Population Structure and Mating System in Tachigali Versicolor, a Monocarpic Neotropical Tree

Heredity 81 (1998) 134–143 Received 28 May 1997, accepted 24 November 1997

Population structure and mating system in Tachigali versicolor, a monocarpic neotropical tree

M. D. LOVELESS*†, J. L. HAMRICK‡ & R. B. FOSTER§ †Department of Biology, The College of Wooster, Wooster, OH 44691, U.S.A., ‡Departments of Botany and Genetics, University of Georgia, Athens, GA 30602, U.S.A. and §Department of Botany, Field Museum of Natural History, Chicago, IL 60605, U.S.A.

Patterns of genetic differentiation among six populations of Tachigali versicolor (Fabaceae: Papilionoideae) on Barro Colorado Island, Panama were investigated. Average gene diversity

within any one population (He) was 0.073 (SD 0.010), and He over all populations was 0.080. Populations showed relatively little genetic differentiation (mean GST = 0.069), suggesting high levels of gene flow. A direct estimate of pollen flow indicated that 21% of the pollen received by a cluster of five trees had travelled at least 500 m. Genetic analyses of the mating system showed complete outcrossing (tm = 0.998, SE<0.054; mean ts = 1.001, SE<0.063). Estimates of F at different life stages showed a slight deficiency of heterozygotes in the six population samples, a slight excess among the 25 flowering adult trees, and no significant deviation in heterozygosity in progeny arrays. These differences may reflect the monocarpic life history of T. versicolor, in which adults flowering in a given year represent a temporal genetic bottleneck, producing a Wahlund effect in genotype frequencies in the overall population. This interpretation of genetic structure in T. versicolor suggests the overriding importance of ecological factors and life history on genetic processes in natural populations.

Keywords: gene ﬂow, genetic diversity, genetic structure, mating system, monocarpy, Tachigali versicolor.

Introduction evolutionary and ecological dynamics of the population. Population genetic structure and the mating system Recent studies have examined the genetic make- are closely interrelated aspects of a species’ up of tropical tree populations (reviewed in Love- ecological genetics. The amount of genetic variation less, 1992; Hall et al., 1994; Boshier et al., 1995a,b; in a species is affected by various ecological pheno- Gibson & Wheelwright, 1995). In general, tropical mena (Loveless & Hamrick, 1984; Hamrick & Godt, plant species have levels of allozyme variability 1990), as well as by the mating system (Holtsford & equal to those of their temperate counterparts Ellstrand, 1989). Furthermore, the way in which (Hamrick & Godt, 1990; Loveless, 1992). Analyses genetic variation is distributed among populations is of genetic differentiation at a local scale on Barro closely tied to gene flow, both by pollen and seed Colorado Island, Panama (hereafter BCI) show that movement (Levin & Kerster, 1974; Loveless & populations separated by one to several kilometres Hamrick, 1984). Because genetic structure is gener- have relatively little genetic differentiation, implying ated by the dispersal, survival and reproduction of high levels of gene flow. Such patterns of population individuals in a population over time, an apprecia- differentiation can be correlated with breeding tion of the evolutionary genetics of a species must biology, individual distribution, and pollen and seed take into account not only strictly genetic events, dispersal (Hamrick & Loveless, 1989; Hamrick & such as fertilization and the mating system, but also Murawski, 1990). Evidence suggests that gene flow ecological and life history traits as they influence the in tropical forests is often substantial, tying together populations of even relatively rare tropical trees into *Correspondence. E-mail: [email protected] large demes or breeding units (Hamrick & Muraw-

