Generic delimitation in the Hymenoxys complex has long been problematic. One taxonomic extreme would recognize only Hymenoxys, whereas the other would split the obviously related taxa into as many as eight genera. This study examined restriction site variation in both cpDNA and nrDNA in the Hymenoxys complex. Fifty-six populations representing 21 species (four with two varieties each) and six of the eight possible genera were analyzed using 21 enzymes, which resulted in the detection of 358 restriction site changes of which 171 were potentially phylogenetically informative. Wagner parsimony using synapomorphic characters generated 26,600 equally parsimonious trees of 223 steps with a consistency index of 0.76 and a retention index of 0.98. Bootstrap analysis indicated that the major clades were strongly supported. The DNA tree supports the recognition of Tetraneuris as a genus separate from Hymenoxys, and the inclusion in Hymenoxys of taxa that at times have been split into the genera Dugaldia, Macdougalia, Phileozera, Picradenia, and Plummera.
Hymenoxys sensu lato includes taxa that are referable to Dugaldia Cass., Hymenoxys Cass., Macdougalia A. Heller, Phileozera Buckley, Picradenia Hook., Plummera A. Gray, Rydbergia Greene, and Tetraneuris Greene.
A reasonable case can be made on morphologic, cytologic (chromosome number), and chemical grounds for combining all of the taxa discussed here into one genus (Hymenoxys) or splitting them into at least the eight genera listed above (Bierner  provides a comparison of these taxa with regard to morphology, cytology, flavonoid chemistry, monoterpene chemistry, and sesquiterpene lactone chemistry).
Difficulties with generic delimitation in this group are illustrated by differences in the treatments presented by various workers during this century. Rydberg (1915) recognized Dugaldia, Hymenoxys, Macdougalia, Plummera, Rydbergia, and Tetraneuris, and placed the taxa referable to Phileozera and Picradenia in Hymenoxys. Turner and Powell (1977) recognized only Dugaldia and Hymenoxys, and submerged Macdougalia, Phileozera, Picradenia, Plummera, Rydbergia, and Tetraneuris in Hymenoxys. Robinson (1981) recognized Hymenoxys, Macdougalia, Plummera, and Tetraneuris, and submerged Phileozera, Picradenia and Rydbergia in Hymenoxys, and Dugaldia in Helenium L. Bierner (1994) recognized only Hymenoxys and Tetraneuris, and submerged Dugaldia, Macdougalia, Phileozera, Picradenia, Plummera, and Rydbergia in Hymenoxys as subgenera.
As major evidence for this latter treatment, Bierner (1994) cited preliminary unpublished analyses of DNA restriction site data (the information now being published in this paper) that clearly separated the taxa of Tetraneuris into one branch of the DNA phylogenetic tree and grouped the other taxa into another branch. He also cited the presence of 6-methoxy flavone aglycones, flavonol 3-O-acetyl glycosides, and secopseudoguaianolides in all of the groups except for Tetraneuris, and the presence of 6,8-dimethoxy flavone aglycones and monoterpene glycosides only in Tetraneuris. The one exception was Hymenoxys texana (tentatively placed by Bierner  and Spring et al.  in Hymenoxys subgenus Phileozera), which contains monoterpene glycosides and lacks seco-pseudoguaianolides (Spring et al., 1994). In fact, the phenogram produced by Spring et al. (1994) based on presence or absence of chemical components isolated from glandular trichomes placed Hymenoxys texana with Tetraneuris.
This paper presents a phylogenetic analysis of DNA restriction site data from taxa. in this complex. The major questions focus on the status of Dugaldia, Macdougalia, Plummera, and Tetraneuris. The working hypothesis at the beginning of this project (based on all evidence available when the study began in 1992) was to recognize Dugaldia, Hymenoxys, Macdougalia, Plummera, and Tetraneuris as distinct genera, and to recognize Phileozera, Picradenia, and Rydbergia as subgenera of Hymenoxys. Hence, the generic names Dugaldia, Macdougalia, and Plummera are used throughout this article even though the results of this study indicate that they are in fact congeneric with Hymenoxys.
Representatives of Hymenoxys subgenera Hymenoxys and Rydbergia were not available for this study. We feel confident, however, from evidence presented by Bierner (1994) and Spring et al. (1994) that the taxa comprising these subgenera are clearly associated with those referable to Dugaldia, Macdougalia, Phileozera, Picradenia, and Plummera.
