Discussion of preliminary results/evidence A. Fern phylogeny and basal relationships of leptosporangiate ferns. We propose here to expand upon morphological and molecular studies that we have already published (109, 53) in an attempt to provide resolution for basal fern relationships that remain elusive. Our morphological studies will be expanded to include the basal taxa that were not included in earlier analyses (see morphological data set below). Molecular phylogenetic analyses of ferns are based primarily on the chloroplast gene rbcL, though there are now a few studies using other genes that include a small sampling of ferns [18S: 142, 73 and 16S: 90]. Even for angiosperms, there are few phylogenetic studies that have used other gene sequences across a broad range of taxa, and still fewer have compared results from two or more gene sequences for the same taxa (1, 66, 69, 101). The study proposed here will focus on the basal fern relationships using the chloroplast genes atpB and 16S, and the nuclear genes 18S and 26S, in addition to the rbcL data that are already published (53), as a new source of phylogenetically informative data. Cladograms based on all five genes (and morphology) will be evaluated and compared in terms of their resolution and congruence, as well as for various measures of phylogenetic signal, and for the molecular data sets: transition/transversion bias, sequence divergence, and homoplasy. The genes we have selected for analysis are the best studied at present and most likely to yield important phylogenetic results for higher level vascular plant taxa. rbcL. Analyses of rbcL [see symposium issue of American Fern Journal 85(4)] provided many new insights into fern phylogeny, but was notably weak in providing new knowledge on basal fern relationships. In order to compare to our smaller preliminary analyses for "new" genes shown here that are sequenced for fewer taxa, we show an rbcL analysis in Fig. 4 of 17 fern taxa using the seed plant Cycas as an outgroup. "Polypodiaceous" ferns are underlined. Bootstrap support for basal taxa with rbcL is very weak (<50%). This is not improved by adding more taxa to the analysis (53). atpB. The atpB gene is located in the large single copy region of the chloroplast genome, and is contiguous with the atpE gene and downstream from the rbcL gene. atpB encodes the b subunit of ATP synthase, which has a highly conserved structure that couples proton translocation across membranes with the synthesis of ATP (66). Hoot et al. (66) demonstrated the utility of atpB for comparative sequence studies at higher taxonomic levels. It has a rate of evolution very similar to that found for rbcL, and like that gene it has no insertions or deletions, no introns, and is readily aligned (144). P. Wolf's preliminary attempts to amplify atpB from ferns using published primers (66) consistently failed. Success was finally achieved using cloned fragments of Adiantum chloroplast DNA (kindly provided by M. Hasebe, Univ. of Tokyo) and the published physical map of the Adiantum chloroplast genome (50). S tandard M13 primers were then used to sequence the appropriate inserts (numbers 1 and 18 of ref. 50). Sequences were aligned to tobacco and rice atpB sequences, obtained from GenBank, and PCR and sequencing primers were designed. Using this strategy, a preliminary data set was generated by P. Wolf for a 680-bp fragment (tobacco positions 1163-1844) from several fern taxa. The 680-bp fragment of atpB amplified consistently from all fern taxa screened. For all samples a single PCR product was obtained that provided unambiguous sequence data. A prelimininary topology is shown in Fig. 5 for 18 ferns, with Pinus as an outgroup, indicating that for a total of 588 characters, 198 are parsimony informative (34%). This analysis resulted in 3 most parsimonious trees (one of which is shown) of 484 steps. The distribution of random trees was highly skewed to the left (g1 = -0.55, p < 0.05; 62) indicating a nonrandom data set, which is consistent with a phylogenetic signal. Although not an identical data set of taxa to that used in the rbcL analsyis, it is most noteworthy that there is greater bootstrap support for basal branches in the atpB topology. From estimates of sequence divergence values, atpB appears to be evolving slightly slower than rbcL (Wolf, unpubl.). For six taxa examined for both rbcL and atpB the mean sequence divergence was 14.0 for rbcL and 12.0 for atpB. This is consistent with results from a similar comparison between the same genes in angiosperms (66) . 16S: The chloroplast 16S rDNA gene is less divergent than rbcL and has been used to construct phylogenetic trees at very broad levels, to include cyanobacteria and plastid sequences (45). Mishler et al.'s (93) analysis of 370 bp of the chloroplast 16S and 23S rDNA genes from a diverse sampling of 11 bryophytes indicated that these sequences may be able to resolve deep branches of land plant phylogenies. An initial study using 25 land plants (including 10 ferns) has just been published by Manhart (90), indicating their utility in vascular land plants (see also attached letter from J. Manhart). As with atpB, a leptosporangiate fern clade was strongly supported with 16S, which was not the case with rbcL. This is also shown here in Fig. 6 for a smaller analysis of the 10 fern taxa, rooted with seed plants. In addition, we also show strong bootstrap support for basal branches of the topology, as in Fig. 5 for atpB. 18S: Preliminary studies suggest that 18S nrDNA may be useful for examining relationships among the main groups of pteridophytes (73, 74) and also for examining the base of the leptosporangiate fern clade (110, 142). Although taxon sampling for 18S nrDNA is considerably less than for rbcL (less than 20 pteridophytes examined to date), preliminary phylogenetic analyses result in trees similar to, though somewhat less resolved, than those for rbcL , though the bootstrap support is better than for rbcL (see Fig. 7). The g1 statistic for this dataset is comparable to that for 16S. 26S: Primers for 26S amplification were designed from published sequences (145) by P. Wolf. Nuclear 26S nrDNA did not amplify in all fern taxa. Amplification was often unreliable and multiple PCR products were common. Only two long sequences (each 1377 bp) could be determined with confidence: one for Vandenboschia and one for Blechnum. These differed at 173 nucleotide positions, corresponding to 12.6% sequence divergence. 