Gene flow and geographic variation in natural populations of Alnus acumi1'lata ssp. arguta (Fagales: Betulaceae) in Costa Rica and Panama

Seventeen natural populations in Costa Rica and Panama were used to asses geneflow and geographic patterns of genetic variation in tbis tree species. Gene flow analysis was based on the methods of rare alleles and FST (Index of genetic similarity M), using the only four polymorphic gene loci among 22 investigated (PGI-B, PGM-A, MNR-A and IDH-A). The geographic variation analysis was based on Pearson 's correlations between four geographic and 1 4 genetic variables. Sorne evidence of isolation by distance and a weak gene flow among geographic regions was found. Patterns of elinal variation in relation to altitude (r = -0.62 for genetic diversity) and latitude (r = -0.77 for PGI-B3) were also observed, supporting the hypothesis of isolation by distance. No private alleles were found at the single population level.

Pollen and seed moverrient among subdivided popuIations in a tree speeies, eounteraet the effeet of genetic divergenee promoted by seIeetion and random genetie drift within small isoIated popuIations (HartI 1980).Spatial distribution, flowering phenology, ineompatibility meehanism, and pollen and seed dispersion patterns, all eontribute to genetie strueturing of populations.The status of a population at any moment and its rate of ehange will be the result of the eombination of all these factors (Hamrick and Murawski 1991).Alnus acuminata shows a midsuccessional species pattern, subject to repeated coIonization, extinction, and recolonization episodes associated to natural catastrophies.In such species, where neighborhood sites may be relatively small, interactions among neighborhoods in space and time may counteract any potential for genetic drift (Hamrick 1992in Boshieret al. 1995).SimilarIy as described by Cordia alliodora (Boshier et al. 1995), "from observation during rain on sloping terrain, the formation of temporary ri vulets will lead to additional seed dispersal and sometimes to c1umped areas of regeneration".Slnce such temporary water courses carry seed into rivers, dispersal distances in the order of kilometers are feasible.Such long-distance dispersal events are very important in coIonization and in breaking down , popuIation isoIation vía gene flow.
Gene flow and its ímportance in evolution has been debated for a long time under a variety of strong víews, as pointed out by Slatkin (1981): "One view states that gene flow is common and that a small amount of gene flow among different parts of a species' range effectively unifies th e species and affects signifieantly the genetie ehanges in eaeh part of the range".The other view is that gene flow is uncommon and that natural seleetion acts more or less independently in each part of the range of the species.Besides, the estimates of gene flow can be severely affected sinee there is always a non notieeable amount of gene flow, as well as, a large amount that may have migrated but not breed.Therefore, methods of estimatíng gene flow among populations have been proposed and diseussed by Slatkin (1981Slatkin ( , 1985) ) and found Jater that FST and rare-alleles methods yield comparable estimates under a wide variety of conditions.However, since there are practica!problems in estimating the frequencies of rare alleles in electrophoretic studies, especially when sampling diploid genotypes (Gregorius 1980), is therefore FST likely to be more useful under realistic conditions (Slatkin andBarton 1989, Slatkin 1993).
Gene flow is often analyzed under ecogeographical or elinal patterns of species' range distribution.The origin of clinal variation in natural populations has been debated since Mayr (1942 in Yang andYeh 1995) postulated a differentiation between zones of primary and secondary intergradation.Mayr defined prímary intergradation as elinal variation that develops within a continuous series of populations and secondary intergradation as the consequence of junction between populations that ha ve beeome differentiated in allopatry.Sorne empírical evidenee suggests that current patterns of allele frequency elinal variation reflect secondary contaet between populations that were isolated in refugia during the Pleistocene (post-Pleistocene secondary contact) and possessing di fferent prevalent alleJes (espeeially rare alleles) at a gene locus (Gallant et al. 1993, Hattemer et al. 1993, Konnert and Bergmann 1995).Besides, for several european tree speeies, an important genetic effect has been caused by forest utilization in the last eenturies (Hattemer, et al. 1993).However, other research suggest that elina! or eeogeographieal variation is mostly the result of spatial adaptation's patterns of enzyme phenotypie (genotypic) polymorphisms to gradually varying environmental faetors (Bergmann 1978, Bergmann andGregorius 1993), and thus, reflects pnmary intergradation.However, it is important to know whether the extent of gene flow between populations depends on their geographic distances, "isolation by distance", as originally proposed by Wright (1943 in Yang andYeh 1995).Therefore, it is intended to elucidate in this study the possible patterns of gene flow throughout the Alnus acuminata natural occurrence in Costa Rica and Panama, as well as the possibility of isolation by distance among the different geographieal regions.
Alnus acuminata is one of the most widely distributed species of the genus.It covers a large discontinuous geographical range throughout the Americas, stretching from central Mexico to northern Argentina (Lamprecht ] 990).This tree speeies was recently subdivided into three subspeeies, namely acuminata, arguta and glabrata (Furlow 1979a(Furlow , 1979b)).The speeies oceurs through wide ecological gradients, from around 1400 to 3200 m in elevation (Anonyrnous 1995, Furlow 1979a, Lamprecht 1990), from 2000 until 5000 mm in annual precipitatíon (Anonymous 1995, Coen 1983, Vargas 1994) and from deep and rieh vo1canie soils (andosols) to shallow, heavy-textured, low fertility and poorly draíned inceptisols (Anonymous 1995, Vázquez 1983) .These highly variable environmental conditions c1early enhanee large differences in the eeologieal eonditions over short distanees throughout the distribution of this tree species, which may have promoted an important amount of genetic variation among populations, with large potential benefits.The distribution of this species in Costa Rica is discontinuous, as it is found along the Central Volcaníc Mountain Range and the Talamanca Mountain Range, where the populations are separated one from another by valleys (Alvarez 1956, Camacho andMurillo 1986).

