The phylogenetic position of Carex section Phyllostachys and its implications for phylogeny and subgeneric circumscription in Carex (Cyperaceae) ^†

Abstract

Wide speculation surrounds the origin and phylogenetic relationships of the most highly reduced sections in the genus Carex. In order to gain a better understanding of phylogeny in Carex, the relationship of the reduced sect. Phyllostachys to 12 putatively related sections, representing all four subgenera (Primocarex, Indocarex, Carex, Vignea), was inferred from sequences of the ITS (internal transcribed spacer) region of nrDNA. Phylogenetic reconstructions identified two main clades: (1) a “compound” clade composed of sections from subg. Indocarex, Primocarex, and a portion of subg. Carex, and (2) a “reduced” clade consisting of sections from subg. Carex (Phyllostachys) and Primocarex (Filifoliae and Firmiculmes). Subgenus Indocarex was paraphyletic within the “compound” clade supporting classifications that have merged it within a wider subg. Indocarex/Carex/Primocarex line. Subgenus Primocarex was polyphyletic. This result was consistent with theories that extreme reduction has occurred along several different evolutionary lines in Carex. Phylogenetic theories inferred from the presence or abnormal growth of the rachilla were not supported by tree topologies. Difficult sectional circumscriptions, such as the separation of sections Laxiflorae and Careyanae, were strongly upheld by sequence data. The ITS region is an effective tool for defining sectional limits and for estimating relationships among sections in Carex, but does not provide enough phylogenetic information to fully resolve relationships below the sectional level.

The genus Carex L. is one of the largest (∼2000 spp.) and most widespread of all flowering plant genera (Reznicek, 1990). Although cosmopolitan, most of its species are north temperate in distribution with centers of diversity in North America and eastern Asia (Ball, 1990; Reznicek, 1990; Naczi, 1992). The habitats of Carex species are as diverse as the genus and range from dry open savannas and rain forests, to wet meadows, deciduous and coniferous forests, and Arctic tundra. The genus and its tribe Cariceae Kunth ex Dumort. are clearly distinguished within the Cyperaceae Juss. by the possession of consistently unisexual flowers and a perigynium, which is a sac-like structure of prophyllar origin that surrounds the naked gynoecium (Blaser, 1944). The perigynium provides the primary source of taxonomic characters in Carex, and it is largely through its subtle differences in shape, size, texture, and nervation that the enormous number of species in Carex are distinguished (Nelmes, 1951).

Despite the global distribution and ecological importance of this genus, evolutionary relationships within Carex are poorly understood (Reznicek, 1990). This lack of understanding can be attributed to the nature of morphological and anatomical characters in Carex. Floral reduction (Smith and Faulkner, 1976), uniform vegetative morphology and anatomy (Metcalfe, 1971), and repeated events of parallelism and reversal (Reznicek, 1990; Naczi, 1992; Starr, 1997) have obscured phylogenetic trends and have led to the recognition of many artificial taxa at the sectional and subgeneric level. The extreme reduction in the genus and the almost complete lack of knowledge of its ontogeny (Alexeev, 1988), have further raised homology problems that complicate phylogenetic reconstruction at all levels.

DNA characters offer an obvious source of variation whose use is just beginning to be explored in Carex. The advantages afforded by these characters in terms of character number, homology assessment, reduced homoplasy, and ease of scoring character states suggest that DNA characters could make a substantial contribution to the systematics of this genus.

In this study we investigate the usefulness of DNA characters in the systematics of Carex by using sequences from the internal transcribed spacer (ITS) regions of nuclear ribosomal DNA (nrDNA). These loci have become the sequences of choice for addressing phylogenetic questions within genera because they are easily aligned and amplified, they have a high mutation rate, and they are relatively small (for a review see Baldwin et al., 1995).

Because of the size of the genus, we chose to focus our study on Carex section Phyllostachys (J. Carey) L.H. Bailey. Several arguments favor this approach. First, the small size of the section (eight species) allows for both interspecific and intraspecific variation to be assessed. Second, its clear delimitation from other sections and the recent resolution of a number of its taxonomic problems (Catling, Reznicek, and Crins, 1993; Naczi, Reznicek, and Ford, 1998; Ford et al., 1998) suggest that the interpretation of results should be not confounded by a poorly resolved taxonomy. Third, the evolutionarily reduced and unconventional inflorescence morphology of the Phyllostachys has led many authors to consider the section as pivotal to our understanding of subgeneric limits and evolution in the genus. Thus a clarification of the phylogenetic position of this section will allow us to address these wider systematic issues in Carex.

The specific objectives of this study were therefore to: (1) assess the utility of the ITS region for resolving both phylogenetic and taxonomic questions in Carex; (2) determine whether Carex section Phyllostachys is monophyletic; and (3) determine the systematic position of this section within the genus Carex by comparing sequence data of section Phyllostachys with species representative of putatively related sections.

SYSTEMATIC BACKGROUND

Subgeneric classification

In the only worldwide monograph of the genus, Ku¨kenthal (1909) divided Carex into four subgenera: (1) Primocarex Ku¨k., characterized by solitary terminal spikes; (2) Vignea (P. Beauv. ex Lestib. f.) Peterm., distinguished by sessile, bisexual inflorescence units, the lack of a cladoprophyll (a tubular structure subtending an inflorescence), and distigmatic flowers; (3) Indocarex Baillon, a mainly tropical group with paniculate inflorescences, bisexual inflorescence units, cladoprophylls, inflorescence prophylls (perigynium-like organs subtending lateral inflorescence units), and tristigmatic flowers; and (4) Carex, characterized by mostly peduncled, unisexual inflorescence units, cladoprophylls, and usually tristigmatic flowers.

