Aliso: a Journal of Systematic and Evolutionary Botany Phylogenetic Relationships in Phoradendreae (viscaceae) Inferred from Three Regions of the Nuclear Ribosomal Cistron. Ii. the North American Species of Phoradendron Phylogenetic Relationships in Phoradendreae (viscaceae) Inferred from Three Regi

A parsimony analysis based on sequences from the ITS region and two partitions of the 26S subunit of nuclear ribosomal DNA was used to infer phylogenetic relationships among the North American species of Phoradendron. A strongly supported clade united all but one of the species typically lacking ca taphylls, a character used previously to distinguish the northern species from those of Central and South America. The divergent placement of P. califo rnicum rel ative to the members of this " north ern" clade confirmed the hypothesis that s peci es lacking cataphylls are polyphyletic. Four of five species parasitic on conifers formed a well-supported clade. However. a strongly supported relati onship between P. rhipsa/inum and P. bra chystachyum, the former a parasite of co nifers, renders con ifer par-asitism homoplastic. A s ister group rel ationsh ip between these two species is not appa re nt from morphological evidence. A clade unit ing P. serotinum, P. tomentosum, and P. velutinum was strongly suppo rted. A broad host range characterized two of the three lineages of the basal trit omy in the northern clade, whereas the third line age united species specialized in parasitism of oaks or conifers.


INTRODUCTION
The two mistletoe genera Phoradendron Nutt. and Dendrophthora Eichler together compose the New World tribe Phoradendreae Engler. Both genera occur in Central America, the West Indies, and tropical latitudes of South America. However, only Phoradendron extends into the subtropical and temperate latitudes of North and South America. Phoradendron species parasitize a wide range of angiosperms and less commonly conifers. Trelease (1916) distinguished 240 species, but more recent estimates fluctuate between some 100 species or less (Kuijt 1982) and over 200 species (1. Kuijt pers. comm.). This discrepancy is due to the continual discovery of new species, but above all to the difficulty in circumscribing species, reflecting an unusual morphology in which few characters can be scored as discrete and many cannot be scored at all in a substantial subset of species. This paper represents the second of two contributions on phylogenetic relationships in Phoradendreae, the first dealing with major clades and the second focusing on Wiens's revision (1964) of Phoradendron species in North America.

Morphology
Phoradendreae are characterized by an articulated spike, whereby successive internodes of the inflores- cence are arranged in an opposite decussate pattern (Fig. 1). Inflorescence internodes vary in number from one ( Fig. 2) to ca. 10. Each internode consists of two flower areas on opposite sides of the inflorescence axis (Fig. 1). These flower areas are studded with tiny, sessile, mostly trimerous flowers arranged in vertical rows (series), the apical flower either remaining the only flower in its row (biseriate inflorescence; Fig. 3) or crowning a central row (triseriate inflorescence; Fig.  4). The number of series cannot be discerned in species where each side of the inflorescence internode holds an insufficient number of flowers (one to four; e.g., Fig. 2). Among the Phoradendron species in North America, triseriate inflorescences are the rule, the biseriate type predominating in species farther south.
Vegetatively, the North American species of Phoradendron are characterized by the predominant absence of cataphylls (Fig. 5) , These small leaflike or bractlike structures are inserted in pairs on the first internode of lateral branches (Fig. 6) and are typical of tropical species. However, occasionally they are formed in several North American species said to lack cataphylls. Usually, cataphylls are not associated with floral organs (Fig. 6), but some species produce fertile cataphylls that subtend inflorescences, Cataphylls differ from foliage leaves by their much reduced size, In some species, however, foliage leaves are reduced to scales. The distinction between scalelike foliage leaves and cataphylls is then often difficult to diagnose. Primary orientation refers to the orientation of the lowermost (primary) foliar organs (cataphylls, if present, otherwise leaves) on lateral branches in relation to the main branch axis . In the majority of tropical species and several species in North America, the primary orientation is consistently median (Fig. 7), whereas most North American species have a consistently transverse orientation (Fig. 8). Two species, however, exhibit both orientation types.

