Gene cooption and convergent evolution of oxygen transport hemoglobins in jawed and jawless vertebrates

Abstract

Natural selection often promotes evolutionary innovation by coopting preexisting genes for new functions, and this process may be greatly facilitated by gene duplication. Here we report an example of cooptive convergence where paralogous members of the globin gene superfamily independently evolved a specialized O₂ transport function in the two deepest branches of the vertebrate family tree. Specifically, phylogenetic evidence demonstrates that erythroid-specific O₂ transport hemoglobins evolved independently from different ancestral precursor proteins in jawed vertebrates (gnathostomes) and jawless fish (cyclostomes, represented by lamprey and hagfish). A comprehensive phylogenetic analysis of the vertebrate globin gene superfamily revealed that the erythroid hemoglobins of cyclostomes are orthologous to the cytoglobin protein of gnathostome vertebrates, a hexacoordinate globin that has no O₂ transport function and that is predominantly expressed in fibroblasts and related cell types. The phylogeny reconstruction also revealed that vertebrate-specific globins are grouped into four main clades: (i) cyclostome hemoglobin + cytoglobin, (ii) myoglobin + globin E, (iii) globin Y, and (iv) the α- and β-chain hemoglobins of gnathostomes. In the hemoglobins of gnathostomes and cyclostomes, multisubunit quaternary structures provide the basis for cooperative O₂ binding and allosteric regulation by coupling the effects of ligand binding at individual subunits with interactions between subunits. However, differences in numerous structural details belie their independent origins. This example of convergent evolution of protein function provides an impressive demonstration of the ability of natural selection to cobble together complex design solutions by tinkering with different variations of the same basic protein scaffold.

Natural selection often promotes evolutionary innovation by coopting preexisting genes for new functions. Gene cooption may have played a role in major episodes of adaptive change in multicellular organisms, and it appears to be an important mechanism for generating morphological and physiological diversity (1–5). Gene duplication may be an especially important facilitator of cooptive evolution (6). This is well illustrated in the vertebrate globin gene superfamily, because there are several well-documented cases where paralogous gene copies have acquired distinct physiological functions and/or patterns of expression (7–11).

Globins are ancient proteins that are present in each of the three domains of life (11–13). Throughout the 20th century, myoglobin (Mb; an O₂ storage protein in muscle) and hemoglobin (Hb; an O₂ transport protein in red blood cells) were the only known globin proteins in vertebrates (8, 14). Early in the 21st century, comparative genomic studies revealed a surprising diversity of novel globin genes in vertebrates, including neuroglobin (Ngb) (15), cytoglobin (Cygb) (16–18), globin-E (GbE) (19), globin-X (GbX) (20), and globin-Y (GbY) (21). The discovery of these novel globin genes has motivated experimental studies to elucidate their physiological functions and evolutionary studies to assess their phylogenetic affinities and taxonomic distributions (22–28).

Phylogenetic studies have revealed that vertebrate globins fall into two distinct clades. One clade contains GbX and Ngb, two highly divergent genes, which appear to be more closely related to annelid intracellular globins than to any other vertebrate globins (20, 21). The other clade contains a set of genes that are products of vertebrate-specific duplication events: Cygb, GbE, GbY, Mb, and the Hbs of jawed vertebrates (gnathostomes) and jawless fish (cyclostomes, represented by lampreys and hagfish) (20–22, 28). The monophyly of these vertebrate-specific globins is well supported (20, 21), but phylogenetic relationships within this group remain highly uncertain.

Because the passive diffusion of O₂ in blood plasma is not generally sufficient to meet the metabolic demands of large, active animals, the evolution of Hb-mediated blood–O₂ transport represented a key physiological innovation in vertebrate life that opened up new opportunities for the evolution of aerobic metabolism. In gnathostomes, Hb is a tetrameric protein assembled from two α-chain and two β-chain subunits. The progenitors of the α- and β-globin gene families arose via tandem duplication of an ancestral, single-copy globin gene approximately 450–500 mya, after the gnathostome common ancestor diverged from jawless fishes (29–31). In the α₂β₂ Hb tetramers of most extant gnathostomes, the cooperativity of O₂ binding stems from an oxygenation-linked transition in quaternary structure. The origin of cooperativity was preceded by the gene duplication that gave rise to structurally distinct α- and β-chain subunits (11, 30, 32). By contrast, in the Hbs of extant cyclostomes, cooperativity of O₂ binding stems from oxygenation-linked dissociation of multimers into ligated monomers (33–39). For this reason, cyclostome Hbs have been considered “… a transition stage between invertebrate and vertebrate hemoglobins” (40). Phylogenetic studies of vertebrate globins have presented tree topologies that are not consistent with a single origin of O₂ transport Hbs (22, 28, 41). However, incomplete sampling of taxa and gene lineages has not permitted any definitive conclusions.

