Genetic Diversity and Reproductive Biology in Warea Carteri (Brassicaceae), a Narrowly Endemic Florida Scrub Annual Author(S): Margaret E

Home , Autogamy, Warea (plant), Warea carteri

Genetic Diversity and Reproductive Biology in Warea carteri (Brassicaceae), a Narrowly Endemic Florida Scrub Annual Author(s): Margaret E. K. Evans, Rebecca W. Dolan, Eric S. Menges, Doria R. Gordon Source: American Journal of Botany, Vol. 87, No. 3 (Mar., 2000), pp. 372-381 Published by: Botanical Society of America Stable URL: http://www.jstor.org/stable/2656633 . Accessed: 22/10/2011 09:41

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Botanical Society of America is collaborating with JSTOR to digitize, preserve and extend access to American Journal of Botany.

http://www.jstor.org AmericanJournal of Botany 87(3): 372-381. 2000.

GENETIC DIVERSITY AND REPRODUCTIVE BIOLOGY IN WAREA CARTERI (BRASSICACEAE), A NARROWLY ENDEMIC FLORIDA SCRUB ANNUAL1

MARGARET E. K. EVANS,25 REBECCA W. DOLAN,3 ERIC S. MENGES,2 AND DORIA R. GORDON4

2ArchboldBiological Station,PO. Box 2057, Lake Placid, Florida 33862 USA; 3FriesnerHerbarium, Butler University,Indianapolis, Indiana 46208 USA; and 4The Nature Conservancy,Department of Botany,PO. Box 118526, Universityof Florida, Gainesville, Florida 32611 USA

Carter's mustard(Warea carteri) is an endangered,fire-stimulated annual endemic of the Lake Wales Ridge, Florida, USA. This species is characterizedby seed banks and large fluctuationsin plant numbers,with increases occurringin postdisturbancehabitat. We investigatedthe matingsystem, patterns of isozyme variation,and effectivepopulation sizes of W. carteri to betterunderstand its populationbiology and to commenton reservedesigns and managementproposals relevant to this species. Warea carteriis self-compatibleand autogamous,and probablylargely selfing. Measures of geneticvariation in W. carteri were lower than values reportedfor species with similar ecological and life historytraits (6.6% of loci polymorphicwithin populations, 1.87 alleles per polymolphiclocus, and 0.026 and 0.018 expected and observed heterozygosity,respectively). The high average value for Nei's genetic identity(0.989) reflectsthe paucity of genetic diversity. Genetic variationwithin populations was not correlatedwith aboveground population size, effectivepopulation size estimates (Ne), or recentdisturbance history. Much of the diversitydetected was foundamong populations(FST = 0.304). A significant cline in allele frequenciesat one locus and a significantnegative correlation between geographic distance and Nei's genetic identityalso point to spatial organizationof geneticdiversity. As a resultwe propose thatreserve design should include the entiregeographic range of W. carteri. We also recommendthat the naturalfire regime be mimicked.

Key words: Brassicaceae; effectivepopulation sizes; genetic diversity;mating system; Warea carteri.

The Lake Wales Ridge (LWR; Highlands and Polk one of seven LWR endemics forwhich we are gathering Counties) of south-centralFlorida is both a hotspot of detailed geneticand demographicdata to evaluate current biodiversityin the continentalUnited States and highly reserve designs and make recommendationsabout fire endangered (Noss and Peters, 1995). Among higher management. plants, there are 16 federallylisted species endemic to Warea carteriis a federallyendangered annual occur- the LWR. The ongoingcreation of an archipelagoof fire- ringin xeric sandhillsand scrubbyflatwoods (community managed reserve "islands" is key to theirconservation. delineationsfollow Abrahamson, 1984). There are his- Designinga reservesystem and land managementprogram torical records of W. carteri fromoutside the LWR, in- that will supportviable populations or metapopulations cluding Brevard, Broward, Dade, De Soto, and Glades of these species requires knowledge of theirautecology Counties (Al-Shebaz, 1985). In his descriptionof the spe- and populationstructure. Warea carteri (Brassicaceae) is cies, Small (1909) stated that W. carteri was the "most common species of Warea, occur[ring]in great abun- 1 Manuscriptreceived 25 August 1998; revision accepted 24 June dance." Warea carteri has almost certainlybeen extir- 1999. The authors thank Butler undergraduatestudents Halle McKinney, pated fromthese areas, indicatingit has already suffered Aaron Mast, Bill Thornberg,Jennifer Copeland, and Matt Halfhill for a reductionin its range. The xeric communitiesof the assistance in conducting the electrophoresis;Dawn Berry, Deborah LWR were historicallydistributed in over 200 patches Graves, Stacey Halpern, Nancy Kohfeldt,George Landeman, Marina rangingin area fromfour to 1200 hectares(ha) along the Morales-Hernandez,Pedro F Quintana-Ascencio,Joyce Voneman, and -160 X 20 km north-southridge (Christman,1988). Per- Rebecca Yahr forfield assistance at ArchboldBiological Station(ABS); The NatureConservancy's staff, including Geoff Babb, Steve Mol-rison, oni and Abrahamson (1985) estimatedthat 84% of the Bea Pace, and Steve Shattlerfor help withcollecting census data at the presettlementextent of xeric, upland communities in Tiger Creek Preserve; Rebecca Yahr for organizingthe leaf sampling; Highlands County (which contains close to half the Pedro F Quintana-Ascenciofor assistance withnested ANOVAs; Russ LWR) had been convertedto citrusgroves or residential Lande for his recommendationsregarding the effectivepopulation size developmentsby 1981. methodology;Scott Bergen forcreating Figures 1 and 4; and Christine V. Hawkes, Lucinda McDade, Michael Nachman,D. Lawrence Venable, Demographic data indicate thatW. carterithrives in a and Pedro F Quintana-Ascenciofor helpful commentson the manu- temporallyrestricted, postdisturbance niche (Menges and script.This work was fundedby The NatureConservancy's Ecosystem Gordon, 1996). Large fluctuations(up to threeorders of Research Program,the Florida Chapterof The NatureConservancy, the magnitude)in the numberof plants are associated with Florida Division of Forestry,Archbold Biological Station (ABS), and fire,with increases the year afterfire followed by de- Butler University. 5Author forcorrespondence, current address: Departmentof Ecology creases the second year afterfire (Menges and Gordon, and EvolutionaryBiology, Biological Sciences West, Room 310, Uni- 1996). Oscillations in 2-yr cycles oftenoccur, damping versityof Arizona, Tucson, Arizona 85721 USA (520-621-4022). over time. Warea carteri may appear to be extirpated 372 March 2000] EVANS ET AL.-GENETICS AND ECOLOGY OF AN ENDANGERED ANNUAL 373

