Cladistics

Cladistics 27 (2011) 559–580 10.1111/j.1096-0031.2011.00352.x

A phylogenetic analysis of morphological and molecular characters of Lithospermum L. (Boraginaceae) and related taxa: evolutionary relationships and character evolution

James I. Cohen*

Department of Biology and Chemistry, Texas A&M International University, LBVSC 379E, 5201 University Blvd, Laredo, TX 78041, USA Accepted 12 January 2011

Abstract

Lithospermum (Boraginaceae) includes ca. 60 species and exhibits a wide range of floral, palynological, and vegetative diversity. Phylogenetic analyses based on 10 chloroplast DNA regions and 22 morphological characters were conducted in order to (i) examine evolutionary relationships within Lithospermum and among related genera of Boraginaceae, and (ii) investigate patterns of morphological evolution. Several morphological features, such as long-funnelform corollas, faucal appendages, reciprocal herkogamy, and evident secondary leaf venation, have evolved multiple times within the genus. In contrast, other morphological features, including the presence of glands and the position and number of pollen pores, are less plastic and tend to characterize larger clades. Some features, including the presence of glands, are interpreted as symplesiomorphic for Lithospermum, while others, such as evident secondary leaf venation, appear to have originated repeatedly. The range of structural diversity that occurs among the species of Lithospermum suggests the potential utility of this genus as a model for integrative studies of evolution, development, and molecular biology. The Willi Hennig Society 2011.

Lithospermum L., a genus in the family Boraginaceae, the other New World members of Lithospermeae, previ- comprises ca. 60 species, with a centre of diversity in ously placed in Lasiarrhenum I.M. Johnst., Macromeria Mexico and the south-western United States. Recently, D. Don, Nomosa I.M. Johnst., Onosmodium Michx., Cohen and Davis (2009) and Weigend et al. (2009) Perittostema I.M. Johnst., and Psilolaemus I.M. Johnst. reconstructed phylogenies of Lithospermum and related (Cohen and Davis, 2009). With this broader generic genera of Lithospermeae, the tribe to which Lithosper- circumscription, Lithospermum exhibits a wide range of mum belongs. Cohen and Davis (2009) utilized 10 floral, palynological, and vegetative diversity. chloroplast DNA (cpDNA) regions, and Weigend et al. In light of the broad range of structural diversity (2009) concatenated two cpDNA regions along with the among its species, Lithospermum may be a useful taxon nuclear ribosomal internal transcribed spacer (ITS). Both in which to investigate character evolution. Some of these analyses resolved Lithospermum as non-mono- features, such as flowers with long-funnelform corollas phyletic, with members of related genera nested among its (> 4 cm in length) and exserted anthers and stigmas species. In light of these relationships, Cohen and Davis (termed Macromeria-type flowers), appear to have (2009) expanded the circumscription of Lithospermum, originated multiple times (Cohen and Davis, 2009). and under this broader interpretation the genus appears However, the phylogenetic distributions of other char- to be monophyletic. Lithospermum currently includes the acters, including type of herkogamy, pollen shape, and species that traditionally have been recognized as mem- pattern of leaf venation, remain unexamined. bers of the genus (Johnston, 1952, 1954a,b) as well as all The goals of the present study are twofold: (i) to reconstruct a phylogeny of Lithospermum (herein *Corresponding author: the ingroup) and other members of Lithospermeae E-mail address: [email protected] and Boraginaceae (herein the outgroup) through a

The Willi Hennig Society 2011 560 J.I. Cohen / Cladistics 27 (2011) 559–580 combination of cpDNA sequence data and morpholog- www.ebi.ac.uk/Tools/muscle/index.html) using the de- ical character data, and (ii) to investigate patterns of fault settings, with subsequent adjustments made by eye. morphological character evolution. The results of the The matrices that include the 10 cpDNA regions are present study will allow for the identification of char- missing ca. 20% of the sequence data. This number is due, acters that diagnose larger clades and that may prove in part, to failed amplifications for some cpDNA regions predictive for future taxonomic classifications. for a limited number of species, but the majority of missing A great deal of discussion has taken place during the data are confined to a set of 10 species for which data were past 20 years regarding the inclusion of morphological obtained from GenBank. With the exclusion of these characters in phylogenetic analyses (e.g. Eernisse and species, only 5.7% of the data are missing. The topology of Kluge, 1993; Luckow and Brunneau, 1997; Scotland the consensus tree, with respect to the remaining species, et al., 2003; Wiens, 2004). I believe that the most does not differ when these 10 species are excluded. comprehensive tests of phylogenetic relationships and Sequences generated for the present study are deposited patterns of character evolution result from the integra- in GenBank (GenBank numbers in Table S1), and the tion of multiple types of critically examined data (e.g. matrix is available at TreeBASE (http://purl.org/phylo/ Eernisse and Kluge, 1993; Luckow and Brunneau, treebase/phylows/study/TB2:S1130). 1997), and the present study employs this holistic approach to phylogenetic analysis and the investigation Morphological coding of character evolution. The morphological matrix includes 22 characters (Table 1). Half of the characters are binary, while the Materials and methods other half are multistate. Morphological character data were gathered from Taxon sampling living plant material, herbarium specimens from BH, CAS, F, GH, MEXU, MICH, NY, TEX ⁄LL, US, and The taxon sampling employed in the present study WISC, and digital images of species. Published descrip- includes 67 species (Table S1). Thirty-seven belong to the tions also were consulted (Johnston, 1952, 1953a,b, ingroup, and this sampling represents both the morpho- 1954a,b; Valentine and Chater, 1972; Al-Shehbaz, 1991; logical and geographical range of variation within Litho- Zhu et al., 1995), as were other peer-reviewed publica- spermum sensu Cohen and Davis (2009). The outgroup tions (Dı´ez et al., 1986; Jian-Chang et al., 1995; Boyd, comprises 30 species from related genera of Boragina- 2003; Selvi and Bigazzi, 2003; Aytas Akc¸ in and Ulu, 2007; ceae: three from Boragineae, seven from Cynoglosseae, Thomas et al., 2008; Ferrero et al., 2009). I attempted to one from Echiochileae, and 19 from Lithospermeae. observe at least 20 specimens for each species; however, given the limited quantity of representative material for DNA sequence data some taxa, fewer specimens were sometimes examined. This occurred more frequently with outgroup than with Sequence data from one protein-encoding cpDNA ingroup species. If a species included multiple states for a region, matK; and from nine non-encoding cpDNA character, the species was scored with all applicable states regions: ndhF–rpl32, psbA–trnH, psbJ–petA, the rpl16 for that character. In a few cases where it was not possible intron, trnK–rps16, trnL–rpl32, trnQ–rps16, ycf6– to collect data (often related to pollen) for a particular psbM, and trnL–trnF, were included in the present species, information for a specific character was based on study. The majority of the species were collected from observations of congeneric species. For the morpholog- wild populations. For these taxa, herbarium specimens ical matrix, 1.5% of the cells are scored as missing. were collected and deposited at the Bailey Mortorium Herbarium (BH) at Cornell University, and leaf tissue Phylogenetic analysis was dried and preserved in silica gel for subsequent DNA extraction. Taxa not collected from natural Three matrices were constructed: the cpDNA matrix populations were obtained from gardens (Cornell Plan- comprises cpDNA sequence data plus scores for tations, Missouri Botanical Garden, and National structural features of DNA regions (gaps, inversions, Botanic Garden of Belgium) as leaf samples preserved and unusual nucleotide motifs1); the morphological in silica gel, or as DNA isolations from either the DNA bank of the Royal Botanic Gardens, Kew, UK or the 1An unusual nucleotide motif is defined as a short (between seven and South African National Biodiversity Institute (SANBI). 10 nucleotides in length) non-inverted, non-length-variable (or shorter DNA isolation and PCR and sequencing protocols are by one nucleotide) sequence of nucleotides that occurs in two or more taxa and differs substantially in composition (more than 50% of the the same as those described in Cohen and Davis (2009). aligned nucleotides) from the sequences of most other species. Each Sequence data were aligned using the MUSCLE server unusual nucleotide motif was treated as a single character, as each is at the European Bioinformatics Institute (EBI) (http:// treated provisionally as having arisen from a single event. Table 1 Morphological characters, their states, and additional information

Character Character states Comments 1 Position of leaves (0) cauline and basal (1) only cauline (2) cauline and A pseudobasal rosette is defined as a rosette that is sometimes present and pseudobasal may be ephemeral. This type of rosette includes leaves that, although they may have short internodes between them, are not necessarily from the base of the stem. 2 Pattern of leaf venation (0) only midvein evident (1) midvein and secondary Sunken venation refers to veins that are not flush with the surface of the leaf veins evident, veins sunken (2) midvein and (Fig. 4e). State 3 refers to leaves that either have trinerviate venation with secondary veins evident, veins not sunken one or two pairs of secondary veins branching from the middle of these three (3) trinerviate with 1 or 2 pairs of secondary veins veins or a venation pattern with a midvein and one or two pairs of veins or midvein and 1 or 2 pairs of secondary veins branching from the midvein. 3 Naphthoquinones (0) present (1) absent Species that sometimes produce small amounts of naphthoquinones are coded as polymorphic. 4 Abaxial trichomes on corolla (0) present (1) absent

