Evolutionary bedfellows: reconstructing the ancestral state of autotomy and regeneration

Abstract

Some form of regeneration occurs in all lifeforms and extends from single-cell organisms to humans. The degree to which regenerative ability is distributed across different taxa, however, is harder to ascertain given the potential for phylogenetic constraint or inertia, and adaptive processes to shape this pattern. Here, we examine the phylogenetic history of regeneration in two groups where the trait has been well-studied: arthropods and reptiles. Because autotomy is often present alongside regeneration in these groups, we performed ancestral state reconstructions for both traits to more precisely assess the timing of their origins and the degree to which these traits coevolve. Using an ancestral trait reconstruction, we find that autotomy and regeneration were present at the base of the arthropod and reptile trees. We also find that when autotomy is lost it does not re-evolve easily. Lastly, we find that the distribution of regeneration is intimately connected to autotomy with the association being stronger in reptiles than in arthropods. While these patterns suggest that decoupling autotomy and regeneration at a broad phylogenetic scale may be difficult, the available data provides useful insight into their entanglement. Ultimately, our reconstructions provide important groundwork to explore how selection may have played a role during the loss of regeneration in specific lineages.

Graphical Abstract

Introduction

Long before On the Origin of Species (Darwin, 1859) introduced the world to the theory of natural selection, naturalists were busy exploring regenerative phenomena in animals and plants (Réaumur, 1712; Lenhoff and Lenhoff, 1986; Dinsmore, 1996; Hijmans and Elith, 2016). As Darwin’s ideas took root among different scientists, evolutionary thinking came to influence regenerative biologists who began offering explanations for the persistence or loss of regenerative ability across the tree of life (Cravens and Allen, 1980). In parallel with the influence of evolutionary thinking, a shifting trend in embryological research catalyzed a more rigorous experimental approach to the study of embryonic development and regeneration. At its height, the dichotomy between naturalists and experimentalists embodied a key argument concerning the distribution of regenerative abilities among animals. The argument centered on whether regeneration was an adaptive trait or instead, an irreducible and fundamental property of life. During this period, August Weismann and Thomas Hunt Morgan embodied opposing viewpoints in this debate: “…whereas for Weismann, animal regeneration was a contingent, naturally selected adaptation, Morgan insisted that animal regeneration was neither contingent nor a selected adaptation but rather a general property of the organism as a whole: a universal organic quality characterizing any organism occurring in different degrees” (page 516, Esposito 2013).

Evidence in favor of tissue regeneration as a fundamental organismal property comes from the fact that on its surface, regeneration has close links to embryogenesis (reviewed in Sánchez Alvarado, 2000) and asexual reproduction (reviewed in Bely and Nyberg, 2010); these links suggest regeneration has deep evolutionary origins dating back to the advent of multicellularity or even earlier. Given that pattern restoration extends to the scale of single-cell organisms and their internal structures (Needham, 1953a; Tartar, 2011), regeneration may simply be the expression or augmentation of a more basic process of pattern formation and elaboration. This idea rests with regeneration predating the evolution of the embryo where basic cellular-level regenerative processes are amplified in postembryonic animals as tissue (multi-cell) regeneration. Regardless of whether regenerative processes originated before or after the evolution of multicellular organisms, it likely emerged early in the evolution of complex life on earth. Thus, the distribution of regenerative ability in different lineages is likely shaped by phylogenetic constraint or inertia where the physical and genetic mechanisms underlying early regenerative processes constrained the evolution of later regenerative mechanisms. Arguments for regeneration as an adaptive trait rely on fitness advantages gained by regenerating a whole organism or a part thereof. For example, if a piece of a planaria or Hydra is severed, regrowth via regeneration can be so substantial that it can create new progeny where the parent organism regenerates and the severed piece regenerates into a daughter organism. Agonistic interactions with predators or with conspecifics can cause tissue damage which regeneration can serve to mitigate, thus minimizing fitness declines (Morgan, 1901a; Maginnis, 2006; Slack, 2010). In the context of appendages, the loss of limbs or tails is so common, and minimizing the damage so important, that in some species these tissues are preferentially lost through autotomy. During autotomy, the structure is removed at a weak point or breakage plane allowing a more limited wound which facilitates regeneration of the structure. In the absence of direct testing, there is a potential role for both selection and phylogenetic inertia in the evolution of regeneration.

How can we determine the relative strength of selection and constraint in the evolution of regeneration? One tool is the comparative method from evolutionary biology where we look for patterns that emerge from analyzing the presence and extent of tissue regeneration among different taxa across the tree of life. When considering multicellular animals, regenerative ability varies substantially (Needham, 1953a; Bely and Nyberg, 2010). For instance, some cnidarians show near limitless regenerative ability with the power to regenerate entire individuals from dissociated cells (Goss, 1969). This stands in contrast to animals that can only regenerate specific structures, such as the tail but not limbs in lizards (Vitt et al., 1977), and further still to animals like mites that appear to lack any appreciable regenerative capacity whatsoever (Maruzzo et al., 2005). This variation suggests that either the adaptive value of regeneration varies across taxa, resulting in its selective maintenance in some taxa and selective elimination in others, or the developmental and pattern formation processes that produce regeneration in some taxa evolve complexities that lead to its elimination in other groups. Thus, distinguishing the relative roles of adaptation and constraint remains a crucial task in understanding the evolution of regeneration (Sánchez Alvarado, 2000; Bely and Nyberg, 2010; Tiozzo and Copley, 2015).