Ku¨kenthal (1909) considered subg. Primocarex primitive within Carex because its spikelets typically possessed a conspicuous secondary axis (rachilla): a trait common to the other genera in the Cariceae, but normally lacking in subg. Indocarex, Carex, and Vignea. Most subsequent authors (Reznicek, 1990, and papers cited therein) have criticized this arrangement, believing that the species in subg. Primocarex had evolved independently from multispicate species in the other subgenera or from closely related genera in the Cariceae. Based on these conclusions, most recent authors recognize three subgenera (Carex, Indocarex, and Vignea) with members of Primocarex divided between subg. Vignea or Carex or placed in closely related genera (Reznicek, 1990). Contrasting opinions, however, have been expressed by researchers such as Ohwi (1936) and Koyama (1962) who recognized only two subgenera, Carex and Vignea, based on the observation that some species in sections Hymenochlaenae (Drejer) L.H. Bailey and Frigidae Christ subsect. Decorae Ku¨k. are transitional in their morphology between subg. Carex and Indocarex.

Carex sect. Phyllostachys and its position within the genus Carex

Carex section Phyllostachys is a clearly defined group of eight species (C. backii Boott; C. juniperorum Catling, Reznicek, & Crins; C. latebracteata Waterfall; C. jamesii Schwein., C. saximontana Mack.; C. willdenowii Willd.; C. basiantha Steud.; and C. superata Naczi, Reznicek, & B. A. Ford) restricted to North America north of Mexico. The section is easily distinguished by features such as large foliaceous pistillate scales and culms that are dilated at the apex (Crins, 1990; Catling, Reznicek, and Crins, 1993).

The strongly marked nature of the Phyllostachys and the lack of an obvious outgroup have led to wide speculation regarding its position within the genus. Bailey (1885, 1886) postulated that Phyllostachys was composed of two subsections: (1) Bractoideae L.H. Bailey, which included three of the species (C. backii, C. jamesii, and C. willdenowii) presently recognized in section Phyllostachys; and (2) Phyllostachyae L.H. Bailey, composed of species currently placed in section Firmiculmes Ku¨k. (C. multicaulis L.H. Bailey, C. geyeri Boott). Bailey postulated that Phyllostachys was an offshoot of sect. Acrocystis Dumort. and that the section was connected to sect. Laxiflorae Kunth via C. multicaulis. Features such as dilated culm apices in many species of the Laxiflorae and its nearest allies (Mackenzie, 1935) lend some credibility to Bailey's hypothesis.

Holm (1900, 1903) argued that sect. Phyllostachys was not closely related to sections Acrocystis, Laxiflorae, or Firmiculmes, but was instead an extreme form of sect. Hymenochlaenae that demonstrated transitional characters between subg. Carex and Indocarex. His hypothesis was partly based on the fact that sect. Phyllostachys emulates the indocaricoid condition when the abnormal growth of the rachilla within basal perigynia leads to the production of lateral androgynous inflorescence units.

Kreczetovicz (1936) expressed views similar to Holm's but believed that sect. Phyllostachys was a highly derived group that had evolved only recently (Pliocene or Pleistocene) from subg. Indocarex via transmutive reduction. He defined transmutive reduction as reduction where fusion and/or a transformation of parts (e.g., of a bract into a scale) obscured a recent origin for species from subg. Indocarex. He cited the leafy pistillate scales and dilated culms of the Phyllostachys as clear evidence that such a reduction had taken place.

Based on smut host-parasite studies, Savile and Calder (1953) proposed that section Phyllostachys was a highly derived section within a lineage that included the multispicate section Limosae (Tuckerm.) Christ and the unispicate section Scirpinae Tuckerman. Although they conceded that subgenus Primocarex sensu Ku¨kenthal was artificial, their studies of smut infections suggested that nearly half its species comprised a natural group. Other reduced taxa, like sections Phyllostachys and Scirpinae, were considered highly evolved because their smuts appeared to be related to smuts that infected multispicate carices typical of advanced subgenera. In addition, species in these sections seemed to lack a rachilla, a feature present in “true” primocarices.

Although authors rarely discuss the intricacies of their evolutionary system, the arrangement of taxa within their monographs presents an implicit phylogenetic hypothesis. The position of sect. Phyllostachys in the monographs of Mackenzie (1935) and Ku¨kenthal (1909) is worthy of mention owing to the great influence of these works on the present systematics of the genus. Mackenzie (1935) did not divide his 71 North American sections into subgenera but arranged them in a system where sectional affinities were vaguely expressed through a conspectus. Although the first five sections, referable to subg. Primocarex, were thought to be “primitive,” other solitary-spiked species were regarded as “closely related to more developed sections” and aligned with representatives of the other subgenera. Section Phyllostachys was positioned between the unispicate sections Polytrichoideae (Tuckerm.) Mack. and Filifoliae (Tuckerm.) Mack. at the beginning of a large group of sections variously placed in subg. Indocarex, Carex, and Primocarex. Mackenzie (1935) did not directly align the Phyllostachys or Polytrichoideae with any other section, but grouped the Filifoliae with sections Obtusatae (Tuckerm.) Mack., Praecoces Christ, and Acrocystis. Ku¨kenthal (1909) placed sect. Phyllostachys between two Old World sections, the Rhomboidales Ku¨k. and the Elatae Ku¨kenthal. Species in these sections are generally characterized by many-flowered, multispicate inflorescences, sheathing inflorescence bracts, and many-nerved perigynia. These characters are either absent or weakly developed in sect. Phyllostachys.

MATERIALS AND METHODS

Choice of taxa

Specimen citations for all individuals sequenced in this study and the nomenclature of the taxa employed are given in Tables 1 and 2. Taxa were chosen as representatives of specific groups or putative lineages to which Carex section Phyllostachys has previously been associated (see Systematic Background section). Some of the taxa placed close to sect. Phyllostachys by Mackenzie (1935) and Ku¨kenthal (1909) could not be included in this study due to the unavailability of material or to problems encountered in amplification and sequencing.