134 ©1998 The Genetical Society of Great Britain. GENETIC STRUCTURE IN TACHIGALI VERSICOLOR 135 ski, 1990; Chase et al., 1996; Nason et al., 1996; Stacy peccaries and fungal infection) mortality (Kitajima et al., 1996). & Augspurger, 1989). Information on the reproductive biology and Seeds lack dormancy, and germination occurs at mating systems of tropical trees provides an inde- the start of the rainy season, in late April and early pendent corroboration of high levels of gene flow. May. Germination of undamaged seeds is high, and Initial studies of tropical plant mating systems seedling shadows may be dense beneath adults. demonstrated that dioecy and other sexual systems Seedlings are shade-tolerant, persisting for many which enforce outcrossing were more common in years in the understorey. However, seedlings in tropical than in temperate species (Bawa & Opler, light-gaps grow rapidly, forming a cohort that is 1975). High levels of outcrossing have been heterogeneous in size (Kitajima & Augspurger, confirmed by genetic analyses of tropical tree mating 1989). To reach canopy size, several gap-mediated systems (reviewed in Loveless, 1992). Fewer data are growth spurts are probably necessary. available for seed dispersal; however, the preponder- ance of animal-mediated seed dispersal suggests that dispersal should have a high variance, permitting Field collections long-distance seed movement (Hamrick & Loveless, Genetic diversity on BCI was sampled at two spatial 1986; Hamrick et al., 1993). scales. Four collection areas located about 200 m In this paper, local patterns of genetic organiza- apart were chosen within a 50 ha mapped forest plot tion in populations of Tachigali versicolor (Fabaceae: established in 1980 (the Forest Dynamics Plot, or Papilionoideae) on BCI, Panama, are described and FDP; Hubbell & Foster, 1983). Within each area, 72 the mating system is analysed using allozyme individuals were sampled. Two additional off-plot markers. These data allow calculation of indirect sites, each located at least 1 km from the FDP, were and direct estimates of gene flow in a natural popu- also identified. Fifty individuals were sampled in lation of this tropical tree. these sites. In all locations, individuals of different size classes were sought over an area of 6–8 ha. Leaf Materials and methods collections were kept cool in the field, vacuum-dried for at least 48 h, packaged with dessicant, and The study species returned to the U.S.A. Dried samples were kept at Tachigali versicolor (Fabaceae: Papilionoideae) is 70°C until used for electrophoresis. found from Costa Rica to western Colombia (Croat, Seed collections were made in February and 1978). It is a relatively common canopy tree in March 1985, when mature fruits from the 1984 secondary and primary forests on BCI (Foster, 1977; flowering episode (Fig. 1) were being dispersed. Kitajima & Augspurger, 1989). Tachigali versicolor is Seeds were collected directly beneath 25 adult trees. monocarpic: adults die after flowering (Foster, In those cases where seed shadows of different 1977). Other, South American species of Tachigali adults might have overlapped, collections were made are also apparently monocarpic (Gentry, 1993). Age in a direction away from other seed sources. Seeds at flowering is unknown, but growth rates vary were germinated in a growing house on BCI, and widely, depending on the light environment of seed- leaf tissue was vacuum-dried. lings and saplings (Augspurger, 1984; Kitajima & Augspurger, 1989), and trees may flower in a wide Laboratory procedures range of size classes (Foster, pers. obs.; Loveless, pers. obs.). Flowering is pulsed at intervals of 4 to Horizontal starch gel electrophoresis was performed 6 years (Foster, 1977, pers. obs.). On BCI, flowering on the samples. The dried leaves were crushed occurred in 1970, 1974, 1978, 1983, 1984, 1989 and under liquid nitrogen, mixed with an extraction 1994 (Foster, 1977, pers. obs.). Although individuals buffer and adsorbed onto filter paper wicks. A total sometimes flower (and die) out of synchrony, in of 31 allozyme loci in 17 enzyme systems were most off-years there is no flowering. resolved in the population samples. Isozymes and Flowering occurs from January to July. Individual buffers used in this study are given in Table 1, and trees flower for 6–12 weeks. The mature fruit is a are modified from Soltis et al. (1983). large, wind-dispersed, single-seeded samara. Most Four loci (Idh, Dia, Fe1 and Fe3) were used to fruits fall 100 m from the parent (Kitajima & assess mating systems. In all cases, progeny banding Augspurger, 1989). Fruits suffer both pre- (bruchid patterns were consistent with a simple Mendelian beetles and parrots) and postdispersal (rodents, interpretation of the allozyme loci.

Table 1 Enzyme systems resolved in electrophoretic analyses of Tachigali versicolor. Buffer systems are taken from Soltis et al. (1983)

Abbreviation and loci Buffer system Enzyme system scored (n)

6 Fluorescent esterase Fe (3) Peroxidase Per (2) Leucine amino peptidase Lap (2) Acid phosphatase Acp (1) Malic enzyme Me (1) Colorometric esterase Ce (3) Triose-phosphate isomerase Tpi (3) Phosphogluco isomerase Pgi (2) Phosphoglucomutase Pgm (2) 4 6-Phosphate dehydrogenase 6Pgdh (2) Aldolase Ald (1) 5 Isocitrate dehydrogenase Idh (1) Diaphorase Dia (1) Fig. 1 Map showing the location of flowering individuals Malate dehydrogenase Mdh (4) and sampled trees (squares) of Tachigali versicolor on 7 Glutamic-oxaloacetic Got (1) Barro Colorado Island, Panama, in 1984. The forest transaminase dynamics plot is shown by the box in the centre of the Diaphorase Dia (1) island. The five trees indicated by the star are those used in the direct estimate of pollen movement. 8* b-Galactosidase Bgal (1) *Modified from Soltis et al. (1983). Statistical analysis Standard statistics of genetic diversity were calculated for individual population samples and for were also calculated between each pair of popula- BCI as a whole by pooling over all populations. For tion samples. each population, we determined the proportion of The mating system was analysed according to the polymorphic loci (P), the mean number of alleles methods of Ritland & Jain (1981), using a program per locus (A) and per polymorphic locus (AP), and from Ritland (1990). The procedure estimates multi- 2 locus outcrossing (t m), mean single-locus outcrossing the mean genetic diversity (He =1Sp i where pi is the frequency of the ith allele; Nei, 1973). Statistics (ts) and maternal genotypes, and calculates pollen were calculated for each locus and then averaged allele frequencies and individual outcrossing rates over all loci. Observed and expected heterozygosities for each maternal family. Variances for individual for each polymorphic locus were compared by family outcrossing rates were based on 100 boot- calculating Wright’s fixation index (where straps; variances for population outcrossing esti- F =1[H /H ]) to determine deviations from mates were based on 500 bootstraps. Differences in o e pollen allele frequencies among families were tested random-mating expectations. Deviations of F from x 2 zero were tested using x 2 (Li & Horovitz, 1953). using a heterogeneity (Workman & Niswander, Differences in allele frequencies among popula- 1970). tions were examined using a x 2 analysis (Workman & Niswander, 1970). For each polymorphic locus, Results statistics of gene diversity were calculated (Nei, Genetic variation within and among populations 1973, 1977), including total gene diversity (HT), mean gene diversity within populations (HS), gene Of 31 loci surveyed for T. versicolor on BCI, 13 diversity among populations (DST) and GST, the (41.9%) were polymorphic in at least one popula- proportion of the total gene diversity found among tion. On average, however, only 29.6% of the loci populations (GST = DST/HT). Overall means were were variable in any given population (Table 2). calculated by averaging over all polymorphic loci. Despite the relative proximity of the populations on Nei’s coefficients of genetic distance (Nei, 1972) BCI, there were six private alleles (Slatkin, 1985)

