Investigating the effects of topography and clonality on genetic structuring within a large Norwegian population of Arabidopsis lyrata

Abstract

Background and Aims

The gene flow through pollen or seeds governs the extent of spatial genetic structure in plant populations. Another factor that can contribute to this pattern is clonal growth. The perennial species Arabidopsis lyrata ssp. petraea (Brassicaceae) is a self-incompatible, clonal species found in disjunctive populations in central and northern Europe.

Methods

Fourteen microsatellite markers were employed to study the level of kinship and clonality in a high-altitude mountain valley at Spiterstulen, Norway. The population has a continuous distribution along the banks of the River Visa for about 1·5 km. A total of 17 (10 m × 10 m) squares were laid out in a north–south transect following the river on both sides.

Key Results

It is shown that clonal growth is far more common than previously shown in this species, although the overall size of the genets is small (mean diameter = 6·4 cm). Across the whole population there is no indication of isolation by distance, and spatial genetic structure is only visible on fine spatial scales. In addition, no effect of the river on the spatial distribution of genotypes was found.

Conclusions

Unexpectedly, the data show that populations of small perennials like A. lyrata can behave like panmictic units across relatively large areas at local sites, as opposed to earlier findings in central Europe.

INTRODUCTION

Limited gene flow within a population may give rise to a correlation between genetic and geographic distance. Documenting this structure is important in order to understand the detailed population dynamics and microevolution. In species with leptokurtic pollen and seed dispersal curves, most of the gametes are deposited close to their site of origin (Schaal, 1980; Hardy and Vekemans, 1999; Austerlitz et al., 2004; Vekemans and Hardy, 2004; Otero-Arnaiz et al., 2005; Clauss and Mitchell-Olds, 2006; Vaughan et al., 2007). Vegetative propagation through surface or subsurface structures can also contribute to spatial structure of populations, the exact pattern depending on the growth strategy (see, for example, Jonsson et al., 1996; van Rossum et al., 2004; Honnay and Jacquemyn, 2008). Such clonal plant populations consist of genets, which are formed by groups of genetically identical units (ramets), which can be more or less connected through, for example, stolons and rhizomes (Oborny and Cain, 1997). Thus, all ramets originating from the same zygote together constitute a genet, even if they become detached from each other at a later stage (Eriksson, 1993). Clonality may give rise to two distinct growth forms, referred to as the so-called guerrilla and phalanx strategy, respectively (Doust, 1981). The guerrilla strategy is in its extreme characterized by an individual sending out its vegetative propagation organs in several directions and far from each other (e.g. Parks and Werth, 1993; McLellan et al., 1997). Contrary to this is the phalanx strategy, consisting of slower advancement through shorter and denser groups of vegetative fragments (e.g. Maddox et al., 1989; Reisch et al., 2007). Although many plant species have been found to exhibit one or the other growth form, numerous species have been shown to fall into the continuum between these extremes (Slade and Hutchings, 1987; Jonsson et al., 1996; McLellan et al., 1997).

Clonal propagation is expected to give rise to high allelic diversity, higher levels of linkage disequilibrium and higher differentiation between populations compared with sexual reproduction, as well as heterozygote excess relative to Hardy–Weinberg expectations (Delmotte et al., 2002; Halkett et al., 2005). These effects should be most prominent in strictly clonal species, as heterozygosity can become fixed within populations and the effect of genetic drift can thus be reduced (Eriksson, 1993; Balloux et al., 2003). For species with both asexual and sexual reproduction, efficient clonal propagation may result in a reduction in sexual reproduction, especially for self-incompatible species or for species experiencing strong inbreeding depression (Ricardo et al., 2006; Honnay and Jacquemyn, 2008). The lack of sexual reproduction also removes the opportunity for creation of new genotypes through fusion and recombination, and slightly deleterious mutations can accumulate (Muller, 1964). Another disadvantage of clonal reproduction is low dispersal ability and a higher susceptibility to fragmentation (see, for example, Leimu et al., 2006). In contrast, clonality can be beneficial in small populations during colonization of new environments and in peripheral populations when sexual reproduction can be suboptimal due to a lack of suitable mates (Silvertown, 2008). As a consequence, a mixture of both clonal spread and sexual reproduction may be beneficial in a heterogeneous environment, where one or the other mode of reproduction can be emphasized depending on local or microsite characteristics (Gaudeul et al., 2007; Gonzales et al., 2008).

