Taxonomy. Linnaeus included two species of Caribbean Pectis (Pt. ciliaris L. and Pt. linifolia L.) when describing the genus in 1759. Pectis was included in the original circumscription of tribe Tageteae (Cassini, 1819) along with other former Heliantheae genera with leaves and phyllaries bearing glands. Based on molecular and morphological similarities, Panero (2007) included Tageteae in his Heliantheae Alliance, placing the traditional Tageteae genera, including Pectis and Porophyllum Guett., into subtribe Pectidinae.

The most extensive taxonomic treatments of Pectis are by Gray (1849, 1852, 1884, 1888), Fernald (1897), and Keil (1975, 1977a, 1978). Gray described six subgeneric divisions within Pectis based largely on differences in pappus elements (1849, 1852). By 1883 Gray recognized just three sections, Pt. sect. Eupectis, Pt. sect. Pectothrix, and Pt. sect. Pectidium. In 1897 Fernald elevated the sections to subgenera, and used pappus characters to assign 34 North American species to five subgenera: Eupectis, Heteropectis, Pectothrix, Pectidopsis (resurrected from Gray 1852) and Pectidium. Fernald’s treatment remains the most complete revision to date.

In the mid-70s, Keil revised four of the Pectis subgenera. For Pectis subg. Heteropectis (1975) and Pectidium ( =  Pt. sect. Pectis, Keil, 1978) he followed Fernald’s treatment of the species (two species in each section) but reduced the subgenera to sections. Although various workers used pappus characters to segregate Pectis into different genera (Lessing, 1830; de Candolle, 1836) or divide it into sections or subgenera (Gray, 1849, 1884; Fernald, 1897), these characters can be variable, even within populations. In 1977, Keil dismantled subgenera Pectothrix and Pectidopsis, using some species from each (as well as a few previously-unassigned taxa) to form a redefined Pt. sect. Pectothrix. Keil included species in Pt. sect. Pectothrix based on a combination of characters – position of foliar glands, shape of capitula, number of ray and disc florets, and corolla pubescence. In spite of these significant revisions, fewer than half of Pectis species have been assigned to sectional rank, and no treatment has covered the full geographic range of the genus.

Porophyllum is a genus of about 25 species of annual herbs and perennial shrubs (or subshrubs) found from the southwestern United States to southern Brazil, including the Caribbean islands. Some are arid-adapted but, unlike Pectis, many species of Porophyllum occur in mesic areas. Like Pectis, they have prominent oil glands on their leaves and phyllaries but in Porophyllum the scent is usually described as strong and unpleasant. The genus differs from others in the Tageteae by its combination of well-developed leaves, discoid heads, and a pappus entirely of bristles. In the most recent treatment of the genus, Johnson (1969) placed Porophyllum species into Pr. sect. Hunteria Moc. & Sessé (DC.) and Pr. sect. Porophyllum based primarily on leaf characters and habitat. Species of Pr. sect. Hunteria have thick leaves that are sessile to short-petioled, with narrow blades and are distributed in arid or semi-arid Mexico and the southwestern U.S. Species of Pr. sect. Porophyllum have thin, petioled leaves that are filiform to broad, and are distributed in South and Central America and in more mesic regions of North America.

Two molecular studies have included various Pectis and Porophyllum species. Baldwin et al. (2002) analyzed the helenioid Heliantheae using ITS data, and included one species each of Pectis and Porophyllum. Loockerman et al. (2003) used ITS and ndhF sequences to infer relationships within Tageteae and included six Pectis and four Porophyllum species. Both studies suggested a close relationship between Pectis and Porophyllum but Loockerman et al. (2003) found three Porophyllum species nested within Pectis, calling into question the monophyly of the two genera. However, both analyses had very small sample sizes and lacked strong support for their relationships. In the Loockerman et al. (2003) study, Porophyllum tridentatum was strongly supported as being sister to Leucactinia bracteata and Urbinella palmeri, and thus a new genus, Bajacalia, was erected for the three taxa.

Chromosome numbers in Pectis and Porophyllum. The base chromosome number for Pectis is x  =  12 (Keil, 1977). Of the 54 Pectis species and varieties for which chromosomes have been counted, 39 (72%) are diploid and 15 are polyploid (Appendix S2 in Supplemental Data with the online version of this article). Eight species are tetraploid, two of which (Pt. longipes and Pt. repens) have been reported as having diploid individuals as well. Six Pectis are hexaploid, among them Pt. saturejoides, which has also been reported as diploid. Pectis ericifolia is the only octoploid reported in the genus. Most Porophyllum species have a base number of x  =  12, but five species (Pr. lanceolatum, Pr. macrocephalum, Pr. punctatum, Pr. ruderale and Pr. viridiflorum) have been counted as x  =  11. Of these, Pr. punctatum and Pr. ruderale have also been reported with x  =  12 counts. Of the 14 Porophyllum species for which the ploidy level is known, nine are diploids, four are tetraploid and Pr. ruderale has been reported with diploid, triploid, and tetraploid counts. Porophyllum greggii is the only known hexaploid in the genus.

