Repeated evolution of similar phenotypes: Integrating comparative methods with developmental pathways

Abstract

Repeated phenotypes, often referred to as ‘homoplasies’ in cladistic analyses, may evolve through changes in developmental processes. Genetic bases of recurrent evolution gained attention and have been studied in the past years using approaches that combine modern analytical phylogenetic tools with the stunning assemblage of new information on developmental mechanisms. In this review, we evaluated the topic under an integrated perspective, revisiting the classical definitions of convergence and parallelism and detailing comparative methods used to evaluate evolution of repeated phenotypes, which include phylogenetic inference, estimates of evolutionary rates and reconstruction of ancestral states. We provide examples to illustrate how a given methodological approach can be used to identify evolutionary patterns and evaluate developmental mechanisms associated with the intermittent expression of a given trait along the phylogeny. Finally, we address why repeated trait loss challenges strict definitions of convergence and parallelism, discussing how changes in developmental pathways might explain the high frequency of repeated trait loss in specific lineages.

Keywords:
Repeated evolution; recurrent phenotypes; comparative methods; developmental pathways

Introduction

Similar phenotypes may emerge several times along the evolutionary history of a given lineage, characterizing phylogenetically-discontinuous traits that are often referred to as ‘homoplasies’ in cladistic analyses (see West-Eberhard, 2003; Wake et al., 2011; Orgogozo, 2015). The intermittent expression of a given trait along an evolutionary trajectory is developmentally feasible because regulatory changes may modulate genetic pathways and also turn on and off the signaling cascades related to the establishment of that phenotype (West-Eberhard, 2003). Genetic mechanisms involved in the repeated evolution of specific traits have puzzled researchers for decades (e.g. Hunt et al., 1998; Shi and Yokoyama, 2003; Schluter et al., 2004; Rosenblum et al., 2010; Davies et al., 2012; Guerreiro et al., 2013; Projecto-Garcia et al., 2013; Liu et al., 2014; Nery et al., 2016; Mohammadi et al., 2016; Hu et al., 2017; Liu et al., 2022). Recent advances in modern analytical phylogenetic tools and the stunning assemblage of new information on developmental mechanisms in the past years enable us to evaluate the topic under an integrated perspective and also to revisit major concepts and classical examples of phenotypic recurrence in nature. We start this review by reassessing the classical definitions of convergence and parallelism at different biological levels. Then, we detail the principal comparative methods used to evaluate repeated evolution of similar phenotypes, focusing on phylogenetic inference, estimates of evolutionary rates and reconstruction of ancestral states. Together with the synthetic presentation of each method, we provide a few examples to illustrate how that methodological approach can be used to identify evolution patterns and evaluate developmental mechanisms associated with the intermittent expression of a given trait along the phylogeny. Finally, we discuss why repeated trait loss challenges strict definitions of convergence and parallelism, and address how changes in developmental pathways might explain the high frequency of repeated trait loss in specific lineages. Across this discussion, we adopt the expression ‘recurrent phenotypes’ to refer to similar traits that emerged several times along a given phylogeny regardless of the genetic mechanism underlying the evolution of such similarities, so that the term per se does not imply a distinction between parallelism or convergence at the phenotypic level (as further explained, see also West-Eberhard, 2003 for an extensive discussion on ‘recurrence’).

Revising concepts: Convergence and parallelism

The extensive interest on how similar phenotypes repeatedly evolved in nature has motivated researchers from different fields to intensively investigate the mechanisms associated to these similarity patterns and to propose concepts delimiting the processes that explain recurrent phenotypes. Two concepts - convergence and parallelism - have appeared with increasing frequency in evolutionary studies along the past three decades (Figure 1), and are addressed in this section. Equivalent selective pressures are often claimed to be a possible explanation for the recurrent evolution of similar phenotypes among phylogenetically-distant lineages (see Wake, 1991). Given that the same phenotype might result from different genetic trajectories (a concept known as ‘many-to-one’ mapping of genotype to phenotype), the repeated evolution of similar phenotypes turns into an even more interesting event (Storz, 2016). Accordingly, the concepts of convergence and parallelism ultimately focus on how similar are the mechanisms underlying a recurrent phenotype.

Figure 1 -
Number of publications with the terms “convergence/convergent evolution” (light gray) and “parallelism/parallel evolution” (dark gray) from 1990 to 2021. Data retrieved from Web of Science (https://www.webofscience.com/).

At the phenotypic level, evolutionary similarities observed among different lineages (here termed ‘recurrent phenotypes’) have been classically defined as parallel or convergent evolution (see Scotland, 2011; Rosenblum et al., 2014) based initially on the distances among taxa. Specifically, similar phenotypes among closely related lineages agree with the definition of parallelism, while those among distantly related taxa would correspond to convergence (Figure 2). The criterion for differentiating ‘distance’ among lineages, however, may be vague (see Davis and Heywood, 1963; Conte et al., 2012; Rosenblum et al., 2014). Other studies provided alternative definitions for both terms (reviewed in Gompel and Prud’homme, 2009; Wake et al., 2011), until completely removing the term parallelism from the classification of evolutionary similarities at the phenotypic level (see Arendt and Reznick, 2008). In this review, we opted for not distinguishing convergent and parallel evolution at the phenotypic level; instead, we adopt the term ‘recurrent phenotypes’ and untangle this discussion from the main focus of our review, which are the genetic mechanisms underlying evolution of phenotypic similarities among different lineages.

Figure 2 -
Application of the terms “convergence” and “parallelism” at the phenotypic level was originally based on the phylogenetic distance among lineages that evolved similar phenotypes.

The molecular processes associated with recurrent phenotypes are often unknown, and several studies aim to elucidate whether repeated evolution is usually settled on the same or in different developmental pathways (as further discussed in this review). We can evaluate the molecular bases of phenotypic recurrence at two levels: 1) the locus level, which concerns the molecules (e.g. DNA sequence or protein) as a whole; and 2) the site level, which considers each site (e.g. nucleotide or amino acid) independently. At the locus level, recurrent phenotypes from trait changes involving different metabolic pathways are defined as convergence (see box in the left at Figure 3a), while those involving changes in the same metabolic pathway are referred to as parallelism (see box in the right at Figure 3A). Cases interpreted as parallelism can be also evaluated regarding whether the identified changes reside in the same genome regions or not (see Figure 3A). At the site level, two or more lineages can independently have the same nucleotide or amino acid at the same position (Figure 3B). When the ancestral basis or the ancestral amino acid is the same for both lineages, it is considered a parallel substitution. In the case of different origins, these substitutions are referred to as convergent substitutions (Storz, 2016).

Figure 3 -
The terms “convergence” and “parallelism” are used to describe the genetic basis of recurrent phenotypes at two different levels: (A) locus and (B) nucleotide or amino acid sites. (A) At the locus level, species 1 (sp1) and species 3 (sp3) share a recurrent phenotype. In the box at the left (‘Convergence’), the red ray indicates the molecular basis (gene1 and TF2, respectively) associated with the recurrent phenotype in sp1 and sp3, illustrating a case of molecular convergence in which genetic changes in the species reside at different signaling pathways. In the box at the right (‘Parallelism’), the example along the column ‘same metabolic pathway/same sequence’ illustrates a genetic basis of the recurrent phenotype in sp1 and sp3 settled at the enhancer (red ray), while that the column ‘same metabolic pathway/different sequence’ illustrates a case where genetic changes in sp1 and sp3 locate at different components of the same signaling pathway (red rays at the gene and the TF, respectively). (B) Site substitutions from different ancestral nucleotides or amino acids represent a convergence (left), while those resulting from the same trajectory are defined as a parallelism (right)

Molecular patterns can be also categorized based on their location within the genome. In this case, changes in protein-coding regions are often regarded as ‘genetic’, while changes in non-coding genomic loci are frequently referred to as regulatory or epigenomic. For instance, both have potential effects on the phenotype - the former by directly modifying the protein sequence and structure, and the latter by influencing gene expression (Bulger and Groudine, 2011; Meddens et al., 2019).

