UC Davis Talk Bapst 05-11-17

transcript

From Cambrian Trilobites to Modern Brachiopods:

New Approaches in Phylogenetic Paleobiology

David Bapst, PhD

University of California Davis

It All Starts With Morphology

Bapst et al., 2012, PNAS

Paleontologist relies primarily on morphology for discriminating taxon units and inferring relationships

• In many groups in the fossil record, we find specimens of different geologic age that share identical or very similar morphological features

Morphology, Taxa, Trees

• In many groups in the fossil record, we find specimens of different geologic age that share identical or very similar morphological features

• We use those features to define morphospeciesthat can span millions of years, to infer relationships among lineages (phylogenies)

Systematic Paleontology

• Code taxa for discrete morphological traits (characters)

• Cladistic analyses use maximum-parsimony methods, which attempt to find cladograms (undated trees) implying the fewest character changes

• More generally, we refer to evolutionary trees as phylogenies, particularly if they are dated trees

• Taxa closer together on a cladogram share more traits

~0.5 mm

Maletz

; C.E. M

itchell

Phylogenies let us Answer Questions

• When do new species or lineages first evolve?

• Which species are ancestors and which are descendants?

• If organisms have similar morphologies, is that similarity due to a shared evolutionary history?

• Or does shared morphology reflect similarities in ecology and function?

These questions are key to understanding the history of life

New Approaches to Phylogenetic Paleobiology

1. Use combined molecular and morphological datasets in living brachiopods to determine which dataset better reflects evolutionary history

2. Reconstructing relationships between direct ancestors and their descendants across the tree of life

Convergence: When Morphological Similarity Misleads Us

Maletz and Zhang, 2003

~0.5 mm

1. Combined Data Phylogenetics in Rhynchonellids

Mitchell, 1991

Silurian

Ordovician

~0.5 mm

Testing Morphological Phylogenetics

• For most fossil groups, we are forced to assume features reflect shared evolutionary history (homology)

• Test the reliability of morphology-based phylogenies by comparing to independent molecular data for extant taxa• Two datasets will infer incongruent phylogenies, if one or both

datasets have poor phylogenetic signal

• A fossil-rich group with extant diversity:• Rhynchonellida: 19 living genera (500+ extinct genera)

• Important to understand morphology to place extinct lineages

Rhynchonellida• Articulated brachiopods with spirolophe lophophores

supported on crura, features of which are considered important to systematics of the group

• 10/19 extant genera live at bathyal or abyssal depths and are difficult to collect

Hard to visualize how these differ… let’s look at tanglegrams

Note: All topologies in this section are single MPTs or majority-rule / half-compat summaries

Morphology(56 Characters, Reweighted Parsimony MPT)

Schreiber et al. 2013

18S and 28S rDNA(3435 base pairs, Bayesian)

Cohen & Bitner 2013

Scattered Superfamilies

Why Phylogenetic Analyses Disagree

• Differences in character data used (molecular & morphology)

• Differences in taxa included

• Differences in outgroups used for rooting

Outgroups Determine Where We Root Trees

• Phylogenetic analyses need to include more than just the group of interest (the ingroup taxa)

• Need to include ‘outgroup’ taxa that are strongly assumed to outside the ingroup to give directionality (‘rooting’ the tree)

Mitchell et al., 2013

Outgroups

Trees have no directionalitywithout a root

Cohen & Bitner 2013

• Revise morphological data to include all rhynchonellides that we have rDNA data for

• Use non-rhynchonellide outgroups• Compare molecular and morphological analyses

using a standardize dataset

Bapst, Schreiber & Carlson, in press, Syst. Biol.1. Combined Data Phylogenetics in Rhynchonellids

with Sandy Carlson & Holly Schreiber

Is incongruence due to differences in character data?

Cohen & Bitner 2013

Revised Morphological Matrix (66 characters, Bayesian)

This Study

1. Combined Data Phylogenetics in Rhynchonellids Bapst, Schreiber & Carlson, Syst. Biol.

InarticulateBrachiopods

Terebratulids

Rhynch.

Morphology-only analyses don’t even distinguish the rhynchonellides and the

terebratulids as separate clades

Cohen & Bitner 2013

This Study

InarticulateBrachiopods

Rhynch.

Terebratulids

• Which dataset carries the most signal of shared evolutionary history?

• What if the two datasets both have hidden support for a third alternative topology?

• We can answer both by combining our morphological and molecular datasets, and analyzing them simultaneously

Why do the datasets still disagree?