MATERIALS AND METHODS
Allozyme data: the 17 sampled populations represent geographicalIy and ecologically the distinct natural regions of Alnus acuminata ssp.arguta in Costa Rica and Panama (Anonymous, 1995, Murillo et al. 1993) as shown in Table 1.The various single populations sampled are distribuited along five natural population regions.Each single population is represented by a random sampie of 31 to 63 adult trees (with an average of 52 trees) separated by a minimum of 50 m distance.Different materials from each tree were electrophoretically assayed for variation in 10 enzymes systems.The samples were hand homogenized in a mortar after the addition of two drops of the following extraction buffer (Murillo 1997) (Cheliak andPitel 1984, Soltis andSoltis 1989) were initially tested for their suitability to improve the zyrnograms.The staining of enzymes was performed according to recipes given by Cheliak and Pitel (1984) and Soltis and Soltis (1989).The enzyme polymorphisms and its inheritance pattern was previously developed (Murillo and Hattemer 1997).
Gene flow assessment: Gene flow assessment was based on the methods of rare alleles and FST proposed by Slatk:in (1981Slatk:in ( , 1993)).There were considered the only four polymorphic gene loci.The total number of alleles observed in a population was (1) k == (he number of polymorphic alleles observed in a sample of n ij genes at the ith locus and the jth population.i == the ¡th locus (i == 1, 2, .. , l) The same procedures were utilized for the analysis of gene flow at the regions of populations level.Single populations were pooled in their regional level, according to their geographic distribution.The allele frequencies were divided into rare (p < 0.105 in average across populations) and common (p > 0.105 in average) alleles.The alleles were then registered in rare ( A R ) and total number of observed alleles (aT) ' Slatkin (1981) defines occupancy number of an allele (i) to be the number of demes (ni) in which it is presento Consequently, the conditional average frequency p (i) is the average frequency of all alleles having the same occupancy number It has been proposed that populations sharing rare alleles i .aremore likely to have a recent coancestry ei.ther because of mutations in recent history, 01.' because more distantly related populations ¡'o st such alleles following divergence and isolation (Bergmann and Gregorius 1993).Besides, it has been shown that the average frequency of an allele conditioned on the number of local populations it appears in (named as the conditional average frequency), is a sensitive predictor of gene flow, since it Ís independent of both the selection ' intensity and mutation rates (Slatkin 1981). .
The index of genet , ic similarity (M) from Slatkin (1993) lS estimated from the relationship: It has been demostrated to be a direct measure of gene exchange among populations, independently from the spatial arrangement of locations sampled and to be independent of mutation rate when it is small (Slatk:in and Barton 1989).Thus, it is appropriate for characterizing the extent and pattern of gene .dispersal between populations within a single species in an island mode!.In estimating M, FST was computed as a genetic divergence between all possible pairs of populations (136 pairs).If there was a pattern of isolation by distance, Slatkin predicted a linear relationship between loglo (M) and 10gIO (d) without assuming either one-dimensional or two dimensional dispersal (Slatkin 1993): Then, M was plotted against geographic distance (in km) on a lag-Iog graph in order to show this relationship.With this information the regression coefficient (b) was estimated using equatíon (4).A significant regression (b < O) would indicate the presence of isolation by dístance, as postulated by Slatkin (1993).However, there was no parametric test for significance of the b because values of M from different population pairs were not independent as required by the regression.Therefore, it was obtained an empirical distribution of b by bootstrapping as suggested by Yang and Yeh (1995) using the Statistical Analysis System (SAS).Each bootstrap sarnple was drawn from the data by sampling (with replacement) n times from the n pairs of 10gIO (M) and logIO (d), where n = 136 for the populations' data-structure and n = 10 for the region of populations' data structure.The estimate of b \ Vas' computed directly from the data and from each of 1000 bootstrap samples.A 95% confidence ínterval for b was constructed as the interval between the 26th to the 975th of 1000 otdered bootstrap estimates of b, after controlling normality in the dístribution of data.