Here we report a comprehensive phylogenetic reconstruction of the vertebrate globin gene superfamily that includes representatives from each of the major lineages of gnathostomes as well as cyclostomes. Results of this analysis revealed that the erythroid Hbs of cyclostomes and gnathostomes are not orthologous proteins. Instead, the functionally similar O₂ transport proteins were coopted from phylogenetically distinct and anciently diverged globin protein precursors. This represents an example of “cooptive convergence,” where paralogous members of the same gene family independently evolve the same specialization of function in different lineages. After being pressed into service as O₂ transport proteins, the two paralogous globins independently evolved similar biochemical properties in extant cyclostomes and gnathostomes.

The Hbs of cyclostome and gnathostome vertebrates are encapsulated in red blood cells and both proteins provide a highly efficient means of O₂ transport from the respiratory surfaces to the cells of metabolizing tissues while also contributing to the transport of CO₂ back to the gas exchange organs. The efficiency of both Hbs as O₂ transport proteins stems from subunit–subunit interactions (homotropic effects), which account for the cooperativity of O₂ binding, and the deoxygenation-linked binding of allosteric ligands (heterotropic effects), which provides a mechanism for the cellular regulation of Hb–O₂ affinity. In the Hbs of cyclostomes and gnathostomes, cooperativity and allosteric regulation are made possible by oxygenation-linked changes in quaternary structure (42). Thus, the O₂ transport Hbs of both taxa convergently evolved distinct forms of both homotropic and heterotropic cooperative effects from different ancestral protein monomers that lacked cooperativity.

Results

Description of Data.

We estimated phylogenetic relationships among all vertebrate-specific members of the globin protein superfamily (Cygb, GbE, GbY, Mb, and Hbs), with special attention to the relationship between cyclostome and gnathostome Hbs. We used a set of vertebrate Ngb sequences to root the tree. To compile the globin sequence dataset, we interrogated the genome assemblies of nine gnathostome vertebrates and used bioinformatic tools to annotate the entire globin gene repertoire of each species. These nine species included representatives of all major gnathostome lineages present in the genome databases (teleost fish, amphibians, squamate reptiles, birds, and mammals). We compiled additional sequences from cartilaginous fish and cyclostomes. When possible, we included more than one species per lineage. Our sample included globin sequences from three cartilaginous fish (red stingray, gummy houndshark, and Port Jackson shark), three teleost fish (medaka, pufferfish, and zebrafish), one amphibian (western clawed frog), one squamate reptile (green anole lizard), two birds (chicken and zebra finch), two mammals (human and platypus), and 12 sequences of functional Hbs from three different cyclostome species: sea lamprey (5 paralogous sequences), Arctic lamprey (3 paralogous sequences), and hagfish (4 paralogous sequences), which cover the two extant cyclostome subclasses Myxini and Hyperoartia. We also retrieved a set of Ngb outgroup sequences from a representative set of gnathostome taxa. A complete description of all sequences used is included as SI Appendix, SI Materials and Methods, and SI Appendix, Table S1.

Most of the gnathostome species included in this study possess multiple paralogous copies of α- and β-like globin genes. Because the monophyly of the α- and β- globin gene families has been well established (30, 31), we only included a representative subset of the α- and β-globins from each species in our analyses.

Phylogenetic Relationships Among Vertebrate Globins.

