sign in sign up
<<
Home , Inkayacu, Kairuku, Megadyptes, Pachydyptes, Palaeeudyptes, Palaeospheniscus

Divergence time estimation

Rachel Warnock

001011

Computational Evolution http://www.bsse.ethz.ch/cevo Testing hypotheses in evolutionary biology & macroevolution

t speciation or times

r rates of morphological or molecular evolution

λ rate of speciation

μ rate of extinction Macroevolutionary parameters of interest are phylogenetic parameters

r t λ μ Macroevolutionary parameters of interest are phylogenetic parameters

r t λ μ Macroevolutionary parameters of interest are phylogenetic parameters

r t λ μ

ψ ρ

sampling rate extant sampling A straightforward model of evolution and sampling

λ — speciation rate μ — extinction rate A straightforward model of evolution and sampling

●● λ — speciation rate ● ● ●● μ — extinction rate ● ● ● ● ψ — fossil sampling rate ● ●

●●●●

● ●

● ● ● A straightforward model of evolution and sampling

●● λ — speciation rate ● ● ●● μ — extinction rate ● ● ● ● ψ — fossil sampling rate ● ● ● ρ — extant species sampling ●●●● ● ● ● ●● ● ● ● ● ● The fossilized birth-death process

●● λ — speciation rate ● ● ●● μ — extinction rate ● ● ● ● ψ — fossil sampling rate ● ● ● ρ — extant species sampling ●●●● ● ● ● ●● ● ● ● ● ●

Stadler, 2010 A straightforward model of evolution and sampling

●● r λ — speciation rate ● ● r ●● μ — extinction rate ● ● ● ● ψ — fossil sampling rate ● ● r ● ρ — extant species sampling ●●●● r ● ● ● ●● ● ● ● ● ●

tsampling

textinction tspeciation A straightforward model of evolution and sampling

●● λ — speciation rate ● ● ●● μ — extinction rate ● ● ● ● ψ — fossil sampling rate ● ● ● ρ — extant species sampling ●●●● ● ● ● ●● ● ● ● ● ● The molecular clock hypothesis

Zuckerkandl, Pauling. 1962, 1965 Building the tree of life

● ATGCATGC ● TTGCCTGC ●

● TTGCATCG

● ATGCATCG

● ATGCATGG ● TTGCCTGG ●

● TAGCGTGC ● TAGCGAGC

branch lengths = rate x time ν = rt Dating the tree of life

● ATGCATGC ● TTGCCTGC ●

● TTGCATCG

● ATGCATCG

● ATGCATGG ● TTGCCTGG ●

● TAGCGTGC ● TAGCGAGC

branch lengths = time Calibrating the molecular clock

● ATGCATGC ● TTGCCTGC ●

● TTGCATCG

● ATGCATCG

● ATGCATGG ● TTGCCTGG ●

● TAGCGTGC ● TAGCGAGC 6.5 branch lengths = time Calibrating the molecular clock

● ATGCATGC ● TTGCCTGC ● 20 ● TTGCATCG 70 ● ATGCATCG ● ATGCATGG ● TTGCCTGG 55 ●

● TAGCGTGC ● TAGCGAGC 6.5 branch lengths = time Bayesian divergence time estimation

P( data | model ) P( model ) P( model | data ) = P( data )

Slides adapted from Louis du Plessis Taming the BEAST Workshop 2016 Bayesian divergence time estimation

likelihood priors

P( data | model ) P( model ) P( model | data ) = P( data )

posterior marginal likelihood of the data

Slides adapted from Louis du Plessis Taming the BEAST Workshop 2016 Bayesian divergence time estimation

ACAC... λ μ TCAC... ACAG... ψ ρ

tree calibration substitution clock alignment tree model priors model model

P( data | model ) P( model ) P( model | data ) = P( data )

Slides adapted from Louis du Plessis Taming the BEAST Workshop 2016 Bayesian divergence time estimation

ACAC... λ μ TCAC... ACAG... ψ ρ

tree calibration substitution clock alignment tree model priors model model

ACAC... λ μ λ μ TCAC... P( A CAG... | ψ ρ ) P( ψ ρ ) λ μ ACAC... TCAC... ψ ρ P( | A C AG... ) = ACAC... TCAC... P( A C AG... )

Slides adapted from Louis du Plessis Taming the BEAST Workshop 2016 Bayesian divergence time estimation

ACAC... λ μ TCAC... ACAG... ψ ρ

tree calibration substitution clock alignment tree model priors model model

ACAC... λ μ λ μ TCAC... P( A CAG... | ψ ρ ) P( ψ ρ ) λ μ ACAC... TCAC... ψ ρ P( | A C AG... ) = ACAC... TCAC... P( A C AG... )

λ μ Coming up… ψ ρ

Slides adapted from Louis du Plessis Taming the BEAST Workshop 2016 Bayesian divergence time estimation

ACAC... λ μ •TCSequenceAC... data do not contain information about absolute time ACAG... ψ ρ

tree calibration substitution clock alignment tree model priors model model

ACAC... λ μ λ μ TCAC... P( A CAG... | ψ ρ ) P( ψ ρ ) λ μ ACAC... TCAC... ψ ρ P( | A C AG... ) = ACAC... TCAC... P( A C AG... )

λ μ Coming up… ψ ρ Bayesian divergence time estimation

ACAC... λ μ •TCSequenceAC... data do not contain information about absolute time ACAG... ψ ρ

•This has several importanttree consequences:calibration substitution clock alignment tree model priors model model •We need strong prior information about the divergence

times or the substitution ArateCAC... λ μ λ μ TCAC... P( A CAG... | ψ ρ ) P( ψ ρ ) λ μ ACAC... TCAC... ψ ρ P( • Model selection | A C AG... cannot) = be used to select among possible ACAC... TCAC... calibration strategies P( A C AG... )

•If there is uncertainty in the calibrations, even an infinite amount of sequence data won’t completely eliminate