Here we attempt to address this issue by looking at the phylogenetic history of regeneration and its co-occurrence with autotomy. Studying autotomy is crucial for understanding if (or how) the fitness costs and benefits of regeneration vary across different species because it is necessary to separate the potential costs and benefits of regeneration from the costs of autotomy that may precede regeneration in many species. Numerous arthropods are capable of autotomizing their limbs following damage or predation attempts (Fleming et al., 2007). Since the growth of many arthropods is tied to molting, many of them develop a regenerated limb in a papilla sac that later unfolds after molting (e.g., fiddler crab, Hopkins 1993; cockroach, Truby 1983). Moreover, autotomy can be beneficial in arthropods since molting can lead to appendages becoming stuck in the exuviate (i.e., shed exoskeleton) and autotomy ensures that the exoskeleton is shed even in these cases. Reptiles are another group of organisms heavily studied for their regenerative capacities and a group in which autotomy is widespread (Bellairs and Bryant, 1985). Notably, lizards are well-known to autotomize their tail when attacked by predators and a number of scincid and gekkonid lizards can shed parts of their integument in response to physical contact (Bauer et al., 1989; Zani, 1996; Scherz et al., 2017). Although lizards can regenerate their autotomized tail, the replacement is not a perfect facsimile since the missing vertebrae and spinal cord are replaced by a cartilaginous rod and ependymal tube respectively (Simpson, 1968; Bateman and Fleming, 2009; Gilbert et al., 2013; Lozito and Tuan, 2016). While not perfect, autotomy followed by regeneration would seem to offer an advantage as tail autotomy alone causes diminished locomotor performances (McElroy and Bergmann, 2013). However, other reptiles, such as chameleons, are neither capable of tail autotomy nor tail regeneration, which highlights the phylogenetic variation in both of these traits (Bellairs and Bryant, 1985; Anderson and Higham, 2014).

Here we use this variation in regenerative ability and autotomy to reconstruct the ancestral states of these two traits across arthropods and reptiles. Knowing these ancestral states is crucial for determining the evolutionary timing for the origin of autotomy relative to regeneration, and this timing in turn provides evidence for the value of regeneration in rebuilding structures lost by autotomy. For example, if autotomy evolved first, this could have set the stage for the adaptive evolution of regeneration. On the other hand, if regeneration evolved first, this would suggest the function of regenerative abilities must be broader than simply appendage replacement. In fact, our results show that autotomy and regeneration were present at the base of the arthropod and reptile trees. While this result suggests that untangling the relative role of adaptation and constraint in the evolution of tissue regeneration will require a more sophisticated approach than the one used here, our reconstructions provide important groundwork to explore how selection may play a role in the loss of regeneration in particular lineages.

Methods

Definitions

Regeneration

The term regeneration refers to the functional replacement of tissue that is lost or damaged. Importantly, we consider reparative regeneration in this paper as it occurs in response to injury or autotomy, either self-induced or caused by an external agent. More precisely, we refer to regeneration as the restoration of a pattern that occurs when the organism’s body system is disrupted. This definition separates regeneration from scar formation that occurs in some animals (e.g., replacement of lost tissue with dissimilar, non-functional tissue). This definition also applies across multiple levels of biological organization and unites regeneration in single-cell organisms with the higher-level combination of processes that occurs during replacement of an appendage. As a singular event, regeneration encompasses individual processes at different levels of biological organization. For example, barrier restoration, homeostasis, and morphogenesis broadly describe regeneration from beginning to end. Thus, for the purposes of identifying regeneration as a trait, this necessitates considering these processes together.

Autotomy

Autotomy has been broadly defined as the loss of tissue induced directly by an individual itself (as a reflex severance – autotomy – or by pulling on the appendage – autotilly) or indirectly by the action of outside agent (autospasy) (Wood and Wood 1932). In this paper we refer to autotomy broadly, including all of the aforementioned definitions. Autotomy can occur voluntarily or as the result of extensive damage to a limb or appendage. In either case, an autotomy plane prevents hemolymph or blood loss and facilitates regeneration of the missing appendage (Bellairs and Bryant, 1985; Maginnis, 2006; Fleming et al., 2007). Following autotomy, tissue either regenerates or heals with a scar. Regeneration in arthropods and reptiles generally proceeds via hemostasis, inflammation, re-epithelialization, blastema formation, and morphogenesis (Seifert et al., 2012).

Ancestral State Reconstruction

Ancestral state or trait reconstruction in a phylogenetic context was proposed as early as 1938 (Dobzhansky and Sturtevant, 1938) and is an approach that improves on similar methodologies appropriated from cladistics. What distinguishes ancestral state reconstruction is how it considers shared evolutionary history between taxa. Character mapping traits on a cladogram using parsimony to manually reconstruct ancestral states ignores shared evolutionary history and favors a mild view of a traits’ history because it ignores the effect of branch lengths in shaping the topology of the tree. Formal ancestral state reconstruction uses a more standardized approach in that it applies a hypothetical model of evolution to a phylogeny and in doing so takes into account evolutionary time.

Character Coding

In our data gathering, we code each experimental or observational report of bonafide regeneration, or its absence, with a value of 1 or 0, respectively; autotomy was similarly recorded. In the absence of experimental or observational information explicitly specifying trait presence of absence, we attributed a value of NA. If genera or species within a higher taxon level scored differently (e.g., some regenerate, others do not) then a branch tip on the tree would receive a 1/0 as was the case for Pterygota (see Figure 1). Although some authors use regeneration in referring to embryos, for the purposes of our reconstructions we only considered regeneration for post-embryonic stages of development. Most embryos can regenerate parts and tissue regeneration in embryos can be influenced by the degree of autonomous or conditional development present in a given taxa. Because life stage, tissue type, and organismal traits can influence regenerative ability (Seifert et al., 2012b), we considered data for presence or absence of autotomy and regeneration when it was reported for (1) any post-embryonic life stage and (2) for any appendage or body part. An exception to these criteria applied to the Pterygota where we only considered regeneration if reported for adults due to the role of regenerative ability in early development for these insects, which undergo partial or complete metamorphosis. While regenerative ability can vary across life-stage, it should be noted that the majority of data we found scored the presence or absence of these traits in adults. Even when life stages clearly separate regenerative capacities in some clades (as in holometabolous insects), we decided to embrace this fact and explore the differences between those groups and other closely related groups (see below). Since many species within a clade lack direct information on whether they autotomize or regenerate, we assigned presence or absence values for autotomy and regeneration to a tip of a clade even when only one or two species in the clade had been investigated. A potential bias in our dataset is that studies are more likely to report presence rather than absence of regeneration and autotomy. This particular limitation cannot be avoided until more research is conducted as advocated for below.

Figure 1:

Evidence of regeneration and autotomy manually mapped onto an arthropod phylogeny (Regier et al., 2010). Common names of organisms at each branch tip are on the right-hand side for convenience (see Table 1 for data). Red and blue boxes at the tips of the tree indicate presence or absence, respectively, for regeneration. Red and blue branches on the phylogeny indicate the presence or absence, respectively, of autotomy. Dashed branches indicate data are not available (NA).

