Evolution, 55(4), 2001, pp. 677±683 EVOLUTIONARY RATES AND SPECIES DIVERSITY IN FLOWERING PLANTS TIMOTHY G. BARRACLOUGH1 AND VINCENT SAVOLAINEN2 1Department of Biology and NERC Centre for Population Biology, Imperial College at Silwood Park, Ascot, Berkshire SL5 7PY, United Kingdom E-mail: [email protected] 2Molecular Systematics Section, Jodrell Laboratory, Royal Botanic Gardens, Kew, Richmond Surrey TW9 3DS, E-mail: [email protected] Abstract. Genetic change is a necessary component of speciation, but the relationship between rates of speciation and molecular evolution remains unclear. We use recent phylogenetic data to demonstrate a positive relationship between species numbers and the rate of neutral molecular evolution in ¯owering plants (in both plastid and nuclear genes). Rates of protein and morphological evolution also correlate with the neutral substitution rate, but not with species numbers. Our ®ndings reveal a link between the rate of neutral molecular change within populations and the evolution of species diversity. Key words. Angiosperms, DNA, molecular evolution, speciation, species richness. Received July 17, 2000. Accepted October 31, 2000. Speciation is dependent on genetic change: changes at the Chase et al. (1993) based on DNA sequences of rbcL, a plastid DNA level allow populations to diverge and ultimately to gene encoding the large subunit of ribulose-1,5-biphosphate- form new species (Harrison 1991; Coyne 1992; Coyne and carboxylase/oxygenase (RUBISCO), we found evidence for Orr 1999). However, the relationship between rates of spe- a positive relationship between rates of DNA change and ciation and molecular evolution remains uncertain. Many au- species diversi®cation (Barraclough et al. 1996; Savolainen thors argue that rates of genetic change and speciation are and Goudet 1998). However, this result relied on a prelim- closely linked. For example, bursts of genetic change may inary phylogenetic analysis, in terms of the classi®cation used occur at speciation events, as a result of population structures and incomplete sampling of angiosperm families. Further- associated with speciation (Mayr 1963; Carson and Temple- more, subsequent power analyses raised serious doubts over ton 1984; Gavrilets and Hastings 1996; Templeton 1996). the robustness of the pattern (Savolainen and Goudet 1998). Alternatively, the divergence of new species may be driven Since then, comprehensive multigene and nonmolecular data by ``background'' rates of genetic change within populations have been produced for angiosperms, as well as a new clas- (Coyne 1994; Orr 1995). However, there may also be no si®cation based on the combined data (Watson and Dallwitz relationship between speciation and rates of genetic change, 1991, 1999; APG 1998; Soltis et al. 1999; Qiu et al. 1999). for example, if speciation is limited by the formation of iso- These data provide the opportunity for comprehensive and lated populations rather than by rates of genetic divergence robust tests of the correlation between the rate of DNA evo- (Allmon 1992), or if population structure at speciation has lution and species richness. no in¯uence on rates of genetic change (Coyne 1994). In In addition, we use the recent data to perform new analyses addition, the nature of DNA change typically involved in investigating the possible role of adaptive changes in ex- speciation remains unknown. Recent work highlights the role plaining the observed relationship. One explanation for a pos- of ecological and adaptive changes in promoting species di- itive relationship between sequence change and species num- vergence (Orr and Smith 1998; Rundle et al. 2000), but in- bers might be that genetic change at the chosen loci is as- trinsic factors affecting the origin of neutral genetic variation sociated with the adaptive radiation of angiosperm families. within lineages may also play a major role (Orr 1995). Re- For example, substitutions at the rbcL locus have been linked solving these issues is fundamental for understanding the link to differences in photosynthetic pathways among plants oc- between evolution occurring within populations and the or- cupying different habitats (Hudson et al. 1990). Hence, igin of species diversity (Williams 1992), but to date evidence changes at the rbcL locus might, in principle, be associated has been scarce (Mindell et al. 1989; but see Sanderson 1990). with the adaptive radiation of plant lineages among habitat Here we use sister-group comparisons to demonstrate a types. If this were the case, we might expect the rate of amino positive relationship between species richness and the rate acid substitutions (nonsynonymous mutations) at the rbcL of neutral molecular evolution in ¯owering plants. Flowering locus to display