Temporal changes in population genetic diversity and structure in red and white clover grown in three contrasting environments in northern Europe

Abstract

Backgound and Aims

Extending the cultivation of forage legume species into regions where they are close to the margin of their natural distribution requires knowledge of population responses to environmental stresses. This study was conducted at three north European sites (Iceland, Sweden and the UK) using AFLP markers to analyse changes in genetic structure over time in two population types of red and white clover (Trifolium pratense and T. repens, respectively): (1) standard commercial varieties; (2) wide genetic base (WGB) composite populations constructed from many commercial varieties plus unselected material obtained from germplasm collections.

Methods

At each site populations were grown in field plots, then randomly sampled after 3–5 years to obtain survivor populations. AFLP markers were used to calculate genetic differentiation within and between original and survivor populations.

Key Results

No consistent changes in average genetic diversity were observed between original and survivor populations. In both species the original varieties were always genetically distinct from each other. Significant genetic shift was observed in the white clover ‘Ramona’ grown in Sweden. The WGB original populations were more genetically similar. However, genetic differentiation occurred between original and survivor WGB germplasm in both species, particularly in Sweden. Regression of climatic data with genetic differentiation showed that low autumn temperature was the best predictor. Within the set of cold sites the highest level of genetic shift in populations was observed in Sweden.

Conclusions

The results suggest that changes in population structure can occur within a short time span in forage legumes, resulting in the rapid formation of distinct survivor populations in environmentally challenging sites.

INTRODUCTION

The current emphasis on sustainable agriculture requires the cultivation of high-yielding temperate forage legumes such as red and white clover (Trifolium pratense and T. repens, respectively) in northern Europe. However, their widespread adoption is currently hampered by a lack of sufficiently reliable cultivars, especially in the colder, more northern regions (Helgadóttir et al., 2001). Extending the use of these valuable species through plant breeding requires not only the exploitation of a wider range of genetic diversity (Greene et al., 2004), but also information about the rate and direction of genetic change within populations being cultivated close to the margin of their natural species distribution.

The intraspecific genetic diversity present in temperate sown grasslands has been eroded over the last half-century or so, largely as a result of the replacement of the original landraces and ecotypes of grassland species by productive commercial varieties with a more restricted genetic base (Kölliker et al., 2003; Herrmann et al., 2005; van Treuren et al., 2005). Forage legume varieties released in the EU must conform to strict standards of distinctness, uniformity and stability, and programmes of germplasm improvement are often based on small numbers of parental genotypes. Although this approach has enhanced the agronomic performance of legume varieties, the resulting restriction of genetic diversity could be detrimental in terms of the adaptive potential of agricultural grassland populations in a changing climate, as a certain degree of genetic variability within plant populations increases the genome flexibility that is important for adaptive evolution (Li et al., 2009). The presence of high levels of intraspecific genetic diversity in natural ecosystems has encouraged plant breeders to consider the development of populations with similarly high diversity through the use of blends of varieties, composite populations or heterogeneous varieties. The performance of such populations is expected to be favoured in environments characterized by large spatial and/or temporal heterogeneity (Annicchiarico and Piano, 1997). Thus, the creation of forage legume populations with deliberately increased levels of diversity is a strategy that could be used to improve yield stability in environments that are prone to periods of biotic or abiotic stress.

Understanding the basis of adaptive genetic differentiation in survivor populations in diverse environments contributes to the development of breeders' models of plant varieties, and survivor populations themselves may also provide a useful source of germplasm for breeding programmes (Wedderburn et al., 2005). The process of adaptation implies changes in population allele frequency over time in response to environmental changes (Pérez de la Vega, 1997), and this is commonly referred to as ‘genetic shift’. Environment plays a major role in determining the genetic structure of populations (Greene et al., 2004), and the outbreeding propensity and generally high levels of genetic heterogeneity found in red and white clover favour the occurrence of genetic shift in response to environmental factors (Wedderburn et al., 2005). Rapid genetic shift has previously been observed in morpho-physiological traits in two white clover varieties grown at a number of sites in Europe encompassing a range of climatic conditions (Collins et al., 2001, 2002). In those studies the varieties were grown for 3 years in environments that differed substantially from those in which they were initially selected, and it was hypothesized that climatic conditions had acted as a selection pressure resulting in genetic differentiation between ‘original’ and ‘survivor’ populations. Although genetic variation within and between forage legume populations has been successfully estimated using morpho-physiological traits (e.g. Collins et al., 2001, 2002), such assessments are indirect and have required laborious field experimentation. Additionally, significant morpho-physiological variation can result from the action of only a small number of genes, and such divergence may have little relationship with the overall degree of genetic differentiation between populations (Arroyo-García et al., 2001). The alternative approach of molecular marker analysis offers a more efficient method, as genetic relationships are estimated on the basis of plant genotype rather than phenotype (Kölliker et al., 2001). The amplified fragment length polymorphism (AFLP) technique (Vos et al., 1995) is a useful tool for DNA fingerprinting and genetic mapping (Roldán-Ruiz et al., 2000) in species where genomics resources are less developed. It is particularly suitable for genetic diversity analysis due to its high reproducibility and capacity to detect multiple loci in a single assay (Guthridge et al., 2001).