In surveying the existing literature for robust information about autotomy and regeneration in arthropods, we relied on comprehensive reviews for arthropods (Maruzzo et al., 2005), invertebrates (Fleming et al., 2007), and extensive compendiums of these traits (Morgan, 1901b; McVean, 1982). Similarly, in order to determine the ancestral state for autotomy and regeneration in squamate reptiles, we surveyed the existing literature and relied heavily on a monograph that compiled information on these traits (Bellairs and Bryant, 1985) which we supplemented with species-level studies when possible (see below).

Phylogenies

In order to map both autotomy and regeneration in arthropods, we selected one of the most recent arthropod phylogenies (Regier et al., 2010). Although this phylogeny provides the most complete coverage of arthropod families and is congruent with recent discussions of arthropod relationships, it is not a chronogram whose branch lengths are proportional to time, which are preferred when reconstructing ancestral states using maximum likelihood (Litsios and Salamin, 2012). Thus, for the ancestral state reconstruction, we used the most recent arthropod phylogeny with calibrated divergence times (Rehm et al., 2011). Ultimately, this phylogeny was trimmed to include only species for which we had regeneration and autotomy data from the literature search before running the ancestral state reconstruction. However, this new tree did not include any species where regeneration was absent rendering it impossible for us to run a formal ancestral state reconstruction for regeneration. Nevertheless, this analysis might be unnecessary in view of the lack of variation in regeneration capacities across the arthropod tree.

For squamate reptiles, we used a recent phylogeny based on morphological and genetic data as our backbone for character mapping and for the ancestral state reconstruction of autotomy and regeneration (Pyron, 2017). Due to a lack of data on regeneration and autotomy for the Serpentes, we omitted this infra-order from our analysis. Regeneration is probably absent in snakes, but there are sporadic reports of caudal autotomy (or tail breakage) in a few colubrids (Arnold, 1984). Whether a tail fragility in these species represents true autotomy or a unique trait remains unresolved (Etheridge, 1967) and underscored the decision to omit them from our analyses. When focusing on Iguanidae, we used a less recent, but more specific phylogeny to manually map autotomy and regeneration in this group (Wiens and Hollingsworth, 2000).

Analyses

To examine the evolutionary history of autotomy and regeneration in arthropods and reptiles, ancestral state reconstruction analyses were done using R version 3.0.1 (R Core Team, 2013) and the ape (Paradis et al., 2004), caper (Orme et al., 2012), geiger (Harmon et al., 2008), phytools (Revell, 2010), phangorn (Schliep, 2011), and xlsx (Dragulescu, 2014) packages. Each ancestral reconstruction figure depicts a density map of 100 stochastic trait histories generated from our dataset using an MCMC approach (stochastic character mapping; Huelsenbeck et al. 2003). The reconstructed ancestral state at an internal node of the tree for autotomy and regeneration is the probability of that traits occurring among the 100 stochastic trait histories.

Results

Ancestral state for autotomy and regeneration in arthropods

Mapping regeneration and autotomy for all arthropods revealed that regeneration, while nearly ubiquitous across arthropod orders, appears to have been independently lost in three orders; in the Opiliones (harvestmen), Geophilomorpha (soil centipedes) and then again within the Endopterygota comprising all the holometabolous insects (Figure 1 and Table 1). However, based on the phylogenetic resolution of our branches, there are a few additional fine scale losses (i.e., genus and species) that are not present on the tree and represent isolated losses of regenerative ability (e.g., mites). In contrast with regeneration, autotomy was lost in ten major lineages containing approximately 23,000 species (Figure 1 and Table 1; Resh and Cardé 2009).

Table 1.

Data for autotomy and regeneration in arthropods associated with Figure 1. Each tree branch is presented with the corresponding order and class. Autotomy and regeneration are characterized as present (1), absent (0), or unknown (NA). Unless indicated by other superscripts, the reference for the character state is ¹Maruzzo et al., 2005; ²Wood and Wood, 1932; ³Korschelt, 1907; ⁴McVean, 1982 and ⁵Fleming et al., 2007.

Tree branch	Class	Subclass or Order or Infraorder or Family	Genus	Regen1	Autotomy1
Winged insects	Insecta	Pterygota, Paleoptera	NA	1	1
		Pterygota, Neoptera	NA	0/1	0/1
Silverfish & Firebrats	Insecta	Zygentoma	NA	1	15
Jumping bristletails	Insecta	Archaeognatha	Machilis, Thermobia & Lepisma	1	15
Springtails	Entognatha	Entomobryomorpha	Orchesella & Tomocerus	1	0
Two-pronged bristletails	Entognatha	Diplura	Campodea	1	15
Horseshoe shrimps	Cephalocarida	Brachypoda	NA	NA	NA
Barnacles	Maxillopoda	Thecostraca	Lepas	1	NA
Crabs	Malacostraca	Brachyura	Uca	13	12
Crayfish & Lobster	Malacostraca	Pleocyemata	Cambarus & Homarus	13	12
Isopods	Malacostraca	Peracarida	Asellus & Porcellio	15	14
Mantis shrimps	Malacostraca	Stomatopoda	Squilla	1	0
Nebalia	Malacostraca	Leptostraca	Nebalia	1	0
Plankton	Hexanauplia	Cyclopoida	NA	1	0
Water fleas	Branchiopoda	Cladocera	Triops	1	NA
Fairy shrimps	Branchiopoda	Anostraca	Branchipus	1	NA
Fish lices	Maxillopoda	Arguloida	Argulus	1	0
Seed shrimps	Ostracoda	Halocyprida	NA	12	04
Millipedes	Diplopoda	Spirostreptida	Julus & Spirostreptus	15	NA
Garden centipedes	Symphyla	Scolopendrellidae	NA	NA	NA
Soil centipedes	Chilopoda	Geophilomorpha	NA	03/4	04
Other centipedes	Chilopoda	Scutigeromorpha	Scutigera	1	14
Spiders & Amblypygi	Arachnida	Amblypygi	Centaurus	1	1
False scorpions	Arachnida	Pseudoscorpiones	NA	NA	NA
Harvestmen	Arachnida	Opiliones	Leiobunum	0	1
Mites & Ticks	Arachnida	Opilioacarida	Opilioacarus	1	0
Horseshoe crabs	Merostomata	Xiphosura	Limulus	1	NA
Sea spiders	Pycnogonida	Pantopoda	Phoxichilus & Nymphon	1	1
Velvet worms*	Udeonychophora	Peripatopsidae	NA	NA	NA
Velvet worms*	Udeonychophora	Peripatidae	NA	NA	NA
Water bears*	Eutardigrada	Parachaela	NA	NA	NA
	Eutardigrada	Apochela	NA	NA	NA

^*Velvet worms and water bears are outgroups in this phylogeny.