The pattern of genetic structure on a continental (Muller et al., 2008; Ansell et al., 2010) and regional level (Clauss and Mitchell-Olds, 2006; Gaudeul et al., 2007) has already been documented in the perennial Arabidopsis lyrata (Brassicaceae), a close relative of the model plant A. thaliana. Two subspecies are currently recognized (Al-Shehbaz and O'Kane, 2002), with A. l. lyrata occurring in North America and A. l. petraea in Europe. In contrast to its relative, A. lyrata is outcrossing (van Treuren et al., 1997; Schierup, 1998) and self-incompatible (Charlesworth et al., 2000; Clauss and Mitchell-Olds, 2006). However, self-compatible populations have been documented in North America (Mable and Adam, 2007). Like several other Arabidopsis species, it is commonly found in highly disturbed habitats such as river beds and edges, screes and roadsides. Arabidopsis lyrata is increasingly used in plant evolutionary and ecological genetics, mainly because of the close phylogenetic relationship to A. thaliana, enabling comparative studies of differences in life history traits (Savolainen et al., 2000; Kuittinen et al., 2004; Abel et al., 2009).

Physical obstacles such as surface topography and water bodies can act as barriers to gene flow on large spatial and temporal scales and result in genetic differentiation within a species (Futuyma, 2005). On smaller scales, environmental features can lead to differentiation between populations due to, for example, topography (Walker et al., 2009), moisture (Xie et al., 2001), and availability of pollinators (Kulbaba and Worley, 2008). The presence of physical obstacles may explain differences in genetic structure between mountainous and widely dispersed Norwegian localities compared with a set of geographically closer lowland populations in Sweden (Gaudeul et al., 2007). Furthermore, in central Europe, populations of A. lyrata have been found to be significantly differentiated over distances <16 km, and fine-scale structure has been documented within populations (Clauss and Mitchell-Olds, 2006). However, the topographic features differ between the different areas, and the results of one area cannot necessarily be generalized. In particular, in Norway, large continuous populations may have a different genetic structure than the forest populations of central Europe, e.g. due to different pollinator behaviour or substrate quality.

In addition to extensive outcrossing, A. lyrata ssp. petraea has been shown to exhibit varying degrees of clonal propagation or facultative inbreeding on a regional scale (Schierup, 1998; Mable et al., 2005; Clauss and Koch, 2006; Gaudeul et al., 2007). Gaudeul et al. (2007) showed that the level of clonality varied among Scandinavian populations, whereas no clonality was documented in the study by Clauss and Mitchell-Olds (2006). Clonality has also been shown in another perennial relative occurring in Europe, A. halleri (van Rossum et al., 2004; Llaurens et al., 2008). Genetic neighbourhood sizes (Wright, 1943) have been shown to be inversely correlated to plant density in populations (Schmitt, 1983). Thus, in a clonal species like A. lyrata a negative relationship could be expected between neighbourhood size and the total number of genets. Due to the clonality, this relationship should be stronger when comparing neighbourhood size with number of genets than with census size (number of ramets). In addition, the level of clonality may be density dependent. In sparsely populated patches, variation in clonality could be due to, for example, local variation in nutrient availability, leading to an increased focus on producing large ramets to obtain resources more efficiently (see Alpert and Stuefert, 1997, and references therein). As A. lyrata throughout Europe is shown to be outcrossing (Clauss and Mitchell-Olds, 2006), high levels of clonality may result in insects transferring incompatible pollen in small and scattered populations. This can then lead to lower seed set and a reduction in fitness within populations. One could still expect that plant individuals limited by pollen should allocate more energy to vegetative growth, which could lead to increased clonal propagation, resulting in a positive feedback mechanism (Wolf et al., 2000; Honnay and Jacquemyn, 2008). Thus, to investigate how self-incompatibility influences fitness, the contribution from both seed and ramet production has to be considered.