Our primary aims were to reconstruct the globin gene repertoire of the vertebrate common ancestor and to clarify the relationship between cyclostome and gnathostome Hbs. It was traditionally assumed that Hb and Mb originated via duplication of an ancestral, single-copy globin gene before the cyclostome/gnathostome divergence, such that each of these two vertebrate lineages inherited orthologous copies of the same “proto-Hb” gene (11, 30, 32, 40, 43). Under this scenario, we would expect a phylogeny in which the Hbs of cyclostomes are sister to the clade of gnathstome α- and β-Hb genes: [Mb(cyclostome Hb, gnathostome Hb)]. Contrary to this expectation, our maximum likelihood and Bayesian analyses supported a phylogeny in which cyclostome Hb was sister to Cygb, with maximum likelihood bootstrap support of 70% and Bayesian posterior probability of 1.00 (Fig. 1). Cyclostome Hbs, Cygb, GbE, GbYs, and Mb were all placed in strongly supported monophyletic groups, with maximum likelihood bootstrap support values that ranged from 96% to 100% and Bayesian posterior probabilities ≥0.99. Our phylogeny reconstructions also grouped the vertebrate-specific globins into four distinct clades: (i) cyclostome Hb + Cygb, (ii) Mb + GbE, (iii) GbY, and (iv) the α- and β-chain Hbs of gnathostomes (this latter clade is sister to the other three clades of vertebrate-specific globins) (Fig. 1).

Fig. 1.

Maximum likelihood phylogram describing relationships among the Cygb, GbE, GbY, cyclostome Hb, gnathostome α- and β-Hb, and Mb genes of vertebrates. Ngb sequences were included to root the tree. Numbers above the nodes correspond to maximum likelihood bootstrap support values, and those below the nodes correspond to Bayesian posterior probabilities.

We performed a comprehensive sensitivity analysis to evaluate how the phylogenetic results were affected by the use of different alignment algorithms, the use of different amino acid substitution models, and the use of different outgroup sequences (e.g., vertebrate GbX or globins from basal chordates such as the sea squirt, Ciona intestinalis). To do this, we performed phylogenetic searches for 10 alternative alignments of our sequences under three different models of amino acid substitution. In each of these different analyses, vertebrate globins consistently fell into the four main clades described above, and cyclostome Hb was invariably placed as the sister group to gnathostome Cygb. The bootstrap support value for the node joining cyclostome Hb and Cygb ranged from 48% to 70% among maximum likelihood analyses, whereas posterior probabilities for the same node were far less variable, ranging from 0.99 to 1.00. In all analyses, the trees depicting a sister relationship between cyclostome Hb and gnathostome Cygb had uniformly higher likelihood scores than any of the alternative topologies (SI Appendix, Table S2). Finally, we added vertebrate Globin X and Ciona globins as additional outgroup sequences, and, again, the tree depicting a sister relationship between cyclostome Hb and gnathostome Cygb had a higher likelihood score than any of the alternatives (SI Appendix, Table S3). Full results of the sensitivity analysis are provided in SI Appendix, SI Results.

The phylogeny reconstruction shown in Fig. 1 provides the basis for two important conclusions: (i) precursors of the four main globin gene lineages were all present in the common ancestor of extant vertebrates; and (ii) the Hbs of cyclostomes and gnathostomes did not descend from the same ancestral protein in the cyclostome/gnathostome common ancestor. Instead, the cyclostome Hbs are orthologous to the hexacoordinate Cygbs of gnathostome vertebrates. These results are not congruent with any of the previously hypothesized relationships among vertebrate globins. The traditional view regarding the orthology of cyclostome and gnathostome Hbs (11, 30, 32) was based on phylogeny reconstructions that did not include Cygb or other more recently discovered members of the globin protein superfamily (SI Appendix, Fig. S1A). Alternative phylogenetic relationships have been suggested by more recent work, which included a wider coverage of vertebrate globin diversity. For example, trees presented by Burmester et al. (22, 28) depict a close relationship between GbE and Cygb and a basal position for cyclostome Hbs (SI Appendix, Fig. S1B). This phylogeny implies that either the O₂ transport functions of cyclostome and gnathostome Hbs were coopted in parallel from a hexacoordinate ancestral state, or alternatively, the O₂ transport function evolved once and was secondarily lost in the lineage that gave rise to the remaining gnathostome-specific globins. Finally, Katoh and Miyata (41) presented a tree in which Cygb was sister to the cyclostome Hbs, and Mb was the most basal of the vertebrate-specific globins (SI Appendix, Fig. S1C).