Crosstalk between convergence-parallelism and regulatory networks-gene interactions

As aforementioned, molecular patterns at the locus level associated with recurrent phenotypes are usually defined as parallelism when involving the same sequences, and as convergence when related to different sequences. The comparison of orthologous sequences or proteins has been a central point for several studies that evaluated molecular bases of recurrent phenotypes (e.g. Rosenblum et al. 2010; Davies et al., 2012; Guerreiro et al., 2013; Projecto-Garcia et al., 2013; Liu et al., 2014; Mohammadi et al., 2016; Nery et al., 2016; Hu et al., 2017; Liu et al., 2022). Nonetheless, genes are part of regulatory networks, interacting with cis-regulatory elements (such as enhancers and promoters) and transcription factors that control the expression of one gene or a group of genes (Babu et al., 2004; Wagner and Lynch, 2008; Voordeckers et al., 2015). A greater number of sequences working together, as in complex regulatory networks, might confer flexibility to developmental interactions and eventually facilitate repeated evolution of similar phenotypes in different lineages (see Orr, 2005; Rosenblum et al., 2014; Yeaman et al., 2018; Pereira et al., 2022).

Pleiotropy is also an important topic to be considered in discussions regarding the molecular bases of recurrent phenotypes and associated regulatory pathways. Several genes are pleiotropic, which means that a given gene is involved in the establishment of different phenotypic traits (Lobo, 2008). Changes in that gene, therefore, likely affect several processes simultaneously. In highly pleiotropic genes, changes in cis-regulatory elements might be a powerful tool in evolution because the modular architecture of these regions enable that changes affecting gene expression in specific tissues or cells and also modifying developmental times of specific structures do not compromise other phenotypic traits (Prud’homme et al., 2006; Monteiro and Podlaha, 2009; Feigin et al., 2019, Morris et al., 2020).

Despite several studies focusing on cis-regulatory convergent evolution (e.g. Booker et al., 2016; Kvon et al., 2016; Partha et al., 2017; Tollis et al., 2018; Feigin et al., 2019; Sackton et al., 2019), some questions remain central to this discussion. Do different changes in the same regulatory pathway challenge strict definitions of convergence and parallelism? After all, when changes occur in different sequences that are involved in the same regulatory network, but also associated with other developmental pathways, shall we classify them as convergence, or parallelism?

Comparative methods: Molecular associations of recurrent phenotypes

In this section, we focus on phylogenetic comparative methods (PCMs) based on a phylogenetic inference that are frequently used to address the molecular bases of recurrent phenotypes. Phylogeny and ancestral character reconstructions are essential to evaluate repeated evolution of a given phenotype among different lineages (see Speed and Arbuckle, 2017 for a review in methods of studies addressing recurrent phenotypes). Phylogenetic inferences aim to recover information from the topology (=the relative branching order) and branch lengths (=evolutionary distance or probability of character change) related to a given group (Baum and Smith, 2013). Several methods have been developed for phylogenetic inference (e.g., distance and statistical or probabilistic methods), and this step is considered essential to evaluate evolutionary patterns of recurrent phenotypes (Garland et al., 2005). Probabilistic methods are represented by the maximum likelihood (Felsenstein, 1981, 1985) and Bayesian (Rannala and Yang, 1996; Mau et al., 1999) approaches.

The increasing availability of genomic data makes it possible to perform a comprehensive search for signatures of similarities in a genomic scale (Speed and Arbuckle, 2017). Several studies use tools for a genomic search (e.g. Thomas and Hahn, 2015; Chikina et al., 2016; Hu et al., 2017; Sackton et al., 2019), while others focus on certain genes or specific regulatory pathways already known to be related with the studied phenotype (e.g. Mohammadi et al., 2016; Pereira et al., 2022). Significant progress in the fields of comparative genomics and functional genomics recently provided a deep understanding of regulatory mechanisms likely involved in these evolutionary processes (Lamichhaney et al., 2019).

Gene/site tree and species tree incongruence

The phylogeny inference based on one genetic locus results in a gene tree, or genealogy. This approach contrasts with that used for a species tree, which contains several, if not all, gene trees (Maddison, 1997). In practice, the species tree based on molecular data can be built using a group of concatenated genes [supermatrix approach] or as a summary of dozens of gene trees [multispecies coalescent approach] (Rannala et al., 2020). Some of the software used to perform these analyses are synthesized at Table 1. Incongruence between the genealogy and a species tree can result from diverse biological factors, including incomplete lineage sorting [ILS], introgression, and lateral gene transfer (see Maddison, 1997). These factors are called hemiplasy, a term used to define a pattern similar to homoplasy but produced by a non-homoplasy event, which may result in an apparent similarity in the genealogy and also affect reconstructions of the ancestral sequence (Avise and Robinson, 2008; Mendes et al., 2016).

Table 1 -
Comparative analyses used to evaluate convergent and parallel evolution, with most used software and associated references.

Analysis		Softwares		References
Phylogenetic Tree Inference (Topo) Analysis of Evolutionary Rates (BL) Concordance factor of genes (gCFs) Concordance factor of sites (sCFs) Phylogenetic signal (PS)	Topo and BL	RaxML		Stamatakis, 2006
		IQTree		Nguyen et al., 2015; Minh et al., 2020b
		MrBayes		Ronquist & Huelsenbeck, 2003
		BEAST		Drummond & Rambaut, 2007; Bouckaert et al., 2014
	BL	aaML	PAML	Yang, 2007
	gCFs	IQTree		Minh et al., 2020a
	sCFs	IQTree		Mo et al., 2023
	PS	SH-test	CONSEL	Shimodaira and Hasegawa, 2001
Correlations between morphotypes and sequence rates		Forward Genomics		Hiller et al., 2012; Prudent et al., 2016
		TraitRateProp		Levy et al., 2017
		TraitRELAX		Halabi et al., 2020
		RERconverge		Kowalczyk et al., 2019
		Coevol		Lartillot & Poujol, 2011
		PhyloAcc		Hu et al., 2019
		PhyloAcc-GT		Yan et al., 2022
Ancestral state reconstruction		make.simmap/phytools	R	Revell, 2012
		baseML or codeML	PAML	Yang, 2007
Selection tests (Branch/Clade Model)		codeML	PAML	Yang, 2007
		aBSREL	HyPhy	Smith et al., 2015
		BUSTED		Murrell et al., 2015
		RELAX		Wertheim et al., 2015
Selection Tests (Branch-site Model)		codeML	PAML	Yang, 2007
		FEL	HyPhy	Pond et al., 2005
		FUBAR		Murrell et al., 2013
		MEME		Murrell et al., 2012
		SLAC		Pond et al., 2005

Incongruence between topologies may also represent genetic convergence or parallelism (homoplasy) and, in this case, the comparison of gene and species trees represents an effective approach, for both coding and regulatory sequences. As phylogenetic analyses compare site-by-site similarities, convergence or parallelism in one or more sites (as illustrated in Figure 3) may erroneously group species, possibly influencing the phylogenetic inference analyses and causing a genetic tree discordance (i.e. clustering phylogenetically unrelated species in the gene tree), which is also known as phylogenetic incongruence. Therefore, the comparison between a gene topology and the most-accepted species tree is a tool used to detect possible effects of molecular similarity (Davies et al., 2012). Some methods have been developed to assist identification of the proportion of genes (gene support frequency or gene concordance factor) and sites (site concordance factor) that align with a given species tree (Ané et al., 2007; Minh et al., 2020a; Mo et al., 2023), as synthesized in Table 1.