Several studies have documented spatial genetic structure (SGS) at a range of spatial scales in A. lyrata, from continental down to local scales (Clauss and Mitchell-Olds, 2006; Gaudeul et al., 2007; Schierup et al., 2006, 2008; Ansell et al., 2010). Clauss and Mitchell-Olds (2006) investigated fine scale structure in a German forest population, where plants are found on separate rocks. The detailed genetic structure has not been studied in other kinds of landscapes, such as the open alpine habitat found in the Norwegian mountains. Furthermore, the degree of clonality may vary between and within populations, as well as how this contributes to any observed spatial genetic pattern. In the present study, the following questions are addressed. (a) What is the level and spatial distribution of clonality within an extensive population of A. lyrata? (b) How is clonality related to population genetic parameters within this population? (c) Is the population random mating, or do limited seed or pollen dispersal result in SGS? Also, it was evaluated to what extent topographic features cause isolation by distance in the population as a whole in the presence of SGS.

MATERIALS AND METHODS

Site and sampling design

Leaf material was sampled in August 2006 within a large Arabidopsis lyrata population at Spiterstulen, Norway (61°37′N, 008°24′E; 1104 m a.s.l.) covering approx. 0·4 km² (Fig. 1). The population occurs mainly along both sides of the river Visa on fluvial and glacial depositions with the growth substrate dominated by sand or gravel. On the fluvial depositions, the most common species group is bryophytes, although several species of graminoids occur together with A. lyrata. However, plant density is low, and non-vegetated ground constitutes on overall a sizable part of the habitat. The glacial river has high water levels due to melting ice during the main flowering months in July–August. The river is in itself not very wide, but as it is glacier fed, the current is rapid, creating vortices of cold air above the water. This can potentially prohibit or reduce insect movement across it. As most of the A. lyrata patches are close to the river, the plants are prone to flooding every year. This may lead to seeds and plants being washed away, but the seeds may be able to establish at downstream sites. Few potential pollinators were seen, but some Dipteran and Lepidopteran species were observed feeding on A. lyrata flowers during sampling.

Fig. 1.

The Arabidopsis lyrata locality at Spiterstulen, Norway (61 °37′N, 008 °24′E; 1104 m a.s.l.). Squares (10 × 10 m; n = 17) with 50 individuals sampled in each are denoted by letters A–Q. The River Visa flows in a northward direction, as depicted by the arrow. Source: Norge Digitalt/gislink.no.

Squares (10 × 10 m) containing at least 50 plants each were marked throughout the population, along both sides of the river, and in a west–east transect across it. In total, 17 squares with a sufficient density of plants were established. In each square, 50 individual rosettes were sampled, and their positions in two-dimensional space (x, y) measured to the nearest centimetre. If possible, only sexually reproducing rosettes were sampled. Sampling within the squares was designed to cover the spatial distribution of plants, i.e. both rosettes close to each other and far from each other were sampled, and more samples were taken from dense patches than from sparse patches. This was done to ensure that clonal growth on different scales could be detected. The total number of A. lyrata rosettes, i.e. the census size, per square varied from 52 to 369. A total of 850 individuals was sampled throughout the population, and plant material was then dried at 35 °C for 24 h and afterwards stored at room temperature. A summary of census sizes, altitudes and co-ordinates for each square is shown in Table 1.

Table 1.

Description of the 17 squares of A. lyrata investigated in this study

Square	Census size	Altitude (m a.s.l.)
A	307	1104
B	108	1104
C	369	1107
D	70	1107
E	188	1110
F	225	1111
G	102	1112
H	55	1142
I	161	1137
J	52	1127
K	58	1105
L	100	1105
M	109	1116
N	314	1117
O	68	1131
P	55	1173
Q	59	1160
Mean ± s.d.	141·2 ± 103·4

Census size is the number of rosettes (ramets) per square (10 m × 10 m).

Genotyping

Plant DNA was extracted using the EZNA™ Plant Kit (Omega Bio-Tek. Inc., Norcross, GA, USA) and the NucleoSpin^® 96 Plant kit (Macherey-Nagel; Düren, Germany), following the manufacturer's protocols. The PCR reaction was carried out in a 10-μL volume consisting of 1 µL DNA template (5 ng μL), 5 µL PCR master mix (QIAGEN Multiplex PCR Master Mix; Qiagen, Hilden, Germany), 3 µL H₂O and 1 µL primer mix. Plant individuals were genotyped for 14 microsatellite loci in two separate multiplex reactions (Table 2). Amplification was performed on an Eppendorf Mastercycler Gradient (Eppendorf, Hamburg, Germany), using the following procedure: 95 °C for 15 min, 34 cycles of 94 °C for 30 s, 50 °C for 1 min 30s and 72 °C for 1 min, with a final extension for 30 min at 60 °C. Capillary electrophoresis was performed on a 3730 DNA Analyser (Applied Biosystems, Foster City, CA, USA). GeneMapper 4·0 (Applied Biosystems) was applied to analyse the fragments and score the alleles.

