ECOLOGICAL DIFFERENTIATION AND DIPLOID SUPERIORITY ACROSS A MOVING PLOIDY CONTACT ZONE

Authors


Abstract

Plant polyploid complexes provide useful model systems for distinguishing between adaptive and nonadaptive causes of parapatric distributions in closely related lineages. Polyploidy often gives rise to morphological and physiological changes, which may be adaptive to different environments, but separate distributions may also be maintained by reproductive interference caused by postzygotic reproductive isolation. Here, we test the hypothesis that diploid and descendent polyploid races of the wind-pollinated herb Mercurialis annua, which are found in parapatry over an environmental gradient in northeast Spain, are differentiated in their ecophysiology and life history. We also ask whether any such differences represent adaptations to their different natural environments. On the basis of a series of reciprocal transplant experiments in the field, and experiments under controlled conditions, we found that diploid and polyploid populations of M. annua are ecologically differentiated, but that they do not show local adaptation; rather, the diploids have higher fitness than the polyploids across both diploid- and polyploid-occupied regions. In fact, diploids are currently displacing polyploids by advancing south on two separate fronts in Spain, and previous work has shown that this displacement is being driven to a large extent by asymmetrical pollen swamping. Our results here suggest that ecophysiological superiority of the diploids may also be contributing to their expansion.

Polyploidy has contributed substantially to the diversification of land plants, with up to 70% of extant lineages estimated to have a polyploid history and more than 4% of speciation events likely to have involved genome duplication (Otto and Whitton 2000). The clear importance of polyploidy in land-plant evolution, however, stands in stark contrast to the difficulty that new polyploids are expected to have in becoming established. Because mating between polyploids and diploids typically gives rise to sterile offspring, the establishment of a new polyploid lineage in sympatry with its diploid progenitors can be severely compromised by the process of minority cytotype exclusion, whereby the rarer cytotype (typically the new polyploid lineage) mates mainly with incompatible partners (Levin 1975; Husband 2000). Unless the new polyploid also evolves ways of avoiding mating with sympatric diploids, for example, through self-fertilization or through differences in flowering time, successful polyploid establishment will necessarily depend on the occupation of habitat that is geographically separate from that of the related diploids (Stebbins 1971; Rausch and Morgan 2005). It is thus not surprising that related lineages with different ploidy levels typically exist in allopatry or parapatry.

Given that the different ploidy levels of a species complex are commonly spatially separated, we might expect them also to occupy different environments, and thus to be ecologically divergent from one another. This might occur in two ways. First, polyploids that are dispersed into a new environment may respond to natural selection and become locally adapted. Certainly, polyploids may evolve at least as rapidly as their progenitor diploids, because they have greater potential genetic variation on which selection can act (Bingham 1980; Stebbins 1980; Bever and Felber 1992; Soltis and Soltis 1993). Moreover, isolation of polyploids from their diploid progenitors will protect them from the effects of gene flow, which otherwise can act to prevent adaptation at range boundaries (Kirkpatrick and Barton 1997; Lenormand 2002; Pannell et al. 2004).

Second, polyploidization itself might confer upon the new lineage a predisposition toward survival in certain new environments, that is, polyploids might be preadapted to a new environment. For instance, the so-called gigas traits (i.e., large cells and organs, Randolph 1941; Bretagnolle et al. 1998) and/or slower rates of cell division that often follow polyploidization (Noggle 1946; McArthur and Sanderson 1999) can directly affect a polyploid population's ecophysiology (Stebbins 1971; Masterson 1994; Ramsey and Schemske 2002). Polyploidization alters gene dosage and gene expression (Guo et al. 1996; Hegarty et al. 2005), which can change phenotypes and may cause the breakdown of self-incompatibility systems (Lewis 1960; Stone 2002), or altered sex determination (Westergaard 1958); these latter changes can afford polyploids with an increased ability to colonize new marginal habitats (Hagerup 1931; Löve and Löve 1943; Stebbins 1950; Ehrendorfer 1980; Vogel et al. 1999; Pannell et al. 2004).

If polyploidization per se preadapts polyploid lineages to certain environments, we might expect nonrandom associations between ploidy and the environment. Some patterns have indeed been found in the distribution of recently formed polyploid plants, such as their increased frequency at high latitudes relative to their diploid progenitors (Stebbins 1950; Johnson and Packer 1965; Jackson 1976; Brochmann et al. 2004) or their occurrence in marginal habitats (Löve and Löve 1943; Stebbins 1950; Vogel et al. 1999). However, there are many exceptions, and no clear pattern has yet emerged (Stebbins 1971; Lewis 1980; Thompson and Lumaret 1992).

Regardless of whether genome duplication per se has contributed to ecological divergence among polyploid lineages, any subsequent process of adaptation is predicted to give rise to a “home-range” advantage to each respective ploidy level as a response to natural selection locally (e.g., Clausen et al. 1940; Linhart and Grant 1996; Wang et al. 1997; Fritsche and Kaltz 2000; Bronson et al. 2003). Such local adaptation is most effectively and convincingly revealed by reciprocal transplant experiments (RTEs), whereby genotypes from different localities are transplanted among localities, including that of their own provenance. The RTEs assess adaptation by a genotype to a broad spectrum of its possible environmental range, including natural disease challenges, and can show local adaptation at both large (Schmidt and Levin 1985; Jordan 1992) and small spatial scales (Antonovics 1976; Schmitt and Gamble 1990; Bell et al. 2002).

Evidence of local adaptation has been inferred for several diploid–polyploid contact zones (Blaise et al. 1991; Felber-Girard et al. 1996; Husband and Schemske 1998; Hardy et al. 2000), but RTEs have, to our knowledge, only been carried out between diploids and descendent polyploids in the forb Ranunculus adoneus (Baack and Stanton 2005) and the grass Anthoxanthum alpinum/A. odoratum (Flegrová and Krahulec 1999). In R. adoneus, diploid and autotetraploid populations show strict spatial segregation in alpine Colorado (Baack 2004). Here, RTEs involving 10 populations over three years found no evidence for local adaptation between the cytotypes during the early phases of establishment (Baack and Stanton 2005). In Anthoxanthum, the diploid A. alpinum and its tetraploid relative A. odoratum occupy contrasting ranges in Europe. However, results from RTEs involving two populations over two years suggested a higher fitness for A. odoratum in both environments and gave little clear evidence of differential local adaptation (Flegrová and Krahulec 1999). We are thus still far from a general understanding of the relative importance and interactions of adaptive and nonadaptive processes as regulators of the distribution of different ploidy levels within polyploid complexes.