Here we describe the use of AFLP markers to analyse changes in population diversity over time (i.e. analysis of ‘original’ vs. ‘survivor’ populations) in red and white-clover populations grown in three climatically contrasting northern European sites (Iceland, Sweden and UK). The investigation was part of a wider multinational research collaboration carried out under the auspices of EU COST Action 852 (‘Quality legume based forage systems for contrasting environments’) which operated from 2002 to 2007. It was based on a common experiment using four-species mixtures and monocultures of forage legumes and grasses repeated at a number of sites across Europe and in Canada and Australia (Kirwan et al., 2007). The COST 852 experiment included an optional treatment in which the impact of the level of intraspecific variation in the legume species (red clover and white clover) was assessed by the use of single varieties vs. wide genetic base (WGB) composites, comprising mechanical seed mixtures of many varieties and populations (Collins et al., 2004; Frankow-Lindberg et al., 2009). The WGB populations were developed with the intention of testing the effect of increased phenotypic diversity on the agronomic performance of important forage legumes. The single varieties in the experiment were chosen on the basis of their suitability for agricultural use in the local area, and were therefore different at each site. Every batch of seed of the WGB populations supplied to participating sites was drawn from the same seed stock, making the starting (‘original’) populations comparable at each site.

The two species analysed in this experiment differ substantially in growth habit: red clover has an upright growth habit and is regarded as a short-lived perennial, whereas white clover spreads along the ground via a network of perennating stolons. Under the managements imposed at all sites in this experiment it is unlikely that significant seedling recruitment occurred in either species due to the removal of inflorescences by cutting. Therefore, we consider that genetic input from non-experimental genotypes was minimal.

We hypothesised that: (a) in both species the single varieties display a lower level of ‘within population’ genetic diversity than the WGB germplasm; (b) survivor populations contain a lower level of genetic diversity than the original populations; (c) at each site survivor populations within the respective species are genetically closer to each other than the original populations, i.e. the environment within each site promotes genetic similarity; (d) survivor populations of the same original population selected at different sites are genetically distinct from each other, i.e. environmental differences between sites have the potential to promote genetic dissimilarity.

Although only a small number of sites contributed to this study, we investigated correlations between climatic variables and values of population differentiation obtained from AFLP analysis in an attempt to identify aspects of the sites' climates that might promote genetic shift.

MATERIALS AND METHODS

Plant material

The single varieties used in this study were chosen because they were considered to be agronomically well adapted to their experimental sites. For red clover (Trifolium pratense) the varieties were: in the UK, ‘Merviot’ (early-flowering, productive diploid variety bred in Belgium); in Sweden, ‘Fanny’ (a selection for nematode resistance and improved seed-set made from high-yielding tetraploid lines based on landraces from southern Sweden). For white clover (T. repens) the varieties were: in Iceland, ‘Norstar’ (a variety bred in Norway for improved winter hardiness based on a selection from a wild population); in Sweden, ‘Ramona’ (a selection for increased flowering made from the Swedish variety ‘Sonja’).

Four WGB composite populations were constructed at IBERS, Aberystwyth, UK and at Graminor AS, Ilseng, Norway from a mixture of commercial varieties and gene pool material. The red clover central European (WGB-C) population, containing seed of 11 commercial varieties bred for use in temperate regions, plus a small amount of winter-hardy germplasm, was constructed as follows: a bulk mixture was made with 100 g of ‘Lemmon’ and ‘Amos’; 60 g of ‘Astra’, ‘Norseman’, ‘Britta’, ‘Merviot’, ‘Marcom’, ‘Deben’ and ‘Grasslands Pawera’; 46 g of ‘Milvus’ and 30 g of ‘Sabtoron’. To 495 g of this mixture was added 150 g of ‘northern red clover composite’ from Norway/Denmark (P. Marum, Graminor AS, Norway, pers. comm.; see details below), resulting in a 23 % ‘gene pool’ concentration by weight. This population encompassed all germplasm categories employed by red-clover breeders (early/late flowering; diploid/tetraploid). The red clover northern European (WGB-N) population contained seed from the ‘northern red clover composite’ created in the Nordkløver project of NordGen (the Nordic Genetic Resource Centre) by the inter-crossing for two generations of the 283 accessions comprising their complete red-clover collection, plus varieties from other parts of the world (mainly the former USSR) (for more details, see Helgadóttir et al., 2000). The white clover WGB-C population, containing seed of 13 commercial varieties bred for use in temperate regions plus a small amount of winter hardy ‘gene pool’ germplasm from central Asia, was constructed as follows: a bulk mixture was made with 100 g of ‘Nesta’, ‘Olwen’, ‘Alice’, ‘Menna’, ‘Siwan’, ‘Sabeda’, ‘AberHerald’, ‘Klondike’ and ‘Merida’; 75 g of AberDairy blend (comprising 70 % ‘AberDai’ and 30 % ‘AberHerald’), and 10 g each of ‘AberVantage’, ‘Donna’ and ‘Katrina’. To 570 g of this mixture was added 30 g of ‘gene pool’ germplasm from central Asia (L. van Soest, Centre for Genetic Resources, The Netherlands, pers. comm.), resulting in a 5 % gene-pool concentration. The white clover WGB-N population contained seed from nine lines and varieties bred for use in northern Scandinavia, plus ten commercial varieties from southern Scandinavia and Latvia bred for use in southern Scandinavia (P. Marum, pers. comm.). It comprised the following: 60 % by weight of the winter-hardy lines and varieties ‘LøKv9601’, ‘KvKv9001’, ‘Undrom’, ‘Norstar’, ‘Snowy’, ‘HoKv9238’, ‘HoKv9240’, ‘Sandra’, ‘Jögeva 4’; 40 % of the winter-hardy varieties ‘Sonja’, ‘Lena’, ‘Gandalf’, ‘Milkanova’, ‘Milo’, ‘Klondike’, ‘Rivendel’, ‘Ramona’, ‘SW Hebe’ and ‘Priekulu 61’.

