RECLASSIFICATION OF NORTH AMERICAN HAPLOPAPPUS ( COMPOSITAE : ASTEREAE ) COMPLETED : RAYJACKSONIA GEN . NOV . 1

Rayjacksonia R. L. Hartman & M. A. Lane, gen. nov. (Compositae: Astereae), is named to accommodate the "phyllocephalus complex," formerly of Haplopappus Cass. sect. Blepharodon DC. The new combinations are R. phyllocephalus (DC.) R. L. Hartman & M. A. Lane, R. annua (Rydb.) R. L. Hartman & M. A. Lane, and R. aurea (A. Gray) R. L. Hartman & M. A. Lane. This transfer completes the reclassification of the North American species of Haplopappus sensu Hall, leaving that genus exclusively South American. Rayjacksonia has a base chromosome number of x = 6. Furthermore, it shares abruptly ampliate disk corollas, deltate disk style-branch appendages, and corolla epidermal cell type, among other features, with Grindelia, Isocoma, Olivaea, Prionopsis, Stephanodoria, and Xanthocephalum. Phylogenetic analyses of morphological and chloroplast DNA restriction site data, taken together, demonstrate that these genera are closely related but distinct.

With the breakup of Haplopappus sensu lato, all of the taxa recognized by Hall (1928) have been accommodated within previously named genera, except for H. phyllocephalus DC. and its variety (at which rank Hall treated H. annuus [Rydb.] Cory) and H. aureus A. Gray, also called the "phyllocephalus complex" (Venugopalan, 1966;Hartman, 1976). In this paper, we present evidence for the generic status of this group, formally describe the new genus, and present an analysis of its phylogenetic relationships with certain other Astereae. With the establishment of this genus, all of the taxa formerly of Hap-1 Manuscript received 4 May 1995; revision accepted 27 June 1995. The authors thank Billie L. Turner, Guy Nesom, and herbarium TEX-LL for loan of specimens; John Semple and Gregory Brown for valuable discussion of Haplopappus and related genera; David Morgan and Youngbae Suh for the use of cpDNA restriction site data; David Keil for assistance with Latin; Stacie Kawaguchi for her assistance in the construction of Table 1 and John Strother for his donation of its first incarnation; Christopher Haufler, Gregory Brown, John Strother, and Theodore Barkley for critical readings of the manuscript; and Raymond C. Jackson for the use of his name. This study was supported by grants to MAL from NSF (BSR 8508631 and BSR 8908963) and the University of Kansas Graduate Research Fund. 2Author for correspondence (Fax: 913-864-5093; e-mail: MLA-NE@UKANS.EDU). lopappus sensu Hall (1928) are reclassified and are currently being treated as indicated in Table 1.

MATERIALS AND METHODS
Morphological data-In Table 2 we compare morphological features of our new genus and the six other genera that we consider to be most closely related. Even though we have hypothesized (Hartman, 1990;Lane and Li, 1993) that they might be part of the "x = 6" group of Astereae, Pyrrocoma, Vanclevea and Xylorhiza are not included because phylogenetic analysis of a large cpDNA data set that contains 163 taxa of Astereae and 729 restriction sites obtained from digests with 16 enzymes (M. Lane et al., 1996;mutation ;t --1:41 0 C.) 00 RI , ! -.,2-      CB, w w w w w w rA ry) rA rA ri) ry) cn cn rn ry) 0 r-4 w ry) rA :a9 9 9 to az N~~~~~~~~~~. H--that these three genera are not a part of the clade with abruptly ampliate disk corollas and deltate style-branch appendages. Certain of the features presented in Table 2 are ceded for phylogenetic analysis. The characteristics in Table 2 are consistent among all the species of a genus; we have encapsulated the variation within the genera in the descriptions of characters and in the scorings used for the phylogenetic analysis (e.g., polymorphisms are so coded). Hazardia Greene was chosen as the outgroup (codings as indicated in the caption of Table 2) because the cpDNA analysis mentioned above indicated that Hazardia is a member of a clade that is sister to the clade containing the genera of interest in this study. A data matrix was constructed using MacClade Version 3.0 (Maddison and Maddison, 1992). The matrix comprised seven ingroup and one outgroup taxa vs. the 17 coded characters from Table 2.
Molecular data-The cpDNA data set of Lane and coworkers (1996)  ) was drawn from the large data set using MacClade. Of the 729 characters in the whole data set, 38 were found to be phylogenetically informative for the taxa of interest here; they were extracted from the larger data set and used in our analyses.
Combined data-Following the analysis of molecular data, a third matrix was compiled; this included the 17 morphological characters plus the 38 molecular characters. Two exhaustive searches for shortest trees were conducted using the third matrix: one that excluded Olivaea because no data were available for the 38 molecular characters, and one that included Olivaea, coded as missing data for the molecular characters.
Phylogenetic analyses-Exhaustive searches for shortest trees were made using Phylogenetic Analysis Using Parsimony Version 3.1.1 (PAUP; Swofford, 1993). Multistate characters were treated as unordered and as representing polymorphisms; ACCTRAN optimization was used in all cases. Trees were drawn for presentation using MacClade Version 3.0 (Maddison and Maddison, 1992).