© The Genetical Society of Great Britain, Heredity, 81, 134–143. GENETIC STRUCTURE IN TACHIGALI VERSICOLOR 137 present. Populations were similar in levels of poly- mean AP = 2.23), and these alleles were rare where morphism, number of alleles per locus and per poly- they were found (frequencies of 0.02). Mean morphic locus, and observed and expected expected heterozygosity was 0.073, and the pooled heterozygosities (Table 2). Only three of the 13 He-value for all BCI sites was only slightly higher polymorphic loci had more than two alleles (BCI (0.080). Observed heterozygosities were consistently smaller than expected, indicating a slight deficiency of heterozygotes (Table 2). When allele frequencies at each locus and in each population were tested Table 2 Estimates of genetic variation for six populations against Hardy–Weinberg expectations, 11 of 55 fixa- of Tachigali versicolor on Barro Colorado Island (BCI), tion indices (20%) were significantly different from Panama (see footnote for definition of abbreviations) zero. In every instance, the deviation from zero was positive, indicating heterozygote deficiencies at these Population NP A APHo He loci. FDP 1 72 0.290 1.350 2.220 0.058 0.072 Allele frequencies at 10 of the 13 polymorphic loci FDP 2 71 0.323 1.320 2.000 0.057 0.062 differed significantly among the six populations FDP 3 72 0.355 1.390 2.090 0.078 0.089 (Table 3). However, DST values overall were rela- FDP 4 72 0.323 1.350 2.100 0.055 0.064 tively low, resulting in small GST values (mean = TBarbour 50 0.226 1.260 2.140 0.061 0.070 0.069). Values for GST are additive and can be Armour 52 0.258 1.320 2.250 0.071 0.082 subdivided hierarchically within BCI to compare the Mean* 0.296 1.330 2.130 0.063 0.073 four populations within the FDP with the two SD 0.047 0.044 0.092 0.009 0.010 samples from outside the FDP. Hierarchical GST Within BCI 0.419 1.520 2.230 — 0.080 statistics (Table 3) indicate, however, that population differentiation does not conform to expectations N, number of plants in the population sample; P, for these spatial patterns. The four neighbouring proportion of polymorphic loci; A, mean number of alleles per locus; AP, mean number of alleles per polymorphic populations within the FDP have more than twice the among-population differentiation than they have locus; Ho, observed heterozygosity; He, gene diversity, or Hardy–Weinberg expected heterozygosity; FDP, forest with populations at the level of BCI. Calculations of dynamics plot. genetic identity among the six populations also *Means include all 31 loci, 13 of which are polymorphic. reflect this; FDP1 is relatively different from all the

Table 3 Heterogeneity x 2-values and Nei’s statistics of genetic diversity for 13 polymorphic loci in Tachigali versicolor (Nei, 1973)

2 Locus x (d.f.) HT HS DST GST GSTB* GSTP†

Fe2 21.8 (3) 0.487 0.480 0.007 0.014 0.004 0.011 Fe3 80.9 (1) 0.079 0.071 0.008 0.104 0.015 0.089 Me 17.9 (1) 0.010 0.010 0.000 0.023 0.000‡ 0.020 Ce1 4.7 (1) 0.003 0.003 0.000 0.006 0.000‡ 0.000‡ Ce2 156.4 (1) 0.084 0.067 0.017 0.201 0.017 0.184 Ce3 21.0 (1) 0.266 0.259 0.007 0.027 0.004 0.024 Pgi1 28.0 (3) 0.210 0.206 0.004 0.018 0.009 0.009 Pgm1 7.0 (1) 0.013 0.013 0.000 0.009 0.008 0.001 6Pgdh2 69.2 (1) 0.252 0.230 0.022 0.089 0.030 0.058 Ald 117.5 (1) 0.233 0.198 0.035 0.151 0.120 0.031 Idh 74.7 (6) 0.215 0.208 0.007 0.032 0.029 0.004 Dia 168.8 (1) 0.409 0.320 0.089 0.217 0.029 0.188 Mdh3 10.1 (1) 0.205 0.202 0.003 0.013 0.001 0.012 Mean 0.190 0.174 0.015 0.069 0.020 0.049

*GSTB is that portion of the total GST that differentiates populations within the forest dynamics plot from the other two populations on Barro Colorado Island.

†GSTP is that portion of the total GST that occurs between populations within the forest dynamics plot. ‡Because of rounding.