Table 2.

Summary of marker information for the two multiplexes used in the study

Marker name	Colour	Primer sequences	Allele range	No. of alleles	Reference
Multiplex 1
AthCDPK9	Blue	TCAATCATTGTCCAAAACTTGG	83–93	5	Clauss et al. (2002)
		GAAACTGACTTGGAGAAGGCA
ELF3	Green	CGGAAGGACTGATATACAAGC	294–326	9	Kuittinen et al. (2004)
		TGTTGGGTGTTCTGAAGAT
F20D22	Green	CCCAAGTGACGTCTGGTTTC	170–178	6	Clauss et al. (2002)
		AACAAAATGAGTTTCTCTGCATG
ICE13	Blue	GATCCTTCACCGGGTCTTG	242–248	5	Clauss et al. (2002)
		GTGGTGGAGACTCTTCGAGC
ICE3	Yellow	GACTAATCATCACCGACTCAGCCAC	76–111	14	Clauss et al. (2002)
		ATTCTTCTTCACTTTTCTTGATCCCG
MSAT2·22	Blue	CGATCCAATCGGTCTCTCT	209–226	10	Loudet et al. (2002)
		TGGTAACATCCCGAACTTC
Multiplex 2
AthZFPG	Yellow	TTGCGTTTCCACATTTGTTT	150–166	9	Clauss et al. (2002)
		TGGGTCAATTCACATGTAGAGA
F19K23-483	Blue	GGTCTAATTGCCGTTGTTGC	187–190	4	Clauss et al. (2002)
		GAATTCTGTAACATCCCATTTCC
ICE14	Green	TCGAGGTGCTTTCTGAGGTT	231–237	4	Clauss et al. (2002)
		TACCTCACCCTTTTGACCCA
MHJ24	Blue	CCGTCCTTGATCCTTGAGATTCTGAG	125–130	6	Clauss et al. (2002)
		CAATTCCGAAAATCATATTCATGCACC
nga112	Yellow	TAATCACGTGTATGCAGCTGC	176–204	16	Bell and Ecker (1994)
		CTCTCCACCTCCTCCAGTACC
nga151	Green	GTTTTGGGAAGTTTTGCTGG	93–95	2	Bell and Ecker (1994)
		CAGTCTAAAAGCGAGAGTATGATG
ATTSO392	Blue	TTTGGAGTTAGACACGGATCTG	136–148	8	Clauss et al. (2002)
		GTTGATCGCAGCTTGATAAGC
T15M6	Blue	CATCCATGAATCTTGACTTC	196–201	4	J. Leppälä, unpubl. res.
		GAACAATGCAGAAACTGTG

For each marker, marker name, dye colour (blue = FAM, green = YY, yellow = NED), primer sequences, allele range, number of alleles found in this study, and reference for the marker are shown.

Population genetic structure

In the analyses the number of sampled rosettes, N, is given as number of ramets, while the number of independent multilocus genotypes are defined as genets, G. Unless otherwise specified, analyses and results are based on genet data. For each square, the following measures of genetic diversity were calculated: observed heterozygosity (H_O), expected heterozygosity (H_E), the inbreeding coefficient (F_IS) and proportion of polymorphic loci (PPL). H_O, H_E and PPL were estimated in GenAlex 6·2 (Peakall and Smouse, 2006), while F_IS was estimated with Arlequin 3·11 (Excoffier et al., 2005). Correlation analyses were performed between census size, proportion of genets (G/N) and H_E (Spearman rank correlation); whereas differences in H_E, H_O and F_IS between genets and ramets were tested for using paired Wilcoxon tests. All association analyses, as well as comparisons between genets and ramets, were computed using SPSS version 14·0 (Windows version 14·0, SPSS Inc., Chicago, IL, USA).

