COMMENTARY COMMENTARY Trait-based extinction catches the Red Queen napping during the Cambrian P. David Polly1 changes, estimating the rates of speciation Department of Geological Sciences, Indiana University, Bloomington, IN 47405 and extinction inside and outside that node, and repeating the exercise for all traits (8). The tenuous balance between speciation and or otherwise adapting faster than the Red The number of candidate trait changes in extinction governs the rise and fall of di- Queen’s pace of constant extinction expected even one clade is large, 275 in the early tet- versity within clades, from which have in resource-limited environments (5, 6). In rapod dataset alone (9). Furthermore, flaws emerged the sweeping changes in Earth’s PNAS, Wagner and Estabrook (7) ask in the phylogeny can lead to misinterpreta- standing biodiversity since life’sorigin(1, whether diversification was associated with tions about the link between diversification 2). In a clade where speciation and extinction the evolution of new traits and, if so, did and traits, and phylogenetic analyzes of ex- are equally likely, net diversity remains con- the probability of speciation go up in clades tinct metazoan taxa are sparse compared with their immense Phanerozoic diversity. stant; when the rate of speciation exceeds that have the new trait, did the group’srateof To meet these challenges, the authors extinction, diversity increases exponentially; trait evolution go up, did extinction go up in approached the problem indirectly by look- and when the chance of extinction outweighs the clades that do not have the trait, or some ing at the sequence of trait combinations in speciation, then diversity falls, resulting ulti- combination of these. They analyzed 319 trait temporal series of fossil taxa and the times mately in the demise of the clade. Metazoan datasets from metazoan phyla with good fos- when clades reached their peak diversity. life diversified rapidly in the Cambrian and sil records—arthropods, brachiopods, mol- Wagner and Estabrook used the concept of Ordovician periods (542–443 Mya), after luscs, echinoderms, and chordates—sampled character compatibility (10) to simplify their which short-term extinction events and from the early radiation in the Cambrian to analyses and to avoid relying too heavily on bursts of radiation overshadowed long-term the present day. They found that diversifica- the topologies of potentially erroneous phy- increases in diversity (3, 4). This pattern sug- tion was linked to trait evolution in all post- logenetic trees. Compatibility analysis was de- gests that rates of speciation exceeded extinc- Cambrian clades, but not in the earliest trilo- veloped in the late 1960s and 1970s as part of tion during the first phase of metazoan bites,andthatitwaslinkedtoanincreasein the explosion of interest in computerized diversification, after which the balance shifted extinction among the clades that did not phylogenetic analysis. Estabrook, who died back and forth. A long-standing question is evolve new traits. in 2011, was one of the major contributors whether clades that have successfully diversi- The analytical challenges for determining to the theory, methods, and algorithms used fied have done so by evolving new traits that whether traits are related to diversification to identify compatible characters and the so- confer an advantage allowing them to out- across 319 extinct clades are enormous. called “convex” groups that share them (10– perform or outcompete other groups, either Tackled directly, the problem requires iden- 12). Compatible characters are best under- by winning battles in evolutionary arms races tifying the phylogenetic node at which a trait stood as those character combinations that do not logically require homoplasy to explain them. Most approaches to phylogenetics test A Incompatible homoplasy (the independent evolution of Compatible a trait in two or more taxa) in the context 2: 1 0 2: 0 1 of a specific tree topology. If a character ABCD ABC C D change occurs only once on a tree, it is A 00 00 a uniquely derived synapomorphy, but if it 2: 0 1 changes more than once it is a homoplasy. B 10 01 2: 0 1 Phylogenetic algorithms typically search tree C 10 10 space for the topology that minimizes the 1: 0 1 D 11 11 1: 0 1 number of homoplasies using one criterion or another (13). Parsimony, for example, finds the topology that strictly minimizes ho- B moplasy and maximizes synapomorphy (14) 11 10 Youngest and maximum likelihood finds the topology 10 01 10 11 that maximizes the probability of the charac- ter distributions taking into account that if 00 00 00 Oldest

Incomp- Author contributions: P.D.P. wrote the paper. HSC DSC atible The author declares no conflict of interest.