5 Cleistogamy (0) present (1) absent 559–580 (2011) 27 Cladistics / Cohen J.I. 6 Adaxial trichomes on the corolla (0) present (1) absent Johnston (1952, 1954a) identified only a few species of Lithospermum with trichomes inside the corollas. However, some of the species that he did not mention as bearing trichomes on the inside of their corollas are polymorphic for this character, with only some individuals producing trichomes in this position. 7 Faucal appendages (0) present (1) absent 8 Glands inside corolla (0) present (1) absent 9 Corolla colour (0) cream (1) purple (2) blue (3) yellow (4) white Some species are coded as polymorphic for this character. In some species, the (5) white with a yellow centre (6) green-yellow corolla is of two different colours (two-toned). In other species, the flowers of (7) orange (8) cream-green (9) pink ⁄ red one individual bear corollas that are of one colour, but those of other individuals are of a different colour. Both these instances result in polymorphic coding. Often, if a species is polymorphic for corolla colour among individuals, it also bears two-toned corollas. 10 Corolla shape (0) salverform (1) salverform-funnelform Macromeria-type corollas include states 7 and 8. (2) funnelform (3) tubular (4) campanulate–Onosmodium-type (5) campanulate–Lasiarrhenum-type (6) urceolote (7) long-funnelform that gradually taper from the base to the apex (8) long-funnelform flaring open (9) rotate 11 Corolla lobes (0) reflexed (ca. 180) (1) flared (ca. 90) Some species have their corolla lobes oriented in different manners at different (2) erect ⁄ ascending (< 45) stages of development, or throughout their geographic range. 12 Type of herkogamy (0) reverse herkogamy (1) non-herkogamy Some species exhibit multiple types of herkogamy. (2) approach herkogamy (3) reciprocal herkogamy 13 Filament shape in cross-section (0) circular (1) elliptical (2) triangular 14 Anther exsertion ⁄ insertion (0) exserted (1) inserted 15 Trichomes on the abaxial (0) present (1) absent surface of anthers 16 Pollen shape (0) ellipsoid (1) spherical (2) cylindrical (3) prolate with a constricted equator (hourglass) (4) ovoid 561 562 J.I. Cohen / Cladistics 27 (2011) 559–580

matrix is composed of scores for 22 morphological characters (Table 1); and the combined matrix concat- enates the cpDNA matrix and the morphological matrix. All characters were treated as non-additive, and all characters were weighted equally for cladistic analysis. Maximum parsimony phylogenetic analyses were conducted with each of the three data matrices and with various permutations of the data matrix described below (with a character or group of characters removed). The following search strategy was applied in all analyses: the data were analysed using TNT (Golob- off et al., 2008), with 1000 000 trees held in memory, and 1000 independent iterations, with 20 trees held per iteration, of 1000 parsimony ratchet iterations (Nixon, 1999), with 10% probability of upweighting and 10% probability of downweighting, followed by 1000 cycles of tree drifting; afterwards, 100 rounds of tree fusion and random sectorial searches were performed (Golob- off, 1999a), followed by TBR-max, swapping among all the most parsimonious trees until completion. Clade support for the combined matrix was measured with pores tends to be fixedbear in pollen species, grains but variation withfixed tends a to within greater occur a number in certain of species(i.e. range: that pores. 12). either However, 6–8 most pores, taxa 3–5, are or more than 10 pores development; therefore these speciesare are coded polymorphic for as this such. character and producing terminal stigmas, andstigmas. other individuals developing subterminal after the fruit is dispersed (Johnston, 1953b; Thomas et al., 2008). TNT (Goloboff et al., 2008). Ten thousand jackknife These character states are defined in this manner because the presence of three Most species that are tardily exserted often also have inserted styles earlier in replicates (36% removal probability; Farris et al., 1996) were conducted. For each replicate, 10 TBR searches were conducted, with 10 trees held after each replicate, and a total of 100 000 trees held in memory for the duration of the entire jackknife resampling. The same search strategy was employed for the cpDNA matrix, but the data were analysed with NONA (Goloboff, 1999b). Consistency indices (CI) and retention indices (RI) were calculated after removal of parsimony-unin- formative characters.

Results

Sequence variation

The cpDNA and combined matrices each include a total of 10 036 aligned nucleotides from the 67 species. (3) three to five exserted (3) precociously exserted A total of 918 informative characters were obtained from the aligned sequenced data. This number includes 811 informative nucleotides and 107 informative gaps, inversions, and unusual nucleotide motifs.

Phylogenetic analysis

Eighty most parsimonious (MP) trees of 1810 steps (CI = 0.61, RI = 0.80) were discovered for the cpDNA matrix, and seven MP trees of 2054 steps (CI = 0.57, Character Character states Comments RI = 0.76) were found for the combined matrix. The

) strict consensus tree from the combined matrix is provided in Fig. 1, and that from the cpDNA matrix 1718 Pollen pore position Pollen pore number (0) equatorial (1) sub-equatorial (0) three (1) six to eight (2) more than 10 19 Style position (0) exserted at anthesis (1) inserted (2) tardily 2021 Stigma22 Nutlet attachment Nutlet bases attachedis (0) (0) basal present (1) (1) sub-medial (0) absentpresented terminal (1) subterminal in Fig. 2. Analysis Some of species are polymorphic for thisthe State character, 0 with refers some to individuals the base 22 of the nutlet remaining attached morphological to the gynobase Continued Table 1 ( characters alone resulted in 177 MP trees of 165 steps J.I. Cohen / Cladistics 27 (2011) 559–580 563

Fig. 1. Strict consensus of combined matrix (L = 2054; CI ⁄ RI = 0.57 ⁄ 0.76). Numbers above branches are jackknife values > 50%.

(CI = 0.29–0.30, RI = 0.66–0.67). The strict consensus L. and Maharanga DC. This poorly supported four- from this matrix had little resolution (Table S2). taxon clade is sister to the rest of Lithospermeae. In Outgroup relationships for the consensus trees of the contrast, analysis of the cpDNA matrix yields a cpDNA and the combined matrices are nearly identical, consensus tree in which Podonosma is sister to the rest with three exceptions. The first involves the clade that of Lithospermeae, and a clade comprising Echium, includes Mairetis I.M. Johnst., Halacsya Do¨ rfl., and five Onosma, and Maharanga is resolved as sister to the additional species. This clade is more fully resolved, remainder of the tribe. The third relationship that differs although with less support (67% jackknife support, JK) between the two analyses involves two species of in the tree from the combined matrix than in that from Glandora D.C. Thomas, Weigend, & Hilger. The con- the cpDNA matrix. The second contrasting relationship sensus of the cpDNA matrix resolves Glandora oleifolia concerns the placement of Podonosma Boiss. The (Lapeyr.) D.C. Thomas as sister to Lithospermum,and analysis of the combined matrix results in a phylogeny Glandora diffusa (Lag.) D.C. Thomas as sister to the in which Podonosma and Echium L. (61% JK) are clade composed of G. oleifolia and Lithospermum.In sisters, with this clade sister to one that includes Onosma contrast, the consensus of the combined matrix resolves 564 J.I. Cohen / Cladistics 27 (2011) 559–580

Fig. 2. Strict consensus of 10 cpDNA-region matrix (L = 1810; CI ⁄ RI = 0.61 ⁄ 0.80). Numbers above branches are jackknife values > 50%. these two species of Glandora as sisters (64% JK). Even DC., are resolved as sisters, and two Eurasian species, though the topology of the outgroup is similar between L. officinale L. and L. erythrorhizon Siebold & Zucc., the consensus trees of the two matrices, there is greater are sisters. The South African species are resolved as jackknife support for most relationships in the phylog- sister to the rest of the genus, and the Eurasian species eny of the combined matrix. are sister to the New World members of Lithospermum Both the cpDNA and combined matrices resolve (Fig. 2). In contrast, the combined matrix yields a Lithospermum sensu Cohen and Davis (2009) as mono- different set of relationships at the base of the ingroup. phyletic, but the combined matrix provides greater The clade that is sister to the rest of the genus is support for this clade (89% versus 29% JK). Apart from composed of either L. notatum (I.M. Johnst.) J.I. different jackknife support values, the consensus tree of Cohen, L. flavum Sesse´ & Moc., L. johnstonii J.I. each matrix resolves different relationships at the base of Cohen, and L. exsertum (D. Don) J.I. Cohen, or these the ingroup. In the cpDNA consensus tree, two South four species and L. obovatum J.F. Macbr., L. strictum African species, L. scabrum Thunb. and L. cinereum Lehm., and five other species (Fig. 1). J.I. Cohen / Cladistics 27 (2011) 559–580 565

In the analyses of each matrix, nine species pairs Pattern of leaf venation (L = 19, CI = 0.15). The (clades)—L. johnstonii and L. exsertum, L. scabrum and ancestral pattern of leaf venation in Boraginaceae L. cinereum, L. revolutum B. L. Rob. and L. calycosum cannot be reconstructed unambiguously (Fig. 3). (J.F. Macbr.) I.M. Johnst., L. leonotis (I.M. Johnst.) J.I. Regardless of the ancestral condition, secondary vena- Cohen and L. nelsonii Greenm., L. mirabile Small and tion, either with or without sunken veins, has evolved L. mirabile · incisum, L. molle Muhl. and L. helleri multiple times among members of the family [Figs 3 (red (Small) J.I. Cohen, L. macromeria J.I. Cohen and and blue hash marks) and 4e,f, respectively]. In the L. multiflorum Torr. ex A. Gray, L. trinervium (Lehm.) ingroup, eight species are polymorphic for one of two J.I. Cohen and L. discolor M. Martens & Galeotti, and conditions: (i) leaves with only a midvein (state 0) or L. officinale and L. erythrorhizon—were obtained, all of with a midvein and secondary venation of either type them supported by 50% JK or greater, with most (state 0 and either state 1 or 2); or (ii) leaves with either supported by > 85% JK. The inclusion of morpholog- type of secondary venation (states 1 and 2) (Table 1). ical data in the analyses results in increased jackknife Sunken secondary veins are a synapomorphy for a large support for three of these clades and decreased support clade composed of L. molle, L. tuberosum, and 13 other for one of these clades. In addition, analyses of the species. Within this clade, secondary venation either has combined matrix yield increased support for two clades: been lost four times, or has been lost three times and one composed of L. cobrense Greene and L. tubuliflo- subsequently gained once. A venation pattern consisting rum Greene (5% versus 87% JK), and another that of either a midvein and one or two pairs of secondary includes L. distichum Ortega, L. oblongifolium Greenm., veins, or three equal or subequal primary veins and one and L. gayanum I.M. Johnst (48% versus 75% JK). or two pairs of secondary veins, either originated twice Despite weak support for many clades of the ingroup, or arose once, and was subsequently lost (Fig. 3, green jackknife support tends to increase with the addition to hash marks). This type of venation characterizes both the analyses of morphological characters. L. rosei (I.M. Johnst.) J.I. Cohen and L. trinervium. Although the evolutionary relationships resolved by the two matrices are similar, only one clade composed of Naphthoquinones (L = 8, CI = 0.12). The production more than two species is resolved with the same phylo- of naphthoquinones has originated at least 11 times genetic relationships. These clade include L. viride among members of Lithospermeae, and most of these Greene, L. nelsonii, and four other species. Additionally, instances occur in Lithospermum. Five of these origins one clade occurs only in the results of the combined involve species, such as L. multiflorum, L. officinale,and matrix (Fig. 1). This clade, which is composed of L. tu- L. revolutum, that frequently produce observable quan- berosum Rugel ex. DC., L. canescens Torr., and six other tities of naphthoquinones. species, receives weak jackknife support. Most of the members of this clade are scattered throughout the Morphological characters—floral ingroup in the phylogeny of the cpDNA matrix (Fig. 2).