Map co-ordinates from each square were imported in MapSource (Garmin, Olathe, KS, USA) to calculate a geographical distance matrix for the squares. This was combined with a genetic matrix based on Nei genetic distance (Saitou and Nei, 1987) to test for isolation by distance across the whole population using a Mantel test (Mantel, 1967) with 999 permutations as implemented in GenAlex (Peakall and Smouse, 2006).

Hardy–Weinberg equilibrium (HWE) was tested across the whole population and within each square using GenePop (Rousset, 2008). AMOVA (analysis of molecular variance) was performed in Arlequin (Excoffier et al., 2005) using geographical position (east or west side of the river; river bank or moraine ridge) as a grouping variable, yielding four groups. To get an initial idea about the degree of substructure in the whole population, Wright's F_ST (Weir and Cockerham, 1984) was calculated in a pairwise manner between squares using Arlequin (Excoffier et al., 2005). The same software was also used to perform a global differentiation test (Weir and Hill, 2002) to investigate significant differences between squares. The proportion of linkage disequilibrium, P_D, based on number of locus pairs in linkage disequilibrium in relation to all locus pairs, was calculated as described in Stenøien and Såstad (1999).

Clonality

When investigating clonality, there is a possibility that two individuals which are genetically similar may have arisen separately through sexual reproduction and are identical by state. This can tested for by calculating the probability of observing a particular genet in an individual through the formula

1

where p_i is the frequency of each allele at a specific locus of the genotype in the sample population, and h is the number of heterozygous loci (Parks and Werth, 1993). When encountering genets containing more than two ramets, the extension

2

has to be added to the calculation (Sydes and Peakall, 1998), where n is the number of times ramets belonging to the same genotype are sampled. This gives the probability of sampling (n–1) copies of the specific genotype. The approach is conditioned on the population being in HWE. It can be assumed that the occurrence of identical multilocus genotypes in the sample is not due to random resampling if P < 0·05 in eqn (2). P_cgen and P_r were calculated for all clonal genets (n = 131) encountered in the total population.

Spatial genetic structure within each square was investigated using the software SPAGeDi 1·2 (Hardy and Vekemans, 2002). The program estimates SGS using pairwise comparisons of individuals in relation to predefined geographical distances. The kinship coefficient described by Loiselle et al. (1995) for comparisons of plants within different distance classes was applied. As we expected, geographically close individuals to be more related than more distant ones, distance classes were set to 5, 10, 15, 20, 25, 50, 100, 250, 500 and 1000 cm. In addition, the software calculated a final distance class (maximum 1414 cm) in squares where ramets could be found >1000 cm apart. For genets in which several ramets were sampled, a median position based on all ramets within the genet was calculated. Kinship was regressed against distance class and 95 % confidence intervals were computed using 10 000 permutations.

Neighbourhood size, Nb, is defined as the effective number of individuals in a random breeding neighbourhood (Wright, 1946). This has lead to the interpretation of Nb as a panmictic unit, although Epperson (2007) pointed out that this is not the case. Hardy and Vekemans (1999) formulated Nb as a function of the variation of dispersal distance,

3

where D is the effective density of individuals, and σ² is the axial variance of dispersal distance. Where direct dispersal estimates are not available, indirect estimates can be calculated as

4

where F is the average kinship coefficient and blog is the regression slope based on the log of spatial distance (Hardy et al., 2006; Hardy and Vekemans, 2007). F and blog were estimated for each square with SPAGeDi 1·2 (Hardy and Vekemans, 2002). Correlations between Nb and census size were tested for, as well as with the number of genets, respectively (Spearman rank correlation). Also the number of genets (G), proportion of genets (G/N) and mean number of ramets per genet were estimated. When screening the data set for potential clones, all individuals containing missing data were excluded. This will cause the measures of clonality to be minimum estimates. Clone size was calculated as the maximum distance between ramets within each genet.

RESULTS

Out of the 850 individuals sampled, 827 could be genotyped. All 14 microsatellite markers were polymorphic in the population as a whole, as well as in all squares, although F19K23 was monomorphic (i.e. frequency of most common allele >0·95) in 15 of the 17 squares. The total number of alleles (± s.d.) per marker varied from 2 (nga151) to 16 (nga112), with an average of 7·3 ± 3·9 alleles (Table 2).