Taxonomy
Two taxonomic treatments (Trelease 1916;Wiens 1964) cover the species of Phoradendron in North America. Trelease's monograph of the entire genus places the majority of North American species in " Boreales," an informal group comprising 57 taxa (50 species). Its members are characterized by the absence of cataphylls. However, the monophyly of the North American species based on the absence of cataphylls has been questioned, in that certain predominantly acataphyllous species occasionally develop cataphylls (Kuijt 1996), while the acataphyllous P. californicum shares other morphological traits with southern species (Kuijt 1959(Kuijt , 1996Wiens 1964). Trelease placed P. californicum, a scale-leaved species, together with two other scale-leaved species in " A p hyllae, " a subdivision of section Pauciflorae Engler. Wiens's revision (1964) is confined to the acataphyllous species of Phoradendron in North America. It covers the same suite of taxa as Trelease's "Boreales," whereby Wiens's 24 taxa (18 species) are a result of placing many of Trelease's taxa in synonymy. Wiens retained P. californicum in his revision, but placed it in monotypic section Phoradendron. The two taxonomic treatments also differ in the placements of P. longifolium, P. robinsonii, and P. scaberrimum in Pluriseriales Engler by Trelease (1916), but in Calyculatae Trelease by Wiens (1964).
Here, I develop a phylogenetic hypothesis for the North American species of Phoradendron using molecular sequences of the nuclear rDNA. The phylogenetic framework is then used to compare the taxonomic treatments by Trelease and Wiens in a phylogenetic context. The distribution of several morphological and life history characters reported in the literature (Wiens 1964;Kuijt 1996) on the molecular phylogeny is used to infer additional phylogenetic structure.

MATERIALS AND METHODS
Of the 43 taxa sampled, 34 belong to Phoradendron, six to Dendrophthora sen su Kuijt (1961) and three are outgroup species belonging to Viscaceae (Korthalsella latissima) and Santalaceae (Acanthosyris asipapote and Thesium carinatum ). Dendrophthora species are considered part of the ingroup because re-cent evidence suggests that Phoradendron may not be monophyletic unless Dendrophthora is included within it (Nickrent and Duff 1996;Ashworth 2000). The 34 Phoradendron taxa include 19 North American representatives of Wiens's and Trelease's treatments, one species (P. rhipsalinum) that was unknown to Trelease and Wiens, and 14 other species, primarily from Central and/or South America and the West Indies (Table  I). All plants collected by the author in Mexico were sent to J. Kuijt for identification. For convenience, nomenclature follows Wiens (1964) for the North American acataphyllous species and Trelease (1916) for the remainder, even though some nomenclatural modifications have been made since then (e.g., P. serotinum is now P. leucarpum; Reveal and Johnston 1989). In most cases, a species is represented by a single sequence and for multiply sequenced species nucleotide polymorphisms at a particular site are encoded in terms of standard ambiguity codes. However, multiple sequences were retained for several species in order to represent separate subspecies (P. bolleanum, P. juniperinum, P. tomentosum, and P. villosum) and geographic differentiation (P. tomentosum from San Luis Potosi, Mexico, and Texas and P. velutinum from San Luis Potosi and Michoacan, Mexico).
The sequences used in this study comprise the ITS region and two partitions of the 26S nuclear rDNA (Fig. 9). Rapid sequence evolution in Viscaceae (Nickrent and Franchina 1990;Nickrent and Starr 1994;Nickrent et aI. 1994;Nickrent and Soltis 1995;Nickrent and Duff 1996;Molvray et aI. 1999) prompted the use of a composite gene region in which slowerevolving sequences enabled alignment with outgroups and the more divergent ingroup species, while the faster-evolving sequences resolved relationships between closely related taxa. A detailed account of the rationale dictating taxon sampling, outgroup selection, and choice of an appropriate gene region is given elsewhere (Ashworth 2000).
Methodology pertaining to DNA extraction followed the 2X CTAB buffer protocol by Doyle and Doyle (1987), with DNA extracts diluted to a final concentration of 10 ng/f..L1. Double-stranded DNA template was amplified by the polymerase chain reaction (Mullis et aI. 1986) using amplification conditions detailed in Ashworth (2000). The three DNA regions were amplified and cycle sequenced using various combinations of forward and reverse primers ITS2, ITS3, ITS4, ITS4i, ITS5, ITS5i (White et al. 1990), 1643f (CCGCCCGTCGCTCCTACCG) and ITS4-172 (CCCAAACAACCCGACTCACCAACAG) (Ashworth 2000) for the ITS region, and 641 r, 1715f and 2134r (D.L. Nickrent), S I' and S2 (Bult and Zimmer 1993) for 26S rDNA. The two regions of 26S rDNA (D2 and D8; Fig. 9) are delimited by primer pairs S2 + 641r and 1715f + 2134f, respectively. They corre- Table I. Specimens used in the study. Unless otherwise stated, all are deposited at RSA. Nomenclature follows Wiens (1964) for species from North America and Trelease (1916) 9. Diagram of the ITS region and 26S large subunit of nuclear rONA drawn to scale. Gene regions sequenced in this study (ITS I and ITS2 spacers, the S.8S portion of the small-subunit nuclear rONA, and the 02 and 08 regions) are printed in bold. The 02 region is bounded by primers S2 and 641 r and the 08 region by primers 171Sf and 2134r. The 02 and 08 divergent domains (light gray) are regions with increased numbers of nucleotide substitution relative to the more conserved flanking sequences of the 26S nuclear rONA. They are located within the 02 and 08 regions (dark gray), respectively. spond to positions 1128-1495 and 2583-2983 in the 26S sequence of Arabidopsis thaliana (GenBank X52320; Unfried and Gruendler 1990).
Sequences were generated on an Applied Biosystems 373A automated DNA sequencer using a Sequagel-6 polyacrylamide gel (National Diagnostics). The sequences were assembled, edited, and combined into a consensus sequence using Sequencher 3.0 (Gene Codes Corporation, Inc ., Ann Arbor, MI). All sequences are available from GenBarIk and accession numbers are provided in Table I. Aligned sequences were submitted to parsimony analysis using PAUP* version 4.0 131 (Swofford 1998). Several alternative alignments, including computer-assisted (ClustalW version 1.4; Thompson et al. 1994) and visual alignments, were tested. The visually aligned matrix is archived in Ashworth (1999 ;Ph .D. thesis) and posted on the Internet at the website: http://www.science.siu.edu/parasitic-plants/ AlignmentslPhor.align.htrnl. Parsimony analyses used heuristic searching with accelerated tran sformation optimization and TBR branch swapping. Starting trees were obtained by simple addition or 100 cycles of random addition, and branch swapping restricted to the best trees only. During branch swapping, the STEEP-EST DESCENT option was switched off. All optimal trees were saved (MULPARS).
Branch support was calculated using 1000 bootstrap replicates (Felsenstein 1985), as implemented on PAUP* ver sion 4.0131 in the "Fast stepwise addition" function, and using the decay approach (Bremer 1988;Donoghue et al. 1992). Bootstrap values of at least 70% are taken to indicate strong branch support (Hillis and Bull 1993; but see Efron et al. 1996). Homoplasy of the sequence data was calculated using the consistency index (Kluge and Farris 1969;Sanderson and Donoghue 1989) and the retention index (Farris 1989). Gaps were coded as missing data.