uncertainty in the posteriors λ μ Coming up… ψ ρ REVIEWS

Neutral theory illustrates the Bayesian clock dating of equation (2) in a the Bayesian estimates converge to the true parameter Also termed the neutral two-species case. values as more and more data are analysed. However, mutation-random drift theory; Direct calculation of the proportionality constant convergence on truth does not occur in divergence time claims that evolution at the z in equation (2) is not feasible. In practice, a simula- estimation. The use of priors to resolve times and rates molecular level is mainly Markov Chain Monte Carlo random fixation of mutations tion algorithm known as the has two consequences. First, as more loci or increasingly that have little fitness effect. algorithm (MCMC algorithm) is used to generate a longer sequences are included in the analysis but the sample from the posterior distribution. The MCMC calibration information does not change, the posterior Neutral mutations algorithm is computationally expensive, and a typi- time estimates do not converge to point values and will Mutations that do not affect cal MCMC clock-dating analysis may take from a few instead involve uncertainties31,54,62. Second, the priors on the fitness (survival or reproduction) of the individual. minutes to several months for large genome-scale data times and on rates have an important impact on the pos- sets. Methods that approximate the likelihood can terior time estimates even if a huge amount of sequence Advantageous mutations substantially speed up the analysis29,57,58. For technical data is used62,63. Errors in the time prior and in the rate Mutations that improve the reviews on Bayesian and MCMC molecular clock dating prior can lead to very precise but grossly inaccurate time fitness of the carrier and are REFS 59,60 62,64 favoured by natural selection. see . estimates . Great care must always be taken in the con- Nearly a dozen computer software packages cur- struction of fossil calibrations and in the specification Deleterious mutations rently exist for Bayesian dating analysis (TABLE 1), all of of priors on times and on rates in a dating analysis65,66. Mutations that reduce the which incorporate models of rate variation among lin- As the amount of sequence data approximates fitness of the carrier and are eages (the episodic or relaxed clock models envisioned genome scale, the molecular distances or branch removed from the population 61 by negative selection. by Gillespie) . All of these programs can also analyse lengths on the phylogeny are essentially determined multiple gene loci and accommodate multiple fossil without any uncertainty, as are the relative ages of the Substitution calibrations in one analysis. nodes. However, the absolute ages and absolute rates Mutations that spread into the cannot be known without additional information (in population and become fixed, Limits of Bayesian divergence time estimation driven either by chance or by the form of priors). The joint posterior of times and natural selection. Estimating species divergence times on the basis of rates is thus one-dimensional. This reasoning has been uncertain calibrations is challenging. The main diffi- used to determine the limiting posterior distribution Relaxed clock models culty is that molecular sequence data provide informa- when the amount of sequence data (that is, the number Models of evolutionary rate tion about molecular distances (the product of times of loci or the length of the sequences) increases without drift over time or across 31,54 lineages developed to relax the and rates) but not about times and rates separately. In bound . An infinite-sites plot can be used to deter- molecular clock hypothesis. other words, the time and rate parameters are unidenti- mine whether the amount of sequence data is satur- fiable. Thus, in Bayesian clock dating, the sequence ated or whether including more sequence data is likely Rate and timedistances areare resolved only into absolute semi-identifiable times and rates to improve the time estimates (FIG. 2). The theory has through the use of priors. In a conventional Bayesian been extended to the analysis of large but finite data estimation problem, the prior becomes unimportant and sets to partition the uncertainties in the posterior time

a Prior f(t,r) b Likelihood L(D|t,r) c Posterior f(t,r|D)

6

5

4 s/s/My) 3 (× 10 r

2 Rate,

1

0 0 10 20 30 40 50 0 10 20 30 40 50 0 10 20 30 40 50 Time, t (Ma) Time, t (Ma) Time, t (Ma) | Figure 1 Bayesian molecular clock dating. We estimate the posterior likelihood (part b) is calculated under the Jukes–CantorNature model Reviews. The | posteriorGenetics distribution of divergence time (t) and rate (r) in a two-species case to surface (part c) is the result of multiplying the prior and the likelihood. The illustrate Bayesian molecular clock dating. The data are an alignment of the data are informative about the molecular distance, d = tr, but not about t and 12S RNA gene sequences from humans and orang-utans, with 90 r separately. The posterior is thus very sensitive to the prior. The blue line differences at 948 nucleotides sites. The joint prior (part a) is composed of indicates the maximum likelihood estimate of t and r, and the molecular two gamma densities (reflecting our prior information on the molecular rate distance d, with ˆtˆr = ˆd. When the number of sites is infinite, the likelihood and on the geological divergence time of human–orang-utan), and the collapses onto the blue line, and the posterior becomes one-dimensional62.

dos Reis et al. 2016. Nature Genetics Reviews 74 | FEBRUARY 2016 | VOLUME 17 www.nature.com/nrg

© 2016 Macmillan Publishers Limited. All rights reserved The global molecular clock model

• The substitution rate is constant over time

• All lineages share the same rate

branch length = substitution rate low high The global molecular clock model

exponential

δc c clock rate

to the phyloCTMC

branch length = substitution rate low high The independent uncorrelated rates model

• Lineage-specific rates are uncorrelated

• The rate assigned to each branch is drawn independently from an underlying distribution branch length = substitution rate low high

Drummond et al. 2006; Rannala & Yang 2007; Lepage et al. 2007 The independent uncorrelated rates model

exponential exponential

δν ν ri branch rates

i ∈ N

to the phyloCTMC

branch length = substitution rate low high Many molecular clock models

• Global clock (Zuckerkandl & Pauling, 1962)

• Uncorrelated/independent rates models (Drummond et al. 2006; Rannala & Yang 2007; Lepage et al. 2007)

• Log-normally distributed autocorrelated rates (Thorne, Kishino & Painter 1998; Kishino, Thorne & Bruno 2001; Thorne & Kishino 2002)

• Local clocks (Hasegawa, Kishino & Yano 1989; Kishino & Hasegawa 1990; Yoder & Yang 2000; Yang & Yoder 2003, Drummond and Suchard 2010)

• Mixture models on branch rates (Heath, Holder, Huelsenbeck 2012)

• Punctuated rate change model (Huelsenbeck, Larget and Swoford 2000) Vagaries of the rock and fossil records

Bonaparte Basin, AU Vagaries of the rock and fossil records

• The fossil record is extremely incomplete & highly uneven

• Preservation gets (approximately) worse as you go further back in time

• % of original habitat area that survives today at outcrop: • Marine shelf habitats 2% • Terrestrial habitats 0.5% (of which = 20% and Neogene =30% of the total)

Wall et al. 2011. Geological Society Special Publication Controls on the probability of preservation

Modified from Smith, 2007. J. of the Geological Society provide minimum estimates of species divergence time

Molecular evolution:

Morphological evolution:

Fossil preservation:

Time Fossils provide minimum estimates of species divergence time

genetic Molecular evolution: divergence

Morphological evolution:

Fossil preservation:

Time Fossils provide minimum estimates of species divergence time

genetic Molecular evolution: divergence

Morphological evolution: apomorphy

Fossil preservation:

Time Fossils provide minimum estimates of species divergence time

genetic Molecular evolution: divergence

Morphological evolution: apomorphy

Fossil preservation: earliest fossil

Time Fossils provide minimum estimates of species divergence time

1. Fossil minimum 2. Acquisition of apomorphy 3. Most probable divergence time 3. 2.

1.

