Molecular basis of hemoglobin adaptation in the high-flying bar-headed goose

During the adaptive evolution of a particular trait, some selectively fixed mutations may be directly causative and others may be purely compensatory. The relative contribution of these two classes of mutation to adaptive phenotypic evolution depends on the form and prevalence of mutational pleiotropy. To investigate the nature of adaptive substitutions and their pleiotropic effects, we used a protein engineering approach to characterize the molecular basis of hemoglobin (Hb) adaptation in the high-flying bar-headed goose (Anser indicus), a hypoxia-tolerant species renowned for its trans-Himalayan migratory flights. To test the effects of observed substitutions on evolutionarily relevant genetic backgrounds, we synthesized all possible genotypic intermediates in the line of descent connecting the wildtype bar-headed goose genotype with the most recent common ancestor of bar-headed goose and its lowland relatives. Site-directed mutagenesis experiments revealed one major-effect mutation that significantly increased Hb-O2 affinity on all possible genetic backgrounds. Two other mutations exhibited smaller average effect sizes and less additivity across backgrounds. One of the latter mutations produced a concomitant increase in the autoxidation rate, a deleterious side-effect that was fully compensated by a second-site mutation at a spatially proximal residue. The experiments revealed three key insights: (i) subtle, localized structural changes can produce large functional effects; (ii) relative effect sizes of function-altering mutations may depend on the sequential order in which they occur; and (iii) compensation of deleterious pleiotropic effects may play an important role in the adaptive evolution of protein function.


Introduction 28
During the adaptive evolution of a given trait, some of the selectively fixed mutations will be directly 29 causative (contributing to the adaptive improvement of the trait itself) and some may be purely 30 compensatory (alleviating problems that were created by initial attempts at solution). Little is known 31 about the relative contributions of these two types of substitution in adaptive phenotypic evolution and 32 much depends on the prevalence and magnitude of antagonistic pleiotropy (Burch & Chao, 1999 mutations that produce an adaptive improvement in one trait have adverse effects on other traits, then the 36 fixation of such mutations will select for compensatory mutations to mitigate the deleterious side effects, 37 and evolution will proceed as a 'two steps forward, one step back' process. In systems where it is possible 38 to identify the complete set of potentially causative mutations that are associated with an adaptive change 39 in phenotype, key insights could be obtained by using reverse genetics experiments to measure the direct 40 effects of individual mutations on the selected phenotype in conjunction with assessments of mutational 41 pleiotropy in the same genetic background.