Materials and Methods
DNA restriction site variation was examined in samples from 56 populations representing 25 taxa of Hymenoxys sensu lato (Table 1): one of the three taxa of Dugaldia, two of the three taxa of Hymenoxys subgenus Phileozera, 10 of the 11 taxa of Hymenoxys subgenus Picradenia, the one taxon of Macdougalia, both taxa of Plummera, and nine of the 11 taxa of Tetraneuris. DNA restriction site variation was compared to two taxa included as outgroups (Table 1): Helenium drummondii and Psilostrophe villosa.
Total DNAs were isolated from leaf material, all of which was field collected except for that of Hymenoxys texana (greenhouse-grown at the University of Texas at Austin). DNAs from the Turner collection and the Bierner 1988 collections were extracted from fresh material by Ki-Joong Kim, those from the Bierner 1989 and 1991 collections were extracted by the first author from material stored at -70°C, and those from the Bierner 1992 collections were extracted by the first author from fresh material. In all cases, the CTAB method of Doyle and Doyle (1987) was used, and the DNAs were then further purified in cesium chloride/ethidium bromide gradients as described by Sambrook et al. (1989). Restriction enzyme digestions, agarose gel electrophoresis, bi-directional transfer of DNA fragments from gels to Zetabind (CUNO Inc., Meriden, Connecticut) nylon filters, labeling of recombinant plasmids by nick-translation, filter hybridizations, and autoradiography were performed as described in Palmer (1986), Jansen and Palmer (1987), and Palmer et al. (1988). Enzyme digestions were done using 21 sixbase pair restriction endonucleases: AvaI, AvaII, BamIII, BanI BanII, BeII, BgIII, BstNI, BstXI, ClaI, DraII, Eco0l09, EcoRI, EcoRV, HaeII, HincII, HindIII, NeiI, NsiI, SspI and XmnI. Restriction fragment data were obtained from hybridizations done with 15 cloned SacI restriction fragments of lettuce cpDNA (Jansen and Palmer, 1987) and three subclones of the Helianthus argophyllus Torrey & A. Gray ribosomal repeat (M. Arnold, unpublished). We did not map sites for either cpDNA or nrDNA; however, this was unnecessary because levels of variation were low enough to allow us to reliably interpret restriction site changes (Jansen et al., in press). We were also able to distinguish between restriction site and length changes, because the latter were evident with most of the 21 enzymes examined.
Sources of DNA for taxa of Hymenoxys sensu lato, Helenium drummondii, and Psilostrophe villosa. Multiple samples of a taxon are given unique numbers within the taxon that follow the scientific name. All vouchers are deposited at TEX.
Restriction site data were subjected to Wagner parsimony (Farris, 1970) analyses using a Macintosh Quadra 700 microcomputer and PAUP version 3.1.1 (Swofford, 1993). Tree Bisection Reconnection (TBR) branch-swapping and Mulpars options were used to search for the most parsimonious Wagner trees, and 100 random additions were performed to search for islands of equally parsimonious trees (Maddison, 1991). Helenium drummondii and Psilostrophe villosa were selected as outgroups, but it became obvious after a first computer run with PAUP that Psilostrophe villosa belonged in the ingroup. Helenium drummondii, therefore, served as the outgroup for subsequent analyses. The PAUP CONTREE option was used to construct a strict consensus tree. Bootstrap (Felsenstein, 1985) analyses (1000 replicates) in PAUP were used to derive confidence intervals using the same options as the parsimony analyses except that only one random addition was performed without Mulpars.
We were able to detect 358 restriction site changes, of which 187 were autapomorphic and 171 were potentially synapomorphic. The list of all character changes and the data matrix for the 171 synapomorphic characters have been deposited under Hymenoxys at TEX and are also available from the first author.
Of the 187 autapomorphic characters, 154.(82%) were confined to three taxa: 73 (39%) in Helenium drummondii, 39 (21%) in Psilostrophe villosa, and 42 (22%) in Hymenoxys texana. The other 33 autapomorphic characters were distributed among 19 populations representing 15 taxa; eight populations had one autapomorphy, nine populations had two, one population had three, and one population had four.
Wagner analyses of the 171 synapomorphic characters, of which 144 (84%) were in the chloroplast genome, produced 26,600 equally most parsimonious trees of 223 steps with a consistency index (excluding autapomorphic characters) of 0.76, and a retention index of 0.98. Because of limitations of computer memory, it is likely that not all of the equally parsimonious trees were detected. Topology of the consensus tree using all 171 synapomorphic characters is congruent with that of the tree produced without using nrDNA characters. The consensus tree along with character support and bootstrap values, and with autapomorphic characters added to it, is shown in Figure 1.