Time 314TaxonomicOpinion uncertainty:TRENDS in Ecology crown and Evolution versusVol.20 No.6 June stem 2005 groups

100 90 80 70 Stem Gnathostomes Crown 60 50

Diversity 40 30 20 10 Tunicata Cephalochordata Myxinoidea Petromyzontida Conodonta Pteraspidomorphi Anaspida Thelodonti Galeaspida Osteostraci Placodermi Chrondrichthyes Acanthodii Actinopterygii Sarcopterygii

Duplication

Donoghue,TRENDS Purnell. in Ecology & 2005. Evolution TRENDS

Figure 1. Gnathostome origins and the timing of gen(om)e duplication. The cladogram shows the hypothesis of relationships among living (black lines) and extinct (white lines) clades of lower vertebrates. The grey box shows the lack of precision in placing the gnathostome duplication event after the origin of Petromyzontida (lampreys) but before the origin of Chondrichthyes (sharks, skates and rays), and the range offossil clades that fall within this interval. The vertical bars and scale above the cladogram show diversity (number of families, total for each clade). Diversity is based on data for the Palaeozoic because much of the increase in chondrichthian, actinopterygian and sarcopterygian diversity occurred during the Mesozoic and Cenozoic, hundreds of millions of years after the genomic doubling event implicated in gnathostome origin. Relationships based on [28,31]; familial diversity from [50]. all, although most of the groups that we now place among living species [37] teleosts are the most diverse and stem gnathostomes have been known since the 19th century successful group of vertebrates, and their phenotypic [33], others have been discovered or resolved as being diversity, number of species [8,15,38],andeven vertebrates only relatively recently (galeaspids in 1965 increased phenotypic complexity [8] have all been linked [34],pituriaspidsin1991[35] and conodonts in 1993 [36]). to gen(om)e duplication. Closer scrutiny of the apparent The significance of stem relatives of extant higher taxa congruence between duplication and increases in diver- also extends to inferred patterns of diversity change, and sity and complexity, however, reveals a different picture. the adaptive radiations that are identified as a result. For The timings and topology advocated by Hoegg et al. [8], instance, many of the clades of stem gnathostomes are for example, appear to resolve the position of the extremely speciose and would do nothing to diminish the duplication to a point intermediate between teleosts inferred burst of diversity in the crown were they extant and their nearest living relatives, but the authors make (and, therefore, members of the crown) (Figure 1). Indeed, no mention of the 11 extinct clades [39] that intercalate until their demise (most became extinct during the Late between these two lineages. These stem teleosts fill not , w375 million years ago), they were numerically only the phylogenetic, but also the morphological chasm superior to contemporary lineages that now lie within the separating living teleosts from other living actinoptery- crown group. It is not until well after the decline of the gians [39,40], smoothing patterns of character acqui- major groups of stem gnathostomes (i.e. pteraspido- sition previously taken to suggest an evolutionary burst morphs, galeaspids, osteostracans and placoderms) that or sudden increase in phenotypic complexity at the crown gnathostomes became more diverse [41]. origin of teleosts. Furthermore, the phylogenetic posi- The same pattern can also be demonstrated for the pre- tioning of the gen(om)e duplication event (Box 1)does teleost portion of actinopterygian fish phylogeny in which not even coincide with the major teleost radiation, which afurtherepisodeofgen(om)eduplicationhasbeen occurred within the derived acanthomorph sub-clade implicated (Box 1). Proponents of this event make the (PolymixiiformesCParacanthopterygiCAtherinomorphaC evolutionary consequences clear. With almost 24 000 Percomorpha). www.sciencedirect.com 314TaxonomicOpinion uncertainty:TRENDS in Ecology crown and Evolution versusVol.20 No.6 June stem 2005 groups

•Crown group: all descendants of the last common ancestor of 100the living members of a group 90 80 70 Stem Gnathostomes Crown •60Stem group: all species more closely related to the living 50

Diversity 40members of a group than to any other 30 20 10

•TotalTunicata group:Cephalochordata Myxinoidea stemPetromyzontida +Conodonta crownPteraspidomorphi groupAnaspida Thelodonti membersGaleaspida Osteostraci Placodermi Chrondrichthyes Acanthodii Actinopterygii Sarcopterygii

•Early crown and stem group members may be difficult to distinguish

Duplication

TRENDS in Ecology & Evolution

Figure 1. Gnathostome origins and the timing of gen(om)e duplication. The cladogram shows the hypothesis of relationships among living (black lines) and extinct (white lines) clades of lower vertebrates. The grey box shows the lack of precision in placing the gnathostome duplication event after the origin of Petromyzontida (lampreys) but before the origin of Chondrichthyes (sharks, skates and rays), and the range offossil clades that fall within this interval. The vertical bars and scale above the cladogram show diversity (number of families, total for each clade). Diversity is based on data for the Palaeozoic because much of the increase in chondrichthian, actinopterygian and sarcopterygian diversity occurred during the Mesozoic and Cenozoic, hundreds of millions of years after the genomic doubling event implicated in gnathostome origin. Relationships based on [28,31]; familial diversity from [50]. all, although most of the groups that we now place among living species [37] teleosts are the most diverse and stem gnathostomes have been known since the 19th century successful group of vertebrates, and their phenotypic [33], others have been discovered or resolved as being diversity, number of species [8,15,38],andeven vertebrates only relatively recently (galeaspids in 1965 increased phenotypic complexity [8] have all been linked [34],pituriaspidsin1991[35] and conodonts in 1993 [36]). to gen(om)e duplication. Closer scrutiny of the apparent The significance of stem relatives of extant higher taxa congruence between duplication and increases in diver- also extends to inferred patterns of diversity change, and sity and complexity, however, reveals a different picture. the adaptive radiations that are identified as a result. For The timings and topology advocated by Hoegg et al. [8], instance, many of the clades of stem gnathostomes are for example, appear to resolve the position of the extremely speciose and would do nothing to diminish the duplication to a point intermediate between teleosts inferred burst of diversity in the crown were they extant and their nearest living relatives, but the authors make (and, therefore, members of the crown) (Figure 1). Indeed, no mention of the 11 extinct clades [39] that intercalate until their demise (most became extinct during the Late between these two lineages. These stem teleosts fill not Devonian, w375 million years ago), they were numerically only the phylogenetic, but also the morphological chasm superior to contemporary lineages that now lie within the separating living teleosts from other living actinoptery- crown group. It is not until well after the decline of the gians [39,40], smoothing patterns of character acqui- major groups of stem gnathostomes (i.e. pteraspido- sition previously taken to suggest an evolutionary burst morphs, galeaspids, osteostracans and placoderms) that or sudden increase in phenotypic complexity at the crown gnathostomes became more diverse [41]. origin of teleosts. Furthermore, the phylogenetic posi- The same pattern can also be demonstrated for the pre- tioning of the gen(om)e duplication event (Box 1)does teleost portion of actinopterygian fish phylogeny in which not even coincide with the major teleost radiation, which afurtherepisodeofgen(om)eduplicationhasbeen occurred within the derived acanthomorph sub-clade implicated (Box 1). Proponents of this event make the (PolymixiiformesCParacanthopterygiCAtherinomorphaC evolutionary consequences clear. With almost 24 000 Percomorpha). www.sciencedirect.com Taxonomic uncertainty: preservation biases

NATURE | Vol 463 | 11 February 2010 LETTERS

Chordata Cephalochordata Vertebrata Petromyzontida

No decay: crown cephalochordate No decay: crown petromyzontid (juvenile)