4
of the affinity-enhancing αP119A mutation? Given that the substitutions in question involve closely 79 linked sites in the same gene, another possibility is that neutral mutations at the other sites simply 80 hitchhiked to fixation along with the positively selected mutation. Since the other mutations in bar-headed 81 goose Hb have not been tested, we do not know whether αP119A accounts for all or most of the evolved 82 change in O 2 affinity. Moreover, the original studies tested the effect of αP119A by introducing the 83 goose-specific amino acid state into recombinant human Hb. One potential problem with this type of 84 'horizontal' comparisonwhere residues are swapped between orthologous proteins of contemporary 85 speciesis that the focal mutation is introduced into a sequence context that is not evolutionarily 86 relevant. If mutations have context-dependent effects, then introducing goose-specific substitutions into 87 human Hb may not recapitulate the phenotypic effects of the mutations on the genetic background in 88 which they actually occurred (i.e., in the ancestor of bar-headed goose). An alternative 'vertical' approach 89 is to reconstruct and resurrect ancestral proteins to test the effects of historical mutations on the genetic 90 background in which they actually occurred during evolution (Harms & Thornton, 2010; Hochberg & 91 Thornton, 2017). 92 Here we revisit the functional evolution of bar-headed goose Hb, a classic text-book example of 93 biochemical adaptation. We reconstructed the α Aand β A -chain Hb sequences of the most recent common 94 ancestor of the bar-headed goose and its closest living relatives, all of which are lowland species in the 95 genus Anser. After identifying the particular substitutions that are specific to bar-headed goose, we used a 96 combinatorial approach to test the functional effects of each mutation in all possible multi-site 97 combinations. To examine possible pleiotropic effects of causative mutations, we also measured several 98 properties that potentially trade-off with Hb-O 2 affinity: susceptibility to spontaneous heme oxidation 99 (autoxidation rate), allosteric regulatory capacity (the sensitivity of Hb-O 2 affinity to modulation by 100 anionic effectors), and measures of both secondary and tertiary structural stability. Measuring  Results and Discussion 107 Direction of amino acid substitutions 108 Using globin sequences from bar-headed goose, greylag goose, and other waterfowl species in the 109 subfamily Anserinae, we reconstructed the α-and β-chain sequences of the bar-headed goose/greylag 110 goose ancestor, which we call 'AncAnser' because it represents the most recent common ancestor of all 111 extant species in the genus Anser ( Figure 1A). The principle of parsimony clearly indicates that all three 112 of the α-chain substitutions that distinguish the Hbs of bar-headed goose and greylag goose occurred in 113 the bar-headed goose lineage (Gα18S, Aα63V, and αP119A), whereas each of the two β-globin 114 substitutions occurred in the greylag goose lineage (βT4S and βD125E)( Figures 1A,B).
Ancestral protein resurrection and functional testing 117 It is often implicitly assumed that the difference in Hb-O 2 affinity between bar-headed goose and greylag 118 goose is attributable to a derived increase in Hb-O 2 affinity in the bar-headed goose lineage (Black & 119 Tenney, 1980;Gillespie, 1991;Li, 1997;Hochachka & Somero, 2002). In principle, however, the pattern 120 could be at least partly attributable to a derived reduction in Hb-O 2 affinity in the greylag goose lineage, 121 even if αP119A does account for the majority of the change in bar-headed goose. To resolve the polarity 122 of character state change, we synthesized, purified, and functionally tested recombinant Hbs (rHbs) 123 representing the wildtype Hb of bar-headed goose, the wildtype Hb of greylag goose, and the 124 reconstructed Hb of their common ancestor, AncAnser. Functional differences between bar-headed goose 125 and AncAnser rHbs reflect the net effect of three substitutions (αG18S, αA63V, and αP119A) and 126 differences between greylag goose and AncAnser reflect the net effect of two substitutions (βT4S and 127 βD125E; Figure 1B). 128 Since genetically based differences in Hb-O 2 affinity may be attributable to differences in 129 intrinsic O 2 -affinity and/or changes in sensitivity to allosteric effectors in the red blood cell, we measured binding and dissociation rates under the same conditions. 138 The O 2 -equilibrium measurements confirmed the results of previous studies (Petschow et al.,139 1977; Rollema & Bauer, 1979) by demonstrating that the Hb of bar-headed goose has a higher intrinsic 140 O 2 -affinity than that of greylag goose (as revealed by the lower P 50 for stripped Hb)( Figure 2A, Table 1). 141 This difference persisted in the presence of Clions (P 50(KCl) ), in the presence of IHP (P 50(IHP) ), and in the 142 simultaneous presence of both anions (P 50(KCl+IHP) )( Figure 2A, Table 1). The difference in Hb-O 2 affinity 143 between bar-headed goose and greylag goose is mainly attributable to differences in intrinsic affinity, as 144 there were no appreciable differences in sensitivities to allosteric effectors (Table 1). This is consistent 145 with a previous report that native Hbs of bar-headed goose and greylag goose have identical binding 146 constants for inositol pentaphosphate (Rollema & Bauer, 1979). Pairwise comparisons between each of 147 the two modern-day species and their reconstructed ancestor (AncAnser) revealed that the elevated Hb-O 2 148 affinity of the bar-headed goose is a derived character state. O 2 -equilibrium properties of greylag goose 149 and AncAnser rHbs were generally very similar ( Figure 2A). The triangulated comparison involving rHbs 150 from the two contemporary species (bar-headed goose and greylag goose) and their reconstructed 151 ancestor (AncAnser) revealed thatin the presence of physiological concentrations of Cland IHP -72% 152 of the difference in Hb-O 2 affinity between bar-headed goose and greylag goose is attributable to a 153 derived increase in the bar-headed goose lineage and the remaining 28% is attributable to a derived 154 reduction in the greylag goose lineage (Figure 2A). This demonstrates the value of ancestral protein 155 resurrection for inferring the direction and magnitude of historical changes in character state.

156
Kinetic measurements demonstrated that the increased O 2 -affinity of bar-headed goose rHb (i.e., 157 the lower ratio of dissociation/association rate constants, k off /k on ) is attributable to a lower rate of 158 dissociation, k off , in combination with a faster rate of O 2 -binding, k on , relative to the Hbs of both greylag 159 goose and AncAnser ( Figure 2B,C). The Hbs of greylag goose and AncAnser exhibited highly similar 160 rates of both k off and k on ( Figure 2B,C).