In this study, we use RTEs to test competing hypotheses regarding the distribution of lineages in the polyploid complex Mercurialis annua, in which diploid populations are parapatric to hexaploid populations at two contact zones in northwest and northeast Spain. Adaptive ecological divergence between these ploidy levels was hypothesized by Durand (1963), who noted that the hexaploids are generally found in warmer and drier areas than diploids in both Spain and North Africa. In support of his hypothesis, Durand (1963) recorded broad phytosociological associations between the hexaploid populations and species characteristic of drier sites, and between the diploid populations and species characteristic of moister sites. He also noted that several other species had their range boundaries close to the diploid–hexaploid transition (Durand 1963).

Adaptive ecological divergence in the M. annua complex could be due to an association between ploidy level and sexual system (Pannell et al. 2004), confirmed across the geographic range of M. annua by means of chromosome counts (Durand 1963; Obbard et al. 2006a) and measurements of C-values (Obbard et al. 2006a). Diploids are exclusively dioecious and polyploid populations variously monoecious or androdioecious (where males co-exist with monoecious individuals). One idea is that these differences are the result of selection at the metapopulation level: in environments where populations are small and turnover is high, recurrent colonization favors monoecy (and thus hexaploidy) because of the advantage of reproductive assurance through self-fertilization (Pannell 1997; Pannell and Barrett 1998); in contrast, dioecy (and thus diploidy) might be favored in environments where populations are larger and more stable (Pannell 1997, 2000; Barrett and Pannell 1999). There is some support for the metapopulation hypothesis (Eppley and Pannell 2007; Obbard et al. 2006b), but the hypothesis that adaptive ecophysiological differences also exist between ploidy levels has not previously been subjected to empirical test.

Two recent findings provide an additional framework within which to interpret results of RTEs. First, patterns of genetic diversity indicate that the diploid–hexaploid contact zones of M. annua in northern Spain are of secondary origin, with the diploids and hexaploids having migrated from a southeastern European and a southern Spanish or north African refugium, respectively (Obbard et al. 2006b). And second, both of the diploid–hexaploid contact zones in Spain are currently moving, with the diploids displacing the hexaploids as a result of asymmetrical reproductive interference and pollen swamping associated with differences in their sexual systems (Buggs and Pannell 2006, and see below). Diploids therefore occupied their current range in northern Spain relatively recently and might be expected to be less well adapted there, ecophysiologically, than the hexaploids that they displaced by pollen swamping.

Our study here used RTEs at both the seed and seedling levels to test the hypothesis that diploid and hexaploid populations are locally adapted across the contact zone in northeast Spain. Because of logistical difficulties involved in RTEs at the seed level, our seed RTE used seed sourced from several diploid and several hexaploid populations, pooled respectively, and grown at each of 15 localities on either side of the current ploidy transition. The seed RTE thus accounted for variation among sites within regions, but it did not assess variation among source populations within ploidy levels specifically. In contrast, in the seedling RTE, we established seedlings at only three sites on either side of the ploidy transition, but here we accounted for variation among source populations within ploidy level (rather than just between individuals). Presumably for logistical reasons, RTEs have often only considered one site, or one source population, per region (e.g., Antonovics and Primack 1982; Jordan 1992; Bennington and McGraw 1995; Flegrová and Krahulec 1999; Fritsche and Kaltz 2000; Eckhart et al. 2004; Baack 2005). In contrast, our seed and seedling RTEs increased sampling over planting sites and seed sources, respectively; together they allow well-supported inferences about the evolution of local adaptation between the lineages of interest. Because we had a priori reasons to expect differences in drought tolerance between diploids and hexaploids, with the latter expected to be more drought tolerant (Durand 1963), we augmented our RTEs with drought experiments under controlled conditions to test this hypothesis. To examine the effect of the same ploidy difference in different environments, we used seed from both the northeast Spain and the northwest Spain contact zones in glasshouse experiments.

Materials and Methods

STUDY SPECIES AND POPULATIONS

Mercurialis annua s.l. (Euphorbiaceae) is a wind-pollinated annual plant occupying ruderal habitats throughout central and western Europe and around the Mediterranean Basin (Tutin et al. 1964). Throughout the majority of its range, M. annua is diploid, with hexaploids existing in the coastal lowlands of southern Spain, and tetraploid and higher ploidy levels found in areas of North Africa and Mediterranean islands (Durand 1963; Pannell et al. 2004). In mesic climates, plants flower throughout the year, but the species is a winter annual in the Mediterranean region. Primary seed dispersal is ballistic, with secondary dispersal by ants (Lisci and Pacini 1997) and humans. This study was based on natural populations visited in the winter of 2002. Most were growing on recently disturbed soil on roadsides, building sites, rubbish dumps, arable fields, and neglected gardens and olive groves. Occasional populations were found in walls and rocky crevasses. Seed was collected by harvesting large numbers of whole adult plants that were dried in perforated plastic bags, threshed, sieved, and winnowed. The ploidy of all seeds was inferred from the location and sexual system of their parental populations. If plants were established from the seed in the experiments below, the ploidy of source populations was confirmed by the measurement of C-value by flow cytometry, following the protocol of Obbard et al. (2006a).

RECIPROCAL TRANSPLANT OF SEEDS IN NORTHEAST SPAIN

We sourced seed from multiple families in three populations located in the hexaploid area, about 80 km southwest of the contact zone (labeled HEX-far), and three populations in the diploid area, 90 km northeast of the contact zone (DIP-far). Locations are shown in Figure 1 and Table 1. We sowed seeds directly into disturbed soil at 30 sites in November 2002, with 15 sites selected in area HEX-far and 15 in area DIP-far (Fig. 1). We did not use sites currently occupied by M. annua, because a rain of local seed would have contaminated our plots. Instead, we selected empty sites occupied by the same suite of ruderal species typically found growing with M. annua (R. J. A. Buggs and J. R. Pannell, unpubl. ms). It is of course possible that the sites chosen were in some cryptic way unsuitable for M. annua, but this is unlikely. Moreover, because we used the same criteria to select sites in all transplant areas, our possible ignorance is unlikely to have been biased. At each site, we cultivated five 50 cm × 50 cm square quadrats with a fork, and in two we planted 200 seeds from a single source area, so that each of the source areas had its own quadrat. We left the third quadrat empty as a control to measure the seed bank of the site. The seeds we planted from each source area were pooled in equal numbers from the three populations (see Introduction for justification). At each site we augmented the RTE by planting, in the two remaining quadrats, seed sourced from three hexaploid and three diploid populations at the contact zone (Fig. 1; labeled HEX-near and DIP-near, respectively) using the same sampling and planting method as above.