Morphological characters—vegetative Abaxial trichomes on the corolla (L = 7, CI = 0.14). Plants of all the species of both Cynoglosseae and Echiochileae included in the present analyses have Position of leaves (length, L = 8, CI = 0.25). All the corollas that are abaxially glabrous, but some members species of Cynoglosseae and Boragineae included in the of Boragineae, as well as some early diverging members present analysis develop a basal rosette, with the exception of Lithospermeae, have trichomes on the abaxial surface of Amsinckia tessellata A. Gray (Cynoglosseae), which of their corollas. The ancestral condition of this char- produces a pseudobasal rosette (Table 1). In contrast, the acter cannot be reconstructed unambiguously from the presence of only cauline leaves is resolved as the plesio- available results. In Lithospermeae, glabrous corollas morphic condition for Lithospermeae. Among the are a synapomorphy for a seven-species clade that members of the tribe, the presence of basal rosettes is a includes Cerinthe major L. and Lithodora zahnii (Heldr. synapomorphy for the clade composed of Halacsya ex Hala´ csy) I.M. Johnst. Presence of abaxial trichomes and Neatostema I.M. Johnst. Outside of Lithospermum, on the corollas is a synapomorphy for the clade that this is the only occurrence of this state within the tribe. includes Buglossoides Moench., Glandora, and Litho- Basal rosettes originated at least three times among spermum. species of Lithospermum (e.g. in the ancestor of L. cob- rense and L. tubuliflorum). Pseudobasal rosettes origi- Cleistogamy (L = 3, CI = 0.33). Cleistogamy arose nated at least four times, as in L. mirabile and in three times within Lithospermeae: twice in Lithosper- L. matamorense DC. Three of the four origins of a mum (once in L. calycosum and once in the ancestor of pseudobasal rosette are among species that do not L. mirabile and L. mirabile · incisum), and once in the always produce rosettes. monospecific genus Neatostema. 566 J.I. Cohen / Cladistics 27 (2011) 559–580

Fig. 3. Phylogenetic distribution of patterns of leaf venation optimized on one most parsimonious tree. Orange hash marks, acquisition of leaves in which the only evident vein is the midvein; red, origin of leaves that produce sunken secondary veins; blue, acquisition of leaves that produce secondary veins flush with the surface of the leaf; green, origin of leaves with a venation pattern consisting of either an evident midvein and one or two pairs of secondary veins, or three equal or subequal primary veins and one or two pairs of secondary veins. Hash marks in two colours represent species or clades that are polymorphic for this character.

Adaxial trichomes on the corolla (L = 6, least eight times within Boraginaceae (Fig. 5, green hash CI = 0.16). Trichomes on the adaxial surface of corol- marks). Four of these origins are within Lithospermum, las originated at least nine times within Lithospermeae. and only one species, L. californicum A. Gray, has lost This character state tends to arise in a single species, and faucal appendages. Additionally, the presence of faucal four of the origins are within species that are polymor- appendages is a synapomorphy for Boragineae. phic for this character. The presence of corollas with adaxial trichomes is a synapomorphy for the clade Glands inside corolla (L = 8, CI = 0.12). The presence composed of L. cobrense and L. tubuliflorum. of glands (Fig. 4d) arose independently at least three times among species of Boraginaceae: once in Lindelofia Faucal appendages (L = 9, CI = 0.11). The presence Lehm., once or twice among members of Boragineae, of faucal appendages (Fig. 4c) appears to have arisen at and once in the clade containing Buglossoides, Glandora, J.I. Cohen / Cladistics 27 (2011) 559–580 567

(a) (b)

(c) (d)

(e) (f)

Fig. 4. Morphological features of species of Lithospermum. (a) Subterminal stigma, with apical projection. (b) Terminal stigma. (c) Five faucal appendages located at intersection of base of corolla lobes and apex of corolla tube. (d) Three-celled glands on adaxial surface of corolla. (e,f) Leaves with evident secondary venation. Veins in L. tuberosum are sunken (e), while veins in L. exsertum are not (f). and Lithospermum (Fig. 5, orange hash marks). Most The salverform corolla shape (Fig. 6b) is the most species in this large clade have retained the presence of common type in the ingroup. This corolla shape is the glands, but this state has been lost at least four times ancestral condition for two large clades: one includes (Fig. 5, purple hash marks), with three of these losses L. scabrum, L. gayanum, and four additional species, occurring in Lithospermum. and another is composed of L. calcicola B. L. Rob., L. caroliniense MacMill., and 20 other species. Corolla color (L = 23, CI = 0.30). I recognized 10 In the present analyses, two different types of Macrom- distinct corolla colours (Fig. 6) among the species of eria-type flowers (states 7 and 8) have been recognized. Boraginaceae. Each of these colours has originated One type includes long-funnelform corollas that flare out multiple times throughout the family. Blue corollas are (state 8; Fig. 6c), and this shape originated once in the resolved as ancestral for the present phylogeny, while clade that includes L. notatum, L. flavum, L. johnstonii, yellow corollas are ancestral for Lithospermum. and L. exsertum. The other type is composed of long- funnelform corollas that gradually taper from the base to Corolla shape (L = 27, CI = 0.33). I identified 10 the apex (state 7; Fig. 6d). This corolla shape arose three different corolla shapes among the species of Boragin- times independently: in L. oblongifolium,inL. macrom- aceae. The salverform-funnelform shape (Fig. 6a) is eria, and in L. leonotis (Fig. 5, red hash marks). plesiomorphic for Lithospermeae and Lithospermum. Additionally, the present phylogeny provides evidence Corolla lobes (L = 20, CI = 0.10). The present phy- for multiple losses of this character state, particularly logeny provides evidence that flared corolla lobes are the among early diverging members of Lithospermeae, such plesiomorphic condition for the family. Reflexed corolla as Echium and Arnebia Forssk. lobes originated at least seven times, and ascending 568 J.I. Cohen / Cladistics 27 (2011) 559–580

Fig. 5. Phylogenetic distribution of various floral features optimized on one most parsimonious tree. Red hash marks, acquisition of Macromeria- type flowers; green, acquisition of faucal appendages; orange, origin of glands; purple, loss of glands.

corolla lobes arose at least ten times. Within Lithosper- Type of herkogamy (L = 16, CI = 0.18). Approach mum, reflexed corolla lobes originated multiple times, herkogamy (stigmas positioned above the anthers) is the often among individual species, such as L. californicum ancestral condition for Lithospermeae and for Litho- and L. viride. However, this type of corolla lobe spermum. This type of herkogamy has been lost multiple orientation is a synapomorphy for the clade composed times in both the ingroup and the outgroup. of L. johnstonii and L. exsertum. In contrast, ascending Although reciprocal herkogamy (Fig. 7, blue hash corolla lobes characterize only one clade that comprises marks) is present in some species of Boragineae and L. molle, L. trinervium, and five other species, although Cynoglosseae, the present study included only species of other species, such as L. revolutum and L. notatum,bear Lithospermeae that exhibit this floral feature. Within corollas with ascending lobes. In the combined matrix, Lithospermeae, reciprocal herkogamy originated at least five species are coded as polymorphic for the presence of eight times, with five of these origins occurring in both flared and ascending corolla lobes. Lithospermum. The structure of the phylogeny of the J.I. Cohen / Cladistics 27 (2011) 559–580 569

(a) (b)

(c)

(d)

Fig. 6. Corolla colour and shape of species of Lithospermum (a) Yellow, salverform-funnelform corolla of L. cobrense. (b) White with a yellow centre, salverform corolla of L. nelsonii. (c) L. flavum bears orange, long-funnelform corollas that flare out. (d) L. macromeria produces green-yellow to green-cream, long-funnelform corollas that taper from the base to the apex. combined matrix (Figs 1 and 7) does not support any elliptical in cross-section have arisen 3–5 times. In the losses of reciprocal herkogamy. outgroup, filaments with this shape in cross-section Reverse herkogamy (stigmas positioned below the evolved twice: once in Cerinthe L. and once in anthers) as a fixed state characterizes only one outgroup Podonosma. In the ingroup, this character state either species, Mairetis microsperma (Boiss.) I.M. Johnst., but evolved independently in each of three species, some species, such as Buglossoides purpureo-caerulea (L.) L. macromeria, L. rosei,andL. trinervium, or was I.M. Johnst., are polymorphic for this condition. In the acquired once at the base of the clade that includes ingroup, the presence of reverse herkogamous flowers is a these three species and subsequently lost twice. Only one synapomorphy for only one clade. This clade includes species included in this study, Trachystemon orientalis L. tuberosum, L. matamorense, and six other species. (L.) G. Don, bears filaments that are triangular in cross- Although other species of Lithospermum exhibit reverse section. herkogamy, they do not do so as a fixed state. Anther exsertion ⁄insertion (L = 9, CI = 0.11). Inser- Filament shape in cross-section (L = 6, CI = 0.33). tion of anthers in the corolla tube is the ancestral Most species of Boraginaceae produce filaments that are condition for Lithospermeae and for Lithospermum. circular in cross-section. This state is resolved as Anther exsertion originated at least eight times. Most plesiomorphic for Lithospermeae. Filaments that are of these origins are isolated instances, such as in 570 J.I. Cohen / Cladistics 27 (2011) 559–580

Fig. 7. Phylogenetic distribution of heterostyly and of stigma position optimized on one most parsimonious tree. Blue hash marks, origin of heterostyly; red, origin of flowers with stigmas exserted at anthesis from corolla tube; yellow, origin of stigmas exserted early from corolla tube; green, origin of stigmas exserted late from corolla tube; orange, origin of stigmas inserted in corolla tube. Hash marks in two colours represent species or clades that are polymorphic for this character.