Choice of outgroup

Efforts to obtain complete sequences for the ITS 1 and ITS 2 regions of a member of the closely related genus Kobresia Wahlenb. were unsuccessful. As a result, the two species sequenced from subgenus Vignea were chosen as the outgroup. This decision was based on the general agreement that this subgenus is natural (e.g., Nelmes, 1951; Koyama, 1962; Reznicek, 1990; etc.) and not closely related to sect. Phyllostachys. Unrooted searches supported the assumption of monophyly for the ingroup, i.e., a single branch partitioned the ingroup from the outgroup (Swofford et al., 1996).

DNA isolation and PCR amplification

Total genomic DNA was extracted from live or dried leaf tissue of a single individual (Table 1) following a modified CTAB method (Doyle and Doyle, 1987) with 1.0% beta-mercaptoethanol in the extraction buffer. To assess the amount of infraspecific variation that might be expected for the ITS region in Carex, two to four individuals from across the range of each of the eight species in section Phyllostachys were sequenced. In two cases, individuals from single populations (C. saximontana, Ford 9501 & Starr; C. superata, Naczi 4013 et al.; Table 1) were sequenced to determine whether any variation might exist at the population level.

PCR amplification and sequencing of ITS

Double-stranded DNA for the complete ITS region (3′ 18S to 5′ 26S fragment) was PCR (polymerase chain reaction) amplified from total genomic DNA using the forward primer ITS-L (Hsiao et al., 1994), and the reverse primer ITS-4 (White et al., 1990). As in grasses (Hsiao et al., 1994), ITS-L provided better amplification than the forward primer ITS-5 (White et al., 1990). Each 100-μL reaction mixture contained 10 μL of 10X Taq reaction buffer, 6 μL of 25 mmol/L magnesium chloride solution, 65–70 μL of de-ionized water, 5 μL of each of the primers in 5-pmol concentrations, 2 μL of a 10 mmol/L dNTP solution in equimolar ratio, 1μL of Taq DNA polymerase (1 unit), and 10–50 ng of template DNA. Double-stranded PCR products were produced on a GeneE® thermal cycler (Techne Inc., Princeton, New Jersey) via 30 cycles of DNA denaturation at 95°C for 1 min, primer annealing at 48–55°C for 1 min, and DNA strand extension by Taq DNA polymerase at 72°C for 2 min. The PCR was terminated at the end of 30 cycles by a final extension at 72°C for 7 min. The double-stranded DNAs were then purified by differential filtration using ULTRAFREE®-MC filter units (30000 NMWL, Millipore Corporation, Bedford, Massachusetts) and diluted with de-ionized water based on the strength of the amplification.

Cycle sequencing of the purified amplification product was performed using the dideoxy chain termination method (Sanger, Nicklen, and Coulson, 1977) employed by the fmol®*1 Sequencing System (Promega Corporation, Madison, Wisconsin). All sequencing primers were 5′ end-labeled in a preliminary reaction involving T4 polynucleotide kinase and [γ-³²P]-dATP (Amersham Life Sciences Inc., Cleveland, Ohio). Termination products for both ITS 1 and ITS 2 were produced from the double-stranded template DNA with the terminal primers ITS-L and ITS-4, and the internal primers ITS-2 and ITS-3 of White et al. (1990). Fragments were separated electrophoretically on 0.4 mm, denaturing, 6% polyacrylamide gels (8mol/L urea) at 2300 v/60 W. Gels were fixed in 10% acetic acid for 20 min, washed in distilled water, and patted dry with paper towels to remove excess fluid. The gels were then placed in an oven at 65°C for 3 h to dry, and exposed to Kodak Biomax-100 or Bio-Max MR film for 24–48 h depending on the activity of the gel.

Sequence analysis

The boundaries of the coding (18S, 5.8S, and 26S rDNA) and spacer regions were determined by comparison to the published sequences for rice (Takaiwa, Oono, and Sugiura, 1985). Complete sequences for the entire ITS region (ITS 1, 5.8S gene, and ITS 2) were not obtained for all taxa, therefore, only ITS 1, ITS 2, and a small, 17-bp portion at the 3′ end of the 5.8S gene that included two point mutations, were included in the analysis. DNA sequences were aligned initially using CLUSTAL V (Higgins, Bleasby, and Fuchs, 1992), then adjusted manually to minimize gap number using SeqApp version 1.8a (Gilbert, 1992). Regions of ambiguous alignment were excluded from all distance calculations and phylogenetic analyses in order to reduce systematic error (Swofford et al., 1996). Absolute pairwise distances between sequences were determined in PAUP 3.1.1. (Swofford, 1993) using the DISTANCE MATRIX option. Gaps were coded as missing for all phylogenetic analyses as recommended by Wojciechowski et al. (1993). Primary sequence lengths and GC contents were determined in Amplify 1.2 (Engels, 1993). These values were manually recalculated for those sequences with ambiguous nucleotide characters (e.g., N, Y, R), which are unacceptable to the program.