Rare-alleles method:
There were no single private alleles found at the single population nor at the geographical regíon leve!.In Table 3 ít can be observed that there were only 3 alJeJes registered that were shared by a few populations, while the other 7 alleles were present in all populations.However no single population nor regíon contained all 10 polymorphíc alleles.The distributional pattern of rare ane\es wasvariable.Alle1ePGI-B 1 was observed only in Poás 1 and Poás II regions, while allele MNR-A2 was found in 10 populations belonging to different regions.At the regionallevel thereis a sligthly lower gene tlow arnong tegions, since there are more rare alleles concentrated in the lowest occupancy number (Table 3).
Within the Poás n and Talamanca regions the results suggest a very high gene tlow among their populations (Table 3).In both regions it occurred that all 10 single alleles were shared by all populations, with similar frequency values.Meanwhile, in Po a s 1 and lrazú regions an intermediate gene tlow pattern appeared, since there were sorne alleles found that were present only in a few or a single population.Resides, sorne contrasting allele frequencies were found in the populations.For instance, between Zarcero and Bajos del Toro populations (Poás 1 region) a strongsignificative difference was observed in gene locus PGM-A(X 2 = 10.236***).   5.In general, there were negative relationships among all geographical and genetic variables.Latitude did not correlate significantly with any of the gene tic variables, with the exception of a slightly but significative negative association to the percentage of heterozygosity in PGI-B.At higher elevation a weak tendency to diminution in percentage of heterozygosity in PGM-A was observed as well as in the degree of heterozygosity.On the other hand, longitude showed a significative negative assoclatlOn with subpopulation differentiation (Dj) and hypothetical gametic diversity (V gam), as well as tendency to reduction in percentage of heterozygosity in PGM-A and in the degree of heterozygosity.Meanwhile, altitude showed a high negative correlation with the number of multilocus genotypes, as well as a significative and negative association to diversity (V), total population differentiation (O T in Figure 2) and hypothetical gametic diversity (Vgam).Eventhough the relationship between subpopulation differentiation (at the regional level) and the geographic vector was not significative, due to the few degrees of freedom (r = -0.85,p = 0.065), the coeffícient of deterrnination obtained reached 74% (r2 = 0.74 in Figure 3).With an increment in altitude there was a tendency to reduction (negative association) in the proportion of heterozygosity in gene locus PGM-A observed as well as in the degree of heterozy gosity.�. 0,00 -1----+-----1-----1----+----1 7,00 8,00 9,00 10,00 11,00 12,00 A c1ear c1inal pattern in dependence of geographic variables was observed for sorne of the alleles at gene loci PGM and POI (Table 5).Meanwhile PGM-Al-allel frequency was negatively correlated with altitude (r =: -0.55*).PGM-A3 showed a strong positive association with altitude (Figure 4).Allel frequencies from POI-B2 and POI-B3 displayed opposing patterns with respect to latitude (Figure 5 .. lA ...  -+--+--+--+--+---1-----. 1----., 1) 1---- scale.Genetic similarity (M) and geographic distance (d) varied considerably among pairs of populations and regions (Tables 6 and 7).

