Specialization has often been claimed to be an evolutionary dead end, with specialist lineages having a reduced capacity to persist or diversify. In a phylogenetic comparative framework, an evolutionary dead end may be detectable from the phylogenetic distribution of specialists, if specialists rarely give rise to large, diverse clades. Previous phylogenetic studies of the influence of specialization on macroevolutionary processes have demonstrated a range of patterns, including examples where specialists have both higher and lower diversification rates than generalists, as well as examples where the rates of evolutionary transitions from generalists to specialists are higher, lower or equal to transitions from specialists to generalists. Here, we wish to ask whether these varied answers are due to the differences in macroevolutionary processes in different clades, or partly due to differences in methodology. We analysed ten phylogenies containing multiple independent origins of specialization and quantified the phylogenetic distribution of specialists by applying a common set of metrics to all datasets. We compared the tip branch lengths of specialists to generalists, the size of specialist clades arising from each evolutionary origin of a specialized trait and whether specialists tend to be clustered or scattered on phylogenies. For each of these measures, we compared the observed values to expectations under null models of trait evolution and expected outcomes under alternative macroevolutionary scenarios. We found that specialization is sometimes an evolutionary dead end: in two of the ten case studies (pollinator-specific plants and host-specific flies), specialization is associated with a reduced rate of diversification or trait persistence. However, in the majority of studies, we could not distinguish the observed phylogenetic distribution of specialists from null models in which specialization has no effect on diversification or trait persistence.
Estimated tree using data from Fernandez et al., (2013)
Our estimated tree generated using data from Fernandez et al., (2013). The tree is based on the ITS DNA region (accession numbers from the supplementary material of Fernández-Mazuecos et al., (2013)) for 29 species of herbaceous plants in the genus Linaria and one outgroup, Antirrhinum graniticum. The tree was estimated using BEAST v2.1.3 (Bouckaert et al., 2014) on the constrained topology from Fernández-Mazuecos et al., (2013). We used a GTR + Γ model of nucleotide evolution, separate partitions for each gene, a relaxed log-normal clock, and birth-death priors. Further information and full references can be found in the supplementary material of our paper.
Fernandez_et_al_2013.tree
Estimated tree using data from Janz et al., (2001)
Our estimated tree generated using data from Janz et al., (2001). The tree is based on the mitochondrial gene (nd1) and nuclear gene (wingless) for 25 species in the butterfly tribe Nymphalini and one outgroup, Argynnis paphia. The accession numbers were taken from Nylin et al., (2001). The tree was estimated using BEAST v2.1.3 (Bouckaert et al., 2014) on the constrained topology from Janz et al., (2001). We used a HKY + Γ model of nucleotide evolution with separate partitions for each gene, a relaxed log-normal clock, and a coalescent prior. Further information and full references can be found in the supplementary material of our paper.
Janz_et_al_2001.tree
Estimated tree using data from Johnson et al., (2009)
Our estimated tree generated using data from Johnson et al., (2009). The tree is based on the nuclear gene (EF-1α) and mitochondrial genes (12S and COI) for 38 species in the genus Columbicola and one outgroup, Oxyliperurus chiniri. The accession numbers were taken from Johnson et al., (2007). The tree was estimated using BEAST v2.1.3 (Bouckaert et al., 2014), allowing the topology to vary. We used a GTR + Γ model of nucleotide evolution with separate partitions for each gene and codon position, a relaxed log-normal clock, and a birth-death prior. Further information and full references can be found in the supplementary material of our paper.
Johnson_et_al_2009.tree
Estimated tree using data from Larkin et al., (2008)
Our estimated tree generated using data from Larkin et al., (2008). The tree is based on the mitochondrial genes (COI and COII including the intervening tRNA-leucine) and the nuclear gene (EF-1α) for 35 species of bees in the genus Andrena and one outgroup, Ancylandrena larreae. The accession numbers were taken from Larkin et al., (2006). The tree was estimated using BEAST v2.1.3 (Bouckaert et al., 2014) on the constrained topology from Larkin et al., (2008). We used a HKY + Γ model of nucleotide evolution, separate partitions for each gene, a relaxed log-normal clock, and a yule prior. Further information and full references can be found in the supplementary material of our paper.