Population genetic structure

A summary of population genetic parameters for each square and across all squares can be found in Table 3. No correlation of population genetic parameters (N, G/N, H_E) was significant after Bonferroni correction (Spearman rank correlation; P < 0·05). Across the entire population, expected heterozygosity (± s.d.), H_E, was found to be 0·480 ± 0·031 and 0·484 ± 0·027 in ramets and genets, respectively. For observed heterozygosity, H_O, the population average was 0·462 ± 0·062 and 0·464 ± 0·045 for ramets and genets, respectively. H_E and H_O were not significantly different between ramets and genets (pairwise Wilcoxon tests; P = 0·29 and P = 0·98 for H_E and H_O, respectively). The inbreeding coefficient for genets, F_IS, varied from –0·17 to 0·19 in the squares, with values positive and significantly different from zero in 8 of 17 squares (1023 permutations). In ramets, the same procedure resulted in F_IS ranging from –0·35 to 0·18, with 9 of 17 squares having values positive and significantly different from zero. The PPL was, on average ± s.d., 0·94 ± 0·03, ranging from 0·93 to 1·00 among squares. It was found that two out of 14 loci were not in HWE (P < 0·05) in the total population. Repeating the same test for each individual square, ten squares had a few loci (1–3) not in HWE. The global test showed only square Q not to be in HWE. The proportion of linkage disequilibrium ± s.d., P_D, in genets was 0·26 ± 0·11.

Table 3.

Summary of genetic parameters for each square and across the whole population, split into ramets (all rosettes included, n = 827) and genets (only independent multilocus genotypes included, n = 546)

					Genets			Ramets
Square	N	G	G/N	PPL	He	Ho	FIS	He	Ho	FIS
A	45	38	0·84	92·9	0·479	0·435	0·093	0·486	0·442	0·080
B	49	41	0·84	92·9	0·512	0·518	−0·028	0·521	0·521	0·127
C	50	23	0·46	92·9	0·489	0·466	0·018	0·470	0·485	−0·025
D	50	38	0·76	92·9	0·512	0·450	0·112	0·515	0·452	− 0·055
E	50	41	0·82	100·0	0·491	0·437	0·096	0·491	0·436	0·106
F	50	39	0·78	92·9	0·512	0·482	0·057	0·510	0·483	0·092
G	47	40	0·85	92·9	0·507	0·520	−0·032	0·515	0·524	0·042
H	46	36	0·78	92·9	0·453	0·409	0·102	0·444	0·406	−0·036
I	49	34	0·69	92·9	0·496	0·441	0·097	0·481	0·428	0·079
J	46	20	0·43	92·9	0·460	0·445	0·057	0·442	0·386	0·092
K	49	38	0·78	92·9	0·522	0·473	0·098	0·523	0·466	0·104
L	47	35	0·74	92·9	0·477	0·417	0·122	0·480	0·409	0·135
M	49	31	0·63	92·9	0·480	0·395	0·190	0·475	0·390	0·180
N	50	21	0·42	92·9	0·443	0·474	−0·071	0·453	0·484	−0·082
O	50	37	0·74	100·0	0·471	0·451	0·034	0·470	0·405	0·124
P	50	11	0·22	100·0	0·494	0·578	−0·137	0·471	0·631	−0·351
Q	50	23	0·46	92·9	0·424	0·489	−0·172	0·415	0·503	−0·234
Mean	48·6	32·1	0·66	94·1	0·484	0·464	0·037	0·480	0·462
s.d.	1·7	9·0	0·19	2·8	0·027	0·045	0·097	0·031	0·062

N = number of ramets genotyped, G = number of genets detected, G/N = proportion of genets, PPL = proportion of polymorphic loci, H_E = expected heterozygosity, H_O = observed heterozygosity, F_IS = inbreeding coefficient.

Bold numbers indicate values significantly different from 0 (P < 0·05).

Pairwise F_ST values between squares (n = 136) had a median value of 0·039, ranging from 0·004 to 0·123. A global test of square differentiation showed, however, that none of the comparisons were significantly different from zero (P > 0·05). Furthermore, there was no pattern of isolation by distance across the population (r = 0·069, P = 0·28) (Fig. 3). Separating squares into groups based on location (side of river and river bank/moraine ridge – four groups) in an AMOVA analysis explained the highest proportion of variation among groups, although the component of variation was still very low (1·7 %; P < 0·001). Almost all variation (95·3 %) was due to variation among individuals within squares. This supports the global differentiation analysis, showing very low differentiation between squares within the population.