We next used ancestral state reconstruction to resolve the phylogenetic history of autotomy. In the case of regeneration, there was no variation in regenerative ability (all remaining taxa were capable of regeneration) using the Rehm et al. (2011) tree; thus, the basal condition was presence of regeneration. Even in the Regier et al. (2010) tree, regeneration is the most parsimonious ancestral state for arthropods because only three derived groups showed a lack of regenerative ability (Figure 1 and Table 1). We determined that autotomy was the ancestral state in arthropods based on the available data (scaled likelihood of autotomy at the root: presence > 0.99, absence = 5.9×10⁻⁴; Figure 2).

Figure 2:

Ancestral state reconstruction of autotomy (see text for methodological details) using a recent arthropod phylogenetic tree (Rehm et al. 2011 based on Meusemann et al. 2010). Red and black circles at the tips of the tree indicate autotomy or no autotomy, respectively (see Table 1 for data). The ancestral reconstruction recovers probabilities for the presence or absence of autotomy along the tree branches showing an ancestral state of autotomy in the present phylogeny. Along those tree branches, absence of autotomy is represented in blue and presence in red. Major arthropods’ phyla or orders have been layered on top of the phylogeny for convenience.

Next, we sought to investigate two arthropod groups where autotomy and regeneration appeared to be more evolutionarily labile. First, we examined the insects (class Insecta) in more detail (Figure 3 and 4). Despite limited data given the number of species within this class, the available data for adult Odonata (Figure 3; Table 2) strongly suggests that most, if not all, hemimetabolous insects (incomplete metamorphosis) are capable of autotomy and regeneration. In contrast, examining holometabolous insects (complete metamorphosis), we found no instances of regeneration in adults (note that Frankliniella and Dictyoptera are hemimetabolous and outgroups in this tree; Figure 4; Table 3). As for autotomy, although the data are incomplete, they support that all holometabolous insects conserve their capacity to autotomize limbs (Figure 4; Table 3). That some holometabolous insects exhibit regenerative ability as larvae suggests that the radical developmental shift to complete metamorphosis created a negative trade-off with regenerative ability.

Figure 3:

Modified phylogeny of adults Odonata (hemimetabolous insects; from Figure 2 in Dumont et al., 2010) onto which both regeneration and autotomy have been mapped (see text for methodological details and Table 2 for data). Taxonomic groups are on the right-hand side for convenience. Red and blue boxes at the tips of the tree indicate presence or absence, respectively, for regeneration. Red and blue branches on the phylogeny indicate the presence or absence, respectively, of autotomy. Dashed branches indicate data are not available (NA).

Figure 4:

Modified phylogeny of adults holometabolus insects (from Figure 2 in McKenna and Farrell, 2010) onto which both regeneration and autotomy has been mapped (see text for methodological details and Table 3 for data). Taxonomic groups are on the right-hand side for convenience and genera are included where known. Red and blue boxes at the tips of the tree indicate presence or absence, respectively, for regeneration. Red and blue branches on the phylogeny indicate the presence or absence, respectively, of autotomy. Dashed branches indicate data are not available (NA).

Table 2.

Autotomy and regeneration data for adult Odonata associated with Figure 3. Each taxon is presented with the associated literature reference used for characterization of autotomy and regeneration as present (1), absent (0), or unknown (NA). ¹Child and Young, 1903; ²Robinson et al., 2006; ³Maruzzo et al., 2005; ⁴Fleming et al., 2007; ⁵Stoks, 2003; ⁶Tennessen, 2009; ⁷Madhavan and Schneiderman, 2007; ⁸Moore and Tabashnik, 1989; ⁹French, 1976 and ¹⁰Maginnis, 2006. Outgroups used for Figure 3 are: Pyralidae, Curculionidae, Trichogrammatidae, Ectobiidae, Acrididae, Sminthuridae and Baetidae.

Genus or Species name	Suborder	Family	Regeneration	Autotomy
Argia moesta	Zygoptera	Coenagrionidae	11/10	12
NA	Zygoptera	Calopterygidae	NA	13
Burmargiolestes sp.	Zygoptera	Megapodagrionidae	NA	13
Noguchiphaea sp.	Zygoptera	Calopterygidae	NA	13
Pseudolestes sp.	Zygoptera	Pseudolestidae	NA	13
NA	Zygoptera	Hetaerinidae	NA	13
NA	Zygoptera	Rimanellidae	NA	13
NA	Zygoptera	Heliocharitidae	NA	13
NA	Zygoptera	Protoneuridae	NA	13
NA	Zygoptera	Platycnemididae	NA	13
NA	Zygoptera	Philogeniidae	NA	13
Stenocnemis sp.	Zygoptera	Platycnemididae	NA	13
Ceriagrion sp.	Zygoptera	Coenagrionidae	11/10	14
NA	Zygoptera	Pseudostigmatidae	NA	13
Nephallenia sp.	Zygoptera	Coenagrionidae	11/10	14
Devadatta sp.	Zygoptera	Amphipterygidae	NA	13
Rhinagrion sp.	Zygoptera	Megapodagrionidae	NA	13
Coeliccia sp.	Zygoptera	Platycnemididae	110	13
Philoganga sp.	Zygoptera	Amphipterygidae	NA	13
NA	Zygoptera	Euphaeidae	NA	13
NA	Zygoptera	Chlorocyphidae	NA	13
NA	Zygoptera	Megapodagrionidae	NA	13
NA	Zygoptera	Megapodagrionidae	NA	13
NA	Zygoptera	Plastystictidae	NA	13
Lestes viridis	Zygoptera	Lestidae	15	14
NA	Epiprocta	Libellulidae	NA	16
NA	Epiprocta	Macromiidae	NA	16
Macromidia sp.	Epiprocta	Corduliidae	NA	16
NA	Epiprocta	Corduliidae	NA	16
NA	Epiprocta	Gomphidae	NA	16
Anax imperator / A. cyanea	Epiprocta	Aeshnidae	13	14
NA	Epiprocta	Petaluridae	NA	16
NA	Epiprocta	Chlorogomphidae	NA	16
NA	Epiprocta	Neopetaliidae	NA	16
NA	Epiprocta	Cordulegastridae	NA	16
NA	Epiprocta	Epiophlebiidae	NA	16
Galleria mellonella	Heteroneura	Pyralidae	17	18
Sitophilus zeamais	Polyphaga	Curculionidae	NA	NA
Trichogramma minutum	Apocrita	Trichogrammatidae	NA	NA
Blattella germanica	Cockroaches	Ectobiidae	19	19
Oxya chinensis	Caelifera	Acrididae	13	13
Sminthurus viridis	Collembola (order)	Sminthuridae	13	03
Callibaetis ferrugineus	Pisciforma	Baetidae	13	13

