ECOTYPIC AND ALLOZYME VARIATION OF CAPSELLA BURSA-PASTORIS AND C. RUBELLA (BRASSICACEAE) ALONG LATITUDE AND ALTITUDE GRADIENTS ON THE D3ERIAN PENÍNSULA

NEUFFER, B. & R. HOFFROGGE (2000). Variación ecotípica y alozímica de Capsella bursapastoris y C. rubella (Brassicaceae) a lo largo de gradientes latitudinales y altitudinales en la Península Ibérica. Anales Jard. Bot. Madrid 57(2): 299-315 (en inglés). Se han comparado diversos caracteres biológicos relacionados con la capacidad colonizadora (inicio de la floración, número de hojas, diámetro de la roseta, altura de la planta, número de ramas, dimensiones del fruto y número de semillas) de plantas de Capsella procedentes de la Península Ibérica mediante un experimento de bloques aleatorios en el campo. Los datos se evaluaron con un análisis de componentes principales. También se registraron el tipo de hojas y el perfil aloenzimático de las plantas. Las plantas de Capsella bursa-pastoris procedentes de altitudes altas y bajas de la zona climática Mediterránea de verano seco (Sierra Nevada) mostraron ser de floración temprana, mientras que las plantas de los Pirineos, con un clima alpino, presentaron una floración tardía. En C. bursa-pastoris el tipo de hoja "rhomboidea" resultó ser el más frecuente, en tanto que en C. rubella lo fue el tipo "heteris". Se observó un cambio en las frecuencias de los tipos de hojas a lo largo de una clina geográfica, lo que se explicaría por los componentes adaptativos que posee la forma de la hoja. Las aloenzimas presentaron un patrón de distribución geográfico y en C. bursa-pastoris un determinado genotipo multilocus podría ser un marcador molecular para el ecotipo de floración temprana.


INTRODUCTION
Colonizing species have received considerable attention during the past decades (BAKER, 1965;CLEGG & BROWN, 1983; BARRETT & RICHARDSON, 1986), and they share a number of common genetic features often including polyploidy (BROWN & MARSHALL, 1981).Their adaptation may be due to a "general purpose genotype" of high phenotypic plasticity (BAKER, 1965) or pronounced ecotypic variation.It has been shown that both concepts are not mutually exclusive.
A strong correlation between the time to flowering and elevation above sea level was observed for Capsella populations from European alpine regions.Alpine populations comprise an array of genotypes with different susceptibility to low temperatures (NEUFFER & BARTELHEIM, 1989).It was shown that summer annual "early flowering" genotypes were replaced by winter annual "late flowering" genotypes along an altitudinal gradient.This topocline was paralleled by an ecocline expressed as a shortening of the period of vegetative growth.Furthermore with higher elevation the plants remain smaller and the frequency of to the midrib dissected leave lobes increased (NEUFFER & BARTELHEIM, 1989).A similar pattern of variation was also observed for extra-European regions (e.g.Sierra Nevada California, USA, NEUFFER & HURKA, 1999).Differentiation along altitudinal gradients studied so far concentrated on températe climates.Question arises whether in Mediterranean climate types variation patterns along altitudinal gradients are basically different from those in températe climates.
The Iberian Peninsula is subdivided into two major vegetation zones: the colddeciduous mesophytic broadleafed forest in the north and the sclerophyllous forest in the south (SCHMITHÜSEN, 1976).The Pyrenees in the northern part have an alpine climate, whereas the southern part of the central mountains and the Sierra Nevada in the south with an elevation above 3000 m show a summer-dry Mediterranean climate.
The present study highlights the genetic composition and adaptation strategy in populations of C. bursa-pastoris and C. rubella Reut.from different mountain, valley and coastal habitats on the Iberian Peninsula.C. rubella is a diploid species in contrast to C. bursa-pastoris which is tetraploid.Both species are closely related.The geographical range of C. bursa-pastoris is much wider than that of C. rubella which has been attributed to the greater genetic flexibility gained by the polyploidization (NEUFFER & ESCHNER, 1994).In the Mediterranean región, both species often occur together in the same sites.It will be demonstrated that ecotypic differentiation of Capsella in summerdry Mediterranean mountains is different from variation patterns observed in mountains with alpine climates.