1 1

Stage-1 decay: stem cephalochordate Stage-1 decay: stem petromyzontid (juvenile)

2 2

Stage-2 decay: stem cephalochordate Stage-2 decay: crown vertebrate

3 3

Stage-3 decay: crown Stage-3 decay: stem vertebrate

4 4

Stage-4 decay: stem chordate Stage-4 decay: stem chordate

5 5

Stage-5 decay: stem chordate Stage-5 decay: stem chordate

Figure 3 | Morphological decay stages of Branchiostoma (left) and larval taphonomic bias exists in identification of fossil . Characters are Lampetra (right) and the phylogenetic position of each stage if interpreted colour coded according to the hierarchical level for which they are as a fossil. Rectangles on branches of the phylogeny are morphological informative (green, chordate; yellow, cephalochordate; blue, vertebrate; characters, their shade indicating of loss (white, early; dark, late). As purple, cyclostome and vertebrate (sensu ref. 7); red, petromyzontid). each organism decays, its phylogenetic position moves down the tree; thus, Samson et al. 2010. Nature

information during decay of chordates is non-random, which will position could be much more derived (Supplementary Fig. 3). Our affect the preservation potential of characters in the fossil record. results therefore highlight the dangers inherent in assuming that The more phylogenetically informative anatomical features decay placement of a non-biomineralized fossil in the stem of a major extant before the more plesiomorphic characters (that is, chordate synapo- clade necessarily reveals significant information about the time of morphies). Significantly, this means that the fossil record will be origin and patterns of character acquisition of that group. Putative biased towards preservation of symplesiomorphies because characters fossil chordates that are placed in stem positions (stem chordate or that decay very rapidly are generally less likely to be preserved, as the stem vertebrate) because they possess only decay-resistant chordate window of time during which preservation processes can act is of characters (that is, they lack crown-group synapomorphies) must be shorter duration. This is especially true for purported non-vertebrate treated with caution. fossil chordates, all of which are known from Burgess Shale- The prevalence of this decay bias of stem-ward slippage among type deposits, in which the dominant mode of preservation of non- clades that include soft-bodied fossils is unknown. Experimental decay biomineralized anatomy is through stabilization of recalcitrant tissues data for the polychaete Nereis, for example, reveal that the most decay- as organic films18–20. These circumstances will strongly favour preser- resistant features are jaws and chaetae15. Previously these characters vation of decay-resistant symplesiomorphies, and analysis of chordate were considered to be phylogenetically informative, but questions remains will suggest phylogenetic placement of fossils in a more basal have subsequently been raised about the utility of jaws and chaetae position than is correct (Fig. 3), potentially in the stem of the crown in polychaete and scolecodont (fossil polychaete) phylogeny and group to which they belong. Thus, our decay data strongly suggest that as both are strongly tied to function22,23.Morerecentpoly- the fossil record of non-vertebrate chordates is skewed by a systematic chaete phylogenies indicate that it is the labile, soft-tissue characters bias of stem-ward slippage and should be reviewed in light of this. that are more phylogenetically informative24,25; it is these characters Two examples illustrate our point. The fossil Cathaymyrus has that are the first to decay15. characters interpreted as a notochord, pharyngeal ‘striations’, myo- If this decay bias is widespread, the many important evolutionary meres and few other informative anatomical features21. Parsimony episodes that are understood from the fossil record of exceptionally therefore constrains its phylogenetic position as a stem chordate. preserved soft-tissue remains will need careful reconsideration. More However, when viewed in light of the decay bias we have identified, character-based decay analyses of more taxa are needed before we can Cathaymyrus is comparable to Branchiostoma at an advanced stage of develop widely applicable taphonomic models, but this new decay (Supplementary Fig. 3) and, as a result, the absence of more approach will help to constrain anatomical interpretations of what synapomorphic characters could be the result of non-preservation. are currently the most controversial yet potentially critical fossils A ‘cone of phylogenetic uncertainty’ therefore extends the range of known. Every fossil represents the result of interactions between potential placements of Cathaymyrus to encompass the crown groups the processes of decay and the processes of preservation. Decay data of the non-biomineralized chordates (that is, cephalochordates, show clear associations of characters that have similar decay resis- urochordates and juvenile cyclostomes; Fig. 1). Metaspriggina tances and are consequently lost at about the same time. Fossils that similarly exhibits stem-chordate characters only (chevron-shaped seem to preserve an ensemble of anatomical characters that are lost at myomeres; the homology of apparently metameric anterior structures very different stages of decay, and that lack associated characters of is uncertain)7, but decay biases indicate that its true phylogenetic comparable decay resistance, are therefore worth closer scrutiny: 799 ©2010 Macmillan Publishers Limited. All rights reserved Stratigraphic age uncertainty

Benton et al. 2009. TimeTree of Life Soft maximum constraints on divergence times

Neoproterozoic Phanerozoic 107 Ixodes 106 Acanthoscurria 99 Anoplodactylus Scutigera Litopenaeus 100 Onychiurus Arthropoda 98 103 105 Tribolium 101 Rhodnius 97 102 104 Gryllus Daphnia 96 Euperipatoides Onychophora Caenorhabditis Nematoda Protostomia Priapulis Priapulida 82 95 Spadella Flaccisagittaitta Chaetognatha 94 Philodina Brachionuss Rotifera 83 Euprymna 900 93 Mytilus 84 Crassostreaa Mollusca 91 Lottia 92 Aplysia 85 89 Tubifex 86 Helobdella Bilateria 63 88 Pomatoceros Annelida Capitella 87 Alvinella 81 Xenoturbella Symsagittifera Xenacoelomorpha 76 80 Saccoglossus 77 Ptychodera 79 Strongylocentrotus Ambulacraria Holothuria 78 Patiria Deuterostomia 75 Eptatretus Petromyzon Eumetazoa 68 Leucoraja 58 64 Xenopus 71 73 Ornithorhynchus 66 69 74 Mus Homo Chordata 70 72 Gallus Danio 57 65 67 Molgula Ciona Branchiostoma 62 Hydractinia Hydra 55 61 Nematostella Cnidaria Metazoa 59 Anemonia 60 Acropora Trichoplax Placozoa 56 Suberites Amphimedon Porifera Cryogenian Ediacar. Cam O S Dev Carb Pr Tr Jr K Cen. 850 635 541 485 443 419 359 299 252 201 145 66 0 Time (Ma)