L. macromeria and L. leonotis; however, anther exser- Pollen shape (L = 16–17, CI = 0.23–0.25). I recog- tion is a synapomorphy for the clade composed of nized five pollen shapes among species of Boraginaceae, L. notatum, L. flavum, L. johnstonii, and L. exsertum. and all occur in Lithospermum. The present phylogeny provides evidence that ellipsoid pollen (Figs 8 and 9, Trichomes on the abaxial surface of anthers (L = 3, orange hash marks) is the plesiomorphic state for CI = 0.33). Three species of Lithospermum—L. macro- Lithospermeae. This character state has been lost meria, L. rosei, and L. trinervium—produce trichomes multiple times, including at least seven times among on the abaxial surface of their anthers. As with the members of the outgroup. optimization of filament shape in cross-section, the Ellipsoid pollen is reconstructed as ancestral for structure of the present phylogeny does not differentiate Lithospermum, and this shape has been lost multiple unambiguously between three origins of this state and times. Prolate pollen with a constricted equator (termed one origin followed by two losses. hourglass-shaped, as the pollen closely resembles this J.I. Cohen / Cladistics 27 (2011) 559–580 571

(b) mum. This clade includes L. molle, L. canescens, and 13 other species. In this clade, a reversion to pollen with (a) equatorial pores has occurred in only one species, L. multiflorum.

Pollen pore number (L = 5, CI = 0.60). Presence of three pores originated three times, once each in Anchusa leptophylla Roem. & Schult., Halacsya, and the clade composed of Podonosma, Onosma, Echium,andMaha- ranga (d) . Most other species produce pollen with 6–8 pores, with the only exceptions being species of Arnebia and Symphytum L. Species of these two genera bear pollen with more than eight (often at least 10) pores. This derived condition evolved independently in each (c) taxon.

Style position (L = 20, CI = 0.15). Style insertion has arisen or been lost numerous times throughout the (e) evolution of the genus, tribe, and family. Multiple origins of style exsertion, whether precocious (three gains) (Fig. 7, yellow hash marks), tardy (at least two gains) (Fig. 7, green hash marks), or at anthesis (at least five gains) (Fig. 7, red hash marks), occur among members of Lithospermum.

Stigma (L = 5, CI = 0.20). Terminal stigmas (Fig. 4b) are plesiomorphic for Lithospermeae, with Fig. 8. Five different pollen shapes of species of Lithospermum. (a) the presence of subterminal stigmas (Fig. 4a) originating Prolate with a constricted equator (hourglass shape); (b) cylindrical; (c) at least nine times. Most of these origins occur within ovoid; (d) ellipsoid; (e) spherical. species that are polymorphic for this character. Only a limited number of species, including L. distichum, object) evolved independently at least five times within L. leonotis, L. nelsonii, L. obovatum, and all members of the ingroup (Fig. 9, purple hash marks). This pollen Buglossoides, exclusively develop subterminal stigmas. shape is a synapomorphy for the clade that includes L. latifolium Michx., L. canescens, and five other spe- Morphological characters—fruit cies. Apart from this clade, this type of pollen is produced only by isolated species or one morph of a heterostylous species. Ovoid pollen (Fig. 8c) originated Nutlet attachment (L = 1, CI = 1). Sub-medial nutlet once in Lithospermum, and subsequently was lost at attachment characterizes Cynoglosseae, and basal nutlet least three times (Fig. 9, red hash marks). Spherical attachment characterizes Boragineae and Lithosper- pollen (Fig. 8e) characterizes the clade that includes meae. L. viride, L. mirabile, and three other taxa. In addition, this type of pollen arose independently in both Nutlet bases attached (L = 2, CI = 0.5). Nutlet bases L. exsertum and L. strictum (Fig. 9, green hash marks). that remain attached to the gynobase evolved twice: Cylindrical pollen is a synapomorphy for a clade once in Glandora and once in Lithodora Griseb. This that includes two Mexican species, L. distichum and condition is restricted to the members of these genera. L. oblongifolium, and one South American species, L. gayanum (Fig. 9, blue hash marks). Discussion Pollen pore position (L = 7, CI = 0.14). Pollen with equatorial pores is the ancestral condition for Litho- Phylogenetic analysis spermeae (Fig. 9, aqua hash marks). Pollen with sub- equatorial pores originated at least seven times (Fig. 9, Analyses of the cpDNA and combined matrices grey hash marks). Most of these instances occur in only resolve the same outgroup relationships, with two one or two species, but sub-equatorial pollen pores are a notable exceptions—the basal relationships of Litho- synapomorphy for a large clade of species of Lithosper- spermeae and the monophyly of Glandora. The basal 572 J.I. Cohen / Cladistics 27 (2011) 559–580

