Morphological Stasis abd Molecular Divergence in the Intercontinental Disjunct Genus Datisca (Datiscaceae)

ABSTRACf The genus Datisca comprises two species and has an intercontinentally disjunct distribution: D. cannabina is native to southwest and central Asia, whereas D. g/omerata is distributed from northern California to northern Baja California. In 1975, Axelrod proposed a geohistorical scenario to account for such "Madrean-Tethyan links," suggesting that these disjunctions resulted from migration across the mid-Atlantic from the Paleogene up to the Neogene, approximately 23 to 65 m.y.a. The two species are quite similar in most phenotypic traits which have been studied to date. The major difference between the two involves their breeding system: D. cannabina is dioecious while D. g/omerata is apparently androdioecious. Despite these similarities, Nei's mean genetic identity between the two species is I = 0.142. This is one of the lowest values yet reported for congeneric flowering plants and provides evidence for an ancient origin of the disjunction. Furthermore, the fact that the western populations of D. cannabina have a much higher genetic identity value with D. glomerata than does the eastern population supports the idea that dispersal occurred across the Atlantic. In addition, the population genetic structure of D. glomerata is consistent with an androdioecious breeding system.


INTRODUCTION
One of the primary goals of biogeography is the interpretation of disjunct distributions (Brown and Gibson 1983). To date, most biogeographical studies of disjuncts have relied on morphological comparisons. The results of such studies, however, may be ambiguous, since rates of morphological change are not readily correlated· with time of divergence. Comparative molecular studies provide new data sets which can contribute to our understanding of biogeographical problems.
For example, molecular data has the potential to resolve the most basic question in the analysis of a particular disjunct distribution (Carlquist 1983): is it of recent origin (presumably through long distance dispersal); or is it attributable to geohistorical processes? A low degree of genetic divergence will favor the first alternative, whereas a high degree will favor the second.
Disjuncts can also serve as a model for evolutionary biologists who are interested in comparative rates of evolution. All discrete populations are, by definition, disjunctly distributed (Brown and Gibson 1983). Intercontinental disjunctions simply present the most extreme manifestation of this. As such they allow us to observe evolution in the absence of gene flow, and thereby provide an ideal situation for the study of evolutionary rates.
Molecular data have been successfully applied to biogeographical questions in animals such as the Old and New World monkeys (Sarich and Cronin 1980), Pacific island-dwelling mosquitoes (Pashley, Rai, and Pashley 1985), and within the Drosophila melanogaster species complex (Lachaise, Cariou, David, Lemeunier, Tsacas, and Ash burner 1988). In plants, isozyme electrophoresis has been used extensively to quantify genetic variation; but only limited use has been made of these data in examining intercontinental disjunctions (V ogelmann and Gastony 1987;Parks and Shan-An 1988;Hoey and Parks 1988;Liston, Rieseberg, and Elias 1989). This is unfortunate, for these data make it possible to evaluate different biogeographical models and to characterize the types and evolution of the genetic systems involved in these 0 411 1000 km disjunctions. Here we use electrophoretic evidence to examine an intercontinental disjunct in the genus Datisca (Datiscaceae).
Datisca is a genus of two species. Datisca glomerata (Presl.) Baill. is distributed from Siskiyou County in northern California to the Sierra San Pedro Martir in northern Baja California, Mexico (Fig. 1). Datisca cannabina L. is distributed from the island of Crete in the east Mediterranean to the foothills of the Himalayas in northwestern India ( Fig. 2) but is absent from an approximately 1 000-km stretch of desert from east of the Caspian Sea to northwestern Afghanistan (Rechinger 1966).
Both species are long-lived, tall perennial herbs of riparian habitats and are morphologically and anatomically quite similar (Davidson 1973(Davidson , 1976Boesewinkel 1984). Bohm (1988) has recently shown that the two species have identical flavonoid profiles. In addition, both species possess root nodules containing the nitrogen-fixing microsymbiont Frankia (Chaudhary 1979;Akkermans, Roelofsen, Blom, Huss-Danell, and Harkink 1983) and are apparently diploids with a gametic chromosome number n = 11 (Sinoto 1929;Snow 1959). Both taxa also show a high degree of intraspecific uniformity. For example, no intraspecific flavonoid variation was found among 12 populations of D. glomerata from throughout California (Bohm 1987).
The only apparent difference between the two species is that D. cannabina is strictly dioecious, while D. glomerata is apparently androdioecious, possessing hermaphrodite and male individuals. Mature leaves of D. cannabina are pinnatisect, while those of D. glomerata are incised-serrate, sometimes deeply divided near the base but never to the midrib. However, seedlings and young leaves of the two species are virtually indistinguishable. Likewise, plants of D. cannabina grown in reduced light produce glomerata-like leaves (Fig. 3). Thome (1972) has summarized the major disjunct ranges of vascular plants. He classifies Datisca and approximately 35 other genera with similar disjunctions as Mediterranean-American disjuncts. Axelrod (1975) refers to these disjuncts as Madrean-Tethyan links and has presented a vicariance hypothesis explaining their present distribution in the two regions. Summarizing the fossil data, he concludes that during the Paleogene a sclerophyllous flora adapted to subhumid climate inhabited lower-middle latitudes along the shores of the Tethyan region and across southern North America (Axelrod 197 5, fig. 1 ). Extant links are relicts of effective migration across the middle Atlantic which was then narrower (Phillips and Forsyth 1972;Dietz and Holden 1970). At the beginning of the Oligocene, sea level fell, and it is probable that discontinuous parts of the Mid-Atlantic rise were emergent (Tarling 1980). Volcanic islands along the Mid-Atlantic Rise and its flanks could have provided additional "stepping stones" facilitating dispersal (Axelrod 1975;Tarling 1980). In addition, the lower latitude ofthe east coast of North America and the ENE orientation of the Appalachian axis (Phillips and Forsyth 1972;Walper and Rowett 1972) would have provided numerous xeric sites for migration, particularly on granite-gneiss domes, "shale barrens," and other igneous substrates found in the Piedmont (Axelrod 1975).
In the present study, we have attempted to answer the following questions for the intercontinental disjunct species pair of Datisca cannabina and D. glomerata: 1) does the amount of genetic divergence conform with hypotheses concerning the age of the disjunction? 2) what is the relationship between the genetic and morphological divergence of the two species? and 3) is the population genetic structure of the two species consistent with their putative breeding systems?