Figure 6. The Timetree of the Metazoa Encompassing Major Sources of Uncertainty in Time Estimates Node ages are plotted at the posterior mean for the calibration strategy 1, one partition, IR, and LG + G analysis. The node bars are composites extending from the minimum 2.5% HPD limit to the maximum 97.5% limit across all analyses (excluding results from calibration strategies 3 and 4 and from alternative topologies). dos Reis et al. 2015. Current Biology Cen, Cenozoic; K, Cretaceous; Jr, ; Tr, ; Pr, ; Carb, ; Dev, Devonian; S, ; O, ; Cam, ; Ediacar, Ediacaran. by a global sea level rise associated with increased tectonic ac- appear to be premature. They fail to integrate different sources tivity leading to the destruction of older rock sequences by of uncertainties, which make accurate and precise divergence erosion and subduction. Although this may have promoted the time estimates impossible with current data and methods. innovation and radiation of skeletonizing [14], it will Progress may be possible through analysis of combined also have diminished the fossil record of their forebears [9]. morphological and molecular data, which allow fossil species That said, there remains a record of metazoan- and bilaterian- to be integrated into divergence time analyses on par with like fossil remains and traces in the Ediacaran that we consid- their living relatives [69, 70]. Combined analyses are expected ered insufficiently robust to substantiate a minimum constraint to reduce uncertainty in prior node ages as compared to tradi- on metazoan clades but that invariably informed maxima. tional analysis based on simplistic fossil-based constraints [71, Further insights into the biology of these organisms and others 72]. However, most such analyses conducted to date have like them may well explain away apparent inconsistencies be- yielded unacceptably old divergence time estimates, even tween molecular clock estimates of deep metazoan clade ages older than traditional node-calibrated studies [73]. Otherwise, and their fossil record. statistical analyses of fossil stratigraphic data may yield Nevertheless, attempts to build evolutionary narratives of more objective time priors (e.g., [74–76]) and more informative evolution based on recent molecular clock studies calibrations. Above all, establishing unequivocal evidence for

Current Biology 25, 2939–2950, November 16, 2015 ª2015 The Authors 2947 Soft maximum constraints on divergence times 2009 POINTS OF VIEW 369