Provenances, sowing andgrowing conditions
Seed samples of up to 10 individual plants of 25 populations of Capsella (Brassicaceae) were randomly taken from natural populations from different regions of the Iberian Península (Table 1, Fig. 1).Three of these populations (No. 1249, 1252, 1290, Table 1; • in Fig. 1) were "two species sites" (C.bursa-pastoris and C. rubella growing together), the others pure C. bursa-pastoris stands.Seeds were stored for at least six months at minus 20 °C.Progenies raised from the mother plants were grown in a random block design in an open field experiment at the Botanical Garden of the University of Osnabrück, Germany (52° 18' N, 8 o 00' E, 90 m above sea level).Seeds were germinated in plástic pots (7 cm in diameter) on top of a substrate mixture containing TKS1 (Torfkultursubstrat) and crystal sand (2:1) in an unheated glasshouse on April 20, 1995.Young plants were planted into the experimental field on May 29, 1995.It was intended to analyze 10 families (= progenies) per population and 10 individuáis per family.Different group sizes resulted from either low germination (fewer than 10 individuáis per family), low survival of seedlings, or small population sizes (fewer than 10 plants per population).

Quantitative parameters
The following parameters were recorded in tiie experimental field: -onset of flowering: time from sowing to breaking the first flower bud (Fl); -rosette maximum diameter, recorded in most cases shortly after the onset of flowering (RD); -number of rosette leaves at maximum rosette diameter (LN); -total plant height of the inflorescence (PH) at the end of the growing season; -branching number: number of rosette shoots (BB) and number of stem branches (SB); -fruit dimensions: width (FW) and length (FL); -number of fruits per 10 cm of one well developed branch (FN).The number of stem branches was observed  in one individual from each family.The other parameters were observed in each individual.

Statistical evaluation
Mean valué, standard deviation, range, median valué (m) and the coefficient of variability (cV%) were computed.Since the data did not show a normal distribution (KOLMOGOROV-SMIRNOV Test) and variances were not equal, a non-parametric one-way analysis of variance for unbalanced group numbers was employed (H-test of KRUSKAL & WALLIS).The test assesses the phenotypic homogeneity (-) and heterogeneity (+) within and between populations.Plant-specific parameters including median valúes and coefficient of variability have been evaluated by a principie component analysis (PCA) and a regression analysis.The data were statistically evaluated with the SPSS program (Superior Performance Software System, Versión 6.0, ANOVA).

Quantitative parameters
All populations were heterogeneous for at least one character (Table 2).One population was homogenous with only one exception (fruit length, pop.1283, Sierra Nevada), other populations were variable in most characters (pop. 830, 1229, 1238, 1281, 1289, 1291, 1295, and Capsella rubella plants of 1249).For all C. bursa-pastoris plants the rosette diameter was the most homogeneous (Table 2, Total, cV% 14.6) and the rosette leaf number the most variable character (cV% 43.4).For C. rubella onset of flowering was the most homogeneous (cV% 9.4) and plant height the most variable character (cV% 47.7).The populations 1252, 1278 and 1283 showed no significant intrapopulational differences for onset of flowering which in general was the most variable character (Table 2).14 from 25 populations were homogenous in fruit number.
In comparison with C. bursa-pastoris, C. rubella plants were generally flowering later, developed more rosette leaves, and the fruits were smaller.A regression analysis (Table 3, Fig. 4) demonstrated for both species a high positive correlation between    Populations from the northern part of Spain were slightly later flowering and more variable (Fig. 4).
Figure 5 demonstrates the relationships between populations and the three factors given by PCA analysis.The Atlantic population 1249 and the Pyrenean population 830 are separated from the other populations by factor 1 but pulled apart by factor 2. The populations from the different mountain regions form subgroups.The Sierra de Gredos populations and the Sierra Nevada populations are connected by the overlapping group of the Sierra Cazorla.The Sierra de Gredos populations are more similar to the Pyrenean populations than to the more southern populations.The populations from the mesetas (named "Central") are in between.The Sierra Nevada population 1282 from the highest elevation (SNevada Mountain, Fig. 5) is different from the other populations of the Sierra Nevada región.

Leaf morphology
Capsella rubella plants were predominantly "heteris" (110 individuáis "heteris", 40 individuáis "rhomboidea"), whereas within C. bursa-pastoris the leaf type "rhomboidea" was prevalent (1235 out of 1944 individuáis).The "tenuis" type occurred in very low frequency (only 6 individuáis within C. bursa-pastoris, none within C. rubella).The frequency of "rhomboidea" type increased in the mountainous regions (Pyrenees, Sierra Cazorla, Sierra de Gredos, Sierra Nevada, Fig. 3).The "heteris" type was frequent in Atlantic and central regions and in the Sierra de Gredos (Fig. 3).The "simplex" type predominanted at the Sierra Nevada, Cazorla and central regions, and disappeared completely at the Atlantic coast in favour of the "heteris" type (Fig. 3).It was also rare in the Pyrenees and in high mountain regions of the Sierra Nevada.A geographic differentiation pattern was obvious (Fig. 3).