fromsites for years, only to have abovegroundindivid- uals recruitafter fire. Sites associated withanthropogenic disturbanceoften support more stable and persistentbut small populations. Historically,fires were frequentin peninsular Florida. Estimates of the natural fire return . 4 v : S :- intervalsof the communitiesW. carteri inhabitsspan 2- 20 yr,based on the life historiesof plants in these com- munitiesand theirpatterns of abundance in response to fire(Abrahamson and Hartnett,1990; Myers, 1990; Os- tertagand Menges, 1994; Menges and Kohfeldt, 1995; Menges and Hawkes, 1998). Fires have now been sup- pressed on the LWR for -60 yr. Orangc The ephemeralnature of populationsof W. carterisug- Osceola gests that seed banks are present.Seed dormancyof at least 2 yr has been demonstratedunder laboratorycon- ditions(N. Kohfeldtand E. Menges, ArchboldBiological Station [ABS], unpublisheddata). Fruitsand seeds of W. carterido not have any specialized structuresto facilitate dispersal, and spatial congruenceof abovegroundpopu- lations over time suggests limiteddispersal (E. Menges, ABS, personalobservations). Seeds of W. carteriaverage 0.282 ? 0.067 mg, and theirpotential horizontal travel is 1.89 m in a wind speed of 2.17 m/s(D. Gordon, The Nature Conservancy [TNC], unpublisheddata). Conservation of genetic diversityis considered to be .~~~~~~ S. importantfor the long-termevolutionary potential of species (Barrettand Kohn, 1991; Huenneke, 1991). The pri- marypurpose of thisstudy was to assess geneticdiversity in W. carteri, by surveyingisozymes in 23 populations throughoutthe species' range (Fig. 1), and make recommendationsabout how to capturethis diversityin reserves. Since genetic diversityis a dynamic,not a static phenomenon,we assessed processes and traitsthat might affect the population genetics of W. carteri. Using Wright'sF statistics,we assessed populationsubstructure and gene flow.Since the matingsystem of W. carteriwas unknown and likely to influencepatterns of genetic diversity(Loveless and Hamrick,1984; Hamrickand Godt, 1990), we determinedthis traitand discuss its implica- tions for demographicand genetic dynamics. Given the large fluctuationsof populationswith fire and persistence between fires via in situ seeds, we estimatedeffective populationsizes of W. carteri and relatepopulation-level genetic diversitystatistics with census population sizes, effectivepopulation sizes, and recenthistory of disturbance. Thus our approach to the conservationof W. carterihas been to obtain genetic,ecological, and demographicdata to reconstructthe dynamics of past and presentpopula- tions and evaluate reserve designs and firemanagement regimesthat will affectpopulation dynamics and genetic ,.,V.,V variationin the future.

MATERIALS AND METHODS

Floral biology-Flowers of W. carteri are perfectand offerboth nectar and pollen as floralrewards. They are open and radially sym- metrical,with fourwhite petals and six white stamensof equal length. The flowersare clusteredin round,dense racemes,and thereare several inflorescencesper plant. By observing marked flowerswe found that Fig. 1. Known populations of Warea carteri on the Lake Wales the flowersare protandrous.The stamensabscise beforestigmas become Ridge (LWR, shaded) in Lake, Polk, and Highlands Counties, Florida, receptive.Stigmatic receptivity was assessed using hydrogenperoxide, USA. Populations sampled for isozymes are shown with a filledcircle a standardtest (Kearns and Inouye, 1993), and by observing changes and populationsnot sampled are shown with an open circle. in the surfacearea and textureof the stigma. 374 AMERICAN JOURNAL OF BOTANY [Vol. 87