The ingroup taxa are separated into two clades. One includes taxa referable to Hymenoxys subgenus Picradenia, H. subgenus Phileozera, Dugaldia, Macdougalia, and Plummera. The monophyly of this clade is strongly supported by the presence of 33 shared characters and a bootstrap value of 100%. The other clade contains taxa referable to Tetraneuris and Psilostrophe. It too is strongly supported as monophyletic by 12 shared characters and a bootstrap value of 88%.
Within the first clade, the separation of Hymenoxys texana from the other taxa is very strongly supported by a bootstrap value of 100%, and the separation of H. odorata is supported by a bootstrap value of 77%. In addition, separation of H. brachyactis is weakly supported by a bootstrap value of 62%, the two Plummera taxa are associated by a bootstrap value of 64%, and Dugaldia hoopesii 2 and Hymenoxys helenioides 2 are strongly united with a bootstrap value of 99%. There is also a weak subgroup of Hymenoxys rusbyi, H. subintegra, and H. cooperi 3, 4, and 5 supported by a bootstrap value of 49%.
In the second clade, 12 characters and a bootstrap value of 88% unite Psilostrophe and Tetraneuris. It should be noted, however, that Psilostrophe is separated from Tetraneuris by 87 characters, two more than separate the Tetraneuris clade and the clade containing Hymenoxys, Dugaldia, Macdougalia, and Plummera.
The monophyletic nature of Tetraneuris by itself is very strongly supported by the presence of 40 synapomorphies and a bootstrap value of 100%. Within Tetraneuris, five synapomorphies and a bootstrap value of 88% unite Tetraneuris ivesiana, T. scaposa var. scaposa 4, T. scaposa var. argyrocaulon, T. turneri, T. linearifolia var. linearifolia, and T. linearifolia var. arenicola. The T. linearifolia varieties are strongly associated by five synapomorphies and a bootstrap value of 100%, and T. scaposa var. argyrocaulon is associated with T. turneri by two synapomorphies and a bootstrap value of 77%. Among the other Tetraneuris taxa, there is a strong association of T. argentea with T. acaulis var. acaulis supported by a bootstrap value of 81%, and a strong association of T. scaposa var. scaposa 1, 2, and 3 (note that T. scaposa var. scaposa 4 is well separated) supported by a bootstrap value of 98%.
In several cases, multiple populations of the same taxon were identical with regard to DNA restriction site changes. Hymenoxys richardsonii var. floribunda 4 and 5 were identical to one another, as were T. turneri 1 and 4, T. linearifolia var. linearifolia 1, 3, and 4, and all four populations of T. linearifolia var. arenicola. More often, multiple populations of the same taxon exhibited differences in restriction sites. Dugaldia hoopesii 1 and 2 were different from one another, as were Hymenoxys helenioides 1 and 2, H. richardsonii var.floribunda 1, 2, 3, and 4/5, H. cooperi 1, 2, 3, 4, and 5, H. odorata 1 and 2, Tetraneuris ivesiana 1 and 2, T. scaposa var. scaposa 1, 2, 3, and 4, T. scaposa var. argyrocaulon 1, 2, and 3, T. turneri 1/4, 2, 3, and 5, T. linearifolia var. linearifolia 1/3/4 and 2, T. argentea 1 and 2, T. acaulis var. acaulis 1 and 2, and T. acaulis var. arizonica I, 2, and 3.
DNA restriction site variation strongly supports the notion that Dugaldia, Macdougalia, and Plummera are congeneric with Hymenoxys. In fact, these taxa align much more closely with the taxa of H. subgenus Picradenia than do H. odorata and H. texana. As a matter of consistency, if Dugaldia, Macdougalia, and Plummera were to be recognized as separate genera, it would be necessary to recognize Hymenoxys odorata and H. texana at the generic level. Taxonomically, Macdougalia has been included in Hymenoxys by most recent workers (e.g., Turner and Powell, 1977; Bierner, 1994), but submersion of Plummera in Hymenoxys has been suggested only by Turner et al. (1973), Turner and Powell (1977), and Bierner (1994), and only Bierner (1994) has submerged Dugaldia in Hymenoxys.
Conversely, the taxa of Tetraneuris are clearly separated from the other taxa, forming a monophyletic clade supported by the presence of 40 shared characters and a bootstrap value of 100%. In fact, Tetraneuris is separated from Hymenoxys, Dugaldia, Macdougalia, and Plummera by a total of 85 character changes and is separated from Psilostrophe by 87 character changes, degrees of separation that we believe merit recognition at the generic level.
