− ions, (iii) in the presence of inositol hexaphosphate (IHP) (a chemical analog of the endogenously produced inositol pentaphosphate) and (iv) in the simultaneous presence of Cl− and IHP. This latter treatment is most relevant to in vivo conditions in avian red blood cells. In each treatment we estimated P50 (the O2 partial pressure [PO2] at which Hb is half-saturated), and the Hill coefficient, n50, an index of subunit cooperativity in the Hb tetramer.
The experiments revealed that HbD has a uniformly higher O2 affinity than HbA in both species (fig. 1 and table 1), a pattern of isoform differentiation that is consistent with data from other bird species (Grispo et al. 2012; Cheviron et al. 2014; Galen et al. 2015; Natarajan et al. 2015b; Opazo et al. 2015; Natarajan et al. 2016). Comparisons between the two nightjar species revealed no appreciable differences in the intrinsic O2 affinities of either HbA or HbD (as indicated by P50’s for the “stripped” Hbs); however, in the presence of IHP, O2-affinities of both isoforms were significantly higher (i.e., P50’s were lower) in the highland H. longirostris than in the lowland H. decussata (fig. 1 and table 1).
By cloning and sequencing the full complement of adult-expressed globin genes in all nightjar specimens, we deduced that the observed difference in HbA O2 affinity between H. longirostris and H. decussata is attributable to the independent or combined effects of four amino acid substitutions (three in the αA-globin gene and one in βA-globin; fig. 2). Likewise, the observed difference in HbD O2 affinity between the two species is attributable to the independent or combined effects of eight substitutions (seven in αD-globin and one in βA-globin; fig. 2). Since HbA and HbD share the same β-chain subunit, the same β112Ile→Val (Iβ112V) substitution at site 112 distinguishes both isoforms of the two species. Given that the major HbA isoform accounts for ~80% of total adult Hb in both species, we mainly focus on the functional evolution of HbA from this point forward.
To infer the polarity of character state changes at each of the sites that differ between the Hb isoforms of H. decussata and H. longirostris, we cloned and sequenced orthologs of αA-, αD-, and βA-globin from 11 other New World nightjar species, including four additional species in the genus Hydropsalis (fig. 2). Phylogenetic reconstructions of the globin sequences from the 13 nightjar species revealed a high level of genealogical discordance between the αA/αD and βA-globin gene trees (supplementary fig. S2, Supplementary Material online). This pattern of genealogical discordance has been documented for other nuclear and mitochondrial genes in previous studies of nightjar systematics (Larsen et al. 2007; Han et al. 2010; Sigurdsson and Cracraft 2014) and is likely attributable to incomplete lineage sorting and/or introgressive hybridization.
In spite of the discordant tree topologies for the unlinked α- and β-type globin genes, the polarities of the four HbA amino acid changes were unambiguous (fig. 2 and supplementary fig. S2, Supplementary Material online). At each of the four sites that distinguish the HbA isoforms of H. decussata and H. longirostris, parsimony clearly suggests that H. decussata possesses the derived amino acid. Thus, even though H. decussata and H. longirostris conform to the well-documented adaptive trend where high-altitude bird species tend to have higher Hb-O2 affinities than their lowland relatives (Storz 2016b), this particular pair of nightjar species is unusual in that the low-altitude species evolved a derived reduction in Hb-O2 affinity and the high-altitude species retained the ancestral, high-affinity character state. In other studies of birds in which altitude-related changes in Hb function have been documented, the high-altitude species have evolved derived increases in Hb-O2 affinity (Projecto-Garcia et al. 2013; Galen et al. 2015; Natarajan et al. 2015b, 2016). There is no independent evidence to suggest that H. decussata descended from highland ancestors or that the evolved reduction in Hb-O2 affinity in this species constitutes an adaptation to low-altitude.