Table 4 Estimates of population outcrossing rates in occurred between allele frequencies in the repro- Tachigali versicolor. Standard errors are shown in ductive gene pool (pollen and ovules) and the vege- parentheses tative population in five of eight comparisons (Table 5). Pollen and allele frequencies at all four loci also Multilocus outcrossing rate (tm) = 0.998 (0.054)* differed significantly among the 25 adults (Table 6). Single locus outcrossing rates (ts) Based on the multilocus outcrossing estimate, the Idh 1.026 (0·090)† equilibrium inbreeding coefficient [Feq =(1 t)/ ǹ Dia 0.983 (0·123)† (1 t); Wright, 1965] for this population is Fe1 1.052‡ Feq = 0.001, reflecting the absence of inbreeding Fe3 0.769‡ among the progeny arrays. The mean fixation index Average ts 1.001 (0·063)* in the sample of adults flowering in 1984 [calculated as F =1(Ho/He) and averaged over the four loci] *SE based on 500 bootstraps for the entire model. †SE based on 100 bootstraps. was negative (F = 0.106), indicating a slight excess ‡Fit of the model did not permit variance estimates. of heterozygotes. A value of the inbreeding coefficient for the population samples was also obtained from FIS, which represents within-population other locations (mean I = 0.982 vs. I = 0.995 for all inbreeding. The average FIS for the 13 polymorphic other pairwise comparisons). loci in the six multi-aged population samples was positive (FIS = 0.187), and significantly different from the fixation index in the flowering adult sample Mating system analysis (Student’s t-test, t15 = 2.434, P0.05). There is a Analyses of progeny arrays from 25 adult trees that deficiency of heterozygotes in the population flowered in 1984 indicate that T. versicolor is samples, and a moderate excess of heterozygotes completely outcrossed. The multilocus outcrossing among the flowering adults. rate was tm = 0.998 and the mean single-locus outcrossing rate was t = 1.001 (Table 4). For two of s Measures of gene flow the loci (Fe1 and Fe3), the frequency of the second allele was very low, and the iteration algorithm did Two indirect methods were used to estimate Nm, not permit variance estimates. Such individual ts the number of migrants per generation from popula- rates are probably not very dependable. Bootstrap tion structure. Wright’s (1951) equation based on variances were calculated for Idh and Dia, and indi- FST (GST) gave a value for Nm = 3.32. The Slatkin cate complete outcrossing (Table 4). If single-locus method (Slatkin, 1985) gave Nm = 6.65. These esti- outcrossing rates are lower than multilocus rates, mates are rather different in magnitude, but both this difference can be attributed to biparental suggest relatively large amounts of migration among inbreeding (Shaw et al., 1980; Ritland & Jain, 1981). T. versicolor populations on BCI. In T. versicolor, therefore, there is no evidence for A more direct method of measuring gene flow by biparental inbreeding. pollen was possible because of the spatial distribu- Chi-squared values were calculated to test the fit tion of genotypes in the 1984 flowering adults. Five of progeny arrays to the mixed-mating model. Both flowering trees located in a cluster were homozygous Idh and Dia showed slight deviations from the for the common allele at the Idh locus. However, 10 2 mixed-mating model (x 1 = 5.79, 0.01P0.025 and of the 240 progeny were heterozygous for the rare 2 x 1 = 3.93, 0.025P0.05, respectively), indicating allele, which must have originated from beyond an excess of heterozygous progeny from homozygous these five trees. The value of m, the rate of gene parents. Progeny arrays at Fe1 and Fe3 were not flow, can be calculated from the equation: significantly different from those predicted by the m =(q q )/(q Q), mixed-mating model. 0 1 0 Adult genotypes for these 25 trees were inferred where q0 is the frequency of the rare allele in the from their progeny arrays, permitting separate esti- five adults (0.00), q1 is the frequency of the allele in mations of allele frequencies in the ovule and pollen their progeny (0.021), and Q is the frequency of the pools. Our population samples for BCI, using trees allele in the flowering population as a whole (0.200). of all size classes, provided a measure of allele This gives a value of m = 0.105. Thus, 21% of the frequencies in the population as a whole. Compari- effective pollen migrated from beyond this cluster of sons between ovule and pollen allele frequencies five trees. In 1984, the nearest flowering tree to this were not significant. However, significant differences group was 500 m distant (Fig. 1).

Table 5 Comparison of allele frequencies among different source groups in Tachigali versicolor on Barro Colorado Island (BCI). ‘Adults’ refers to the 25 trees sampled from the 1984 ﬂowering cohort. ‘Pollen’ is the frequency of alleles in the total pollen pool sampled (1017 gametes). ‘BCI’ represents the overall allele frequencies generated by pooling the six populations shown in Table 2. Standard errors are shown in parentheses for those values which were estimated using maximum likelihood methods. x 2-values test for differences among different sources using a contingency x2

Locus and allele

No. alleles Idh Dia Fe1 Fe3 sampled

Source 24232312 N

Adults 0.200 0.800 0.120 0.880 0.980 0.020 0.000 1.000 50 (ovules) (0.049) (0.049) (0.045) (0.045) (0.026) (0.026) (0.000) (0.000) Pollen 0.130 0.869 0.194 0.806 0.964 0.036 0.008 0.992 1017 (0.028) (0.023) (0.030) (0.030) (0.014) (0.014) (0.005) (0.005) BCI 0.122 0.878 0.287 0.713 1.000 0.000 0.041 0.959 778 2 2 2 2 2 x Adults vs. pollen x 1 = 2.036 x 1 = 1.681 x 1 = 0.372 x 1 = 0.396 2 2 2 2 2 x Adults vs. BCI x 1 = 2.574 x 1 = 6.520* x 1 = 15.579*** x 1 = 2.139 2 2 2 2 2 x Pollen vs. BCI x 1 = 0.236 x 1 = 21.236*** x 1 = 28.901*** x 1 = 22.387***

*P0.05, ***P0.001.