Larkin_et_al_2008.tree
Estimated tree using data from Mendlová and Šimková (2014)
Our estimated tree generated using data from Mendlová and Šimková (2014). The tree is based on two partial ribosomal RNA genes (SSU and LSU) and the entire ITS1 nuclear gene region for 25 species in the genus Cichlidogyrus and 3 species in the genus Scutogyrus. The accession numbers were taken from Mendlová et al. (2012) . The tree was estimated using BEAST v2.1.3 (Bouckaert et al., 2014) on the constrained topology from Mendlová and Šimková (2014) . We used a GTR + Γ model of nucleotide evolution, separate partitions for each gene, a relaxed log-normal clock, and birth-death priors. Further information and full references can be found in the supplementary material of our paper.
Mendlova_&_Simkova_2014.tree
Estimated tree using data from Schweizer et al., (2014)
Our estimated tree generated using data from Schweizer et al., (2014). The tree is based on three partial nuclear genes (c-mos, RAG-1 and Zenk) and one mitochondrial gene (ND2) for 78 parrot species within the order Psittaciformes and one outgroup, Falco tinnunculus. The accession numbers were taken from the supplementary material of Schweizer et al., (2014). The tree was estimated using BEAST v2.1.3 (Bouckaert et al., 2014) on the constrained topology from Schweizer et al., (2014). We used a HKY + Γ model of nucleotide evolution with separate partitions for each gene, a relaxed log-normal clock, and a coalescent prior. Further information and full references can be found in the supplementary material of our paper.
Schweizer_et_al_2014.tree
Estimated tree using data from Šimková et al., (2006)
Our estimated tree generated using data from Šimková et al., (2006). The tree is based on partial 18S rDNA and the entire nuclear ITS1 gene region for 51 species in the genus Dactylogyrus and two outgroups, Thaparocleidus vistulensis and Clediodiscus pricei. The accession numbers were taken from the Šimková et al., (2004). The tree was estimated using BEAST v2.1.3 (Bouckaert et al., 2014) on the constrained topology from Šimková et al., (2006). We used a GTR + Γ model of nucleotide evolution with separate partitions for each gene, a relaxed log-normal clock, and a birth-death prior. Further information and full references can be found in the supplementary material of our paper.
Simkova_et_al_2006.tree
Estimated tree using data from Tripp & Manos (2008)
Our estimated tree generated using data from Tripp & Manos (2008). The tree is based on the nuclear ribosomal ITS + 5.8S region and the chloroplast trnG-trnR region for 115 species in the genus Ruellia and one outgroup, Sanchezia speciosa. We obtained the alignment of these genes from TreeBASE under study ID S1995. The tree was estimated using BEAST v2.1.3 (Bouckaert et al., 2014) on the constrained topology from Tripp & Manos (2008). We used a GTR + Γ model of nucleotide evolution with separate partitions for each gene and codon position, a relaxed log-normal clock, and a birth-death prior. Further information and full references can be found in the supplementary material of our paper.
Tripp_&_Manos_2008.tree
Estimated tree using data from Stireman (2005)
Our estimated tree generated using data from Stireman (2005). The tree is based on EF-1α and 28S rRNA genes for for 49 species of Tachinid flies in the subfamily Exoristinae and one outgroup, Metoposarcophaga sp. The accession numbers were taken from the Stireman (2002). The tree was estimated using BEAST v2.1.3 (Bouckaert et al., 2014) on the constrained topology from Stireman (2005). We used a GTR + Γ model of nucleotide evolution with separate partitions for each gene and codon position, a relaxed log-normal clock, and a birth-death prior. Further information and full references can be found in the supplementary material of our paper.
Stireman_2005.tree