Converting fossil evidence into a point calibration is Neoproterozoic Phanerozoic 107 Ixodes only appropriate if the fossil taxon represents the ac- 106 Acanthoscurria 99 Anoplodactylus Scutigera tual common ancestor of 2 extant lineages (e.g., Nodes Litopenaeus 100 Onychiurus Arthropoda A and B in Fig. 1), which is highly unlikely. In any •Minimum98 constraint:103 105 552.85 Ma TriboliumKimberella, interpreted as a 101 Rhodnius case, the age of the fossil needs to be estimated, and 97 102 104 Gryllus Daphnia 96 Euperipatoides Onychophora this can come with errors relating to stratigraphic in- protostome Caenorhabditis Nematoda Protostomia 82 Priapulis Priapulida terpretation and radiometric dating (Magallon´ 2004; 95 Spadella POINTS OF VIEW 2009 Flaccisagittaitta Chaetognatha 369 94 Philodina Gandolfo et al. 2008). For some fossils, these uncer- Brachionuss Rotifera 83 Euprymna tainties can be of almost negligible magnitude because • 900 93 Mytilus Maximum84 constraint: 833 Ma LagerstättenCrassostreaa Mollusca in the Bitter SpringsConverting fossil evidence into a point calibration is 91 Lottia accurate stratigraphic constraints are available (Benton 92 Aplysia only appropriate if the fossil taxon represents the ac- 85 89 Tubifex and Donoghue 2007), but in most instances there can be and Svanbergfjellet86 Formations,Helobdella preserve in 3 dimensions at Bilateria 63 88 Pomatoceros Annelida tual common ancestor of 2 extant lineages (e.g., Nodes Capitella alargedegreeofuncertaintyowingtotheunavailability 87 Alvinella A and B in Fig. 1), which is highly unlikely. In any the cellular 81level prokaryotes, sphaeromorphXenoturbella acritarchs, of isotopic or paleomagnetic dates. One notable exam- Symsagittifera Xenacoelomorpha 76 80 Saccoglossus case, the age of the fossil needs to be estimated, and 77 Ptychodera ple is the Tingamarra fossil fauna in , which multicellular algae,79 but nothing thatStrongylocentrotus couldAmbulacraria be interpretedthis as can a come with errors relating to stratigraphic in- Holothuria includes the oldest Australasian marsupial fossils. The 78 Patiria Deuterostomia 75 Eptatretus terpretation and radiometric dating (Magallon´ 2004; Petromyzon original Early ( 55 Ma) radiometric date for the totalEumetazoa group metazoan68 Leucoraja Gandolfo et al. 2008). For some fossils, these∼ uncer- 58 64 Xenopus site means that the fauna includes the oldest known 71 73 Ornithorhynchus 66 69 74 Mus tainties can be of almost negligible magnitude because Homo Chordata songbirds in the world and, potentially, the oldest bat 70 72 Gallus accurate stratigraphic constraints are available (Benton Danio fossil (Godthelp et al. 1992; Boles 1995). These claims 57 65 67 Molgula Ciona and Donoghue 2007), but in most instances there can be Branchiostoma were placed in doubt, however, by the interpretation 62 Hydractinia alargedegreeofuncertaintyowingtotheunavailability Hydra of Woodburne and Case (1996) that the Tingamarra 55 61 Nematostella Cnidaria of isotopic or paleomagnetic dates. One notable exam- Metazoa 59 Anemonia 60 Acropora fauna is Late , less than 30 Ma old. Substantial Trichoplax Placozoa ple is the Tingamarra fossil fauna in Australia, which 56 Suberites recent corroborating evidence appears to lay this argu- Porifera Amphimedon includes the oldest Australasianment to marsupial rest in favor fossils. of the The Early Eocene date (Beck et al. Cryogenian Ediacar. Cam O S Dev Carb Pr Tr Jr K Cen. original Early Eocene ( 55 Ma) radiometric date for the 850 635 541 485 443 419 359 299 252 201 145 66 0 Time (Ma) ∼ 2008). site means that the fauna includesAmajordeficiencyofpointcalibrationsisthattheir the oldest known Figure 6. The Timetree of the Metazoa Encompassing Major Sources of Uncertainty in Time Estimates songbirds in the world and, potentially, the oldest bat Node ages are plotted at the posterior mean for theCalibrations: calibration strategy 1, one Benton partition, IR, and et LG al. + G analysis.2015. The Palaeontologica node bars are composites extending Electronica. from the Plots: Ho, Phillips. 2009, Sys Bio employment leads to divergence time estimates that minimum 2.5% HPD limit to the maximum 97.5% limit across all analyses (excluding results from calibration strategies 3 and 4 and from alternative topologies). fossil (Godthelp et al. 1992; Boles 1995). These claims Cen, Cenozoic; K, Cretaceous; Jr, Jurassic; Tr, Triassic; Pr, Permian; Carb, Carboniferous; Dev, Devonian; S, Silurian; O, Ordovician; Cam, Cambrian; Ediacar, display illusory precision. This is also a problem when Ediacaran. were placed in doubt, however,secondary by calibrations the interpretation (i.e., molecular date estimates ob- of Woodburne and Casetained (1996) from that other, the Tingamarra independent analyses) are converted by a global sea level rise associated with increased tectonic ac- appear to be premature. They fail to integrate different sources fauna is Late Oligocene, lessto than point 30 values Ma old. without Substantial due consideration of associ- tivity leading to the destruction of older rock sequences by of uncertainties, which make accurate and precise divergence recent corroborating evidence appears to lay this argu- erosion and subduction. Although this may have promoted the time estimates impossible with current data and methods. ated error (Graur and Martin 2004). Ignoring calibration innovation and radiation of skeletonizing animals [14], it will Progress may be possible through analysis of combined ment to rest in favor of theuncertainty Early Eocene has date profound (Beck et consequences al. for evolution- also have diminished the fossil record of their forebears [9]. morphological and molecular data, which allow fossil species 2008). That said, there remains a record of metazoan- and bilaterian- to be integrated into divergence time analyses on par with ary hypothesis testing involving time scales because like fossil remains and traces in the Ediacaran that we consid- their living relatives [69, 70]. Combined analyses are expected Amajordeficiencyofpointcalibrationsisthattheir FIGURE 2. Methods for incorporating calibrations into a dat- it increases the probability of type I errors. The use of ered insufficiently robust to substantiate a minimum constraint to reduce uncertainty in prior node ages as compared to tradi- employment leads to divergence time estimates that on metazoan clades but that invariably informed maxima. tional analysis based on simplistic fossil-baseding analysis: constraints a) point [71, calibration; b) hard minimum bound; c) hard point calibrations appears to have declined in recent Further insights into the biology of these organisms and others 72]. However, most such analysesmaximum conducted to bound; date have d) soft maximum bound;display e) normal illusory distribution; precision.years, This is primarily also a problem owing to when strong criticisms (Graur and like them may well explain away apparent inconsistencies be- yielded unacceptably old divergencef) lognormal time estimates, distribution; even g) exponentialsecondary distribution. calibrations Relative prob- (i.e., molecular date estimates ob- tween molecular clock estimates of deep metazoan clade ages older than traditional node-calibratedability studies is [ measured73]. Otherwise, on the vertical axis of each graph. Martin 2004) as well as the availability and accessibility and their fossil record. statistical analyses of fossil stratigraphic data may yield tained from other, independentof more analyses) sophisticated are converted methods. Nevertheless, they re- Nevertheless, attempts to build evolutionary narratives of more objective time priors (e.g., [74–76]) and more informative to point values without due consideration of associ- animal evolution based on recent molecular clock studies calibrations. Above all, establishing unequivocal evidence for main a popular technique in divergence dating studies concerning calibration methodology,ated error particularly (Graur and the Martin 2004). Ignoring calibration Current Biology 25, 2939–2950, November 16, 2015 ª2015 The Authors 2947 (for a survey, see Ho 2007). use of multiple calibrations. Finally,uncertainty we present has a profound brief consequences for evolution- case study examining the effectsary of using hypothesis different testing cali- involving time scales because FIGURE 2. Methods for incorporatingbration techniques. calibrations into a dat- it increases the probability of type I errors. The use of ing analysis: a) point calibration; b) hard minimum bound; c) hard point calibrations appears to have declined inHard recent Bounds maximum bound; d) soft maximum bound; e) normal distribution; years, primarily owing to strong criticisms (Graur and f) lognormal distribution; g) exponential distribution.CALIBRATIONS Relative AT prob-INTERNAL NODES Hard minimum bounds (Fig.2b).—It has long been rec- ability is measured on the vertical axis of each graph. Martin 2004) as well as theognized availability that and fossil accessibility evidence is only able to provide Point Calibrationsof more sophisticated methods.hard minimum Nevertheless, bounds they on re- divergence times. Correctly Traditionally, divergence timemain estimates a popular have technique been identified in divergence fossils dating can provide studies confirmation that a lin- concerning calibration methodology,calibrated by particularly fixing the age the of(for at least a survey, 1 node see in Ho the 2007).eage existed at a certain point in time, but using them use of multiple calibrations.tree Finally, to a point we value present (Fig. a brief 2a). For example, in order to to make precise statements about divergence events case study examining theestimate effects of the using timing different of mammalian cali- globin gene duplica- would be highly inadvisable because the actual lineage bration techniques. tions, Zuckerkandl and Pauling (1965) assumed that theHardcould Bounds have been in existence well before the appear- most recent common ancestor of placental mammals ance of the calibrating fossil. In Figure 1, for example, CALIBRATIONS ATexistedINTERNAL 80 MaNODES ago. This calibrationHard wasminimum not derived bounds (Figthis.2b would).—It has be longequivalent been rec- to assuming that Fossil 1 is ognized that fossil evidence is only able to provide Point Calibrationsfrom specific fossil evidence but was an estimate of the positioned exactly at Node A. It would be more appro- basal mammalian divergences basedhard on minimum the Tertiary bounds fos- onpriate divergence to use Fossil times. 1 Correctly to place a minimum age constraint Traditionally, divergence time estimates have been sil record. Subsequent studies haveidentified tended fossils to provide can provideon Node confirmation A. Prior to that the a development lin- of methods that calibrated by fixing the age of at least 1 node in the more explicit justifications of fossil-basedeage existed calibration at a certain pointcould in take time, this but characteristic using them into account, some key tree to a point value (Fig. 2a). For example, in order to choices. to make precise statementsdating about studies divergence attempted events to avoid the problem by using estimate the timing of mammalian globin gene duplica- would be highly inadvisable because the actual lineage tions, Zuckerkandl and Pauling (1965) assumed that the could have been in existence well before the appear- most recent common ancestor of placental mammals ance of the calibrating fossil. In Figure 1, for example, existed 80 Ma ago. This calibration was not derived this would be equivalent to assuming that Fossil 1 is from specific fossil evidence but was an estimate of the positioned exactly at Node A. It would be more appro- basal mammalian divergences based on the Tertiary fos- priate to use Fossil 1 to place a minimum age constraint sil record. Subsequent studies have tended to provide on Node A. Prior to the development of methods that more explicit justifications of fossil-based calibration could take this characteristic into account, some key choices. dating studies attempted to avoid the problem by using Probabilistic divergence times priors

Uniform (min, max)

Normal (µ, σ)

Log Normal (µ, σ)

Gamma (α, β)

Exponential (λ)

Species divergence Fossil time minimum Modified from Heath 2012. Sys Bio Cryg Ediacaran Camb Ord Sil Devon Carb Perm Triassic Jurassic Cretaceous Mite 2 Varroa destructor Tick Ixodes scapularis Water flea Daphnia pulex 4 The impact of different A 1 Copepod Lepeophtheirus salmonis Head louse Pediculus humanus calibration priors 3 6 Triatomid bug Rhodnius prolixus 7 Pea aphid Acyrthsiphon pisum Jewel wasp 5 Nasonia vitripennis 9 Honey bee Apis mellifera 10 Carpenter ant 8 Campanotus floridianus Red flour beetle Tribolium castaneum Uniform (min, max) 11 Silkmoth Bombyx mori 12 Mosquito Anopheles gambiae

Mandibulata 13 Fruitfly Drosophila melanogaster 14 Eumetabola Hessian fly Mayetiola destructor Log Normal (µ, σ)