Matinzgsystem-We testedthe matingsystem of W. car-ter-iin a nat- froman initial survey of a population we obtained an estimateof the ural population at the Lake Placid Scrub Wildlife and Environmental populationsize, and on a second sweep throughthe populationwe sam- Area (LPSWEA) south of Lake Placid, Florida, in the 2nd and 3rd wk pled an appropriatefraction of the plantsto obtain a total sample of 30 of October 1994. This population consisted of -300 individuals bor- individuals.Leaves were shipped to the laboratorywithin 24 h, where deringa sand road in scrubbyflatwoods that had not burned in about they were processed for horizontalstarch gel electrophoresis.Tissue ten years.The matingsystem experiment tested for agamospermy (asex- was extractedin a modifiedWendel's sorghumbuffer (Morden, Doe- ual reproductionvia seeds), autogamy (spontaneous self-pollination), bley, and Schertz, 1987). Gels stained for acid phosphotase(ACP, En- and self-compatibility.Outcross-pollinated and open-pollinatedflowers zyme Commission designation3.1.3.2), alcohol dehydrogenase(ADH, served for comparisons.We applied each pollinationtreatment to one 1.1.1. 1), colorimetricesterase (EST, 3.1.1.-), peroxidase(PER, 1.11.1.7), inflorescenceper plant, such that all 20 plants in the experimentre- and phosphoglucomutase(PGM, 5.4.2.2) were run in sodium borate ceived each pollinationtreatment. bufferpH 8.6 (Kephart, 1990) at 200 v for 2.5 h. Glyceraldehyde-3- We removedall flowersand fruitsfrom the inflorescencesthat would phosphate dehydrogenase(G-3-PDH, 1.2.1.12), isocitratedehydroge- be treatedso that only flowerbuds remained and randomlyassigned nase (IDH, 1.1.1.41), glucose-6-phosphateisomerase (GPI, 5.3.1.9) and each to a pollinationtreatment. Plastic mesh bags (Delnet nonwoven 6-phosphogluconicacid dehydrogenase(6-PGD, 1.1.1.44) gels were run polyolefin,Applied ExtrusionTechnologies, P.O. Box 852, Middletown, in a histidine/citratebuffer system (Ellstrand, 1984) at 25 mA for 3 h. Delaware, 19709; mesh size 0.8-0.9 mm) secured withtwist-ties were Aconitase (ACO, 4.2.1.3), malate dehydrogenase (MDH, 1.1.1.37), used to isolate all but the controlinflorescences. Newly opened flowers mendadionereductase (MNR, 1.6.99.-), and triosephosphateisomerase in the agamospermy,self, and outcross treatmentswere emasculated (TPI, 5.3.1.1) gels were run on tris-citratepH 8.0 (Kephart,1990) at 50 beforeanther dehiscence daily for4 d. Flowers in the self-compatibility mA for 5 h. and outcross treatmentswere pollinated as their stigmas matured,by When more than one putative locus was observed for an enzyme, directcontact between anthers of first-dayflowers protected from insect loci were numberedsequentially, with the most anodally migratinglo- visitationand the stigmas.Generally, we used a single pollen donorper cus designated"one." We assigned sequentialletters to enzymevariants plantin the outcrosstreatment. The average distancebetween the plants forindividual loci in the same manner.No formalgenetic analyses were donatingand receivingpollen was 6.0 m (SE = 0.8, range = 1.5-13.1, conducted to documentthe patternof inheritanceof putative alleles. N = 20). On the fourthday, we removedthe remainingbuds and apical Allelic assignmentswere based on the observed patternof population meristemfrom all inflorescencesin all pollinationtreatments, ensuring variation,the known subunitstructure of the enzymes,and the cellular that no additional flowerswould release self-pollenand confoundthe compartmentalizationgenerally observed for plant enzymes (Kephart, treatments. 1990). Standardmeasures of levels of genetic variation(percentage of Fruitswere harvestedas theymatured and scored undera dissecting loci polymorphic,number of alleles per locus, and observed and ex- microscope for numberof aborted and full-sized,mature seeds. Seeds pected heterozygosity)were calculated forpopulations and forthe spe- were considered aborted if they appeared to have less than half the cies as a whole. Wright'sF statistics(Wright, 1965, 1978) were cal- endospermof a full-sizedseed. All full-sizedseeds fromeach fruitwere culated for all loci, and deviationsfrom Hardy-Weinberg expectations weighed togetherto obtaina mean seed mass. We used a Kruskal-Wallis were detected using chi-square tests. Pairwise genetic similaritybe- test to detect pollinationtreatment effects in the proportionalfruit set tween populationswas calculated using Nei's unbiased geneticidentity data, since these data could notbe transformedto approximatea normal (Nei, 1978). Hierarchicalcluster analysis of geneticidentity values was distribution(Sokal and Rohlf, 1995). Proportionalseed set data were performedusing arithmeticaveraging (UPGMA; Sneath and Sokal, arcsine transformedand a nested analysis of variance was performedto 1973). test for treatmenteffects. We testedfor differencesin mean seed mass per fruitamong the pollinationtreatments with a one-wayANOVA. All Effectivepopulation sizes-Because W. car-ter-ihas bothoverlapping ANOVA analyses were evaluated using Type IV sum of squares since generationsand fluctuatingpopulation sizes, we used models thatcor- differentfruit numbers were treatedin each plant and pollinationtreat- rect for these factors sequentially.Hill (1972) demonstratedthat the ment combination,resulting in emptycells in the data matrix(SAS, effectivesize per generationof an equilibriumpopulation with overlap- 1989). Pairwise differencesamong treatmentswere evaluated using Tu- ping generationsis equivalentto the effectivesize of a populationwith key's Honestly SignificantDifference. Significance was evaluated at P discretegenerations. We used the followingform of the "Hill" model < 0.05 throughout. (Eq. 11 in Caballero, 1994) to correctfor variance in reproductivesuccess and the matingsystem in populationswith overlapping generations: Electrophoresis-We visitedall knownextant sites for W. car-ter-ion Nv = (4NcL)/[2(1 -F1s) + Sk2(1 + F1s)] (1) the LWR for which we were able to obtain permissionto collect leaf materialduring the 1994 or 1995 floweringseason (October); additional whereNc is thecohort size, L is the generationlength, Sk2 is the variance areas of suitablehabitat were also surveyedfor populations. Populations in reproductivesuccess and Fls is Wright's(1969) statisticthat measures were defined by isolation from conspecifics by at least 50 m. Leaf the deviation fromHardy-Weinberg expectation within populations. It collectionswere made frompopulations that were at least 1.6 km distant is importantto note that ve used the floweringplant stage of the life fromother sampled populations.We sampled a total of 23 populations cycle as the referencepoint forour calculation of Nc and Sk2(this issue throughoutthe range W. ca-teri currentlyoccupies, with the exception is discussed clearlyby Rockwell and Barrowclough,1995). Specifically, of an isolated site known fromLake County,Florida, thatwe were not "cohortsize" refersto the numberof floweringplants expected to result able to access (Fig. 1). The northernmostsampled populationwas 115 froma coholt of seeds, and 'reproductivesuccess" refersto thenumber km fromthe southernmostsampled population. Abovegroundpopula- of floweringplants produced per floweringplant. The average effective tion size averaged 93 plants (median = 37, range = 18-703). History size per generation(Nv) was convertedto an average numberper year of fireand anthropogenicdisturbance were perfectlycolrelated among (Nv,*)by multiplyingby the generationlength (Hill, 1972; Lande and the populations we studied: all five populations occurringin natural Barrowclough,1987). We thentook the harmonicmean of these values vegetationhad burnedrecently, and all 18 populationsoccurring along to colrect for fluctuationin population size and age structure(Lande sand roads had not. and Barrowclough, 1987), giving our final estimate of effectivesize For genetic analysis, we collected a small numberof leaves from (Ni) every plant in populations containingbetween 15 and 29 plants, and The data requiredfor the Hill model include the numberof individ- froma spatiallystratified random sample in populationswith 30 or more uals (here,flowering plants) entering the population per generation,their plants. The stratifiedsamples were collected in the followingmanner: variance in reproductivesuccess, and Fls. Following Heywood (1986) March 2000] EVANS ET AL.-GENETICS AND ECOLOGY OF AN ENDANGERED ANNUAL 375

TABLE 1. Census data and estimated average effective population sizes for selected populations of Warea carteri at Archbold Biological Station (ABS) and The Nature Conservancy's Tiger Creek Preserve (TCP). These are populations censused for at least 6 yr and for which the maximum aboveground population size among all census years was at least 20 individuals. For this analysis, we defined populations as groups of plants that were isolated from conspecifics by at least 50 m and had the same fire history. NH, NA and N, are defined in the text.