the strongest relationship. Our previous work plants (angiosperms) are an ideal study group for investi- did not investigate this possibility. Thus, we test whether gating these questions. Not only are they the product of one synonymous or nonsynonymous changes display the stronger of the major radiations of organisms on earth, but they have relationship with species numbers. We also perform related also been the focus of the most comprehensive molecular analyses comparing sites which differ in their effects on the systematics project carried out on any equivalent-sized group functional secondary structure of 18S rDNA (details in Meth- of organisms (Chase et al. 1993, Soltis et al. 1999, 2000). In ods). addition, the rate of DNA evolution in both plastid and nu- A second possible adaptive explanation would be if fast clear genes has been shown to vary widely among lineages rates of sequence change lead to increased rates of functional (Clegg et al. 1994, 1997), although the causes are still not phenotypic change within lineages. In this case, phenotypic fully understood. Previously, using an early phylogeny by change could be an intermediate factor linking rates of mo- 677 q 2001 The Society for the Study of Evolution. All rights reserved. 678 T. G. BARRACLOUGH AND V. SAVOLAINEN lecular evolution to the adaptive radiation of families. To where N1 and N2 are the numbers of species in sister families investigate this possibility, we consider rates of morpholog- 1 and 2. We use (X1)±(X2) as the branch length contrast, where ical change among angiosperm families. Morphological rates X1 and X2 are the branch lengths. Signi®cance of the asso- have been found to correlate with rates of DNA change in ciation was assessed by least squares regression forced some other groups (Omland 1997), but the correlation with through the origin (Harvey and Pagel 1991). species numbers has not been tested in previous studies. In We perform additional analyses to investigate the effects addition, morphology is one aspect of the phenotype that has of several possible sampling artifacts on our results. First, been scored for all plant families. Therefore, we use mor- our analyses rely on accurate knowledge of sister family re- phological characters from an online database of angiosperm lationships. The three-gene phylogeny does not include all families (Watson and Dallwitz 1991, 1999) to test whether angiosperm families, and therefore missing families would morphological rates of change are correlated with species be likely to disrupt some of our sister pairs if they were richness and/or molecular rates within angiosperms. included in the phylogeny. We assessed the likely impact of incomplete sampling by repeating our analyses including METHODS only those sister family pairs also found in analyses with complete taxonomic representation, but based mostly on few- We use the recently published phylogenetic analysis by er genes (Qiu et al. 1999; Chase et al. 2000; Savolainen et Soltis et al. (1999) and Soltis et al. (2000), based on three al. 2000). Second, the three-gene phylogeny may contain genes from both plastid and nuclear genomes (rbcL, atpB, errors in topology simply due to weak signal or sampling 18S rDNA, total 4811 base pairs). This new phylogeny of error in the data. We assessed the possible impact of these angiosperms is the most encyclopedic so far: the tree is well errors by repeating our analysis including only those sister resolved, well supported, and includes 75% coverage of plant families judged to be strongly supported by Soltis and co- families as de®ned by the new classi®cation by the Angiosperm workers (jackknife support . 85%; Soltis et al. 1999). Fi- Phylogeny Group (APG 1998). The matrix is available from nally, one possible bias affecting our analyses is that max- the website http://www.wsu.edu:8080/;soltilab/threepgenep imum parsimony (and possibly maximum likelihood) may trees/nature.html. We identi®ed all sister family pairs from reconstruct longer branch length in families in which more the phylogeny (listed in the Appendix), where sister families terminals were sampled, the so-called node density effect are de®ned as two families derived directly from a single (Sanderson 1990). If the angiosperm phylogeny tended to common ancestor. We compare sister families so as to max- include more terminals for families with more species, this imize the number of comparisons in our analyses, without could lead to a bias. We controlled for this effect by repeating including nested comparisons. Our analysis only relies on the the analysis including only those comparisons with equal families being sister clades and does not depend upon their numbers of terminal taxa sampled in each sister family. status as families.