Geographic vector
There was only one case for which M < 1 (between populations Boquete and Turrialba, M = 0.95, d = 222 km) and several very close Geographic distance (in Km) Fig. 6.Association between the genetic similarity (M) and the geographical distance (km) on 10 garithmic scales , for all possible pairs between the 5 geographical regions of natural occurrence for Alnus acuminata in Costa Rica and P anama.
10,00 1,00 Geographic distance (in Km) 1000 ,00  Genetic similarity values M (under the diagonal) and geographical distll(lces (above the diagonal in �) for all possible pairs of geographical regions of natural occurrence for Alnus acuminata in Costa Rica and Panama  Between pairs of geographical regions there were no single value of M < 5, indicat i ng a large gene flow at this level.
Estimates of regression of log 1 0 (M) on 10g I O (d) were in both cases negative and small (Table 8).Consistently, their confidence intervals were also negative and s ignificantly deviated from ze r o.Even through the broad range of values were all b negative, with a slight exception a t the regional level.

DISCUSSION
Clinál variation: The extreme variation in edaphic and climatic conditions among  1993), so that among-population fluctuation can o nly be partially explained by the geographic local adapt¡¡,tion of the populations.However, there was a weak but significative pattern of reduction in genetic parameters as long it goes south.This pattern of clinal variation is partially obscured by the disjunction between geographical occurence of this tree species within Costa Rica.The lack of a natural continuum may create sorne physical obstacules to gene flow among regions and populations.This can be observed in the results shown in Table 3.In the T alamanca and Poás 11 reg i ons still exists a continu i ty of natural forests among the local populations and a lower anthropogenic disturbance in comparison to the other regions.This is reflected in an equal occupancy number for all their alleles.If the Poás 1 and Poás II regions were pooled together, arguing that they are physically close to each other, then there would be a private allele (POI-BI with p = 0.075) for this pooled region.Then, the respective analysis would come out with a clear pattern of a tree species with an intermediate gene f10w (Slatkin 1981).
According to Slatkin (1981), in the high migration species (Nm > 1) there are numerous alle1es that have intermediate occupancy numbers with very low frequencies, as shown in Table 3.Then, the smooth form of the curve in Figure 1 corresponds to a high migration species.This type of situations support the arguments in favor of the FST over the rare-alleles method for studying gene flow, as has been latter pointed out by the same author (Slatkin and Barton 1989).The FST method uses al! of the gene-frequency data, thereby making it less sensitive to misreading electrophoretic gels for particular loci.Besides, it is les s sensitive of errors that would come with the identification of a private allele as belonging to another al!elic type, or like in this study, where different possible definitions of a region of demes could bring out totally different frequencies for their rare and private al!eles.In the same direction, the rare alleles method would be more sensitive to the common problem 01' sampling size and sampling procedures in genetic inventories (Gregorius 1980, Hattemer et al. 1982).Thus, the results obtained with the p (i) method in this study should not be considered as conclusive.The patterns observed tend to favor a model of a high gene f10w among populations within regions and a high-intermediate gene f10w among regions 01' populations.