Discussion In fact, although T. versicolor is ecologically common on BCI (61.9 individuals1.0 cm d.b.h. Genetic variation within and among populations ha 1; Condit et al., 1996), its unique life history Compared to 15 other common tree species on BCI, makes it more similar, genetically, to rare tropical T. versicolor has low levels of genetic variation tree species. Because only a fraction of the popula- (Tables 2 and 7). Values of P in T. versicolor are tion reproduces in any flowering year, the number of lower than those for 16 rarer species on BCI reproductive individuals is much less than the actual (Hamrick & Murawski, 1991) and those found in a numerical population. Flowering cohorts on BCI in broad survey of tropical woody plants (Loveless, the last six flowering cycles ranged from 12 trees (in 1992). Mean heterozygosity is also lower in T. versi- 1983, out of synchrony) to around 75 (in 1974). In color than in other tropical tree species (Table 7). effect, T. versicolor has reproductive densities of 0.05 Heterozygosity in this species is similar to values individuals ha1 or less, with seedling cohorts of found in some of the rare species studied on BCI different years being generated in genetic isolation but is considerably below the mean for the rare from one another. Such small flowering populations species (Hamrick & Murawski, 1991). are likely to experience genetic drift relative to the

Table 6 Range in pollen allele frequencies and in multilocus outcrossing rates among 25 progeny families of Tachigali versicolor. Allele frequencies of the rare allele at each locus are shown

Idh Dia Fe1 Fe3 tm

Range in values 0·00–0·65 0·00–0·90 0·00–0·08 0·00–0·16 0.63–2.00* 2 x 24 306.10 477.03 92.15 137.65 P0.001 P0.001 P0.001 P0.001

*Estimates of tm close to 2.00 are an artifact of the Newton–Raphson algorithm used in the computer analysis. These are typical of families with multiply heterozygous parents, about which little information on t can be obtained (K. Ritland, program documentation).

Table 7 A comparison of estimates of genetic parameters in Tachigali versicolor and other tropical trees. Because samples for this species came from a limited geographical area (Barro Colorado Island, BCI), we compare the data to within- population values from other studies. Standard errors are shown in parentheses

Average Population No. of No. of Polymorphic heterozygosity differentiation

species loci loci (P) (%) (He)(GST) Source

Tachigali versicolor — 31.0 29.6 0.073 0.069 This study 16 common woody species 16 34.5 60.9 0.211 0.055 Hamrick & Loveless on BCI (1989) 16 rare woody species on BCI 16 — 42.1 0.142 — Hamrick & Murawski (1991) Native woody tropical plants 81 — 39.0 (2.7) 0.109* (0.011) 0.109† (0.019) Loveless (1992)

*He was based on 74 species for which data were available. †GST was based on 37 species for which data were available population as a whole, with consequent loss of 7). In fact, it has more differentiation within the genetic variation. The loss of genetic diversity would four sites on the 50-ha FDP than most of the other be enhanced if whole cohorts subsequently failed to species studied (Hamrick & Loveless, 1989). This reproduce. This may explain the relatively low level may reflect limited seed dispersal and seedling of genetic variation in T. versicolor on BCI. Because clumping in T. versicolor (Foster, 1977; Kitajima & possibilities for recombination and the spread of Augspurger, 1989; Hamrick et al., 1993). alleles between cohorts are limited, with long lag times, genetic diversity can be lost to future genera- Mating systems tions by drift in every reproductive episode. Past evolutionary events (such as bottlenecks or specia- Tachigali versicolor is completely outcrossed. In this tion events) could also contribute to low levels of respect, the species resembles most tropical tree genetic diversity in this species. species for which mating systems have been The six local populations on BCI all had a defi- described. Of 13 tropical woody angiosperm species ciency of heterozygotes, which is usually explained (including T. versicolor) summarized by Loveless by inbreeding, but T. versicolor is completely (1992), only two [Cavanillesia platanifolia (Murawski outcrossed. Because each of the population samples et al., 1990), and Ceiba pentandra (Murawski & included individuals of different sizes and ages, and Hamrick, 1992)] had mixed-mating systems. Even because flowering behaviour in T. versicolor strongly among species with mass canopy flowering, where structures gene flow by pollen, population samples geitonogamous pollen transfer must occur lumped together individuals from different, geneti- frequently, outcrossing is the rule. cally isolated cohorts (i.e. a temporal Wahlund There was also no evidence for biparental effect). Even within a single reproductive cohort, the inbreeding in T. versicolor. Although some flowering seed are grouped around the maternal tree, which trees are spatially close to one another, their should to lead to fine-scale genetic structure in the progeny do not show apparent inbreeding that seedlings (Hamrick & Loveless, 1986; Hamrick et al., would indicate genetic relatedness among nearby 1993). Where such existing population subdivision is adults. This is consistent with the assumption that not recognized, the result is a Wahlund effect: a individuals flowering in a given year are not from a deficiency of heterozygous genotypes relative to the common seedling cohort, but represent a genetically expectations of the composite population. unrelated mixture of adults that happen to flower at GST calculations (Table 3) indicated only small a particular time. amounts of differentiation among these six popula- There were no significant differences between tions. This differentiation did not, however, reflect ovule and pollen allele frequencies at any of the four spatial separation of the populations, suggesting that loci (Table 5), suggesting that flowering trees contri- some process other than simple isolation by distance buted more or less equally to reproduction through is acting to structure local populations. pollen and ovules. However, there were significant Tachigali versicolor is similar to other tropical trees differences between allele frequencies in the repro- in having limited population differentiation (Table ductive gene pool (pollen and ovules) and the vege-