Fig. 3.

Distribution of pairwise F_ST-values (n = 136) from 17 squares of Arabidopsis lyrata at Spiterstulen, Norway. NS = comparisons not significantly different from 0. The distance between squares A and H was 1·2 km.

Clonality

The total sample consisted of 546 genets, of which 415 were present in a single ramet each. After excluding all genets with missing data, the 423 remaining genets contained on average (± s.d.) 1·7 ± 1·8 ramets (range 1–20; Fig. 2A). Mean genet size (± s.d.) was 6·4 ± 22·5 cm (range 0–267·2 cm; Fig. 2B). In most of the squares (n = 12), only a subsample of all rosettes was sampled. For those squares (C, J, N, P and Q; Fig. 1 and Table 1) where most of the rosettes (>80 %) within a square were sampled, genet size was on average (± s.d.) 10·4 ± 33·4 cm (range 0–267·2 cm) and the number of ramets in each clone was 2·1 ± 2·8 (range 1–20). Most genets were small, and the growth form of ramets within genets condensed (Fig. S1). Still, for some genets, the ramets were spread out or intermixed with other genets (Fig. S1). The proportion of genets, G/N, varied from 0·22 in square P to 0·85 in square G, with a grand mean (± s.d.) of 0·66 ± 0·19 (Table 3). For genets containing more than one ramet, the probability of encountering a second, identical genotype, P_cgen, varied from 7·93 × 10⁻¹² to 0·007, whereas the probability of encountering more than two identical genotypes, P_r, varied from 6·9 × 10⁻⁶¹ to 0·0003, suggesting that identical genotypes are ramets of the same clone in all cases considered. The mean neighbourhood size (Nb) ± s.d. across all squares was 22·6 ± 9·7, ranging from 10·4 to 38·5. The number of genets per square, G, had almost the same range (11–41) as Nb, although its mean ± s.d. was higher (32·1 ± 9·0). If only considering squares which were exhaustively sampled, Nb ± s.d. was 20·0 ± 9·1. The number of genets, G, was negatively correlated with Nb (P = 0·027), but not after Bonferroni correction. All squares showed indications of SGS at a fine scale according to SPAGeDi analyses (Fig. 4). Genetically similar plants were found at distances up to 15–25 cm, but in a few squares there was significant SGS up to 25–50 cm, and even 100–250 cm in square Q.

Fig. 2.

Summary of genet size in the Arabidopsis lyrata population at Spiterstulen, Norway. Individuals with missing data are excluded. (A) Size distribution of genets based on the number of ramets within each genet; (B) genet size shown as the greatest distance measured between two ramets in a genet. A genet size of 0 means the genet was found in only one ramet.

Fig. 4.

Kinship coefficient plotted against distance interval in all 17 squares from the Spiterstulen population. Kinship coefficient from Loiselle et al. (1995).

DISCUSSION

Extensive clonality was found in A. lyrata at Spiterstulen, resulting in a fine-scale spatial genetic pattern. Despite this clonality, the population as a whole is random mating without any specific substructure.

Clonality

Stochastic events due to biotic or abiotic factors are expected to affect species capable of clonal growth less than non-clonal species (Honnay and Bossuyt, 2005), thus loss of genetic diversity through genetic drift can be reduced in clonally reproducing species (Balloux et al., 2003). In this way, clonality can be viewed as a form of insurance, in the same way as a seed bank, under conditions that are not optimal for sexual reproduction. Frequent disturbances should not favour clonality (Silvertown, 2008), although A. lyrata in Norway occurs in habitats prone to seasonal flooding. Through clonal reproduction, plants can increase the chances of survival in harsh environments, a clear advantage at high altitudes and alpine areas (Bliss, 1971). Clonal propagation is likely to influence SGS, and can do so over considerable distances, depending on the size and abundance of clones (see, for example, Parks and Werth, 1993; Vaughan et al., 2007). Gaudeul et al. (2007) showed that clonality in A. lyrata was evident but highly variable in Scandinavian populations, whereas no indication of clonality was found in a study from central Europe (Clauss and Mitchell-Olds, 2006). This difference in clonality between Spiterstulen and central Europe can thus be a reflection of reduced levels of sexual reproduction in the species' distribution margin in Europe (Eckert, 2002; Rasmussen and Kollmann, 2008).