AFLP analysis of ‘survivor’ vs. ‘original’ populations

Sampling procedure

AFLP markers were used to assess the genetic diversity within and between red- and white-clover populations grown at three COST 852 sites: Iceland (Korpa: 64 °09'N, 21 °44'W; 30 m a.s.l.), Sweden (Svalöv: 55 °55'N, 13 °07'E; 55 m a.s.l.) and UK (Aberystwyth: 52 °26'N, 4 °01'W; 30 m a.s.l.). At each site monoculture plots of red and white clover were established from seed of (a) one single variety of each species; (b) WGB populations of each species (for a description of the experimental design, see Frankow-Lindberg et al., 2009). Post-establishment the plots were managed by cutting between two and four times per growing season (depending on site) to a stubble height of around 5 cm. After a number of years (3–5 depending on site) leaf samples were collected from random survivor genotypes within a subset of the sown populations. In Iceland, samples were collected after 4 and 5 years from the red and white clover WGB-N populations, respectively, and after 4 years from the white clover variety ‘Norstar’. At this site, samples were also collected from a wild population of white clover adapted to local conditions. In Sweden, samples from red clover survivors were collected after 3 years from the populations WGB-C and WGB-N and from the variety ‘Fanny’; samples from white clover survivors were collected from the populations WGB-C and WGB-N and from the variety ‘Ramona’. In the UK, survivor samples were collected after 4 years from the red clover variety ‘Merviot’. The number of survivor plants sampled in each population ranged from 31 to 96. ‘Original’ populations within both species were created by growing plants from seed in a heated glasshouse in IBERS, Aberystwyth, using seed randomly drawn from exactly the same seed-lots employed in the COST 852 experiment. The randomly chosen genotypes sampled at sites after the experiment had finished were those remaining after natural selection had operated for a number of years, and thus constituted ‘survivor’ populations for those sites. These could subsequently be compared in terms of genetic structure with the original populations sampled randomly from the initial seed stock, thus potentially providing useful insights into the processes of selection operating at different sites. It should be pointed out that, ideally, ‘original’ samples for AFLP analysis would comprise DNA extracted from newly sown genotypes in the experimental plots. However, we considered that using DNA extracted from a large number of genotypes grown from exactly the same seed batch sown in the field experiments in the various sites would constitute an acceptable substitute for a truly original population. In addition, all the ‘original’ populations for this study were grown in a common glasshouse environment, thus removing any effects of environmental variation on their genetic structure. All the ‘WGB original’ (WGB-O) populations comprised 96 genotypes; the single variety original populations comprised 31–48 genotypes. A list of populations, their allocated codes and the number of genotypes sampled at each site are given in Table 1.

Table 1.

List of populations sampled, with analysis codes

Red clover	White clover
(96) WGB-C-O	(96) WGB-C-O
(31) WGB-C-SE	(31) WGB-C-SE
(96) WGB-N-O	(96) WGB-N-O
(48) WGB-N-IS	(48) WGB-N-IS
(31) WGB-N-SE	(31) WGB-N-SE
(47) F-O	(47) R-O
(31) F-SE	(31) R-SE
(48) M-O	(32) N-O
(48) M-AB	(32) N-IS
	(32) W-IS

The number of genotypes per sample is in brackets.

WGB, wide genetic base composite; C, Central European population, N, Northern European population. Single varieties: M, ‘Merviot’, F, ‘Fanny’, R, ‘Ramona’, N, ‘Norstar’, W, wild ecotype. O,original population, IS, Iceland survivor population, SE, Sweden survivor population, AB, UK survivor population.

