Contingency and determinism in replicated adaptive radiations of island lizards

Science

Volumn 279, Issue 5359, 1998, Pages 2115-2118

  Losos, Jonathan B ^a   Jackman, Todd R ^a   Larson, Allan ^a   De Queiroz, Kevin ^b   Rodríguez Schettino, Lourdes ^c  

^a WASHINGTON UNIVERSITY   (United States)

^b DEPARTMENT OF PALEOBIOLOGY   (United States)

^c INSTITUTO DE ECOLOGÍA Y SISTEMÁTICA   (Cuba)

Author keywords

[No Author keywords available]

Indexed keywords

ADAPTIVE RADIATION; EVOLUTIONARY RADIATION; LIZARD;

ARTICLE; DNA SEQUENCE; EVOLUTION; LIZARD; MORPHOMETRICS; NONHUMAN; PHYLOGENY; PRIORITY JOURNAL;

GREATER ANTILLES; WEST INDIES;

ANOLIS;

EID: 0032571568     PISSN: 00368075     EISSN: None     Source Type: Journal    
DOI: 10.1126/science.279.5359.2115     Document Type: Article

References (36)

1. S. J. Gould, Paleobiology 1, 2 (1985).
2. J. E. Cadle and H. W. Greene, in Species Diversity in Ecological Communities: Historical and Geographical Perspectives, R. Ricklefs and D. Schluter, Eds. (Univ. of Chicago Press, Chicago, 1993), pp. 281-293.
3. M. Travisano, J. A. Mongold, A. F. Bennett, R. E. Lenski, Science 267, 87 (1995).
4. S. J. Gould, Wonderful Life: The Burgess Shale and the Nature of History (Norton, New York, 1989).
5. G. H. Orians and R. T. Paine, in Coevolution, D. Futuyma and M. Slatkin, Eds. (Sinauer, Sunderland, MA, 1983), pp. 431-458.
6. J. A. Wiens, The Ecology of Bird Communities (Cambridge Univ. Press, Cambridge, 1989).
7. R. E. Ricklefs and D. Schluter, Eds., Species Diversity in Ecological Communities: Historical and Geographical Perspectives (Univ. of Chicago Press, Chicago, 1993).
8. W. J. Bock, in Major Patterns in Vertebrate Evolution, M. K. Hecht, P. Goody, B. Hecht, Eds. (Plenum, New York, 1977), pp. 57-82; P. H. Harvey and M. D. Pagel, The Comparative Method in Evolutionary Biology (Oxford Univ. Press, Oxford, 1991).
9. W. J. Bock, in Major Patterns in Vertebrate Evolution, M. K. Hecht, P. Goody, B. Hecht, Eds. (Plenum, New York, 1977), pp. 57-82; P. H. Harvey and M. D. Pagel, The Comparative Method in Evolutionary Biology (Oxford Univ. Press, Oxford, 1991).
10. E. E. Williams, in Lizard Ecology: Studies of a Model Organism, R. B. Huey, E. R. Pianka, T. W. Schoener, Eds. (Harvard Univ. Press, Cambridge, MA, 1983), pp. 326-370.
11. T. C. Moermond, Behaviour 70, 147 (1979); Ecology 60, 152 (1979).
12. T. C. Moermond, Behaviour 70, 147 (1979); Ecology 60, 152 (1979).
13. J. B. Losos, Ecol. Monogr. 60, 369 (1990).
14. We measured live adult mates of 46 species of Greater Antillean Anolis. Species were assigned to an ecomorph class a priori on the basis of morphological, behavioral, and habitat data. The following measurements were made of (usually) 10 to 15 individuals of each species: mass; number of subdigital lamellae on the second and third phalanges of pedal digit IV; and snout-to-vent length (svl), forelimb length, hindlimb length, and tail length. Mean values were used for each species. To remove the effects of body size, interspecific regressions were conducted for each variable against svl (all variables were In transformed); residual values were used in subsequent analyses. To reduce the dimensionality of the data, a principal components (PC) analysis [H. Hotelling. J. Educ. Psychol. 24, 417 (1933); H. H. Harman, Modern Factor Analysis (Univ. of Chicago Press, Chicago, ed. 3, 1976)] was conducted on the residual variables; In svl was also included so that the analysis considered both size and shape. The first four PC axes accounted for 95.8% of the variance. We then used the unweighted paired group method using arithmetic averages (UPGMA) [ R. R. Sokal and C. D. Michener, Univ. Kans. Sci. Bull. 38, 1409 (1958)] to visualize the similarity relationships of species in a morphological space defined by these four PC axes.
15. We measured live adult mates of 46 species of Greater Antillean Anolis. Species were assigned to an ecomorph class a priori on the basis of morphological, behavioral, and habitat data. The following measurements were made of (usually) 10 to 15 individuals of each species: mass; number of subdigital lamellae on the second and third phalanges of pedal digit IV; and snout-to-vent length (svl), forelimb length, hindlimb length, and tail length. Mean values were used for each species. To remove the effects of body size, interspecific regressions were conducted for each variable against svl (all variables were In transformed); residual values were used in subsequent analyses. To reduce the dimensionality of the data, a principal components (PC) analysis [H. Hotelling. J. Educ. Psychol. 24, 417 (1933); H. H. Harman, Modern Factor Analysis (Univ. of Chicago Press, Chicago, ed. 3, 1976)] was conducted on the residual variables; In svl was also included so that the analysis considered both size and shape. The first four PC axes accounted for 95.8% of the variance. We then used the unweighted paired group method using arithmetic averages (UPGMA) [ R. R. Sokal and C. D. Michener, Univ. Kans. Sci. Bull. 38, 1409 (1958)] to visualize the similarity relationships of species in a morphological space defined by these four PC axes.