Fig. 1. (A) Character transitions on a phylogeny of compatible (Left) and incompatible (Right) sets of character See companion article 10.1073/pnas.1406304111. combinations in four taxa. (B) Transitions in stratigraphically arranged character pairs. 1Email: [email protected].

www.pnas.org/cgi/doi/10.1073/pnas.1419138111 PNAS Early Edition | 1of2 Downloaded by guest on October 2, 2021 a character evolves once it is likely to evolve Other more direct methods exist for 20), which Wagner and Estabrook found to again. Compatibility analysis starts with the estimating whether diversification is linked be the most common mode of trait-based character combinations themselves, some of to trait evolution. Most approaches involve diversification in the fossil record. What which require homoplasy no matter what the fitting a model of trait-dependent speciation Wagner and Estabrook sacrificed in terms of topology (Fig. 1A). Compatible characters are and extinction to a phylogenetic tree and biological specificity, they gained in terms of those that could have evolved without homo- traits observed in the tip taxa (15–19). For broad applicability to a wide variety of fossil plasy on at least some phylogenies, even example, likelihood can be used to fit a model data sets. though they may turn out to have evolved Wagner and Estabrook found that diver- more than once depending on the topology Wagner and Estabrook sification was linked to trait changes in all ofthetruetree. found that diversifica- clades except Cambrian trilobites, but by By placing compatibility in a stratigraphic tion was linked to trait increasing the rate of extinction, not specia- framework in which taxa are distributed tion. The evolution of new characters appears through sequences of older and younger changes in all clades not to have directly conferred a speciation rocks, Wagner and Estabrook were able to except Cambrian trilo- advantage on the taxa that acquired them; distinguish cases where lineage splitting (di- bites, but by increasing instead, it conferred a disadvantage on those versification) must have occurred from cases that did not. This finding is compatible with where the observed character changes were therateofextinction, Van Valen’s Red Queen hypothesis that most compatible with a single lineage evolving not speciation. evolution is driven by competition and that through time. Any two characters are com- species are constantly in danger of losing patible if the taxa in the analysis record them in which the speciation and extinction rates their advantage and thus becoming extinct, in only three of their four possible combina- covary with one or more quantitative traits because of new adaptations in their compet- tions. An ordering can always be found in on a phylogenetic tree of known topology itors (5). Van Valen’sanalogywasthattaxa which the transformation from one pair of (19). This method assesses directly whether must run to keep up, just like Lewis Carroll’s characters to the next never requires a re- speciation and extinction are linked to spe- ’ versal. These compatible orderings are strati- cific traits. Other variants assess whether Red Queen. Van Valen shypothesiswas, graphically consistent if they also occur in a specific character transition, the evolution however, based a constant rate of extinction temporal order in the fossil record (Fig. 1B). of a key innovation like opposable thumbs, is within clades. Wagner and Estabrook con- If the intermediate pair of characters occurs linked with diversification (17). The advan- firm that extinction is linked to character temporally between the others, then the char- tages of these methods over Wagner and adaptations in other taxa, but their results acters have hierarchical stratigraphic compat- Estabrook’s is that they localize on a tree suggest that the probability extinction ibility (HSC). If the intermediate pair occurs where diversifying events took place and increases when other taxa evolve. Further- before the other two, then the characters have identify what characters or trait states are more, their results suggest that the competi- divergent stratigraphic compatibility (DSC), associated with them; however, their dis- tive disadvantage conferred by improvements which suggests that character change and di- advantages are that they require a fully re- in neighboring species did not exist in the versification occurred. The intermediate pair solved tree with taxa found only at the tips, Cambrian, at least not in trilobites. Resources can also occur later than the other two, in which makes them inapplicable to fossil data and niche space may have been plentiful which case it is stratigraphically incompati- that lie deep within the nesting structure of enough in the first phase of metazoan evolu- ble. Taking the probabilities of sampling taxa the phylogeny. These methods are also weak tion that the Red Queen did not apply, no in the fossil record into account, Wagner and at detecting changes in extinction rates (19, matter what traits evolved in other species. Estabrook used simulations of diversification and character evolution to estimate the pro-