These hypotheses make mutually exclusive predictions regarding the evolutionary origins of erythroid O₂ transport Hbs in vertebrates, and these predictions can be tested statistically by using phylogenetic topology tests (44–46). Under the “single cooption” hypothesis (Fig. 2A), the Hbs of cyclostomes and gnathostomes descend from the same ancestral precursor protein, and hence, the O₂ transport function evolved only once in the gnathostome/cyclostome common ancestor. This hypothesis predicts that the Hbs of cyclostomes and gnathostomes should form a monophyletic group to the exclusion of all other vertebrate-specific globins, as shown in Fig. 2A. If the cyclostome Hbs were the basal vertebrate-specific globin (a “parallel cooption or single cooption/secondary loss” scenario) (22, 28), cyclostome Hbs would be expected to be sister to a clade that contains the gnathostome α- and β-Hbs, Cygb, GbE, GbY, and Mb, as shown in Fig. 2B. Finally, if cyclostome and gnathostome Hbs are products of different duplication events (a “convergent cooption” scenario), as suggested by our results and those of Katoh and Miyata (41), we would expect cyclostome Hbs and Cygbs to appear as sister lineages, as shown in Fig. 2C. Results of constrained searches favored the independent origin of cyclostome and gnathostome Hbs in all cases (Fig. 2C and SI Appendix, Table S2). The parametric bootstrapping tests (46, 47) were highly significant in all cases, favoring the convergent cooption scenario (P ≤ 0.001), and the Shimodaira-Hasegawa and approximately unbiased topology tests lacked power to distinguish among the three hypotheses.

Fig. 2.

Alternative hypotheses regarding the phylogenetic relationships between gnathostome and cyclostome Hbs. Under the single cooption hypothesis (A), the Hbs of cyclostomes and gnathostomes derive from a proto-Hb precursor protein that acquired an O₂ transport function in the vertebrate common ancestor. Under the parallel cooption or single cooption/secondary loss hypothesis (B), the O₂ transport function evolved independently in both Hb lineages, or, alternatively, the function was ancestral and was then secondarily lost in the remaining gnathostome-specific globins. Finally, under the convergent cooption hypothesis (C), the O₂ transport function evolved independently in the Hbs of cyclostomes and gnathostomes. Our results favor the convergent cooption hypothesis.

Previous studies have suggested that GbE is more closely related to Cygb than to any other globin (26, 28). Given that GbE has thus far been found only in birds, it was hypothesized to derive from a bird-specific duplication. By contrast, our phylogeny indicates that GbE is more closely related to Mb than to any other globin, and the fact that GbE and Mb are located on the same chromosome in birds is consistent with the phylogenetic results. It thus appears that Mb and GbE derive from a duplication event that predated the gnathostome radiation, and subsequently, the GbE ortholog was independently lost in all gnathostome lineages other than birds. We postulate a similar scenario for GbY, as this gene was probably present in the ancestor of all extant vertebrates and was independently lost in all lineages other than amphibians (as represented by Xenopus), squamate reptiles (as represented by Anolis), and monotreme mammals (as represented by the platypus). Interestingly, the multiple Hbs of lampreys and hagfish do not form reciprocally monophyletic groups (Fig. 1). The phylogenetic patterns indicate that both lineages inherited at least two Hb paralogs from the cyclostome common ancestor (approximately 450 mya) (48, 49), and the globin gene repertoires of lampreys and hagfish were then further expanded by subsequent rounds of lineage-specific gene duplication and divergence.

Discussion

Vertebrate-specific globins can be grouped into four distinct lineages, as represented by (i) cyclostome Hb + Cygb, (ii) Mb + GbE, (iii) GbY, and (iv) the α- and β-chain Hbs of gnathostomes. The common ancestor of extant vertebrates possessed a globin gene repertoire that included progenitors of each of these four distinct gene lineages. Representatives of the first three of these lineages have not been found in cyclostomes, whereas gnathostomes appear to have retained representatives of all four paralogous gene lineages. Subsequent gene duplications and gene losses have occurred in different gnathostome lineages, as illustrated by the independent loss of GbE in all gnathostomes other than birds. These results demonstrate that variation in the globin gene repertoire among extant vertebrates can be attributed to differential retention and loss of an ancestral gene set that was inherited from the vertebrate common ancestor roughly 600 mya in the Cambrian Period.

Cooptive Convergence of Protein Function.