162
Effects of individual mutations in bar-headed goose Hb 163 In combination with the inferred history of sequence changes ( Figure 1A,B), the comparison between the 164 rHbs of bar-headed goose and AncAnser indicates that the derived increase in Hb-O 2 affinity in bar-165 headed goose must be attributable to the independent or joint effects of the three substitutions at sites α18, 166 α63, and α119. To measure the effects of each individual mutation in all possible multi-site combinations, 167 we used site-directed mutagenesis to synthesize each of the six possible mutational intermediates that 168 connect the ancestral and descendant genotypes ( Figure 1B). In similar fashion, we synthesized each of 169 the two possible mutational intermediates that connect AncAnser and the wildtype genotype of greylag 170 goose ( Figure 1B). 171 The analysis of the bar-headed goose mutations on the AncAnser background revealed that 172 mutations at each of the three α-chain sites (αG18S, αA63V, and αP119A) produced significant increases 173 in intrinsic Hb-O 2 affinity (indicated by reductions in P 50(stripped) )( Figure 3, Table 1). The Pα119A 174 mutation had the largest effect on the ancestral background, producing an 18% reduction in P 50(stripped) 175 (increase in intrinsic Hb-O 2 affinity). On the same background, αG18S or αA63V produced 7% and 14% 176 reductions in P 50(stripped) , respectively. In the set of six (=3!) possible mutational pathways connecting the 177 low-affinity AncAnser genotype (GAP) and the high-affinity bar-headed goose genotype (SVA), the 178 αP119A mutation produced a significant increase in Hb-O 2 affinity on each of four possible backgrounds 179 6 (corresponding to the first step in the pathway, two alternative second steps, and the third step; Figure 3). 180 When tested on identical backgrounds, αP119A invariably produced a larger increase in intrinsic Hb-O 2 181 affinity than either αG18S or αA63V. Nonetheless, of the six possible forward pathways connecting GAP 182 and SVA, αP119A had the largest effect in four pathways and αA63V had the largest effect in the 183 remaining two. The two pathways in which αA63V had the largest effect were those in which it occurred 184 as the first step. In fact, αG18S or αA63V only produced significant increases in Hb-O 2 affinity when 185 they occurred as the first step. The effects of these two mutations were always smaller in magnitude when 186 they occurred on backgrounds in which the derived α119-Ala was present. In addition to differences in 187 average effect size, αP119A also exhibited a higher degree of additivity across backgrounds than the other 188 two mutations. For example, the affinity-enhancing effect of αP119A on the AncAnser background is 189 mirrored by a similarly pronounced reduction in O 2 -affinity when the mutation is reverted on the wildtype Given that the AncAnser and greylag goose rHbs exhibit similar equilibrium and kinetic O 2 -binding 208 properties ( Figure 2A,B,C), the two greylag goose mutations (βT4S and βD125E) obviously do not 209 produce an appreciable net change in combination. Interestingly, however, each mutation by itself 210 produces a slightly reduced sensitivity to IHP (  O 2 -transport molecule. Although mutational changes in intrinsic O 2 affinity (∆log P 50(stripped) ) were not 230 7 significantly correlated with changes in autoxidation rate in the full dataset (r = -0.311), analysis of the 231 bar-headed goose rHb mutants revealed a striking pairwise interaction between mutations at α18 and α63 232 (residues which are located within 7 Ǻ of one another). The αA63V mutation produced a >2-fold increase 233 in the autoxidation rate on backgrounds in which the ancestral Gly is present at α18 ( gating function, resulting in an increased susceptibility to heme oxidation. The increased autoxidation rate 239 caused by αA63V is fully compensated by αG18S ( Figure 5), a highly unusual amino acid replacement 240 because glycine is the only amino acid at this site (the C-terminal end of the A helix) that permits the 241 main chain to adopt the typical Ramachandran angles (Figure 4figure supplement 1). Introduction of 242 the serine side chain at α18 in bar-headed goose Hb forces this residue to undergo a peptide flip relative to 243 human Hb, so the carbonyl oxygen points in the opposite direction. This unusual replacement at α18 may 244 be required to accommodate the bulkier Val side chain at α63, thereby alleviating conformational stress. 245 Aside from the compensatory interaction between mutations at α18 and α63, we observed no 246 evidence for trade-offs between O 2 -affinity and any of the other measured functional or structural 247 properties. There were no significant correlations between ∆log P 50(stripped) and changes in allosteric 248 regulatory capacity (Table 1) 2B, source data 2) and there were no significant correlations between ∆log P 50(stripped) and changes in 252 stability over the physiological range (pH 6.5, r = -0.357; pH 7.5, r = -0.052). Likewise, the rHbs 253 exhibited very little variation in the stability of tertiary structure as measured by UV-visible spectroscopy 254 ( Figure 3 figure supplement 2C, source data 3) and there were no significant correlations between ∆log 255 P 50(stripped) and changes in stability over the physiological range (pH 6.5, r = -0.511; pH 7.5, r = -0.338). In 256 summary, we found no evidence for pleiotropic trade-offs between intrinsic O 2 -affinity and any measured 257 properties of Hb structure or function other than autoxidation rate.