Bapst, Schreiber & Carlson, accepted with revisions, Syst. Biol.1. Combined Data Phylogenetics in Rhynchonellids

Quantitative Phylogenetic Methods• Maximum Parsimony (PAUP)

• Bayesian (MrBayes)• Model-based phylogenetics, uses likelihood and prior

probability distributions to sample topologies via an MCMC

• Treat different data types as distinct partitions• Same model of sequence change as Cohen & Bitner 2013

• Two morph model configurations: (1) relaxed assumptions, versus (2) strict assumptions, to maximize information content

• Unlike maximum-parsimony which counts character changes, Bayesian phylogenetics depend on likelihood of character data, not directly connected to differences in # of characters

• Applied both to combined datasets, all taxa versus only those taxa with rDNA

Quantitative Phylogenetic Methods• Maximum Parsimony (PAUP)

• Bayesian (MrBayes)• Model-based phylogenetics, uses likelihood and prior

probability distributions to sample topologies via an MCMC

• Treat different data types as distinct partitions• Same model of sequence change as Cohen & Bitner 2013

• Two morph model configurations: (1) relaxed assumptions, versus (2) strict assumptions, to maximize information content

• Unlike maximum-parsimony which counts character changes, Bayesian phylogenetics depend on likelihood of character data, not directly connected to differences in # of characters

• Applied both to combined datasets, all taxa versus only those taxa with rDNA

…adds up to 12 analyses! How to compare?

Summarizing phylogenetic analyses often results in polytomies

• When character data equally support multiple hypotheses of relationships, analyses return multiple trees, which can be very large

• Summary trees collapse those areas of uncertainty as nodes that aren’t fully bifurcating

• Referred to as polytomies, or a loss of phylogenetic resolution

So We Invented A New Measure…

• A measure to quantify how different two summary trees are

• Most metrics treat polytomiesin summary trees as a difference, even though polytomies reflect uncertainty

• Contradiction difference (CD) simply measures incongruence between two summary trees

Pairwise Contradiction Difference

• Values at one mean trees have no similarity

• Values at zero mean trees have no disagreement

Pairwise Contradiction Distance

The original molecular and morphological analyses produced completely contradictory trees

Reanalysis in this Study

Combined DatasetsMrBayes, Maximum Information

Shared Taxa Only

Molecular-Only Agrees With Morphology+Molecular

CD = 0.06

Combined DatasetsMrBayes, Maximum Information

CD = 0.31

Morphology-Only disagrees with Morphology+Molecular

Poor Phylogenetic Signal of Morphological Data Implies Convergence

• Bayesian morphological+molecular analyses were all congruent with each other and the molecular-only tree

• Morphology-only analyses, across different methods and sets of taxa, disagreed with each other

If the molecular-only and morphological+molecular analyses are right, are there any morphological characters that are good indicators of relatedness?

• Calculate consistency indices for each character

• Characters that vary within the Rhynchonellida all have lots of convergence when mapped on the tree from the molecular-only analysis• Including characters

classically used to distinguish traditional taxonomic groups

Classic diagnostic traits for taxa within Rhynch.

Characters for distinguishing articulate and inarticulate taxa

Remainder

More Informative

Not the Same Story for Short-Loop Terebratulids

Molecular Only Analysis(2574 Sites, Bayesian)

Morphology Only Analysis(85 Characters, Bayesian)

1. Combined Data Phylogenetics in Rhynchonellids Carlson, Bapst & Schreiber, in prep.

Could fossil data improve Rhynchonellide Relationships?• Code large number of fossil taxa within Rhynchonellida to

more finely reconstruct evolutionary lineages and resolve the appearance of convergence across modern groups

Want to read more?

This just out…

2. Ancestor-Descendants in the Fossil Record

The Question of Ancestors in the Fossil Record

The problem is, very rarely can we read the fossil record as literally as this

How do we infer the relationships among ancestors & their descendants,

given the incompletenessof the fossil record?

The Effects of UnsampledLineages In the Fossil Record

• Incompleteness of the rock record means we don’t observe all taxa during all intervals

• Limits our ability to reconstruct precisely when lineages originated, and what the relationships are between morphotaxa

• Are some morphotaxa ancestral to other morphotaxa?