Census Population 1988 1989 1990 1991 1992 1993 1994 1995 N11 NA N, TCP 1 458b 0 23b 4 1 0 78b 2.1 174.7 4.6 TCP 4a 2 47b 0 1 25b 7 2.2 29.9 4.8 TCP 11 1 25 0 41b 11 3 2.4 29.5 5.2 TCP 21 - lb 17 43b 2 1 1 0 gb 3.7 29.0 8.1 TCP lOa 2 10 5 0 703b 164 3.3 318.7 7.2 ABS 7a 18 0 21 4 23 11 4.0 28.1 8.7 ABS 2/3a 64 2 35 4 54 30 70 0 4.3 70.2 9.3 ABS 4 16 11 0 72b 6 135 13 88 5.6 92.4 12.1 ABS 1 58 2 44 12 18 30 7 10 8.4 48.9 18.1 TCP 5 12b 19 15 7 50b 7 11.8 39.6 25.5 ABS 10 77 18 134 16 45 21 28.8 112.0 62.2 TCP 3a 140 95 223 213 32 45 94b 77.0 259.9 166.3 ABS 12/13a - - 49 326 161 64 352 59 83.0 34.0 199.1 a These populations were also surveyed for isozyme diversity. b Fire occurred the previous year. and Husband and Barrett(1992), we obtained a species-wide estimate size estimatestested were the arithmeticand harmonicmeans of the of variance in reproductivesuccess fromdata on variance in fecundity average effectivepopulation sizes per year (Nv,*), describedabove, and from 11 populations in 3 yr. Only one and two of these populations the harmonicmean of census populationsizes (NA,Ne, and NH,respec- could be sampled for threeand two sequential years,respectively; the tively).Fire was consideredrecent if it had occurredwithin 3 yrbefore remainingpopulations did not appear abovegroundmore than one of leaves were sampled. the threeyears theywere sampled. Our estimateof seed productionper plantwas the productof the numberof fruitsper plant(N = 8-30 plants RESULTS per populationand year) and the average numberof full-sizedseeds per fruitper populationand year (N = 1 fruitper plant). Species values for Mating system-Warea carteriis self-compatibleand the mean and variance of seed productionper plant were obtained by autogamous,but not agamospermous(Fig. 2). Ninety-two averagingvalues across years withinpopulations and then across pop- percentof fruitand 61% of the ovules in these fruitde- ulations. We then used the analytical solution described by Barrow- veloped in response to hand-pollinationwith self pollen. clough and Rockwell (1993) to estimatevariance in reproductivesuc- Autogamous fruitset averaged 51 %, and seed set within cess: these fruitsaveraged 33%. Few fruitdeveloped in the Sk2 = P2VF + XFVP (2) agamospermy treatment(2%), and these fruitsset no set lower in the where p is the probabilityof survivingfrom seed to floweringplant seeds. Proportionalfruit was significantly (0.068; E. Menges and C. Weekley,ABS, unpublisheddata), is the autogamytreatment compared to eitherof the hand-pol- VP linated treatments(self and outcross); autogamous pro- variance associated with this probability[VP = p(l - p)], and XF and VF are species-level estimatesof the mean and variance in seed proportionalseed set was significantlylower than self, out- duction per plant described above, respectively.We obtained an equi- cross, and controltreatments. The proportionof fruitand libriumestimate of the numberof floweringplants enteringthe popu- seed set in the self and outcross treatmentswas signifi- lation per generation(N0L) withthe followingexpression: cantly higher than in the open-pollinatedcontrol treatment. Self and outcross pollinationtreatments were not N0L = NXFpL (3) significantlydifferent for fruit or seed set. Differencesin where N is the numberof floweringplants per year and L is the gen- mean seed mass per fruitwere not significantlyrelated to erationlength (mean age of parents,including residence time of seeds pollinationtreatment. in the soil, 4 yr). We used census data fromselected populationsof W. carteri at ABS and The Nature Conservancy's Tiger Creek Preserve Electrophoresis-Of 25 putative loci that could be (near Frostproofin Polk County,Florida) for values of N (Table 1). clearly and consistentlyresolved, six (24%) were poly- Abovegroundpopulation sizes of zero were replaced withones in these morphicat the species level: Acp-2, Est-], Idh-], Idh-2, calculations. A species-wide value for Fls was obtained fromthe iso- Pgm-2, and Pgm-3. Three of these,Est-], Idh-], and Idh- zyme data presentedhere. at Allele We used a nonparametricMann-Whitney U testto detectdifferences 2, had only one allele present >5% frequency. between estimatesof average effectivepopulation size in populations frequenciesvaried more widely for the remainingthree with and withoutfire management(where fire managementincludes loci (Table 2). Nei's genetic identityvalues for all pop- prescribedburning). We used Pearson and Spearmancorrelations to test ulations averaged 0.989 (range = 0.946-1.00). Within for relationshipsbetween census population size, effectivepopulation thisrange, the populationsgrouped into threemain clus- size, or disturbancehistory and the proportionof loci polymorphic,the ters and one outlier(Fig. 3). Banding patternswere con- numberof alleles per polymorphiclocus, and expected heterozygosity sistentwith the number of loci usually associated with per locus. The colrelation between these population genetic statistics diploids (Kephart, 1990), with the exception of PGM, and effectivepopulation sizes was testedwith data fromthe six popu- which had three stainingloci, representingan apparent lations for which we had both types of data. The effectivepopulation duplication.No private alleles were found. Three addi- 376 AMERICAN JOURNAL OF BOTANY [Vol. 87

1.1 ~ - _ I ______A cies in the numberof heterozygotes.The mean value of 1.0 FlS across all loci was 0.325. Genetic differentiationamong populations of W. carteri,assessed withWright's (1969) statisticFST, was mod- o I erate.Mean FST averaged across all polymorphicloci was 0') .8' 0.304, indicatingthat -30% of the total geneticvariance we detected in W. carteri is found among populations. a) .6- Based on this value, Nm, the number of migrantsper CD) generationis 0.57 (where Nm = [(1/FST) - 1]/4, from a .5- , Wright[1951]). This expressionfor Nm measureshistor-

