POLYPLOIDY: A FACTOR IN THE EVOLUTION OF DICERANDRA BENTH. (LABIATAE)

Chromosome counts are reported for the first time in Dicerandra Benth. (Labiatae), a genus endemic to the south-eastern United States. Representative taxa are tetraploid (« = 16) and hexaploid (n = 24). Diploids were not found. The systematic, phytogeographic and conservation implications of polyploidy in this genus are discussed. The monophyletic nature of the genus is upheld by the discovery that both sections have a presumed base number of x = 8. Hexaploidy has apparently evolved independently in all branches of sect. Dicerandra, but not in sect. Lecontea.

genus have different pollination syndromes. Exclusive of the anthers, the floral morphology of each section is so distinct as almost to be considered in the informal super-groups of the Labiatae, Ocimoid-grex and Stachyoid-grex of Hedge (1992)! Section Lecontea has a tubular corolla, inserted stamens and style under the upper lip, and sect. Dicerandra has an infundibular corolla, and exerted stamens and style, often declinate along the lower lip (Huck, 1987). In artificial crossing experiments to produce Fl hybrids, annual species were remarkably interfertile when crossed with other annuals and perennials, whereas crosses between perennials were unsuccessful (Huck, 1987). Natural hybrids between annuals are rare, however, and pollen viability counts of individuals in those naturally occurring hybrid swarms was near 90% (Huck, 1987). Myxocarpy, a situation in which hydrophilic cells, known as slime cells, cover the pericarp wall, characterizes Dicerandra (Huck, 1987) as well as the other taxa of subfamily Nepetoideae (Cantino & Sanders, 1986). Annual species are located on xeric ridges or on sandy, disturbed areas, usually beside functioning water-courses, in Georgia, Alabama, South Carolina and northern Florida; perennial species are found on isolated dune refugia in central Florida below 29°N latitude (Fig. 1). Habitat disturbance is a factor in the ecology of the genus.

MATERIAL AND METHODS
Buds for chromosome counts were collected in autumn 1994 throughout the range of the genus by the two authors. Pre-flowering buds were fixed in the field in 6 chloroform: 3 ethanol 95%: 1 glacial acetic acid for 48 hours, rinsed in 70% ethanol, and then transferred to 70% ethanol for storage in a freezer. A random collection of 25-30 buds was made from approximately 25 plants at each population site, except for D. thinicola, where each bud collection represents a single plant. Plants of this latter species were found in a large population with much variation in flowering time and flower colour. The buds were stained in Snow's alcoholic-hydrochloric acid-carmine (Snow, 1963) for 4-7 days at 50-60°C. Single flowers were placed on a slide in a drop of 45% acetic acid and the calyx was removed. A drop of Hoyer's mounting medium was mixed with the acetic acid and the anthers were excised, cut in half, and spaced before the coverslip was placed. The slide was heated on a warming tray (63°C) for approximately 3-4 minutes and the coverslip was then pressed to spread the pollen mother cells. When cells with countable chromosomes were observed, they were sketched and, when suitable, camera lucida drawings were made. Slides are quite permanent and will be available for reference as this chromosome study continues. One count is reported for each taxon, except for two counts for D. frutescens and D. odoratissima from buds from different populations, and three counts from D. thinicola from buds from different plants collected in the same population. Voucher specimens have been deposited at FLAS.

RESULTS
Fifteen chromosome counts for 11 taxa of Dicerandra are reported (Table 1) and representative cells are illustrated in camera lucida drawings (Figs 2-12). Of the eleven taxa studied seven have n = 16 and four have rc = 24 (Table 1). These numbers suggest x = 8. However, if this is the case, there are no known diploid species and the chromosome levels represented by the above are tetraploid and hexaploid respectively (Table 1). A base chromosome number of 8 is not unusual in the Labiatae. Other genera in the subfamily Nepetoideae with chromosomes that suggest this base number are Elsholtzia Willd. and Salvia L. (Goldblatt & Johnson, 1994), Satureja L. and Melissa L. (Federov, 1969) and Monarch. L. (Irving, 1976). Ranging from 1-4, B chromosomes were observed in most taxa (Table 1; Figs 2-12), including 2-4 in D. thinicola; they were visible at diakinesis but not at later stages, and thus they may occur in other cells.