Photosynthetic pathway. C4 photosynthesis is a complex trait in which anatomical, chemical and regulatory modifications reorganize the first steps of carbon assimilation found in the C3 pathway. In C3 plants, the first enzyme involved in photosynthesis is Rubisco. In C4 photosynthesis, the first enzyme is PEP-C. Rubisco discriminates more against C13 than PEP-C does, and as a result, C4 plants have a higher proportion of C13 than C3 plants. This difference allows determination of the pathway used by a given plant by measuring the proportion of C13 in a tissue sample. Although C4 photosynthesis has arisen repeatedly and takes many different forms, there is a phylogenetic component to its distribution. The 65 C4 lineages occur in just 19 families (Sage et al., 2012). In the Asteraceae, one of the largest families of flowering plants, there are five C4 lineages that occur in just two tribes: Coreopsidae (in Chrysanthellum Pers., Glossogyne Cass., and Isostigma Less. of subtribe Chrysanthellinae) and Tageteae (in Flaveria Juss. of subtribe Flaveriinae and in Pectis of subtribe Pectidinae). The C4 genera in Chrysanthellinae form a monophyletic group and thus are thought to represent a single acquisition of the syndrome (Kimball & Crawford, 2004). However, based on phylogenetic analysis, C4 photosynthesis is thought to have multiple origins in Flaveria (McKown et al., 2005). Muhaidat et al. (2007) included Pectis glaucescens in their survey of C4 eudicots, and reported its chemical subtype as NADP-ME and its Kranz type as atriplicoid. This clustered phylogenetic distribution suggests an underlying predisposition toward C4 photosynthesis in certain clades. An evolutionary change in a common C3 ancestor might facilitate additional modifications down the line and could explain multiple origins of C4 among close relatives.

Just as there is a phylogenetic signal in the occurrence of C₄ photosynthesis, the origin of the syndrome is clustered in particular geographic areas as well. The majority of New World C₄ eudicot lineages arose in North America (Sage et al., 2011). The evolution of C₄ lineages has been linked to the Oligocene decline of atmospheric CO₂, but environmental factors such as heat, drought, and fire regime may have played a role (Osborne, 2011). Pollen records show that C₄ grasses began to dominate just 8-3 mya (Cerling, 1999) but only recently have molecular studies allowed an estimate of the timing of their origin. The rise of C₄ photosynthesis in monocots is estimated at 32.0–25.0 mya for Chloridoideae (Christin et al., 2008) and 10-20 mya for Cyperaceae (Besnard et al., 2009). Portulaca is the oldest known lineage of C₄ eudicots, having diverged ± 30 mya (Ocampo & Columbus, 2010; Christin et al., 2011).

Much of our knowledge of the evolution of C₄ photosynthesis is from Flaveria (Engelmann et al., 2003; Westhoff & Gowik, 2004; McKown et al., 2005; McKown & Dengler, 2007), which has both C₃ and C₄ species, as well as C₃-C₄ intermediates. The modifications in C₃-C₄ intermediates are thought to confer evolutionary benefits in their own right, and may leave C₃-C₄ intermediates in a more or less stable state of photosynthetic efficiency appropriate to their environment, with no momentum toward one or the other state. However, the development of fully functional C₄ photosynthesis involves as a series of modifications, from changes in leaf anatomy to up-regulation and cell-specificity of C₄ chemicals. C₃-C₄ intermediates may also represent snapshots of the process of evolution toward full C₄ photosynthesis, with each state a precursor to full C₄ functionality. With the benefit of a phylogeny to determine ancestral vs. derived states, one can trace anatomical and biochemical alterations that may predispose plants with intermediate traits to develop the full C₄ pathway. This analysis has been done in Flaveria (Engelmann et al., 2003; Westhoff & Gowik, 2004; McKown et al., 2005; McKown & Dengler, 2007) with the resulting acquisition path outlined in Sage (2003) and Gowik and Westhoff (2011).

By expanding the focus to other C₄ eudicots, we may see anatomical and physiological commonalities in the evolution of the C₄ syndrome. Such studies are continuing in Amaranthaceae (Kadereit et al., 2003; Sage et al., 2007), Cleome (Brown et al., 2005; Marshall et al., 2007), Heliotropium (Vogan et al., 2007), and Molluginaceae (Christin et al., 2010). Each lineage examined will offer insights into C₄ biology and evolution.

Taxon and marker selection. Our strategy was designed to sample as widely as possible both Pectis and Porophyllum, with the primary focus on Pectis. Roughly 230 names have been applied to various taxa in Pectis, ca. two thirds of which are considered taxonomic synonyms. Recent treatments and floras were utilized when deciding which species to include and which to treat as synonyms. For previously synonymized taxa that are wide-ranging, accessions from each region were included (when possible) to assess whether molecular data support their placement in synonymy. Since no single treatment covers Pectis throughout its range, Pectis species recognized by Aristeguieta (1964), Bautista (1987), Cabrera (1978), Jørgensen and León-Yánez (1999), Keil (1975, 1977a, 1978, 1996), and Liogier (1962, 1996, 2000) were sampled. Porophyllum species recognized by Johnson (1969) and Turner (1996) were also examined. Porophyllum tridentatum Benth., ( =  Bajacalia tridentatum (Benth.) Loockerman, B.L. Turner & R.K. Jansen was not included in our analysis of Porophyllum, as the molecular analysis of Loockerman et al. (2003) and cytological data (x  =  15) of Reveal & Moran (1977) suggested it is not closely related to the rest of Porophyllum.

Not all species were sampled, due either to rarity in the field, paucity of herbarium collections, or both. The final sampling included 78 species and varieties of Pectis, 22 of Porophyllum, and one species each of Chrysactinia A. Gray, Nicolletia A. Gray and Tagetes L. as outgroups. Thus, approximately 75% of both Pectis and Porophyllum species were included. Outgroups were chosen based on relationships indicated in Loockerman et al. 2003. Sampled taxa and voucher information with GenBank numbers are listed in Appendix 1.

For the molecular analyses we used loci that were 400-2,000 bases long and sufficiently variable to resolve relationships at the species level. The internal transcribed spacer (ITS) region of nuclear ribosomal DNA (nrDNA) was included because of its prior use in elucidation of sub-generic Asteraceae relationships (Baldwin, 1993; Baldwin et al., 2002) and ease of amplification from herbarium material. The chloroplast (CP) loci selected based on length and p-distance between species (Shaw et al., 2007; Timme et al., 2007; Hansen et al., 2009) were coding regions matK and 3’ ndhF, and the CP non-coding areas rpl16 intron, trnL-rpl32 spacer, 3’ trnV-ndhC spacer, 5’ trnY-rpoB spacer). Primer and locus information are summarized in Appendix S1 (see Supplemental Data with the online version of this article).