Using the inferred character polarity for each of the four amino acid changes described above, we reconstructed and experimentally resurrected the HbA isoform of the H. decussata/H. longirostris ancestor. We then used a combinatorial protein engineering approach to synthesize genotypes representing all possible mutational intermediates in each of 24 forward pathways connecting the ancestral genotype “GVAI” to the quadruple-mutant genotype of H. decussata, “SIVV”. This required the synthesis, expression, and purification of 24=16 recombinant Hb (rHb) mutants. We characterized the oxygenation properties for each of these 16 rHbs and we also measured several additional biochemical and biophysical properties that potentially trade-off with Hb-O2 affinity (autoxidation rates and indices of structural stability). The goal was to test whether changes in Hb-O2 affinity were consistently associated with changes in other properties and to characterize pleiotropic effects of affinity-altering mutations.
Consistent with the analysis of the native HbA isoforms, the quadruple-mutant SIVV genotype (identical to H. decussata HbA) exhibited a significantly lower O2-affinity than the ancestral GVAI genotype (identical to H. longirostris HbA) in the presence of allosteric effectors (P50(KCl+IHP) was 1.6-fold higher; supplementary table S1, Supplementary Material online). Data for the full set of 16 rHbs revealed very little variation in intrinsic O2 affinity (indexed by P50(stripped)) but substantial variation in O2 affinity in the presence of allosteric effectors (indexed by P50(KCl+IHP); fig. 3A); this variation was largely attributable to unusually low IHP sensitivities in a subset of double and triple mutants. In particular, two double-mutants (SIAI and SVAV) and one triple-mutant (SIVI) exhibited aberrant O2-binding properties, with P50(KCl+IHP) values that were drastically lower than that of the ancestral GVAI genotype (supplementary table S1, Supplementary Material online). The P50(KCl+IHP) values for these three genotypes are far below the range of naturally occurring values for avian HbA isoforms measured under identical conditions (Grispo et al. 2012; Natarajan et al. 2015b, 2016). The combination of an aberrantly high Hb-O2 affinity in the presence of IHP and low cooperativity (i.e., values of n50 close to 1.0) indicate that tissue O2 delivery would be impaired because Hb would remain highly saturated even at the low PO2 prevailing in systemic capillary beds. An increase in Hb-O2 affinity reduces the PO2 at which O2 is released, resulting in a diminished gradient for O2 diffusion from capillary blood to the tissue mitochondria. This is why human Hb mutants with unusually high O2-affinity are typically pathological (Wajcman and Galactéros 1996; Steinberg and Nagel 2009).
Of the 24 possible forward pathways that connect the ancestral, high-affinity GVAI genotype and the derived, low-affinity SIVV genotype, a total of 12 pathways include the aberrant SIAI, SVAV, and/or SIVI genotypes as intermediate steps. Using the heuristic metaphor introduced by Malcolm et al. (1990), these pathways fall outside the so-called “neutral corridor”. For a given functional or structural property of a protein, the neutral corridor is defined by the range of values that fall within the interval between the lower (or upper) 95% confidence limit of the ancestral, start-point genotype and the upper (or lower) 95% confidence limit of the descendant, endpoint genotype (Malcolm et al. 1990). As shown in figure 3B, values of P50(KCl+IHP) for the SIAI, SVAV, and SIVI genotypes fell well below the lower 95% confidence limit of the ancestral, high-affinity GVAI genotype, and an additional double-mutant genotype, GIAV, exhibited a marginally significant reduction in P50(KCl+IHP) below this same threshold. These four genotypes exhibited lower IHP sensitivities than all the other 12 genotypes (supplementary table S1, Supplementary Material online), indicating reduced capacities for allosteric regulatory control. In the remaining forward pathways, each successive mutational step either results in a significant reduction in Hb-O2 affinity or a more-or-less unchanged Hb-O2 affinity (where observed differences were within the range of measurement error). The existence of these two distinct pathway types—those contained within the neutral corridor and those that include putatively deleterious intermediate steps—is attributable to the fact that the sign of some mutations’ phenotypic effects depend on which other substitutions have already