RESULTS
The distinctions among genera presented in Table 2 are those that we have found to be most taxonomically informative during several years of study of these members of Astereae. In addition, the analysis of morphological and cpDNA data provides strong support for the segregation of the "phyllocephalus complex" from Haplopappus sensu Hall (1928). Therefore, we here establish the genus Rayjacksonia to accommodate R. phyllocephala (type species), R. annua, and R. aurea. The formalities of classification and nomenclature are presented below.
A key to the species of Rayjacksonia, descriptions, and discussion of their relationships will appear in a forthcoming thesis and publication by one of our students.
The exhaustive search of the morphological data set resulted in five fully resolved, equally parsimonious trees (Figs. 1-5) (Table 2) for Grindelia, Isocoma, Olivaea, Prionopsis, Rayjacksonia, Stephanodoria, and Xanthocephalum. The outgroup, Hazardia, was scored as indicated in the caption of Table 2. One of these trees (Fig. 5) is equivalent to a 60% majority rule consensus tree. Fig. 6. The strict consensus tree of the five most parsimonious trees (Figs. 1-5) derived from the analysis of morphological data.
consistency index (RC) of 0.493. One of the five trees (Fig. 5) is equivalent to a 60% majority rule consensus tree. The strict consensus, however, is completely unresolved (Fig. 6).
Analysis of the cpDNA data for six of the seven genera of this clade (excluding Olivaea) also resulted in five most parsimonious trees    ) for Grindelia, Isocoma, Prionopsis, Rayjacksonia, Stephanodoria, and Xanthocephalum (data for Olivaea were unavailable; the outgroup is Hazardia). One of these trees (Fig. 11) is equivalent to a 60% majority rule consensus tree. Fig. 12. The strict consensus tree of the five most parsimonious trees (Figs. 7-11) derived from the analysis of cpDNA restriction site data. trees (Fig. 11) is equivalent to a 60% majority rule consensus. In this case, the strict consensus (Fig. 12) is slightly more resolved, with Stephanodoria and Xanthocephalum forming a clade, as do Grindelia and Prionopsis.
The first "total evidence" analysis (using six of the genera and 55 characters) resulted in two equally parsimonious trees (Figs. 13, 14) of 71 steps (CI = 0.845, RI = 0.676, RC = 0.572). The topology of Fig. 13, in which Isocoma is a sister group to the Stephanodoria-Xanthocephalum clade, while Rayjacksonia is sister to the Grindelia-Prionopsis clade, is unlike any of the trees shown in Figs. 1-12. The topology of the tree shown in Fig. 14 is equivalent to that of Fig. 7, which places Isocoma and Rayjacksonia in clades distinct from and basal to the Stephanodoria-Xanthocephalum and Grindelia-Prionopsis clades. The second "total evidence" analysis, which included all seven genera and 55 characters although the molecular data are missing for Olivaea, resulted in a single most parsimonious tree (Fig. 15) Fig. 15. The single most parsimonious tree (73 steps, CI = 0.849, RI = 0.703, RC = 0.597) that results from an analysis of Grindelia, Isocoma, Olivaea, Prionopsis, Rayjacksonia, Stephanodoria, and Xanthocephalum (the outgroup is Hazardia) using a combined matrix that included 17 morphological and 38 cpDNA restriction site characters. Numbers below branches indicate unambiguous changes that support the indicated clade. Only morphological characters were available for Olivaea, so lowercase letters are used for the Olivaea label, as in Figs. 1-6; genera for which molecular data were available are labeled in uppercase, as in  to Fig. 8, except that Olivaea is included as sister group to the Stephanodoria-Xanthocephalum clade.

DISCUSSION
The 15 trees resulting from the analyses conducted in this study are equally well supported, with CIs > 0.82 and RCs > 0.49 in all cases. Only two of the 15 trees (Figs. 7,14) are fully congruent, which is attributable to the larger number of characters in the molecular data set (38) vs. the morphological one (17), and the absence of Olivaea from the analysis that produced Fig. 14. Two other trees (Figs. 8,15) might be congruent but must be questioned because Olivaea was excluded from the analysis from which Fig. 8 was derived, and because data were missing for Olivaea for 38 of the 55 characters used to produce Fig. 15. Additional molecular (e.g., ITS sequence) data might allow for complete resolution of these relationships, but all that the current analyses can show is that these seven genera are closely allied with one another. The lack of resolution in the strict consensus trees (Figs. 6,12) indicates that different pairings of the genera share different combinations of characteristics derived from their common ancestor. It seems likely that these groups did not arise in strictly dichotomous fashion.
Rayjacksonia is variously allied with Prionopsis, the Grindelia-Prionopsis clade, or Isocoma, or it is maintained as a distinct clade (Figs. 1-15). Therefore, in our opinion, Rayjacksonia should be treated as an independent genus and not be united with Isocoma (Jackson and Dimas, 1981) or regarded as more closely related to Isocoma (Hartman, 1976) than it is to any of the other genera in this group. Further, Xanthocephalum and Isocoma are separated by several unambiguous changes ). As we have stated previously (Hartman and Lane, 1991), they should be maintained as distinct genera, even though an intergeneric hybrid is known.
[Vol. 83 Stephanodoria and Xanthocephalum are sister groups in all of the analyses that include molecular characters, but are not always united in the trees resulting from the morphological data alone (e.g.,. It is tempting to suggest that Stephanodoria should be submerged within Xanthocephalum (and indeed the combination exists to accommodate this interpretation; Robinson, 1893), except that two of the equally parsimonious trees include Isocoma in this clade (Figs. 2,13) and two include Olivaea in it (Figs. 1, 15). No tree includes all four of these genera in a single clade. Therefore, we believe that the distinct status of these genera should be maintained until data are obtained that unequivocally support one or another of the possible mergers.
The genus is named for Dr. Raymond C. Jackson, who first reported the correlation in the Astereae between a base chromosome number of x = 6 and the abruptly ampliate or "goblet-shaped" disk floret corollas. Furthermore, he has spent much of his professional career investigating species of Haplopappus sensu lato.