DNA isolation, PCR amplification, and sequencing. Total genomic DNA was extracted from ± 20 mg of dried leaves using Qiagen DNeasy Plant Mini Kits (Qiagen, Valencia, California). Dilution of the genomic DNA to 1:10 provided the best amplification of both nuclear and chloroplast loci. PCR methods followed Loockerman et al. (2003). PCR products were visualized under UV light in a 1.5% agarose gel containing SYBR safe DNA gel stain (Invitrogen, Carlsbad, CA, USA). Amplicons were cleaned by adding 4.0% Shrimp Alkaline Phosphatase and 1.0% Exonuclease I to the PCR tube and heating to 37° C for 30 min followed by 80° for 15 min (Werle et al., 1994), and sequenced using BigDye (v.3.1) Terminator Cycle Sequencing (Applied Biosystems, Foster City, Ca, USA) at the Institute for Cell and Molecular Biology Core Facility, The University of Texas at Austin.

Phylogenetic analysis. Individual sequences were trimmed and edited using Sequencher (Gene Codes CorPt., Ann Arbor, Michigan), and contigs aligned with MacClade 4.08 OSX (Maddison & Maddison, 2005). If a particular accession would not amplify for a certain locus, that region of the combined dataset was coded as missing data. All sequences are deposited in GenBank (http://www.ncbi.nlm.nih.gov/genbank/), with accession numbers listed in Appendix 1.

Because ITS sequences can present polymorphisms through hybridization, introgression, and incomplete lineage sorting (Alvarez & Wendel, 2003), PCR products of some accessions were cloned to test for variation between and within species. Since cloning all accessions was not feasible, only those samples for which direct sequencing showed evidence of polymorphisms (13 accessions) were cloned. Successful amplifications were visualized on an agarose gel and cloned using the TOPO-TA cloning kit (Invitrogen, Carlsbad, CA, USA). Ten colonies were chosen from each plate and amplified using the M13 plasmid primers provided in the cloning kit. All 5.8S motifs were screened to identify pseudogenes (Harpke & Peterson, 2008), and this resulted in the elimination of a total of 5 clones.

Clone copies from the same accession almost never appeared in separate well-supported clades. For accessions that had different clone copies, the clones were either monophyletic or appeared with other related species in polytomies (tree not shown). Because the support for such clades was so low, we did not believe any particular copy to be more representative of a species than any another copy, and so chose the clone from the first of ten colonies selected from each plate to include in the final analyses. An exception, Pectis multiflosculosa, had two clone copies in two distinct clades, so a copy of each type was included in the ITS and the combined CP+ITS analyses. The Pt. multiflosculosa CP sequence was duplicated and added to each ITS clone sequence.

Before further analysis, duplicate samples of taxa for which the sequences were identical across all loci were eliminated. Sequence length of each locus and percentage of missing and parsimony-informative (P.I.) characters are shown in Table 1.

Table 1. 

Statistics for datasets used including results from ML searches.

The ITS and the combined CP datasets were analyzed separately and together, using maximum likelihood (ML) with RAxML (Stamatakis, 2006) and Bayesian inference (BI) with MrBayes 3.1.1 (Huelsenbeck & Ronquist, 2001). All characters were weighted equally, character state transitions were treated as unordered, and gaps were treated as missing data.

RAxML allows for individual partitions of a dataset to be run with their own model of molecular evolution, including partitioning by codon position. The CP and CP+ITS datasets were partitioned as follows: chloroplast non-coding, chloroplast coding (by codon position), and ITS. The combined partitions were run together under the same GTR model, with parameters estimated separately for each partition. Each analysis was performed ten times from a random starting tree, with 500 bootstrap replicates. As the final likelihood values of each run were very similar (within 0.003% of final score), bootstrap replicates from each run were combined to estimate support for the tree with the best ML score. Clades with bootstrap values above 70% are considered well-supported (Hillis & Bull, 1993).

Prior to conducting the Bayesian analyses, the Akaike information criterion (AIC) was used via Modeltest 3.7 (Posada and Crandall, 1998) to determine the most appropriate model of DNA sequence evolution for each of the seven loci. The results were incorporated into the three analyses: ITS, CP, and CP+ITS combined. Bayesian analyses were performed using default priors, with two simultaneous runs with four Markov chains with heating values of 0.15, 0.2 (default), or 0.3, sampling every 100 generations. Each chain was run for at least 10 million generations and up to 20 million generations, depending on how long it took to reach stationarity (the average standard deviation of split frequencies between runs ≤0.01), and convergence was confirmed by using AWTY graphical analysis (Wilgenbusch et al., 2004). Burn-in trees (30%) were discarded, and the remaining trees and their parameters saved. The frequency of inferred relationships was used to represent estimated posterior probabilities (PP). Clades with PP ≥ 0.95 are considered strongly supported (Wilcox et al., 2002).

Hypothesis testing. Conflicting topologies between analyses of CP and ITS datasets led to testing the results of each analysis against the topology of the other. Hypotheses of alternate topologies were tested using the approximately unbiased (AU) test (Shimodaira, 2002) implemented in CONSEL (Shimodaira & Hasegawa, 2001), comparing constrained vs. best trees from 500 RAxML bootstrap replicates. Specifically, conflicts were addressed between the ITS and the CP datasets regarding the placement of three clades/taxa: Porophyllum amplexicaule + Pr. scoparium, Pectis linifolia + Pt. coulteri, and Pt. papposa var. papposa. This was done by analyzing the ITS dataset with RAxML under a constraint of the CP topology, and comparing the best topology under the constraint with the best unconstrained topology recovered from the ITS dataset. A reciprocal test was then conducted using the CP dataset, comparing the best topologies recovered from the CP dataset, both unconstrained and under the constraint of the ITS topology.