Biogeographically, Alnus acuminata has mígrated from the northern to southern hemisphere throughout Latin America, leaving several disjunct groups of populations isolated during the glacial events at hígh elevations (Furlow 1979a).The dynamics of repeated glacial and interglacial events during the last million years may explain Alnus migration through the Central American filter (Gómez 1986, Rich andRich 1983), as well as the loss and reestablishment of the contact between populations.Today, there is a clear pattern 01' several isolated regions 01' populations along the mountains in Costa Rica and Pan ama.This implies the loss of a continuum distribution.Thus, it is possible to think in a pattern of north to south of slow reduction in genetic variability along its geographical distribution in Costa Rica and Panama.
The altitudinal pattern 01' reduction in genetic variability with elevation could be explained by the frequency of natural catastrophies.North 01' these mountain range several active voIcanoes are located that have produced repeated events 01' major impact during the last centuries (Al varado 1994).This is particular clear in the Irazú region, where the last eruptions destroyed a large portion of Alnus habitats, enhancing severe and very recent bottlenecks in the populations from this area.Similarly, the relatively high frequency of earthquakes episodes in Central America during the last centuries (Al varado 1994) has produced large impact and landslides up in the mountain ranges.Therefore, it may partially explain the reduced levels of genetic variation in populations from Irazú region, and clearly for the populations located near the summit of the mountains (Irazú, Turrialba, and Siberia populations).Also, with an increase in altitude there is a decrease in avaiJable surface.It is clear that the chances of establishing population depends largerly on suitable habitat areas (Mosseler 1995).Therefore, it is possible to think about a direct association between population size potential as well as population perpetuation potencial with available surface.Besides, with an increase in elevation there is a tendency in reducing the length 01' the dry season (Vargas 1994).During the dry season most of the pollen shedding in this wind pollinated species occurs (Vílchez and Murillo 1994), which could be severely reduced during the rain season.

T ABLE 3
Occupancy numbers and conditional average frequenciesJ) detected for JO al/e/es al four polymorphic loó in na illral populalions of Alnus acuminata in Cosla Rica and Panama According to Slatkin (1981) 2) n = number of populations within (he regionAllele PGM-A2 was found only in the Bajos del Toro population and allele PGM-Al showed a deviation in frequency between both populations, larger than their respective confidence intervals.Similarly, among populations within the Irazú region a significative difference (X2 test) in all gene loci was found.

6
Patterns 01 genl! dispersal among populations and regions: Estimates of correlation between geographic (latitude, longitude and altitude) and genetic variables are presented in Table Fig. l.Oecupaney numbers and eonditional average frequencies detected for 10 afieles at four polymorphic loei in 17 natural populations of Alnus acuminata in Costa Rica and Panama.

Fig. 3 .
Fig. 3. Association between the subpopulation differentiation and a geographic vector in natural populations of Alnus acuminata from five geographical regions in Costa Rica and Pan ama (r = -0.85,p = 0.065).

Fig. 7 .
Fig. 7. Association between the genetic similarity (M) and the geographical distance (km) on logarithmic scales, for áll possible pairs between the 17 natural populations of Alnus acuminata in Costa Rica and Panama.
adjusted with

TABLE 5
Pearson correlations between genetic and geographic variables in 17 natural populations of Alnus acuminata from Costa Rica and Panama (P values in parenthesis)

TABLE 7
r

TABLE 8
Statistics (Mean, range and conjidence intervals) of the estima tes (obtained by bootstraping) of intercept (a) and slope (b) from the regression of genetic similarity (M) on geographic distance (d) on a log-log scale in Alnus acuminala populationsfrom Costa Rica and Panama