occurred. All 24 pathways have the same ancestral starting point (GVAI), they all have the same endpoint (SIVV), and they all involve the same combination of four mutations; the different pathways are only distinguished by the sequential order in which the substitutions occurred. If the aberrant functional properties of the above-mentioned genotypes confer a reduced fitness, then sign-epistasis for Hb-O2 affinity would have the effect of restricting the selective accessibility of pathways leading from GVAI to SIVV, as depicted in a four-dimensional genotype space in fig. 4. These inferences about pathway accessibility are premised on two assumptions about evolutionary dynamics. First, the depictions in figures 3B and 4 assume that mutations are fixed sequentially on invariant genetic backgrounds, so mutational trajectories can be envisioned as a succession of discrete transitions between adjacent points in sequence space (Weinreich 2010). The second assumption is that once a mutation is fixed, it does not later revert to the ancestral state. The simultaneous fixation of two or more mutations and mutational reversions can potentially produce new connecting pathways in addition to the direct, forward pathways shown in figures 3B and and44 (Kimura 1985; Weinreich and Chao 2005; Weinreich et al. 2005; DePristo et al. 2007; Meer et al. 2010; Weinreich 2010; Franke et al. 2011; Covert et al. 2013; Palmer et al. 2015; McCandlish et al. 2016; Wu et al. 2016).
What is the biophysical basis of the observed epistasis for Hb-O2 affinity in the presence of allosteric effectors? IHP binds with 1:1 stoichiometry between the β-chain subunits of deoxyHb via charge-charge interactions with a set of highly conserved cationic residues (Arnone and Perutz 1974). Amino acid replacements that directly or indirectly inhibit phosphate binding will typically increase Hb-O2 affinity by shifting the allosteric equilibrium in favor of R-state oxyHb (Storz et al. 2009, 2010; Revsbech et al. 2013; Natarajan et al. 2013; Janecka et al. 2015; Natarajan et al. 2015a). However, none of the observed substitutions in H. decussata HbA directly affect IHP-binding sites in the central cavity. Instead, the combination of low cooperativity and reduced responsiveness to IHP (supplementary table S1, Supplementary Material online) suggest the hypothesis that the high-affinity rHb mutants are partly dissociating into α1β1 and α2β2 dimers. This is because oxygenation-linked transitions in quaternary structure provide the basis for both cooperative O2-binding and allosteric regulation via IHP-binding. If these allosteric properties are compromised, it suggests that the Hb samples may contain an equilibrium mixture of functionally intact Hb tetramers (whose O2-affinities are significantly reduced in the presence of IHP) and dissociated α1β1 and α2β2 dimers (whose O2-affinities are unaffected by the presence of IHP). To test this hypothesis, we performed a gel-filtration experiment to assess whether the rHb mutants with the three lowest P50(KCl+IHP) values (SIAI, SVAV, and SIVI) exhibited higher rates of tetramer–dimer dissociation relative to the quadruple-mutant, SIVV (which had the highest P50(KCl+IHP) value). Results of this experiment clearly demonstrated that the rHb mutants with low IHP sensitivity had elution volumes intermediate between values for tetrameric human Hb and monomeric myoglobin—which is indicative of tetramer–dimer dissociation—whereas SIVV had an elution volume similar to that of human Hb, the control for an intact, tetrameric assembly (fig. 5). These results indicate that the exceedingly low IHP sensitivities of SIAI, SVAV, and SIVI (and the sign epistasis for P50(KCl+IHP)) are not directly attributable to mutational effects on IHP binding. Instead, the unusually high O2-affinities in the presence of IHP are attributable to the indirect effects of tetramer–dimer dissociation. The Hb concentrations in our in vitro O2 equilibrium experiments are considerably lower than in vivo concentrations in circulating red blood cells, but if the same variation in quaternary structural stability is manifest in vivo, then our findings suggest that the subset of mutational pathways that include the unstable genotypes as intermediate steps may represent evolutionary dead ends. Even if the rate of tetramer–dimer dissociation is lower under in vivo conditions, our results demonstrate how—in principle—epistasis for particular functional properties of proteins may be mediated indirectly by mutational effects on stability of the tetrameric assembly.