Table 3.

Autotomy and regeneration data for adult holometabolous insects associated with Figure 4. Each taxon is presented with the associated literature reference used for characterization of autotomy and regeneration as present (1), absent (0), or unknown (NA). ¹Wigglesworth, 1944; ²Maginnis, 2006; ³Fleming et al., 2007; ⁴Haynie and Bryant, 1976; ⁵Marchand, 1917; ⁶Shcherbakov et al., 1995; ⁷ Yang et al., 2016 and ⁸Hermann, 1971. ⁹Frankliniella and Dictyoptera are hemimetabolous insects and outgroups in this phylogeny.

Genus or Species name	Order or Suborder	Family	Regeneration	Autotomy
Pterostichus	Adephaga	Carabidae	NA	NA
Bembidion	Adephaga	Carabidae	NA	NA
Laccophilus	Adephaga	Dytiscidae	NA	NA
Hydroscapha	Myxophaga	Hydroscaphidae	NA	NA
Tenomerga	Archostemata	Cupedidae	NA	NA
Strangalia	Polyphaga	Cerambycidae	01	NA
Tribolium	Polyphaga	Tenebrionidae	02	13
Chauliognathus	Polyphaga	Cantharidae	01	NA
Halictophagus sp.	Strepsiptera	Halictophagidae	NA	NA
Mengenilla	Strepsiptera	Mengenillidae	NA	NA
Kempynus	Hemerobiiformia	Osmylidae	NA	NA
Austroneurorthus	Neuroptera	Nevrorthidae	NA	NA
Platystoechotes	Neuroptera	Polystoechotidae	NA	NA
NA	Megaloptera	NA	NA	NA
Mongoloraphidia	Raphidioptera	Raphidiidae	NA	NA
Microchorista	Mecoptera	Nannochoristidae	NA	NA
Nannochorista	Mecoptera	Nannochoristidae	NA	NA
Panorpa	Mecoptera	Panorpidae	NA	NA
Bittacidae	Raptipeda	Bittacidae	NA	NA
Boreus	Mecoptera	Boreidae	NA	NA
Neotyphloceras	Hystrichopsyllomorpha	Hystrichopsyllidae	NA	NA
Ctenocephalides	Pulicomorpha	Pulicidae	NA	15
Drosophila	Brachycera	Drosophilidae	04	13
Musca	Brachycera	Muscidae	NA	NA
Tipulidae	Nematocera	Tipulidae	NA	13
Anopheles	Nematocera	Culicidae	NA	16
Noctuidae	Glossata	Noctuidae	07	13
Bombyx	Glossata	Bombycidae	02	13
Hydropsyche	Annulipalpia	Hydropsychidae	NA	NA
Pteromalidae	Apocrita	Pteromalidae	NA	18
Tenthredinidae	Symphyta	Tenthredinidae	NA	18
Apis	Apocrita	Apidae	NA	18
Frankliniella9	Terebrantia	Thripidae	NA	NA
NA	Dictyoptera9	NA	13	12

Ancestral state for autotomy and regeneration in reptiles

We next investigated the evolutionary history of these two traits in reptiles and assigned each tip of the phylogeny with a value for regeneration and autotomy based on family and infra-order information (Figure 5; Table 4). Character mapping of both traits revealed that they were widespread across reptiles (Figure 5; Table 4). Similar to arthropods, we determined autotomy and regeneration as the ancestral state for squamate reptiles based on our data (autotomy: scaled likelihood of autotomy at the root: presence = 0.9747, absence = 0.0253, standard error = 7.74×10⁻⁴; regeneration: Scaled likelihood of regeneration at the root: presence = 0.9942, absence = 0.0058, standard error = 4.940×10⁻⁴; Figure 6 and Figure 7). To investigate the association of regeneration and autotomy in more detail, we examined the Iguanidae family where both traits showed independent losses (Figure 8; Table 5). Our analysis revealed that loss of one trait did not necessarily predict the other. For example, Iguana delicatissima is capable of regeneration but not autotomy, whereas a sister species Iguana iguana is capable of autotomy but not regeneration (Figure 8). Together, our results show that regeneration and autotomy are the basal condition in squamate reptiles and that tail regeneration, albeit imperfect, has persisted in most major reptilian lineages.

Figure 5:

Modified phylogeny of squamate reptiles (Pyron, 2017) onto which regeneration and autotomy have been mapped (see text for methodological details and Table 4 for data). Full names of the taxa are on the right-hand side with family color coded for convenience. Red and blue boxes at the tips of the tree indicate presence or absence, respectively, for regeneration. Red and blue branches on the phylogeny indicate the presence or absence, respectively, of autotomy. Dashed branches indicate data are not available (NA).

Table 4.

Autotomy and regeneration data for squamate reptiles associated with Figures 5, 6, and 7. Each taxon is presented with the associated literature reference used for characterization of autotomy and regeneration as present (1), absent (0), or unknown (NA). ¹The reptile database (http://www.reptile-database.org/). Unless indicated by other superscripts, the reference for the character state is from ²Bellairs and Bryant 1985; ³Arnold 1994; ⁴Anderson and Higham 2014; ⁵Cope 1967; ⁶Zug et al. 2006; ⁷Wiens and Etheridge 2006; ⁸Zani 1996; ⁹Laspiur et al. 2007 and ¹⁰Poe 2006.