Beyond a certain body size threshold, simple diffusion of O₂ in blood plasma is generally not sufficient to meet the metabolic demands of animal life. Here we report the surprising discovery that similar physiological problems have called forth similar solutions in different lineages during the basal radiation of vertebrates. We discovered that the ancestors of cyclostome and gnathostome vertebrates independently invented erythroid-specific O₂ transport Hbs as a means of enhancing blood–O₂ transport (Fig. 3). In this context, applying the name “Hb” to both proteins simply denotes functional analogy and not homology (8, 11, 28). Although cyclostomes and gnathostomes make use of functionally similar respiratory pigments for blood–gas transport, the superficial similarities in protein function do not reflect continuity of inheritance from a common ancestral protein. Our phylogeny reconstruction indicates that cyclostome Hbs are most closely related to gnathostome Cygb, a hexacoordinate globin protein that is predominantly expressed in the cytoplasm of cells that are actively engaged in the production of extracellular matrix components in visceral organs. The protein may also play a role in intracellular signaling pathways or other functions related to cellular O₂ metabolism (22–28). Some of the functionally similar features related to homo- and heterotropic interactions have a different structural basis in cyclostome and gnathostome Hbs (38, 39), as might be expected if the functions were coopted and modified from different precursor proteins (i.e., different ancestral states). In both cases, multisubunit structures provided the basis for cooperative O₂ binding by coupling the effects of ligand binding at individual subunits with interactions between subunits, but differences in numerous structural details belie their independent origins.

Fig. 3.

An evolutionary model describing the independent evolution of erythroid-specific O₂ transport Hbs in gnathostomes and cyclostomes. The gnathostomes are here represented by amniotes, amphibians, and teleost fish. According to this model, the duplication of an ancestral, single-copy globin gene in the stem lineage of vertebrates produced one descendant gene lineage that eventually gave rise to the α- and β-Hbs of gnathostomes (left branch) and another gene lineage that gave rise to the common ancestor of Mb, Cygb, and cyclostome Hb (right branch). In the latter gene lineage, a subsequent duplication before the vertebrate radiation gave rise to Mb (which was secondarily lost in cyclostomes) and Cygb (= cyclostome Hb). In the gnathostome and cyclostome Hb gene lineages, independent origins of O₂ transport functions are denoted by orange lines. In the case of gnathostome Hb, cooperativity of O₂ binding stems from oxygenation-linked transitions in quaternary structure of an α₂β₂ heterotetramer. The origin of cooperativity and allosteric regulation was preceded by duplication and divergence of the proto α- and β-globin genes. In the case of cyclostome Hb, cooperativity stems from oxygenation-linked dissociation of multimers into ligated monomers. The O₂ transport function of gnathostome Hb preceded the divergence of cartilaginous fish from the ancestor of teleosts and tetrapods approximately 500 mya (48, 78), and the O₂ transport function of cyclostome Hb preceded the divergence between representatives of the two extant cyclostome subclasses, Myxini and Hyperoartia, approximately 450 mya (49). Thus, the convergent evolution of erythroid-specific O₂ transport Hbs appears to have occurred in the early Paleozoic (approximately 450–600 mya), after the split between gnathostomes and cyclostomes, but before the split between cartilaginous fish and the ancestor of teleosts and tetrapods and before the split between hagfish and lamprey. Estimated divergence dates are taken from Hedges (78).

In contrast to the tetrameric Hbs of most gnathostomes, the Hbs of cyclostomes exist as monomers in the oxygenated state and self-associate into dimers or tetramers upon deoxygenation (33–39). This oxygenation-linked reversible aggregation accounts for a modest degree of cooperativity, and the release of Bohr protons upon dissociation into monomers provides a mechanism of allosteric regulation. Heterotetrameric Hbs of the hagfish Eptatretus burger exhibit significant cooperativity (50), and evidence for similar subunit interactions have been documented in multimeric Hb isoforms of the hagfish Myxine glutinosa (37). The formation of heteromultimers composed of unlike subunits appears superficially similar to the α₂β₂ heterotetramers of most gnathostomes, but the oxygenation-linked transition in quaternary structure is completely different. The intersubunit contacts of heterotetrameric gnathostome Hbs primarily involve the C, G, and H helices of the globin chain subunits (51), whereas the contact surfaces of the deoxygenated, homodimeric cyclostome Hbs involve the E helix and the AB corner, such that the heme groups are in almost direct contact (52–55). The heme–heme interactions of cyclostome Hbs are intriguingly similar to those of the homodimer of Cygb (23, 38, 39, 56, 57), an observation that makes sense in light of our inferred phylogenetic relationship between these two globin proteins (Fig. 1).