To decipher the causative effects of the four amino acid mutations, it is clear that we need to distinguish between their direct effects on Hb-O2 affinity and their indirect effects on Hb-O2 affinity that are mediated by effects on tetramer–dimer dissociation.
Results of homology-based structural modeling suggest specific biophysical mechanisms by which each of the four amino acid substitutions contribute to the reduced O2 affinity of H. decussata HbA. The two substitutions in the α-chain A-helix (Gα4S and Vα13I) are predicted to reduce Hb-O2 affinity by eliminating specific atomic contacts in oxyHb, thereby reducing the relative stability of the R-state (supplementary fig. S3, Supplementary Material online). The other two amino acid substitutions, Aα34V and Iβ112V, affect intradimeric α1β1 and α2β2 contacts. The Aα34V substitution results in the gain of two additional intradimer atomic contacts per Hb tetramer in both the R- and T-states, as α34V forms a van der Waals contact with β128A in each of the constituent αβ dimers (fig. 6A and B). This stabilization of intradimeric contacts is predicted to restrict allosteric motion in the T→R transition in quaternary structure, thereby reducing O2-affinity by shifting the allosteric equilibrium in favor of the deoxy T-state. The Iβ112V substitution reduces Hb-O2 affinity by simultaneously destabilizing the R-state and stabilizing the T-state. The ancestral β112I forms an intradimeric van der Waals contact with α107V in oxy (R-state) Hb (fig. 6C), and it forms an intradimeric contact with α110A in deoxy (T-state) Hb. The Iβ112V substitution eliminates the intradimer contact with α107V in oxyHb (fig. 6D), which contributes to the destabilization of the R-state, and it produces an additional intradimeric contact between β112V and α106L in deoxyHb, which contributes to the stabilization of the T-state.
In addition to contributing to the reduced Hb-O2 affinity in H. decussata, the Iβ112V substitution also appears to affect the propensity for tetramer–dimer dissociation, although its effects are highly context dependent. On some backgrounds, the Iβ112V substitution reduced P50(KCl+IHP) values below the floor of the neutral corridor (e.g., SVAI→SVAV and GIAI→GIAV); in other cases the same substitution produced equally dramatic, compensatory increases in P50(KCl+IHP) (e.g., SIAI→SIAV and SIVI→SIVV). For example, the SIAI and SIVI genotypes exhibited the lowest P50(KCl+IHP) values and the lowest sensitivities to IHP (supplementary table S1, Supplementary Material online); on these two backgrounds, the Iβ112V mutation increased P50(KCl+IHP) 2.1-fold and 2.6-fold, respectively, and it increased IHP sensitivity 1.6-fold and 2.6-fold, respectively (supplementary table S1, Supplementary Material online). This dramatic rescue effect is likely not attributable to direct effects on IHP-binding; rather, the Iβ112V substitution increased responsiveness to IHP by inhibiting tetramer–dimer dissociation. In other words, restoration of allosteric regulatory capacity followed automatically from a restoration of the intact tetrameric assembly. These documented effects are consistent with other functional studies of β112 mutations in human Hb. For example, the Cβ112G mutation produces a 4-fold reduction in the tendency of human Hb to dissociate into dimers (Fronticelli et al. 1994). By contrast, Cβ112R and Cβ112F increase this tendency and are associated with hemolytic anemia (Adams et al. 1979; Brennan et al. 2002). Since tetramer–dimer dissociation involves the interdimeric α1β2/α2β1 interfaces, the data for human Hb mutants and the engineered nightjar Hb mutants indicate that perturbations of the intradimeric contacts are transmitted to contacts between dimers (Fronticelli et al. 1994; Vasquez et al. 1999; Yasuda et al. 2002; Shikama and Matsuoka 2003; Bellelli et al. 2006).