26S is evolving slower than rbcL but faster than 18S. A central region of 334 bp was successfully sequenced from eight taxa and based on patterns of divergence observed between distant taxa, we are confident that 26S is worthwhile pursuing for resolving basal relationships of ferns. B. Character Evolution and Diversification. In the last ten years, the study of character evolution has become increasingly sophisticated with the application of phylogenetic criteria (5, 23, 49, 83, 85, 104, 105, 126). The combined study of phylogeny and character evolution provides a key framework for understanding the relationship among the different components of species diversification and ecological adaptation. This has permitted the reevaluation, in a rigorously analytical fashion, of many classic hypotheses regarding the importance of character change patterns as a factor in diversification (3, 4, 18, 19, 31, 39, 119, 120, 121). Rothwell (121) noted that in terms of species richness, homosporous pteridophytes are more successful than all other non-angiospermous grades combined, particularly when compared to non-angiospermous seed plants (Fig.2). His phylogenetic comparison evaluated the diversity of growth forms and life history patterns that likely were associated with pteridophytes being successful throughout time. This in turn prompted us to examine the current fern phylogenetic framework (109), for species richness patterns. We observe that by far the greatest species diversity within ferns is restricted to the "polypodiaceous" clade (Fig. 3). Of the 18 most speciose genera of ferns, all except two fall in the "polypodiaceous" clade, for a total of about 80% of extant ferns in this clade (Fig. 3). The basal clades of ferns comprise ca. 70% of the total number of fern families (following 75), but only about 20% of the total number of species. Why is the "polypodiaceous" clade more speciose? Were there key innovations associated with this evolutionary success? Is there a correlation between certain character state transformations or ecological conditions (e.g., epiphytism) and diversification? In a symposium held nearly 25 years ago, pteridologists lamented the absence of studies correlating structure and function or the significance of various morphological features in ecological terms, and the lack of a broad conceptual framework (92, 139,140). "Our greatest overall challenge in future fern research, the most unexploited frontier, is that of explanation of characters-states in terms of their biological significance" (139, pp. 94-95). Due to the lack of a phylogeny for ferns, their questions regarding character evolution and diversification could not be posed and tested in a rigorous fashion as they can now. Phylogenetic trees specify an order of origination of characters and clades and can therefore be used to test hypotheses on causal relationships between characters and changes in diversity. Different phylogenetic hypotheses can have rather different implications. The following characters in particular will be optimized using parsimony onto topologies derived from analyses of morphology and the cp and nrRNA sequences: Sporangial characters. Patterns in various sporangial characters such as sporangial stalk width, stalk length, and position of annulus are used to circumscribe various groups of ferns. A well-supported topology will permit assessment of polarity transformations and provide insight into the evolution of the sporangium and its impact on the diversification of ferns. Spore characters. Patterns in exospore structure suggest a clearly defined reduction sequence in the number of layers, which has been used to delimit basal from more derived taxa. The number of exospore layers may have been subject to homoplasy. Observation of exospore structure requires laborious transmission electron microscope studies, and so relatively few species have been investigated (137). The results of this study may suggest priorites for further TEM studies. Gametophytic characters. Characters related to the gametangia such as antheridium/archegonium position, number of antheridium wall cells/number of archegonial neck cell tiers, are likely to show transfomation patterns related to phylogenetic position that can be reexamined in terms of their potential impact on fern diversification. Sporophytic vegetative characters. Various features (e.g., vein fusion, hydathodes, rhizome scales) that have been important in the classification of various fern groups can be investigated for patterns associated with basal vs. more derived ferns. Preliminary analyses of character evolution in Pryer et al. (109) suggest that character transformations towards a narrower sporangium stalk (Fig. 8), an interrupted annulus (Fig. 9), and a reduction in number of antheridium wall cells (Fig. 10) may have been associated with the diversification of the "polypodiaceous" clade. Likewise, a transition to fewer exospore layers (Fig. 11) includes the "polypodiaceous" clade and also some of the more basal clades, leading one to surmise the associated and possibly correlated patterns. The identification of such patterns is a first step. Next, one can test whether the patterns observed are evidence for correlated evolutionary change (see section on "Comparative methods analyses" in proposed research plan). N.B. Numbers in parentheses refer to literature citations listed in References. Top of page