D I S C U S S I O N
The phytogeographic pattern of Dicerandra with hexaploid taxa sympatric or parapatric with tetraploids, missing diploids and isolated populations of perennial species coincident with major ridge systems in Florida (Fig. 1) implies a complicated evolutionary history for this genus. The successive fall and rise of sea level during the Pleistocene (Cooke, 1939;Neill, 1957;White, 1970) presumably opened up disturbed habitats for colonizing genera such as Dicerandra. The buoyant mucilaginous fruit (myxocarpy) of this genus was perhaps an important factor in the movement of propagules along ancient waterways in the south-eastern United States. Deposition and isolation on dune 'islands' presumably set the stage for genetic divergence and speciation. Isolated North American genera such as Conradina A. Gray, Stachydeoma Small, Acanthomintha A. Gray, Monardella Benth. and Pogogyne Benth., as well as other isolated genera discussed by Hedge (1992), are common in the Labiatae. Extinct diploids of Dicerandra probably occupied habitats no longer present today, such as those habitats discussed by Watts (1983). Missing diploids in genera such as Hyptis (Harley & Hey wood, 1992) are also well known in the Labiatae.
Polyploidy can be viewed as a stabilizing force in the evolution of taxa in Dicerandra. Tetraploids («=16) in Dicerandra are presumed to be allopolyploids, formed by a doubling of base chromosome number (n = 8) after hybridization events (see Stebbins, 1971;De Wet, 1975). Dicerandra may be an example of secondary  hybrid polyploidy discussed by Stebbins (1971), in which hybrid tetraploids (« = 16) backcross to ancestral diploids (n = 8). Such a hypothetical cross in Dicerandra would yield triploid hybrids (« = 12), which, if doubled, could produce hexaploids (« = 24) with good chromosomal pairing and normal meiosis. Presence of B chromosomes in a majority of Dicerandra species is consistent with their known occurrence in the Labiatae (Jones, 1995). Although their significance is not completely understood, B chromosomes may imply continuing evolution in the genus as suggested by Harley and Heywood (1992) for subgenus Calophace in Salvia.
Our cytological study aids in understanding relationships in the genus and highlights some taxa for review and discussion. Previously, Krai (1982) has suggested that the section now delineated as sect. Lecontea is so different from sect. Dicerandra that it could be treated as a new genus. However, chromosome numbers of «=16 and a presumed base number of x = 8 are found in both sections of Dicerandra (Table 1). Continued inclusion, at least for the time being, of sect. Lecontea within the genus Dicerandra is supported by these studies.
Dicerandra odoratissima Harper and D. radfordiana Huck, both annuals with chromosome counts of n= 16 in sect. Lecontea, are found in a small area of Georgia and South Carolina ( Fig. 1; Table 1). Although it was hypothesized that D. radfordiana might have a higher chromosome number since its corolla dimensions, fruit diameter and pollen size are approximately twice the size of D. odoratissima (Huck, 1987), this is not the case. Discontinuities in flavonoid profiles between the two species have been found (Huck, unpublished data). Barely 100 plants of D. radfordiana remain in the wild.
Dicerandra linearifolia (Ell.) Benth. var. linearifolia, an annual with pale pink to white corollas, yellow anthers and narrow leaves, is found in northern Georgia and north-west Florida (Fig. 1). Chromosome counts of the type variety are « = 24, a hexaploid ( Table 1). The type for the species and for the genus as collected by Elliott in 1821 presumably, then, was a hexaploid. The variety D. linearifolia (Ell.) Benth. var. robustior Huck, with dark purple corollas, reddish brown anthers and wider leaves found in northern Florida, is a tetraploid, with counts of «=16 (Table 1; Fig. 1). Differences in morphology of floral features, geography and, now, chromosome number, suggest that these two varieties may be better addressed as subspecies of Dicerandra linearifolia. Further chromosome counts of this species throughout its range are needed.
The distinctive Dicerandra densiflora Benth., with small lavender corollas, a short tube and anthers with brush-like spurs, is a hexaploid (« = 24; Table 1). This annual is found in the Suwannee River Basin between northern annuals and southern perennial taxa of the genus (Fig. 1).
In 1962 L.H. Shinners described the first perennial in the genus, Dicerandra frutescens Shinners. Distribution of this species as recognized by Shinners was from north to south along the central ridges in Florida and terminated at the Lake Wales Ridge (sensu White, 1970), a distance of some 180 miles (257km). As it is now circumscribed, D. frutescens occurs only on the western flank of the Lake Wales Ridge (Fig. 1). This species is a hexaploid (« = 24; Table 1), has a white corolla with purple marks on the standard, flavonoids that vibrantly fluoresce under UV light (Huck, unpublished data) and a unique antifeedant chemical (Eisner et al., 1990).
