American Journal of Botany 101(2): 287–299. 2014. P HYLOGENY OF LAMIIDAE 1 N ANCY F . R EFULIO-RODRIGUEZ 2 AND R ICHARD G. OLMSTEAD 2,3 2 Department of Biology, Box 355325, University of Washington, Seattle, Washington 98195 USA • Premise of the study: The Lamiidae, a clade composed of approximately 15% of all fl owering plants, consists of fi ve orders: Boraginales, Gentianales, Garryales, Lamiales, and Solanales; and four families unplaced in an order: Icacinaceae, Metteniusi- aceae, Oncothecaceae, and Vahliaceae. Our understanding of the phylogenetic relationships of Lamiidae has improved signifi - cantly in recent years, however, relationships among the orders and unplaced families of the clade remain partly unresolved. Here, we present a phylogenetic analysis of the Lamiidae based on an expanded sampling, including all families together, for the fi rst time, in a single phylogenetic analyses. • Methods: Phylogenetic analyses were conducted using maximum parsimony, maximum likelihood, and Bayesian approaches. Analyses included nine plastid regions ( atpB , matK , ndhF , psbBTNH , rbcL , rps4 , rps16 , trnL - F , and trnV - atpE ) and the mitochondrial rps3 region, and 129 samples representing all orders and unplaced families of Lamiidae. • Key results: Maximum Likelihood (ML) and Bayesian trees provide good support for Boraginales sister to Lamiales, with successive outgroups (Solanales + Vahlia) and Gentianales, together comprising the core Lamiidae. Early branching patterns are less well supported, with Garryales only poorly supported as sister to the above ‘core’ and a weakly supported clade composed of Icacinaceae, Metteniusaceae, and Oncothecaceae sister to all other Lamiidae. • Conclusions: Our phylogeny of Lamiidae reveals increased resolution and support for internal relationships that have remained elusive. Within Lamiales, greater resolution also is obtained, but some family interrelationships remain a challenge. Key words: Boraginales; chloroplast; Garryales; Gentianales; Lamiales; Lamiidae; mitochondrial; Solanales. Current understanding of Lamiidae phylogeny comes from been unable to establish precise sister group relationships two decades of molecular phylogenetic research aimed at un- among major clades within Lamiidae, beyond fi nding that Gar- derstanding the phylogeny of Asteridae and angiosperms in ryales, Icacinaceae, Oncothecaceae, and Metteniusiaceae repre- general ( Olmstead et al., 1992 , 1993 , 2000 ; Chase et al., 1993 ; sent basal lineages, and that Lamiales, Gentianales, Solanales, Soltis et al., 2000 , 2011 ; Albach et al., 2001 ; Kårehed, 2001 ; Boraginales, and Vahliaceae form a well-supported clade. Bremer et al., 2002 ; González et al., 2007 ). Through these stud- Lamiidae are one of two major clades, along with Campanu- ies, the monophyly of Lamiidae is well established, and in- lidae ( Cantino et al., 2007 ; Tank and Donoghue, 2010 ), com- cludes four large clades: (1) Boraginales (forget-me-nots, prising the core asterids (sensu the Angiosperm Phylogeny heliotropes); (2) Gentianales (coffee, gentians, milkweeds); (3) Group (APG), 1998 ; APG II, 2003 ; APG III, 2009 ; Soltis et al., Lamiales (African violets, foxglove, mint, olives, sesame, trum- 2011 ). Lamiidae (sensu Olmstead et al., 1992 ) are roughly pet creeper, verbena); and (4) Solanales (morning glory, petu- equivalent to the subclass Lamiidae of Takhtajan (1987, but not nia, potato, tomato), as well as a series of smaller clades of less 1997). In several early studies Lamiidae were referred to as as- well-known plants that mostly form a basal-grade relative to the terids I ( Chase et al., 1993 ), euasterids I ( APG, 1998 ; APG III, large clades (Garryales, Icacinaceae, Metteniusiaceae, Oncoth- 2009 ), or lamiids ( Bremer et al., 2002 ; Judd and Olmstead, ecaceae, and Vahliaceae). Despite a progressive increment in 2004 ). Adding to this confusion, Cantino et al. (2007) used the number of genes and/or taxa sampled, these studies have Lamiidae for the smaller, less-inclusive clade (Lamiales, Genti- anales, Solanales, and Boraginales.), coining a new name, Gar- ryidae, for the clade recognized here as Lamiidae. This stemmed 1 Manuscript received 6 November 2013; revision accepted 7 January from the exclusion of Garryales from Lamiidae because of poor 2014. resolution in an early paper based on rbcL ( Olmstead et al., The authors thank Doug and Pam Soltis for coordinating the Angiosperm Tree of Life Project; Mike Schoenborn for his help with the sequencing of the 1993 ). Priority, and common, informal recognition under the rps3 mitochondrial region; Bruce Godfrey (Director of the Comparative variety of names listed above, suggests that Lamiidae is best Genomics Center-University of Washington) for assistance in the sequencing