DNA extraction and data analysis

In each population, samples of approx. 50 mg fresh weight (f. wt) were collected in the form of one fresh actively expanding leaf per plant. Each sample was placed in a labelled Eppendorf tube and kept on ice (UK, Sweden) or submerged in liquid nitrogen (Iceland) until the required number of genotypes in the population had been processed. Tubes were subsequently stored at –80 °C prior to DNA extraction. Leaf samples from Iceland and Sweden were freeze dried and sent to IBERS, Aberystwyth for processing. Total genomic DNA was extracted from the leaf samples with a DNeasy™ 96 plant kit (QIAGEN GmbH, Hilden, Germany). Fluorescent AFLP analysis was carried out on 100 ng of DNA from each sample using an ABI 3730 Genetic Analyzer as described by Skøt et al. (2005). The AFLP analysis on all the samples for this study was carried out over a period of 2 months to avoid issues with drift in peak size over long time periods. The four selective primer pairs used are shown in Table 2. Analysis of the electropherograms was performed using the GeneMapper^® v3·7 (Applied Biosystems, Warrington, UK) software program with manual editing. Only peaks above a threshold intensity value of 100 were scored. Markers were scored as present (1) or absent (0) and the data were exported to Microsoft Excel format. AFLP bands below a frequency level of 5 % were removed from the dataset prior to further data analysis. The presence/absence data matrix produced in GeneMapper was analysed using the program AFLP-SURV (Vekemans, 2002). The following genetic diversity and population genetic structure parameters were calculated by AFLP-SURV using the method of Lynch and Milligan (1994): number and proportion of polymorphic loci at the 5 % level; expected heterozygosity within a population (Hj) (also called ‘Nei's gene diversity’; Nei, 1987) and its variance components; and the proportion of the total gene diversity occurring among, as opposed to within, populations (population differentiation: F_st). F_st was calculated as H_b/H_t, where H_t was the total gene diversity in the overall sample and H_b was the average gene diversity among populations in excess of that observed within populations (Reynolds et al., 1983). Allele frequencies were estimated using the method of Zhivotovsky (1999), and confidence intervals on genetic distances were estimated using bootstrap analysis (1000 replicates). A t-test was used for comparisons of Hj values for original vs. survivor populations. Analysis of molecular variance (AMOVA) was carried out with the program GenAlEx (Peakall and Smouse, 2006) to assess genetic variation within and between populations. To assist in visualizing the data, principal co-ordinates analysis (PCO) was carried out using Genstat v13·3 (Genstat, 2010). Input to the PCO analysis was from a similarity matrix formed from pairwise F_st values for the populations in each species separately.

Table 2.

AFLP analysis: sequences of primer pairs for selective amplification

EcoRI	MseI
5'-GAC TGC GTA CCA ATT CAC A-3'	5'-GAT GAG TCC TGA GTA ACA G-3'
5'-GAC TGC GTA CCA ATT CAC A-3'	5'-GAT GAG TCC TGA GTA ACC T-3'
5'-GAC TGC GTA CCA ATT CAC T-3'	5'-GAT GAG TCC TGA GTA ACC T-3'
5'-GAC TGC GTA CCA ATT CAC A-3'	5'-GAT GAG TCC TGA GTA ACG A-3'

Nucleotides underlined are selective sequences.

Site characterization

Meteorological data for each site were recorded over the duration of the experiment at weather stations located near the field plots, and these provided a set of bioclimatic variables. Average daily values for rainfall, minimum, mean and maximum air temperatures, and daily growing degree-days (DGDD) were obtained for winter, spring, summer and autumn (as defined by each site) (Table 3). Seasonal values of DGDD were calculated using the daily mean air temperature (T_mean) minus a base temperature for plant growth (T_base) of 5 °C. To prevent the occurrence of negative values of DGDD the formula was adjusted so that days on which T_mean ≤ T_base were assigned a score of zero.

Table 3.

Daily meteorological data averaged over seasons for each site during the experiment

Season and location	Date	Rainfall (mm d−1)	Min. air temp. (°C)	Mean air temp. (°C)	Max. air temp. (°C)	DGDD
Winter
Iceland	1 Nov–30 Apr	3·67	–1·95	1·18	4·32	0·34
Sweden	25 Nov–4 Mar	1·21	–1·84	0·80	2·94	0·18
UK	1 Dec–28 Feb	2·71	2·63	5·85	8·60	1·94
Spring
Iceland	1 May–25 Jun	1·45	4·59	8·61	12·63	3·82
Sweden	5 Mar–15 May	1·17	1·05	5·45	9·89	2·33
UK	1 Mar–31 May	2·19	5·67	9·31	12·98	4·51
Summer
Iceland	26 Jun–31 Aug	2·14	8·08	12·00	15·93	7·00
Sweden	16 May–15 Sep	2·37	11·07	15·57	20·21	10·57
UK	1 Jun–31 Aug	2·03	12·33	15·86	19·63	10·86
Autumn
Iceland	1 Sep–31 Oct	3·67	2·56	6·09	9·62	2·33
Sweden	16 Sep–24 Nov	1·58	4·54	8·24	11·73	3·83
UK	1 Sep–30 Nov	3·68	7·88	11·22	14·73	6·42

Daily growing degree-days (DGDD) were calculated as the average of (daily mean air temp – 5 °C), with negative values assigned a score of zero.

Correlation between genetic differentiation and climate

Due to the limited number of sites in this study it was not possible to carry out a comprehensive correlation analysis of values of F_st with climatic variables. However, regressions were carried out between F_st values for a subset of pairs of original and survivor populations, and the climatic variables in Table 3. This was possible for red, but not white-clover populations, since white clover survivors were only collected at two sites (Iceland and Sweden). On the basis of the results of this analysis the relationship between DGDD for each season and F_st was also examined for red clover through principal components analysis (PCA) based on a correlation matrix using Genstat v13·3 (Genstat, 2010).