Plants
Twenty-three populations of Datisca glomerata encompassing the entire geographic range of the species were sampled for electrophoretically detectable genetic variation (Table 1, Fig. 1). Twenty-one of the populations were sampled by field collections of leaves from individual plants. Since most populations were very small (1-30 individuals) all plants could be examined. Two larger populations (DG 1 and DG 3) of over 100 individuals were sampled at random. Two additional populations were grown up from bulk seed collections (DG 18 and DG 19). Leaves were either used immediately or stored at -70 C for later analysis.
Three collections of D. cannabina were available for study (Table 1, Fig. 2). All were derived from bulk seed collections. Although the sampling of populations was not as complete as for D. glomerata, the sites of D. cannabina represented the eastern, central and western parts of the species' range.
Cytosolic and plastid isozymes were distinguished for TPI by the enrichment of intact chloroplasts following the methodology of Mills and Joy (1980). The enriched chloroplast pellet was lysed in one drop of extraction buffer, absorbed into a paper wick, and used for enzyme electrophoresis.
Genotype frequencies were inferred directly from isozyme phenotypes based on the knowledge of the generally conserved enzyme substructure, compartmentalization, and isozyme number in higher plants (Gottlieb 1981(Gottlieb , 1982. The lack of excessive variation within populations also aided in the genetic interpretation of isozyme patterns.
If more than one locus was present for an enzyme, loci were numbered sequentially with the most anodally migrating locus designated 1. Likewise, enzyme variants at individual loci were given letters with the fastest allozyme designated a.