It is worth noting that this approach detects similarity but does not distinguish convergence from parallelism. Subsequent tests estimating the phylogenetic signal can provide a statistical value of how much the alternative topology (gene tree) is supported given the expected species phylogeny (see Blomberg et al. 2003; Münkemüller et al. 2012). A more quantitative approach is, however, necessary to estimate evolutionary parameters and test competing hypotheses (Ansari and Didelot, 2016).

An example of phylogenetic inference: repeated evolution of laryngeal echolocation in bats

A topic that exemplifies the application of phylogenetic tree inference is the repeated evolution of echolocation among bats. Echolocation is a biological sonar that evolved independently in lineages as distant as bats and whales (Shen et al., 2012; Liu et al., 2014; Thomas and Hahn, 2015). Within Chiroptera (bats), this phenotype is observed in two non-related lineages: the suborder Yangochiroptera and the superfamily Rhinolophoidea (suborder Yinpterochiroptera). In addition to the superfamily Rhinolophoidea, Yinpterochiroptera also includes the Pteropodidae family of non-echolocating Old World fruit bats (Liu et al., 2014). As specialized hearing co-evolves with echolocation, two genes (Tmc1 and Pjvk) associated with nonsyndromic hearing loss in mammals are particularly interesting to understand the evolution of echolocation among bats (Vater and Kössl, 2004; Xu et al., 2013). Phylogenetic inference estimating gene trees for Tmc1 and Pjvk erroneously group laryngeal echolocating bat lineages in a monophyletic clade (see Davies et al., 2012), suggesting molecular similarity of these genes among groups. Subsequent studies (see Liu et al., 2022) revisited the topic and found evidence for a single origin of laryngeal echolocation in bats and an eventual loss in the Pteropodidae family, and hemiplasy may also explain the patterns of evolutionary similarity observed in these bats.

Evolutionary rates analyses

Phylogenetic analyses may also provide information regarding Evolutionary Rates (ER), which are very useful to evaluate molecular bases associated with the repeated evolution of similar phenotypes. ERs are estimated from the amount of nucleotide or amino acid changes in a given lineage over a specific period of time (Baum and Smith, 2013). Phenotypic transitions may involve changes in selection forces on the genes or proteins related to those phenotypes, causing a shift in the evolutionary rates of the sequences (Kowalczyk et al., 2019). One approach often used consists of investigating shifts in the ER occurring independently on the branches of lineages with recurrent phenotypes (Partha et al., 2017; Kowalczyk et al., 2019). The branch lengths are calculated for each gene, so these rates are gene-specific, termed as Relative Evolutionary Rates (RER) by Kowalczyk et al. (2019). These RER for each gene are then correlated with the evolution of a recurrent phenotype across the phylogeny (Partha et al., 2017; Kowalczyk et al., 2019).

As aforementioned, topologies corresponding to gene trees may encompass homoplasy, an effect detected by conflicts between gene trees and species trees. Topology differences may also derive from other factors, including gene evolutionary rates. Genes that evolve rapidly are more prone to involve conflicts attributed to ILS (incomplete lineage sorting), which may result in discrepancies between gene and species trees (Degnan and Rosenberg, 2006), especially if estimated lengths of internal branches are shorter in the species tree than in gene trees (Guerrero and Hahn, 2018). Branch lengths may differ between the gene tree and the species tree even in identical topologies (Edwards, 2009). The positioning of tips associated with long branches may also be imprecise due to an artifact named ‘long branch attraction’ (Degnan and Rosenberg, 2009). Estimates of the “hemiplasy risk factor” - given by the ratio between homoplasy and hemiplasy - can be a valuable tool to estimate the likelihood of incongruence resulting from homoplasy or hemiplasy (Guerrero and Hahn, 2018). Ignoring the mismatch between gene and species trees may result in incorrect estimates of substitution rates when mapping sequences from conflicting loci in the species tree (Mendes et al., 2016). To overcome such a challenge, some programs consider gene tree heterogeneity in their approach (Guerrero and Hahn, 2018; Yan et al., 2022). Despite the vast majority of models treating phenotypes as binary, there are some models that consider associations between genomic substitution rates and continuous phenotypes in the analyses implemented (see Kowalczyk et al., 2019).

Another approach using estimates of ER consists of traditional methods of selection tests hypotheses. These methods are based on codons and therefore useful for coding sequences, and include site (Massingham and Goldman, 2005; Yang et al., 2000), branch (Yang and Nielsen, 2002), branch-site (Zhang et al., 2005) and clade (Yang and Nielsen, 2002; Bielawski and Yang, 2004) models (see Huerta-Cepas et al., 2016; Gao et al., 2019). However, the model that takes into account only the changes among sites (site model) has little utility for analysis of recurrent phenotype. This approach can be used in only one lineage, with a specific trait or set of traits, but may also be implemented to evaluate recurrent phenotypes. Since phenotypic changes are often explained by positive selection, these methods are able to evaluate whether branches or clades with recurring phenotypes likely involve changes in selection regimes (Yang, 1998).

These analyses usually compare the likelihood of neutral models (which reflect genetic drift, for example) with alternative models of evolution, according to which sequence patterns would reflect adaptive evolution or scenarios of constrained changes (Yang, 2007; Smith et al., 2015). A key variable for these selection tests is the ω value (an indicator of selective pressure), which corresponds to the ratio between nonsynonymous [dN] and synonymous [dS] substitution rates (Nei and Gojobori, 1986; Li, 1993, Yang and Nielsen, 2000). In the branch and clade models, the software compare the one-ω ratio model, which assumes the same ω values for all branches, and the two (or more)-ω ratio model, which admits different ω values for some pre-established lineages (Yang, 2002). The ω indicates the type of selection regime acting on a protein-coding gene (ω < 1: purifying selection; ω = 1: neutral evolution; and, ω > 1: positive selection; see Zhang et al., 2005; Yang, 2007). The branch-site model approach combines different ratios across sites and across branches (Zhang et al., 2005). In addition to detecting episodic selection along pre-specified branches in the tree, this analysis identifies the sites of a coding gene evolving under purifying, neutral or positive selection (Zhang et al., 2005; Gharib and Robinson-Rechavi, 2013). It should be taken into account, however, that the analysis considering distantly related species can be misinterpreted due to saturation of sites or amino acids (Lamichhaney et al., 2019).