RESULTS
The total length of the aligned data matrix was 1602 nucleotides, of which 813 nucleotides resided in the ITS region, 382 in the D2 region and 407 in the D8 reg ion. Of 800 variable sites, 594 were potentially parsimony-informative. Missing data accounted for 3.96% of the data matrix. Lengths of sequences of the ingroup taxa (excluding partially sequenced taxa and gaps) ranged from 1380 to 1496 nucleotides, with an average of 1412.
Sequence alignment revealed the presence of 36 indels. Fourteen are relevant in the context of this study. Eight support branches on the strict consensus tree (Fig. 10) and six are homoplastic (e, g', n, y, I, J), of which one (n) is shared by two of three branches in a tritomy. Ten of the indels are 1-2 nucleotides in length (five each), three are 3 nucleotides long and one is four nucleotides long. All but two are located in the ITS region. Figure 10 depicts the strict consensus of four equally most parsimonious trees (MPTs) of 2234 steps. Of a total of 37 ingroup nodes, 27 have bootstrap values of~70%, 28 have decay values of~3, and 25 have both. A bootstrap value of 88% and a decay value of 10 support a clade, henceforth called the northern clade (Fig. 10), including P. rhipsalinum and all members of Trelease's "Boreales" and Wiens' revision except P. californicum. Phoradendron californicum appears as sister to a clade that includes not only members of the northern clade, but several other species (Dendrophthora guatemalensis, P. brevifolium, P. forestierae, P. reichenbachianum, and P. vernicosum) that have cataphylls. The precise placement of P. californicum is uncertain owing to weak bootstrap support (67%; Fig. 10) and a long branch (146 nucleotide substitutions; Fig. II). Interestingly, the branch decays only when all trees more than five steps longer than the most parsimonious solution are included in the analysis.
Within the northern clade, several clades receive strong support (Fig. 10). The first of these combines P. brachystachyum and P. rhipsalinum, the second P. serotinum, P. tomentosum, and P. velutinum (henceforth the P. se rotinum clade), the third P. galeottii and P. longifolium, the fourth P. villosum subsp. coryae and P. scaberrimum, and the fifth P. bolleanum subsp. bolleanum to P. capitellatum, seven taxa parasitic on conifers occurring in western North America (henceforth the conifer-parasite clade). In the P. serotinum clade, P. serotinum and P. tomentosum form a clade whose sister species is P. velutinum. Within the conifer-parasite clade, three clades each unite (a) the two subspecies of P. juniperinum, (b) two of the three subspecies of P. bolleanum (subspp. den sum and paucifiorum) , and (c) P. minutifolium and P. bolleanum subsp. bolleanum.