Molecular dating. To infer divergence times, data were combined with sequences from GenBank to create a dataset of matK and 3’ ndhF sequences. This included 21 sequences from the five major Pectis clades recovered from the phylogenetic analysis, as well as several Porophyllum and other species of tribe Tageteae. Data from six of the 12 major Asteraceae clades recovered in Panero and Funk (2008) were included, for a total of 46 taxa. The phylogeny was rooted on the branch leading to Barnadesia and Doniophyton (Barnadesioideae), which is well-supported as sister to the rest of Asteraceae (Jansen & Palmer 1987; Kim et al., 2005; Panero & Funk, 2008).

Modeltest 3.06 (Posada & Crandall, 1998) with the Akaike Information Criterion was used to determine the most appropriate model of sequence evolution for the dataset. The Likelihood Ratio Test (LRT) was used to determine whether the data satisfied the assumptions of a molecular clock, with the formula of LR  =  2*(lnL1-lnL2), where lnL1 is the likelihood of the tree with a molecular clock enforced, and lnL2 is the likelihood of the tree without the clock constraint, and degrees of freedom of n-2, where n is the number of taxa (Felsenstein, 1988). The LRT resulted in a significant difference between the trees, so a Bayesian relaxed uncorrelated lognormal molecular clock was used to account for rate heterogeneity across lineages.

Not all lineages evolve at the same rate, and when using molecular data to infer dates for lineages, it is best to use multiple fossils to constrain various nodes throughout the tree to lessen the margin of error associated with rate smoothing. Fossils provide only approximate dates, can be difficult to associate accurately with extant taxa, and are necessarily younger than the lineage they represent. Most paleobotanical evidence of the Asteraceae is from fossil pollen, but a macrofossil from Patagonia of 47.5 mya was allied with the Mutisioideae (Barreda et al., 2010). Although constraining a tree at multiple nodes is best, due to the uncertainty of the relationships between some Asteraceae subfamilies (Panero & Funk, 2008), only the node between the Barnadesioideae and the remainder of the Asteraceae was constrained.

The aligned data matrix was analyzed in BEAST 1.7.0 (Drummond & Rambaut, 2007) from an input file created in BEAUti 1.7.0 (packaged with BEAST). Two independent runs of 10,000,000 were conducted. Settings were as follows: a substitution model of GTR+G+I (based on Modeltest results), with base frequencies estimated; relaxed uncorrelated lognormal clock with rates estimated; a Yule process speciation tree prior with the starting tree randomly generated; prior distributions were set at default except for the ingroup. The node separating Barnadesioideae from the rest of the Asteraceae was calibrated based on the Mutisioideae fossil date of 47.5 mya. Assuming the Barreda et al. (2010) fossil date is a minimum age of the split, a lognormal prior distribution with a mean of 2.0, standard deviation of 0.5, and offset of 44.5 mya was applied. These settings provide a 5% probability of 47.75 mya, set the median probability at 51.89 MY, and a 95% probability of 61.32 mya for the most recent common ancestor (MRCA) of Barnadesioideae and the rest of the Asteraceae. Tracer v1.5 (Rambaut & Drummond, 2007) was used to assess convergence of the runs, confirm an appropriate estimated sample size, and determine the appropriate number of burn-in trees. Tree information from each run was combined with LogCombiner (packaged with BEAST), with the first 20% (2,000 trees) of each run discarded. A maximum clade credibility tree was constructed with TreeAnnotator (packaged with BEAST), and FigTree v1.3.1 (Rambaut, 2008) was used to visualize the estimations of divergence dates on the tree.

Carbon Isotope Ratios. Since plants differentially use carbon isotopes (C12 and C13) depending on whether they use C3 or C4 photosynthesis, the proportion of C13 in leaf tissue can be used as a proxy for the type of photosynthesis employed by that plant. Because rubisco discriminates against C13, C4 plants have a higher proportion of C13 than C3 plants. This proportion is expressed as delta (∂) in parts per thousand (‰). The metric ∂C13‰ is the difference between the tissue sample reading and a reference reading. The reference used for C13 is Pee Dee Belemnite (PDB), a Cretaceous marine fossil (Belemnitella americana, †Belemnitellidae) from the Pee Dee formation of South Carolina. PDB has a C13/C12 ratio that is higher than other natural samples. By convention, this standard is set to zero, so the amount of carbon measured in plants and animals is expressed as a negative number (Petersen & Fry, 1987). As C4 plants have more C13 than C3 plants, their ∂C13‰ signature (-15‰ to -10‰) is higher than that of C3 (-21‰ to -30‰) plants (Cerling, 1999; Marchese et al., 2005), so ∂C13‰ provides an indirect method for inferring C4 photosynthesis.

To infer the photosynthetic pathway for each sample, 2 mg of plant tissue (stem or leaves) were assayed for carbon isotope ratio using an Integramass spectrometer with a PDB standard. Carbon isotope ratios were determined by the University of California stable isotope facility ( http://stableisotopefacility.ucdavis.edu).

CP dataset. The final CP dataset comprising all CP loci contained 157 accessions. The models indicated by Modeltest were GTR+G+I for the rpl16 intron, GTR+G for the 5’ trnY-rpoB spacer, and TVM+G for the remaining four CP loci. As MrBayes does not allow for the TVM submodel, GTR+G was substituted for the Bayesian analysis. Prior to combining the CP loci, areas of ambiguous alignment were excluded from each CP locus, with a total of 1,321 bases excluded. The aligned dataset comprised 6,201 bp, and ten RaxML runs resulted in a best tree of -ln  =  18,348.32 (Fig. 1A). Table 1 provides a summary of loci and dataset statistics.

Fig.1.