An example of analyses based on evolutionary rates: Repeated evolution of aquatic mammals

The transition of mammalian lineages to aquatic environments occurred several times and evolved similar phenotypic traits associated to the aquatic lifestyle, including modifications in the hindlimb configuration (Fish and Hui, 1991; Fish et al., 2008), body elongation, and changes in the nostrils relative positioning (Uhen, 2007). Some previous studies have used the ER approach to identify shifts in evolutionary rates among dozens or hundreds of genes (Chikina et al., 2016; Nery et al., 2016), providing evidence for parallel evolution in the evolutionary rates of hundreds of genes during the evolution of three marine mammalian lineages (Cetacea, Pinnipedia and Sirenia; see Chikina et al., 2016). Analyses using selection tests that focused on evolution of Hox genes, a family of genes which encodes transcription factors related to the body plans and development (Carroll, 1995), identified that each aquatic mammalian lineage encompasses a different set of positively-selected Hox genes, which remarkably overlap in their functions during the development of some of these phenotypic traits (Nery et al., 2016).

Ancestral sequence reconstruction (ASR)

Ancestral sequence reconstructions (ASR) are used to statistically infer the ancestral sequences of genes, non-coding regions or proteins within the nodes of a given phylogenetic tree, using present-days homologous sequences (Thornton, 2004; Merkl and Sterner, 2016). These methods are useful for studies of recurrent evolution of similar phenotypes, and allow distinguishing nucleotide or amino acid changes as representing convergent or parallel evolution. In this approach, homologous sequences are aligned and each site or amino acid has its evolutionary history reconstructed using a species phylogeny through a variety of software (Table 1). While this approach can be used to study both coding and regulatory sequences, it is particularly advantageous to evaluate mutations occurring at the same site, as the comparisons are performed site-by-site.

The computational methods of ASR use approaches that were originally developed for phylogenetic analyses (Gumulya and Gillam, 2017). The first method used was maximum parsimony (Fitch, 1971), which assumes to be more likely a reconstruction encompassing the minimum number of substitutions. Development of these methods was followed by the advance of probabilistic approaches - maximum likelihood (‘ML’, Yang et al., 1995; Koshi and Goldstein, 1996) and Bayesian reconstructions (Ronquist and Huelsenbeck, 2003; see Gumulya and Gillam, 2017 for a review). The probabilistic methods deal better with unequal ER (what is expected in cases of recurrent phenotypes) and to estimate the confidence of each inferred ancestral state (Gumulya and Gillam, 2017). The Maximum Likelihood methods are classified in two categories: marginal and joint (Yang et al., 1995). The joint reconstruction is considered more suitable for studies of phenotypic recurrence because it adequately accounts for changes in each site, while the marginal reconstruction is preferred for studies aiming to evaluate the molecular sequences in a particular node, being more often used in studies that aim to reconstruct ancestral proteins (Yang et al., 1995; Gumulya and Gillam, 2017).

Examples of ASR: 1) Ribs in the posterior trunk region of snakes, caecilians and manatees, 2) Resistance to toxic effects of bufadienolides in snakes, 3) Bamboo-eating pandas, 4) Hemoglobin-Oxygen affinity in hummingbirds

Good examples of how ASR analyses contribute to evaluating the repeated evolution of similar phenotypes are illustrated by studies conducted with several animal taxa. The first example we provide relates to the development of ribs in the posterior trunk region (i.e., lumbar region) in some amniote lineages. Most vertebrates exhibit morphologically-distinct regions along the axial skeleton, being the lumbar region characterized by the absence of ribs (Wellik and Capecchi, 2003; Carapuço et al., 2005; McIntyre et al., 2007). Some lineages, however, have ribs associated to the vertebrae in the posterior trunk region - this is the case of iconic animals such as the manatees and elephants (mammals), the snakes (reptiles), and the caecilians (amphibians). The genetic mechanism associated to this rib-associated lumbar morphotype is a recurrent polymorphism that evolved in lineages as distant as snakes, ​​Afrotheria mammals and the lissamphibians Gymnophiona and Urodela (Guerreiro et al., 2013; Pereira et al., 2022). This nucleotide change occurred in the H1 enhancer, a region that regulates the expression of MYF5, a gene involved in rib development in vertebrate embryos. An ancestral sequence reconstruction analysis demonstrated that this is an example of parallelism, with the three substitutions identified occurring from the same base T to the same nucleotide C (Pereira et al., 2022).

Another good example of the ASR approach is provided by studies evaluating snake lineages that are resistant to toxic steroids named bufadienolides. These steroids bind to cell membranes and disable the Na+/K+-ATPase pumps, but some predators evolved resistance to these chemical defenses of toads involving toxic steroids bufadienolides (Mohammadi et al., 2016). Toxic resistance apparently evolved in association with mutations observed even in species that do not appear to prey frogs often, and have originated multiple times in predatory lineages (Mohammadi et al., 2016). Two parallel amino acid changes in the H1-H2 extracellular loop of the Na+/K+-ATPase apparently explain the toxin resistance observed in snakes (Mohammadi et al., 2016), being one the Q[Glutamine]111L[Leucine], and the other a G[Glycine]120R[Arginine].

As another example of ASR analyses, we also can cite the study of repeated diet transitions to bamboo-eating in carnivores. Two non-phylogenetically related species, the giant panda (Ailuropoda melanoleuca, Carnivora, Ursidae) and the red panda (Ailurus fulgens, Carnivora, Ailuridae), evolved diets specialized in bamboos and an adaptive pseudothumb (Hu et al., 2017). Signs of adaptive changes in the genes dync2h1 and pcnt, probably involved in the development of a pseudothumb, have been identified from ancestral reconstructions of protein sequences implemented using thousands of orthologs (Hu et al., 2017). From these analyses, the authors proposed two parallelisms in the dync2h1 gene: R3128K[Lysine] (in giant and red pandas) and K3999R (in giant and red pandas and also in the Weddell seals and walrus). Moreover, the analyses indicate a possible pseudogenization of the umami taste receptor gene tas1r1 in both panda lineages (Hu et al., 2017).

Finally, variation in the Hemoglobin-Oxygen affinity in birds provides a fourth example of ASR analyses applied to the study of recurrent phenotypes. The Hemoglobin-Oxygen affinity varies according to the atmospheric partial pressure, and animals with high levels of aerobic activity under hypoxic conditions often have optimizing blood-O₂ affinity (Projecto-Garcia et al., 2013). In South American hummingbirds, colonization of new elevation zones occurred in association with similar amino acid substitutions that changed the respiratory properties of hemoglobin (Projecto-Garcia et al., 2013). Ancestral reconstruction of such changes provide evidence for two parallel amino acid substitutions: G13S[Serine] and G83S (Projecto-Garcia et al., 2013).

Repeated trait loss: How ‘absence’ evolved multiple times, and why it challenges strict definitions of convergence and parallelism

Disjunctive expression of phenotypic traits is developmentally feasible, especially when trait expression is settled on switch-regulated mechanisms (see West-Eberhard, 2003). Repeated loss of specific phenotypic traits is very common in evolution, and defies strict definitions of convergence and parallelism because modifications in different components of a signaling pathway may silence developmental processes and result in the absence of that trait in a given lineage. Repeated loss is particularly likely if the structure being lost has some developmental and functional independence from other traits and, therefore, is less subjected to pleiotropic trade-offs (Womack et al., 2018). Phenotypic traits are established during development through intricate signaling pathways encompassing several genes that interact with each other. Accordingly, changes in either component of these signaling cascades might silence the developmental pathway, resulting in the absence of that given trait in the lineage. Given the strict definitions of convergence and parallelism (see Figures 2 and 3), one may ask how to classify changes settled on different components of a given developmental pathway (see West-Eberhard, 2003 for a review).