Since 1962 two other Dicerandra species, both tetraploids (n=16), have been discovered within the original range for Dicerandra frutescens as delineated by Shinners (Table 1; Fig. 1). In 1981 Dicerandra cornutissima Huck occurring in north central Florida was recognized. This species has a mat of basal leaves, reddish purple corollas and long, slender spurs on the white anthers. In 1989, yet another species was recognized, Dicerandra christmanii Huck & Judd, with an oil profile of 1,8-cineole and oc-terpineol unlike the pulegone found in other perennial Dicerandra species (Huck et al., 1989). This latter species with pale yellow anthers and dangling corollas was found only 7 miles (11.2km) from D. frutescens. Now it appears that there is yet another undescribed taxon, an apparent new subspecies of D. frutescens, on the eastern flank of the Lake Wales Ridge, a tetraploid, at n= 16 ( Fig. 1; Table 1). This latter taxon resembles hexaploid D. frutescens, and the corolla undergoes two colour phases: yellow, in anther-dehiscent stage, and white fading to dark pink, in the stigma-receptive stage. The spurs are longer than D. frutescens, and the taxon has somewhat wider leaves. Further research is needed to clarify this undescribed taxon.
The Atlantic Coastal Ridge in Florida, another site of great endemism (Austin et al., 1987), runs parallel to the species-rich Lake Wales Ridge (Fig. 1), but is of more recent origin than the interior ridge (Tanner, 1992). On the southern part of the Atlantic Coastal Ridge Dicerandra immaculata Lakela, a hexaploid, with n = 24 ( Fig. 1; Table 1) is found. This species has a pale pink corolla, an immaculate standard and intensely hairy spurs on the anthers. Dicerandra thinicola H.A. Miller, a recently named species just to the north on the same ridge, is a tetraploid, with n = 16 (Table 1). This latter species can be distinguished by a wide corolla throat in which the palate holding the stigma is clearly visible during anthesis. Considerable variability in floral colour, and an autumn blooming period concurrent with northern species, are apparent. Some pollen inviability and corolla shape differences characterize this species, but chromosome numbers are constant ( Table 1).
Addition of new character information on chromosome numbers requires a refinement of earlier phylogenetic analyses presented by Huck (1987) and Huck and co-workers (1989) for Dicerandra. The monophyletic nature of the genus which includes sect. Dicerandra and sect. Lecontea as proposed in those analyses is upheld by the discovery that the base number for both sections is presumably JC = 8. The hypothetical ancestors of these two clades are possibly the missing diploids. Hexaploidy has evolved independently in both perennial and annual groups in all four branches of sect. Dicerandra but not in sect. Lecontea. Since polyploid differences in hexaploids and tetraploids may be used to imply relative ages of the taxa (Kruckeberg & Rabinowitz, 1985), possible speculative relationships may be suggested in Dicerandra: hexaploid D. linearifolia var. linearifolia younger than tetraploid D. linearifolia var. robustior and hexaploid D. frutescens in Highlands County in Florida younger than a new tetraploid subspecies in Polk County ( Fig. 1; Table 1). The Lake Wales Ridge group including D. frutescens, D. christmanii and a suspected new subspecies (or species, pending results of further research) of D. frutescens appears to form an evolutionary unit; populations with their respective ploidy levels should be represented in conservation programmes.
A speculative phytogeographic scenario for the genus might include an origin of the diploid Dicerandra in the Appalachian Province, with movement into the Coastal Plain, as has occurred with other species (Thome, 1993). A re-radiation of taxa from Apalachicola refugia to the north and south in the late Pleistocene is possible. This general pattern might have included the early separation of the annuals and perennials in sect. Dicerandra with hexaploid D. densiflora more recently evolved and occupying the ancient area known by biogeographers such as Webb (1990) as the Suwannee Strait in Florida. In areas of sympatry, possibly of recent origin, rare hybrids occur between annuals hexaploid D. densiflora and tetraploid D. linearifolia var. robustior and between hexaploid D. linearifolia var. linearifolia and tetraploid D. odoratissima (Fig. 1). An early radiation centre is now apparent in the Lake Wales Ridge.

C O N C L U S I O N S
The pattern of missing diploids and extant hexaploids and tetraploids with many isolated endemics suggests that Dicerandra is an old polyploid complex. Our interpretation of the cytological data further implies that development of tetraploids and hexaploids occurred when diploids were present early in the history of Dicerandra, possibly in response to a radically changing environment. While polyploidy certainly plays a role in isolation of species, it is seen as a stabilizing factor following hybridization, an apparent theme in outcrossing Dicerandra. Further, the tremendous variability carried in polyploid taxa masks the expression of recessive alleles (Briggs & Walters, 1984) and serves as a genetic reservoir for the continuing vitality of small populations in genera such as Dicerandra. Further research will include a chromosome count survey of the widespread D. linearifolia, as well as counts of any newly discovered populations of Dicerandra.