Data Analysis
Nei's (1972) mean genetic distance (D) and identity (/) were computed for all pairwise population comparisons both within and between species using the GENEST AT program (Lewis and Whitkus 1988). The proportion of polymorphic loci, mean number of alleles per locus, and mean heterozygosity were also computed for each taxon.
F, the inbreeding coefficient or fixation index (Wright 1965(Wright , 1969, which measures deviation of heterozygote proportions from Hardy-Weinberg expectations, was determined for all polymorphic loci in each population and for both species. F can range from -1.0 to 1.0, with the former indicating an excess and the latter a deficiency of heterozygotes, respectively, when compared to Hardy-Weinberg expectations. Significance of deviations at each locus was tested by the x 2 test against Hardy-Weinberg expectations. The distribution of genetic variation within and among populations ofthe two species was analyzed using gene diversity statistics (Nei 1975), as calculated by GENEST AT (Lewis and Whitkus 1988). Total gene diversity (HT) is subdivided into gene diversity within populations (Hs) and gene diversity among populations (DsT) where Differentiation among populations is calculated as where GsT can vary between 0 (Hs = HT) and 1 (Hs = 0), i.e., populations are fixed for different alleles.
Quantitative estimates of interpopulational gene flow were calculated using the model of Slatkin (1985). This method provides an indirect estimate of Nm, the amount of interpopulational gene flow, using a linear relationship between the logarithm of Nm and the logarithm of the average frequency of alleles restricted to a single population (private alleles). This relationship is given as where ft(l) is the average frequency of private alleles, N is the effective population size, m is the migration rate among populations, a= -0.505, and b = -2.440.
Estimates of Nm less than 1.0 indicate relatively little gene flow, while Nm values of 1.0 or greater suggest high levels of gene flow (Slatkin 1985).
The loci Sod and Tpi-1 were not expressed in young tissue, and therefore were not scored in populations DC2 and DC3 of Datisca cannabina. Both PGD loci and Pgm-2 were inconsistently expressed in frozen material and thus data are missing at these loci for some D. glomerata populations.

Interspecific Allozyme Variation
Allele frequencies at the 21 interpreted loci were totalled for each species (Table  2). Allele frequencies for each population are available from the authors upon request. No locus was monomorphic over all populations of both species. Nei's mean genetic identity (Nei 1972) between D. cannabina and D. glomerata was 0.142 (D = 1.96). This is one ofthe lowest values recorded to date for congeneric plant species. All populations of D. glomerata had a higher genetic identity value with the two western populations of D. cannabina (DC2 and DC3) than with the central Asian population (Table 3).
Three populations of D. glomerata (DG8, DG14 and DG23) had no alleles in common with the central Asian population of D. cannabina (DC 1) and a genetic identity of zero (Table 3). Only three alleles  were shared between interspecific populations. Nineteen populations of D. glomerata possessed one or two of these alleles. One population (DG 15) shared all three alleles with D. cannabina.

Intraspecific Allozyme Variation
The mean genetic identity for intraspecific population comparisons within D.
glomerata was I = 0.84 7. Genetic identity values ranged from 0.617 to 1.000 (Table 3). Populations in close geographic proximity had the highest genetic identities, but some widely separated populations were also quite close. For example, population DG 3 from Mendocino Co. had high genetic identity values with DG 22 and DG23 from Baja California (I= 0.937 and 0.935, respectively). The genetically most divergent populations within D. glomerata were DG9 and DG10. Among the three populations of D. cannabina sampled, the two western populations from the island ofLesvos (DC3) and the Transcaucasus mountain range (DC2) shared a high genetic identity value(/= 0.893). Both of these populations had a very low genetic identity value with the Central Asian population DCl (Table 3).
Genetic variability within the two species of Datisca was characterized by several standard methods (Table 4). The mean percentage of polymorphic loci However, in D. cannabina only population DC2 had any polymorphic loci (P = 21.0%). The mean number of alleles per locus (A) was also similar in both species (1.05 in D. glomerata and 1.06 in D. cannabina). No more than two alleles per population were found at any polymorphic locus.
Consistent with the low levels of allelic variation in Datisca, values of observed (H 0 ) and expected (He) heterozygosity were also low (Table 4) in both species. The total value of F. the fixation index (Wright 1969), for polymorphic loci in D. glomerata reflected a highly significant heterozygote deficiency (Table 4). The comparable value of F for D. cannabina reflected a frequency of heterozygotes not significantly different from Hardy-Weinberg expectations.
The apportionment of genetic diversity as measured with the gene diversity statistics of Nei (197 5) is similar in D. cannabina and D. glomerata (Table 4). Both species show low levels of total genetic diversity (HT), and most genetic diversity is found among populations (DsT) rather than within populations (H 8 ).