16. We measured live adult mates of 46 species of Greater Antillean Anolis. Species were assigned to an ecomorph class a priori on the basis of morphological, behavioral, and habitat data. The following measurements were made of (usually) 10 to 15 individuals of each species: mass; number of subdigital lamellae on the second and third phalanges of pedal digit IV; and snout-to-vent length (svl), forelimb length, hindlimb length, and tail length. Mean values were used for each species. To remove the effects of body size, interspecific regressions were conducted for each variable against svl (all variables were In transformed); residual values were used in subsequent analyses. To reduce the dimensionality of the data, a principal components (PC) analysis [H. Hotelling. J. Educ. Psychol. 24, 417 (1933); H. H. Harman, Modern Factor Analysis (Univ. of Chicago Press, Chicago, ed. 3, 1976)] was conducted on the residual variables; In svl was also included so that the analysis considered both size and shape. The first four PC axes accounted for 95.8% of the variance. We then used the unweighted paired group method using arithmetic averages (UPGMA) [ R. R. Sokal and C. D. Michener, Univ. Kans. Sci. Bull. 38, 1409 (1958)] to visualize the similarity relationships of species in a morphological space defined by these four PC axes.
17. The UPGMA analysis indicates that members of each of the ecomorph classes cluster together in morphological space. UPGMA and other clustering methods sometimes distort similarity relationships [K. de Queiroz and D. A. Good, Q. Rev. Biol. 72, 3 (1997)]; however, direct examination of the Euclidean distances (that is, the distance separating two species in the multivariate space determined by the four PC axes) confirms that every species is closer to a member of its own ecomorph class than it is to any member of any other ecomorph class. The probability that random assignment of species to ecomorph clusters will produce an arrangement in which the species are clustered by ecomorph class to the extent seen in the real data (that is, with no species out of place) is much less than 0.0001.
18. The uncertainty depends on which ecomorph is ancestral. For example, if the ancestor were a trunk anole, then no instances of trunk anole evolution would be required, but four instances of trunk-ground anoles would be required. By contrast, if the ancestor were a trunk-ground anole, only two instances of trunk anole evolution would be required, because this ecomorph occurs only on two islands, whereas trunk-ground and most other ecomorphs occur on all four islands. These figures also assume, conservatively, that one ecomorph is ancestral to all of the others and that each ecomorph evolved only once per island.
19. The amplification and sequencing of 1455 alignable base positions followed previously published procedures [J. R. Macey, A. Larson, N. B. Ananjeva, Z. Fang, T. J. Papenfuss, Mol. Biol. Evol. 14, 91, (1997); J. R. Macey, A. Larson, N. B. Ananjeva, T. J. Papenfuss, J. Mol. Evol. 44, 660 (1997)], except that annealing of primers for sequencing reactions was done at 53° to 61°C. Alignments of sequences encoding five transfer RNAs (Trp, Ala, Asn, Cys, and Tyr) were constructed on the basis of secondary structural models [Y. Kumazawa and M. Nishida, ibid. 37, 380 (1993); J. R. Macey and A. Verma, Mol. Phylogenet. Evol. 7, 272 (1997)]. All sequences have been deposited in GenBank.
20. The amplification and sequencing of 1455 alignable base positions followed previously published procedures [J. R. Macey, A. Larson, N. B. Ananjeva, Z. Fang, T. J. Papenfuss, Mol. Biol. Evol. 14, 91, (1997); J. R. Macey, A. Larson, N. B. Ananjeva, T. J. Papenfuss, J. Mol. Evol. 44, 660 (1997)], except that annealing of primers for sequencing reactions was done at 53° to 61°C. Alignments of sequences encoding five transfer RNAs (Trp, Ala, Asn, Cys, and Tyr) were constructed on the basis of secondary structural models [Y. Kumazawa and M. Nishida, ibid. 37, 380 (1993); J. R. Macey and A. Verma, Mol. Phylogenet. Evol. 7, 272 (1997)]. All sequences have been deposited in GenBank.