Species name	Suborder1	Family1	Regeneration2	Autotomy2
Acontias_percivali	Scincomorpha	Scincidae	1	1
Aeluroscalabotes_felinus	Gekkota	Eublepharidae	1	1
Agama_agama	Iguania	Agamidae	1	1
Amphiglossus_splendidus	Scincidae	Amphiglossus	1	1
Amphisbaena_fuliginosa	Lacertoidea	Amphisbaenidae	1	1
Anniella_pulchra	Anguimorpha	Anniellidae	1	1
Anolis_carolinensis	Iguania	Dactyloidae	1	1
Aspidoscelis_tigris	Sauria	Teiidae	1	1
Basiliscus_basiliscus	Iguania	Corytophanidae	1	1
Bipes_biporus	Lacertoidea	Bipedidae	1	1
Bipes_canaliculatus	Lacertoidea	Bipedidae	1	1
Brachylophus_fasciatus	Iguania	Iguanidae	1	03
Brachymeles_gracilis	Scincidae	Scincidae	1	1
Brookesia_brygooi	Iguania	Chamaeleonidae	0	04
Callopistes_maculatus	Sauria	Teiidae	1	1
Calotes_emma	Iguania	Agamidae	16	1
Celestus_enneagrammus	Anguimorpha	Diploglossidae	1	1
Chalarodon_madagascariensis	Iguania	Opluridae	1	1
Chamaeleo_laevigatus	Iguania	Chamaeleonidae	0	04
Coleonyx_variegatus	Gekkota	Eublepharidae	1	1
Corytophanes_cristatus	Iguania	Corytophanidae	0	08
Cricosaura_typica	Scincoidea	Xantusiidae	1	1
Crotaphytus_collaris	Iguania	Crotaphytidae	0	03
Delma_borea	Gekkota	Pygopodidae	1	1
Diplometopon_zarudnyi	Amphisbaenia	Trogonophidae	1	1
Dipsosaurus_dorsalis	Iguania	Iguanidae	1	1
Elgaria_multicarinata	Anguimorpha	Anguidae	1	1
Enyalioides_laticeps	Iguania	Hoplocercidae	1	17
Eublepharis_macularius	Gekkota	Eublepharidae	1	1
Eugongylus_rufescens	Scincoidea	Scincidae	1	1
Feylinia_polylepis	Scincoidea	Scincidae	1	1
Gambelia_wislizenii	Iguania	Crotaphytidae	1	1
Gekko_gecko	Gekkota	Gekkonidae	1	1
Geocalamus_acutus	Lacertoidea	Amphisbaenidae	1	1
Gonatodes_albogularis	Gekkota	Sphaerodactylidae	1	1
Heloderma_horridum	Anguimorpha	Helodermatidae	1	03
Heloderma_suspectum	Anguimorpha	Helodermatidae	1	03
Lacerta_viridis	Lacertoidea	Lacertidae	1	1
Lanthanotus_borneensis	Anguimorpha	Lanthanotidae	1	1
Leiocephalus_barahonensis	Iguania	Leiocephalidae	1	1
Leiolepis_belliana	Iguania	Agamidae	0	07
Leiosaurus_catamarcensis	Iguania	Leiosauridae	0	09
Lepidophyma_flavimaculatum	Scincoidea	Xantusiidae	1	1
Lialis_burtonis	Gekkota	Pygopodidae	1	1
Liolaemus_bellii	Iguania	Liolaemidae	1	1
Morunasaurus_annularis	Iguania	Hoplocercidae	1	1
Oplurus_cyclurus	Iguania	Opluridae	1	1
Petrosaurus_mearnsi	Iguania	Phrynosomatidae	1	1
Phelsuma_lineata	Gekkota	Gekkonidae	1	1
Phrynosoma_platyrhinos	Iguania	Phrynosomatidae	0	0
Phymaturus_palluma	Iguania	Liolaemidae	1	1
Physignathus_cocincinus	Iguania	Agamidae	1	07
Platysaurus_pungweensis	Scincoidea	Cordylidae	1	1
Plestiodon_fasciatus	Scincoidea	Scincidae	1	1
Plica_plica	Iguania	Tropiduridae	1	1
Pogona_vitticeps	Iguania	Agamidae	1	1
Polychrus_marmoratus	Iguania	Polychrotidae	0	07
Pristidactylus_torquatus	Iguania	Leiosauridae	1	1
Pseudopus_apodus	Anguimorpha	Anguidae	1	1
Rhacodactylus_auriculatus	Gekkota	Diplodactylidae	1	1
Rhineura_floridana	Amphisbaenia	Rhineuridae	1	05
Saltuarius_cornutus	Gekkota	Carphodactylidae	1	1
Sauromalus_ater	Iguania	Iguanidae	1	1
Sceloporus_variabilis	Iguania	Phrynosomatidae	1	1
Scincus_scincus	Scincoidea	Scincidae	1	1
Shinisaurus_crocodilurus	Anguimorpha	Shinisauridae	1	1
Smaug_mossambicus	Scincoidea	Cordylidae	1	1
Sphenodon_punctatus	Rhynchocephalia	Sphenodontidae	1	1
Sphenomorphus_solomonis	Scincoidea	Scincidae	1	1
Stenocercus_guentheri	Iguania	Tropiduridae	1	1
Strophurus_ciliaris	Gekkota	Diplodactylidae	1	1
Takydromus_sexlineatus	Lacertoidea	Lacertidae	1	1
Teius_teyou	Gymnophthalmoidea	Teiidae	1	1
Teratoscincus_przewalskii	Gekkota	Sphaerodactylidae	1	1
Tiliqua_scincoides	Scincoidea	Scincidae	1	1
Trachylepis_quinquetaeniata	Scincoidea	Scincidae	1	1
Trogonophis_wiegmanni	Lacertoidea	Trogonophidae	1	1
Tupinambis_teguixin	Gymnophthalmoidea	Teiidae	1	1
Uma_scoparia	Iguania	Phrynosomatidae	1	1
Uranoscodon_superciliosus	Iguania	Tropiduridae	1	1
Uromastyx_aegyptia	Iguania	Agamidae	1	1
Urostrophus_vautieri	Iguania	Leiosauridae	0	010
Uta_stansburiana	Iguania	Phrynosomatidae	1	1
Varanus_acanthurus	Anguimorpha	Varanidae	1	03
Varanus_exanthematicus	Anguimorpha	Varanidae	1	03
Varanus_salvator	Anguimorpha	Varanidae	1	03
Xantusia_vigilis	Scincoidea	Xantusiidae	1	1
Xenosaurus_grandis	Anguimorpha	Xenosauridae	1	03
Xenosaurus_platyceps	Anguimorpha	Xenosauridae	1	03

Figure 6:

Ancestral state reconstruction of autotomy (see text for methodological details and Table 4 for data) using a recent reptile phylogenetic tree (Pyron, 2017). Colors are the same as Figure 2 in terms of presence and absence of the trait. The ancestral reconstruction shows an ancestral state of autotomy for the whole phylogeny. Major squamate reptile families have been layered on top of the phylogeny for convenience; (*) denotes Lacertoidae and (**) denotes Rhynchocephalia.