An emblematic example of repeated trait loss refers to the multiple origins of snakelike phenotypes in Tetrapoda. Snakelike phenotypes are characterized by elongated bodies and reduced or absent limbs, and entirely limbless species are observed in clades as distant as Lissamphibia and Lepidosauria (see Woltering, 2012). Several studies aimed to identify the genetic bases associated with limb loss in specific groups (e.g. Singarete et al., 2015; Guerreiro et al., 2016; Kvon et al., 2016; Leal and Cohn, 2016; Ovchinnikov et al., 2022; Roscito et al., 2022; also reviewed in Leal and Cohn, 2018), and comparisons among clades provide evidence that different changes in developmental pathways may independently produce the same phenotype characterized by absence of limbs. For example, molecular evolution analyses in three limbless lineages - snakes, amphisbaenians and caecilians - suggest five sites in the first exon of the gene Hoxa13 evolving under positive selection in snakes (Kohlsdorf et al., 2008), a pattern not identified for this gene in amphisbaenians and caecilians (Singarete et al., 2015). On the other hand, limb loss in snakes and caecilians also seems related to deletions in the ZRS limb-specific enhancer (Kvon et al., 2016; Ovchinnikov et al., 2022). This enhancer regulates the expression of sonic hedgehog in developing limbs, a gene that modulates the production of the SHH morphogen in the zone of polarizing activity (ZPA), playing a key role in the establishment of the anterior-posterior axis in developing limbs (Petit et al., 2017; Jin et al., 2019). Snakes that are completely limbless (i.e. without vestigial limbs) exhibit a 17-base deletion in ZRS and accelerated evolutionary rates in the sequence of this enhancer (Kvon et al., 2016). This deletion and the high evolutionary rates of the snake ZRS suggest impairment of the enhancer function with consequent relaxed selection, which was confirmed by experiments inserting the snake ZRS into mice that generated individuals with severe limb reduction (Kvon et al., 2016). In caecilians, the ZRS enhancer element seems to be entirely absent from the genomes sequenced so far (Ovchinnikov et al., 2022), suggesting a similar mechanism involved in limb loss in Lissamphibia. However, other limbless squamate species do not exhibit such deletion in the ZRS (Roscito et al., 2022), and the ZRS patterns differ even among closely-related lizard species that exhibit limb reduction and digit loss (Kohlsdorf, 2021), suggesting that the phenotype of absent limbs might also evolve through changes in other genes or cis-regulatory elements.

Another example of recurrent loss is observed in fossorial mammals that spend most of their lives under the surface. Adaptation of the subterranean lifestyle usually involves eye reduction or loss and impairment of the sense of sight (Partha et al., 2017). Recent studies identified accelerated evolutionary rates in genes and enhancers related to eyes in non-phylogenetically related subterranean lineages of moles and mole-rats (lens intrinsic membrane protein 2 [lim2] and retinal proteins: retinal outer segment membrane protein 1 [rom1] and rod cell-specific G protein, subunit alpha [gnat1]) suggesting an intricate mechanism associated to the loss of visual function in these animals (Partha et al., 2017).

These examples illustrate how repeated trait loss defies the identification of developmental changes underlying the absence of a given phenotypic feature in different lineages, especially in studies aiming to classify the associated genetic patterns as convergence or parallelism. Trait loss involves two complicating aspects for such studies: 1) part of the sequence variation associated to a silenced pathway may correspond to degeneration of the signaling cascade, instead of the mechanism ‘responsible’ for switching off the developmental process; 2) part of sequence conservation observed in a silenced pathway may indicate molecular stability due to pleiotropy. This discussion could be significantly expanded by novel studies considering complete signaling networks, instead of focusing on candidate genes, combined with conceptual discussions addressing the developmental processes underlying a disjunct expression of phenotypic traits along the phylogeny.

Research in the past decade produced a considerable number of studies addressing the processes and mechanisms related to the repeated evolution of similar phenotypes, which nurtured discussions about homoplasy and encouraged reexamination of key concepts, including convergence and parallelism. In this review, we use an integrated approach to discuss this topic, which consists of revisiting the classical definitions of convergence and parallelism, describing some comparative methods used to assess the evolution of repeated phenotypes, and examining how repeated trait loss challenges strict definitions of convergence and parallelism. To illustrate how different methodological approaches can be used to evaluate such evolutionary patterns, we provide examples of studies focusing on various lineages. A major goal of this review is to highlight the importance of combining modern analytical phylogenetic tools with knowledge about developmental pathways and regulatory mechanisms to completely understand the repeated evolution of similar phenotypes. Despite challenges for the study of developmental pathways in biological systems that are not experimental models, the growing number of genomes available and the proliferation of analytical tools designed to operate large amounts of data stimulate significant progress in the field. As the depth of knowledge increases, so does its ability to reveal the genetic and molecular mechanisms enabling recurrent evolution in biological lineages.

Acknowledgements

This study was supported by a FAPESP-Brazil fellowship to AGP (2019/21712-5) and a FAPESP-Brazil Thematic Grant awarded to TK (2020/14780-1), and a CNPq productivity fellowship awarded to TK (304170/2022-4).