Maletz & Mitchell (1996)

Qualitative Reconstructions of Ancestor-Descendant

Relationships

Kennett and Srinivasan (1983) from Pearson (1998)

Inferring Ancestors in Stratocladistics

Bloch et al., 2001

• Stratocladistic methods treat implied gaps in the fossil record as interchangeable with morphological changes under maximum parsimony (Fisher, 1991; 1994)

• Need formal model of diversification and incompleteness in the fossil record to calculate likelihood of stratigraphic gap of a given duration

Three-Rate Calibrated Time-Scaling (cal3)• Takes an existing undated cladogram, sample potential

divergence dates for nodes under a likelihood function of diversification and incompleteness of the fossil record

• Treats taxa as persistent morphotaxa, allowing for different types of ancestor-descendant relationships based on the overlap of their stratigraphic durations

• Created and implemented in paleotree for R (Bapst, 2012)

2. Ancestor-Descendants in the Fossil Record Bapst, 2012; Bapst, 2013; Bapst, 2014

‘Budding’ Cladogenesis

Anagenesis

Modes of Differentiation

Notice that budding can look like anagenesis (but not vice versa)

in an incomplete record

Anagenesis

‘Budding’

Ancestor-Descendant (AD) Relationships in Cambrian pterocephaliid trilobites

• Hopkins (2011) reviewed qualitative ancestor-descendant pairs previously suggested for this group

• Does cal3 find support for those pairs, and does it match the mode inferred by earlier studies?

• Apply cal3 to the cladistic analysis from Hopkins (2011)• Obtain 100 dated phylogenies, quantify support for a given

AD pair as the proportion of trees on which that pair is inferred

Bapst & Hopkins, 2017, Paleobio.

with Melanie Hopkins

• Each pair is a stacked barplot

• Dots indicate putative pairs

• Support for alla priori AD pairs, & a few extra

• cal3 finds very little support for anagenesis

• Support for budding suggests globally instantaneous origins of new morphotaxa

Bapst & Hopkins, 2017, Paleobio.2. Ancestor-Descendants in the Fossil Record

Bayesian sampled-ancestor tip-dating• Infer dated phylogenies from character and stratigraphic

data simultaneously in a Bayesian MCMC, under likelihood models for morphological change and the fossilized-birth-death model (Heath et al., 2014)

• Taxa are instantaneous points but can be placed as sampled-ancestors (Gavryushkina et al., 2014)

Tip-Dating with Mesozoic Theropods• We used a somewhat infamous dataset to compare

tip-dating methods with cal3, for ancestor-descendant relationships, divergence dating, estimating evolutionary rates, etc...• Do the methods agree?

• The support for particular taxa to be probable ancestors were fairly correlated across methods

• So… Is Archaeopteryx really the ancestral bird?

Bapst, Wright, Matzke & Lloyd, 2016; Biol. Lett.2. Ancestor-Descendants in the Fossil Record

withApril Wright Graeme LloydNick Matzke

• Significant rank-order pair-wise correlations of ancestral placement between methods• Strongest between MrBayes

and BEAST2

• Considerable differences despite similar model

• Median # of ancestors per tree for tip-dating = 1-2

• With cal3 (using entire taxon durations) = 17• Always buddingBeast2

(PP)MrBayes(PP)

cal3(prop)

Bapst, Wright, Matzke & Lloyd, 2016

Whither the Ancestral Bird?

• Archaeopteryx rarely placed as a sampled ancestor

• Never placed as ancestor on lineage leading to extant birds, but rather as a sampled ancestor to itssister taxon / possible synonym Wellnhoferia

Bapst, Wright, Matzke & Lloyd, 2016

Individual Occurrences as Operational Taxon Units

Hunt, 2007

20 Poseidonomicus species

Time (Mya)

Tip-Dating Ostracod Occurrences

3 previously defined morphospecies are paraphyletic (budding!)

12 sampled ancestors

Outgroups

Tip-Dating with FADs and LADs• Ancestors among

Devonian TerebratulideBrachiopods?

• 9 paraphyletic genera among 72 in-group taxa

• 21 sampled ancestors (FADs ancestral to LADs)

FAD = First Appearance DatumLAD = Last Appearance Datum

A New Era of Paleo-Phylogenetics

Thanks for listening! Questions?

• The rapid development and deployment of new methods allows us to leverage phylogenies to better understand evolution in deep time

• Bayesian phylogenetics allows us to model the potential complexities of morphological evolution

• Methods like cal3 and tip-dating take into account what we know about incompleteness of the rock record, and likely existence of ancestral taxa to create dated phylogenies from the fossil record

This research was funded by NSF grant EAR-1147537.

Gaps in Densely-Sampled Fossil Records

Maletz and Zhang, 2003; Vandenberg, 2003; C.E. Mitchell

• Closest relatives separated by a 15 to 20 million year gap in this lineage:

• Were the intermediates living somewhere else? Open ocean?

BergstromgraptusMiddle Darrwillian

SinoretiograptusLatest Katian

Branching Times from Time-Scaled Phylogenies

Lloyd, Bapst, Friedman& Davis, 2016 Biol. Lett.

cal3 and a non-parametic dating method agree: accounting for incompleteness of their records, dinosaurs likely diverged millions of years earlier than suggested by the fossil record

UC Davis Talk Bapst 05-11-17

Science