used to refer to the more inclusive clade. facility; and Yin-Long Qiu and D. Tank for sharing primers information. We Within Lamiidae, the largest and most intractable clade is La- are grateful to the following colleagues: Mark Chase, Laszlo Csiba, Claude miales (sensu Olmstead et al., 1993 , 2000 , 2001 ; Soltis et al., dePamphillis, Barry Hammel, Maribeth Latvis, Jeffrey Paul Benca, Jim Smith, 2000 , 2011 ; Bremer et al., 2002 ; Schäferhoff et al., 2010 ). Sasa Stefanović, Lena Struwe, Eva Wallander, Alexandra Wortley, Gregory As currently delimited, Lamiales is composed of 25 families Young, Andrea D. Wolfe, and Dave Tank; as well as the Missouri Botanical (Acanthaceae, Bignoniaceae, Byblidaceae, Calceolariaceae, Garden, Rancho Santa Ana Botanic Garden, Royal Botanic GardensUK), and Carlemanniaceae, Gesneriaceae, Lamiaceae, Lentibulari- the Smithsonian Institution (US) for providing DNA or leaf tissue for DNA aceae, Linderniaceae, Martyniaceae, Mazaceae, Oleaceae, extraction. This study was supported by a National Science Foundation (NSF) Assembling the Tree of Life grant EF-0431184 to RGO, and NSF Systematics Orobanchaceae, Paulowniaceae, Pedaliaceae, Phrymaceae, Plan- grants DEB-0542493 and DEB 1020369 to RGO. taginaceae sensu APG, Plocospermataceae, Rehmanniaceae, 3 Author for correspondence (e-mail: [email protected]) Schlegeliaceae, Scrophulariaceae, Stilbaceae, Tetrachondraceae, Thomandersiaceae, and Verbenaceae), and over 20 000 species doi:10.3732/ajb.1300394 ( Olmstead et al., 2000 ; Olmstead, 2012, http://depts.washington. American Journal of Botany 101(2): 287–299, 2014 ; http://www.amjbot.org/ © 2014 Botanical Society of America 287 288 AMERICAN JOURNAL OF BOTANY [Vol. 101 edu/phylo/Classifi cation.pdf). Lamiales phylogeny has proven similarity criterion ( Simmons et al., 2004 ) using the program Se-Al (vers. diffi cult to resolve ( Wortley et al., 2005 ). Even though it is 2.0a11; Rambaut, 1996–2002). Gaps were treated as missing data. Ambigu- strongly supported as a monophyletic group, analyses of up to 17 ously aligned regions ( matK ~450 nt; psbBTNH ~175 nt; trnV ~525 nt; trnL-F ~430 nt; rps16 ~1100 nt; rps3 ~300 nt; ndhF ~30 nt) were excluded from genes ( Soltis et al., 2011 ) have failed to provide a fully resolved analyses. phylogeny. The recent study of Schäferhoff et al. (2010) , with three chloroplast gene regions, included the broadest taxon sam- Maximum parsimony— Maximum parsimony (MP) analyses were per- pling to date, but also left many relationships among core fami- formed for each marker separately (results not shown) and combining all 10 lies poorly resolved. genes using PAUP* (version 4.0b10; Swofford, 2002 ). Heuristic searches were Here, we provide a Lamiidae phylogeny with the broadest performed with 100 random stepwise-addition replicates, using tree bisection- taxon sampling to date. Ten gene regions (~17 214 bp) were reconnection branch swapping (TBR), and with MULTREES on and 10 trees sequenced, including both chloroplast and mitochondrial ge- held at each step. This was followed by a second set of searches using the con- sensus of the trees obtained as a reverse constraint and 1000 replicates to search nomes. Within Lamiidae, we specifi cally focused on Lamiales, for other islands of equal or shorter trees ( Catalan et al., 1997 ). Bootstrap values because of the lack of resolution in previous studies. (BP) for clades were obtained using similar settings described above, but sav- ing no more than fi ve trees and with MULTREES off ( DeBry and Olmstead, 2000 ). Gaps were treated as missing data. MATERIALS AND METHODS Model selection— Nucleotide substitution models for each gene were deter- Taxon sampling — Taxon sampling consisted of 129 samples representing mined by the program jModelTest ( Posada, 2008 ) using the Akaike information all four orders and the fi ve additional families unplaced to order in Lamiidae. criterion (AIC). The nucleotide substitution models obtained for each gene are Sampling was concentrated on resolving family relationships within Lamiales. as follows: atpB (TVM + I + G), matK (GTR + G), ndhF (GTR + I + G), Six taxa were chosen as the outgroups based on previous studies (e.g., psbBTNH (TVM + I + G), rbcL (TVM + I + G), rps3 (TVM + G), rps4 (TVM + G), Olmstead et al., 2000 ; Bremer et al., 2002 ; Soltis et al., 2011 ). Because it has rps16 (GTR + G), trnL - F (TVM + G), and trnV - atpE (TVM + G). been shown that Campanulidae are sister to Lamiidae (together forming the core Asteridae, Judd and Olmstead, 2004 ), four taxa were selected representing Bayesian analysis— Bayesian Markov Chain Monte Carlo (MCMC, Yang the main clades of campanulids, Helwingia (Aquifoliales), Aralia (Apiales), and Rannala, 1997 ) posterior probabilities (PP) were calculated