References

• Ané C, Larget B, Baum DA, Smith SD and Rokas A (2007) Bayesian estimation of concordance among gene trees. Mol Biol Evol 24:412-426.
• Ansari MA and Didelot X (2016) Bayesian inference of the evolution of a phenotype distribution on a phylogenetic tree. Genetics 204:89-98.
• Arendt J and Reznick D (2008) Convergence and parallelism reconsidered: What have we learned about the genetics of adaptation?. Trends Ecol Evol 23:26-32.
• Avise JC and Robinson TJ (2008) Hemiplasy: A new term in the lexicon of phylogenetics. Syst Biol 57:503-507.
• Babu MM, Luscombe NM, Aravind L, Gerstein M and Teichmann SA (2004) Structure and evolution of transcriptional regulatory networks. Curr Opin Struct Biol 14:283-291.
• ​​Baum DA and Smith SD (2013) Tree thinking: An Introduction to Phylogenetic Biology. Roberts and Company Publishers, Colorado.
• Bielawski JP and Yang Z (2004) A maximum likelihood method for detecting functional divergence at individual codon sites, with application to gene family evolution. J Mol Evol 59:121-132.
• Blomberg SP, Garland T and Ives AR (2003) Testing for phylogenetic signals in comparative data: Behavioral traits are more labile. Evolution 57:717-745.
• Booker BM, Friedrich T, Mason MK, VanderMeer JE, Zhao J, Eckalbar WL, Logan M, Illing N, Pollard KS and Ahituv N (2016) Bat accelerated regions identify a bat forelimb specific enhancer in the HoxD locus. PLoS Genet 12:e1005738.
• Bouckaert R, Heled J, Kühnert D, Vaughan T, Wu C H, Xie D, Suchard MA, Rambaut A and Drummond AJ (2014) BEAST 2: A software platform for Bayesian evolutionary analysis. PLoS Comput Biol 10:e1003537.
• Bulger M and Groudine M (2011) Functional and mechanistic diversity of distal transcription enhancers. Cell 144:327-339.
• Carapuço M, Nóvoa A, Bobola N and Mallo M (2005) Hox genes specify vertebral types in the presomitic mesoderm. Genes Dev 19:2116-2121.
• Carroll SB (1995) Homeotic genes and the evolution of arthropods and chordates. Nature 376:479-485.
• Chikina M, Robinson JD and Clark NL (2016) Hundreds of genes experienced convergent shifts in selective pressure in marine mammals. Mol Biol Evol 33:2182-2192.
• Conte GL, Arnegard ME, Peichel CL and Schluter D (2012) The probability of genetic parallelism and convergence in natural populations. Proc Biol Sci 279:5039-5047.
• Davies KTJ, Cotton JA, Kirwan JD, Teeling EC and Rossiter SJ (2012) Parallel signatures of sequence evolution among hearing genes in echolocating mammals: An emerging model of genetic convergence. Heredity (Edinb) 108:480-489.
• Davis PH and Heywood VH (1963) Principles of Angiosperm Taxonomy. Oliver and Boyd, Edinburgh.
• Degnan JH and Rosenberg NA (2006) Discordance of species trees with their most likely gene trees. PLoS Genet 2:e68.
• Degnan JH and Rosenberg NA (2009) Gene tree discordance, phylogenetic inference and the multispecies coalescent. Trends Ecol Evol 24:332-340.
• Drummond AJ and Rambaut A (2007) BEAST: Bayesian evolutionary analysis by sampling trees. BMC Evol Biol 7:214.
• Edwards SV (2009) Is a new and general theory of molecular systematics emerging?. Evolution 63:1-19.
• Feigin CY, Newton AH and Pask AJ (2019) Widespread cis-regulatory convergence between the extinct Tasmanian tiger and gray wolf. Genome Res 29:1648-1658.
• Felsenstein J (1981) Evolutionary trees from DNA sequences: A maximum likelihood approach. J Mol Evol 17:368-376.
• Felsenstein J (1985) Confidence limits on phylogenies: An approach using the Bootstrap. Evolution 39:783-791.
• Fish FE and Hui CA (1991) Dolphin swimming - a review. Mamm Rev 21:181-195.
• Fish FE, Howle LE and Murray MM (2008) Hydrodynamic flow control in marine mammals. Integr Comp Biol 48:788-800.
• Fitch WM (1971) Toward defining course of evolution: Minimum change for a specific tree topology. Syst Biol 20:406-416.
• Gao F, Chen C, Arab DA, Du Z, He Y and Ho SY (2019) EasyCodeML: A visual tool for analysis of selection using CodeML. Ecol Evol 9:3891-3898.
• Garland T, Bennett AF and Rezende EL (2005) Phylogenetic approaches in comparative physiology. J Exp Biol 208:3015-3035.
• Gharib WH and Robinson-Rechavi M (2013) The branch-site test of positive selection is surprisingly robust but lacks power under synonymous substitution saturation and variation in GC. Mol Biol Evol 30:1675-1686.
• Gompel N and Prud’homme B (2009) The causes of repeated genetic evolution. Dev Biol 332:36-47.
• Guerreiro I, Nunes A, Woltering JM, Casaca A, Nóvoa A, Vinagre T, Hunter ME, Duboule D and Mallo M (2013) Role of a polymorphism in a Hox/Pax-responsive enhancer in the evolution of the vertebrate spine. Proc Natl Acad Sci U S A 110:10682-10686.
• Guerreiro I, Gitto S, Novoa A, Codourey J, Huynh TH, Gonzalez F, Milinkovitch MC, Mallo M and Duboule D (2016) Reorganisation of Hoxd regulatory landscapes during the evolution of a snake-like body plan. Elife 5:e16087.
• Guerrero RF and Hahn MW (2018) Quantifying the risk of hemiplasy in phylogenetic inference. Proc Natl Acad Sci U S A 115:12787-92.
• Gumulya Y and Gillam EM (2017) Exploring the past and the future of protein evolution with ancestral sequence reconstruction: The ‘retro’ approach to protein engineering. Biochem J 474:1-19.
• Halabi K, Levy Karin E, Guéguen L and Mayrose I (2020) TraitRELAX - A codon model for associating phenotypic traits with altered selective patterns of sequence evolution. Syst Biol 70:608-622.
• Hiller M, Schaar BT, Indjeian VB, Kingsley DM, Hagey LR and Bejerano G (2012) A “forward genomics” approach links genotype to phenotype using independent phenotypic losses among related species. Cell Rep 2:817-823.
• Hu Y, Wu QI, Ma S, Ma T, Shan L, Wang X, Nie Y, Ning Z, Yan L, Xiu Y et al (2017) Comparative genomics reveals convergent evolution between the bamboo-eating giant and red pandas. Proc Natl Acad Sci U S A 114:1081-1086.
• Hu Z, Sackton TB, Edwards SV and Liu JS (2019) Bayesian detection of convergent rate changes of conserved noncoding elements on phylogenetic trees. Mol Biol Evol 36:1086-1100.
• Huerta-Cepas J, Serra F and Bork P (2016) ETE 3: Reconstruction, analysis, visualization of phylogenomic data. Mol Biol Evol 33:1635-1638.
• Hunt DM, Dulai KS, Cowing JA, Julliot C, Mollon JD, Bowmaker JK, Li WH and Hewett-Emmett D (1998) Molecular evolution of trichromacy in primates. Vision Res 38:3299-3306.
• Jin L, Wu J, Bellusci S and Zhang JS (2019) Fibroblast growth factor 10 and vertebrate limb development. Front Genet 9:705.
• Kohlsdorf T (2021) Reversibility of digit loss revisited: Limb diversification in Bachia lizards (Gymnophthalmidae). J Exp Zool B Mol Dev Evol. DOI: 10.1002/jez.b.23024.
  » https://doi.org/10.1002/jez.b.23024.
• Kohlsdorf T, Cummings MP, Lynch VJ, Stopper GF, Takahashi K and Wagner GP (2008) A molecular footprint of limb loss: Sequence variation of the autopodial identity gene Hoxa-13. J Mol Evol 67:581-593.
• Koshi JM and Goldstein RA (1996) Probabilistic reconstruction of ancestral protein sequences. J Mol Evol 42:313-320.
• Kowalczyk A, Meyer WK, Partha R, Mao W, Clark NL and Chikina M (2019) RERconverge: An R package for associating evolutionary rates with convergent traits. Bioinformatics 35:4815-4817.