None of the mutations—singly or in combination—produced any detectable perturbations of overall secondary or tertiary structure, as measured by means of circular dichroism and UV–visible spectroscopy, respectively (supplementary fig. S4A and B, Supplementary Material online). Thus, mutational effects on the propensity for tetramer–dimer dissociation are apparently localized to intra- and interdimeric contact surfaces. The Gα4S mutation produced consistent reductions in autoxidation rate on all eight backgrounds (supplementary fig. S4C, Supplementary Material online) and was associated with reductions in Hb-O2 affinity. None of the other mutations produced a detectable change in heme autoxidation rate on any genetic background.
Results of the protein engineering experiments revealed that 1/2 of all possible forward pathways connecting functionally distinct genotypes included mutational intermediates with severely compromised functional capacities. Mutational pathways that pass through these functionally aberrant intermediates appear to be selectively inaccessible, and (to the extent that in vitro experiments approximate in vivo conditions) likely represent unrealized historical pathways of functional evolution. In addition to elucidating accessibility relationships between functionally distinct ancestral and descendent genotypes, results of this analysis also demonstrate how epistasis for particular functional properties of proteins may be mediated indirectly by mutational effects on structural stability. In the case of the nightjar Hbs, sign epistasis for Hb-O2 affinity is indirectly attributable to mutational effects on tetramer–dimer dissociation.
We collected nightjar specimens from numerous localities in the Peruvian Andes and adjacent lowlands. All specimens were preserved as vouchers in the ornithological collection of the Museum of Southwestern Biology of the University of New Mexico and the Centro de Ornitología y Biodiversidad (CORBIDI), Lima, Peru. Complete specimen data are available via the ARCTOS online database (supplementary table S2, Supplementary Material online). All birds were live-trapped in mistnets and were sacrificed in accordance with protocols approved by the University of New Mexico Institutional Care and Use Committee (Protocol number 08UNM033-TR-100117; Animal Welfare Assurance number A4023-01). All collections were authorized by permits issued by management authorities of Peru (004-2007-INRENA-IFFS-DCB, 135-2009-AG-DGFFS-DGEFFS, 0377-2010-AG-DGFFS-DGEFFS, 0199-2012-AG-DGFFS-DGEFFS, and 006-2013-MINAGRI-DGFFS/DGEFFS).
For each nightjar specimen, we collected 20–60μl of whole blood from the brachial or ulnar vein using heparinized microcapillary tubes. Red blood cells were separated from the plasma fraction by centrifugation, and the packed red cells were flash-frozen in liquid nitrogen. We collected pectoral muscle from each specimen as a source of both genomic DNA and globin mRNA. Muscle samples were flash-frozen or preserved using RNAlater. All tissue and blood samples were subsequently stored at −80°C. To increase taxon sampling in our survey of globin sequence variation, we obtained tissue samples for additional nightjar species from the University of Kansas Natural History Museum (Hydropsalis maculicauda, H. parvula, and H.anomala) and the Museum of Southwestern Biology (Hydropsalis segmentata, Chordeiles minor, Caprimulgus rufigena, Antrostomus carolinensis, Antrostomus arizonae, Nyctipolus nigrescens, Phalaenoptilus nuttalli, and Nyctiphyrnus ocellatus). Complete specimen data are provided in supplementary table S2, Supplementary Material online.
Genus and species names generally follow the South American Classification Committee (Remsen et al. 2016) and North American Classification Committee (Chesser et al. 2016) of the American Ornithologists' Union, with the exception that we adopt the broadly defined genus Hydropsalis as proposed by Sigurdsson and Cracraft (Sigurdsson and Cracraft 2014). Phylogeny estimates for the New World nightjars demonstrate that alternative classifications would include genera that are non-monophyletic [as would be the case for the generic names Systellura, Hydropsalis, and Setopagis proposed by Remsen et al. (2016)].