Phoradendron californicum
The divergent position of P. californicum argues strongly against the monophyly of "Boreales" sensu Trelease (1916) or a monophyletic acataphyllous lineage sensu Wiens (1964) (Fig. II). Instead, it agrees with morphological evidence (shared biseriate inflorescences) that suggests a closer relationship of P. californicum with cataphyllous tropical species (Kuijt 1959(Kuijt , 1996. Regrettably, P. olae, its putatively closest relative (Kuijt 1997), is not part of this analysis. The data also suggest two independent origins of scale leaves (Fig. 10). Wiens (1964)    11) may reflect accelerated sequence evolution, extinctions that eliminated close relatives, or incomplete sampling. Incomplete sampling is perhaps the simplest explanation, in that this analysis includes less than half of Phoradendron species and ca. 10% of Dendrophthora species.

Parasites of Conifers
Pauciflorae sensu Wiens (excluding P. californicum; Fig. 11) is monophyletic, and its members form a fairly uniform group in terms of morphology (small to scalelike leaves and few-flowered spikes), host preference (conifers, usually forming highly specific associations; Fig. 12) and nucleotide sequence data (bootstrap 96%, decay of 9). Strongly supported sister group relationships for three species pairs suggest that (1) P. juniperinum is monophyletic, but (2) P. bolleanum is not. Further evidence for the polyphyly of P. bolleanum comes from indel data. Although node support is weak (bootstrap of 67%, decay of 4), the clade uniting the three P. bolleanum subspecies and P. minutifolium is supported by insertion b. Insertion d unites P. bolleanum subspp. densum and pauciflorum but not P. bolleanum subsp. bolleanum (Fig. 10, 12). Phoradendron bolleanum is monophyletic only if P. minutifolium is placed in synonymy or subspp. densum and pauciflorum are treated as species. Interestingly, plants presumed to have originated by hybridization between P. bolleanum subsp. densum and P. juniperinum resemble P. minutifolium (Vasek 1966;Wiens and DeDecker 1972).
The sister group relationship between the two subspecies of P. juniperinum is strongly supported. Phoradendron juniperinum subsp. libocedri differs from the typical subspecies by a more pendulous habit and longer internodes that mimic the morphology of its host, Calocedrus decurrens (Torr.) Florin. The typical variety is a parasite of Juniperus and Cupressus species. Three MPTs (not shown) support a sister group relationship between P. juniperinum and P. capitellatum. The latter is the most divergent within Pauciflorae, by virtue of numerous autapomorphic nucleotide substitutions (Fig. 11) and a deletion of 44 nucleotides in the ITS 1 region. Phoradendron capitellatum is also the only species in the conifer-parasite clade to flower in winter (December-February; Wiens 1964), although reports of summer flowering do exist (1. Kuijt pers. comm.). Only one of the four MPTs (not shown) shows P. capitellatum as sister to a clade composed of the three P. bolleanum subspecies and P. minutifolium. Fosberg (1941) placed P. capitellatum as subspecies under P. bolleanum whilst retaining P. minutifolium at the species level.