Comparison of CP and ITS topologies. A: ML cladogram from CP dataset, ln = -18348.3266. B: ML cladogram from ITS dataset, ln. = -7478.8052. Branches in bold are well-supported (≥ 70 bootstrap or ≥ 0.95 PP). Colors refer to Pectis clades A-E, PH=Porophyllum sect. Hunteria, PP=Porophyllum sect. Porophyllum.

ITS dataset. The final ITS dataset was reduced to those accessions for which sequences for the chloroplast loci were obtained. Because of uncertainty of the alignment, 154 bases were excluded prior to analysis. The final aligned dataset comprised 156 accessions of 684 bp. Ten RaxML runs resulted in a best tree with a likelihood score of -ln  =  7,478.80 (Fig. 1B). Table 1 shows a summary of results statistics for the datasets.

CP+ITS dataset. An incongruence length difference (ILD) test indicated a significant conflict between the ITS and CP datasets (p  =  0.02). However, visual inspection of the majority rule trees from ML bootstrap and Bayesian analyses showed that most of the conflicts were not well supported. The ILD test is thought to be very conservative, falsely rejecting congruence (Cunningham, 1997; Darlu & Lecointre, 2002) showed that when the ILD test resulted in a p-value of greater than 0.01, combining the datasets improved or did not diminish phylogenetic accuracy. Therefore, the CP and ITS datasets were combined. The final CP+ITS dataset contained 157 accessions, with 6,885 aligned bp. Ten RaxML runs of this dataset resulted in a best tree of -ln  =  27,363.46 (Figs. 2-4).

Fig.2.

ML cladogram from CP+ITS analysis, showing species of Pectis, with clades A-D color-coded. ML bootstrap support is shown above branches, and Bayesian PP support is shown below. An asterisk indicates support below 50% bootstrap or 0.95 PP. Thick-lined branches lead to well-supported clades (≥ 70 bootstrap, ≥ 0.95 P). ln = −27363.4645. Phylogram of same analysis appears below the cladogram.

Phylogenetic analyses—Pectis. Analyses of the CP, ITS, and CP+ITS datasets provide strong support for the monophyly of Pectis. Figure 1 shows a comparison of the CP and ITS topologies, and the topology recovered from the combined CP+ITS dataset is shown in Figs. 2 and 3 (Pectis) and Fig. 4 (Porophyllum). All datasets recover five well-supported clades within Pectis (clades A-E, Fig. 1). The CP and ITS differ in their placement of either clade A (ITS) or clade B (CP) as sister to the rest of the genus. In all analyses, the remainder of the genus is split between two well-supported Sonoran and Chihuahuan Desert clades (C and D in Fig. 1) and a large clade (E) with species that occur from central Mexico to San Luis, Argentina.

Fig.3.

ML cladogram from CP+ITS analysis, showing species of Pectis clade E.

Fig.4.

ML cladogram from CP+ITS analysis, showing Porophyllum species, collection site, and sectional affiliations.

Phylogenetic analyses—Porophyllum. The inclusion of 22 of roughly 30 Porophyllum species allowed conclusions to be made regarding the relationships of Porophyllum to Pectis and within Porophyllum. The results suggest that Pr. amplexicaule and Pr. scoparium do not belong to Porophyllum (sensu Johnson, 1969). They form a strongly-supported clade that is the sister-group to a combined Pectis+Porophyllum clade. Porophyllum filiform and Pr. greggii form a well-supported clade sister to the rest of Porophyllum, which appears as a grade of clades (Fig. 4). Pr. sect. Hunteria and Pr. sect. Porophyllum are not recovered as monophyletic.

Hypothesis testing. The CP and ITS datasets provided incongruent results in three groups of note: Pectis imberbis + Pt. linifolia (clade A), Pt. papposa var. papposa, and Porophyllum amplexicaule + Pr. scoparium. Reciprocal AU tests were performed to assess whether, given a dataset, there was a significant difference in likelihoods between the best tree obtained from that dataset and the alternative topology (the constraint) obtained from the other dataset. The constraint topologies and results are shown in Fig. 5.

Fig.5.

Reciprocal tests of alternate topologies. Figures on the left (1A, 1B, 1C) show the topologies recovered with the CP dataset, and figures on the right (2A, 2B, 2C) show the topologies recovered with the ITS dataset. *p-values of ≥ 0.05 indicate that the topology shown has a significantly lower likelihood than the best tree recovered using that dataset.

The ITS topology (Fig. 1B) shows clade A (Pectis imberbis+Pt. linifolia) as a sister group to the rest of the genus, followed by clade B (Pt. coulteri + Pt. multiseta). These relationships are swapped in the CP dataset, which recovers clade B sister to the rest of Pectis, and clade A appearing next in the grade (Fig. 1A). The latter relationship is not strongly supported (59% ML bootstrap and 0.67 Bayesian PP support), and the AU test shows that, when using the CP dataset, the best CP topology is not significantly better than the best ITS topology. The CP+ITS dataset strongly supports the position of clade A as sister to the rest of the genus.

Pectis papposa var. papposa groups with Pt. vollmeri in the topology suggested by the CP dataset (Clade C of Fig. 1A). However, the ITS topology (Fig. 1B) shows Pt. papposa var. papposa grouped in Clade D with Pt. filipes var. subnuda and Pt. barberi. Given the ITS dataset, the CP topology can be rejected; likewise, given the CP dataset, the ITS topology can be rejected. The CP+ITS dataset strongly supports the position of Pt. papposa var. papposa in clade C with Pt. vollmeri.

The CP dataset recovered a clade of Porophyllum amplexicaule+Pr. scoparium at the base of Pectis+Porophyllum (Fig. 1A), whereas the ITS dataset showed this clade to be sister to the rest of Porophyllum (Fig. 1B). Given the ITS dataset, the CP topology cannot be rejected, but given the CP dataset, the ITS topology can be rejected. In the combined CP+ITS dataset, Pr. amplexicaule and Pr. scoparium are well-supported as sister to the combined Pectis+Porophyllum.