Fig. 9. Phylogenetic distribution of pollen shape optimized on one most parsimonious tree. Purple hash marks, origin of prolate pollen with a constricted equator (hourglass shape); orange, origin of ellipsoid pollen; red, origin of ovoid pollen; green, origin of spherical pollen; blue, origin of cylindrical pollen; aqua, origin of pollen with equatorial pores; grey, origin of pollen with subequatorial pores. Hash marks in two colours represent species or clades that are polymorphic for this character. relationships of Lithospermeae differ between the trees The present analysis includes representatives of only of the two matrices. The cpDNA matrix reconstructs five of the nine genera of Lithospermeae that produce Podonosma as sister to the rest of Lithospermeae, a pollen with three pores (Johnston, 1952, 1953a,b, relationship also recovered by Thomas et al. (2008) and 1954a,b). The clade composed of Podonosma, Echium, Weigend et al. (2009). In contrast, the combined matrix Onosma, and Maharanga contains four of the sampled resolves as sister to the remainder of the tribe a clade taxa, while the fifth, Halacsya, is nested among members composed of Podonosma, Echium, Onosma, and Maha- of the tribe that produce pollen with 6–8 pores. The ranga. Three pollen pores is a synapomorphy for this inclusion of additional triporate representatives of the clade. With the exclusion of the character ‘‘pollen pore tribe, such as Lobostemon Lehm. and Vaupelia Brand number’’ from analyses of the combined matrix, the (Johnston, 1954b), will help to determine if pollen with species or clade that is sister to the rest of Lithospermeae different numbers of pores characterizes two separate cannot be resolved unambiguously. clades. If this is the case, then the tribe would include J.I. Cohen / Cladistics 27 (2011) 559–580 573 two clades, each diagnosed by pollen with a specific relationships also differ from other, recent phylogenies number of pores. However, if these two clades are not of Lithospermum (Cohen and Davis, 2009; Weigend resolved, then this result would provide evidence that et al., 2009). In an overall comparison with previous pollen with three pores is the ancestral state for the tribe. phylogenetic analyses of Lithsopermum, the present Another difference in the outgroup relationships phylogenies include more species of the genus and are obtained with and without the morphology included better resolved. The relationships derived from the involves the relationship between two species of Glan- cpDNA matrix are congruent with those of Cohen and dora. The combined matrix resolves the two species as Davis (2009), but differ from those of Weigend et al. sisters, a result also found by Thomas et al. (2008) and (2009). The results of the combined matrix differ Ferrero et al. (2009); however, the cpDNA sequence from those of both previous studies. Despite these data alone do not. In analyses of the combined matrix, differences, some species pairs, such as L. officinale and the inclusion of two characters—‘‘nutlet base attach- L. erythrorhizon, are recovered in all analyses. ment’’ and ‘‘type of herkogamy’’—is necessary to Cohen and Davis (2009), Weigend et al. (2009), and resolve unambiguously G. diffusa and G. oleifolia as the present analyses resolve the New World genera that sisters. If these characters are excluded from the had previously been segregated from Lithospermum, analysis, the remainder of the combined matrix cannot such as Onosmodium and Macromeria, as nested among determine whether or not the two species are sisters. species of the genus. These analyses reconstruct the Although the present analysis includes 10 cpDNA former Onosmodium, but not the former Macromeria,as regions and morphological character data, it included monophyletic. Both Cohen and Davis (2009) and only two of the six species of Glandora. Utilizing more Weigend et al. (2009) resolved the species previously character data and more complete taxon sampling, such included in Onosmodium as closely related to L. tubero- as that employed by Thomas et al. (2008) and Ferrero sum, and these species all bear leaves with sunken, et al. (2009), may help resolve the relationships among secondary veins. However, the present analyses resolve the species of this genus. different relationships. Analyses of the cpDNA matrix Although the cpDNA and the combined matrices recover L. canescens as sister to L. molle and L. helleri, resolve similar outgroup relationships, the ingroup rela- two species formerly included in Onosmodium, while tionships that they resolve differ to a greater extent. One analyses of the combined matrix result in the two latter of the most striking differences involves relationships at species in a clade with L. macromeria, L. discolor,and the base of the ingroup. Analyses of the combined matrix, three other species. Regardless of the placement of the but not those of the cpDNA matrix, yield a clade at this species previously included in Onosmodium, the group is position composed of four species with Macromeria-type resolved as monophyletic. flowers. Consequently, some character states that One notable difference between the present phyloge- are derived in analyses of the cpDNA matrix, such as nies and those of Weigend et al. (2009) involves bioge- the absence of faucal appendages, and the presence of ographic relationships. Although Weigend et al. (2009) long-funnelform corollas that flare open (state 8), are and the present analyses recover the African species and resolved as ancestral in analyses of the combined matrix. the Eurasian species as two separate monophyletic One notable difference between the trees obtained groups, the relationships among the species within each with the cpDNA and combined matrices is that, with the of these regions differ. Using a matrix composed of ITS inclusion of morphological data, jackknife support and trnL–trnF sequence data, Weigend et al. (2009) increases for the relationships of the close relatives of recovered the African and Eurasian species of Litho- Lithospermum, such as Buglossoides and Glandora.In spermum in the same clade, and this clade is nested contrast, jackknife support for larger clades of the among New World members of the genus. However, this ingroup does not tend to increase with the inclusion of clade of Old World species is not reconstructed in morphological data. This lack of increased support, analyses of their other matrices (i.e. ITS and ITS ⁄trnL– along with the structural diversity among species of trnF ⁄trnS–trnG). Similar relationships also are recov- Lithospermum, is consistent with observed patterns for ered in the present analyses. In the combined matrix, the rapid radiations, or island colonizations followed by African and Eurasian species are resolved among New subsequent diversification (Bateman, 1999). Lithosper- World species of Lithospermum (Fig. 1). However, the mum may have undergone this type of diversification cpDNA matrix reconstructs the clade of South African throughout the mountains and higher-elevation regions species as sister to the clade that includes the Eurasian of Mexico and the western United States. It is important species and the New World members of the genus, and to note that both matrices resolve many of the same the Eurasian species are sister to the New World species sister-species relationships as well as some of the same of Lithospermum (Fig. 2). The relationships of the larger clades composed of five or six species. cpDNA matrix are congruent with those of Cohen and In addition to differences between the phylogenies Davis (2009). Given the conflicting relationships recov- of the cpDNA and combined matrices, the ingroup ered, it is difficult to hypothesize whether Lithospermum 574 J.I. Cohen / Cladistics 27 (2011) 559–580 originated in the Old World and colonized the New L. officinale, and 13 other species. Seven species distrib- World once; or, after an initial colonization of the New uted among other clades bear leaves with evident World, two independent dispersal events occurred, one secondary venation, but three of these species are resulting in the Eurasian species and the other giving rise polymorphic for this character, producing some leaves to the species of southern Africa. with evident secondary venation and others in which Despite a matrix that includes 10 cpDNA regions and only the midvein is evident. Jones et al. (2009) studied 22 morphological characters, support for a many the evolution of leaf venation in South African species relationships in the present phylogenies is weak (Figs 1 of Pelargonium LÕHe´ r. ex Aiton. These authors con- and 2). Weak support can have a variety of causes, cluded that leaf venation is quite plastic in the genus. including conflict among the data, few characters This also appears to be the case in Lithospermum. supporting branches, the manner in which the data are Species that produce leaves with evident secondary generated, and the type of support value calculated venation tend to produce larger plants, and this may be (Alfaro et al., 2003; Hillis and Bull, 1993; Davis et al., influenced considerably by developmental patterns. 2004). In order to determine the cause of the weak Because the plants themselves are larger, this may support values, it is useful to examine the characters cause the leaves to become larger and develop a greater supporting the relationships. In the present case, it number of evident veins. Conversely, instead of being a appears that a lack of characters, rather than conflict, is by-product of development, a greater number of the cause of the weak support. Other studies (e.g. evident veins could allow for the growth of larger Thomas et al., 2008; Ferrero et al., 2009) that focused plants, due to an increased ability of the leaf to on the close relatives of Lithospermum and produced transport larger quantities of water, nutrients, and well resolved phylogenies of other genera of Lithosper- sugars to and from the rest of the plant (Roth- meae, recovered few dichotomous relationships, most Nebelsick et al., 2001; McKown et al., 2010). Therefore with weak support, within Lithospermum. Additionally, it is possible that evident secondary venation either Weigend et al. (2010b) reconstructed a phylogeny allows the plant to grow larger; or that larger plants that included representatives from three of the larger develop larger leaves with a greater number of evident tribes of Boraginaceae—Boragineae, Cynoglosseae and secondary veins. Lithospermeae—and of these three tribes, the shortest Some species of Lithospermum, such as L. tubuliflo- branch lengths occurred in Lithospermeae. Given the rum and L. obovatum, produce basal leaves, but not results from the above-mentioned studies, it is not cauline leaves, with evident secondary venation. The surprising that, compared with other groups in Bora- explanations presented here could also account for this ginaceae, only a small number of characters contribute pattern. The basal leaves of these species tend to grow to branch support values in the present phylogenies of much larger than the cauline leaves, therefore the basal Lithospermum (Figs 1 and 2). leaves would require an increased ability to transport water, nutrients, and sugars to and from the rest of Leaf evolution plant. The above-mentioned hypotheses concerning the I identified four different patterns of leaf venation presence of evident secondary venation do not explain among species of Boraginaceae (Table 1). Most of the why species with larger habits produce leaves in which observed variation in leaf venation occurs in the ingroup the only evident vein is the midvein, or why species with (Figs 3 and 4e,f). All outgroup members of Lithosper- smaller habits produce leaves with evident secondary meae, except Arnebia and Cerinthe, produce leaves in venation. Phylogenetic history provides one possible which the only evident vein is the midvein. This is the explanation for why species with smaller habits, such as ancestral condition for the tribe. In contrast, all species L. tuberosum, produce leaves with evident secondary of Cynoglosseae and Boragineae included in the present venation. The ancestor of this species is resolved to have study (with the exception of A. tessellata) produce leaves produced leaves with evident secondary venation. Con- with evident secondary venation. This is the plesiomor- sequently, even if a plant does not require the putative phic condition for these two tribes. This difference advantages of a greater number of evident secondary between Lithospermeae, on the one hand, and Borag- veins, the presence of this character state is retained as a ineae and Cynoglosseae, on the other, helps to distin- symplesiomorphy. guish between these groups. In Lithospermum, the ancestral state is one in which Floral evolution only the midvein is evident. Leaves with evident secondary venation have arisen multiple times. Evident Corolla shape. I recognized nine different corolla secondary venation characterizes two larger clades: shapes within Lithospermum, with all but one of these one includes L. notatum, L. exsertum, L. flavum states—the salverform corolla shape (Fig. 6b)—char- and L. johnstonii, and the other comprises L. molle, acteristic of fewer than four species. Salverform J.I. Cohen / Cladistics 27 (2011) 559–580 575 corollas have originated at least twice within the bear this type of flower are sister to species that produce genus, and this corolla shape is a synapomorphy for short corollas (1–1.5 cm in length) with inserted anthers two clades, one including L. scabrum, L. gayanum, and and stigmas. Flowers of a form that is intermediate four other species; the other composed of L. calcicola, between these two types have not been observed. It is L. caroliniense, and 20 additional species. Most of the possible that species of intermediate form once existed, eight other corolla shapes, such as tubular or funnel- and are now extinct. form, also have evolved multiple times. These shapes The transition from smaller to larger flowers has been do not characterize any clade, with a few exceptions. hypothesized to occur less frequently than the reverse For example, the presence of a type of campanulate trend (Stebbins, 1957; Johnston and Schoen, 1996; corolla is a synapomorphy for L. molle and L. helleri, Takebayashi and Morrell, 2001; Armbruster et al., two species previously assigned to Onosmodium. Long- 2002; Herlihy and Eckert, 2005). Therefore the pattern funnelform corollas that flare out (Fig. 6c) character- observed in Lithospermum and other taxa, such as ize a clade composed of four species formerly included Lycium L. (Solanaceae) (Miller and Venable, 2003), in Macromeria. However, the present study sampled Penstemon Schmidel (Plantaginaceae) (Wolfe et al., only two of the eight species that were part of 2006), and Ruellia L., along with other species of Onosmodium, and only six of the 11 species included Acanthaceae (Daniel et al., 2008; Tripp and Manos, in Macromeria. 2008), is contrary to expectation. The diversity in corolla shapes among species of Species with Macromeria-type flowers are unusual for Lithospermum is equal to or greater than that of many their putative evolutionary transitions as well as for larger genera. For example, Harrison et al. (1999) their pollination syndrome. Grant and Grant (1970) identified six different corolla types in their study of and Boyd (2002, 2004) have observed hummingbirds Streptocarpus Lindl. and Saintpaulia H. Wendl. (Ges- pollinating the flowers of L. macromeria. Boyd (2002, neriaceae), genera with 130 and 20 species, respectively. 2004) also noted hawkmoths pollinating the flowers of In their analyses of 36 species from these two genera, this plant, but the hawkmoths appear to be less Harrison et al. (1999) found 10 origins of the six corolla important pollinators than the hummingbirds. Addi- shapes. In contrast, the present study resolved 21 tional evidence suggests that some other species with transitions of corolla shape among 37 species of Macromeria-type flowers also are pollinated by hum- Lithospermum. In a study of Campanula L. (Campanul- mingbirds (Fægri and van der Pijl, 1979; Richards, aceae), a genus composed of 350–500 species (6–8 times 1997; Aizen, 2003). The repeated evolution of hum- larger than Lithospermum), Roquet et al. (2008) recog- mingbird pollination is not uncommon (Wolfe et al., nized six different corolla shapes within the genus. In 2006; Tripp, 2007; Daniel et al., 2008; Tripp and Campanula, most species tend to produce tubular- Manos, 2008), but species of Lithospermum do not fit campanulate corollas, but many species are character- the typical hummingbird pollination syndrome, which ized by one of the other five types, each of which has includes long, tubular, red corollas (e.g. Richards, 1997; originated multiple times. Roquet et al. (2008) postulate Tripp, 2007). Even though the corollas of Macromeria- that corolla shape is quite plastic in Campanula, and I type flowers satisfy most of the requirements for have reached a similar conclusion in Lithospermum. stereotypical hummingbird pollination—the corollas However, the evolution of corolla form in Lithospermum are long and tubular—none is red. However, hum- follows different patterns from that in Campanula.In mingbirds also are known to visit flowers with green or Lithospermum, most of the modifications of corolla orange tubular corollas (Richards, 1997; Boyd, 2002, shape involve changes in the length of the corolla tube, 2004), and many species of Lithospermum with the circumference of the apex of the corolla tube, and Macromeria-type flowers produce corollas with one or the orientation of the corolla lobes. However, in both of these colours. Additionally, although L. ma- Campanula, substantial changes occur in the extent of cromeria is pollinated by hawkmoths, it seems unlikely petal fusion and width of the corolla tube (if a tube that most other species with Macromeria-type flowers exists at all). In light of these differences, it appears that are hawkmoth pollinated. This is because the corollas changes in corolla form in Lithospermum tend to be of these species usually are not light-coloured, although quantitative, rather than qualitative. Some changes L. album (G.L. Nesom) J.I. Cohen (a species not involve the overall structure of the corolla, but this is included in this study) and L. johnstonii bear light- infrequent. For example, in L. rosei and L. trinervium, coloured corollas. urceolate and campanulate corolla shapes occur. These shapes are quite rare in Lithospermum, but are common Herkogamy. In Lithospermum, herkogamy can vary in some Old World genera of Lithospermeae, such as both inter- and intraspecifically. Most species, at least Onosma. over parts of their geographic range, or during different Macromeria-type flowers arose four times (Fig. 5, red stages of development, exhibit approach herkogamy, the hash marks). In three of these instances, the species that most common type of herkogamy in angiosperms 576 J.I. Cohen / Cladistics 27 (2011) 559–580

