napc 2014 constraint and convergence on a graptoloid supertree

NAPC 2014

David W. Bapst (SDSMT, UC Davis) & Charles E. Mitchell (SUNY Buffalo)

Functional Constraint in Graptoloid Morphology?

Figures from Bulman, 1970; Mitchell, 1990; Maletz et al., 2009; etc.

Using Phylogeny to Test for Constraints

Convergent traits give opportunity to study biotic and abiotic drivers of trait acquisition across clades

Convergence alone indicates potential

…but isn’t evidence of ecomorph.function or other constraints

Randomly evolving traits will converge

Simulate characters on the phylogeny

Test how often some aspect of the observed data occurs in simulationsWagner and

Erwin, 2006

Major Features Primary Stipe Number

Many, 4, 2, 1.5, 1

Scandency Not Scandent, Partly, Scandent

Reduced Periderm

Extrathecal Threads

Determinant Growth

Cladia

Figures from Bulman, 1970; Mitchell, 1990; Maletz et al., 2009; etc.

GraptoloidPhylogenetics

3D preservation provides detailed characters for cladistics

…but only some taxa: ~49% of genera are on published cladograms

Missing lots of unique colony morphologies

Informal Supertreeof 117 Genera

Trees from:Maletz et al., 2009 Mitchell et al., 2007 Storch et al., 2011Melchin et al., 2011Bates et al., 2005

Adding extra taxa to cladistic ‘skeleton’

Summarize our current knowledge of relationships using all types of evidence

Hierarchical placing of unanalyzed taxa

First, synapomorphies in cladistic analyses

If not, most recent taxonomic placement

Inclusive Summary of Relationships among 245 Graptoloid Genera

The under-analyzed Monograptidae: one single polytomy

Poorly preserved early graptoloids: mostly unresolved

Use soft polytomies (multifurcating nodes) for uncertainty in morphology or taxonomy

Character Analyses Minimum number of gains and losses: How much convergence?

Parsimony ancestral trait reconstruction

Resolve polytomies to minimize character gains under parsimony

Simulation Analyses of constraint

Simulate character evolution on observed phylogeny (randomly-resolved)

Use estimated transition matrices for each trait (assuming indep evolving)

…but no time-scaling?

Character Analyses Minimum number of gains and losses: How much convergence?

Parsimony ancestral trait reconstruction

Resolve polytomies to minimize character gains under parsimony

Simulation Analyses of constraint

Simulate character evolution on observed phylogeny (randomly-resolved)

Use estimated transition matrices for each trait (assuming indep evolving)

…but no time-scaling? Really?

Character Analyses Minimum number of gains and losses: How much convergence?

Parsimony ancestral trait reconstruction

Resolve polytomies to minimize character gains under parsimony

Simulation Analyses of constraint

Simulate character evolution on observed phylogeny (randomly-resolved)

Use estimated transition matrices for each trait (assuming indep evolving)

Analyzing Not Time-scaled Trees

# of character changes / evolutionary ‘length’ per branch: minimal reversals

Are we willing to assume relatively uniform sampling of taxa, nodes

Minimum Number of Transitions Polytomies resolved to minimize

number of character transitions

Minimally, many indep ‘gains’ to derived states with few reversals

High evolvability?

Lack of reversals an indicator of species sorting for some traits?

‘Gains’ Reversals

# of Primary Stipes 15 0

Scandency 6 4

Reduced Periderm 6 0

Extrathecal Threads 3 0

Determinant Growth 6 0

Cladia 3 0 (Sort of a reversal for stipe #)

Relative to Primitive State:

Nonrandom Associations Between Characters Imbalance of by-genus

contingency table for pairs of discrete traits (Cramer’s V)

Proportion of simulations with imbalance as high or higher than observed value

Some traits aren’t evolving independent of each other

# of stipes and scandency

Extrathecal threads with red. periderm and det. growth

Constraints on trait change?

ScandencyReduced Periderm

Extrathecal Threads

Determinant Growth

Cladia

# of Primary Stipes 0.03 0.23 0.14 0.55 0.19

Scandency - 0.32 0.09 0.39 0.81Reduced Periderm - - < 0.01 0.15 0.25

ExtrathecalThreads - - - 0.03 0.34

Determinant Growth - - - - 0.4

But are these ecomorphological constraints?

W&E ‘06: Paleozoic gastropod taxa distributed among fewer morphotypesthan in simulations

Even when the number of potential char combos in simulation limited to the

observed # of combos

Implies ecomorpologicalconstraints, not ‘biotic’ architectural constraints

Observed

Wagner and Erwin (2006)

Ecomorphological Constraint on Graptoloids?

Using same traits except treating cladia as a return to many-stipes…

Observed frequencies are slightly higher than simulations without constraining the number of character combinations

95% Quantile

Ecomorphological Constraint on Graptoloids But when we constrain simulations to

observed number of combinations…

Simulations match observed frequency distribution almost perfectly

I.e., the distribution of taxa among combinations of these traits is closely predicted by a null model where some morphologies are simply unavailable …But not rejecting null ≠ accepting null

95% Quantile

Conclusions Large scale summaries of what we think we know about

relationships in a group can be useful, if we adequately explore phylogenetic uncertainty in our analyses

Derived states of these major morph innovations in graptoloids were repeatedly ‘gained’ independently, implying evolvability

Rarely reverse back… may reflect species sorting (for some)

We can reject these characters evolve independently

But based on the obs frequency of morphotypes, we can’t reject that convergence are due to intrinsic biotic constraints such as construnctional or developmental factors

Thanks to M. Foote, P. Smits, M. Pennell, E. King for useful discussions.

Characters Tied Across Branches…? Sets of traits ( ) seem to repeat in

certain clades … but not always in the same taxon in those clades

Some invisible ‘predictor’ trait?

Use mean patristic distance of taxon with one character to closest relative with other trait

Only works for binary traits

Proportion of simulations with low or lower mean pairwise patristic distance than observed

Extrathecal Threads

Determinant Growth

Cladia

Reduced Periderm < 0.01 0.09 0.31

Extrathecal Threads - 0.16 0.33

Determinant Growth - - 0.32

GlossograptidsCorynoides

Cryptograptus

RetiolitidsSpinograptus

Hard to reject null?

Informing Our Functional Morphology Analyses

Use physics to test a supposed function for morphology

Can test that X could do function Y

Cannot test whether X had any function

Need to start with a carefully chosen specific hypothesis of a specific function

Larger patterns of evolution can be useful for informing what we should investigate

Cladia-like abnormal Amplexograptus latus