259
Conclusions 260 We now return to the two questions we posed at the outset: 261 ( (2) Do function-altering mutations have deleterious pleiotropic effects on other aspects of protein 277 structure or function? 278 On the AncAnser background, the affinity-enhancing mutation, αA63V, produces a pronounced increase 279 in the autoxidation rate. This is consistent with the fact that engineered Hb and myoglobin mutants with 280 altered affinities often exhibit increased autoxidation rates ( Vector construction and site-directed mutagenesis 308 After optimizing nucleotide sequences of AncAnser α Aand β A -globin genes in accordance with E. coli 309 codon preferences, we synthesized the α A -β A -globin cassette (Eurofins MWG Operon). We cloned the incubated them in an ice bath for 30 min. Following sonication of the cells, we added 0.5-1.0% 331 polyethyleneimine solution, and we then centrifuged the crude lysate at 13,000 rpm for 45 min at 4°C. 332 We purified the rHbs by means of two-step ion-exchange chromatography. Using high-333 performance liquid chromatography (Äkta start, GE Healthcare), we passed the samples through a cation 334 exchange-column (SP-Sepharose) followed by passage through an anion-exchange column (Q- spectrum at regular intervals over a 90 h period. We collected spectra between 400nm and 700nm using a 382 BioTek Synergy2 multi-mode microplate reader (BioTek Instruments). We estimated autoxidation rates 383 by plotting the A 541 /A 630 ratio (ratio of absorbances at 540nm and 630nm) vs time, using IGOR Pro 6.37 384 software (Wavemetrics). We used the exponential offset formula in IGOR to calculate the 50% 385 absorbance per half-life (i.e., 0.5AU/half-life).

387
Measurements of structural stability 388 We assessed the pH-dependent stability of the rHbs by means of UV-visible spectroscopy. We prepared 389 20 mM filtered buffers spanning the pH range 2.0-11.0. We prepared 20 mM glycine-HCl for pH 2.0-390 3.5; 20 mM acetate for pH 4.0-5.5; 20 mM phosphate for pH 6.0-8.0; 20 mM glycine-NaOH for pH 8.5-391 10.0; 20 mM carbonate-NaOH for pH 10.5 and phosphate-NaOH for pH 11.0. We diluted the purified 392 rHb samples in the pH-specific buffers to achieve uniform protein concentrations of 0.15 mg/ml. We 393 incubated the samples for 3-4 h at 25°C prior to spectroscopic measurements, and we maintained this 394 same temperature during the course of the experiments. We measured absorbance in the range 260-700 395 nm using a Cary Varian Bio100 UV-Vis spectrophotometer (Varian) with Quartz cuvettes, and we used 396 IGOR Pro 6.37 (WaveMetrics) to process the raw spectra. For the same set of rHbs, we tested for changes 397 in secondary structure of the globin chains by measuring circular dichroism spectra on a JASCO J-815 398 spectropolarimeter using a quartz cell with a path length of 1 mm. We assessed changes in secondary 399 structure by measuring molar ellipticity in the far UV region between 190 and 260 nm in three 400 consecutive spectral scans per sample.

402
Structural modeling 403 We modelled structures of goose Hbs and the various mutational intermediates using the program COOT       Figure 1. Inferred history of amino acid substitution at five sites that distinguish the major Hb isoforms of the bar-headed goose (Anser indicus) and greylag goose (Anser anser). (A) Amino acid states at the same sites are shown for 12 other waterfowl species in the subfamily Anserinae. Of the five amino acid substitutions that distinguish the Hbs of A. indicus and A. anser, parsimony indicates that three occurred on the branch leading to A. indicus (αG18S, αA63V, and αP119A) and two occurred on the branch subtending the clade of all Anser species other than A. indicus (βT4S and βD125E).   The inset graphic shows the environment of the α63-Val residue. When valine replaces the ancestral alanine at this position, the larger volume of the side-chain causes minor steric clashes with two neighboring glycine residues, α25-Gly and α59-Gly and is predicted to increase flexibility of the AB corner. The distances between non-hydrogen atoms (depicted by dotted lines) are given in Ǻ.    . Glycine residues are denoted by triangles, other residues by squares. One residue, α18-Ser, is conspicuous by its unusual backbone angles, and is shown as a green square. This position in the Ramachandran plot is highly unusual for any residue other than glycine. The turn in the backbone between the A and B helices can only be accommodated by a glycine, since the lack of a side-chain avoids the strong steric clash that would develop between a C β atom and the nitrogen atom of residue 19. The serine at α18 is therefore forced to flip the peptide conformation, such that its carbonyl group points in the opposite direction relative to that of Gly 18.