Phylogenetic distribution of polyploids. In Pectis, most of the known polyploids occur in the more terminal clades (Appendix S3, see Supplemental Data with the online version of this article). Clade E4 (Fig. 3) contains five polyploid species, clade E1 has three, and clades D and E3 both have one. Two of the four taxa in clade E2 (tetraploid Pt. repens and hexaploid Pt. saturejoides) have reports of both diploid and polyploid counts. In two cases (Pt. latisquama and Pt. multiflosculosa), the polyploid species appears sister to a diploid species, but most polyploids in the genus are sister to taxa for which the chromosome number is unknown. The four known polyploid Porophyllum species included in this study occur in separate clades throughout the genus. The combined CP+ITS dataset places the hexaploid Pr. greggii with Pr. filiforme (chromosome number unknown), which together are sister to the rest of Porophyllum.

One internal Porophyllum clade comprises all x  =  11 taxa, together with one x  =  12 (Pr. coloratum) and several for which the chromosome numbers are not known.

Molecular dating. The dataset of matK (1,902 bp) and 3’ ndhF (603 bp) sequences comprised 2,505 characters and 46 taxa. The maximum clade credibility tree recovered most clades with greater than 0.95 PP (Fig. 6). The BEAST analysis showed Tageteae and Helianthus diverging 26.55 (19.23–33.73) mya. The MRCA of Tageteae is 24.34 mya, but this node has just 0.92 PP support. Porophyllum amplexicaule diverged from Pectis + Porophyllum at 15.92 MYA (11.03–21.53). The divergence of Pectis and Porophyllum s.s. is estimated at 11.27 (7.56–15.58) mya, and divergence within Pectis began 9.12 (5.75–12.61) mya but most of the nodes split within the last 5 my.

Fig.6.

Chronogram of the maximum clade credibility tree estimated from matK+3’ndhF sequences using BEAST. Horizontal bars are the 95% highest probability density (HPD) for the age of that node (green bars are on Porophyllum nodes, yellow bars are on Pectis nodes). Nodes with HPD bars have PP support of ≥ 0.95. Nodes without HPD bars have <0.95 PP. Mean ages are shown for clades mentioned in the text. A partial timescale is shown at the bottom, with units in millions of years. Epoch dates follow The Geological Society of America (GSA, 2009).

Carbon isotope analysis. We obtained ∂13C‰ values for 80 Pectis and Porophyllum species, as well as Tagetes erecta. All Pectis accessions have ∂13C‰ values consistent with C4 photosynthesis (-15.60‰ to -10.70‰, mean  =  -13.14‰), and all Porophyllum accessions have values consistent with C3 photosynthesis (-30.65‰ to -22.90‰, mean  =  -27.64‰). Tagetes erecta has a ∂13C value of -30.37‰. The frequency distribution of the ∂13C values is presented in Fig. 7, and the average ∂13C value for each species, including previously reported data from species not surveyed by us, is shown in Table 2. Voucher information is given in Appendix 1.

Fig.7.

The frequency distribution of carbon isotope ratios for the species listed in Table 2. All Pectis species sampled have ratios consistent with C₄ plants, all Porophyllum species sampled have ratios consistent with C³ plants.

Table 2. 

Photosynthetic pathways in Pectis and Porophyllum. Data are the average ∂¹³C value for each species (sample size if N>1). Average ∂¹³C value for Pectis  =  -13.14‰; Average ∂¹³C value for Porophyllum  =  -27.64‰. All values are newly reported here unless indicated with an asterisk (*). See Appendix 1 for voucher information.

Infrageneric relationships within Pectis. Six subgeneric divisions have been recognized in Pectis, variously treated as subgenera or as sections. Figure 8 shows the CP+ITS cladogram with these divisions mapped onto the tree according to the revisions of Gray (1852), Fernald (1897) and Keil (1975, 1977a, 1978).

Fig.8.

Taxonomic classifications in Pectis. Cladogram from Fig. 2 with classifications mapped onto the tree. Taxa with two colors were placed into two separate categories under different names but are now considered synonyms. Column 1 shows the genera of Lorentea, Pectidium, Pectidopsis and Pectis at the time that Gray described the former three as sections of Pectis. Thick-lined branches lead to well-supported clades (≥ 70 bootstrap, ≥0.95 PP).

Gray’s (1852) Pectis sect. Eupectis was proposed to include species having uniseriate, paleate, or broad-based and chaffy awns. Aside from Pt. prostrata, which he described in the same publication, Gray did not detail which species were to be included in his section Eupectis, and the species in column 1 of Fig. 8 are the species of Pectis recognized by de Candolle at the time. In the same 1852 treatment, Gray transferred Pectidium punctatum Less. into Pectis, placing it under Pt. sect. Pectidium. Pectis punctatum is synonymous with Pt. linifolia, which was, at the time, in Pt. sect. Eupectis. After the designation of Pt. linifolia (of former Pt. sect. Pectidium) as the type of Pectis (Britton & Millspaugh, 1920), Pt. sect. Pectidium de facto became Pt. sect. Pectis. The species variously recognized within section Eupectis by Fernald or Gray have not been treated since, and do not form a monophyletic group in any of our analyses.

Pectis sect. Heteropectis, comprising Pectis coulteri and Pt. multiseta (sensu Gray 1852, Fernald, and Keil), is recovered as a well-supported clade (Clade B of Figs. 2, 8). Members of this section (Figs. 2, 8 clade B) are restricted to the Sonoran Desert of Baja California and mainland Mexico.