To characterize Hb isoform composition in the red blood cells of the two focal nightjar species, we first separated native Hbs by means of IEF. After performing trypsin digests on excised gel bands, we then performed an MS/MS analysis to identify the resultant peptides. Database searches of MS/MS spectra were performed using Mascot (Matrix Science, v1.9.0, London, UK). Specifically, peptide mass fingerprints derived from the MS/MS analysis were used to query a custom database of avian α- and β-type globin sequences. These amino acid sequences were derived from conceptual translations of the adult-expressed αA-, αD-, and βA-globin genes of Hydropsalis decussata and H. longirostris, in addition to the full complement of embryonic and adult α- and β-type globin genes that have been annotated in the genome assemblies of other birds (Hoffmann et al. 2010, 2011; Grispo et al. 2012; Opazo et al. 2015). We identified all significant protein hits that matched more than one peptide with P<0.05. After deciphering the subunit composition of each Hb isoform by means of MS/MS, we used native gel IEF to densitometrically quantify the relative abundance of HbA and HbD (Opazo et al. 2015).
For each of the 11 nightjar specimens used in the experimental analyses of Hb function, we extracted RNA from pectoral muscle tissue using the RNeasy kit (Qiagen,Valencia, CA) and we amplified full-length cDNAs of the αA-, αD-, and βA-globin genes using a OneStep RT-PCR kit (Qiagen, Valencia, CA). We designed paralog-specific primers using 5′ and 3′ UTR sequences from passerine species, as described previously (Opazo et al. 2015). We cloned reverse transcription (RT)-PCR products using the TOPO® TA Cloning® Kit (Life technologies, Carlsbad, CA), and we sequenced at least five clones per gene in order to recover both alleles. This enabled us to determine full diploid genotypes for each of the three adult-expressed globin genes in each specimen.
We increased taxon sampling for the phylogenetic analyses by sequencing the complete coding regions of the αA-, αD-, and βA-globin genes in 13 nightjar species including H. decussata and H. longirostris. In order to maximize the number of informative characters in our phylogenetic analysis of the αA-globin gene, we used genomic DNA templates for all 13 species to sequence the complete coding region in addition to the introns and noncoding flanking regions. We extracted DNA from pectoral muscle tissue using the DNeasy kit (Qiagen, Valencia, CA).
We estimated phylogenies of the α- and β-type globin genes using maximum likelihood, with support for internal nodes assessed with 500 bootstrap replicates. We estimated the trees using PhyML (Guindon et al. 2010), with the HKY85 substitution model and added parameters for rate heterogeneity among sites (modeled according to a gamma distribution with four rate categories) and invariant sites.
We purified HbA and HbD from hemolysates of wildcaught H. decussata and H. longirostris specimens by means of anion-exchange fast-protein liquid chromatography (FPLC) using an Äkta Pure system and a HiTrap QHP column (GE Healthcare). This procedure also removes endogenous organic phosphates, yielding “stripped” Hb samples. Using purified Hb solutions (0.3mM heme), we measured O2-equilibrium curves at 37°C in 0.1 M HEPES buffer (pH 7.4) in the absence (stripped) and presence of 0.1 M KCl and IHP (at 2-fold molar excess over tetrameric Hb), and in the simultaneous presence of KCl and IHP. We measured O2-equilibria of 3μl thin-film samples in a modified diffusion chamber where absorption at 436nm was monitored during stepwise changes in the equilibration of N2/O2 mixtures generated by precision Wösthoff gas-mixing pumps (Weber 1992). We estimated values of P50 and n50 (Hill’s cooperativity coefficient at half-saturation) by fitting the Hill equation Y=P/(P50n+P) to the experimental O2 saturation data by means of nonlinear regression (Y=fractional O2 saturation; n, cooperativity coefficient) (Grispo et al. 2012; Weber et al. 2013). The nonlinear fitting was based on 5–8 equilibration steps between 30% and 70% oxygenation. Free Cl− concentrations were measured with a model 926S Mark II chloride analyzer (Sherwood Scientific Ltd, Cambridge, UK).
