Some Aspects of Ancestor Reconstruction in the Study of Floral Assembly Christopher Hardy Y James C. Parks Herbarium Dept. of Biology Millersville University of Pennsylvania 7 Jan 2009 Floral Assembly Group Meeting The National Evolutionary Synthesis Center, Durham, NC, USA X 2D Floral Morpho-Space in Commelinaceae 24 categorical characters Non-Metric Multidimensional Scaling 1 Some Aspects of Ancestor Reconstruction in the Study of Floral Assembly 1. Example of Ancestor Reconstruction Methods from Commelinaceae (spiderworts & dayflowers). 2. Assumptions of Parsimony and Model-Based Methods -Emphasis on Branch Lengths. -Programs for Ancestor Reconstruction. 3. Accounting for Uncertainty. X 2D Floral Morpho-Space in Commelinaceae 24 categorical characters Non-Metric Multidimensional Scaling 2 1 1. Example of Ancestor Reconstruction Methods from Commelinaceae (spiderworts & dayflowers). 3 Y X 2D Floral Morpho-Space in Commelinaceae 24 categorical characters Non-Metric Multidimensional Scaling 4 2 Flowers exhibit a high degree of synorganization. Y Cyanotis Aneilema Plowmanianthus © R.B. Faden X Commelina africana (South Africa) Geogenanthus Commelina 5 Flowers exhibit a high degree of synorganization. the intimate (spatial and functional) connection of organs of the same or different types to form a functional apparatus. Endress (1994). 6 3 This implies that floral forms will not be uniformly distributed in morphospace: and they are not. Y Cyanotis Aneilema Plowmanianthus © R.B. Faden X Commelina africana (South Africa) Geogenanthus Commelina 7 This pattern is the product of evolution via natural selection. Y Cyanotis Aneilema Plowmanianthus © R.B. Faden X Commelina africana (South Africa) Geogenanthus Commelina 8 4 Thus, floral evolution is a process shaped by natural selection. Y X 9 10 5 11 12 6 13 14 7 15 16 8 17 18 9 Patterns of floral morphological / biological diversity (the X & Y dimensions) are the products of a history (add the Z dimension). How does one get at that history? 19 Cracking that history: 1. Cladistics. •Tree based on rbcL, 2. DNA as an important ndhF, and 26S (TNT) tool. 3. Theoretical & •Ultrametric (MULTI-DIVTIME) computational advances for •Fossil Pollia provides calibration estimating div. times & (12.2 +/- 5 mya) ancestors. •Ancestor reconstruction (WINCLADA, MULTISTATE) 20 10 Reuniting the cladist with its not-so-distant cousin the pheneticist! Phylogenetic (cladistic) diversification of the Commelinaceae through floral morphological (phenetic) space & time. -useful for inferring the origins of complex floral forms and biology (OUR GOAL!). 21 e.g., floral complexity and buzz pollination in Cochliostema (Commelinaceae) Epiphyte from NW South America lowland rainforests. Large, nectarless, fragrant, asymmetric flowers attract Xylocopine and Euglossine bees. 3 connate fertile stamens. Filament extensions 3 concealed, spirally coiled anthers. conceal & conserve pollen. Stamen hairs mimic pollen mass. Analogous to (convergent with) poricidal anther theca. 1 cm 22 11 What does ancestor reconstruction tell us about the evolution of buzz pollination in Cochlisotema? 5. ??? 4.3. Lower androecial reduction as deception intensifies.intensifies. 3. Hairs evolve as pollen-mimics. Pollinator deception ensues. 2. Filament hairs 1. Six arise in glabrous Tradescantieae, stamens decrease pollen- ancestral. foraging efficiency. 23 There are limits to the phylogenetic approach. 5. ??? 4.3. Lower androecial reduction as deception intensifies.intensifies. 24 12 Stamen morphology in the phylogenetic vicinity of Cochliostema. 25 Stamen developmental morphology in the phylogenetic vicinity of Cochliostema. 26 13 5. Congenital fusion of filament hairs. Acquisition of new color & function. 4.3. Lower androecial reduction as deception intensifies.intensifies. 27 Important side note: There are limits to the phylogenetic approach & ancestor reconstruction. Phylogenetics and developmental studies can be reciprocally illuminating. 28 14 A fascinating story of adaptation & exaptation in Cochliostema. 4. Hairs evolve postgenital & congenital fusion, conceal pollen bearing anthers. 3. Lower androecial reduction as deception intensifies. 2. Hairs evolve as pollen-mimics. 1. Filament hairs arise in Tradescantieae, decrease pollen- foraging efficiency. 29 2. Assumptions of Parsimony and Model-Based Methods -Emphasis on Branch Lengths. -Programs for Ancestor Reconstruction. 30 15 Ancestor Reconstruction: The basic process. Phylogeny reconstruction Molecular Dating Ancestor Reconstruction & or similar analysis of Fossil Calibration character evolution (optional) 31 The first question: Should we choose Parsimony OR Model-based Methods? 32 16 How about: Parsimony AND Model-based Methods. 