4i ical, not current,gene flow and assumes populationsare o at equilibriumwith respect to migrationand drift(Slatkin i:0 .3~ and Barton, 1989). We founda weak but significantneg- 0. ative correlationbetween geographic distance and Nei's 2 .2- genetic identity(r2 = -0.147; P < 0.001), and a strong .11l association between relative population location on the north-southaxis of 0.0 ____ the LWR and a cline in the allele N 19 18 16 17 19 frequenciesat one locus (Fig. 4). This relationshipis sig- AUTO(a) AGAM(b) SELF(c) OUT(c) CONT(a) nificantfor two of threealleles at this locus (Spearman's rank order correlationcoefficient =-0.784, P < 0.001 .8 - _ _ and 0.709, P < 0.001, respectively). B Effectivepopulation sizes-Estimates of average effectivepopulation size were quite small for most popu- 0) lations (Table 1), thoughthere are unresolvedproblems ci~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~with the models and estimationof parameters(see Dis- 6 - cussion). Although the Mann-WhitneyU test did not CO, show any significantdifference in estimatesof average effectiveaboveground population size based on fire-man- CD' agement (P = 0.11, N = 13), the fire-managedpopula- co .4- tionstended to have lower estimatesfor average effective 0 aboveground population size than populations that did 0 not experiencefire management (Table 1). 0. 3- 0 There were no significantcorrelations between within- populationgenetic statistics(proportion of loci polymor- .2 j 1 ______phic, alleles per polymorphiclocus, or expected hetero- N= 613 73 84 zygosityper locus) and abovegroundpopulation size, the AUTO(a) SELF(b) OUT(b) CONT(b) threeestimates of effectivepopulation size, or historyof fireor anthropogenicdisturbance. However, average val- Treatment ues for these statisticsfrom the recentlyburned popula- Fig. 2. Results of the breeding system experiment of Warea carteri: tions in naturalvegetation were consistentlyhigher than proportionalfruit set (A) and proportionalseed set (B) by pollination average values fromthe roadside populationsthat had not = treatmentwith 95% confidenceintervals. AUTO autogamy,AGAM burned vs. of loci = agamospermy,SELF = self-pollination,OUT = outcrosspollination, recently(7.2 6.4% polymorphic,1.12 CONT = open pollination.Different letters beside the treatmentlabels vs. 1.09 alleles per polymorphiclocus, and 0.030 vs. denote significantdifferences in fiuitor seed set (P < 0.05). Sample 0.024 expected heterozygosity,respectively). size (N) is the numberof plants for fruitset and the numberof fruits for seed set. DISCUSSION Mating system,fecundity, and inbreeding-Given that tional loci (Gdh-1, Pgi-1, and Acp-1) were apparently Warea carteri is semelparous and the numberof mates variable,but could not be consistentlyresolved and were available varies in response to a stochasticphenomenon not scored. (fire),we expected to findevidence of selection for re- The mean number of alleles per polymorphiclocus productive assurance. Self-compatibilityand autogamy withinpopulations was 1.87 (range 1.00-2.67) and, on are featuresof the matingsystem of W. carterithat con- average, only 6.6% of loci were polymorphic(the most firmthis expectation. Reproduction via selfing allows common allele being presentat <95% frequency;range contributionsto the seed bank betweenfires. Still, insects 0-12%). Average expected heterozygositywas 0.026 do play a role in maintaininghigh fecundity in W. carteri. (range 0.000-0.045); average direct-countheterozygosity Lower fruitand/or seed set in the autogamy treatment was 0.018. Eight populations had genotypefrequencies compared to the hand-pollinated and open-pollinated thatdeviated significantlyfrom Hardy-Weinberg equilib- treatmentssuggests that insect-facilitated pollination con- rium for all three polymorphicloci; anotherfive popu- tributessignificantly to fecundity(26% totalreduction in lations deviated for at least one locus (R. Dolan, unpub- fecunditywithout insects). Lower fruitand seed set in the lished data). In each case the loci examinedhad deficien- open-pollinated (control) treatmentcompared to the March 2000] EVANS ET AL. GENETICS AND ECOLOGY OF AN ENDANGERED ANNUAL 377

TABLE 2. Allele frequencies by population for the three loci with more than one allele present at >5% frequency detected in War-ea carteri. Populations are alranged from north to south.

Acp-l Pgm-2 Pgm-3 Population Status' a b c a b c a b c Horse Cr 1, 5 0.81 0.15 0.04 0.7 0.83 0.10 1.00 TCP-Libby 1, 3 0.43 0.57 1.00 1.00 TCP-CH 2, 3 0.63 0.37 0.20 0.47 0.33 1.00 Hillcr Hts 2, 4 0.57 0.43 0.95 0.05 0.98 0.02 TCP-Pfund 2, 3 0.17 0.77 0.07 1.00 1.00 TCP-Gate 2, 3 0.70 0.30 1.00 1.00 LASF-Tr 1, 3 0.40 0.60 0.07 0.77 0.16 1.00 LASF-Z Tr 1, 3 0.33 0.67 0.03 0.93 0.03 0.97 0.03 Carter Cr 1, 5 0.27 0.74 1.00 1.00 Van Pelt 1, 4 0.08 0.92 0.33 0.60 0.07 1.00 F Villas 1, 5 0.06 0.94 0.02 0.61 0.37 1.00 T Pouch 1, 4 0.31 0.69 0.81 0.19 1.00 J Creek 1, 5 0.27 0.67 0.06 0.13 0.78 0.09 0.98 0.02 Grand C 1, 4 1.00 0.86 0.14 1.00 L L No 1, 4 1.00 1.00 1.00 L L New 1, 4 1.00 0.19 0.09 0.72 0.91 0.09 High P Est 1, 4 0.06 0.94 - 0.72 0.28 0.88 - 0.12 Lakeview 1, 4 1.00 0.72 0.28 0.96 0.04 Lk PI Scrb 1, 3 1.00 0.62 0.34 0.04 0.88 0.22 ABS-Hufty 2, 3 0.28 0.62 0.10 0.03 0.76 0.21 1.00 ABS-McJ 1, 3 0.37 0.63 0.40 0.50 0.10 1.00 ABS-W 1, 3 1.00 1.00 1.00 ABS-S 1, 3 1.00 0.03 0.97 1.00 a History of disturbance (1 roadside, without recent fire, 2 = natural vegetation, recently burned), and current protection status (3 protected, 4 = not protected, 5 = protection planned) of the populations. hand-pollinatedtreatments is suggestive of pollen limi- stages in selfingspecies (Husband and Schemske, 1996). tation(30 and 21% reductionin fruitset and 11 and 9% The low level of isozyme diversitywe foundwithin pop- reductionin seed set, comparedto self and outcrosstreat- ulations of W. carteriis consistentwith habitual inbreed- ments,respectively). If insect visitationis positivelyre- ing, among other possible causes (such as population lated to plant density,fire could positivelyaffect fecun- fluctuation). dityin a density-dependentmanner (i.e., Allee effect,see Lamont, Klinkhamer, and Witkowski, 1993; Widen, 1993; and Groom, 1998). Species and within-populationpatterns of isozyme of the life historytraits in W. carteri Several aspects of the reproductivebiology of W. car- variation-Several with little the teri suggest that selfingis the common mode of repro- are associated genetic diversity,including gravity-dispersed duction. Warea carteri is protandrous,but because there annual habit,selfing, insect pollination, are manyflowers densely clustered within inflorescences, seeds, as well as rarityand a postdisturbanceniche (Ham- pollen and receptivestigmas are presentsimultaneously rick et al., 1991). The isozyme analysis of W. carteri in close proximity.In addition,many inflorescences flow- revealed levels of variationthat are lower than average er simultaneouslyon a plant.In preliminaryobservations values for species grouped on the basis of each of these (620 min), only a small fraction(3%) of the insecttravel traits.The proportionof loci polymorphicin W. carteri occurredbetween flowers on differentplants of W. carteri is less than six of seven average values foundin a more (M. Evans, ABS, unpublisheddata). Warea carteri was recentstudy that grouped species on the basis of two life foundto be highlyself-compatible, and autogamousfruit historytraits (Hamrick and Godt, 1996). The high aver- set (51%) was not significantlydifferent from open-pol- age value of Nei's genetic identity(0.989) in W. carteri linated fruitset, suggestingautogamy naturally occurs at reflectsthe paucityof geneticvariation at the species and a relativelyhigh rate. All of these factorssuggest that populationlevels. The suite of traitscombined in W. car- pollen transferwithin plants followed by fertilization(au- teri may cause its species- and population-levelstatistics togamyand geitonogamy)is common in W. carteri. The of genetic diversityto be lower than most of the values deficienciesof heterozygotesfound in the isozyme sur- reportedin other studies. Our results contraststrongly vey could be the result of selfing,among other causes with published studies of isozyme variationin some of (e.g., Wahlundeffect). The short-and long-termnegative the perennialherbs endemic to the LWR. These studies impacts of selfinginclude inbreedingdepression and the reportedsurprisingly high levels of geneticvariance from loss of geneticdiversity (Barrett and Kohn, 1991; Huen- smaller samples (Lewis and Crawford,1995; McDonald neke, 1991). The lack of a significantdifference in either and Hamrick,1996). In Polygonella spp., these relatively proportionalfruit or seed set between self and outcross high levels were attributedto the LWR servingas a gla- pollinationsindicates that inbreeding depression does not cial refugiumfor some of the species studied(Lewis and impact early life historystages, althoughinbreeding de- Crawford,1995). A recent study of genetic diversityin pression is expected to be expressed in later life history anotherFlorida scrub endemic,using randomlyamplified 378 AMERICAN JOURNAL OF BOTANY [Vol. 87