21. The amplification and sequencing of 1455 alignable base positions followed previously published procedures [J. R. Macey, A. Larson, N. B. Ananjeva, Z. Fang, T. J. Papenfuss, Mol. Biol. Evol. 14, 91, (1997); J. R. Macey, A. Larson, N. B. Ananjeva, T. J. Papenfuss, J. Mol. Evol. 44, 660 (1997)], except that annealing of primers for sequencing reactions was done at 53° to 61°C. Alignments of sequences encoding five transfer RNAs (Trp, Ala, Asn, Cys, and Tyr) were constructed on the basis of secondary structural models [Y. Kumazawa and M. Nishida, ibid. 37, 380 (1993); J. R. Macey and A. Verma, Mol. Phylogenet. Evol. 7, 272 (1997)]. All sequences have been deposited in GenBank.
22. The amplification and sequencing of 1455 alignable base positions followed previously published procedures [J. R. Macey, A. Larson, N. B. Ananjeva, Z. Fang, T. J. Papenfuss, Mol. Biol. Evol. 14, 91, (1997); J. R. Macey, A. Larson, N. B. Ananjeva, T. J. Papenfuss, J. Mol. Evol. 44, 660 (1997)], except that annealing of primers for sequencing reactions was done at 53° to 61°C. Alignments of sequences encoding five transfer RNAs (Trp, Ala, Asn, Cys, and Tyr) were constructed on the basis of secondary structural models [Y. Kumazawa and M. Nishida, ibid. 37, 380 (1993); J. R. Macey and A. Verma, Mol. Phylogenet. Evol. 7, 272 (1997)]. All sequences have been deposited in GenBank.
23. The most parsimonious phylogeny was estimated using PAUP* [D. L. Swofford, Phylogenetic Analysis. Using Parsimony, version 4.0.0d60 (Smithsonian Institution, Washington, DC, 1997)] with 100 heuristic searches using random addition of sequences and tree bisection and reconnection branch swapping (TBR). To estimate the maximum-likelihood tree, we used the 493 trees within five steps of the most parsimonious tree as starting trees, performing a heuristic search with the TBR branch-swapping option in PAUP* with the following parameters: empirical base frequencies, a transition-transversion ratio of 2 (estimated from the most parsimonious tree as 2.049 and rounded down), and equal rates of substitution among sites. Although we did not include all ecomorph species in this analysis, almost all non-sampled species are closely related to a sampled member of the same ecomorph class from the same island. The two exceptions are rare species from Hispaniola (A. darlingtoni and A. koopmani) that may not be closely related to members of the same ecomorph class on that island [ K. L. Burnell and S. B. Hedges, Caribb. J. Sci. 26, 7 (1990)]. Hence, depending on the phylogenetic relationships of these two species, we may underestimate by two both the number of evolutionary origins of ecomorphs and the instances in which the same ecomorph evolved multiple times on the same island.
24. The most parsimonious phylogeny was estimated using PAUP* [D. L. Swofford, Phylogenetic Analysis. Using Parsimony, version 4.0.0d60 (Smithsonian Institution, Washington, DC, 1997)] with 100 heuristic searches using random addition of sequences and tree bisection and reconnection branch swapping (TBR). To estimate the maximum-likelihood tree, we used the 493 trees within five steps of the most parsimonious tree as starting trees, performing a heuristic search with the TBR branch-swapping option in PAUP* with the following parameters: empirical base frequencies, a transition-transversion ratio of 2 (estimated from the most parsimonious tree as 2.049 and rounded down), and equal rates of substitution among sites. Although we did not include all ecomorph species in this analysis, almost all non-sampled species are closely related to a sampled member of the same ecomorph class from the same island. The two exceptions are rare species from Hispaniola (A. darlingtoni and A. koopmani) that may not be closely related to members of the same ecomorph class on that island [ K. L. Burnell and S. B. Hedges, Caribb. J. Sci. 26, 7 (1990)]. Hence, depending on the phylogenetic relationships of these two species, we may underestimate by two both the number of evolutionary origins of ecomorphs and the instances in which the same ecomorph evolved multiple times on the same island.
25. Those 55 species included the outgroups Diplolaemus darwini and Polychrus acutirostris. Ecomorph species were a subset of those used in the morphometric analysis, with the addition of the Hispaniolan twig anole A. sheplani.
26. To test a hypothesis of ecomorph monophyly, we used PAUP* to find the shortest tree with the constraint that an ecomorph class arose once and did not subsequently give rise to members of any other ecomorph class. The Wilcoxon signed-ranks test was then used to investigate whether the most parsimonious tree required significantly fewer changes than the tree constrained to show monophyly of an ecomorph class [A. Templeton, Evolution 37, 221 (1983);