Log Normal (µ, σ)

Log Normal (µ, σ)

Time Fossil minimum

Warnock et al. 2012. Biology Letters Cryg Ediacaran Camb Ord Sil Devon Carb Perm Triassic Jurassic Cretaceous Mite 2 Varroa destructor Tick Ixodes scapularis Water flea Daphnia pulex 4 The impact of different A 1 Copepod Lepeophtheirus salmonis Head louse Pediculus humanus calibration priors 3 6 Triatomid bug Rhodnius prolixus 7 Pea aphid Acyrthsiphon pisum Jewel wasp 5 Nasonia vitripennis 9 Honey bee Apis mellifera 10 Carpenter ant 8 Campanotus floridianus Red flour beetle Tribolium castaneum Uniform (min, max) 11 Silkmoth Bombyx mori 12 Mosquito Anopheles gambiae

Mandibulata 13 Fruitfly Drosophila melanogaster 14 Eumetabola Hessian fly Mayetiola destructor

Log Normal (µ, σ) B minima only BEAST s=0.5, m=1 Diptera

Log Normal (µ, σ) C Apocrita minima only BEAST s=0.5, m=2.5

D minima only Log Normal (µ, σ) BEAST s=2, m=1

Time Fossil minimum E minima only BEAST s=2, m=2.5

F minima & maxima BEAST s=0.5

600 500 400 300 200 100 Time in millions of years before present (Ma) Cryg Ediacaran Camb Ord Sil Devon Carb Perm Triassic Jurassic Cretaceous Mite 2 Varroa destructor Tick Ixodes scapularis Water flea Daphnia pulex 4 The impact of different A 1 Copepod Lepeophtheirus salmonis Head louse Pediculus humanus calibration priors 3 6 Triatomid bug Rhodnius prolixus 7 Pea aphid Acyrthsiphon pisum Jewel wasp 5 Nasonia vitripennis 9 Honey bee Apis mellifera 10 Carpenter ant 8 Campanotus floridianus Red flour beetle Tribolium castaneum 11 Silkmoth Cambrian Bombyx mori 12 Mosquito Anopheles gambiae 13 explosion Mandibulata Fruitfly Drosophila melanogaster 14 Eumetabola Hessian fly Mayetiola destructor

B minima only BEAST s=0.5, m=1 Diptera Cryogenian C Apocrita minima only BEAST s=0.5, m=2.5

D minima only BEAST s=2, m=1

E minima only BEAST s=2, m=2.5

F minima & maxima Ediacaran BEAST s=0.5

600 500 400 300 200 100 Time in millions of years before present (Ma) node 03 Podocnemis node 04 3 Pelusios node 02 4 Pelomedusa Specified versus effective node 06 2 Chelodina priors node 05 6 Elseya 5 node 07 Chelus 7 node 01 Phrynops node 09 1 Carettochelys node 10 9 Apalone 10 node 08 Lissemys node 13 Chelonia 13 node 12 Dermochelys 8 12 node 14 Chelydra node 15 14 Dermatemys

15 node 16 Staurotypus 16 node 11 Sternotherus node 18 Platysternon

18 node 19 Emys

19 node 20 Graptemys 20 17 Trachemys node 17

Geochelone node 21

21 Mauremys node 22 22 Heosemys 250 200 150 100 50 0 Ma Warnock et al. 2015. Biology Letters A prior for the non-fossil calibrated nodes

Extant species phylogeny

●

●

●

●

●

●

●

●

●

●

A prior for the uncalibrated nodes A prior for the non-fossil calibrated nodes: the tree prior

● ● Yule, 1925 λ = 1 ● ● ● ●

● ● Pr

● ●

● ●

● ●

● ● 0 1 2 3 4 5 t A prior for the non-fossil calibrated nodes: the tree prior

● ● Yule, 1925 λ = 1 ● ● ● ●

● ● Pr

● ●

● ●

● ●

● ● 0 1 2 3 4 5 t

● ● Kendall 1948 ● ● λ = 1, μ = 0.3 ● ● ● ● ● ● Pr

● ● ● ● ● ● ● ●

● ● 0 1 2 3 4 5 t

• Different combinations of λ and μ produce different tree shapes A prior for the non-fossil calibrated nodes: the tree prior

● ● Yule, 1925 λ = 1 ● ● ● ●

● ● Pr

● ●

● ●

● ●

● ● 0 1 2 3 4 5 t

● ● Kendall 1948 ● ● λ = 1, μ = 0.3 ● ● ● ● ● ● Pr

● ● ● ● ● ● ● ●

● ● 0 1 2 3 4 5 t

● ● λ = 1, μ = 0.3, ● ● Yang, Rannala, 1997 ● ρ = 0.5 ● ● ● Pr

● ●

●

● ● ●

● ● 0 1 2 3 4 5 t The fossilized birth-death process

●● λ — speciation rate ● ● ●● μ — extinction rate ● ● ● ● ψ — fossil sampling rate ● ● ● ρ — extant species sampling ●●●● ● ● ● ●● ● ● ● ● ●

Stadler, 2010 Incorporating fossils into the tree prior

● ●

● ● ● ●● Stadler, 2010 λ = 1, μ = 0.3, ● ● ● ● ● ●● ● ρ = 0.5, ψ = 1 ● ● ● ● ● ● ● ●● ● ● ● ● ● ● Pr

● ● ● ● ● ● ●

● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● 0 1 2 3 4 5 t

• The simpler models can be considered special cases of the FBD model

• Traditional node dating may depend on a small number of fossils

• Molecular clock analyses will be sensitive to both the calibration priors & the tree prior Sampled ancestors

● ● ●

● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● • The probability of sampling an ancestor in the fossil record is not● ● ● ● ● ●● ● zero (Foote, 1996. Paleobiology) ● ● ● ● ● ● ●● ● • Bayesian inference of sampled● ancestor● trees now possible in λ =BEAST2, 0.1, μ = 0.03, MrBayes, r = 0.3 RevBayes, DPPDivλ = 0.1, μ = 0.08, r = 0.8

●

•●Different● ● combinations of λ, μ, ρ, ψ affect the probability of having a

sampling an ancestor (Wright,● Heath. in prep)

● ● ● ●

● ● ● ●

● ● ● ●

λ = 0.1, μ = 0.05, r = 0.5 Sampled ancestors

● ● ●

● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ●● ● ● ●

λ = 0.1, μ = 0.03, r = 0.3 λ = 0.1, μ = 0.08, r = 0.8

●

● ● ●

●

● ● ● ●

● ● ● ●

● ● ● ●

λ = 0.1, μ = 0.05, r = 0.5 Wright et al. in prep. The fossilised birth-death process: taxonomic uncertainty

•MCMC is used to propose possible fossil placements

•rjMCMC is used to propose sampled ancestor placements

Fossil occurrence Speciation event The fossilised birth-death process: taxonomic uncertainty

•Each fossil can attach anywhere along the tree, including along unobserved branches

Fossil occurrence Speciation event The fossilised birth-death process: taxonomic uncertainty

•The probability of any given realisation of the FBD process is conditional on the model parameters: λ, μ, ρ, ψ

Fossil occurrence Speciation event Application with or without character data for fossils

• Three alternative scenarios to applying to FBD model in divergence time estimation

• Molecular data for extant taxa only ACAC... TCAC... ACAG...