(Webb and Lloyd, 1986). Reverse herkogamy commonly 1954a,b). I divided the character ‘‘stigma position’’ into occurs among the close relatives of Lithospermum, but four states, and three of these states represent various this type of herkogamy characterizes only one clade in types of stigma exsertion (Table 1; Fig. 7). Most species the ingroup. of Lithospermum that develop exserted stigmas do so at The phylogenetic results indicate that reciprocal anthesis (J.I.C., personal observation); however, preco- herkogamy, a component of heterostyly, originated at cious stigma exsertion arose as a fixed state three times least five times within Lithospermum and at least three in the genus: once in L. revolutum, once in L. rosei,and additional times elsewhere in Lithospermeae, once each once in the ancestor of L. molle and L. helleri (Fig. 7, in Arnebia, Glandora,andLithodora (Fig. 7, blue hash yellow hash marks). Additionally, some individuals of marks). Phylogenetic analyses of other taxa, such as L. trinervium exhibit precocious stigma exsertion. It is Linum L. (Linaceae) (McDill et al., 2009), Lythrum L. not known if the stigma is receptive at this early stage of (Lythraceae) (Morris, 2007), Narcissus L. (Amaryllida- development, but should it be, precocious stigma ceae) (Graham and Barrett, 2004; Pe´ rez-Barrales et al., exsertion would combine herkogamy and dichogamy 2006), and Lithodora and Glandora (Thomas et al., (the temporal separation of pollen dispersal and stig- 2008; Ferrero et al., 2009), have also recovered multiple matic receptivity). This combination of herkogamy and origins of reciprocal herkogamy. Despite evidence from dichogamy also may occur in species that exhibit late these and other phylogenetic analyses, Kohn et al. stigma exsertion (Fig. 7, green hash marks). Future (1996), McDill et al. (2009), Morris (2007), and others crossing studies on species of Lithospermum may pro- have hypothesized that multiple gains of heterostyly vide evidence concerning the timing of stigma receptiv- among close relatives are unlikely. However, Cohen ity. (2010) suggests that, under appropriate conditions, heterostyly may evolve repeatedly in some taxa, includ- Faucal appendages. Faucal appendages are thickenings ing Lithospermum. of the corolla that are located at the intersection of the It should be noted that to infer the pattern of the corolla tube apex and corolla lobe base (Fig. 4c). These evolution of heterostyly, I did not examine gains ⁄losses floral structures may play a role in constricting the apex with either a step matrix or likelihood methods, as of the corolla tube and providing a reward for pollin- others (e.g. Kohn et al., 1996; Ferrero et al., 2009; ators, as glands are produced on the faucal appendages McDill et al., 2009) have done. I did not employ these of some species. These appendages have originated at approaches for two reasons. First, the only component least four times within Lithospermum, and they diagnose of heterostyly included in these analyses was reciprocal two larger clades (Fig. 5, green hash marks). In the herkogamy per se. The other components, self- and present study, this character is coded in binary form intramorph-incompatibility and micromorphological (Table 1), but different types of appendage occur among differences (Barrett et al., 2000), such as morph-specific species of the genus. An examination of the micromor- cell lengths, corolla tube lengths, and pollen sizes, were phology of these appendages may determine if it is not included because insufficient data are available for appropriate to code this character in a more complex these characters in Lithospermum and Boraginaceae. manner. Second, despite the complexity of heterostyly, evidence Species in other groups of Boraginaceae also produce is lacking concerning the rate of evolutionary develop- faucal appendages. The structure of the phylogeny ment of the components of the breeding system. suggests that these appendages have originated at least Therefore it is difficult to hypothesize an appropriate four times among members of the outgroup. Currently, weighting scheme or model for the evolution of hetero- it is not known if the faucal appendages that have styly. Current research has yet to determine if acquisi- resulted from these independent origins have a similar tion of the heterostylous syndrome is two, three, or ten structure. times more ‘‘difficult’’ than its loss, or if it is even more ‘‘difficult.’’ Consequently, I chose to weight gains and Glands. Although not all species of Lithospermum pro- losses of this character equally rather than to employ duce faucal appendages, most members of the genus other analytical strategies. Although the use of equal bear glands on the adaxial surface of their corolla tubes weighting still may be considered an assumption, it (Fig. 4d). These glands are present in the upper portion minimizes the total number of assumptions. of the corolla tube, on the veins that lead to the In addition to flowers that differ in their type of filaments, or both. I coded this character as binary herkogamy, species of Lithospermum exhibit variation in (Table 1), but it is possible to recognize multiple states the position of the stigma relative to the apex of the of this character based on the location of the glands. The corolla tube. The presence of stigmas that are exserted presence of glands is a synapomorphy for two clades, from the corolla tube originated multiple times (Fig. 7), one that includes both Lithospermum and Buglossoides, although the timing of stigma exsertion can occur and one composed of the members of Boragineae before, during, or after anthesis (Johnston, 1952, (Fig. 5, orange hash marks). J.I. Cohen / Cladistics 27 (2011) 559–580 577