1.0 - Similarity Zone 1 0.8 - N= I 0.962 0.975 0.987 1.00 0.6 - + -- + -- + -- + 0.4 -

ABS-McJ 21 0.2 -

Van Pelt 10 0.0 - K MR a b c Lakeview 18 1.0 Lk Pi Scrb 19 Zone 2 0.6 N=5

ABS-S 23 0.6 -

LLNo 15 0.4-

ABS-W 22 0.2 -

GrandC 14 0.0 I a b c F Villas 11 0 . HighP Est 17 ?D 08Zone30.8 TCP-Libby2 I N=2 06 HilIcrHts 4 aL) 4 0.4 -

TCP-Pfund5 -D 0.2-

CarterCr 9 c ABS-Hufty20 a b

1.0- J Creek 13 Zone 4 068- N =8 LASF Tr 7 0,6- T Pouch 12 04 - LASF-ZTr 8 02 -

TCP-C High3 0.0 a b c TCP-Gate6

Horse Cr 1 1.0 zoneZ 5 LL New 16 0.8 - N = 7 Fig. 3. Cluster diagram generatedby the UPGMA method(Swof- 0.6 fordand Selander,1989) based on Nei's unbiased geneticidentity values 0.4 for 23 populationsof Wariiea carteri. The copheneticcorrelation coef- fecientis 0.70. Numbersfollowing population names coiTespondto Ta- 0.2 ble 2 and indicaterelative position north to south in ascending order.