Allozyme analyses
The allelic multilocus structure in the C. bursa-pastoris populations from Spain exhibited low variability (Table 4).Only the AatlB locus, the Lap3B locus and the Gdh2A locus varied considerably between the populations.Some rare alíeles seemed to be restricted to the central and northern part of Spain (e.g.Aat3A-3, Gdh2B-3, Table 4).

Geographical distribution and characterization ofthe populations
The populations observed in this study originated from different vegetation zones, climatic regions, and habitats as shown in Table 1.Population 830 from the Pyrenees was late flowering with large rosette diameter and predominantly "rhomboidea" leaf type (Fig. 3), but remained small with few branches and many small fruits (Table 2).Some allozymes occurring in the Pyrenees were not observed further south except for the Sierra de Gredos (Aat3B-2, Gdh2B-3, 1229, 1295,1301, Table 4).
The populations 1228, 1229, 1295-1301 from the Sierra de Gredos were early to intermediate flowering and were generally tall, but developed only a small number of rosette leaves with different leaf types (Fig. 3).
Pop. 1249 from Sintra (Lisbon) was a "two species site".All plants of both species were late flowering, had small rosettes predominantly with "heteris" leaves, and they remained small with a small number of branches.Their inflorescences had many but small fruits.C. rubella of population 1249 from the western Atlantic coast of Portugal had the most common isoenzyme multilocus genotype of this species (HURKA & NEUFFER, 1997).
The populations 1234,1238 and 1252 from central Spain exhibited various leaf types (Fig. 3).The population 1252 was a "two species site".At this collecting site (Jerez de los Caballeros) we detected one hybrid between C. bursa-pastoris and C. rubella, which is named as C. gracilis Gren. in the literature.This hybrid produced only sterile fruits and therefore was easy to determine.Some C. rubella plants showed an unusual isoenzyme multilocus genotype at the Lap3A -and Gdh2A-loci (Lap3A-ll, instead of Lap3A-66, Gdh2A-22 instead of Gdh2A-U, Table 4) and were later flowering than C. bursa-pastoris.These C. rubella plants were tall and bushy with a few but large rosette leaves and few small fruits.
The populations 1272, 1278, and 1279 from the south coast were early flowering with a relatively small number of rosette leaves showing a high amount of "simplex" types and a low number of small to intermediate fruits.
The populations 1288-1291 from the Sierra Cazorla exhibited predominantly the "rhomboidea" leaf type (Fig. 3).Population 1290 was also a "two species site".Both species had the most common Mediterranean isoenzyme multilocus genotype.
The populations 1280-1287 from the Sierra Nevada were early flowering with an intermediate rosette diameter and a very low number of rosette leaves.Plants produced an intermediate number of large fruits.The plants grew tall with the exception of 1282, which had the smallest rosette diameter, and the lowest number of branches but produced the highest number of fruits and a high number of "rhomboidea" leaves.This population originated from the highest point of all collected sites (2000 m).This stays in contrast to the findings in Rhinanthus glacialis PERSONNAT which is earlier flowering when populations are derived from higher elevations (ZOPFI, 1995).In contrast to our findings from other mountainous regions like the Alps in Europe (NEUFFER & BARTELHEIM, 1989) or the Sierra Nevada in North America (NEUFFER & HURKA, 1999), C. bursa-pastoris was early flowering at high elevations in the Sierra Nevada (pop.1282, Table 2).In alpine climate the higher the altitude of origin, the later the onset of flowering and the more summer annual genotypes were replaced by winter annual genotypes.In the Sierra Nevada, no such clinal variation was observed.We attribute these differences to the different climates in the Alps and in the Sierra Nevada.The latter mountains are summer-dry and have short mild winters.Provenances of the Sierra de Gredos displayed intermediate flower onset (Figure 5).This might be the result of climatic factors which are not as extreme as in the Sierra Nevada (Spain) or in the Pyrenees.
Growth /orms.-Growthperformance of plants from the Pyrenees (pop.830) was similar to what is observed in other mountains appearing low and bushy with large rosettes and many small fruits (NEUFFER & HURKA, 1986b; NEUFFER & BARTELHEIM, 1989).Plants from high elevations in the Sierra Nevada, however, grew taller, had smaller rosettes, and were less branched.They resembled those from the Sierra Nevada valley populations.