© The Genetical Society of Great Britain, Heredity, 81, 134–143. GENETIC STRUCTURE IN TACHIGALI VERSICOLOR 141 tative population on BCI in five of eight local pollen pools. In addition, there is phenological comparisons. If the reproductive gene pool repre- variation among T. versicolor trees flowering in the sented only the 25 sampled maternal trees, such same year (M. D. Loveless, pers. obs.), so trees that differences could result from sampling error. are clustered could experience different pollen pools However, given the likelihood of long-distance as nearby trees initiated or ended their flowering. pollen movement, the pollen pool probably sampled The inbreeding coefficient (F) among different a large fraction of the 61 individuals flowering in groups of T. versicolor showed considerable varia- 1984 (Fig. 1). This suggests that there are real differ- tion. The F among flowering adults in 1984 was ences between trees flowering in 1984 and the negative, indicating an excess of heterozygotes. The overall population, a difference that reflects the equilibrium Feq calculated from the mating system genetic bottleneck through which this species passes estimate of t was essentially zero, indicating during each flowering episode. Because local popu- panmixia among progeny produced in 1984. lations of T. versicolor are slightly different in allele However, the mean FIS value calculated among the frequencies, as GST statistics indicate, adults flower- six population samples was positive, indicating a ing in any year (Fig. 1) would combine individuals moderate deficiency of heterozygotes. This is often from many subpopulations. There are two levels of interpreted as the result of partial selfing within a differentiation that could be involved: spatial differ- mixed-mating system (Brown, 1979), but T. versicolor entiation among different parts of the island, and is completely outcrossed. Another explanation may temporal differentiation among individuals derived be the effect of the life history of T. versicolor in from different cohorts. From this perspective, the generating temporal and spatial genetic structure. reproductive and the vegetative samples represent Separate cohorts, established in different years but different genetic phases of the population of T. versi- each derived from an essentially island-wide mating color on BCI. The 1984 flowering adults represent episode, would produce nonoverlapping (but geneti- the allele frequencies in that reproductive popula- cally correlated) subpopulations. Within each tion as it passes through a temporal bottleneck that cohort, inbreeding might be zero, but different seed will constrain the genetic make-up of the 1984 seed cohorts would come from different reproductive cohort. The vegetative samples, on the other hand, gene pools, each resulting from a temporal bottle- represent individuals of various sizes and ages, as neck created by the subset of flowering adults. seed cohorts from different flowering years are A composite vegetative sample, combining indi- dispersed and grow to adult size in different parts of viduals of different sizes and ages, would generate a the island. temporal Wahlund effect analogous to that Pollen allele frequencies showed highly significant described for spatial subdivision. differences in the pollen pools sampled by different mothers (Table 6). Homogeneity of the pollen pool Gene flow is an assumption of the mixed-mating model (Clegg, 1980; Ritland & Jain, 1981), and the violation of this The indirect estimates of Nm in the present study assumption in T. versicolor may explain why, for Idh suggest substantial gene flow among populations. As and Dia, there was an excess of heterozygous seed dispersal is relatively limited (Kitajima & progeny among homozygous mothers. Where they Augspurger, 1989), this most probably represents have been looked for, differences in pollen allele gene movement by pollen. The direct estimate of frequencies among maternal individuals have been pollen movement indicates that pollinators do, in found in many plant populations. In tropical trees, fact, move pollen relatively long distances (500 m) significant variations in the pollen pool were among flowering trees. The fact that even relatively described for Cavanillesia platanifolia (Murawski et isolated flowering individuals also usually show al., 1990). Stacy et al. (1996) showed that local densi- essentially complete outcrossing provides additional ties of flowering individuals significantly affect evidence that, during fertilization, gene flow is breeding structure, with clumped individuals being substantial. Pollination is not, however, panmictic, more likely to mate with their nearest neighbours. because there were significant differences among Although such spatial patterns would be a logical individuals in the genetic make-up of their pollen explanation for local variations in the pollen pool, pools. little is known about pollinator movement among High rates of gene flow among individuals and the canopies of mass-flowering tropical trees. As we between populations have been reported for other have genotypes for only 41% of the trees that tropical tree species. Webb & Bawa (1983) demon- flowered in 1984, we cannot completely describe strated that pollen may move 225 m between popu-