Phylogenetic analysis

All phylogenetic reconstructions were performed using the computer program PAUP version 3.1.1. (Swofford, 1993) run on a Power Macintosh. Heuristic searches were performed on equally weighted characters using Fitch parsimony (1971) and a SIMPLE stepwise addition of taxa. The “save all minimal trees” (MULPARS), “tree-bisection-reconnection” (TBR), and “collapse all zero length branches (COLLAPSE)” commands were used in searches for optimal trees. Five hundred replicates employing a RANDOM addition sequence were also used in heuristic searches for islands of equally most parsimonious trees (Maddison, 1991). In order to assess the confidence that could be placed in the monophyly of clades, bootstrap (Felsenstein, 1985) and decay analyses (Bremer, 1988; Donoghue et al., 1992) were performed. Decay indices (DI) were estimated using the “converse constraint” method of Baum, Sytsma, and Hoch (1994). Bootstrap (BS) values were determined from 500 replicates using heuristic searches and a SIMPLE stepwise addition of taxa. The amount of phylogenetic information in the parsimony analysis was assessed by use of the consistency index (CI; Kluge and Farris, 1969), and the retention index (RI; Farris, 1989). The number of unambiguous character state changes along branches was determined in MacClade 3.0 (Maddison and Maddison, 1992). A preliminary analysis was conducted to assess the effect of intraspecific variation in Carex section Phyllostachys on the topology of trees. The strict consensus trees of heuristic searches involving all 42 sequences were compared to those trees produced from searches using only the first, second, or third individuals (and combinations thereof) of each of the species sampled more than once (data not shown). No topological differences were observed in these analyses; therefore, only the first individual sequenced from each species (26 sequences in total; see Fig. 1) was used in subsequent analyses to reduce computational time.

The phylogenetic utility of insertions/deletions (indels) was also explored by incorporating indels into the data set as binary characters (bases present = 1/absent = 0). A small portion (17 bp) of the 3′ end of the 5.8S gene that contained the only variability between species in this coding sequence was also included.

To evaluate the contribution of each of the spacers to the analysis and to determine their level of congruence, ITS 1 and ITS 2 sequences were analyzed separately. Heuristic tree searches only employed a SIMPLE addition sequence owing to the length of search times.

RESULTS

Sequence analysis

Aligned sequences for ITS 1, ITS 2, and 17 bp at the 3′ end of the 5.8S gene of the 26 individuals used in the final analysis are presented in Fig. 1. The sequence for the complete ITS region, including both spacers and the entire 5.8S gene of Carex superata, is given in Fig. 2. Summary statistics for all 42 individuals sequenced in this study are given in Table 3.

Spacer lengths for all carices examined ranged from 217 bp (C. peckii and C. baccans) to 223 bp (C. tonsa var. rugosperma) for ITS 1, and from 211 bp (C. filifolia) to 231 bp (C. pensylvanica and C. peckii) for ITS 2. On average, ITS 2 was 3 bp longer than ITS 1. GC contents ranged from 56.4 to 69.5% (mean = 63.4%) for ITS 1 and from 58.8 to 72.7% (mean = 67.8%) for ITS 2. The 5.8S subunit in carices, as determined from three complete sequences (C. superata, Naczi 4013 et al., C. willdenowii, Naczi 4287 & Thieret, and C. eleocharis, Bayer AB-96004 et al.; Tables 1, 2), is 166 bp in length. Alignment of the 5.8S gene with sequences from the Cyperaceae (Kobresia simpliciuscula (Wahlb.) Mack., Keleher & Punter #94/155, WIN; GenBank number GBANAF027490), Poaceae (Oryza sativa; Takaiwa, Oono, and Sugiura, 1985), and two dicot families (i.e., Fabaceae, Vicia faba; Tanaka, Dyer, and Brownlee, 1980; and Cucurbitaceae, Cucumis sativus; Torres, Ganal, and Hemleben, 1990) identified a 3-bp insertion (5′>CAT<3′) near the 3′ end that was shared by the genera Carex and Kobresia (Figs. 2, 3). With the exception of an autapomorphic mutation in C. filifolia, the only variation observed between 5.8S sequences in Carex was a T/C transition at the third position of this insertion. In order to align sequences, an equal number of unambiguous insertion/deletion (indel) events had to be inferred for each spacer (ITS 1 = 11; ITS 2 = 11). All of the indels in ITS 1 were 1 bp in length, whereas five of the 11 indels in ITS 2 were greater than 1 bp, including a unique 10-bp deletion in the ITS 2 spacer of C. filifolia. Ten of the 22 indels in the analysis were autapomorphic, while seven of the remaining 12 had a CI = 1.0.

The levels of nucleotide divergence in Carex ranged from 0.0 to 20.90% in ITS 1 and from 0.0 to 18.72% in ITS 2 (Table 3). When both spacers were considered, the pairwise sequence divergence between carices ranged from 0.0 to 19.70%. With the exception of the Firmiculmes (7.19%), sequence divergence within sections for which more than one species was sequenced was low (0.0–3.83%; Table 4). In contrast, variation among these sections was considerably higher, with the lowest values being found in comparisons between sections Careyanae and Acrocystis (6.41–7.13%; Table 5), and the highest between sections Phyllostachys and Laxiflorae (11.08–14.22%; Table 5). In all cases, the highest divergence values were obtained when members of sections in the “reduced” clade were compared to those in the “compound” clade (see below). Within section Phyllostachys, sequence divergence between species was low, ranging from complete identity (e.g., C. basiantha vs. C. superata; but see below) to 3.83%. Infraspecific variation was extremely low and never exceeded 1% of the sequences compared (Table 6). The highest divergence between populations within a species was seen in C. backii where divergence between the three populations examined ranged from 0.48 to 0.96%. A single point mutation difference was seen in one of the three individuals sampled for both C. saximontana and C. willdenowii. All other variation within species was due to ambiguous characters that may or may not be true polymorphisms. It should be noted, however, that length variation (± 1 bp) between tandem repeats within individuals (e.g., C. juniperorum and C. superata), within populations (C. superata − 2 bp), and within species (C. basiantha and C. superata) was commonly seen at the 3′ poly-A tail of ITS 1. Because of the volatility of this region, it was excluded from pairwise divergence calculations and phylogenetic analyses.