Figure 1.

Map showing location of seed source populations, seed transplant sites, and seedling transplant sites. Filled circles show diploid source populations; filled squares show hexaploid source populations; crosses show seed transplant sites; and small unfilled squares show seedling transplant sites. Areas are annotated with the labels used in the text. Inset: the Iberian Peninsula showing location of the study region and locations of contact zones in 1959 (broken lines) and 2003 (solid lines).

Table 1. Mercurialis annua populations from which seeds were sourced for experiments, showing their location, label, mating system (D: dioecious; A: androdioecious; M: monoecious), 4C-value and standard error, the number of plants for which the 4C-value was measured, and the experiments in which seeds from the source populations were used (1: seed RTE; 2: seedling RTE; 3: first glasshouse experiment; 4: second glasshouse experiment).
Latitude (°N)Longitude (°E)Region of SpainRTE seed source labelMating system4C value (pg)4C value SE4C sample sizeExperiment
  1. 1Confirmed as octoploid by chromosome count.

43.35226−8.40293NW D –3,4
43.21199−8.94477NW D 2.600.01053,4
43.14527−9.08387NW A 7.530.01743,4
43.11745−9.15203NW D 2.590.00753,4
43.09629−9.17756NW A 7.430.02653,4
42.91723−9.14459NW A   3,4
41.27204 1.96954NEDIP–farD 2.860.01431,2
41.26993 1.97654NEDIP-farD 2.900.02131,2
41.26428 1.92745NEDIP-farD 2.750.01131,2
41.14625 1.24371NE D   2,3
41.04990 1.00418NEDIP-nearD 2.680.01131,2
41.03560 0.96737NEDIP-nearD 2.460.01351,2,3,4
40.99485 0.92318NEHEX-nearM 7.860.00651,2,3,4
40.98673 0.91184NEDIP-nearD 2.650.01441,2,3,4
40.93734 0.85592NEHEX-nearM 7.660.01231,2,4
40.88086 0.79949NEHEX-nearM 7.630.02031,2,3,4
40.79583 0.69539NE M11.7510.04763
40.41782 0.41532NEHEX-farM 9.710.01131,2
40.36260 0.39802NEHEX-farM 7.690.02331,2
40.36134 0.30575NEHEX-farM 7.730.00932
40.27919 0.23041NEHEX-farM –1

We revisited each site in January 2003, six weeks after planting, and recorded the number of seedlings established in each quadrat. In April 2003 we counted the survivors and any new germinants, removed them, and measured their aboveground biomass. Again, in October 2003 we recorded the numbers of new seedlings established in each quadrat and in April 2004 we counted and removed survivors. The small size of the plants on their removal meant that contamination of plots with new seed was minimal.

We analyzed early establishment RTE data using a binary logistic regression and biomass measurements using a general linear model (GLM) in a crossed design, with planting area and ploidy level treated as fixed effects, and with mean individual plant mass in each quadrat as the response variable. We included the number of plants as a covariate because density could affect growth. The germination of HEX-near and DIP-near seeds was analyzed separately with a binary logistic regression, but an analysis of their biomass was not carried out due to low plant numbers.

RECIPROCAL TRANSPLANT OF SEEDLINGS IN NORTHEAST SPAIN

We sourced seed from three populations located in the hexaploid area about 80 km southwest of the contact zone (labeled HEX-far), three hexaploid and three diploid populations at the contact zone (HEX-near and DIP-near, respectively), and three diploid populations 90 km northeast of the contact zone (DIP-far). Locations are shown in Figure 1 and Table 1. We established nine experimental gardens at three sites in area HEX-far, three sites in area DIP-far, and three sites close to the contact zone, just north of area DIP-near (labeled CONTACT; see Fig. 1). We were unable to place the latter three gardens at the contact zone itself because habitat suitable for M. annua was very scarce in this area (see Discussion). In each garden we planted, directly into the soil, seedlings that had been grown from seed in compost in a glasshouse at the Centro de Investigaciones sobre Desertificación (CIDE), Albal, Valencia, Spain. Seeds were sown in September 2003 and the seedlings were transferred to the gardens in October 2003. Each garden contained a single array of 120 plants, with 10-cm spacing between plants. Each array consisted of 10 blocks, with each block containing a single plant from each of the 12 seed source populations, in random order in two columns of six rows. We arranged blocks such that six were exposed to edge effects and four were not exposed to edge effects. We irrigated the sites for two days after transplanting, and then left them to grow under natural conditions.

Thirteen weeks after planting in the field, we recorded the survival of plants in all sites. For two sites in each area, we measured gas exchange rates of leaves using an IRGA gas exchange meter, with an ADC Bioscientific portable leaf chamber type PLC4B attached to a leaf chamber analyzer type LCA4 (ADC Bioscientific, Hoddesdon, Herts., UK). We measured the rates of photosynthesis and transpiration on one leaf for each plant from each of the four central blocks by clamping it for 40 sec in natural daylight conditions. We measured gas exchange rates again in April 2004, 23 weeks after planting in the field. We then cut the plants at ground level, dried them, and measured their aboveground biomass and height. We also scored the presence or absence of Melampsora pulcherrima, a common fungal pathogen of M. annua (Longo et al. 1994), for each plant.

We were interested in ecophysiological differentiation due to (1) ploidy by planting-area interactions; (2) ploidy level; (3) planting areas; and (4) distance of seed sources from the contact zone within ploidy levels. To test for these effects, we analyzed photosynthetic rate, transpiration rate, and aboveground biomass and height using a GLM, subjecting datasets to a Box–Cox transformation where necessary. Replicated sites were nested within planting areas, and replicated blocks nested within sites. The distance of seed-source populations from the contact zone (distance class: “near” versus “far”) was nested within ploidy level, and individual seed-source populations were nested within distance class. We treated area, ploidy level, and distance class of source populations from the contact zone as fixed effects, and site and block as random effects. The denominator mean square to test the effect of planting area was provided by sites, and to test the effect of ploidy it was provided by populations. In the analysis of photosynthetic and transpiration rates, we treated photosynthetically active radiation measurements as covariates. We used the central blocks from each site for analysis of aboveground biomass and height, as there was a significant edge effect. We analyzed the presence or absence of M. pulcherrima infection for all plants collected using a binary logistic regression.