RESULTS

AFLP analysis of ‘survivor’ vs. ‘original’ populations

AMOVA analysis of the complete datasets of the two legume species showed that most of the total genetic diversity variation was distributed within populations: 88 % for red clover and 81 % for white clover (Table 4).

Table 4.

Analysis of molecular variance (AMOVA) of red and white-clover populations

Species	No. of populations	Source of variation	Variance component	Percentage of total variation
Red clover	9	Among populations	4·28	12
		Within populations	32·88	88
White clover	10	Among populations	9·97	19
		Within populations	42·83	81

In both species the variance components were significant at P < 0·01.

Within-population genetic structure

The number of loci scored by the four AFLP primer pairs was 289 for red clover and 376 for white clover. In tetraploid species such as white clover, use of AFLP and other dominant markers can overestimate allele frequencies (Lynch and Milligan, 1994). However, since it is an allotetraploid species showing disomic inheritance (Weidema, 1996), this is not a serious issue. The number of polymorphic loci ranged from 187 to 228 in red clover (mean 203) and from 213 to 272 in white clover (mean 245). In red clover, the proportion of polymorphic loci was greater in original populations of the two varieties sampled (‘Merviot’ and ‘Fanny’) than in the original WGB populations (Table 5). In white clover, on the other hand, the proportion of polymorphic loci in the original variety ‘Ramona’ was lower, and that in ‘Norstar’ was the same as in the original WGB germplasm. The wild white-clover population sampled in Iceland had the lowest frequency of polymorphic loci.

Table 5.

Genetic diversity indices (Hj) of red and white clover original and survivor populations of varieties and WGB composite populations

Species/population	Proportion of polymorphic loci	Gene diversity averaged across loci (Hj) ± s.e.	Comparison of Hj for originals vs. survivors (t-test)
Red clover
WGB-N-O	0·66	0·196 ± 0·0090
WGB-N-IS	0·59	0·187 ± 0·0098	n.s.
WGB-N-SE	0·78	0·222 ± 0·0087	P < 0·05
WGB-C-O	0·65	0·201 ± 0·0090
WGB-C-SE	0·79	0·235 ± 0·0095	P < 0·025
M-O	0·68	0·202 ± 0·0094
M-AB	0·70	0·211 ± 0·0093	n.s.
F-O	0·72	0·218 ± 0·0093
F-SE	0·76	0·228 ± 0·0090	n.s.
White clover
WGB-N-O	0·70	0·208 ± 0·0077
WGB-N-IS	0·67	0·209 ± 0·0082	n.s.
WGB-N-SE	0·68	0·200 ± 0·0082	n.s.
WGB-C-O	0·71	0·230 ± 0·0084
WGB-C-SE	0·66	0·208 ± 0·0088	n.s.
R-O	0·63	0·202 ± 0·0087
R-SE	0·57	0·179 ± 0·0088	n.s.
N-O	0·72	0·232 ± 0·0085
N-IS	0·70	0·226 ± 0·0087	n.s.
W-IS	0·57	0·188 ± 0·0096	*

n = 289 for red clover and n = 376 for white clover.

*, W-IS compared with N-O (P < 0·001) and N-IS (P < 0·005)

In red clover, gene diversity (Hj) of the original populations was highest in ‘Fanny’ and lowest in WGB-N-O. There was no significant difference between the other original red-clover populations sampled (Table 5). Among original white-clover populations, ‘Norstar’ and WGB-C-O had the highest gene diversity, while ‘Ramona’ and WGB-N-O had the lowest. Gene diversity of the survivor populations differed in two cases from the respective original populations: in red clover it was significantly higher among the two WGB survivors sampled in Sweden compared with their corresponding original populations. In white clover, gene diversity tended to be lower in survivor populations sampled in Sweden than in their original populations but these differences were not significant. The gene diversity of the wild white-clover population sampled in Iceland was significantly lower than in ‘Norstar’ and its survivor population in Iceland.

Genetic differentiation between populations

In this study, values of F_st between red-clover populations ranged from 0·016 (original population of ‘Merviot’ vs. survivor population of ‘Merviot’) to 0·138 (survivor population of WGB-N sampled in Iceland vs. original population of ‘Fanny’). The mean F_st value for the original red-clover populations was 0·084 (Table 6A). Values for white clover were higher, and F_st ranged from 0·022 (original population of ‘Norstar’ vs. survivor population of ‘Norstar’) to 0·305 (original population of ‘Ramona’ vs. the wild white-clover population sampled in Iceland). The mean F_st value for white clover original populations (excluding the Icelandic ‘wild’ population) was 0·117 (Table 6B).

Table 6.