Phoradendron rhipsalinurn
Pauciflorae sensu Wiens (1964) includes all but one species that parasitize conifers (primarily Cupressaceae s.l.), The discovery of P. rhipsalinum (Calderon de Rzedowski and Rzedowski 1972) occurred after Wiens's revision, but the authors suggested its pl acement in Pluriseriales sensu Wiens (1964). By contrast, Kuijt (1996) suggested it "may be a distant relative of P. bolleanum." Unlike the members of Paucifiorae, however, P. rhipsalinum has elongated inflorescences with several internodes, occasionally develops cataphylls (Kuijt 1996), and has both median and transverse primary orientation. Both morphological and molecular data thus agree that P. rhipsalinum does not ne atly fit into Pauciflorae.
Far more surprising is the strongly supported relationship between P. rhipsalinum and P. brachystachyum, the latter a parasite of angiosperms (Fig. 12).
Three indels (f, x, and C) are uniquely shared by these two species (Fig. 10) . Interestingly, P. brachystachyum also occasionally produces cataphylls (Kuijt 1996). Phoradendron rhipsalinum is an endemic of central Mexico where it is found exclusively on Taxodium mucronatum Ten. Like P. juniperinum subsp. libocedrio it appears to mimic the pendulous branches of its host by means of long internodes and , in this case, linear leaves. Little is known about the biology of this species. Phoradendron brachystachyum, by contrast, is widely distributed in Mexico, grows on a diverse array of angiosperms (especially Fabaceae) and exhibits considerable variation in leaf shape, without apparent host mimicry. Vegetative mimicry (leaves) is well documented among certain Australian Loranthaceae ( Barlow and Wiens 1977).

Parasites of Angiosperms
Within the P. serotinum clade, P. serotinum and P. tomentosum are sister species (Fig. 10). They share medium-sized (ca. 25-30 mm long X 12-20 mm wide), orbicular to obovate leaves having somewhat obtuse apices and inconspicuous venation and sometimes a tomentum that tends to diminish or disappear with maturity (Wiens 1964). They parasitize a wide range of angiosperms, including Betulaceae, Fabaceae, Fagaceae, Oleaceae, and Platanaceae. Two unique synapornorphic insertions (j and E) support the sister group relationship between P. serotinum and P. tomentosum ( Fig. 10 and 12). Sister species to this clade is P. velutinum. This species differs from the preceding two by a yellower color and more elongate leaves with acute tips and tapered bases. Both P. velutinum specimens sampled were found growing on Rosaceae and not oak, the primary host reported by Wiens (1964) Fig. 12. Excerpt o f the strict con sen sus tree from a maximum parsimony ana lys is of thre e regions of nuclear rDNA. Geographic distribution. indel s, morphological traits (Kuijt 1996). life histor y c haracters (Wi en s 1964). and preferred host s (Wiens 1964 a nd personal observations) are indica ted for the me mbers of the northern cl ad e (gray box ) and adja cent parts of the tree . Thickened branches have weak bootstrap support «70%). Onl y indel s pert aining to species in the northern cl ade are mapped onto the tree . Homoplastic ind el s are indicated in bold. Abbrev iat ion s: fc: fertile cata phylls: W: winter flowering , S: summer flowering, F: fall flowering. yr: flowering year-round; t: transverse orientation, m: median orientation, mit: both types of orientation; a: cat aphylls absent, c: ca taphylls present, ale: cataphylls so metimes present. Asterisks denote host ranges gathered from surveys of herbarium specimens; all other host ranges are from Wiens (1964). Geographic distribution is indi cated by a boxed grid. in which one to three blackened squares respectively indicate a distribution that extends no farth er south than 26°, 22°, and 18°N latitude and four blackened squares represent a distribution extending south of--or occ urring exclusivel y south of-18°N latitude . All distribu tion ranges are from Wiens ( 1964). The three latitudes correspond approximatel y to northern Sinaloa, central Agu ascali entes. and southern Puebla, Mexico. distance of 42 nucleotide substitutions between the two P. velutinum seq uences (Fig. II) perhaps reflects a high degree of polymorphism of this species (Trelease 1916).
Sister species P. galeottii and P. longifolium share the same hosts, have overlapping distribution ranges, attain a fairly large size, and have a pendulous habit. Their leaves are linear-(ob)lanceolate, although those of P. longifolium are longer (a mean of 52 rnm , compared with 29 mm in P. galeottii; Wiens 1964) However, they differ in time of anthesis (Fig. 12), P. galeottii sharing summer flowering (ca. July to September; Wiens 1964) with members of the conifer-parasite clade (except P. capitellatum) and with P. villosum, whereas P. longifolium flowers from December to February (Wiens 1964). Wiens (1964) placed P. galeottii in Pluriseriales Engler, a section including species with shorter internodes and smaller leaves than section Calyculatae. However, P. galeottii often has flattened and twisted stems with dilated nodes, characteristic of section Calyculatae (Wiens 1964). Because members of Pluriseriales sensu Wiens (P. galeottii and P. villosum subsp. coryae) each are sister species to a member of Calyculatae sensu Wiens (P. longifolium and P. scaberrimum, respectively), Pluriseriales and Calyculatae are polyphyletic. Pluriseriales sensu Wiens and Trelease is paraphyletic also because Paucifiorae is nested within it (Fig. 11). Trelease's section Calyculatae is monotypic and its sole representative, P. calyculatum, was not included in this study. Figure 12 shows a portion of the strict consensus tree, with geographic distribution, morphological characters, flowering period and host range (Wiens 1964;Kuijt 1996) indicated for each member of the northern clade. Several members of the clade occasionally exhibit morphological characters characteristic of species occurring farther south. Specifically, P. brachystachyum, P. rhipsalinum, P. robinsonii, and P. velutinum occasionally develop cataphylls and have median orientation of the lowermost foliar organs. 'Fertile cataphylls and indel y are shared by P. brachystachyum, P. rhipsalinum, and P. robinsonii. Indels g' and J are shared by P. brachystachyum, P. rhipsalinum, and P. brevifolium. This suggests that the initial branching event within the northern clade may have involved the divergence of the P. brachystachyum clade first, followed by the P. serotinum clade with its sister species P. robinsonii.