Estimates of Gene Flow
Only two private alleles, Mdh-2d and Pgm-2e were restricted to single populations of D. glomerata. Interpopulational gene flow in this species was estimated as Nm = 6.4, when adjusted for sample size (Slatkin 1985), indicating extensive gene flow among populations. Twelve private alleles were found in the three populations of D. cannabina and ten of these were fixed in the population. Thus a value of p(l) = 0.845 is obtained. For this large a value of p(1) the relationship between ln[p(l)] and ln(Nm) is no longer linear and a quantitative estimate of gene flow could not be made (Slatkin 1985). The same is true for the two western populations (DC2 and DC3) where p(l) = 0.535. These results indicate that the three populations of D. cannabina sampled are effectively isolated from each other.

Molecular Divergence
The interspecific mean genetic identity value of I= 0.142 for the two species of the genus Datisca is lower than any of the comparable values summarized in reviews of isozyme data in flowering plants (Gottlieb 1981;Crawford 1983).
However, a lower value (/ = 0.087) has been recently reported for congeneric species of Ceratophyllum (Les 1989). Three populations of D. glomerata have no alleles in common with the central Asian population of D. cannabina. These populations have diverged to such an extent that genetic comparison using allozyme data is no longer feasible (D = oo). The limit of allozyme resolution of genetic divergence usually corresponds to the genus level in animals (A vise 197 5). Thorpe (1982) has summarized intergeneric genetic identity values for 160 populations in animals (primarily vertebrates) and has found a mean /value of0.273 ± 0.107. In plants, intergeneric divergence is sometimes resolvable using allozymes, for example in the Asteraceae (Witter and Carr 1988) and Saxifragaceae (Soltis and Soltis 1986;Riese berg and Soltis 1987). Thus the two species of Datisca v.  Table 4. Percentage of polymorphic loci (P), mean number of alleles per locus (A), mean observed heterozygosity (Ha), mean expected heterozygosity (He), fixation index (F), and gene diversity statistics (HT,H 5 ,DsT,and GsT)  have diverged at genes coding for soluble enzymes as much as typical animal genera and more than some plant genera.
No other plant species pairs with disjunct intercontinental ranges possess genetic identities as low as that found in Datisca. Published values of I range from 0.40 to 0.43 in Liriodendron (Parks and Shan-An 1988), 0.46 in Liquidambar (Hoey and Parks 1988) and from 0.51 to 0.75 in Agastache (Vogelmann and Gastony 1987).
There are two basic processes by which divergence can occur between populations at isozyme loci: 1) changes in frequency can occur between populations by drift and other processes or 2) new alleles can develop via mutation. The latter is a slower process and becomes important in more distantly related taxa (Haverstock, Cole, Richardson, and Watts 1979). In Datisca unique alleles occur at nearly every locus in each species, indicating that mutations most likely account for the divergence (D. Crawford, pers. comm.). This represents strong evidence for an ancient divergence of the taxa.
The time since the divergence of two species can be roughly estimated using genetic distance values derived from allozyme data (Nei 1987;Sarich 1977;Thorpe 1982). However, when the time since divergence is large, the relationship between D and time is no longer linear (Nei 1987). Furthermore, when Dis greater than or equal to one (as in Datisca) its variance becomes very large and the reliability of the dating declines even more (Nei 1987). Nei (1975Nei ( , 1987 calculates that a D value of 1 is equivalent to about five million years. Other estimates of time, given the same value of D, are three to four times larger (Sarich 1977;Thorpe 1982). Using these minimum and maximum values, we can estimate that 10 to 40 million years have passed since the two species of Datisca began to diverge.
Despite the high degree of uncertainty inherent in these estimates, the postulated age of Madrean-Tethyan disjuncts (Axelrod 1975) partly overlaps these broad limits. Axelrod considers that migration across the mid-Atlantic was possible from the Paleogene up to the Neogene, approximately 23 to 65 m.y.a. Additional evidence in favor of the migration of Datisca across the Atlantic is found in the higher genetic identity between D. glomerata and the two western populations of D. cannabina than with the central Asian population.