Phylogenetic analysis

The alignment of ITS spacer sequences resulted in a matrix of 465 characters of which 23 positions in ITS 1 and six positions in ITS 2 were excluded due to alignment ambiguities (marked by asterisks in Fig. 1). Of the remaining 436 characters, 181 (41.5%) were variable (94 in ITS 1; 87 in ITS 2) and 105 (24.0%) were potentially phylogenetically informative (57 in ITS 1; 48 in ITS 2). Heuristic searches of the reduced 26 taxa data matrix using either a STEPWISE or a RANDOM addition sequence of taxa recovered the same five most parsimonious trees of 341 steps each. All of the most parsimonious trees had a consistency index of 0.66 and a retention index of 0.75.

The single tree of five most parsimonious trees, topologically identical to the 50% majority rule tree, is presented in Fig. 4 along with bootstrap values, decay indices, and unambiguous character changes per branch. The only difference between this tree and the strict consensus of all five trees was seen in the loss of resolution in section Phyllostachys (dotted branches in Fig. 4). Two major clades were evident on this tree: (1) a “reduced” clade consisting of sections Phyllostachys, Filifoliae, and Firmiculmes, and (2) a “compound” clade comprising sections Careyanae, Laxiflorae, Hymenochlaenae, Acrocystis, Scirpinae, Indicae, Cruciatae, and Polystachyae. Other strongly supported groups in this analysis included the sister group to the “compound” clade composed of two Southeast Asian indocarices, C. baccans and C. cruciata, and a robust clade (DI = 5; 86% BS) consisting of sections Phyllostachys and Filifoliae. All sections for which more than one individual was sequenced were monophyletic and strongly supported (DI = 6–12; 89–100% BS), except for section Acrocystis (DI = 3; 69% BS). Although the data did not provide firm evidence for the monophyly of section Acrocystis, they did support the existence of two strong clades within it; viz. C. tonsa var. rugosperma and C. albicans, and C. pensylvanica and C. peckii. Most of the internal branches in the “compound” clade were weak, including the Careyanae/Laxiflorae lineage. Within section Phyllostachys three species pairs were recognized: C. backii and C. saximontana, C. juniperorum and C. jamesii, and C. basiantha and C. superata. Carex juniperorum and C. jamesii formed the best supported clade in the section (DI = 2; 86% BS). Little resolution was seen within section Phyllostachys, and the separation of the section into “narrow-” and “wide-scaled” clades (sensu Starr and Ford, 1995; Starr, 1997) was poorly supported. Carex latebracteata, a member of the wide-scaled clade, was present in a group that included the members of the narrow-scaled clade.

Reanalyzing the data with the inclusion of insertion/deletions (indels) and the single variable character found at the 3′ end of the 5.8S gene had little effect on parsimony analyses. The five trees obtained from these searches were topologically identical to the five found using point mutations alone. The trees from both of these analyses had the same consistency index (CI = 0.66) and retention index (RI = 0.75) and differed only marginally by certain bootstrap and decay values and by the number of their steps (data not shown).

Heuristic searches of a separate ITS 1 data set produced 79 trees, 180 steps long with a CI = 0.68 and an RI = 0.78. The strict consensus of this analysis was entirely compatible with the combined analysis and differed only in the level of resolution achieved. The strict consensus of the four trees derived from the analysis of ITS 2 sequences (154 steps; CI = 0.67 and RI = 0.74), however, was slightly different from the combined analysis in the placement of C. polystachya (subg. Indocarex) as sister to the “compound clade” instead of the C. baccans/C. cruciata clade (subg. Indocarex). Nonetheless, the two analyses were complementary, and the combination of data sets resulted in increased resolution and support for clades that were otherwise absent or weak in separate analyses (data not shown).

DISCUSSION

ITS sequence evolution in Carex

The sizes of the ITS region (602–617 bp) and spacers (ITS 1 = 217–223 bp; ITS 2 = 211–231 bp) in Carex are similar to those reported for a wide variety of flowering plants (reviewed in Baldwin et al., 1995). Sequences from the Poaceae showed the greatest similarities to Carex in both the size of the ITS region (Poaceae, 585–603 bp; Hsiao et al., 1995a, b) and its spacers (ITS 1, 214–221 bp; ITS 2, 205–221 bp; Hsiao et al., 1995a, b). However, the range of sequence divergence within Carex is generally much higher than seen in the Poaceae and is comparable to the levels of variation observed between relatively distant genera in grasses (cf. Hsiao et al., 1995a, b). In general, the range of ITS sequence divergence within Carex (0.00–19.70%) was much higher than that observed within other genera (e.g., Epilobium 0.0–12.9%; Baum, Sytsma, and Hoch, 1994; and Antennaria 14%; Bayer, Soltis, and Soltis, 1996; but cf. Arceuthobium; Nickrent, Schuette, and Starr, 1994) and was comparable to the values that are commonly observed between genera within families (e.g., Maloideae 2.7–16.1%; Campbell et al., 1995). The reasons for this are unknown, although the age of the genus and/or poor circumscription (see below) may be involved.

All of the highest pairwise sequence divergence values were observed between C. filifolia and the other members of the genus sampled. The sequence of this species had a number of features that made it distinct. It had the shortest combined ITS 1 and ITS 2 sequence by 7 bp, it possessed a considerably lower GC content (57.5%) than other carices (62–71.1%), it had a unique 10-bp deletion in ITS 2, and it possessed very long branch lengths on all trees. Carex filifolia represents one of many reduced species that have blurred generic limits within the Cariceae. Its generally well-developed rachilla prompted Torrey (1836) to treat it as an Uncinia, while its unusual inflorescence and vegetative habit suggested to Ivanova (1939) that it was best placed in the genus Kobresia. Although this analysis strongly suggests that C. filifolia (sect. Filifoliae) is sister to sect. Phyllostachys (see below), its relationship to the section and to the rest of the genus will not be understood until a larger phylogenetic study of Carex and the Cariceae has been conducted.