portion of HSC, DSC, and incompatible char- 1 Yule GU (1925) A mathematical theory of evolution. Philos Trans R 12 Estabrook GF, Strauch JG, Jr, Fiala KL (1977) An application of acter combinations that are expected when Soc Lond, B 213:21–87. compatibility analysis to the Blackiths’ data on orthopterid insects. diversification and trait evolution are uncor- 2 Raup DM (1985) Mathematical models of cladogenesis. Syst Zool 26(3):269–276. Paleobiology 81(5):525–542. 13 Felsenstein J (2003) Inferring Phylogenies (Sinauer Associates, related, when extinction is elevated in taxa that 3 Sepkoski JJ (1981) A factor analytic description of the Phanerozoic Sunderland, MA). retain ancestral characters, and when specia- marine fossil record. Paleobiology 7(1):36–53. 14 Kluge AG, Farris JS (1969) Quantitative phyletics and the 4 Alroy J, et al. (2008) Phanerozoic trends in the global diversity of evolution of anurans. Syst Zool 18(1):1–32. tion is elevated in taxa that possess derived marine invertebrates. Science 321(5885):97–100. 15 Slowinski JB, Guyer C (1993) Testing whether certain traits characters. When diversification is indepen- 5 Van Valen L (1973) A new evolutionary law. Evol Theory 1:1–18. have caused amplified diversification: An improved method based dent of trait evolution, examples of HSC are 6 Vermeij GT (1994) The evolutionary interaction among species: on a model of random speciation and extinction. Am Nat 142(6): Selection, escalation, and coevolution. Annu Rev Ecol Syst 25:219–236. 1019–1024. uncommon compared with DSC. The times of 7 Wagner PJ, Estabrook GF (2014) Trait-based diversification shifts 16 Paradis E (2005) Statistical analysis of diversification with species maximum diversity (centers of gravity) of taxa reflect differential extinction among fossil taxa. Proc Natl Acad Sci traits. Evolution 59(1):1–12. possessing HSC or DSC characters also differ USA, 10.1073/pnas.1406304111. 17 Ree RH (2005) Detecting the historical signature of key 8 de Queiroz A (2002) Contingent predictability in evolution: Key innovations using stochastic models of character evolution and between models, as do the expected disparities traits and diversification. Syst Biol 51(6):917–929. cladogenesis. Evolution 59(2):257–265. (character differences) among the clade’s 9 Ahlberg PE, Clack JA, Luksevics E, Blom H, Zupin¸s I (2008) 18 Maddison WP, Midford PE, Otto SP (2007) Estimating a binary Ventastega curonica and the origin of tetrapod morphology. Nature character’s effect on speciation and extinction. Syst Biol 56(5): members. Knowing the probabilities of these 453(7199):1199–1204. 701–710. parameters under the different models, Wag- 10 Meacham CA, Estabrook GF (1985) Compatibility methods in 19 FitzJohn RG (2010) Quantitative traits and diversification. Syst ner and Estabrook were able to estimate from systematics. Annu Rev Ecol Syst 16:431–446. Biol 59(6):619–633. 11 Estabrook GF, Johnson CS, Jr, McMorris FR (1976) A 20 Rabosky DL (2010) Extinction rates should not be the character data which models best fit the mathematical foundation of the analysis of cladistics character estimated from molecular phylogenies. Evolution 64(6): observations in the 319 clades of metazoans. compatibility. Math Biosci 29:181–187. 1816–1824.

2of2 | www.pnas.org/cgi/doi/10.1073/pnas.1419138111 Polly Downloaded by guest on October 2, 2021