In Lithospermum, glands have been lost at least four (Johnston, 1952). Despite the small number, the South times (Fig. 5, purple hash marks). The loss of glands American species of the genus display a range of pollen does not appear to follow any pattern related to corolla shape variation similar to that of the species of North shape, colour, or size. The only correlation I can observe America. In order to resolve the ancestor and recon- is that all species that lack glands also produce flowers struct the character evolution of the South American with exserted stigmas. However, not all species with species of Lithospermum, it will be necessary to include exserted stigmas lack glands. Because of the limited these taxa in future phylogenetic analyses. relationship between the absence of glands and other The present phylogenetic analysis provides evidence floral features, it is difficult to infer any benefit from the that pollen shape is quite evolutionarily labile in loss of glands, or any possible linkage between glands Lithospermum. This also appears to be the case in other and the other floral, vegetative, or pollen characters genera of Boraginaceae. In a study of the pollen of investigated in the present study. Boragineae, Bigazzi and Selvi (1998) described 14 different types among members of the tribe. These Pollen evolution authors recognized two genera, Nonea Medik. and Anchusa L., that produce three and five different types An examination of the phylogenetic distribution of of pollen, respectively. pollen shape can inform our understanding of the The diversity in pollen shape among species of evolution of heterostylous species. Two heterostylous Boraginaceae appears to be much greater than that of species of Lithospermum, L. caroliniense and L. tubuli- many other angiosperm groups. For example, families florum, produce pollen that is dimorphic in both size and such as Myristicaceae (Sauqet and Le Thomas, 2003), shape (Johnston, 1952). In both species, the long-style and genera including Gaultheria L. (Lu et al., 2009) and morph produces hourglass-shaped pollen, and the short- Hemerocallis L. (Xiong et al., 1997), produce pollen that style morph produces ellipsoid pollen. Lithospermum varies much more in exine sculpturing than in shape. caroliniense is nested among species that produce Bigazzi and Selvi (1998) noted various types of exine hourglass-shaped pollen; therefore it appears that ellip- ornamentation among species of Boragineae, but other soid pollen arose within the species. In contrast, the pollen characters varied to a greater degree. I did not relatives of L. tubuliflorum produce ellipsoid pollen, so specifically evaluate exine ornamentation, but my lim- hourglass-shaped pollen originated within this species. ited observations suggest that species of Lithospermum Even though the morphs of these species exhibit the do not exhibit evident differences in exine structure. same pollen shapes, the conditions arose in the opposite Regardless of the pattern of variation in exine orna- order in each species (Fig. 9). mentation, pollen shape seems more diverse and more Cylindrical pollen characterizes the clade that evolutionarily labile in Boraginaceae, both at the intra- includes two Mexican species, L. distichum and familial and intrageneric level, than in most families of L. oblongifolium, and one South American species, angiosperms. L. gayanum. In view of this relationship, it is appropri- Two other pollen characters—pollen pore position ate to hypothesize two mechanisms for the origin of the and pollen pore number—were investigated in a phylo- South American species of Lithospermum. The first is genetic context. Most of the species of Boraginaceae that long-distance dispersal of a Mexican species from included in the present study produce pollen with this clade may have given rise to the South American equatorial pores. In the outgroup, pollen with subequ- species. This hypothesis is supported by the phylogeny atorial pores originated independently in Buglossoides of Weigend et al. (2010a). However, their analysis purpureo-caerulea,inArnebia benthamii (Wall. ex G. includes a very limited sampling of species of Lithosper- Don) I.M. Johnst., and in the ancestor of Echium and mum, and the resulting phylogeny is constructed solely Podonosma (Fig. 9, grey hash marks). In Lithospermum, from ITS sequence data, a region with known issues for the equatorial pollen pore position is ancestral (Fig. 9, use in phylogeny reconstruction (Alvarez and Wendel, aqua hash marks). This state has been lost five times, 2003) and with multiple copies in some species of resulting in species that produce pollen with a subequ- Lithospermum (personal observation). In contrast, it atorial pore position. The latter condition characterizes also is suggested that the ancestor of the extant two ingroup clades (Fig. 9, grey hash marks). One L. mediale I.M. Johnst.—a species distributed from includes L. flavum, L. johnstonii, and L. exsertum; the Guatemala to Colombia and with a floral form similar other comprises L. rosei, L. canescens, and 13 additional to that of both L. distichum and L. gayanum—could species. Among the members of this larger clade, have colonized the Andes, and subsequently given rise to equatorial pollen pores have originated secondarily only the Andean members of the genus. However, L. mediale in one species, L. multiflorum. produces ellipsoid, not cylindrical, pollen. Additionally, All species of Lithospermum produce pollen with 6–8 the four other South American species of Lithospermum pores, and this is the ancestral condition for the produce ellipsoid, spherical, or cylindrical pollen phylogeny. This type of pollen has been lost five times 578 J.I. Cohen / Cladistics 27 (2011) 559–580 in Boraginaceae, resulting in pollen with either a larger Garden of Belgium, the Royal Botanical Gardens at or smaller number of pores. One of these instances Kew, and the South African National Biodiversity involves the evolution of pollen with three pores. Institute provided me with plant material or DNA Triporate pollen is a synapomorphy for the clade that isolations which I used to collect DNA sequence data. includes Echium, Podonosma, Maharanga, and Onosma The funding provided by the American Society of Plant (described above). Pollen with more than eight pores Taxonomists, Harold Moore Jr Funds, Molecular and originated twice, once each in Symphytum and Arnebia. Organismic Research in Plant History, Cornell Univer- The pollen grains produced by members of each of these sityÕs Latin American Studies Program, and a research genera differ. In Symphytum, the pores are distributed in travel grant from the Graduate School at Cornell one row around the equator of the pollen grain (Bigazzi University helped support this project, and that was and Selvi, 1998). In Arnebia, the pores are distributed in greatly appreciated. two rows, with one row above and one row below the equator of the pollen grain (Johnston, 1952). This latter condition is uncommon in Boraginaceae. Pollen shape References seems to vary more than pollen pore position or pollen pore number, but this hypothesis should be examined Aizen, M., 2003. Down-facing flowers, hummingbirds, and rain. further with increased sampling of taxa and palynolog- Taxon 52, 675–680. Alfaro, M.E., Zoller, S., Lutzoni, F., 2003. Bayes or bootstrap: a ical characters. simulation study comparing the performance of Bayesian Markov chain Monte Carlo and bootstrapping in assessing phylogenetic confidence. Mol. Biol. Evol. 20, 255–266. Conclusion and future directions Al-Shehbaz, I.A., 1991. The genera of Boraginaceae in the southeast- ern United States. J. Arnold Arbor., Suppl. Ser. 1, 1–169. Alvarez, I., Wendel, J.F., 2003. Ribosomal ITS sequences and plant Within Lithospermum and Boraginaceae, the evolu- phylogenetic inference. Mol. Phylogenet. Evol 29, 417–434. tion of morphological, palynological, and vegetative Armbruster, W.S., Mulder, C.P.H., Baldwin, B.G., Kalisz, S., Wessa, characters follows both simple and complex patterns. B., Nute, H., 2002. Comparative analysis of late floral development The combined analysis presented here provides only a and mating-system evolution in tribe Collinsieae (Scrophulariaceae limited amount of information concerning character s.l.). Am. J. Bot. 89, 37–49. Aytas Akc¸ in, T., Ulu, S., 2007. A morphological and anatomical evolution within Boraginaceae. The present study study on Anchusa leptophylla Roemer & Schultes (Boraginaceae) includes both cpDNA regions and morphological char- distributed in the Black Sea Region of Turkey. Turk. J. Bot. 31, acters that can be aligned or coded for any species in the 317–325. family. Additionally, with more observations of certain Barrett, S.C.H., Jesson, L.K., Baker, A.M., 2000. The evolution of floral and vegetative features, it may be appropriate to stylar polymorphisms in plants. Ann. Bot. (Oxford) 85 (Suppl. A), 253–265. recode some characters, including glands and faucal Bateman, R.M., 1999. Integrating molecular and morphological appendages, in other manners. evidence of evolutionary radiations. In: Hollingsworth, P.M., It is possible to integrate current molecular biology Bateman, R.M., Gornall, R.J. (Eds.), Molecular Systematics and tools, classical developmental techniques, and data Plant Evolution. Taylor & Francis, London, pp. 432–471. Bigazzi, M., Selvi, F., 1998. Pollen morphology in the Boragineae related to the evolution of structural features. For (Boraginaceae) in relation to the taxonomy of the tribe. Plant Syst. example, identification of the genes that affect floral Evol. 213, 121–151. organ length will enhance our understanding of the Boyd, A.E., 2002. Morphological analysis of Sky island populations of evolutionary changes necessary to alter corolla shape Macromeria viridiflora (Boraginaceae). Syst. Bot. 27, 116–126. and type of herkogamy. Given the degree of structural Boyd, A.E., 2003. Phylogenetic relationships and corolla size evolution among Macromeria. Syst. Bot. 28, 118–129. diversity within Lithospermum, the genus may be a Boyd, A.E., 2004. Phylogenetic relationships and corolla size evolution useful model in which to investigate the genetics, among Macromeria (Boraginaceae). Syst. Bot. 28, 118–129. development, and evolution of morphological charac- Cohen, J., 2010. ‘‘A case to which no parallel exists’’: the influence of ters. DarwinÕs Different Forms of Flowers. Am. J. Bot. 97, 701–716. Cohen, J.I., Davis, J.I., 2009. Nomenclatural changes in Lithospermum (Boraginaceae) and related taxa following a reassessment of phylogenetic relationships. Brittonia 61, 101–111. Acknowledgements Daniel, T.F., McDade, L.A., Manktelow, M., Kiel, C.A., 2008. The ‘‘Tetramerium Lineage’’ (Acanthaceae: Acanthoideae: Justicieae): I would like to thank Jerry Davis, Alejandra Nixon- delimitation and intra-lineage relationships based on cp and nrITS Gandolfo, Melissa Luckow, Jim Miller, Sue Sherman- sequence data. Syst. Bot. 33, 416–436. Davis, J.I., Stevenson, D.W., Peterson, G., Seberg, O., Campbell, L.M., Broyles, Caroline Kellogg, Amy Boyd, and an anony- Freudenstein, J.V., Goldman, D.H., Hardy, C.R., Michelangeli, mous reviewer for helpful comments on this manuscript. F.A., Simmons, M.P., Specht, C.D., Vergara-Silva, F., Gandolfo, I very much appreciated all of the herbaria that supplied M., 2004. A phylogeny of monocots, as inferred from rbcL and atpA material for me to examine. The Missouri Botanical sequence variation, and a comparison of methods for calculating Garden, the Cornell Plantations, the National Botanical jackknife and bootstrap values. Syst. Bot. 29, 467–510. J.I. Cohen / Cladistics 27 (2011) 559–580 579