Pectis sect. Pectis ( =  Pt. sect Pectidium Less. sensu Fernald, and later, Keil), is well supported and monophyletic (Clade A Figs. 2, 8). Pectis sect. Pectis has two species, Pt. imberbis and Pt. linifolia (the type of Pectis). Both are tall, erect plants with sparse, cylindrical leaves and elongated internodes. This section (Figs. 2, 8 clade A) includes Pt. imberbis, a rare species of southern Arizona and northern Sonora region, and Pt. linifolia. Pectis linifolia is divided into two varieties. Pectis linifolia var. hirtella is a narrow endemic of Guerrero and Michoacan, Mexico, and Pt. linifolia var. linifolia is a weedy species widely distributed throughout Pectis’ range.

Pectis sect. Lorentea has a complicated history. In 1797, the genus Lorentea was described by Ortega, and referred to a plant that was later identified as a member of the genus Sanvitalia. In 1816, Lagasca published the name Lorentea for species of Pectis that he considered separate from Pectis. The type of Lorentea Lag. was based on a specimen of Pectis prostrata Cav. from Cuba (this specimen was later identified as Pt. humifusa Sw.). In 1830, Lessing segregated a group of Pectis species into a new genus, for which he also used the name Lorentea. Lorentea Lag. and Lorentea Less. were both superfluous, but in practice, both names were used. Gray (1852) described section Lorentea A. Gray for species of Pectis with a biseriate pappus and a ray pappus sometimes greatly diminished or absent. He did not specify which species would be included. Schultz Bipontinus (Seemann et al., 1852) followed Gray’s suggestion that Lorentea sensu Lessing was best included with Pectis, and transferred the Lorentea of de Candolle (1836) and Gardner (1846) into Pectis (Seemann et al., 1852). Although Keil (1977b) mentioned several species as members of Pt. sect. Lorentea, to our knowledge no one has treated the section with a listing of the species to be included. Those noted as “Lorentea” in Fig. 8 are the Lorentea species known at the time that Gray described Pt. section Lorentea.

The sole member of Gray’s original Pectis subg. Pectidopsis was Pt. angustifolia (Gray, 1849). Gray later added Pt. filipes and Pt. uniaristata. Fernald followed Gray’s definition of Pectidopsis as those Pectis species with a pappus that is coroniform or has a few slender but rigid, scabrid awns. Fernald (1897) expanded the subgenus to include 12 taxa, which do not form a monophyletic group in the combined CP+ITS analyses (Appendix S4).

Fernald followed Gray’s 1884 expanded definition of Pectis sect. Pectothrix as a taxon with a pappus (of the disc florets, if not the ray florets) of many equal or unequal bristles, which are often broad at the base, but not true scales. Neither Gray’s nor Fernald’s concept of section Pectothrix forms a monophyletic group in our analyses. On the contrary, their species appear scattered throughout the tree. However, all members of clade C correspond to Pt. sect. Pectothrix sensu Keil (1977a), with additional members of Keil’s Pt. sect. Pectothrix found in clade D (Figs. 2, 8). Pappus morphology therefore seems to be homoplastic in Pectis.

Most species of clades C and D (Figs. 2, 8) occur in or adjacent to the Sonoran and Chihuahuan Deserts, and clade E contains many species of the Pacific Slope of Mexico, as well as those of South America and the Caribbean.

Sectional relationships within Porophyllum. Results from our analyses suggest that the primary characters used to designate the sections of Johnson (1969), i.e., leaf morphology and habitat, do not define clades in Porophyllum. The species relationships show that the genus does not consist of two clades, which correspond to the two sections (Fig. 1). Geography correlates better with relationship, with Sonoran and Chihuahuan species at the base of the tree and southern Mexico-Central American and South American species forming the derived clades. Of the roughly eight South American species, four out of the five sampled form a well-supported clade. The fifth is Pr. ruderale, a variable species of tropical North and South America that is sister to the Honduran accession of Pr. macrocephalum, nested within the Pr. macrocephalum clade. Johnson (1969) subsumed over 20 described species and varieties into the single, wide-spread and variable Pt. ruderale, with two subspecies, Pr. ruderale subsp. macrocephalum and Pr. ruderale subsp. ruderale. He used Pr. ruderale subsp. macrocephalum to refer to what he called the “northern” taxa (SW U.S. to northern Brazil, southern Peru, and Bolivia), and Pr. ruderale subsp. ruderale to refer to the “southern” taxa (Costa Rica and the West Indies south to southern Peru, through Brazil into northern Argentina). He noted that in northern South America, intermediate forms were common where the two taxa were sympatric. Porophyllum leiocarpum, endemic to Puerto Rico, was originally described by Urban as a variety of Pr. macrocephalum but was elevated to specific status by Rydberg and has been treated as such by subsequent authors (Rydberg, 1916; Johnson, 1969; McVaugh, 1984). In our analyses, we have followed Turner’s (1996) morphological criteria in designating the Mexican and Central American taxa as Pr. macrocephalum, and the South American accession as Pr. ruderale. Although intermediate forms surely exist, Turner noted that Pr. macrocephalum has large heads on more stout peduncles and is diploid, whereas Pr. ruderale has small heads on slender peduncles, and is tetraploid. Ecuadorian and Brazilian Pr. ruderale have been reported as n  =  22, 23, 24, 34, 35 and 36 (Turner et al., 1979; Robinson et al., 1981; Carr et al., 1999) but Dillon et al. (1982) reported n  =  12 for a Peruvian accession. Therefore, although only diploid (n  =  11) specimens of Pr. macrocephalum have been found in Chiapas (Strother, 1983) and Arizona (Keil & Pinkava, 1976), Pr. ruderale of South America has diploid and tetraploid (and possibly hexaploid) members. The North and Central American accessions (Pr. macrocephalum) have distinctly more ovate-oblong leaves, whereas the South American accessions (Pr. ruderale) have leaves that are linear-lanceolate. The accessions of Pr. macrocephalum, Pr. ruderale and Pr. leiocarpum form a well-supported clade, with the Central American accession of Pr. macrocephalum sister to the Pr. ruderale from Ecuador (Fig 4). Porophyllum ruderale from Brazil appears in a clade with Pr. angustissimum, Pr. lanceolatum, and Pr. linifolium. Hind (2002) has noted that Porophyllum of Brazil and Argentina may be under collected, and that there could be diversity in the genus that has, in these areas, been overlooked.