GLASSHOUSE DROUGHT EXPERIMENTS

We conducted two glasshouse experiments to test the hypothesis that hexaploid M. annua plants are more drought tolerant than diploids, using diploid and hexaploid genotypes from replicated natural contact zones. One experiment was carried out in the summer and one in the winter of 2003. For both experiments, we used seeds collected from six populations in northeast Spain and six in northwest Spain, with three monoecious/androdioecious and three dioecious populations sampled in each region (Table 1). We planted seeds into peat-based compost in a glasshouse in Oxford and transplanted seedlings individually into 7.5-cm square pots. We placed these into randomized blocks, with two subplots per block (Cochran and Cox 1957), each containing one plant from each source population, placed in random order on a 42 cm × 32 cm plastic tray. Twelve blocks were used in the first experiment and five in the second. We imposed a differential watering regime, with one subplot in each block given 14 ml water only when plants were wilting (“drought” treatment) and the other subplot watered daily (“control” treatment). We rerandomized the position of blocks and of pots within sub-blocks approximately weekly during the course of the experiments. Plants that died or were severely affected by disease were removed from the experiment.

In this experiment, we were interested in ecophysiological differentiation due to (1) ploidy level within northeast Spain; (2) ploidy level within northwest Spain; and (3) the contrasting environments of northeast versus northwest Spain. To test for these effects, we carried out the measurements and analyses described below. Any datasets that did not fit the assumptions of ANOVA were Box–Cox transformed and all statistical analysis was carried out using the Minitab Software Package (Minitab 2000).

We harvested plants in the first experiment after approximately eight weeks. The plants were affected by red spider mite, so we recorded the level of disease for each plant using a scale, where 0 denoted a plant dead on harvest (or removed early due to disease) and 5 denoted a plant showing no disease symptoms on harvest. We determined water treatment, ploidy level, and region effects on disease with balanced ANOVA. We measured the dry and fresh mass of the plants on harvest, and calculated the dry/fresh mass ratio (succulence). A drought-tolerant plant should maintain higher rates of growth than a drought-sensitive plant under water stress, and maintain higher levels of succulence. To exclude the possible effects of widespread disease on plant growth, analyses were carried out only on plants with a disease rating of 2 or above, by GLM.

We removed two leaves from all plants in the two blocks least affected by disease in the first experiment. On one leaf we determined the ratio of carbon-13 to carbon-12 (δ13C), allowing assessment of the discrimination of the photosynthetic apparatus for 13C, known as Δ. A high value of Δ indicates a low water efficiency when two similar organisms are compared (Farquhar et al. 1982, 1989). The leaves were finely ground and freeze dried overnight and the relative abundance of 13C and 12C was determined using an isotope ratio mass spectrometer. These ratios were converted to discriminations (Δ in ‰) assuming a 13C composition of air in the well-ventilated glasshouse of –8.0‰ relative to PDB (Farquhar et al. 1989). We used the other leaf from each plant to measure stomatal characteristics: Using nail varnish peels from the undersides of the leaves, we counted the numbers of stomata and measured the lengths of five stomatal guard cells under a light microscope, using three randomly chosen 0.15-mm2 fields of view. Water treatment, ploidy level, and region effects were determined for these attributes with general linear model (GLM). Among plants in general, the density of stomata tends to be higher in drought conditions, and their guard cell length shorter (Meidner and Mansfield 1968; Sutcliffe 1979).

We harvested plants in the second experiment approximately 12 weeks after planting, measuring the dry and fresh mass, the dry/fresh mass ratio, and the height of the plants. Immediately before harvesting, and also on the previous day, we measured gas exchange rates of leaves using an IRGA gas exchange meter (see above for details). On the first day of measurements, the photosynthetically active radiation incident on the plants was significantly lower than on the second day (first day mean = 278.5 μmol photons/m2 s; SE = 11.2; second day mean = 362.38 μmol photons/m2 s; SE = 7.25; T=−6.30, P < 0.001, df = 250). Water treatment, ploidy level, and region effects were determined with GLM. A drought-tolerant plant might be expected to have lower rates of transpiration in water-stressed conditions but to maintain levels of photosynthesis.

Results

CONFIRMATION OF PLOIDY LEVELS

Flow cytometry results are shown in Table 1. These again showed the strict association between dioecy and diploidy (4C = 2.46 – 2.90 pg) in M. annua, and confirmed that most monoecious and androdioecious populations were hexaploid (4C = 7.43 – 7.86 pg). However, we unexpectedly found for the first time ploidy levels of higher than hexaploid in Spain: one octoploid population (4C = 11.75 pg), the ploidy of which we confirmed with a mitotic squash (for protocol see Obbard et al. 2006a), and one population with a C-value that was intermediate between that of hexaploids and octoploids (4C = 9.71 pg). These rare populations had the same sexual system as the hexaploids and were morphologically very similar. The octoploid population was found in the first glasshouse experiment, and we excluded it from analyses. The intermediate C-value population was found in both RTEs, and we included the plants from it in analyses of these experiments. We also subsequently compared this population with all hexaploid plants in the seedling RTE using a GLM and it did not differ from them below the P < 0.2 level in any of the traits measured.

RECIPROCAL TRANSPLANT OF SEEDS IN NORTHEAST SPAIN

During the two years of the experiment, five sites were damaged or destroyed in area DIP-far, and six in HEX-far, and these sites were excluded from the analysis. Plant early establishment was more successful in the DIP-far area than in the HEX-far area (Z= 2.64, P < 0.01; Fig. 2a). We could not detect significant differences between the establishment rates of diploid and hexaploid seed, nor was there a planting area by ploidy-level interaction. We found no evidence of a M. annua seed bank at any of the experimental sites.

Figure 2.

Results of the seed reciprocal transplant experiment. (a) Total number of Mercurialis annua plants established from seed over two years. (b) Mean biomass of M. annua upon harvest after the first growing season. Results are shown for seeds sampled from hexaploid (HEX-far, filled circles) and diploid (DIP-far, unfilled circles) populations in northeast Spain, and sown in field conditions in areas HEX-far and DIP-far. Bars show standard errors.

For the analysis of biomass at the end of the first year, we excluded one site that had been irrigated. Analysis of the remaining data showed a significant ploidy level by planting-area interaction (F1,27= 4.32, P < 0.05; Fig. 2b). Diploids grew larger than hexaploids in both areas, but this effect was greatest in the diploid native area.