Phylogenetic Structure Within the Northern Clade
Flowering period suggests a switch from predominant winter (or fall) flowering in members of the P. brachystachyum and P. serotinum clades and in P. robinsonii, to summer flowering in the poorly supported clade (bootstrap value of 53%, decay value of 2; Fig.  10) that includes conifer and oak parasites. Exceptions are the winter-flowering P. capitellatum and P. longifolium. Calder6n de Rzedowski and Rzedowski (1972) report year-round flowering and fruiting for P. rhipsalinum, a common feature of species in Central and South America. Host specificity shows a correlation with flowering period, in that most of the winterflowering taxa form generalist host associations, whereas summer flowering tends to accompany narrow host specificity. Phoradendron bolleanum subsp. bolleanum is unusual by parasitizing not only Cupressaceae s.l. but also Ericaceae. Conversely, P. rhipsalinum is highly host-specific, despite its phylogenetic proximity to a host generalist.

CONCLUSIONS
The molecular sequence data presented in this study supports Kuijt's (1959Kuijt's ( , 1996 view that cataphyll presence/absence and the orientation of the lowermost foliar organs should not be used to circumscribe two major lineages in Phoradendron. However, a record of the variability for these two characters in some species (Kuijt 1996) has here been used to infer additional structure along some of the weakly supported branches of the molecular phylogeny. The role of temperature or day length in influencing cataphyll formation cannot be ruled out, in that acataphyllous species, including P. califomicum, are typically found at higher latitudes.
A latitudinal influence on host specificity and flowering period is likely, in that almost all tropical species are known to be host generalists with year-round flowering and fruiting. However, within the northern clade, host specificity seems to follow a phylogenetic pattern of distribution.
Interestingly, the host -specific associations among the conifer parasites are almost always associated with vegetative mimicry. Diagnosing leaf mimicry is subjective in that it is more readily recognized in mistletoes growing on hosts with distinctive (scalelike or needlelike or linear) leaf shapes (i.e., conifers) than on hosts having (ob)ovate leaves. Barlow and Wiens (1977) hypothesize that high host specificity is a condition for mimicry. The adaptive role of mimicry is speculative, although camouflage from extant of ancient (vertebrate) herbivores is likely. A preference for mistletoes has been documented for white-tailed deer feeding on a range of plants including Phoradendrori spp. (Gallina 1988), and for other vertebrate herbivores (e.g., Choate et al. 1987). Regardless of mimicry, high host specificity is thought to be an advantage in open habitats dominated by one or only a few host species, as is true of many conifer habitats (e.g., pinyon-juniper woodland).
Of the sections delimited by Trelease and Wiens, only Paucifiorae sensu Wiens is monophyletic. However, poor resolution at the base of the northern clade discourages the retention of sectional boundaries. Additional evidence confirming the paraphyly of P. 10mentosum, P. villosum, and P. bolleanum needs to precede nomenclatural adjustments.