• Kvon EZ, Kamneva OK, Melo US, Barozzi I, Osterwalder M, Mannion BJ, Tissières V, Pickle CS, Plajzer-Frick I, Lee EA et al (2016) Progressive loss of function in a limb enhancer during snake evolution. Cell 167:633-642.
• Lamichhaney S, Card DC, Grayson P, Tonini JF, Bravo GA, Näpflin K, Termignoni-Garcia F, Torres C, Burbrink F, Clarke JA et al (2019) Integrating natural history collections and comparative genomics to study the genetic architecture of convergent evolution. Philos Trans R Soc Lond B Biol Sci 374:20180248.
• Lartillot N and Poujol R (2011) A phylogenetic model for investigating correlated evolution of substitution rates and continuous phenotypic characters. Mol Biol Evol 28:729-744.
• Leal F and Cohn MJ (2016) Loss and re-emergence of legs in snakes by modular evolution of Sonic hedgehog and HOXD enhancers. Curr Biol 26:2966-2973.
• Leal F and Cohn MJ (2018) Developmental, genetic, and genomic insights into the evolutionary loss of limbs in snakes. Genesis 56:e23077.
• Levy KE, Wicke S, Pupko T and Mayrose I (2017) An integrated model of phenotypic trait changes and site-specific sequence evolution. Syst Biol 66:917-933.
• Li WH (1993) Unbiased estimation of the rates of synonymous and nonsynonymous substitution. J Mol Evol 36:96-99.
• Liu Z, Qi FY, Zhou X, Ren HQ and Shi P (2014) Parallel sites implicate functional convergence of the hearing gene prestin among echolocating mammals. Mol Biol Evol 31:2415-2424
• Liu Z, Chen P, Xu DM, Qi FY, Guo YT, Liu Q, Bai J, Zhou X and Shi P (2022) Molecular convergence and transgenic evidence suggest a single origin of laryngeal echolocation in bats. iScience 25:104114.
• Lobo I (2008) Pleiotropy: One gene can affect multiple traits. Nature Education 1:10
• Maddison WP (1997) Gene trees in species trees. Syst Biol 46:523-536.
• Massingham T and Goldman N (2005) Detecting amino acid sites under positive selection and purifying selection. Genetics 169:1753-1762.
• Mau B, Newton MA and Larget B (1999) Bayesian phylogenetic inference via Markov chain Monte Carlo methods. Biometrics 55:1-12.
• McIntyre DC, Rakshit S, Yallowitz AR, Loken L, Jeannotte L, Capecchi MR and Wellik DM (2007) Hox patterning of the vertebrate rib cage. Development 134:2981-2989.
• Meddens CA, Van Der List AC, Nieuwenhuis EE and Mokry M (2019) Non-coding DNA in IBD: From sequence variation in DNA regulatory elements to novel therapeutic potential. Gut 68:928-941.
• Mendes FK, Hahn Y and Hahn MW (2016) Gene tree discordance can generate patterns of diminishing convergence over time. Mol Biol Evol 33:3299-3307.
• Merkl R and Sterner R (2016) Ancestral protein reconstruction: Techniques and applications. Biol Chem 397:1-21.
• Minh BQ, Hahn MW and Lanfear R (2020a) New methods to calculate concordance factors for phylogenomic datasets. Mol Biol Evol 37(9):2727-2733.
• Minh BQ, Schmidt HA, Chernomor O, Schrempf D, Woodhams MD, Von Haeseler A and Lanfear R (2020b) IQ-TREE 2: New models and efficient methods for phylogenetic inference in the genomic era. Mol Biol Evol 37:1530-1534.
• Mo YK, Lanfear R, Hahn MW and Minh BQ (2023) Updated site concordance factors minimize effects of homoplasy and taxon sampling. Bioinformatics 39:btac741.
• Mohammadi S, Gompert Z, Gonzalez J, Takeuchi H, Mori A and Savitzky AH (2016) Toxin-resistant isoforms of Na+/K+-ATPase in snakes do not closely track dietary specialization on toads. Proc Biol Sci 283:20162111.
• Monteiro A and Podlaha O (2009) Wings, horns, and butterfly eyespots: how do complex traits evolve?. PLoS Biol 7:e37
• Morris J, Hanly JJ, Martin SH, Van Belleghem SM, Salazar C, Jiggins CD and Dasmahapatra KK (2020) Deep convergence, shared ancestry, and evolutionary novelty in the genetic architecture of Heliconius mimicry. Genetics 216:765-780.
• Münkemüller T, Lavergne S, Bzeznik B, Dray S, Jombart T, Schiffers K and Thuiller W (2012) How to measure and test phylogenetic signal. Methods Ecol Evol 3:743-756.
• Murrell B, Wertheim JO, Moola S, Weighill T, Scheffler K and Pond SLK (2012) Detecting individual sites subject to episodic diversifying selection. PLoS Genet 8:e1002764.
• Murrell B, Moola S, Mabona A, Weighill T, Sheward D, Pond SLK and Scheffler K (2013) FUBAR: A fast, unconstrained Bayesian approximation for inferring selection. Mol Biol Evol 30:1196-1205.
• Murrell B, Weaver S, Smith MD, Wertheim JO, Murrell S, Aylward A, Eren K, Pollner T, Martin DP, Smith DM et al (2015) Gene-wide identification of episodic selection. Mol Biol Evol 32:1365-1371.
• Nei M and Gojobori T (1986) Simple methods for estimating the numbers of synonymous and nonsynonymous nucleotide substitutions. Mol Biol Evol 3:418-426.
• Nery MF, Borges B, Dragalzew AC and Kohlsdorf T (2016) Selection on different genes with equivalent functions: The convergence story told by Hox genes along the evolution of aquatic mammalian lineages. BMC Evol Biol 16:113.
• Nguyen LT, Schmidt HA, Von Haeseler A and Minh BQ (2015) IQ-TREE: A fast and effective stochastic algorithm for estimating maximum-likelihood phylogenies. Mol Biol Evol 32:268-274.
• Orgogozo V (2015) Replaying the tape of life in the twenty-first century. Interface Focus 5:20150057.
• Orr HA (2005) The probability of parallel evolution. Evolution 59:216-220.
• Ovchinnikov V, Uliano-Silva M, Wilkinson M, Wood J, Smith M, Oliver K, Sims Y, Torrance J, Suh A, McCarthy A et al (2022) Caecilian genomes reveal molecular basis of adaptation and convergent evolution of limblessness in snakes and caecilians. bioRxiv. DOI: 10.1101/2022.02.23.481419.
  » https://doi.org/10.1101/2022.02.23.481419.
• Partha R, Chauhan BK, Ferreira Z, Robinson JD, Lathrop K, Nischal KK, Chikina M and Clark NL (2017) Subterranean mammals show convergent regression in ocular genes and enhancers, along with adaptation to tunneling. Elife 6:e25884.
• Pereira AG, Grizante MB and Kohlsdorf T (2022) What snakes and caecilians have in common? Molecular interaction units and the independent origins of similar morphotypes in Tetrapoda. Proc Biol Sci 289:20220841.
• Petit F, Sears KE and Ahituv N (2017) Limb development: A paradigm of gene regulation. Nat Rev Genet 18:245-258.
• Pond SLK and Frost SDW (2005) Not so different after all: A comparison of methods detecting amino acid sites under selection. Mol Biol Evol 22:1208-1222.
• Projecto-Garcia J, Natarajan C, Moriyama H, Weber RE, Fago A, Cheviron ZA, Dudley R, McGuire JA, Witt CC and Storz JF (2013) Repeated elevational transitions in hemoglobin function during the evolution of Andean hummingbirds. Proc Natl Acad Sci U S A 110:20669-20674.
• Prud’homme B, Gompel N, Rokas A, Kassner VA, Williams TM, Yeh S-D, True JR and Carroll SB (2006) Repeated morphological evolution through cis-regulatory changes in a pleiotropic gene. Nature 440:1050-1053.
• Prudent X, Parra G, Schwede P, Roscito JG and Hiller M (2016) Controlling for phylogenetic relatedness and evolutionary rates improves the discovery of associations between species’ phenotypic and genomic differences. Mol Biol Evol 33:2135-2150.