The DIP-near and HEX-near seeds, which were planted at each site to augment the RTE experiment, showed low rates of early establishment. Approximately one fifth of the number of plants established from HEX-near seeds compared with HEX-far seeds (area HEX-far: 21.8%, area DIP-far: 18.5%). Approximately three quarters of the number of DIP-near seeds established as seedlings compared with DIP-far seeds (area HEX-far: 75.1%, area DIP-far: 74.0%). Thus the diploid seeds sourced near the contact zone had a germination advantage over the hexaploid seeds (Z=−3.20, P < 0.001; mean establishment rates: diploid = 17.7 ± 5.38, hexaploid = 5.0 ± 1.0). There was no significant difference between planting areas.

RECIPROCAL TRANSPLANT OF SEEDLINGS IN NORTHEAST SPAIN

In the early stages of this experiment, two sites were severely affected by a flash flood. In one site in area HEX-far, 36/60 diploids and 53/60 hexaploids died; in one site in area CONTACT, 30/60 diploids and 58/60 hexaploids died. This difference in mortality between ploidy levels was significant in both sites (binary logistic regression on HEX-far site: Z=−4.45, P < 0.001; CONTACT site: Z=−3.38, P < 0.01). Throughout all the other sites, 25/420 diploids and 38/420 hexaploids died. Diploids thus suffered lower mortality during the experiment than hexaploids. The results of the planned fitness measurements are given below.

Differentiation due to planting area by ploidy-level interactions

Only plant height showed a significant ploidy level by planting-area interaction. Diploids grew taller than the hexaploids in all areas, and the difference between ploidy levels was similar in areas HEX-far and DIP-far, but greater in area CONTACT where all plants were unusually short (F2,285= 25.9, P < 0.001; Fig. 3a; Table 2).

Figure 3.

Results from the seedling reciprocal transplant: (a) plant height; (b) plant dry mass; (c) photosynthetic rates in January; (d) proportion of plants infected by Melampsora pulcherrima; (e) transpiration rates in January; and (f) photosynthetic rates in April. All gas exchange rates shown (c, e, and f) were adjusted to the level of photosynthetically active radiation (PAR) using the gradient of a regression between PAR and the photosynthetic or transpiration rate. Results are shown for diploid (unfilled circles) and hexaploid Mercurialis annua plants (filled circles) grown in reciprocal transplant experiments from seedlings in three areas of northeast Spain: hexaploid area (HEX-far); area of contact zone (CONTACT); diploid area (DIP-far). Bars show standard errors.

Table 2.  GLM ANOVA tables for height, dry mass, and photosynthesis and transpiration measured in April and in January from the seedling reciprocal transplant experiment. Photosynthetically active radiation was treated as a covariate in the analyses of gas exchange data. Block and site were treated as random effects. All data was Box–Cox transformed.
SourceHeightDry massJanuary photosynthesisJanuary transpirationApril photosynthesisApril transpiration
dfAdj. MSFdfAdj. MSFdfAdj. MSFdfAdj. MSFdfAdj. MSFdfAdj. MSF
  1. *P < 0.05; **P < 0.01; ***P < 0.001.

PAR  12.34517.41***  10.002 4.98*  10.025 0.11  10.00796.31*
Area  21.137 7.20*  21.90439.7**  29.0449.09  20.04214.0*  26.227 2.21  20.01960.726
Site (area)  40.15867.7***  40.048 3.89*  30.9953.56*  30.003 2.23  32.81210.11***  30.02709.84***
Block (site (area)) 210.002 2.13** 210.012 1.57 180.2792.07** 180.001 3.29*** 180.286 1.31 180.00292.35**
Ploidy  10.14743.8***  10.40018.1**  11.0365.99*  10.000 0.09  11.132 3.84  10.00010.029
Distance (ploidy)  20.009 2.76  20.031 1.41  20.0390.22  20.000 0.33  20.041 0.14  20.00130.63
Pop (distance (ploidy))  80.003 3.07**  80.022 2.81**  80.1731.28  80.000 1.03  80.295 1.35  80.00211.66
Area×ploidy  20.02825.9***  20.001 0.19  20.2151.60  20.001 2.21  20.004 0.02  20.00030.21
Error2850.001 2860.008 2480.135 2480.000 2150.219 2140.0013 
Total325326285285252251

Differentiation due to ploidy levels

In all planting areas, diploids were superior to hexaploids in above-ground biomass (F1,8= 18.1, P < 0.01; Fig. 3b; Table 2) and mean photosynthesis rate in January (F1,8= 5.99, P < 0.05; Fig. 3c; Table 2). Diploids were slightly more susceptible to M. pulcherrima infection than hexaploids, but this difference was not significant (Fig. 3d; Z= 2.37, P > 0.3).

Differentiation due to planting areas

Planting area had a significant effect on four of the plant attributes measured. The interactive effect of area and ploidy on plant height is mentioned above. The biomass of all plants varied significantly with planting area (F2,4= 39.7, P < 0.01; Fig. 3b; Table 2), being similar in areas HEX-far and DIP-far, but lower in area CONTACT. In January, transpiration was highest in area HEX-far and lowest in DIP-far (F2,3= 14.0, P < 0.05; Fig. 3e; Table 2). The incidence of M. pulcherrima infection was highest in area DIP-far and lowest in area HEX-far (Fig. 3d; Z= 2.37, P < 0.05).

Differentiation within ploidy levels due to distance of sourcing from contact zone

Within ploidy levels, the distance at which populations were sampled from the contact zone (near/far) did not have a significant effect on any of the plant attributes measured (Table 2).

GLASSHOUSE DROUGHT EXPERIMENTS

Drought treatment had a significant effect on all plant attributes measured except disease symptoms. We carried out two sets of statistical comparisons, first between ploidy levels within regions, and second between regions within ploidy levels. Figure 4 shows results for key plant attributes.

Figure 4.

Results from two glasshouse experiments showing the effect of drought on diploid and hexaploid Mercurialis annua genotypes grown from seed sampled in northeast and northwest Spain. (a) Carbon isotope discrimination (experiment 1); (b) photosynthetic rate (experiment 2); (c) stomatal guard cell length (experiment 1); (d) stomatal density (experiment 1); (e) transpiration rate (experiment 2); (f) disease rating (experiment 1); (g) plant dry mass (experiment 1); (h) plant dry mass (experiment 2); (i) plant height (experiment 2); (j) dry/fresh mass (experiment 2). Bars show standard errors.