The poor support seen in the lower branches of the compound clade is most likely due to poor taxonomic sampling, but it could also be an indication of rapid radiation. Many authors have felt that widespread speciation has taken place as recently as the Pleistocene (Kreczetovicz, 1936; Whitkus, 1981; Kukkonen and Toivonen, 1988; Ball, 1990), and based on phytogeographical evidence, Stebbins (1981) has suggested that most speciation in Carex has taken place within the last 15 million years.

The alignment of Carex 5.8S sequences with representatives of several families including the Poaceae and Cyperaceae, indicated that a 3-bp insertion had occurred at the 3′ end of this gene (Figs. 2, 3). Interestingly this insertion was shared with the genus Kobresia, a member of the Cariceae, but not with any other taxon investigated thus far (Starr, unpublished data). To our knowledge this is the largest insertion known to have occurred within the 5.8S gene of flowering plants, and it may prove useful for helping to delimit the family, the Cariceae, or a clade within the tribe itself.

Subgeneric circumscriptions in Carex

Two major clades in Carex were identified in this study: (1) an evolutionarily “reduced” clade consisting of sections that are traditionally placed in either subg. Carex or subg. Primocarex and (2) a “compound” clade comprising sections typically placed in subg. Carex, Primocarex, and Indocarex (Fig. 4). These two groups suggest that three of the four subgenera recognized by Ku¨kenthal (1909) are artificial.

Subgenus Carex with its enormous number of species and high degree of variability in the sexual expression of the spikes is, not surprisingly, polyphyletic in this analysis (Nannfeldt, 1977; Reznicek, 1990). Although the subgenus is unnatural, this is only due to the placement of sect. Phyllostachys within the reduced clade. The multispicate “core” sections (i.e., sections Careyanae, Laxiflorae, Hymenochlaenae, and Acrocystis) typical of subgenus Carex, with terminally staminate and laterally pistillate inflorescence units, still form a weak, albeit consistent clade in our analysis. This clade, however, forms a part of a wider “compound” lineage that includes a paraphyletic subg. Indocarex and a section representative of subg. Primocarex, a result that is consistent with classifications that have merged subg. Indocarex, Carex, and Primocarex in part (e.g., Ohwi, 1936; Koyama, 1962). The placement of sections Cruciatae and Polystachyae (subg. Indocarex) as sister to the remainder of the compound clade is furthermore significant because it is compatible with many arguments based on phytogeography (Kreczetovicz, 1936; Nelmes, 1951; Ball, 1990) and inflorescence structure (Kreczetovicz, 1936; Koyama, 1962; Smith and Faulkner, 1976), that have suggested subg. Indocarex was the possible progenitor of a wider subg. Indocarex/Carex/Primocarex line.

The other major finding of this study is that subg. Primocarex appears to be polyphyletic. In all our analyses sect. Scirpinae (subg. Primocarex) was present in the compound clade and formed a weak group with C. polystachya, a North American species in subg. Indocarex. The clear separation of section Scirpinae from the “reduced” clade is significant in that it confirms the common belief that extreme reduction has occurred along several independent evolutionary lines in Carex (Mackenzie, 1935; Kreczetovicz, 1936; Nelmes, 1952; Smith and Faulkner, 1976). However, a clearer understanding of the relationships of the Scirpinae and the reduced clade to the rest of Carex and the Cariceae await a much wider tribal level study.

Sectional delimitation and taxonomic utility of the ITS region in Carex

One of the main obstacles to a phylogenetic classification of Carex at the subgeneric level is the artificial nature of many sections (Crins, 1990; Naczi, 1992). The discovery in this study of low infrasectional and high intersectional ITS sequence divergence suggests, however, that the ITS region is an effective tool for circumscribing sections in Carex and determining their relationships. This point is best exemplified by the differences found between species in sections Laxiflorae (C. laxiflora and C. blanda) and Careyanae (C. plantaginea and C. careyana). Most modern authors (e.g., Ku¨kenthal, 1909; Mackenzie, 1935; Fernald, 1950) have not recognized section Careyanae as being distinct from the Laxiflorae. Recent morphological (Bryson, 1980) and biochemical (Manhart, 1986) treatments of the Laxiflorae have likewise continued to recognize only a single section, but do concede that the Laxiflorae s.l. are composed of two distinctive groups. Naczi (1992) recognized the Laxiflorae and Careyanae as distinct sections and demonstrated in a phylogenetic analysis that the Careyanae were sister to section Griseae and not to the Laxiflorae s.s. The present phylogenetic analysis supports the conclusions of Naczi (1992) in that it indicates weak support for a Careyanae/Laxiflorae clade, but provides strong evidence (DI = 10–15; 100% BS) for the recognition of two distinct sections. Sequence divergence within (0.95%, Laxiflorae; 2.86%, Careyanae; Table 4) and between (8.15–8.59%; Table 5) sections Laxiflorae and Careyanae further supports this conclusion.

Despite the low levels of sequence divergence observed within sections, the ITS region can, in some instances, provide conserved characters that are useful for circumscribing critical groups below the sectional level. The clear separation of C. albicans from C. peckii (sect. Acrocystis) provides a good example. Although these two taxa are distinct at the species level, some authors (Gleason, 1952; Gleason and Cronquist, 1963) have treated them as varieties of C. nigromarginata Schw. The present analysis, however, places these taxa into two separate and strongly supported clades (DI = 3–4; > 90% BS), and identifies at least 12 mutational differences in their ITS regions. While the above example may seem trivial, it does suggest that DNA sequences could provide important characters to future taxonomic studies in the genus. Section Acrocystis, in particular, has numerous unresolved taxonomic problems (e.g., the C. tonsa and C. rossii Boott complexes) where the ITS region could be an effective aid in resolving the intricate taxonomy of this group.