Pairwise genetic differentiation (F_st) between (A) red and (B) white clover original and survivor populations of varieties and WGB composite populations

	WGB-N-O	WGB-N-IS	WGB-N-SE	WGB-C-O	WGB-C-SE	M-O	M-AB	F-O
(A) Red clover
WGB-N-IS	0·076
WBG-N-SE	0·077	0·087
WGB-C-O	0·039	0·042	0·058
WGB-C-SE	0·099	0·087	0·029	0·058
M-O	0·109	0·127	0·082	0·072	0·069
M-AB	0·096	0·099	0·064	0·066	0·063	0·016
F-O	0·108	0·138	0·074	0·082	0·090	0·092	0·086
F-SE	0·105	0·137	0·037	0·083	0·062	0·096	0·090	0·036
Lower 95 % confidence limit = –0·0028; upper limit = 0·0014.
	WGB-N-O	WGB-N-IS	WGB-N-SE	WGB-C-O	WGB-C-SE	R-O	R-SE	N-O	N-IS
(B) White clover
WGB-N-IS	0·030
WBG-N-SE	0·106	0·115
WGB-C-O	0·047	0·052	0·085
WGB-C-SE	0·099	0·093	0·037	0·052
R-O	0·155	0·174	0·196	0·146	0·197
R-SE	0·155	0·174	0·057	0·132	0·089	0·172
N-O	0·071	0·065	0·127	0·095	0·127	0·189	0·195
N-IS	0·095	0·087	0·153	0·116	0·154	0·202	0·215	0·022
W-IS	0·210	0·208	0·248	0·213	0·260	0·305	0·301	0·169	0·136

Lower 95 % confidence limit = –0·0029; upper limit = 0·0015.

A value of 0 indicates identical populations; 1 indicates no alleles in common. Lower and upper confidence limits around zero were estimated by 1000 random permutations of individuals among populations.

In the PCO analysis the first two dimensions explained 68 % and 84 % of the data variance in red and white clover, respectively. PCO analysis showed that the original populations of the red-clover varieties ‘Merviot’ and ‘Fanny’ were clearly separated from each other (F_st = 0·092) (Fig. 1A). However, their survivor populations, sampled in the UK and Sweden, respectively, responded similarly in that ‘Merviot’ survivors were genetically very similar to their original population (F_st = 0·016), and there was only slightly more genetic distance between ‘Fanny’ originals and survivors (F_st = 0·036). The original populations of the two white clover varieties ‘Norstar’ and ‘Ramona’ were also clearly separated from each other (F_st = 0·189) (Fig. 1B), and the wild population sampled in Iceland could be distinguished from both varieties, although it was genetically closer to the two ‘Norstar’ populations. The survivor population of ‘Ramona’ (sampled in Sweden) was genetically more distant from the corresponding original population (F_st = 0·172) than was the survivor population of ‘Norstar’ (sampled in Iceland) (F_st = 0·022). In terms of site responses, WGB germplasm in both species behaved similarly to the single variety populations. Thus, the original WGB-C and WGB-N populations of both species tended to be separated from the corresponding survivor populations sampled in Sweden more than in Iceland (Fig. 1A, B). In white clover, the Swedish survivors of both WGB-N and WGB-C were more similar genetically (F_st = 0·037) than were the original populations (F_st = 0·047), suggesting that the environment at this site promoted genetic similarity in survivor germplasm (Fig. 1B).

Fig. 1.

Scatter plot of genetic relationships among original and survivor populations of (A) red and (B) white clover determined on the basis of principal co-ordinate analysis (PCO) of AFLP data. Data for the PCO derived from a matrix of between-population pairwise values of F_st.

Site climatic data

The three sites differed substantially in terms of climate (Table 3). Regression analyses of values of F_st for original and survivor populations of red clover (Table 6A) on climatic variables showed that F_st primarily correlated with temperature conditions in the autumn (P < 0·08 for daily minimum, mean and maximum air temperatures, and DGDD). The relationship was negative for all these variables. The parameter DGDD has the advantage of being of biological relevance and is widely used to predict crop development in different climates (Russelle et al., 1984). A biplot display of the PCA for DGDD values for all four seasons and F_st for the relevant red-clover population pairs reinforces the result that genetic differentiation was largely influenced by low values of DGDD in the autumn (Fig. 2). On the diagram, the arrows show the loadings for each variable in each component. The arrows also show correlations between measurements: those pointing in the same direction show positive correlations between measurements; those pointing in opposite directions show negative correlations; those that are perpendicular to each other are not correlated. However, it should be emphasized that the dataset was small, so these results must be treated with caution.

Fig. 2.

Biplot display of principal components analysis of DGDD (daily growing degree-days) during winter, spring, summer and autumn, and F_st (genetic differentiation between populations) for the following pairs of red-clover populations: WGB-N-O/IS in Iceland; WGB-N-O/SE, WGB-C-O/SE and F-O/SE in Sweden; M-O/AB in UK.

DISCUSSION

Analysing survivor population germplasm derived from selected sites of the COST 852 common experiment, in which all sites adopted a similar management protocol, allows a generality of interpretation of our results that would not be possible from individual site analyses. The experimental design used here therefore allows us to assess unambiguously the effects over time of site conditions on the populations' genetic structure.