The αA- and βA-globin sequences were synthesized by Eurofins MWG Operon (Huntsville, AL, USA) after optimizing the nucleotide sequences in accordance with E. coli codon preferences. The synthesized αA–βA globin gene cassette was cloned into a custom pGM vector system along with the methionine aminopeptidase (MAP) gene, as described previously (Natarajan et al. 2011, 2013). We engineered each of the α- and β-chain codon substitutions using the QuikChange® II XL Site-Directed Mutagenesis kit from Stratagene (La Jolla, CA, USA). Each engineered codon change was verified by DNA sequencing.
We carried out recombinant Hb expression in the JM109 (DE3) E. coli strain. Bacterial cells were selected in LB agar with dual antibiotics (ampicillin and kanamycin) to ensure that transformants received both pGM and pCO-MAP plasmids for expression. The expression of each rHb mutant was carried out in 1.5L of TB medium. Bacterial cells were grown in 37°C in an orbital shaker at 200rpm until absorbance values reached 0.6–0.8 at 600nm. The bacterial cultures were induced by 0.2mM IPTG and were then supplemented with hemin (50μg/ml) and glucose (20g/l). The bacterial culture conditions and the protocol for preparing cell lysates are described in Natarajan et al. (2011).
We purified each rHb sample by means of two step ion-exchange chromatography as described previously (Natarajan et al. 2011, 2013; Projecto-Garcia et al. 2013; Cheviron et al. 2014; Galen et al. 2015; Natarajan et al. 2015b; Tufts et al. 2015; Natarajan et al. 2016). Samples were passed through an anion-exchange column (HiTrap™ Q-XL, GE Healthcare, 17-5159-01), equilibrated with 20mM CAPS buffer (0.5mM EDTA, pH 9.7), and eluted using a linear gradient of 0–1.0 M NaCl. The eluted fractions were passed through a cation-exchange column (HiTrapTM SP-XL, GE Healthcare, 17-5161-01) equilibrated with 20mM sodium phosphate buffer (0.5mM EDTA, pH 7.2); the samples were then eluted using a linear gradient of 10mM sodium phosphate buffer (0.5mM EDTA, pH 9.0). The samples were desalted by dialysis against 10mM HEPES buffer (pH 7.4) at 4°C. The eluted fractions of each rHb sample were concentrated by means of centrifugal filtration. The purified rHb samples were analyzed by sodium dodecyl sulphate (SDS)-polyacrylamide gel electrophoresis (supplementary fig. S5, Supplementary Material online) and IEF, which confirmed the absence of subunit heterogeneity in the analyzed rHb samples. After preparing rHb samples as oxyHb, deoxyHb, and carbonmonoxy derivatives, we measured absorbance at 450–600nm to confirm that the absorbance maxima match those of the native HbA samples.
We tested for evidence of tetramer–dimer dissociation by means of gel filtration chromatography using FPLC (ÄKTA Pure, GE Healthcare). We used a Superdex 75 10/300 GL column that was pre-equilibrated with 50mM potassium phosphate buffer, pH 7.0, 0.5mM EDTA, 0.15 M NaCl, and we used horse myoglobin and human Hb as protein markers. Human Hb was used as a marker for tetrameric assembly, but it should be noted that vertebrate Hbs typically show apparent molecular weights<64kDa due to reversible dissociation into dimers during passage through the chromatography medium (Weber et al. 2013; Storz et al. 2015). Using the linear relationship between elution volume and log-transformed molecular weight for the reference proteins, we calculated the log-molecular weight of each nightjar rHb mutant by solving the regression equation y=0.330x+5.329, where x is the measured elution volume. We prepared the protein samples in the same buffer using a heme concentration of 0.16mM. We measured absorbance at 415nm with a flow rate of 0.5ml/min.