• Rannala B and Yang Z (1996) Probability distribution of molecular evolutionary trees: A new method of phylogenetic inference. J Mol Evol 43:304-311.
• Rannala B, Edwards SV, Leaché A and Yang Z (2020) The multispecies coalescent model and species tree inference. In: Scornavacca C, Delsuc F and Galtier N (eds) Phylogenetics in the Genomic Era. Self Published, pp 3.3:1-3.3:21.
• Revell LJ (2012) phytools: An R package for phylogenetic comparative biology (and other things). Methods Ecol Evol 3:217-223.
• Ronquist F and Huelsenbeck JP (2003) Mrbayes 3: Bayesian phylogenetic inference under mixed models. Bioinformatics 19:1572-1574.
• Roscito JG, Sameith K, Kirilenko BM, Hecker N, Winkler S, Dahl A, Rodrigues MT and Hiller M (2022) Convergent and lineage-specific genomic differences in limb regulatory elements in limbless reptile lineages. Cell Rep 38:110280.
• Rosenblum EB, Parent CE and Brandt EE (2014) The molecular basis of phenotypic convergence. Annu Rev Ecol Evol Syst 45:203-26.
• Rosenblum EB, Römpler H, Schöneberg T and Hoekstra HE (2010) Molecular and functional basis of phenotypic convergence in white lizards at White Sands. Proc Natl Acad Sci U S A 107:2113-2117.
• Sackton TB, Grayson P, Cloutier A, Hu Z, Liu JS, Wheeler NE, Gardner PP, Clarke JA, Baker AJ, Clamp M et al (2019) Convergent regulatory evolution and loss of flight in paleognathous birds. Science 364:74-78.
• Schluter D, Clifford EA, Nemethy M and McKinnon JS (2004) Parallel evolution and inheritance of quantitative traits. Am Nat 163:809-822
• Scotland RW (2011) What is parallelism? Evol Dev 13:214-227.
• Shen YY, Liang L, Li GS, Murphy RW and Zhang YP (2012) Parallel evolution of auditory genes for echolocation in bats and toothed whales. PLoS Genet 8:e1002788.
• Shi Y and Yokoyama S (2003) Molecular analysis of the evolutionary significance of ultraviolet vision in vertebrates. Proc Natl Acad Sci U S A 100:8308-8313.
• Shimodaira H and Hasegawa M (2001) CONSEL: For assessing the confidence of phylogenetic tree selection. Bioinformatics 17:1246-1247.
• Singarete ME, Grizante MB, Milograna SR, Nery MF, Kin K, Wagner GP and Kohlsdorf T (2015) Molecular evolution of HoxA13 and the multiple origins of limbless morphologies in amphibians and reptiles. Genet Mol Biol 38:255-262.
• Smith MD, Wertheim JO, Weaver S, Murrell B, Scheffler K and Pond SLK (2015) Less is more: An adaptive branch-site random effects model for efficient detection of episodic diversifying selection. Mol Biol Evol 32:1342-1353.
• Speed MP and Arbuckle K (2017) Quantification provides a conceptual basis for convergent evolution. Biol Rev Camb Philos Soc 92:815-829.
• Stamatakis A (2006) RAxML-VI-HPC: Maximum likelihood-based phylogenetic analyses with thousands of taxa and mixed models. Bioinformatics 22:2688-2690.
• Storz J (2016) Causes of molecular convergence and parallelism in protein evolution. Nat Rev Genet 17:239-50.
• Thomas GW and Hahn MW (2015) Determining the null model for detecting adaptive convergence from genomic data: A case study using echolocating mammals. Mol Biol Evol 32:1232-1236.
• Thornton JW (2004) Resurrecting ancient genes: Experimental analysis of extinct molecules. Nat Rev Genet 5:366-375.
• Tollis M, Hutchins ED, Stapley J, Rupp SM, Eckalbar WL, Maayan I, Lasku E, Infante CR, Dennis SR, Robertson JA et al (2018) Comparative genomics reveals accelerated evolution in conserved pathways during the diversification of anole lizards. Genome Biol Evol 10:489-506.
• Uhen MD (2007) Evolution of marine mammals: Back to the sea after 300 million years. Anat Rec (Hoboken) 290:514-522.
• Vater M and Kössl M (2004) The ears of whales and bats. In: Thomas JA, Moss CF and Vater M (eds) Echolocation in Bats and Dolphins. The University of Chicago Press, Chicago, pp 89-99.
• Voordeckers K, Pougach K and Verstrepen KJ (2015) How do regulatory networks evolve and expand throughout evolution?. Curr Opin Biotechnol 34:180-188.
• Wagner GP and Lynch VJ (2008) The gene regulatory logic of transcription factor evolution. Trends Ecol Evol 23:377-385.
• ​​Wake DB (1991) Homoplasy: The result of natural selection, or evidence of design limitations?. Am Nat 138:543-567.
• Wake DB, Wake MH and Specht CD (2011) Homoplasy: From detecting pattern to determining process and mechanism of evolution. Science 331:1032-1035.
• Wellik DM and Capecchi MR (2003) Hox10 and Hox11 genes are required to globally pattern the mammalian skeleton. Science 301:363-367.
• Wertheim JO, Murrell B, Smith MD, Pond SLK and Scheffler K (2015) RELAX: Detecting relaxed selection in a phylogenetic framework. Mol Biol Evol 32:820-832.
• West-Eberhard MJ (2003) Developmental plasticity and evolution. Oxford University Press, New York.
• Woltering JM (2012) From lizard to snake; behind the evolution of an extreme body plan. Curr Genomics 13:289-299.
• Womack MC, Fiero TS and Hoke KL (2018) Trait independence primes convergent trait loss. Evolution 72:679-687.
• Xu H, Liu Y, He G, Rossiter SJ and Zhang S (2013) Adaptive evolution of tight junction protein claudin-14 in echolocating whales. Gene 530:208-14.
• Yan H, Hu Z, Thomas G, Edwards SV, Sackton TB and Liu JS (2022) PhyloAcc-GT: A Bayesian method for inferring patterns of substitution rate shifts and associations with binary traits under gene tree discordance. bioRxiv. DOI: 10.1101/2022.12.23.521765.
  » https://doi.org/10.1101/2022.12.23.521765.
• Yang Z (1998) Likelihood ratio tests for detecting positive selection and application to primate lysozyme evolution. Mol Biol Evol 15:568-573.
• Yang Z (2002) Inference of selection from multiple species alignments. Curr Opin Genet Dev 12:688-694.
• Yang Z (2007) PAML 4: Phylogenetic analysis by maximum likelihood. Mol Biol Evol 24:1586-1591.
• Yang Z and Nielsen R (2000) Estimating synonymous and nonsynonymous substitution rates under realistic evolutionary models. Mol Biol Evol 17:32-43.
• Yang Z and Nielsen R (2002) Codon-substitution models for detecting molecular adaptation at individual sites along specific lineages. Mol Biol Evol 19:908-917.
• Yang Z, Kumar S and Nei M (1995) A new method of inference of ancestral nucleotide and amino-acid-sequences. Genetics 141:1641-1650
• Yang Z, Nielsen R, Goldman N and Pedersen AK (2000) Codon-substitution models for heterogeneous selection pressure at amino acid sites. Genetics 155:431-449.
• Yeaman S, Gerstein AC, Hodgins KA and Whitlock MC (2018) Quantifying how constraints limit the diversity of viable routes to adaptation. PLoS Genet 14:e1007717.
• Zhang J, Nielsen R and Yang Z (2005) Evaluation of an improved branch-site likelihood method for detecting positive selection at the molecular level. Mol Biol Evol 22:2472-2479.

• License information:

  This is an open-access article distributed under the terms of the Creative Commons Attribution License (type CC-BY), which permits unrestricted use, distribution and reproduction in any medium, provided the original article is properly cited.

Edited by

Associate Editor:

Carlos F. M. Menck

Publication Dates

• Publication in this collection
  21 July 2023
• Date of issue
  2023

History

• Received
  28 Dec 2022
• Accepted
  24 May 2023