Population diversity

Within-population genetic variation expresses a population's capacity for adaptation to environmental challenges (Agrimonti et al., 2007). Genetic diversity within populations of cultivated crops is often reduced due to strong selection for phenotypic uniformity in commercial varieties (Hagen and Hamrick, 1998). However, the high level of within-population marker-based variation, compared with the variation between populations, found in both red- and white-clover varieties in the current study is consistent with that reported in white clover using RAPD markers (Gustine et al., 2002) and in other outbreeding forage species using AFLP markers (e.g. Guthridge et al., 2001; Fjellheim and Rognli, 2005). The proportion of polymorphic loci scored in both species in the current study (59–79 % in red clover; 57–72 % in white clover) was, however, lower than values reported elsewhere; van Treuren et al. (2005) found that almost 94 % of the amplified fragments obtained in their survey of Dutch white-clover populations were polymorphic. The reference varieties used in the latter study displayed lower levels of polymorphism than populations collected in old grassland and in nature reserves, probably due to selection for increased uniformity in bred varieties. The relatively low values obtained in the current study therefore reflect the fact that the material was predominantly bred varieties rather than natural populations and it could be argued that some erosion of genetic diversity has occurred in this collection of germplasm. Genetic differentiation between populations in this study, as measured by values of F_st, was generally rather low in both species and tended to be greater in white than in red clover (mean values for original populations were 0·084 for red clover and 0·117 for white clover). In particular, the values found for red clover were markedly low compared with other data published for this species, e.g. in the study of Ulloa et al. (2003), which analysed diversity in 20 breeding populations and varieties using RAPDs. However, the small number of original populations in our study makes comparisons with other datasets difficult.

Climatic effects on AFLP variation

In general, it is difficult to achieve a full understanding of the effects of site on plant population differentiation because in many cases the environmental factor to which the populations are responding is not evident, or there may be many factors acting in concert. In addition, the environmental variable of greatest importance for adaptive variation at one site may not be such a limiting factor at other sites. At any given time in any site, populations could be affected by a number of environmental factors, amongst which the most intense or stressful will determine the predominant selective trend (Pérez de la Vega, 1997). In the climatic analysis component of our study, in which genetic differentiation in red-clover populations was analysed, there was no evidence of any genetic shift occurring in the red-clover variety ‘Merviot’ at the UK site, whereas considerable population differentiation was identified in both the Icelandic and Swedish environments. Our analyses indicate that this can be explained by the relatively warm autumn environment exerting limited selection pressure in the UK site. In contrast, genetic shift in red-clover populations in the colder autumn environments of both Iceland and Sweden was clearly evident. In the case of the red-clover WGB-N population the genetic distance between the two survivor populations was greater than their distance from the original population. Similarly, for white clover the genetic distance between the two WGB-N survivors was greater than the genetic shift at each site. Thus, the environments in Iceland and Sweden appeared to exert different intensities of selection. In white clover the original WGB-N population appeared to be well adapted to the Icelandic environment and did not differ markedly from its survivor population, in contrast to the substantial change it underwent in Sweden. A similar genetic shift was observed in the white-clover variety ‘Ramona’ in Sweden. Therefore, in our group of sites, the clearest evidence of genetic shift in populations was observed in Sweden within the relatively short time span of 3 years. The Swedish environment showed the greatest annual amplitude in temperature during the experiment (reflected in the seasonal variation in DGDD). The Swedish site was also the driest, and it seems possible that one or both of these factors exerted directional selective pressure on the populations grown there. Our limited dataset does not allow us to say with certainty which climatic factors may be causing this shift. However, one practical implication of these results is that it would probably be beneficial for legume breeding programmes in Sweden to consider collecting survivor germplasm in order to reinforce existing breeding lines.

Hypothesis (a): in both species the single varieties display a lower level of ‘within population’ genetic diversity than the WGB populations

The high level of intra-population phenotypic variation evident in the WGB populations growing in field plots was not reflected in their genetic diversity as measured by the DNA marker-based, genome-wide method used here. Hypothesis (a) was not confirmed, as no consistent differences in average genetic diversity were observed in the two population groups. In red clover, levels of gene diversity were comparable for the single varieties and the WGB germplasm; in white clover the situation was more complex, as diversity was significantly lower in ‘Ramona’ and WGB-N-O compared with ‘Norstar’ and WGB-C-O. The difference between the two varieties may be explained by the fact that ‘Norstar’ is basically a wild population collected in Norway (Rennebu, Sør Trøndelag; 62 °50'N, 550 m a.s.l.) for seed production, whereas ‘Ramona’ has undergone extensive selection by breeders.

The lowest level of gene diversity in any population was found in the wild white-clover population sampled in Iceland. Under the challenging climatic conditions in this site, it could be the case that traits of high adaptive value tend to be forced to uniformity by stabilizing selection in which the strongest selection pressure acts against extreme phenotypes (Tigerstedt, 1994). This mode of selection was found to operate on physiological traits directly linked to survival in cold climates by Collins et al. (2002), which leads us to consider hypothesis (b).

Hypothesis (b): survivor populations contain a lower level of genetic diversity than the original populations

This hypothesis can also be rejected. In white clover the genetic diversity of the survivor populations was not significantly different from that of the original populations, and in red clover higher genetic diversity was found in the two WGB survivor populations in Sweden than was present in the original populations. This result cannot be explained if we assume that selection pressure acts to reduce genetic diversity. Other explanations rely on the occurrence of sexual reproduction followed by seed set and seedling recruitment in the plots, or on the germination of ‘wild’ red clover seed from the soil seed bank. However, the management imposed on the field plots in all sites made seedling recruitment highly unlikely, and plots were always sown into field areas in which a strict crop rotation policy was followed, or in which no clovers had ever been grown, thus minimizing the possibility of contamination by clover seed germinating from the seed bank (the rotation is 6 years between clover crops in Aberystwyth; in Svalöv no clover had been grown in the plot area since before 1996; in Korpa the plots were sown into permanent grassland in which no clovers had previously been cultivated). An alternative approach to explaining this seemingly anomalous result is described below.