Polyploidy in Pectis. Roughly 40% of the known polyploids in Pectis occur on islands. Just nine of the ±20 Pectis species that grow in the Caribbean or the Galapagos Islands have been examined in cytological studies, and of these, three are diploids, and six are polyploids. Thus while polyploids comprise 18% of the total Pectis species, they comprise 77% of the island species (Appendix S2). This pattern is also seen in Hawaii, where 80% of the native plants are polyploid (Carr, 1988). While these ratios could change if more Pectis species were sampled, there is no reason to think that the current sampling of cytological data is skewed toward diploid or polyploid taxa.

Two mechanisms are often proposed to explain why islands may be rich in polyploid species. The first is that there is a general trend toward self-fertilization in polyploids (Barringer, 2007) and autogamy is often proposed as one of the catalysts for a widespread distribution or successful colonization (Baker, 1955). However, the trend is not clear for the Asteraceae, which has sporophytic self-incompatibility (SSI). The breeding systems of most Pectis species are unknown, but Keil (1978) reported that Pt. cylindrica, a tetraploid, and Pt. prostrata, a diploid, are autogamous. A second explanation for the high occurrence of polyploidy on islands is that successful establishment favors plants with high genetic diversity (Carr, 1988) sometimes associated with polyploidy. Finally, many species in the Asteraceae have SSI systems that allow occasional self-fertilization. The ability to self-pollinate increases the odds of reproducing after a colonizing event.

Our results showed that all species of Pectis appear to use C₄ photosynthesis (see below). Genome duplication is one of the preconditioning events proposed for a transition from C₃ to C₄ photosynthesis (Monson, 2003). Although 28% of the known species of Pectis are polyploid, these species are not at the base of the tree (Appendix S3). The four known independent origins of C₄ photosynthesis in the Asteraceae all occur within the Helianthoideae supertribe, in a group called the phytomelanic cypsela clade. Barker et al. (2008) found evidence for at least two paleopolyploidy events in the history of the Asteraceae—one at the base of the family, and another in a lineage that eventually gave rise to Helianthus. They also found that genes related to cellular organization are overrepresented in the paleologs of the Asteraceae. Although these genome duplications may have allowed for novel functions that eventually gave rise to C₄ photosynthesis in the this group, the scarcity of C₄ lineages in the Asteraceae shows that many other conditions must be required.

Origin and extent of C₄ photosynthesis. After Smith and Turner (1975) surveyed 20 Pectis and one Porophyllum species, it was suggested that all Pectis species were C4, and all Porophyllum (and, in fact, the remainder of the Tageteae) were C3. However, the closely-related Flaveria has just 21 species yet shows great variation in photosynthetic pathway, suggesting that similar variability might exist in Porophyllum (∼25spp.) or Pectis (∼90 spp.). The ∂13C‰ values reported here appear to confirm that C4 photosynthesis does not occur in Porophyllum. Furthermore, the switch from C3 to C4 appears to have happened after the generic split between Pectis and Porophyllum, as all Pectis have ∂13C‰ values indicative of C4 photosynthesis.

North America is one of the hotspots of origin for C₄ photosynthesis (Sage et al., 2011), and the North American C₄ lineages for which divergence times have been estimated have appeared since the mid-late Miocene: 13 mya for Tidestromia, 6.1 mya for Allionia, 4.7 mya for Boerhavia, and 3.1 mya for Flaveria (Christin et al. 2011). The anatomical preconditioning that may facilitate C₄ photosynthesis must be selected for, and in general, warm and dry environmental conditions are thought to make such adaptations more advantageous. Did Pectis evolve in such an environment? Keil (1978) suggested that the ancestors of Pt. sect. Pectis diverged in the Mexican Highlands. Porophyllum amplexicaule and Pr. scoparium, sister to the rest of Pectis+Porophyllum, are Chihuahuan desert species. The basal species within Pectis are distributed mostly in north and central Mexico from sea level to 800 m; Pt. sect. Pectis (Pt. imberbis and Pt. linifolia, clade A of Fig. 2) is sister to the rest of Pectis. Pectis imberbis is endemic to the Sonoran and Chihuahuan desert areas at the U.S./Mexico border, and Pt. linifolia var. hirtella is endemic to the Mexican states of Michoacan and Guerrero. After Pt. sect. Pectis, the next diverging clade (Pt. sect. Heteropectis) is endemic to the Sonoran Desert. Fossil evidence shows that the general drying trend since the late Miocene led to a flora of increasing tolerance to aridity, with an altitudinal fluctuation during the pluvial stages of the Pleistocene (Axelrod, 1979; Spaulding et al., 1983). The uplift of the Sierra Madre Occidental and Transvolcanic Belt further increased aridity by providing rain shadows. Becerra (2005) suggested these ranges provided a barrier to the northern cold fronts, thereby allowing the development of cold-intolerant taxa 10-20 mya. Although mesic woodlands existed in the present-day desert regions during the pluvial periods of the Pleistocene, evidence from pack-rat middens suggests that pockets of arid refugia persisted throughout these periods (Elias et al., 1995; Van Devender, 2000).

Our BEAST analysis recovers a mean date of ∼11 mya for the common ancestor of Pectis and Porophyllum (Fig. 6), and given the geographic distributions of the basal species of the genera, we can surmise that they diverged in Central/Northern Mexico. Thus, C₄ photosynthesis in this lineage evolved in an area that was increasingly warm and dry, with a pattern of summer monsoon. Sage and colleagues (2011) suggested that these are the environmental conditions that would increase photorespiration in C₃ lineages, setting the stage for a fitness advantage to C₃-C₄ intermediacy.