© The Genetical Society of Great Britain, Heredity, 81, 134–143. 142 M. D. LOVELESS ET AL. lations of the hummingbird-pollinated shrub ogy, by flowering tree density, and by patterns of Malvaviscus arboreus. Using paternity analysis, pollinator movement among trees. Population Hamrick & Murawski (1990) showed that more than samples are a composite of many seed cohorts, and 25% of the pollinations of the bee-pollinated canopy of other, genetically different future reproductive tree Platypodium elegans on the FDP involved moves cohorts. From a genetic perspective, being in the of more than 750 m. A recent study by Chase et al. same seedling cohort is less important than being in (1996) using microsatellite markers showed pollen the same flowering cohort. Thus, the life history movement of up to 350 m in Pithecellobium elegans. dynamic of T. versicolor has a corresponding genetic Thus T. versicolor is not unusual in the degree of dynamic; the population is repeatedly stratified and moderate-distance pollen dispersal it experiences. homogenized by pollen movement, and then under- goes spatial differentiation because of limited seed dispersal. A life history interpretation of genetic structure In T. versicolor, genetic structure is unlikely to arise Acknowledgements from classical isolation by distance during pollination, as has been suggested for many plant species We appreciate field support from the Smithsonian (Levin & Kerster, 1974). Although neighbouring Tropical Research Institute. Techical assistance was reproductive adults may exchange pollen at higher provided by Ms Sue Sherman-Broyles. Andrew frequencies than more distant individuals, the asyn- Schnabel offered helpful comments on the manu- chrony of eventual flowering within a cohort, script. This research was supported by NSF Grant frequent long-distance pollen movement, and the BSR 82-06946 to J.L.H. and assistance from the outcrossed mating system should generate consider- Luce Foundation and the College of Wooster to able genetic mixing in the seeds produced in a single M.D.L. flowering season. In contrast, seed dispersal in T. versicolor is much References more local. Most fruits fall within 150 m of the adult AUGSPURGER, C. K. 1984. Light requirements of neotrop- tree (Kitajima & Augspurger, 1989). Growth rates ical tree seedlings: a comparative study of growth and among seedlings in a cohort vary widely (R. B. survival. J. Ecol., 72, 777–795. Foster, pers. obs.), masking their similar genetic BAWA, K. S. AND OPLER, P. A. 1975. Dioecism in tropical origins. As a result, different spatial populations are trees. Evolution, 29, 167–179. composed of mixtures of seed cohorts, each geneti- BOSHIER, D. H., CHASE, M. R. AND BAWA, K. S. 1995a. Popu- cally constrained by the genotypes of adults flower- lation genetics of Cordia alliodora (Boraginaceae), a ing in the year it was produced. Those trees that Neotropical tree. 2. Mating system. Am. J. Bot., 82, eventually flower then reconstitute a new, tempo- 476–483. rally constrained gene pool, within which pollen BOSHIER, D. H., CHASE, M. R. AND BAWA, K. S. 1995b. Popu- movement occurs over a spatially much larger scale. lation genetics of Cordia alliodora (Boraginacese), a In this flowering cohort, a set of new seed genotypes Neotropical tree. 3. Gene flow, neighborhood, and population substructure. Am. J. Bot., 8, 484–490. is generated, but they again undergo limited disper- BROWN, A. H. D. 1979. Enzyme polymorphism in plant sal. Spatial genetic structure, such as it is in this populations. Theor. Pop. Biol., 15, 1–42. species, is thus most probably the result of limited CHASE, M. R., MOLLER, C. KESSELI, R. AND BAWA, K. S. 1996. seed movement and temporal genetic stratification Distant gene flow in tropical trees. Nature, 383, among seed and seedling cohorts. 398–399. Given the amount of pollen movement among CLEGG, M. T. 1980. Measuring plant mating systems. flowering individuals, are the six local samples in this Bioscience, 30, 814–818. study really populations? The answer seems to be, CONDIT, R., HUBBELL, S. P. AND FOSTER, R. B. 1996. yes, sort of. They are overlapping dispersal shadows, Changes in tree species abundance in a Neotropical winnowed progressively by mortality, but they are forest: impact of climate change. J. Trop. Ecol., 12, not genetic demes because the mixing of genes is 231–256. CROAT, T. B. 1978. Flora of Barro Colorado Island. Stan- limited in T. versicolor by its monocarpic flowering ford University Press, Stanford, CA. phenology. There is a window — that of flowering in FOSTER, R. B. 1977. Tachigalia versicolor is a suicidal the same year — for the individuals in a local area neotropical tree. Nature, 268, 624–626. to pass through in order to be part of a deme, in the GENTRY, A. H. 1993. A Field Guide to the Families and sense of sharing genes. The actual demic effective Genera of Woody Plants of Northwest South America. population size is determined by flowering phenol- Conservation International, Washington, DC.