The phylogenetic position of Carex section Phyllostachys

The results of our study provide little support for the phylogenetic hypotheses proposed by earlier authors (see subsection “Carex sect. Phyllostachys and its position within the genus Carex” in Systematic Background). Bailey's (1885, 1886) enlarged concept of the Phyllostachys that included species currently placed in section Firmiculmes (subsection Phyllostachyae) is polyphyletic in this analysis. Equally unsupported is Bailey's placement of sect. Phyllostachys close to sections Acrocystis and Laxiflorae; both of these sections are placed in a clade that is separate from the Phyllostachys. Holm's (1900) claim that the Phyllostachys were not “sufficiently characteristic” to be considered a section is clearly contradicted by the frequent recognition of its singularity in Carex (e.g., Bailey, 1886; Mackenzie, 1935; Catling, Reznicek, and Crins, 1993), and by the robust statistical (DI = 6; 99% BS) and character support (seven synapomorphies) substantiating its monophyly in this molecular study. Holm's decision to consider the Phyllostachys as primitive within section Hymenochlaenae is equally untenable and is not supported.

Our results support decisions to merge subg. Indocarex and Primocarex in part with subg. Carex (Holm, 1900, 1903; Kreczetovicz, 1936; Ohwi, 1936; Koyama, 1962), but they do not support arguments that have used the presence or abnormal growth of the rachilla to justify this merger. Kreczetovicz (1936) considered the rachilla to be a derived character whose aberrant growth, among other characters, pointed to a recent and direct origin for sections Firmiculmes and Phyllostachys from subg. Indocarex via transmutive reduction. Similarly, Holm (1900, 1903) used a teratological specimen of C. backii to argue for an expanded subg. Carex, noting that sect. Phyllostachys was “in no wise to be distinguished from the Vigneastra (= subg. Indocarex).” The present analysis, however, fails to find evidence to suggest that the presence of rachillae, or their abnormal growth, is any direct indication of phylogenetic relatedness, atavism, or primitiveness. The same type of abnormal growth of the rachilla present in sections Phyllostachys and Firmiculmes is also prevalent in groups in the compound clade such as sections Hymenochlaenae (Koyama, 1962) and Acrocystis (Svenson, 1972). In fact, the presence of rachillae and the abnormal growth of rachillae occur sporadically in all four of the subgenera in Carex (Snell, 1936; Smith and Faulkner, 1972; Svenson, 1972; Reznicek, 1990), suggesting that these characters do not confer any special insight into the phylogeny of this genus (see Reznicek, 1990).

Equally discordant with our data were Savile and Calder's (1953) phylogenetic hypotheses based on smut parasitism. In their arrangement, sect. Phyllostachys is placed in an advanced lineage within subg. Carex that includes sect. Scirpinae, while sect. Firmiculmes is far removed from the Phyllostachys in subg. Carex, and sect. Filifoliae is distantly situated at the base of subg. Primocarex. Our ITS data indicate that sections Phyllostachys and Filifoliae are sister groups, that they are closely related to sect. Firmiculmes, and that the Scirpinae are at best a distantly related section in a completely separate lineage. Contradictions between smut infection records and other data sets are common (Reznicek, 1990), but they do not mean that smut data are systematically uninformative (e.g., Nannfeldt, 1977; Kukkonen and Timonen, 1979; Savile, 1990). The value of these characters for studying phylogeny in Carex is limited, however, by an uncertain knowledge of smut phylogeny (Kukkonen and Timonen, 1979; Reznicek, 1990) and by a poor understanding of whether smut associations and morphology could be linked to other factors such as the ecology or physiology of host species (Hamlin, 1959; Smith and Faulkner, 1976).

Although Ku¨kenthal's (1909) alignment of sect. Phyllostachys with sections Rhomboidales and Elatae cannot be ruled out due to a lack of data, our results would indicate that a close relationship between these three sections is unlikely. The inflorescence morphology of sections Rhomboidales and Elatae clearly aligns them with the subg. Carex group of the compound clade and not with sect. Phyllostachys and its allies.

Of all the treatments examined, Mackenzie's (1935) is the most similar to our molecular results by its suggestion of a wider subg. Indocarex/Carex/Primocarex line and by its juxtaposition of sections Phyllostachys and Filifoliae. Unfortunately Mackenzie did not recognize subgenera and his views on evolutionary relationships in Carex are only expressed through a terse and vague conspectus. It is therefore difficult to determine the extent of his disagreement with Ku¨kenthal's (1909) subgeneric circumscriptions or whether the proximity of sections Phyllostachys and Filifoliae in his system implies a relationship. Our hypothesis that sections Phyllostachys, Filifoliae, and Firmiculmes might form an evolutionarily reduced clade within the genus that is separate from the members of subg. Carex is novel, but the concept of a wider, evolutionarily reduced group in Carex is not.

Phylogeny in Carex section Phyllostachys

Although our study helps clarify the phylogenetic position of section Phyllostachys within Carex, the ITS region was not sufficiently variable to adequately resolve infrasectional relationships. The topology of our tree does show some correlation to a previous analysis based on morphology (Starr and Ford, 1995) in the recognition of three species pairs (i.e., C. backii/C. saximontana, C. juniperorum/C. jamesii, and C. basiantha/C. superata), but it differs by the placement of Carex latebracteata in a separate clade from C. backii and C. saximontana. These results, however, are unreliable since statistical support for the phylogeny is weak and only two clades, the species pairs of C. juniperorum/C. jamesii and C. basiantha/C. superata, do not collapse in the strict consensus tree.