We assessed the pH-dependent stability of the 16 rHbs by means of UV–visible spectroscopy. We prepared 20mM filtered buffers spanning the pH range 2.0–11.0. We prepared 20mM glycine-HCl for pH 2.0–3.5; 20mM acetate for pH 4.0–5.5; 20mM phosphate for pH 6.0–8.0; 20mM glycine-NaOH for pH 8.5–10.0; 20mM carbonate-NaOH for pH 10.5 and phosphate-NaOH for pH 11.0. We diluted the purified rHb samples in the pH-specific buffers to achieve uniform protein concentrations of 0.15mg/ml. We incubated the samples for 3–4h at 25°C prior to spectroscopic measurements, and we maintained this same temperature during the course of the experiments. We measured absorbance in the range 260–700nm using a Cary Varian Bio100 UV–Vis spectrophotometer (Varian Inc., USA) with Quartz cuvettes, and we used IGOR Pro 6 (WaveMetrics Inc., USA) to process the raw spectra. For the same set of 16 rHb mutants, we tested for changes in secondary structure of the globin chains by measuring circular dichroism spectra on a JASCO J-815 spectropolarimeter (JASCO Corp., Japan) using a quartz cell with a path length of 1mm. We assessed changes in secondary structure by measuring molar ellipticity in the far UV region between 190 and 260nm in three consecutive spectral scans per sample.
We treated purified rHb samples with potassium ferricyanide (K3[Fe(CN)6]) to remove bound CO, and we maintained rHbs in the ferrous (Fe2+) state by treating the samples with sodium dithionite (Na2S2O4). We then prepared oxyHb by passing the samples through a Sephadex G-50 column. For each rate measurement, we used 200μl of 20 µM oxyHb in 100mM potassium phosphate buffer, pH 7.0, containing 1mM EDTA and 3mmol/mol of heme catalase and superoxide dismutase. To measure the spontaneous conversion of ferrous (Fe2+) oxyHb to ferric (Fe3+) metHb we recorded the absorbance spectrum at regular intervals over a 90h period. The spectra were collected between 400 and 700 nm using a BioTek Synergy2 multi-mode microplate reader (BioTek Instruments, Inc., USA). We estimated autoxidation rates by plotting the A541/A630 ratio (ratio of absorbances at 540 and 630 nm) vs. time, using IGOR Pro 6.37 software (Wavemetrics, Inc., USA).
We modeled structures of nightjar Hbs using MODELLER ver. 9.17 (Webb and Sali 2016). We selected human Hb (DPB ID 1hho and 3hhb) as templates for oxy- and deoxyHb conformations. We performed additional calculations using Hydrogen Bond Calculation Web server ver. 1.1 (http://cib.cf.ocha.ac.jp/bitool/HBOND/) and the PyMOL Molecular Graphics System (ver. 1.8; Schrödinger, San Diego, CA).
Supplementary data are available at Molecular Biology and Evolution online.
We thank the Peruvian government agencies INRENA and SERFOR for permits. We thank T. Valqui, E. Bautista, A. B. Johnson, S. G. DuBay, C. J. Schmitt, M. R. Jones, and CORBIDI for assistance in the field, and D. Carter and E. E. Petersen for assistance in the lab. We also thank A. Signore and two anonymous reviewers for comments on the manuscript. This work was supported by the National Institutes of Health/National Heart, Lung, and Blood Institute (HL087216 to J.F.S.), the National Science Foundation (MCB-1517636 to J.F.S., DEB-1146491 to C.C.W., and MCB-1516660 to C.C.W.), and the Danish Council for Independent Research, Natural Sciences (4181-00094 to A.F.).