GIBSON, J. P. AND WHEELWRIGHT, N. T. 1995. Genetic struc- minants of genetic structure in plant populations. Ann. ture in a population of a tropical tree Ocotea tenera Rev. Ecol. Syst., 15, 65–95. (Lauraceae): influence of avian seed dispersal. Oecolo- MURAWSKI, D. A. AND HAMRICK, J. L. 1992. Mating system gia, 103, 49–54. and phenology of Ceiba pentandra (Bombacaceae) in HALL, P., ORRELL, L. C. AND BAWA, K. S. 1994. Genetic Central Panama. J. Hered., 83, 401–404. diversity and mating system in a tropical tree, Carapa MURAWSKI, D. A., HAMRICK, J. L., HUBBELL, S. P. AND guianensis (Meliaceae). Am. J. Bot., 81, 1104–1111. FOSTER, R. B. 1990. Mating systems of two Bombaca- HAMRICK, J. L. AND GODT, M. J. W. 1990. Allozyme diversity ceous trees of neotropical moist forests. Oecologia, 82, in plant species. In: Brown, A. H. D., Clegg, M. T., 501–506. Kahler, A. L. and Weir, B. S. (eds) Plant Population NASON, J. D., HERRE, E. A. AND HAMRICK, J. L. 1996. Pater- Genetics, Breeding, and Genetics Resources, pp. 43–63. nity analysis of the breeding structure of strangler fig Sinauer, Sunderland, MA. populations: evidence for substantial long-distance wasp HAMRICK, J. L. AND LOVELESS, M. D. 1986. The influence of dispersal. J. Biogeog., 23, 501–512. seed dispersal mechanisms on the genetic structure of NEI, M. 1972. Genetic distance between populations. Am. plant populations. In: Estrada, A. and Fleming, T. H. Nat., 106, 283–292. (eds) Frugivores and Seed Dispersal, pp. 211–223. W. NEI, M. 1973. Analysis of gene diversity in subdivided Junk Publ., The Hague. populations. Proc. Natl. Acad. Sci. U.S.A., 70, HAMRICK, J. L. AND LOVELESS, M. D. 1989. The genetic 3321–3323. structure of tropical tree populations: associations with NEI, M. 1977. F-statistics and analysis of gene diversity in reproductive biology. In: Bock, J. H. and Linhart, Y. B. subdivided populations. Ann. Hum. Genet., 41, 225–233. (eds) The Evolutionary Ecology of Plants, pp. 129–146. RITLAND, K. 1990. A series of FORTRAN computer Westview Press, Boulder, CO. programs for estimating plant mating systems. J. Hered., HAMRICK, J. L. AND MURAWSKI, D. A. 1990. The breeding 81, 235–237. structure of tropical tree populations. Pl. Sp. Biol., 5, RITLAND, K. AND JAIN, S. K. 1981. A model for the estima- 157–165. tion of outcrossing rate and gene frequencies using n HAMRICK, J. L. AND MURAWSKI, D. A. 1991. Levels of allo- independent loci. Heredity, 47, 35–52. zyme diversity in populations of uncommon neotropical SHAW, D. V., KAHLER, A. L. AND ALLARD, R. W. 1980. A tree species. J. Trop. Ecol., 7, 395–399. multilocus estimator of mating system parameters in HAMRICK, J. L., MURAWSKI, D. A. AND NASON, J. D. 1993. plant populations. Proc. Natl. Acad. Sci. U.S.A., 78, The influence of seed dispersal mechanisms on the 1298–1302. genetic structure of tropical tree populations. Vegetatio, SLATKIN, M. 1985. Rare alleles as indicators of gene flow. 107/108, 281–297. Evolution, 39, 53–65. HOLTSFORD, T. P. AND ELLSTRAND, N. C. 1989. Variation in SOLTIS, D. E., HAUFLER, C. H., DARROW, D. C. AND GASTONY, outcrossing rate and population genetic structure of G. J. 1983. Starch gel electrophoresis of ferns: a compil- Clarkia tembloriensis (Onagraceae). Theor. Appl. Genet., ation of grinding buffers, gel and electrode buffers, and 78, 480–488. staining schedules. Am. Fern J., 73, 9–15. HUBBELL, S. P. AND FOSTER, R. B. 1983. Diversity of canopy STACY, E. A., HAMRICK, J. L., NASON, J. D., HUBBELL, S. P., trees in a Neotropical forest and implications for FOSTER, R. B. AND CONDIT, R. 1996. Pollen dispersal in conservation. In: Sutton, S. L., Whitmore, T. C. and low density populations of three neotropical tree Chadwick, A. C. (eds) Tropical Rain Forest: Ecology and species. Am. Nat., 148, 275–298. Management, pp. 24–41. Blackwell Scientific Publica- WEBB, C. J. AND BAWA, K. S. 1983. Pollen dispersal by tions, Oxford. hummingbirds and butterflies: a comparative study of KITAJIMA, K. AND AUGSPURGER, C. K. 1989. Seed and seed- two lowland tropical plants. Evolution, 37, 1258–1270. ling ecology of a monocarpic tropical tree, Tachigalia WORKMAN, P. L. AND NISWANDER, J. D. 1970. Population versicolor. Ecology, 70, 1102–1114. studies on southwestern Indian tribes. II. Local genetic LEVIN, D. A. AND KERSTER, H. W. 1974. Gene flow in seed differentiation in the Papago. Am. J. Hum. Genet., 22, plants. Evol. Biol., 7, 139–220. 24–49. LI, C. C. AND HOROVITZ, D. G. 1953. Some methods of WRIGHT, S. 1951. The genetical structure of populations. estimating the inbreeding coefficient. Am. J. Hum. Ann. Eugen., 15, 323–354. Genet., 5, 107–117. WRIGHT, S. 1965. The interpretation of population struc- LOVELESS, M. D. 1992. Isozyme variation in tropical trees: ture by F-statistics with special regard to systems of patterns of genetic organization. New Forests, 6, 67–94. mating. Evolution, 19, 395–420. LOVELESS, M. D. AND HAMRICK, J. L. 1984. Ecological deter-