Dı´ ez, M.J., Valde´ s, B., Ferna´ ndez, I., 1986. Pollen morphology of subfamily Vaccinioideae: taxonomic and evolutionary significance. Spanish Lithospermum s.l. (Boraginaceae) and its taxonomic Rev. Palaeobot. Palynol. 154, 106–123. significance. Grana 25, 171–176. Luckow, M., Brunneau, A., 1997. Circularity and independence in Eernisse, D.J., Kluge, A.G., 1993. Taxonomic congruence versus total phylogenetic tests of ecological hypotheses. Cladistics 13, 145–151. evidence, and amniote phylogeny inferred from fossils, molecules, McDill, J., Repplinger, M., Simpson, B.G., Kadereit, J.W., 2009. The and morphology. Mol. Biol. Evol. 10, 1170–1195. phylogeny of Linum and Linaceae subfamily Linoideae, with Fægri, K., van der Pijl, L., 1979. The Principles of Pollination Ecology. implications for their systematics, biogeography, and evolution of Pergamon, Oxford, UK. heterostyly. Syst. Bot. 34, 386–405. Farris, J.S., Albert, V.A., Kallersjo, M., Lipscomb, D., Kluge, A.G., McKown, A.D., Cochard, H., Sack, L., 2010. Decoding leaf hydraulics 1996. Parsimony jackknifing outperforms neighbor-joining. Cla- with a spatially explicit model: principles of venation architecture distics 12, 99–124. and implications for its evolution. Am. Nat. 175, 447–460. Ferrero, V., Arroyo, J., Vargas, P., Thompson, J.D., Navarro, l., 2009. Miller, J.S., Venable, D.L., 2003. Floral morphometrics and the Evolutionary transitions of style polymorphisms in Lithodora evolution of sexual dimorphism in Lycium (Solanaceae). Evolution (Boraginaceae). Perspect. Plant Ecol. Evol. Syst. 11, 111–125. 57, 74–86. Goloboff, P., 1999a. Analyzing large data sets in reasonable times: Morris, J.A., 2007. A Molecular Phylogeny of the Lythraceae and solutions for composite optima. Cladistics 15, 415–428. Inference of the Evolution of Heterostyly. PhD Thesis, Kent State Goloboff, P., 1999b. NONA (NO NAME) ver. 2. Published by the University, Kent, OH, USA. author, Tucuma´ n, Argentina. Nixon, K.C., 1999. Parsimony ratchet, a new method for rapid Goloboff, P., Farris, J.S., Nixon, K.C., 2008. TNT, a free program for parsimony analysis. Cladistics 15, 407–414. phylogenetic analysis. Cladistics 24, 774–786. Pe´ rez-Barrales, R., Vargas, P., Arroyo, J., 2006. New evidence for the Graham, S.W., Barrett, S.C.H., 2004. Phylogenetic reconstruction of Darwinian hypothesis of heterostyly: breeding systems and pollin- the evolution of stylar polymorphisms in Narcissus (Amaryllida- ators in Narcissus sect. Apodanthi. New Phytol. 171, 553–567. ceae). Am. J. Bot. 91, 1007–1021. Richards, A.J., 1997. Plant Breeding Systems, 2nd edn. Chapman & Grant, V., Grant, K.A., 1970. A hummingbird-pollinated species of Hall, London. Boraginaceae in the Arizona flora. Proc. Natl. Acad. Sci. USA 66, Roquet, C., Sa´ ez, L., Aldasoro, J.J., Susanna, A., Alarco´ n, M.L., 917–919. Garcia-Jacas, N., 2008. Natural delineation, molecular phylogeny, Harrison, C.J., Mo¨ ller, M., Cronk, Q.C., 1999. Evolution and and floral evolution in Campanula. Syst. Bot. 33, 203–217. development of floral diversity in Streptocarpus and Saintpaulia. Roth-Nebelsick, A., Uhl, D., Mosbrugger, V., Kerp, H., 2001. Ann. Bot. (Oxford) 84, 49–60. Evolution and function of leaf venation architecture: a review. Herlihy, C.R., Eckert, C.G., 2005. Evolution of self-fertilization at Ann. Bot. (Oxford) 87, 553–566. geographical range margins? A comparison of demographic, floral, Sauqet, H., Le Thomas, A., 2003. Pollen diversity and evolution in and mating system variables in central vs. peripheral populations of Myristicaceae (Magnoliales). Int. J. Plant Sci. 164, 613–628. Aquilegia canadensis (Ranunculaceae). Am. J. Bot. 92, 744–751. Scotland, R.W., Olmstead, R.G., Bennett, J.R., 2003. Phylogeny Hillis, D.M., Bull, J.J., 1993. An empirical test of bootstrapping as a reconstruction: the role of morphology. Syst. Biol. 52, 539–548. method for assessing confidence in phylogenetic analysis. Syst. Selvi, F., Bigazzi, M., 2003. Revision of Anchusa in Greece. Bot. J. Biol. 42, 182–192. Linn. Soc. 142, 431–454. Jian-Chang, N., Yi-zhen, X., Yu-long, Z., 1995. A comparative Stebbins, G.L., 1957. Self fertilization and population variability in palynological study on Maharanga and Onosma (Boraginaceae). higher plants. Am. Nat. 91, 337–354. Acta Phytotax. Sin. 33, 52–57. Takebayashi, N., Morrell, P.L., 2001. Is self-fertilization an evolu- Johnston, I.M., 1952. Studies in the Boraginaceae, XXIII. A survey of tionary dead end? Revisiting an old hypothesis with genetic the genus Lithospemum. With three plates. J. Arnold Arbor. 33, theories and a macroevolutionary approach. Am. J. Bot. 88, 299–366. 1143–1150. Johnston, I.M., 1953a. Studies in the Boraginaceae, XXIV. A. Three Thomas, D.C., Weigend, M., Hilger, H.H., 2008. Phylogeny and genera segregated from Lithospermum. B. Supplementary notes on systematics of Lithodora (Boraginaceae-Lithospermeae) and its Lithospermum. J. Arnold Arbor. 34, 1–16. affinities to the monotypic genera Mairetis, Halacsya and Para- Johnston, I.M., 1953b. Studies in the Boraginaceae, XXV. A revalu- moltkia based on ITS1 and trnLUAA - sequence data and ation of some genera of the Lithospermeae. J. Arnold Arbor. 34, morphology. Taxon 57, 79–97. 258–300. Tripp, E.A., 2007. Evolutionary relationships within the species-rich Johnston, I.M., 1954a. Studies in the Boraginaceae, XXVI. Further genus Ruellia (Acanthaceae). Syst. Bot. 32, 628–649. revaluations of the genera of the Lithospermeae. J. Arnold Arbor. Tripp, E.A., Manos, P.S., 2008. Is floral specialization and evolution- 35, 1–81. ary dead end? Pollination system transitions in Ruellia (Acantha- Johnston, I.M., 1954b. Studies in the Boraginaceae, XXVII. Some ceae). Evolution 62, 1712–1737. general observations concerning the Lithospermeae. J. Arnold Valentine, D.H., Chater, A.O. (Eds.), 1972. Boraginaceae. In: Tutin, Arbor. 35, 158–166. T.G., Heywood, V.H., Burges, N.A., Moore, D.M., Valentine, Johnston, M.O., Schoen, D.J., 1996. Correlated evolution of self- D.H., Wallers, S.M., Webb, D.A. (Eds.), Flora Europaea. Vol. 3, fertilization and inbreeding depression: an experimental study of Diapensiaceae to Myoporaceae. Cambridge University Press, nine populations of Amsinckia (Boraginaceae). Evolution 50, 1478– London, pp. 83–122. 1491. Webb, C.J., Lloyd, D.G., 1986. The avoidance of interference between Jones, C.S., Bakker, F.T., Schlichting, C.D., Nicotra, A.B., 2009. Leaf the presentation of pollen and stigmas in angiosperms II. Herk- shape evolution in the South African genus Pelargonium LÕHe´ r. ogamy. N.Z. J. Bot. 24, 163–178. (Geraniaceae). Evolution 63, 479–497. Weigend, M., Gottschling, M., Selvi, F., Hilger, H.H., 2009. Kohn, J.R., Graham, S.W., Morton, B., Doyle, J.J., Barrett, S.C.H., Marbleseeds are gromwells—systematics and evolution of Litho- 1996. Reconstruction of the evolution of reproductive characters in spermum and allies (Boraginaceae tribe Lithospermeae) based on Pontederiaceae using phylogenetic evidence from chloroplast DNA molecular and morphological data. Mol. Phylogenet. Evol. 52, restriction-site variation. Evolution 50, 1454–1469. 755–768. Lu, L., Fritsch, P.W., Wang, H., Li, H.-T., Li, D.-Z., Chen, J.-Q., Weigend, M., Gottsching, M., Hilger, H.H., Nu¨ rk, N.M., 2010a. 2009. Pollen morphology of Gaultheria L. and related genera of Five new species of Lithospermum L. (Boraginaceae tribe 580 J.I. Cohen / Cladistics 27 (2011) 559–580

Lithospermeae) in Andean South America: another radiation in the Supporting Information Amotape-Huancabamba zone. Taxon 59, 1161–1179. Weigend, M., Gottschling, M., Selvi, F., Hilger, H.H., 2010b. Fossil and extant western hemisphere Boraginaceae, and the polyphyly of Additional Supporting Information may be found in ‘‘Trigonotideae’’ Reidl (Boraginaceae: Boraginoideae). Syst. Bot. the online version of this article: 35, 409–419. Table S1. Species included in analyses, along with Wiens, J.J., 2004. The role of morphological data in phylogeny collection information and Genbank numbers. reconstruction. Syst. Biol. 53, 653–661. Table S2. Strict consensus of morphological matrix Wolfe, A.D., Randle, C.P., Datwyler, S.L., Morawetz, J.J., Arguedas, N., Diaz, J., 2006. Phylogeny, taxonomic affinities, and biogeog- (L = 165; CI ⁄RI = 0.29–0.30 ⁄0.66–0.67). Numbers raphy of Penstemon (Plantaginaceae) based on ITS and cpDNA above branches are jackknife values > 50%. sequence data. Am. J. Bot. 93, 1699–1713. Please note: Wiley-Blackwell are not responsible for Xiong, Z.-T., Chen, S., Hong, D., Luo, Y., 1997. Pollen morphology the content or functionality of any supplementary and its evolutionary significance in Hemerocallis (Liliaceae). Nord. materials supplied by the authors. Any queries (other J. Bot. 18, 183–189. Zhu, G., Riedl, H., Kamelin, R.V., 1995. Boraginaceae. In: Wu, Z.Y., than missing material) should be directed to the corre- Raven, P.H. (Eds.), Flora of China. Vol. 16 (Gentianaceae through sponding author for the article. Boraginaceae). Science Press, Beijing and Missouri Botanical Garden Press, St Louis, MO, pp. 329–427.