Psilostrophe was chosen to serve as an outgroup and was not a focus of this study. However, its association with Tetraneuris in the DNA phylogenetic tree is noteworthy, because many workers, including Bentham (1873), Rydberg (1914), and Turner and Powell (1977) have placed Psilostrophe in a different subtribe from Hymenoxys sensu lato as discussed here. It is apparent from Fig. 1 that Psilostrophe is distinctly closer to Tetraneuris, with which it shares 12 synapomorphies, than to Hymenoxys; however, 87 characters separate Psilostrophe and Tetraneuris. Keeping in mind that this study was intended to examine relationships among Hymenoxys and its putative closest relatives, many taxa in subtribe Gaillardiinae were not included. The position of Psilostrophe will be much clearer, therefore, when we are able to examine other genera of subtribe Gaillardiinae such as Amblyolepis DC., Baileya Harv. & A. Gray, Balduina Nutt, (including Actinospermum Elliott), Gaillardia Foug. (including Agassizia A. Gray & Engelm. and Guentheria Spreng.), Helenium (including Actinea A. L. Juss., Cephalophora Cav., Hecubaea DC., Leptopoda Nutt., and Tetrodus Cass.), Marshallia Schreb., and Plateilema (A. Gray) Cockerell.
On morphologic grounds, Hymenoxys texana seems to belong within the genus Hymenoxys, and DNA restriction site data clearly associate it with Hymenoxys (33 synapomorphies with a bootstrap value of 100%). But, it is separated from the other taxa by 61 character changes (bootstrap value of 100%), and its relationship to any other taxon in this clade is unclear. Furthermore, different from Hymenoxys and similar to Tetraneuris, H. texana possesses monoterpene glycosides and lacks secopseudoguaianolides (Spring et al., 1994), and its very unusual chromosome numbers of n = 8 and 3 (Strother and Brown 1988) are unlike any reported from other taxa in either Hymenoxys or Tetraneuris (mainly n = 15 with some dysploidy and polyploidy; Bierner, 1994). Despite the conflicting data, it is our opinion that the 33 synapomorphies that H. texana shares with the Hymenoxys taxa is compelling evidence for maintaining it in Hymenoxys, but the 61 character changes by which it differs from the other taxa is compelling evidence for separating it into a different subgenus, Picradeniella Cockerell, as suggested by Cockerell (1904).
Recent work by Anderson et al. (1996) provides convincing evidence that Hymenoxys helenioides is a hybrid between Dugaldia hoopesii and H. richardsonii var. floribunda. In our study, D. hoopesii 2, H. helenioides 1 and 2, and H. richardsonii var. floribunda 1 were all collected at the same locality. Dugaldia hoopesii 2 from Utah differs from D. hoopesii 1 from New Mexico by four character changes, and yet H. helenioides 2 (unlike H. helenioides 1) is identical to D. hoopesii 2 with regard to cpDNA restriction sites. This suggests to us that an individual of D. hoopesii in the Utah population was likely the female parent of H. helenioides 2.
Another finding is the separation of Hymenoxys cooperi 3, 4, and 5 from H. cooperi 1 and 2. Populations 1 and 2 are from California and Nevada in the general vicinity of the type locality of H. cooperi (California, San Bernardino County, Providence Mountains). Populations 3, 4, and 5 are from Utah and Arizona in the general vicinity of the type locality of H. biennis (A. Gray) H. M. Hall (Arizona, Mohave County, Mokiak Pass). On morphologic grounds, the first author cannot find characters that consistently separate these populations and, therefore, has treated H. biennis as conspecific with H. cooperi (Bierner, unpublished). DNA restriction site data suggest that this decision should be reexamined.
Likewise, as noted in the results section, there is considerable restriction site variation among multiple populations of several other taxa. All of these taxa exhibit morphologic variation among their populations, often to the extent that the variants have been described as species or varieties. For example, Hymenoxys richardsonii var. floribunda has eight taxonomic synonyms (sensu Bierner, unpublished). Population 1 from this study is from Garfield County, Utah, and is perhaps referable to H. richardsonii subsp. macrantha (Nelson) Cockerell var. utahensis Cockerell, population 2 is from Coconino County, Arizona, and is perhaps referable to H. floribunda (A. Gray) Cockerell var. arizonica Cockerell or H. floribunda var. intermedia Cockerell, population 3 is from Sandia Crest in Bernalillo County, New Mexico, and is morphologically somewhat distinct from other populations (not described in the literature; Bierner, pers. obs.), and populations 4 and 5 (which were identical to one another with regard to DNA restriction site changes) are from Grant County, New Mexico, and are probably referable to H. metcalfei Cockerell. As with H. cooperi, therefore, DNA restriction site data suggest that populations of several taxa in this complex have diverged from one another to some extent and should be examined more closely to determine whether taxonomic recognition of any of these populations is warranted.