33 How about: Parsimony AND Model-based Methods. -Politically speaking: A pluralistic approach gives you more journal options (e.g., Syst Biol vs. Cladistics). 34 17 How about: Parsimony AND Model-based Methods. -Politically speaking: A pluralistic approach gives you more journal options (e.g., Syst Biol vs. Cladistics). -Cost-benefit analysis: Parsimony is quick, easy, costs nothing to do. Widely available in user-friendly packages. 35 How about: Parsimony AND Model-based Methods. -Politically speaking: A pluralistic approach gives you more journal options (e.g., Syst Biol vs. Cladistics). -Cost-benefit analysis: Parsimony is quick, easy, costs nothing to do. Widely available in user-friendly packages. -Common familiarity: Most evolutionary biologists have and still do use parsimony. 36 18 How about: Parsimony AND Model-based Methods. -Politically speaking: A pluralistic approach gives you more journal options (e.g., Syst Biol vs. Cladistics). -Cost-benefit analysis: Parsimony is quick, easy, costs nothing to do. Widely available in user-friendly packages. -Common familiarity: Most evolutionary biologists have and still do use parsimony. -Expanded Toolkit: Model-based methods are powerful & flexible tools to explore data and to estimate various parameters relating to the evolution of flowers. 37 How about: Parsimony AND Model-based Methods. -Politically speaking: A pluralistic approach gives you more journal options (e.g., Syst Biol vs. Cladistics). -Cost-benefit analysis: Parsimony is quick, easy, costs nothing to do. Widely available in user-friendly packages. -Common familiarity: Most evolutionary biologists have and still do use parsimony. -Expanded Toolkit: Model-based methods are powerful & flexible tools to explore data and to estimate various parameters relating to the evolution of flowers. -Paul O. Lewis (UConn, 2007): 38 19 FYI: A list of widely available programs. SIMMAP 1.0 (Bollback, 2006) BAYESTRAITS 1.0 ( Pagel & Meade, 2004 onwards) 39 FYI: A list of widely available programs. User-friendly GUIs SIMMAP 1.0 (Bollback, 2006) BAYESTRAITS 1.0 ( Pagel & Meade, 2004 onwards) 40 20 What are your assumptions when you choose one method over the other? Where different reconstructions exist, these are due to branch lengths. Parsimony Model-based Hardy, C.R. (2006) Reconstructing ancestral ecologies: challenges and possible solutions. Diversity and Distributions 12: 7-19. 97 41 What are your assumptions when you choose one method over the other? Implicit Assumption Explicit Assumption P(change) equal on all branches P(change) ~ branch length (i.e., branch lengths, if had, are not considered). Parsimony Model-based 42 21 What are your assumptions when you choose one method over the other? Implicit Assumption Explicit Assumption Punctuational Equilibrium? Gradual Evolution? Parsimony Model-based 43 What are your assumptions when you choose one method over the other? Implicit Assumption Explicit Assumption Punctuational Equilibrium? Gradual Evolution? Does this mean that models are inappropriate for ancestor reconstruction in adaptive radiations? Does that mean that such methods should not be used in cases where adaptation is thought to have occurred (i.e., in flower evolution)? Proteaceae example borrowed from Paul Lewis. 44 22 Types of branch lengths. Ultrametric branch lengths (e.g., following a molecular clock, Penalized Likelihood or MultiDivtime analysis): time Default assumption in most programs: probability of change is proportional to time (equilibria and punctuated change more difficult to detect). 45 Types of branch lengths. Raw genetic branch lengths (as seen on a “phylogram”): # substitutions Default assumption in most programs: degree of character evolution is proportional to degree of molecular evolution in the gene(s) analyzed. (punctuated morphological change during adaptive radiations difficult to detect). 46 23 Types of branch lengths. Equal branch lengths (e.g., all branches are assigned the same length): Default assumption in most programs: Change equiprobable on all branches. Parsimony-like. (Punctuated morphological change during adaptive radiations potentially easier to detect). 47 One of BayesTraits’ best kept secret: the kappa (k) parameter. Starting tree time k < 1.0 k = 1.0 k >1.0 (branch lengths (branch lengths remain as is) (branch lengths stretched in shortened, degree in proportion proportion to their original length) to their original length) 48 24 One of BayesTraits’ best kept secret: the kappa (k) parameter. Starting tree time k < 1.0 k = 1.0 k >1.0 (branch lengths (branch lengths remain as is) (branch lengths stretched in shortened, degree in proportion proportion to their original length) to their original length) How to use it (my suggestions): 1. LR Test: Does including k increase likelihood relative to not