Hypothesis (c): at each site survivor populations within the respective species are genetically closer to each other than the original populations

In red clover, the Swedish environment produced clear convergence in AFLP markers in both the WGB populations sampled at this site, i.e. the genetic differentiation between the WGB original populations was greater than that between the Swedish WGB survivors. The genetic distance between the survivor population of variety ‘Fanny’ and both WGB survivor populations in Sweden was also reduced compared with the original populations. Similarly, in white clover, the Swedish environment reduced genetic distances between all survivor populations. Therefore this hypothesis is confirmed for the Swedish site for both species. In the Icelandic and UK sites only one red-clover population was grown, so the hypothesis cannot be tested. In the case of white clover in Iceland there was a small reduction in genetic differentiation between survivor populations, such that the F_st values for the survivor population of variety ‘Norstar’ and original and survivor WGB-N populations were 0·095 and 0·087, respectively. This hypothesis can therefore be accepted for white clover in Iceland, but to a lesser extent than was observed in Sweden.

Hypothesis (d): survivor populations of the same original population selected at different sites are genetically distinct from each other

In both species, the only population grown at more than one site was WGB-N. Genetic differentiation between red clover WGB-N survivor populations in Sweden and Iceland was 0·087, which was higher than the differences between the original population and the survivors in Sweden and Iceland (0·077 and 0·076, respectively). The corresponding F_st value for white clover Swedish and Icelandic WGB-N survivors was 0·115, compared with values of 0·106 and 0·030 for differentiation between the original population and the survivors in Sweden and Iceland, respectively. Therefore this hypothesis was validated, confirming that environmental differences between sites had a strong effect on the genetic structure of this germplasm.

The AFLP technique supplies valuable information on the genetic structure of germplasm and characterizes the genetic variability of populations (Kölliker et al., 2001). The results of the present study show that the AFLP technique was also successful at identifying populations in which genetic shift had occurred. Nevertheless, there were some anomalies in our results, e.g. the observed increase in genetic diversity (Hj) in Swedish survivors of both WGB populations, which require further consideration. In this case it is possible that some loci were more and others less abundant in the survivor populations when compared with the original populations. If the changes in one direction at some loci were balanced by changes in the opposite direction at other loci, then the average Hj values in the survivor populations would not differ significantly from those in the corresponding original populations. Thus, changes in genetic constitution might not necessarily be associated with reductions in genetic diversity across loci and, in some cases, could even be associated with increases in diversity. If the allelic frequency of some loci changes in the survivor populations with respect to the original ones, genetic differentiation (F_st) values become significant (i.e. genetic shift occurs). This implies that if selection driven by environmental conditions or by genetic drift (i.e. random processes, because random survival of genotypes could produce the same effect in a finite population) was the underlying factor, the observed genetic shifts were not due to the selection of particular alleles in the populations, but to the selection of particular multilocus genotypes. This is plausible because sexual recombination did not occur in populations in this experiment due to the cutting management imposed on the field plots. Other molecular techniques such as SSRs (single sequence repeats) or SNPs (single nucleotide polymorphisms) might reveal more detail in the variation between alleles in the populations, although issues with polyploidy would be a complicating factor.

In conclusion, it was apparent from this study that measurable genetic shift occurred in populations of both these species over a relatively short time period when grown in cold sites in Northern Europe. The degree of genetic change that occurred depended on how well adapted the populations initially were to their growing environment, as well as on the climatic characteristics of the site itself. It was interesting that a modification of the well-established agronomic concept of ‘heat units’ (DGDD) appeared to have some value as a predictor of the potential of environments to promote genetic shift. The value of DGDD in autumn was crucial in this respect, and lower values in autumn were associated with higher levels of genetic differentiation. However, in the case of Sweden, levels of genetic shift were higher than in Iceland, despite the fact that Swedish DGDDs in autumn were somewhat higher. The largest seasonal fluctuations in DGDD were recorded at the Swedish site. Focusing on the change from autumn to winter, which for perennial plants in cold climates represents an important challenge, the Swedish site showed the largest decrease in DGDD, with values in winter falling by 95 % from their autumn level. Corresponding decreases in Iceland and the UK were more gradual, at 85 % and 70 %, respectively. In perennial forage species the physiological preparation for overwintering that occurs in autumn is driven by a combination of changes in ambient temperature and daylength (Junttila et al., 1990). The sudden onset of winter in the Swedish site in this study may provide the key to understanding the pattern of genetic shift observed there, and germplasm that is unable to acclimate quickly may undergo rapid natural selection. However, it should be emphasized that this conclusion is based on the response of a small number of populations and therefore requires more detailed investigation, in which a larger group of common populations are grown in different environments.