The notable differences in anatomy, micromorphology, and macromorphology within section Phyllostachys (Catling, Reznicek, and Crins, 1993; Starr, 1997; Naczi, Reznicek, and Ford, 1998) suggested that the rate of evolution in these characters was higher than the mutation rate of ITS. By combining the above characters with ITS sequence data we achieved a much better understanding of infrasectional relationships within the Phyllostachys. The results of this study will be presented in a companion paper (Starr, Ford, and Bayer, unpublished data).

CONCLUSIONS

Our study provides evidence to suggest that subgenera Indocarex, Carex, and Primocarex are artificial as traditionally circumscribed. As many previous authors have believed (e.g., Bailey, 1886; Kreczetovicz, 1936; Ohwi, 1936; Koyama, 1962), subg. Indocarex, Carex (in part), and a portion of subg. Primocarex form a monophyletic group. Phylogenetic hypotheses and relationships based on the presence or abnormal growth of the rachilla are not supported in this analysis. Extreme reduction appears to have occurred along several different lineages in Carex. Our data support the conclusion that an evolutionarily “reduced” clade, made up of sections Phyllostachys, Filifoliae, and Firmiculmes, is separate from a main lineage composed of species in subg. Carex, Indocarex, and Primocarex. Unfortunately, a clearer understanding of the relationships of the reduced clade to the rest of Carex and the Cariceae could not be determined due to the small taxonomic sample. Whether this clade is more closely related to elements within Carex or outside of the genus as suggested for Carex filifolia (e.g., Ivanova, 1939) can only be resolved in a much wider tribal level study.

One of the most enduring problems in the systematics of Carex has been the circumscription of sections. Although the ITS region was not variable enough to resolve relationships within sect. Phyllostachys, it shows considerable potential for clarifying sectional limits and for hypothesizing relationships between sections in Carex. The strong support for the monophyly of the morphologically similar sections Laxiflorae and Careyanae illustrates the potential of the region for sectional delimitation.

Any study that includes only a small fraction of a group as large as Carex must come with a caveat. Many of the relationships that have been enumerated herein can only be expected to change as a wider taxonomic sample is introduced. Phylogenetic relationships within Carex will not be completely resolved by 450 bp, and 26 taxa will not be sufficient to unravel subgeneric problems that have existed for over a hundred years. Phylogenetic relationships within Carex can only be resolved through a multidisciplinary approach, using a much wider survey of taxa. However, the potential of this region, and of DNA characters in general, for helping to resolve many long-standing phylogenetic and taxonomic problems in this genus has been demonstrated.

Table 1. Collection data for populations of species sampled from Carex section Phyllostachys. Vouchers are deposited at WIN unless otherwise noted. Individuals sampled from the same population are numbered (1) and (2). GenBank accession numbers are given in succession for ITS 1 sequences and for combined sequences consisting of ITS 2 and 17 bp at the 3′ end of the 5.8S gene.^a Sequences for the entire ITS region (ITS 1, 5.8S, ITS 2) were determined in those individuals with a single GenBank accession number.

Table 2. Taxonomy and collections of the taxa used in this study. Vouchers are deposited at WIN unless otherwise noted. GenBank accession numbers are given in succession for ITS 1 sequences and for combined sequences consisting of ITS 2 and 17 bp at the 3′ end of the 5.8S gene. Sequences for the entire ITS region (ITS 1, 5.8S, ITS 2) were determined in those species with a single GenBank accession number.

Table 3. General sequence statistics for ITS 1 and ITS 2 (alone and combined).

Table 4. Sequence divergence within sections for which more than one species was sequenced. N denotes the number of individuals sequenced per section.

Table 5. Sequence divergence between sections for which more than one species was sequenced. Percentage divergence is given above the diagonal, and absolute differences are given below the diagonal.

Table 6. Sequence divergence within species of Carex sect. Phyllostachys.

Figure 1

The complete sequence for ITS 1 (positions 1–230), ITS 2 (positions 248–481), and 17 bp at the 3′ end of the 5.8S gene (base pairs 231–247), for each of the 26 taxa used in the final analysis. Regions of ambiguous alignment that were not used in phylogenetic analyses are indicated by asterisks below sequences. Insertions/deletions (indels) used in the analysis are indicated by number below sequences. An autapomorphic insertion (position 113) of a single base pair in one of the three individuals of C. backii (Reznicek 6364 & Reznicek; Table 1) sequenced is not shown. The complete length of the combined ITS 1 and ITS 2 spacers (excluding the 17 bp of the 5.8S gene), for each taxon, is given in brackets at the end of sequences.

Figure 2

The complete sequence for the ITS region of Carex superata, showing the ITS 1 and ITS 2 spacers, the complete 5.8S gene, and portions of the 18S and 26S genes of nrDNA. Coding sequences are boxed. The position of the 5′>CAT<3′ insertion in the 5.8S gene is indicated by underlining.

Figure 3

The aligned 3′ end of the 5.8S gene for the Cyperaceae (Kobresia and Carex; Starr, this study), Poaceae (Oryza sativa; Takaiwa, Oono, and Sugiura, 1985), Fabaceae (Vicia faba; Tanaka, Dyer, and Brownlee, 1980), and Cucurbitaceae (Cucumis sativus; Torres, Ganal, and Hemleben, 1990). Note the 3-bp insertion (5′>CAT<3′) shared by Carex and Kobresia.

Figure 4

The single tree of five most parsimonious trees resulting from the phylogenetic analysis of Carex section Phyllostachys and ten putatively related sections that is topologically identical to the 50% majority rule tree. The members of subgenus Vignea were used as an outgroup. The names of the sections and subgenera are given in succession after the brackets to the right of the specific epithets. Branches that collapse in the strict consensus of the five most parsimonious trees are dotted. Unambiguous base pair changes and decay indices (in parentheses) are given above the branch, and bootstrap values are given below the branch. The consistency index (CI), retention index (RI), tree length, and number of potentially phylogenetically informative characters are given in the box at the top left.