• Molecular data for extant taxa + morphological data for both extant & extinct taxa ACAC... 0101... TCAC... 1101... ACAG... 0100... • Morphological data for both extant and extinct taxa or extinct taxa only 0101... 1101... 0100... Application with or without character data for fossils

• Three alternative scenarios to applying to FBD model in divergence time estimation

• Molecular data for extant taxa only ACAC... TCAC... ACAG...

• Molecular data for extant taxa + morphological data for both extant & extinct taxa ACAC... 0101... TCAC... 1101... ACAG... 0100... • Morphological data for both extant and extinct taxa or extinct taxa only 0101... 1101... 0100...

• The model can also be applied to estimate macroevolutionary parameters when you have no character data Stratigraphic range data

λ — speciation rate μ — extinction rate ψ — fossil sampling rate ρ — extant species sampling

Last appearance (bi)

First appearance (ai) Stratigraphic range data

λ — speciation rate μ — extinction rate ψ — fossil sampling rate ρ — extant species sampling

Last appearance (bi)

First appearance (ai) Putting things in a graphical modelling framework

exponential

δµ µ extinction rate

exponential exponential

δλ λ ψ δψ

speciation rate fossilization rate T

a uniform φ ρ b origin time sampling probability

to the phyloCTMCs Putting things in a graphical modelling framework

λ μ ψ ρ

Time Tree Model

Substitution Model Substitution Model

Site Rate Model Molecular Data Morphological Data Site Rate Model

ACAC... 0101... Branch Rate Model TCAC... 1101... Branch Rate Model ACAG... 0100... Fossil Occurrence Time Data Dating the origin of crown

Late Cretaceous Paleocene Eocene Oligocene Pli.Ple.

Gavryushkina et al. 2016. Sys Bio Dating the origin of crown penguins

Artistic reconstructions by: Stephanie Abramowicz for Scientific American Fordyce, R.E. and D.T. Ksepka. The Strangest Scientific American 307, 56 – 61 (2012) Paleocene Eocene Oligocene Miocene Pli. Ple.

Gavryushkina et al. 2016. Sys Bio Dating the origin of crown penguins

0.33 Delphinornis wimani Delphinornis arctowskii 0.3 Delphinornis larseni 0.08 manneringi Delphinornis gracilis 0.93 devriesi 0.49 0.39 Waimanu tuatahi 0.16 0.35 Marambiornis exilis Mesetaornis polaris 0.32 Anthropornis nordenskjoeldi 0.22 Anthropornis grandis 0.11 Palaeeudyptes klekowskii 0.47 Palaeeudyptes gunnari 0.53 0.32 Burnside “Palaeudyptes” paracasensis salasi Pachydyptes ponderosus 0.37 0.24 Palaeeudyptes antarcticus 0.11 0.35 Kairuku grebneffi 0.37 Kairuku waitaki Archaeospheniscus lowei 0.15 0.15 Duntroonornis parvus 0.14 0.1 antarcticus Archaeospheniscus lopdelli 0.08 patagonicus 0.87 Eretiscus tonnii Marplesornis novaezealandiae 0.26 Platydyptes marplesi adeliae 0.71 0.26 Platydyptes novaezealandiae 1 Pygoscelis papua Palaeospheniscus biloculata 1 Pygoscelis 0.05 0.82 0.58 0.13 Pygoscelis grandis Palaeospheniscus bergi 0.31 patagonicus 1 Aptenodytes forsteri

0.77 minor 0.16 Spheniscus urbinai 0.5 Spheniscus megaramphus Spheniscus mendiculus 1 Spheniscus muizoni Spheniscus humboldti 1 0.28 0.39 Spheniscus magellanicus 0.99 Spheniscus demersus Megadyptes antipodes Eudyptes sclateri 0.95 0.15 Madrynornis mirandus 1 Eudyptes schlegeli 1 Eudyptes chrysolophus 0.46 Eudyptes robustus 1 Eudyptes pachyrhynchus 0.32 Eudyptes moseleyi 1 Eudyptes filholi 1 Eudyptes chrysocome Paleocene Eocene Oligocene Miocene Plio. Plei. Paleogene Neogene Qu.

60 50 40 30 20 10 0 Age (Ma) Gavryushkina et al. 2016. Sys Bio Dating the origin of crown penguins

0.33 Delphinornis wimani Delphinornis arctowskii 0.3 Delphinornis larseni 0.08 Waimanu manneringi Delphinornis gracilis 0.93 Perudyptes devriesi 0.49 0.39 Waimanu tuatahi 0.16 0.35 Marambiornis exilis Mesetaornis polaris 0.32 Anthropornis nordenskjoeldi 0.22 Anthropornis grandis 0.11 Palaeeudyptes klekowskii 0.47 Palaeeudyptes gunnari 0.53 0.32 Burnside “Palaeudyptes” Inkayacu paracasensis Icadyptes salasi Pachydyptes ponderosus 0.37 0.24 Palaeeudyptes antarcticus 0.11 0.35 Kairuku grebneffi 0.37 Kairuku waitaki Archaeospheniscus lowei 0.15 0.15 Duntroonornis parvus 0.14 0.1 Paraptenodytes antarcticus Archaeospheniscus lopdelli 0.08 Palaeospheniscus patagonicus 0.87 Eretiscus tonnii Marplesornis novaezealandiae 0.26 Platydyptes marplesi Pygoscelis adeliae 0.71 0.26 Platydyptes novaezealandiae 1 Pygoscelis papua Palaeospheniscus biloculata 1 Pygoscelis antarctica 0.05 0.82 0.58 0.13 Pygoscelis grandis Palaeospheniscus bergi 0.31 Aptenodytes patagonicus 1 Aptenodytes forsteri

0.77 Eudyptula minor 0.16 Spheniscus urbinai 0.5 Spheniscus megaramphus Spheniscus mendiculus 1 Spheniscus muizoni Spheniscus humboldti 1 0.28 0.39 Spheniscus magellanicus 0.99 Spheniscus demersus Megadyptes antipodes Eudyptes sclateri 0.95 0.15 Madrynornis mirandus 1 Eudyptes schlegeli 1 Eudyptes chrysolophus 0.46 Eudyptes robustus 1 Eudyptes pachyrhynchus 0.32 Eudyptes moseleyi 1 Eudyptes filholi 1 Eudyptes chrysocome Paleocene Eocene Oligocene Miocene Plio. Plei. Paleogene Neogene Qu.

60 50 40 30 20 10 0 Age (Ma) Gavryushkina et al. 2016. Sys Bio