Differentiation due to ploidy level in northeast Spain

The diploids and hexaploids from northeast Spain differed in six attributes, listed below. The first five of these suggest that the diploids were more drought tolerant than the hexaploids. (1) Diploids had lower carbon isotope discrimination than the hexaploids in the drought treatment, indicating a higher water use efficiency; in the control treatment the reverse was true and hexaploids had higher water efficiency, with a significant ploidy level by treatment interaction (Fig. 4a; Table 3; F1,14= 14.0, P < 0.01). (2) Diploids had higher photosynthesis rates than hexaploids in the drought treatment, but lower rates in the control treatment (Fig. 4b; Table 3; F1,63= 10.69, P < 0.01). (3) Diploids had shorter stomatal guard cells than those of the hexaploids, a difference that increased in the drought treatment (Fig. 4c; Table 3; F1,294= 16.8, P < 0.001). (4) Diploids had a higher stomatal density (Fig. 4d; Table 3; F1,54= 38.3, P < 0.001). These differences in stomatal characters may have contributed to the observation that (5) diploids had a lower transpiration rate than the hexaploids (Fig. 4e; Table 3; F1,63= 5.39, P < 0.05) in the second experiment. (6) The diploids were also less affected by red spider mite than the hexaploids (Fig. 4f; Table 3; F1,94= 65.87, P < 0.001). The two ploidy levels showed no significant difference in dry/fresh mass in the first experiment and in photosynthetic and transpiration rates on the first day of measurement in the second experiment, when photosynthetically active radiation (PAR) was low (see Methods).

Table 3.  GLM ANOVA tables for carbon isotope discrimination, photosynthetic rates on the day of harvest, guard cell length, stomatal density, transpiration rates on the day of harvest, disease rating, dry mass, height, and dry/fresh mass measurements for plants from northeast Spain in two glasshouse experiments (denoted by numbers in parentheses). Block was treated as a random effect. Photosynthetically active radiation was treated as a covariate in gas exchange data analyses. Mass, gas exchange, and height data were Box–Cox transformed.
SourceC-isotope discrimination (1)Photosynthetic rate (2)Guard cell length (1)Stomatal density (1)Transpiration rate (2)Disease rating (1)
dfAdj. MSFdfAdj. MSFdfAdj. MSFdfAdj. MSFdfAdj. MSFdfAdj. MSF
PAR– – – 192013.7***– – – – – – 1  8.95 7.99**– – – 
Block (B)1 0.17 0.594200 1.7710.014 0.1310.041 0.074  2.00 1.6911 0.788 1.15
Treatment (T)152.50 1.77*1315029.6**10.077 0.7310.200 0.361109.0094.4***1 1.250 1.80
B×T1 0.30 1.034113 1.6910.10919.0***10.58228.5***4  1.16 1.0411 0.688 0.80
Ploidy (P)1 0.12 0.431427 6.36*11.2020.8***10.78338.3***1  6.03 5.39*156.70065.9***
T×P1 4.0314.0**171710.7**10.09716.8***10.134 6.54*1  3.01 2.691 0.45 0.52
Error14 0.29 6397 2940.006 540.021 63  1.12 94 0.86 
Total19752995975119
  1. *P < 0.05; **P < 0.01; ***P < 0.001.

PARDry mass (1)Dry mass (2)Height (2)Dry/fresh mass (2)
– – –   – – – – – – – –  
Block (B)110.0020.4640.0005 0.654 0.57 1.0340.001 2.7
Treatment (T)12.3065.84***10.18022.3***163.0011.3***10.03279.0***
B×T110.0041.0140.0008 1.244 0.56 1.0840.0004 1.29
Ploidy (P)10.0030.8310.0007 1.131 0.68 1.3210.001 3.29
T×P10.0000.0010.0006 0.851 0.12 0.2310.0002 0.74
Error760.004 460.0006 46 0.52 460.0003 
Total101575757

Differentiation between ploidy levels in northwest Spain

The hexaploids from northwest Spain were generally fitter than the diploids from northwest Spain, but this effect was reduced in the drought treatment. The ploidy levels differed in six attributes. (1) Diploids had a similar carbon isotope discrimination to the hexaploids in the drought treatment, but higher in the control treatment, giving a treatment by ploidy-level interaction (Fig. 4a; F1,14= 5.00, P < 0.05). This indicates that in the control treatment the hexaploids were more water efficient than the diploids, as was also found in plants from northeast Spain. (2) The hexaploids had longer stomatal guard cells, especially in the control (Fig. 4c; F1,354= 33.5, P < 0.001). (3) The hexaploids also had a higher stomatal density in the drought treatment, and diploids in the control (Fig. 4d; F1,66= 18.69, P < 0.001). Taking the previous two results together, the hexaploids thus altered their stomatal characteristics due to drought in a similar manner to the northeast Spanish hexaploids, that is, increasing their density but decreasing their size. In contrast the northwest Spanish diploids decreased stomatal density and increased stomatal size. (4) The hexaploids were more diseased in the control treatment (as found in hexaploids from northeast Spain), whereas the diploids were slightly more diseased than the hexaploids in the drought treatment (Fig. 4f; F1,118= 5.47, P < 0.05). (5) Hexaploids had greater dry mass than the diploids in both treatments in the first experiment (Fig. 4g; F1,84= 4.72, P < 0.05) as well as in the control treatment in the second experiment (Fig. 4h; F1,47= 14.20, P < 0.001). This contrasts with the plants from northeast Spain, where the ploidy levels had a similar dry mass. (6) In the second experiment, the diploid plants were taller in the drought treatment, and the hexaploids were taller in the control treatment (Fig. 4i; F1,47= 2.39, P < 0.001). Unlike plants from northeast Spain, there were no differences in transpiration and photosynthesis rates in plants from northwest Spain.