polymorphicDNA, foundno variation(G. Romano, Uni- a b versityof Florida, unpublisheddata). a Fig. 4. Histogramsillustrating north-south clines in the allele frequencies at the locus ACP- 1 in Warea carteri.The populationssampled and for isozyme diversitywere divided into five zones of equal distance Census and effectivepopulation sizes genetic from northto south. The histogramsindicate the average allele fre- diversitystatistics-The lack of significantrelationships quencies (with 95% confidenceintervals) of the populations in each between population-level patternsof genetic variation zone. The frequenciesof alleles a and b were significantlyrelated to and census population sizes or effectivepopulation sizes populationposition. estimatedfrom census data reflectsthe difficultyof pre- dicting the population genetics of a species that has a complex life historyand is responsiveto a stochasticeco- were probablynot good predictorsof populationgenetic logical phenomenon.Several plant studies have found a statistics(percentage of loci polymorphic,number of al- significantpositive relationshipbetween census popula- leles per polymorphiclocus, and expectedheterozygosity tion size and percentageloci polymorphic,number of al- per locus) in our species because of its seed banks. There leles per polymorphiclocus, and/orgene diversity(i.e., is need, both theoreticallyand empirically,for a better expectedheterozygosity; see referencesin Table 1 in Ells- understandingof how seed banks may influenceplant trand and Elam, 1993; Prober and Brown, 1994; Ra- population geneticdynamics. ijmann et al., 1994; Sun, 1996). Census populationsizes Because of its seed banks, we used effectivepopula- March 2000] EVANS ET AL.-GENETICS AND ECOLOGY OF AN ENDANGERED ANNUAL 379 tion sizes to tryto link the demographicand geneticbe- The lack of diversitywe foundwithin populations is con- havior of populations of W. carteri. Unfortunately,the sistentwith drift,especially in combinationwith the ev- theoryfor calculating effective population size fromcen- idence for inbreeding.The fluctuationin aboveground sus data forpopulations with overlapping generations and population sizes of W. carteri suggestsits effectivepop- fluctuatingpopulation sizes is poorly developed. An im- ulation sizes could be low, which would lead to drift. portantassumption of the "Hill" model (Hill, 1972) that However, divergence among populations that is not ex- we used is that the population is at equilibrium.It is cessive and the cline we observed at one locus contradict assumed thatneither cohort size, inbreeding,nor variance a strongrole for drift.Ne remainsa criticalparameter to in familysize changes fromyear to year; under this as- estimatein orderto understandhow significantdrift was sumption,the effectivesize per generationis the same as or is in populationsof W. carteri.We review alternative the effectivesize for a population with discretegenera- calculations of Ne in a separate manuscript(M. Evans, tions. This was not the case in our species. We used con- unpublisheddata), includingmolecular and demographic stantvalues for inbreedingand variance in familysize, estimatesfor the seven listed endemicplants of the LWR summarizedfrom data thatvaried among loci, years,and/ we are studying. or populations. Our estimatesof cohort size varied because of strongfluctuations in reproduction.There is no Past, present, and future-It is likely that both pol- indicationin the theoreticalliterature as to how inaccu- linationand seed bank dynamicsof W. carterihave been rate estimatesof effectivesize per generationmay be for affectedby recenthabitat loss and firesuppression. Pop- populationsexperiencing such fluctuation.In addition,we ulations have become more isolated both spatially and found that summarizingthe fecunditydata in different temporally,constricting gene flow. Fire suppressionhas ways and differentestimates of p led to very different probably depleted seed banks of W. carteri, both as a estimates of variance in family size, cohort sizes, and source of genetic variation and demographic stability. effectivesizes (M. Evans, unpublisheddata). These rea- Differenttypes of disturbance (anthropogenicvs. fire) sons may explain the lack of significantrelationships be- probably also affect demographic dynamics, effective tween effectivepopulation sizes and population genetic population sizes, and population genetic dynamics, statistics. thoughit is not clear how. For example, we mightexpect less genetic diversityin the fire-disturbedpopulations Gene flow and drift-We found evidence thatgenetic that we sampled, since they experienced strongerpop- variance is structuredin W. carteri,but the data are sub- ulation fluctuationand smallerminimum population siz- ject to alternativeexplanations other than limited gene es, but the average values for genetic diversitystatistics flow. The life historytraits found in W. carteri are ex- (percentage of loci polymorphic,number of alleles per pected to limit gene flow among populationsand are as- polymorphiclocus, and expected heterozygosityper lo- sociated with greatervalues of FST, the proportionof the cus) were consistentlyhigher in the fire-disturbedpop- geneticvariation found among populations.The value we ulations than in the roadside populations. Our conser- found for W. carteri (FST = 0.304) is moderaterelative vative recommendationis that the natural fire regime, to otherspecies witheach of these traits(Hamrick et al., which is still under investigation (Abrahamson and 1991), though an even higherproportion might be ex- Hartnett, 1990; Myers, 1990; Ostertag and Menges, pected since these traitsare combined in W. carteri.Es- 1994; Menges and Kohfeldt,1995; Menges and Hawkes, timatedgene flow (Nm = 0.57) based on this value of 1998), be mimicked.The spatial organizationof genetic FST suggests random differentiationcould occur among diversityfound in this study suggests that preservation populations at neutralloci (Wright,1951). However, us- of scrub habitat throughoutthe range of W. carteri is ing FSTto estimategene flow assumes thepopulations are needed to adequately capture the currentdistribution of in migration-driftbalance. Whitlock (1992) found that genetic variation. Similar results and recommendations temporalfluctuations in demographicparameters, includ- were made for three other LWR endemics (Lewis and ing populationsize and migrationrate, increased both the Crawford,1995; McDonald and Hamrick, 1996). mean and variance of FST. He concluded estimatesof FST and Nm fromsingle samples of genetic markersshould Summary-We found that W. carteri is autogamous be viewed conservativelyin species experiencingdemo- and self-compatibleand predict its mating system in- graphicstochasticity. The cline in allele frequenciesalong cludes habitual selfing.These traits,especially combined the north-southaxis of the LWR is anotherform of spa- with seed banks, probably promote persistence in the tial organizationwe detectedin the isozyme data. How- face of spatial and temporal heterogeneity.We found ever,a cline at one locus could be the resultof selection. little genetic variationin W. carteri and what variation This cline in turnwas responsiblefor another spatial pat- we did detect was spatially organized, thus we recom- tern we found: the negative correlationbetween geo- mend range-wide representationin reserves. Further graphic distance and Nei's genetic identity.These indi- evaluation of reserve design and managementwith re- cations thatgenetic diversity is spatiallystructured in W. spect to geneticdiversity in W. carteriwill requirebetter carteri are consistentwith our understandingof its aut- estimatesof effectivepopulation sizes and gene flow.A ecology (selfingand limitedseed dispersal),but because betterunderstanding of the seed biology of W. carteri, they are subject to otherexplanations and are based on includingthe cues forgermination, germination fraction, few loci, we cannot make strongconclusions about gene and the dynamicsof seed viabilityin the soil, will allow flow. us to bettermodel both effectivepopulation sizes and The data we presenthere also do not provide a clear demographicdynamics. Our work highlightsthe impor- pictureof the role of driftin populations of W. carteri. tance of incorporatingseeds into demographic and ge- 380 AMERICAN JOURNAL OF BOTANY [Vol. 87 netic modeling of certainplant populations and the equi- E. Soul6 [ed.], Viable Populationsfor Conservation, 87-123. Cam- librium assumptions of models that have commonly bridge UniversityPress, New York, New York, USA. LEWIS, P O., AND D. J. CRAWFORD. 