The convergent or parallel evolution of a given trait in different phylogenetic lineages can often be interpreted as evidence that the trait confers an adaptive advantage. For a globin protein with an O₂ transport function (as opposed to O₂ storage, O₂ scavenging, or O₂ sensing functions), cooperativity is advantageous because it permits rapid and efficient O₂ unloading over a relatively narrow range of blood–O₂ tensions. Moreover, cooperativity permits O₂ unloading at higher partial pressures of O₂ than is possible in the absence of cooperativity, thereby maintaining a pressure gradient between capillary plasma and the tissue mitochondria. The pH dependence of Hb–O₂ affinity (Bohr effect) is advantageous in active animals because it increases the efficiency of O₂ delivery to metabolizing tissues (58, 59). This may explain why the magnitude of the Bohr effect is substantially greater in the Hbs of lampreys than in the generally less active hagfish (39, 60–63).

Conclusion

The ancestors of extant cyclostomes and gnathostomes independently evolved O₂ transport globin proteins by exploiting different mechanisms of oxygenation-linked conformational change in a multisubunit structure. In both cases, the underlying genes also independently evolved erythroid-specific expression. In the Hbs of both gnathostomes and cyclostomes, cooperative O₂ binding is made possible by coupling the effects of ligand binding at individual subunits and the interactions between subunits in the quaternary structure. This example of convergent evolution of protein function provides an impressive demonstration of the ability of natural selection to cobble together complex design solutions by tinkering with different variations of the same basic protein scaffold.

Materials and Methods

Phylogenetic Inference.

Because the goal of this study was to estimate phylogenetic relationships among the Cygb, GbE, GbY, Hb, and Mb genes of vertebrates, we included Ngb sequences from human, platypus, frog, medaka, tetraodon, and zebrafish to root the tree. Previous results indicate that Ngb is an appropriate outgroup because it is more closely related to annelid intracellular globins than to any other vertebrate globin (20, 28). Phylogenetic relationships were estimated using maximum likelihood and Bayesian methods. Because the use of different sequence alignments and substitution models may influence the results of phylogenetic analyses (64, 65), we conducted a comprehensive sensitivity analysis. Specifically, we aligned sequences using 10 alternative methods, and for each resulting alignment, we performed maximum likelihood and Bayesian analyses using two different substitution models. Briefly, we aligned sequences using Dialign (66), Kalign2 (67), the E-INS-i, G-INS-i, and L-INS-i strategies from Mafft v6.17 (68), Muscle v3.5 (69), Prank (70), Probalign (71), Probcons (72), and PROMALS3d (73). A data file containing the complete set of sequence alignments is provided in the SI Appendix and Dataset S1. Maximum likelihood searches were performed under JTT (74), LG (75), and mixed models, and Bayesian searches were performed under the JTT (74) and mixed models.

We report primary results that were based on the Muscle alignment and the mixed model of amino acid substitution. We report all other results in SI Appendix, Table S1. Maximum likelihood searches were implemented in Treefinder version October 2008 (76), and support for the nodes was evaluated with 1,000 bootstrap pseudoreplicates. Bayesian analyses were conducted using MrBayes version 3.1.2 (77), setting two independent runs of four simultaneous chains for 10,000,000 generations, sampling every 2,500 generations, and using default priors. Once convergence was verified, support for the nodes and parameter estimates were derived from a majority rule consensus of the last 2,500 trees.

Hypothesis Testing.

We compared alternative hypotheses using the Shimodaira-Hasegawa (45), approximately unbiased (44), and parametric bootstrapping tests (46, 47). In the case of parametric bootstrapping, for each simulated data set, we calculated the difference in likelihood score, Δ, between the null hypothesis maximum likelihood topology and the alternative hypothesis maximum likelihood topology. Using an α level of 0.01, the null hypothesis maximum likelihood topology was rejected if ≥99% of the simulation-based Δ values exceeded the observed value. All tests were carried out in Treefinder.