Differentiation between northeast and northwest Spain

In general, differences in the way plants from northeast versus northwest Spain responded to the experimental treatment were greater for hexaploids than for diploids. Specifically, the hexaploids differed in seven attributes. (1) Carbon isotope discrimination was higher in hexaploids from northeast Spain in both water treatments (Fig. 4a; F1,14= 15.52, P < 0.01), indicating a lower water efficiency. (2) Stomatal guard cells were shorter in hexaploids from northwest Spain in both water treatments (Fig. 4c; F1,289= 24.64, P < 0.001). (3) Stomatal density was lower for northeast Spanish hexaploids, and increased more in the drought conditions than for the northwest Spanish hexaploids, giving an interaction between treatment and region (Fig. 4d; F1,52= 4.42, P < 0.05). The smaller, more dense stomata of the northwest Spain hexaploids may have contributed to the fact that (4) their transpiration rates were lower that those of hexaploids from northeast Spain (first measurement: F1,63= 5.17, P < 0.05; second measurement: Fig. 4e; F1,45= 5.87, P < 0.05). (5) The biomass of northwest Spanish hexaploids in the second experiment was greater than that of northeast Spain in the control treatment, but lower in the drought treatment (Fig. 4h; F1,47= 12.47, P < 0.001). (6) The same was true of height (Fig. 4i; F1,47= 4.23, P < 0.05). (7) The dry/fresh mass ratio of the hexaploids from northwest Spain in the second experiment was greater than that from northeast Spain in the control treatment, but lower in the drought treatment (Fig. 4j; F1,47= 4.79, P < 0.05). Thus the succulence of the hexaploids from northeast Spain was affected more by drought than those from northwest Spain.

As noted above, the diploids sourced from the two regions were more similar in their response to the treatments than the hexaploids, and differed in only three attributes. (1) Those from northeast Spain were less affected by disease in the first experiment (Fig. 4f; F1,118= 28.49, P < 0.001). (2) In the second experiment, the diploids from northwest Spain grew taller (Fig. 4i; F1,48= 5.96, P < 0.05). (3) In the first measurement of photosynthetic rates, there was a significant treatment by region interaction (Fig. 4b; F1,65= 5.32, P < 0.05), with diploids from northeast Spain photosynthesizing faster in the drought treatment, and those from northwest Spain photosynthesizing faster in the control treatment.

Discussion

Overall, our results show clear evidence for ecophysiological and life-history differentiation between diploid and hexaploid populations of M. annua in northern Spain. However, our RTEs do not support the hypothesis that both cytotypes are locally adapted on either side of their contact zone in northeast Spain, despite the existence of an ecologically relevant environmental gradient. Instead, diploids were fitter across all areas in both RTEs. These results, together with those from experiments under controlled conditions in the glasshouse, are more consistent with the hypothesis that diploid M. annua is preadapted to the area it is currently invading in Spain. Below, we discuss evidence that supports this idea, and we review our current understanding of the evolution and distribution of sexual systems and ploidy levels in M. annua in the light of it.

ECOLOGICAL DIFFERENTIATION BETWEEN PLOIDY LEVELS

Several lines of evidence point to physiological and life-history differentiation between diploid and hexaploid lineages of M. annua found in northeast Spain: diploids had a higher biomass than hexaploids in all areas in both RTEs; diploids also grew taller and had higher photosynthetic rates and lower mortality in the seedling RTE; diploids were more drought tolerant than the hexaploids in the glasshouse experiment, with significant differences shown between the two cytotypes in carbon isotope discrimination, stomatal characteristics, transpiration rates, and disease susceptibility. Some of the differences between diploid and hexaploid M. annua are likely to have been due in part to their differences in chromosome number. An indication of this is given in the glasshouse experiments, when the differences found between diploids and hexaploids from northeast Spain were replicated in plants from northwest Spain. Thus, differences in stomatal characters, carbon isotope discrimination, and susceptibility to red spider mite would appear to depend, at least in part, on ploidy. The increases of stomatal guard cell length with ploidy level in all lineages studied here are similar to those commonly found in other polyploid complexes (Ohri 1998; Hetherington and Woodward 2003).

To what extent might the ecological differences that we observed between diploids and hexaploids of M. annua have been shaped by natural selection? We might expect that differences in growth rates and reproductive output would be subject to natural selection under different environmental conditions. However, our RTEs and controlled growth experiments conducted with seedlings and adult plants show that ploidy levels differ in a way that is largely independent of the climatic region. Neither of the RTEs showed ploidy level by environment interactions, which could be interpreted as a clear signature of local adaptation across the contact zone. Rather, diploids had higher fitness in all areas, especially within the hexaploid area. Moreover, the hexaploids from northeast Spain were less drought tolerant than the diploids in the glasshouse experiments, despite the drier environment to which they are exposed.

Given these results, we cannot conclude that parapatry between diploid and hexaploid M. annua in northeast Spain is due to environmental adaptation. We therefore have little reason to reject the hypothesis that the primary cause of parapatry is reproductive interference (Buggs and Pannell 2006). Interestingly, similar experimental investigation of ploidy distribution in Ranunculus adoneus points to a similar conclusion (Baack 2005; Baack and Stanton 2005). The only other RTE between ploidies known to us, namely in diploid Anthoxanthum alpinum and related tetraploid A. odoratum, also gave little clear evidence of differential local adaptation (Flegrová and Krahulec 1999). Thus, although polyploidy is increasingly thought to have played an important role in land–plant evolution (Cui et al. 2006), direct experimental evidence that polyploidy has given rise to adaptive differentiation remains elusive.

ECOLOGICAL DIFFERENTIATION WITHIN THE PHYLOGEOGRAPHIC CONTEXT

Our study is unusual in the extent to which its assessment of ecological differentiation can be set within a robust phylogeographic framework. As noted in the Introduction, analysis of patterns of genetic diversity strongly supports the hypothesis that the contact zones are secondary: the diploid and hexaploid lineages spent the Pleistocene glaciation in refugia in the eastern and western ends of the Mediterranean Basin, respectively, and have since come together in western Europe (Obbard et al. 2006b). Moreover, the diploid lineage is currently displacing the hexaploid lineage at a rate of several kilometers per year in both northeast and northwest Spain, as a result of strongly asymmetrical mating and pollen swamping of the hexaploids by diploid males (Buggs and Pannell 2006; Lexer and Van Loo 2006). This suggests that the diploids are probably more recent residents in their current range in both northeast and northwest Spain than the polyploids (Obbard et al. 2006b). The fact that we found many more differences between hexaploids from northeast and northwest Spain than between diploids from northeast and northwest Spain might suggest that the diploids have had less time in which natural selection and drift could cause differentiation, corroborating the recent migration hypothesis. We might therefore have expected local adaptation of the hexaploids (and their lack in the diploids) to have slowed the diploids' advance. However, our experiments found the diploids to be apparently better suited to growing in the hexaploid environment than the hexaploids themselves. It therefore appears that ecophysiological differences between the two cytotypes will quicken the hexaploids' demise.