1995. Pleistocenerefugium endem- been used to estimate essential microevolutionarypa- ics exhibit greaterallozymic diversitythan widespread congeners rameters(Nm and Ne). in the genus Polygoniella(Polygonaceae). Amiierican?Joutrnal of Bot- any 82: 141-149. LOVELESS,M. D., AND J. L. HAMRICK.1984. Ecological determinants LITERATURE CITED of geneticstructure in plantpopulations. Ainnual Review of Ecology and Systematics15: 65-95. ABRAHAMSON, W. G. 1984. Species responses to fireon the Florida McDONALD, D. B., ANDJ. L. HAMRICK.1996. Geneticvariation in some Lake Wales Ridge. AmericacnJourncal of Botacny71: 35-43. plants of Florida scrub.Americani Journal of Botanyi,83: 21-27. AND D. C. HARTNETT. 1990. Pine flatwoodsand dry prairies. MENGES,E. S., ANDD. R. GORDON. 1996. Three levels of monitoring In R. L. Myers and J. J. Ewel [eds.], Ecosystems of Florida, 103- intensityfor rare plant species. Natural Areas Journ-1al16: 227- 149. UniversityPresses of Florida. Gainesville, Florida, USA. 237. AL-SHEBAZ, I. A. 1985. The genera of the Thelypodieae (Cruciferae; , ANDC. V. HAWKES. 1998. Interactiveeffects of fireand mi- Brassicaceae) in the southeasternUnited States. Joutrncalof theAr- crohabitaton plants of Florida scrub. Ecological Applications 8: nold A rboretumii66: 95-11 1. 935-946. BARRETT, S. C. H., AND J. R. KOHN. 1991. Genetic and evolutionary AND N. KOHFELDT. 1995. Life historystrategies of Florida consequences of small population size in plants: implicationsfor scrubplants in relationto fire.Bulletini of the Tor-r-eyBotanical Club conservation.In D. A. Falk and K. E. Holsinger [eds.], Genetics 122: 282-297. and conservation of rare plants. Oxford UniversityPress. New MORDEN,C. W., J. DOEBLEY,AND K. F SCHERTZ.1987. A manual of York, New York, USA. techniques for starch gel electrophoresisof Sorghumlisozymes. BARROWCLOUGH, G. F, AND R. E ROCKWELL. 1993. Varianceof lifetime Texas AgriculturalExperiment Station, MP-1635, College Station, reproductivesuccess: estimationbased on demographicdata. Amer- Texas, USA. icanI Natutralist 141: 281-295. MYERS,R. L. 1990. Scrub and high pine. In R. L. Myers and J.J. Ewel CABALLERO, A. 1994. Developmentsin the predictionof effectivepop- [eds.], Ecosystemsof Florida, 150-193. Universityof CentralFlor- ulation size. Heredity73: 657-679. ida Press, Orlando, Florida, USA. CHRISTMAN, S. P. 1988. Endemism and Florida's interiorscrub. Final NEI, M. 1978. Estimationof average heterozygosityand genetic dis- Report to Florida Game and Fresh Water Fish Commission,Tal- tance froma small numberof individuals.Genetics 89: 583-590. lahassee, Florida, USA. ContractNumber GFC-84-101. Noss, R. E, AND R. L. PETERS. 1995. Endangeredecosystems: a status ELLSTRAND, N. C. 1984. Multiple paternitywithin the fruitsof wild reporton America's vanishinghabitat and wildlife. Defenders of radish,Raphanius sativus. AmericanNaturalist 123: 819-828. Wildlife,Washington, DC, USA. , AND D. R. ELAM. 1993. Population genetic consequences of OSTERTAG, R., AND E. S. MENGES. 1994. Patternsof reproductiveeffort small populationsize: implicationsfor plant conservation.Anlnual with time since last firein Florida scrub plants. Journal of Vege- Review of Ecology anidSystemnatics 24: 217-242. tationScience 5: 303-310. GROOM, M. J. 1998. Allee effectslimit population viability of an annual PERONI, P A., AND W. G. ABRAHAMSON. 1985. A rapid method for plant. AmericanNaturzalist 151: 487-496. determininglosses of native vegetation.Natural Areas Journal 5: HAMRICK, J. L., AND J. W. GODT. 1990. Allozyme diversityin plant 20-24. species. In A. H. D. Brown, M. T. Clegg, A. L. Kahler,and B. S. PROBER, S. M., AND A. H. D. BROWN. 1994. Conservationof the grass Weir [eds.], Plant populations genetics,breeding, and genetic re- box woodlands: populationgenetics and fragmentationof Eucalyp- sources, 43-63. Sinauer; Sunderland,Massachusetts, USA. tus albens. ConservationBiology 8: 1003-1013. AND . 1996. Effectsof life historytraits on genetic RAIJMANN,L. E. L., N. C. VAN LEEUWEN, R. KERSTEN, J. G. B. Oos- diversityin plant species. Philosophical Transactionsof the Royal TERMEIJER, J. C. M. DEN NIJS, AND S. B. J. MENKEN. 1994. Genetic Society of LonidoniB 351: 1291-1298. variationand outcrossingrate in relationto populationsize in Geni- HAMRICK, J. L., M. J. W. GODT, D. A. MURAWSKI, AND M. D. LOVELESS. tianiapneumonanthe L. ConservationBiology 8: 1014-1026. 1991. Correlationsbetween species traitsand allozyme diversity: ROCKWELL,R. F, AND G. FEBARROWCLOUGH. 1995. Effectivepopulation implicationsfor conservationbiology. In D. A. Falk and K. E. size and lifetimereproductive success. Conser-vationBiology 9: Holsinger [eds.], Genetics and conservationof rare plants,75-86. 1225-1233. OxfordUniversity Press, New York, New York, USA. SAS. 1989. SAS/STAT user's guide, version 6, 4th ed., vol. 1. SAS HEYWOOD, J. S. 1986. The effectof plantsize variationon geneticdrift Institute,Cary, NorthCarolina, USA. in populationsof annuals. AmericaniNaturalist 127: 851-861. SLATKIN, M., AND N. H. BARTON. 1989. A comparisonof threemethods HILL, W. G. 1972. Effectivesize of populationswith overlapping gen- for estimatingaverage levels of gene flow. Evolution?43: 1349- erations.Thleoretical Population Biology 3: 278-289. 1368. HUENNEKE, L. E 1991. Ecological implicationsof genetic variationin SMALL, J. K. 1909. Additionsto the floraof peninsularFlorida. Bulletin plant populations. In D. A. Falk and K. E. Holsinger [eds.], Ge- of the TorreyBotanical Club 36: 159-164. netics and conservationof rare plants. Oxford UniversityPress. SNEATH, P H. A., AND R. R. SOKAL. 1973. Numerical taxonomy.W. New York, New York, USA. H. Freeman. San Francisco,California, USA. HUSBAND, B. C., AND S. C. H. BARRETT. 1992. Effectivepopulation SOKAL, R. R., AND F J. ROHLF. 1995. Biometry,3rd ed. W. H. Freeman, size and genetic driftin tristylousEichhornlia paniculata (Ponted- New York, New York, USA. eriaceae). Evolutioni46: 1875-1890. SUN, M. 1996. Effectsof population size, matingsystem, and evolu- 1 AND D. W. SCHEMSKE. 1996. Evolution of the magnitudeand tionaryorigin on genetic diversityin Spiranthes sinensis and S. timingof inbreedingdepression in plants. Evolution 50: 54-70. honigkongensis.Conserivation Biology 10: 785-795. KEARNS, C. A., AND D. W. INOUYE. 1993. Techniques for pollination SWOFFORD, D. L., AND R. B. SELANDER. 1989. Biosys-1: a computer biologists.University Press of Colorado, Niwot, Colorado, USA. programfor the analysis of allelic variationin populationgenetics KEPHART, S. R. 1990. Starch gel electrophoresisof plant isozymes: a and biochemical systematics. Illinois Natural History Survey, comparativeanalysis of techniques.Americani Journal of Botany Champaign,Illinois, USA. 77: 693-712. WIDEN, B. 1993. Demographic and genetic effectson reproductionas LAMONT, B. R., P. G. L. KLINKHAMER, AND E. T. E WITKOWSKI. 1993. related to population size in a rare, perennialherb, Senecio inte- Population fragmentationmay reduce fertilityto zero in Bacnksia grifolius(Asteraceae). Biological Jouernzalof the Linnean Society goodii, a demonstrationof the Allee effect.Oecologia 94: 446- 50: 179-195. 450. WHITLOCK, M. C. 1992. Temporalfluctuations in demographicparam- LANDE, R., AND G. F BARROWCLOUGH. 1987. Effectivepopulation size, eters and the genetic variance among populations. Evolution 46: genetic variation, and their use in population management. In M. 608-615. March 2000] EVANS ET AL.-GENETICS AND ECOLOGY OF AN ENDANGERED ANNUAL 381

WRIGHT, S. 1951. The geneticstructure of populations.Annals of Eu- theoryof gene frequencies:University of Chicago Press, Chicago, geniics15: 323-354. Illinois, USA. . 1965. The interpretationof populationstructure by F-statistics . 1978. Evolution and the geneticsof populations,vol. IV. Var- with special regardto systemsof mating.Evolution1 19: 395-420. iabilitywithin and among naturalpopulations. University of Chi- . 1969. Evolution and the geneticsof populations,vol. II. The cago Press, Chicago, Illinois, USA.