Leafmorphology
Leaf form has implications for thermoregulation, water-use efficiency, photosynthesis, productivity, branching and rooting strategies, and competitive ability (GATES, 1968;PARKHURST & LOUCKS, 1972;TAYLOR, 1975;GIVNISH, 1979, DUDLEY, 1996).A subdivided leaf is believed to decrease the "effective leaf size", thereby reducing overheating and transpiration in dry and exposed conditions, while a continuous leaf surface increases the photosynthetic área and allows the leaf to warm up above air temperature in cool and shady conditions (GIVNISH, 1987).
Most of the C. rubella plants showed the "heteris" leaf type, which is in full accordance with the findings of ALMQUIST (1907,1929), SHULL (1909,1929), NEUFFER & ESCHNER (1995), and NEUFFER & ALBERS (1996).In C. bursa-pastoris, the "rhomboidea" leaf type occurred at a very high frequency (Fig. 3).Frequency of this leaf type is high in general and the distribution is world-wide, in contrast to "tenuis" which is of low frequency in general (HURKA & NEUFFER, 1997).The present study is in accordance with the general picture (Fig. 3).This unequal distribution of the leaf types may be due to the higher adaptive valué of the "rhomboidea" leaf type in regard to water use efficiency or may be just a stochastic phenomenon.In favour of a non-stochastic distribution of the "rhomboidea" leaf type within the Sierra Nevada región is the observation that in lower elevation the leaf types "heteris" and "simplex" occurred.These types disappeared in the mountains (Sierra Nevada and Pyrenees, Figure 3, see also NEUFFER & BARTELHEIM, 1989) and were replaced by the "rhomboidea" leaf type.

Allozyme analyses
Most Capsella bursa-pastoris plants studied are characterized by the multilocus genotype AatlA-11 ).This genotype was named the Mediterranean Multilocus Genotype (MMG) because of its prevalence in the Mediterranean área (NEUFFER & HURKA, 1999).It coincides with an early flowering genotype (Table 2 and 4 may also be govemed by temperature and humidity regimes.In general, it would appear that certain variation patterns of quantitative traits in Capsella on a macro-and microgeographical scale are adaptive despite all the random processes due to its colonizing abilities.It is unlikely, however, that the observed variation pattern in isozyme allelic composition is also the outcome of direct environmental selection pressures.We rather interpret our data by a linkage between isozyme genotypes and phenotypic traits.
Analyses are currently under progress to investígate cosegregation of allozymes and quantitative traits.

Fig
Fig. l.-Map showing the locations of the different Capsella populations Usted inTable 1: • populations including C. bursa-pastoris and C. rubella individuáis; • populations with C. bursa-pastoris individuáis only.
well-developed leaf of each individual was deposited in the Herbarium at the University of Osnabrück (OSBU).
1.4) by two nucleic loci.Details as to the extraction, electrophoresis and visualization of the enzymes and as to the genetic interpretation of the zymograms are gi ven in HURKA & al. (1989) for AAT, HURKA & DÜRING (1994) for GDH, and HURKA & NEUFFER (1997) for LAP.All loci are duplicated in the tetraploid C. bursa-pastoris (HURKA & NEUFFER, 1997).The allozyme patterns of the two species Capsella bursapastoris and C. rubella are different, and both species can easily be keyed out.
m = median valué over all individuáis; cV% = coefficient of variability; n = number of individuáis Fl = onset of flowering; LN = rosette leaf number; RD = rosette diameter; PH = plant height; BB = number of basal branches: SB = number of stem branches; FL = fruit length; FW = fruit width; FN = number of fruits.Results of H-test within and between populations (ANOVA, a < 0.05), + = significant differences, -= no significant differences
offlowering.-Theonset of flowering in Capsella is influenced by geographical, climatic and biotic factors (for a review see HURKA & NEUFFER, 1997).The greater flexibility of C. bursa-pastoris compared to C. rubella was reflected by a diverse array of flowering ecotypes: Early flowering types in the summerdry and hot Mediterranean climate in the South of Spain (populations 1272-1291), and late flowering types in the alpine climate of the Pyrenees (pop.830).
).It stands for discussion to what extent selection will influence the geographic pattern of allozyme variation.In a study of colonial Californian and ancestral Spanish populations of Avena barbata L. it was argued that the main forcé responsible for geographic patterns of isozyme multilocus structures was selection for particular combinations in response to different environments (PÉREZ DE LA VEGA & al., 1991; ALLARD & ai, 1993).AINOUCHE & al. (1996) detected correlations of isozyme multilocus composition with climatic variables in Bromus lanceolatus Roth, and EMERY & al. (1994) observed distinct clusters of allozymic variation that observed in Mediterranean climate.Temperature, humidity and radiation during the summer period differ sharply between alpine and Mediterranean mountains.These different ecological regimes apparently favour different ecotypes and thus would explain the different variation patterns in alpine and Mediterranean mountains.most striking differences concerns the onset of flowering: whereas in alpine climates late flowering replace early flowering ecotypes along an altitudinal gradient, no such replacement occurred in Mediterranean mountains.High mountain populations of Capsella comprise early flowering plants only.Frequencies of leaf morphology types