27. A. Larson, in Molecular Ecology and Evolution: Approaches and Applications, B. Schierwater, B. Streit, G. P. Wagner, R. De Salle, Eds. (Birkhäuser, Basel, Switzerland, 1994), pp. 371-390]. Each ecomorph class was tested separately. We also determined the phylogeny that required the fewest instances of ecomorph evolution and was not significantly longer than the most parsimonious tree. PAUP* was used to find the shortest tree that included as sister taxa a pair of species (or clades) that were inferred to have evolved the same ecomorph independently on the most parsimonious tree. This constraint generally lowered the number of transitions among ecomorphs from 17 to 16; PAUP* searched all trees constrained in this manner and found the one requiring the fewest DNA changes. This tree was then compared to the most parsimonious tree with the use of the Wilcoxon signed-ranks test. Additional constraints were applied in the same manner to further reduce the number of origins of ecomorphs until all trees requiring fewer transitions were significantly longer than the most parsimonious tree. The same hypotheses were tested with the Kishino-Hasegawa test, comparing the maximum-likelihood tree to trees constrained as above
28. [H. Kishino and M. Hasegawa, J. Mol. Evol. 29, 170 (1989)].
29. To estimate the number of evolutionary transitions among ecomorph classes, we reconstructed ancestral states using MacClade [W. P. Maddison and D. R. Maddison, MacClade: Analysis of Phylogeny and Character Evolution, version 3.0 (Sinauer, Sunderland, MA, 1992)]. Ecomorph state was treated as a partially ordered character, with the number of steps required for a transition from one ecomorph to another corresponding to the distance in morphological space between the centroids for the two ecomorph classes. Treating the states as unordered yielded the same estimate of the number of instances of ecomorph evolution inferred from this phylogeny. Nonecomorph species (21 species), most of which do not occur in the Greater Antilles, were not included in the ecomorph reconstructions. Thus, this method estimates the number of transitions among ecomorphs conservatively, because all ancestral nodes are assigned to one of the ecomorph classes.
30. All trees within four steps of the most parsimonious tree also lead to the inference of 17 transitions, although reconstructions for particular ancestral nodes differ. Shortest trees requiring 15 or 16 transitions are not statistically significantly longer than the most parsimonious tree, but those requiring 14 transitions are (Table 1); the likelihood analysis yields similar results.
31. To investigate whether the phylogenetic topology of ecomorph evolution was the same on all four islands, we examined the four ecomorphs present on all four islands. Figure 1C shows that each island exhibits a different topology. For each island separately, we examined each of the 15 possible topologies for four taxa to determine whether any topology is either the most parsimonious tree or is statistically indistinguishable from the most parsimonious tree for all four islands. If no topology met this criterion, then the data would indicate that ecomorph evolution did not occur in the same manner on all islands. For each island, we found the most parsimonious tree conforming to each of the 15 possible phylogenetic topologies for the four ecomorphs (in the case of the two twig anoles on Hispaniola, we used A. insolitus because A. sheplani is the sister taxon of Cuban twig anoles and is nested within a clade of Cuban species) using the "backbone constraints" option in PAUP*, which constrains the relationships of a subset of the taxa but allows the remaining taxa to occur anywhere on the tree (that is, the subset of constrained taxa does not necessarily form a monophyletic group, but the relationships among these taxa must conform to the constraint). We compared each of these trees to the most parsimonious tree (Fig. 1B) using the Wilcoxon signed-ranks test. In addition, we compared the maximum-likelihood tree with each constraint tree using the Kishino-Hasegawa test. Each of the 15 possible ecomorph topologies was rejected for at least one island. Hence, we conclude that the topology of ecomorph evolution differed among islands. In addition, when ancestral ecomorph states were reconstructed with parsimony, each island exhibited a different order of ecomorph evolution.
32. D. J. Irschick, L. J. Vitt, P. A. Zani, J. B. Losos, Ecology 78, 2191 (1997).
33. J. B. Losos, Annu. Rev. Ecol. Syst. 25, 467 (1994).
34. E. E. Williams, Evol. Biol. 6, 47 (1972).
35. J. H. Zar, Biostatistical Analysis, (Prentice-Hall, Englewood Cliffs, NJ, ed. 2, 1984).
36. This work was supported by NSF, the National Geographic Society, and the David and Lucile Packard Foundation.