Figure 7:

Ancestral state reconstruction of regeneration (see text for methodological details and Table 4 for data) using the reptile phylogenetic tree from Figure 6 (Pyron, 2017). Colors are the same as Figure 2 in terms of presence and absence of the trait. The ancestral reconstruction shows an ancestral state of regeneration for the whole phylogeny. Major squamate reptiles families have been layered on top of the phylogeny for convenience; (*) denotes Lacertoidae and (**) denotes Rhynchocephalia.

Figure 8:

Modified phylogeny of Iguanidae (from “combined data” in Figure 1 in Wiens and Hollingsworth, 2000) onto which both regeneration and autotomy has been mapped (see text for methodological details and Table 5 for data). Full names of the taxa are on the right-hand side for convenience. Red and blue boxes at the tips of the tree indicate presence or absence, respectively, for regeneration. Red and blue branches on the phylogeny indicate the presence or absence, respectively, of autotomy. Dashed branches indicate data are not available (NA).

Table 5.

Autotomy and regeneration data for Iguanidae associated with Figure 8. Each tree tip is presented with the associated literature reference used for characterization of autotomy and regeneration as present (1), absent (0), or unknown (NA). ¹The reptile database (http://www.reptile-database.org/); ²Bellairs and Bryant 1985; ³Arnold 1994; ⁴Smith 2016; ⁵Burger and Gochfeld 2006; ⁶de Queiroz 1987; ⁷Robyn 2013; ⁸Ariano-Sanchez and Gil-Escobedo 2016; ⁹Koleska and Jablonski 2018; ¹⁰Wu et al. 2014 ¹¹Koleska et al. 2017 and ¹²Carter and Hayes 2004.

Species name	Suborder1	Family1	Regeneration	Autotomy
Dipsosaurus_dorsalis	Iguania	Iguanidae	12	12
Brachylophus_fasciatus	Iguania	Iguanidae	12	03
Amblyrhynchus_cristatus	Iguania	Iguanidae	12	03/6
Conolophus_pallidus	Iguania	Iguanidae	02	03/6
Conolophus_subcristatus	Iguania	Iguanidae	02	06
Ctenosaura_hemilopha	Iguania	Iguanidae	14	02
Ctenosaura_palearis	Iguania	Iguanidae	17	16
Ctenosaura_quinquecarinata	Iguania	Iguanidae	18	16
Ctenosaura_similis	Iguania	Iguanidae	15	15
Iguana delicatissima	Iguania	Iguanidae	19	03/6
Iguana iguana	Iguania	Iguanidae	010	12/3
Sauromalus ater	Iguania	Iguanidae	111	16
Sauromalus varius	Iguania	Iguanidae	NA	16
Cyclura cychlura	Iguania	Iguanidae	112	16
Cyclura nubila	Iguania	Iguanidae	112	16
Cyclura ricordii	Iguania	Iguanidae	112	16

Discussion

With few exceptions, biologists continue to debate the adaptive nature of regeneration in the absence of formal evolutionary tests or experiments to directly assess the fitness consequences when regeneration is lost (Sánchez Alvarado 2000; Brockes and Kumar 2008; Bely and Nyberg 2010; Lai and Aboobaker 2018; Maden 2018). Directly measuring fitness can be difficult in regenerative species where lifespans extend for many years (with reproduction occurring over a large portion of the lifespan) and reproductive output can be large. Analyzing specific traits using ancestral state reconstruction provides a rigorous test to assess the evolutionary history for a particular trait. Strong inference is possible when species relationships are supported by robust phylogenies and sampling for a particular character is represented widely across the tree branches. In an effort to begin addressing whether the presence or absence of regenerative ability might be under selection, we endeavored to perform an ancestral state reconstruction of appendage regeneration in two groups where it has been relatively well-studied (arthropods and reptiles). We also reconstructed the evolutionary history of autotomy in order to ascertain whether regenerative ability might be constrained or influenced by this trait since the two are often associated for any given species in these taxa (Tiozzo and Copley, 2015).

First, our data demonstrated that regeneration and autotomy were present in the basal arthropod ancestor and that these traits have persisted during the evolution of most arthropod lineages. However, the data also showed that forces, known and unknown, can lead to loss of regenerative ability across entire clades. For example, a major shift in life-history strategy from incomplete to complete metamorphosis was accompanied by a complete loss of regenerative ability among adult holometabolous insects. In the case of arthropods, we could not find an instance where regeneration was lost, only to later remerge in a more recent lineage. Compared to regeneration, our data also showed that autotomy has been lost more frequently and that once lost within a lineage this change tends to persist in all descendent lineages. Although autotomy and regeneration are present in many arthropod lineages, the data suggest that the association of these traits has relaxed during the evolution of more recent taxa.

In contrast, regeneration and autotomy appear tightly linked in reptiles and our ancestral state reconstructions show that autotomy and regeneration were present in the common ancestor of all reptiles. Thus, when autotomy is lost in a particular reptilian lineage, regenerative ability is often also lost and vice versa. We did find exceptions to this general trend within the Iguanidae and Anguimorpha. The Anguimorpha were the only group in which autotomy was repeatedly lost but where regeneration remained. Similar to our finding in arthropods, when autotomy was lost at deep nodes it did not appear to re-evolve. It is important to note that conflicting molecular and morphological information exists concerning the phylogenetic relationships among reptiles and thus it is possible that topological inaccuracies within the tree may ultimately explain observed discrepancies (Wiens and Hollingsworth, 2000). Nevertheless, that these patterns were present in both trees adds strong support to our general conclusions. We did find intriguing counterexamples to the tight linkage between these traits, and investigating these counterexamples may help disentangle the evolution of autotomy and regeneration.