Within Tetraneuris, T. ivesiana, T. scaposa var. scaposa 4, T. scaposa var. argyrocaulon, T. turneri, T. linearifolia var. linearifolia, and T. linearifolia var. arenicola are strongly associated by DNA restriction site datA. Tetraneuris ivesiana is found in the four-corners area of northeast Arizona, southeast Utah, southwest Colorado, and northwest New Mexico (the populations in this study were collected in Utah), T. scaposa var. scaposa ranges from southern Nebraska south to Kansas, Colorado, Oklahoma, New Mexico, Texas, and northern Mexico, making a close approach to the range of T. ivesiana in northwest New Mexico (population 4 was collected in southcentral Texas west of Austin), T. linearifolia var. linearifolia ranges from northcentral Oklahoma south to south Texas and northern Mexico and west to west Texas and southeastern New Mexico (the populations in this study were collected in south Texas), and T. scaposa var. argyrocaulon, T. turneri, and T. linearifolia var. arenicola are all south Texas endemics. At first glance, therefore, it would seem unusual to relate T. ivesiana from the four-corners area with taxa restricted to south Texas; however, both T. scaposa var. scaposa and T. linearifolia var. linearifolia bridge the geographic gap. DNA restriction site data further group T. scaposa var. scaposa 4, T. scaposa var. argyrocaulon, and T. turneri, with T. scaposa var. argyrocaulon and T. turneri being strongly associated with one another. Also, the T. linearifolia varieties, consistent with their taxonomic treatment, form a very strongly supported clade.
Although Tetraneuris scaposa var. scaposa 4 is clearly related to T. ivesiana, T. scaposa var. argyrocaulon, T. turneri, T. linearifolia var. linearifolia, and T. linearifolia var. arenicola as described above, the T. scaposa var. scaposa 1, 2, and 3 populations are separated into a different clade and associated with one another with a bootstrap value of 98%. As mentioned above, population 4 was collected in southcentral Texas, but population 1 was collected in northcentral New Mexico (Torrence County), and populations 2 and 3 were collected in west Texas (Jeff Davis and Brewster counties). DNA restriction site data, therefore, indicate that considerable divergence has occurred among populations that we are recognizing in this study as T. scaposa var. scaposa, and a thorough examination of the T. scaposa complex should be undertaken.
Finally, Tetraneuris argentea and T. acaulis var. acaulis are monophyletic with a bootstrap value of 81%. The ranges of these taxa overlap in northern New Mexico, and although the former has stem leaves and the latter is scapose, both are characterized by dense appressed silky pubescence, which is not seen in the other taxa of Tetraneuris.
In summary, our DNA restriction site studies support the following conclusions:
Dugaldia, Macdougalia, and Plummera are congeneric with Hymenoxys.
Tetraneuris is a genus distinct from but related to Hymenoxys.
Psilostrophe belongs within subtribe Gaillardiinae and is more closely related to Tetraneuris than to Hymenoxys.
Hymenoxys texana resides within Hymenoxys, but it should be placed in its own subgenus, Picradeniella.
Hymenoxys helenioides appears to be a hybrid between Hymenoxys richardsonii var. floribunda and Dugaldia hoopesii (= Hymenoxys hoopesii [A. Gray] Bierner).
Populations referable to Hymenoxys biennis may be distinct from H. cooperi.
Populations of several taxa in this complex (e.g., Hymenoxys richardsonii var. floribunda sensu Bierner) have diverged from one another to some extent and may warrant taxonomic recognition.
Tetraneuris ivesiana, T. scaposa var. scaposa (from central Texas), T. scaposa var. argyrocaulon, T. turneri, T. linearifolia var. linearifolia, and T. linearifolia var. arenicola form a phylogenetically related subgroup, within which T. scaposa var. scaposa (from central Texas), T. scaposa var. argyrocaulon, and T. turneri are further associated.
Tetraneuris linearifolia var. linearifolia and T. linearifolia var. arenicola are a closely related varietal pair.
Populations recognized in this study as Hymenoxys scaposa var. scaposa have diverged from one another.
Tetraneuris argentea and T. acaulis var. acaulis are phylogenetically closely related.
This work was supported by a National Science Foundation Research Opportunity Award supplement to grant BSR-9200707 (Robert K. Jansen). We thank Drs. Ki-Joong Kim and Billie L. Turner for providing several collections and DNA samples for this study and Dr. John Bain for helpful comments on an earlier version of the manuscript.