Morphological Similarity
The phenotypic similarity of the two species is well documented (Davidson 1973(Davidson , 1976Boesewinkel 1984;Bohm 1988). Nevertheless, detailed morpholog-ical studies may reveal the presence of additional characters distinguishing the two taxa. The only consistent morphological distinction between the two species is the presence of hermaphrodite individuals in D. glomerata and female individuals in D. cannabina. Both species have male individuals.
In plants, the best known case of extremely high genetic divergence with high morphological similarity involves the sibling species Clarkiafranciscana and C. rubicunda (Gottlieb 1973). This divergence has been attributed to the prezygotic karyologic isolation of the two species in conjunction with the likely history of sharp fluctuations in population size occurring during the evolutionary history of C. franciscana and subsequent founder events (Gottlieb 1981 ). Further research has, however, revealed additional morphological differences between the two species (Gottlieb 1981).
Comparable examples of morphological similarity accompanied by molecular divergence are best known in animals. Most cases involve sibling species isolated by strong ecological and/or behavioral differences (Ayala, Tracey, Hedgecock, and Richmond 1974;Johnson and Selander 1971;Nanney 1982;Nixon and Taylor 1977;Webster, Selander, and Yang 1972). The initial isolating mechanism may also be major karyotypic differences as in Drosophila (Ayala et al. 1974) and bats (Greenbaum and Baker 1976). Extensive molecular divergence can also reveal that morphological similarity is in fact due to convergence (Maxson and Wilson 197 4). The example which seems most comparable to Datisca has been reported from sea urchins which are allopatric yet continue to occupy similar environments (Lessios 1981).

Evolutionary Stasis
How can we account for the pattern of morphological stability over long periods of evolutionary time as reported here in Datisca? Evolutionary stasis has been considered to result from either external stabilizing selection or inherent genotypic constraints (Williamson 1987). Van Valen (1982) has discussed additional possibilities, but he himself dismisses these. Although Williamson (1987) considers that both stabilizing and constraining factors act in concert to maintain species integrity, our data suggest that strong stabilizing selection is the cause of phenotypic stability in Datisca, for if genetic constraint were involved we would not expect the high levels of genetic divergence. In this case we need to know how selective forces have remained so constant over evolutionary time (Charlesworth, Lande, and Slatkin 1982).

Genetic Structure
Populations of Datisca g/omerata are usually of small size, show extreme heterozygote deficiency (F = 0.617; P < 0.001), and little intrapopulational genetic variation. These traits characterize the species as predominantly inbreeding (Hamrick, Linhart, and Mitton 1979;Loveless and Hamrick 1984). The high level of inbreeding suggests that D. g/omerata is truly androdioecious, and not functionally dioecious, as has been found in other cases of morphological androdioecy (Charlesworth 1984).
In apparent contradiction, the value of Nm = 6.4 suggests a high degree of interpopulational gene flow (Slatkin 1985). However, Slatkin's model assumes that populations are in genetic equilibrium, and most likely this condition is not met in D. glomerata. For example, the populations examined could be remnants of a once wider distribution in California. The present populations simply contain different subsets of the ancestral gene pool. Interesting in this context is the fact that the two most genetically divergent populations, isolated in Owens Valley and Death Valley, do not possess any private alleles. In isolation, these populations have not evolved new alleles, but rather have only preserved a rare subset of the genetic variation present in the species.
In contrast to D. glomerata, the populations of D. cannabina examined do not show deviations from random mating (F = -0.005) and more intrapopulational variation. These findings are in agreement with the fact that this species is dioecious and thus an obligate outcrosser. However we do have effective isolation between the populations sampled as demonstrated by gene flow estimates. Again, historical factors can be invoked to explain this discrepancy. The 1000 kilometers of desert between the eastern and western populations of D. cannabina must act as a very effective barrier to gene flow, which, considering the degree of genetic differentiation, has been maintained over long periods of evolutionary time. Interestingly, the two western populations, which are separated by 1500 kilometers, show a lesser degree of genetic divergence, comparable to the level found among populations of D. glomerata. The northernmost and southernmost populations of glomerata are also separated by approximately 1500 kilometers. In each of these cases, apparently no phenotypic differentiation has occurred, mirroring the process described for the intercontinental disjunct congeners.