The competitive displacement of one lineage by a related one across a hybrid contact zone may not be rare in nature. A competitive advantage on the part of one lineage is thought to be implicated in the movement of a hybrid zone between two species of lice, Geomydoecus aurei and G. centralis, which has moved 700–900 m in five years (Hafner et al. 1998). Similarly, a hybrid zone between Anartia amathea and its sister taxon A. fatima has moved at an average rate of approximately 50 km over the past 20 years (Dasmahapatra et al. 2002). As a final example, a hybrid zone between hermit warblers (Dendroica occidentalis) and Townsend's warblers (D. townsendi) is estimated to have moved 2000 km in the northwest United States over the past 5000 years (Rohwer and Wood 1998; Rohwer et al. 2001). Although models of endogenous and exogenous selection provide different levels of explanation for the maintenance of hybrid zones, the possibility of interactions between them is well known (Bert and Arnold 1995; Kruuk et al. 1999). Certainly, in the case of M. annua, it would appear that endogenous selection (reproductive interference) and exogenous selection (ecological differentiation) are both contributing to the rapid movement of the diploid–hexaploid hybrid zone.

ECOLOGICAL RELEVANCE OF THE ENVIRONMENTAL GRADIENT

Movement of the hybrid zones between diploid and hexaploid populations of M. annua appears to be occurring along an environmental gradient likely to be important to the life history and ecophysiology of M. annua. Our RTE of seeds showed establishment rates to be higher in the DIP-far area than in the HEX-far area, and in our RTE of seedlings we found higher transpiration rates in the HEX-far region for both ploidy levels. There was also a gradient of M. pulcherrima infection in the seedling RTE experiment, being higher for both ploidy levels toward the north. This may reflect an environmental difference and release from pathogens could be part of the cause of the greater fitness of the diploids in the south (Mitchell and Power 2003).

Our results also show that the environmental factors relevant to the ecology of M. annua do not vary monotonically down the coast of Spain. The area near the point of contact of the two ploidy levels appears to be exceptionally inhospitable to M. annua. Our transplants of seedlings in this area showed a much weaker growth than transplants in both the HEX-far and DIP-far regions. Seed sourced from this area had low success at establishing plants during two years, perhaps due to poor maternal provisioning, although we cannot exclude the possibility that hybridization is playing a role. The coastal strip is particularly narrow here and mainly supports bushy schlerophyllous vegetation (ICC 1993). Surveys also have shown that the natural M. annua population density in this region is low (Eppley and Pannell 2006). Such areas of low population density are expected to impede the movement of hybrid zones that are structured by positive frequency-dependent selection against hybrids (Barton and Hewitt 1985). This might help to explain why movement of the northwest contact zone, although it has recently been rapid, appears to have been slower than the corresponding rate of movement in northwest Spain (Buggs and Pannell 2006).

THE DISTRIBUTION OF PLOIDY LEVELS AND SEXUAL SYSTEMS

The diploid–hexaploid contact zones in M. annua also correspond to transitions in the sexual system (see above). It is well known that outcrossing populations typically possess higher levels of genetic variation than partially selfing populations (reviewed in Charlesworth and Pannell 2001). In a survey of genetic variation among populations with different sexual systems in M. annua, Obbard et al. (2006b) found that monoecious populations were substantially less diverse at isozyme loci than androdioecious populations in which males existed. Lack of genetic variation may have restricted the ability of the monoecious northeast hexaploids to respond to natural selection (Lande and Shannon 1996), providing some explanation for their apparently lower fitness than the androdioecious hexaploids from northwest Spain.

In addition to the factors considered in the current study, persistence of diploid and hexaploid M. annua within larger regions is likely to also be a function of selection at the metapopulation level, mediated by differences in their sexual systems, as mentioned above. The interaction of reproductive interference (Buggs and Pannell 2006), reproductive assurance (Pannell and Barrett 1998; Eppley and Pannell 2006; Obbard et al. 2006b), and ecophysiological differentiation (this article) between ploidy levels is likely to be complex. Preliminary results from computer simulations suggest that the selection for monoecy at the metapopulation level could balance local selection under plausible parameter combinations (M.E. Dorken and J.R. Pannell, unpubl. ms.). A broader set of RTEs including androdioecious areas in the southeast of Spain would shed further light on these interactions.

Conclusion

Our study has shown that diploid and hexaploid M. annua are ecologically differentiated where they meet in northern Spain, and that diploids are competitively superior to hexaploids in several important physiological and life-history traits. These results contribute to our understanding of the remarkably rapid displacement by diploids of hexaploids along two contact zones in Spain (Buggs and Pannell 2006). They are also best interpreted within the historical context of the phylogeographic range expansion of the diploids from Eastern Europe and their Pleistocene refugium in the eastern Mediterranean Basin (Obbard et al. 2006b): much of the ecological differentiation between diploid and hexaploid cytotypes likely evolved in allopatry.

Recent data indicate that M. annua is in fact an allopolyploid species, formed as a result of a cross between an autotetraploid derivative of diploid M. annua and diploid M. huetii, a sister species to M. annua (Obbard et al. 2006a). The lower fitness of allopolyploid M. annua contrasts with patterns observed for allopolyploids in Spartina (Anttila et al. 1998; Hacker et al. 2001) and Tragopogon (Novak et al. 1991), whose expanding ranges suggest they have higher fitness than their diploid progenitors. Although both polyploidy and hybridization represent important evolutionary transitions (Otto and Whitton 2000; Rieseberg et al. 2003), hexaploid M. annua may fit a pattern observed in some genera, where hybrids (Wagner 1970; Mayr 1992) or polyploids (Stebbins 1940, 1950) enjoyed an initial advantage over their respective progenitors, but then failed to adapt further.

Associate Editor: J. Kohn

ACKNOWLEDGMENTS

We thank M. E. Dorken, S. A. Harris, D. J. Obbard, J. S. Parker, and two anonymous reviewers for valuable comments on earlier versions of this article; A. Sing and C. Surman for technical assistance; X. Sans and M. Verdú for logistical assistance and hospitality in Spain; and numerous farmers in Spain for the use of their land. RJAB was supported by a studentship awarded by the Biotechnology and Biological Sciences Research Council, U.K., and JRP was partly supported by grants from the Natural Environment Research Council, U.K. F. Platt and the Wellcome Trust provided flow cytometry facilities.

Ancillary