Broadly examining when autotomy and regeneration were lost in certain lineages, we find that autotomy is generally lost in more basal lineages, whereas loss of regeneration appears to be a more recent phenomenon. Furthermore, when autotomy is lost it appears very difficult to regain in descendant lineages (with two exceptions in each tree). This suggests that loss of autotomy is relatively permanent and does not leave organisms with dormant processes that are easily re-activated under the right conditions. This may reflect how autotomy is dependent on the presence of complex morphological and physiological components (e.g., precisely located fracture plane, contractile mechanism of muscle, nervous induction, etc.) that may be constrained by similar evolutionary pressures. The case of fragile tails in some snakes offers support for this view. Although some authors have referred to caudal autotomy among a few colubrids, other authors reject the presence of autotomy and suggest that their unique mechanism of intervertebral breakage has most likely been independently derived (Etheridge, 1967; Arnold, 1984).

Interestingly, while autotomy could be uncoupled from regeneration, loss of regenerative ability appears strongly linked to autotomy in both groups. In general, in every reptilian clade where regeneration was lost, so too was autotomy. Although the converse pattern, loss of autotomy leading to loss of regeneration, was common, Iguanas are an exception where autotomy has been lost several times but regeneration persists (Arnold, 1994; Bateman and Fleming, 2009; Gilbert et al., 2013). Aside from these losses, autotomy is widespread among iguana species, and these losses provide fertile ground for more empirical studies focused on why species have lost the ability to autotomize but are still able to regenerate. Similarly, the persistence or loss of regeneration in Pterygota (winged insects), while not necessarily connected to autotomy, was highly correlated with another trait: incomplete or complete metamorphosis (Maruzzo et al., 2005; Seifert et al., 2012b). In this case, the association between metamorphosis and regeneration suggests that the selective pressures associated with the radical life history change to complete metamorphosis imposed alterations that were antagonistic to regeneration in adult holometabolous insects.

The discrepancy between the timing observed for autotomy and regeneration losses could be explained in several different ways. First, it is possible that in those taxa that regenerate ancestrally, regeneration provides significant fitness benefits and thus those taxa that lose it are more likely to go extinct. In this case, only species who have lost regeneration relatively recently will be present in a particular phylogenetic tree. This is difficult to test unless a phylogeny contains closely related species with and without regeneration where fitness consequences can be directly measured (Maginnis 2006). That autotomy is lost more frequently in the trees may reflect a lower cost associated with its loss and thus, taxa that lose it persist for longer and are represented in the trees. It is also possible that the more frequent loss of autotomy reflects different rates of gain and loss between autotomy and regeneration. In this scenario regeneration is more easily gained and lost, whereas autotomy may be lost more slowly but is rarely re-gained. As a complex trait with many interacting processes, regeneration can be easily lost and gained in some taxa when mutations inactivate a signaling pathway required for regeneration. Examples include loss (and gain) of head regeneration in annelids and some flatworms (Bely and Sikes 2010). However, it is also possible that mutations which restore regeneration may be likely if the inactivated process can be replaced by the augmented function of a different but related process. Moreover, such mutations might be likely to sweep through populations if regeneration is under strong positive selection. Thus, for those species that lose and regain regeneration, the gain might occur before the species that lost it diversifies, which would make inferring the gain difficult using contemporary data. The loss of regenerative ability has also been associated with the evolution of increased cellular diversification, specific life-histories, differences in growth mode, and alterations in adaptive immunity (rev. in Needham 1953; Sánchez Alvarado 2000; Harty et al. 2003; Mescher and Neff 2005). In contrast, autotomy may be a trait more susceptible to complete loss without easy re-enabling options besides evolving the trait altogether anew. Coupled with weaker selection for mutations that do cause gain of autotomy, this could result in slower rates for gaining autotomy and thus losses would persist in the phylogeny when they do occur.

When considering inferences derived from data presented in the current paper, several cautionary notes are appropriate. One problem concerning comparative regeneration studies has always been a paucity of data from non-model species. Data concerning the absence or presence of autotomy and regeneration are lacking for many species, which underscores the need for more empirical studies collecting data for both traits. Data indicating true absences of autotomy or regeneration are particularly important. Indeed, inconspicuous appendages might be capable of autotomy or regeneration or both despite reports to the contrary for a more prominent appendage. Thus, only a systemic approach testing several appendages and organs can truly reject the presence of autotomy or regeneration or both in one species compared to another. For these reasons, we caution against strong inference based on the data presented here and instead present our conclusions as testable hypotheses in need of more systematic data.

To conclude, we found convincing support for the hypothesis that regeneration and autotomy were present in the common ancestors of arthropods and squamate reptiles and neither likely re-evolved in lineages where they were lost. In addition, our work supports the hypothesis that regeneration is lost relatively quickly possibly due to trade-offs with other costly capacities (Giangrande and Licciano, 2014). Recalling the views of Weismann and Morgan who believed, respectively, that regeneration is adaptive or that it is an ancestral and fundamental property, we suggest that our work finds evidence for both of these views. Finally, we emphasize the need for extensive empirical studies that collect data on regenerative ability and autotomy such that stronger inferences can be drawn from phylogenetic reconstructions. Similarly, we encourage researchers to expand exploration of the ancestral states of autotomy and regeneration in other clades such as birds, amphibians, and mammals. For instance, we know that complex tissue regeneration has evolved independently within mammals at least twice (rabbits and spiny mice) (Vorontsova and Liosner, 1960; Joseph and Dyson, 1966; Seifert et al., 2012a). With these data in hand, we will be better equipped to answer a myriad of questions including: have cases of appendage regeneration across vertebrate lineages evolved independently; does autotomy drive selection for regeneration; does the loss of regeneration always lead to the loss of autotomy; and why is regeneration seemingly more essential to maintaining autotomy than autotomy is to maintaining regeneration? Only with a broader understanding of the origins of regeneration and autotomy will we be able to enhance our understanding of those traits and one day harness the power of regeneration for the benefits of human beings.