Clonal structure in A. lyrata

It should be stressed that even though the present study showed considerable clonality in A. lyrata (mean G/N = 0·66), this is most likely an underestimate due to several reasons. First, clone size may have been underestimated as mostly sexually reproducing rosettes were selected and not all rosettes were sampled within each square. Resource allocation within a genet may enable flowering in some fragments, while others stay vegetative (Alpert and Stuefer, 1997). Also, since all rosettes lacking full data when identifying clonal genets were excluded, some parts of clones may have been missed. The clonality level found in the present study is still comparable to what was found in the relative A. halleri (van Rossum et al., 2004; G/N = 0·35–0·94), although a later study by Llaurens et al. (2008) documented an average G/N of 0·98. In the A. halleri populations, however, plant density was considerably higher [33–89 ramets m⁻² (van Rossum et al., 2004); 14 ramets m⁻² (Llaurens et al., 2008)] than in the present study (0·5–3·7 ramets m⁻²). The high occurrence of identical genotypes at Spiterstulen could be due to either clonality or low ability to identify different genets, but as the markers had very high power to distinguish between the genets, we can be confident in treating identical genotypes as ramets of the same genet. It was found that most genets are rather small (6·4 cm and 10·4 cm on average for all squares and for squares with exhaustive sampling, respectively). Few ramets per genet on average contributed to this, but also a pattern is seen where ramets of the same clone either congregate close to each other, i.e. the phalanx strategy, or more rarely, they send out long runners, following the guerrilla strategy. This is also obvious for squares where close to all individual rosettes were sampled. Thus, clone size and shape varies within and between squares (Fig. S1), showing that A. lyrata, as many other species (Jonsson et al., 1996; McLellan et al., 1997; Alberto et al., 2005; Johansen, 2009) does not adhere to either the guerrilla or the phalanx growth strategy, but lies somewhere in between. The former strategy may be more beneficial in patches where resources are more scarce (Doust, 1981), but as no significant correlation between plant density and level of clonality was found, competition between ramets is most likely not the main cause of the spatial distribution pattern of plants seen in the squares. Another, perhaps more likely factor, may be fine-scale variation in soil texture and resources, where chance establishment in poorer patches could cause an individual plant to put more effort into spreading its ramets far about. Finally, competition from other species could potentially influence the distribution of A. lyrata in this population, but as the cover of other plant species was very low in all the squares investigated, this was not taken into account in this study.

Genetic diversity varied between squares, but was not correlated with the proportion of genets (G/N). As such, clonality does not lead to a loss of alleles in the Spiterstulen population, conforming to the theoretical predictions that genetic diversity can be preserved in clonal species as the effect of genetic drift is reduced (Balloux et al., 2003). Clonality can be considered a way of exploiting existing opportunities to the best possible degree. In favourable patches, a successful genet may perhaps benefit more from expanding its distribution through the production of new ramets rather than creating new, possibly less-adapted genets through sexual reproduction. A potential risk involved in having a high dependence on clonality for a self-incompatible species becomes evident as the clones grow bigger. If pollinator activity is high or pollen transfer efficient, clogging of the stigma with pollen containing a similar S-allele can cause low sexual reproduction in plants (reviewed in Barrett, 2002). Thus, availability of unrelated, compatible pollen can be reduced, resulting in lower seed set. This could lead to seed production in A. lyrata to be more limited by mates than by resources, as was shown in a recent study on the same species (Sandring and Ågren, 2009). Still, the present results show that gene flow is not restricted in the population, but as the effect of pollen transport has not been separated from seed transport, it is not possible to say whether pollen limitation was observed.

The concept of neighbourhoods was originally developed by Wright (1943), and can be said to describe an area in which breeding is more or less restricted for the individuals within it. Neighbourhoods can be useful in describing to some degree spatial structure within a population in terms of the relationship between dispersal and drift, but not when the occurrence of long-range dispersal is high (Fenster et al., 2003). In the present study, neighbourhood size across the whole population was similar to when only considering the squares in which close to all rosettes were sampled. The present estimates of neighbourhoods show a trend for a negative correlation between neighbourhood sizes and level of clonality (P = 0·05, not significant after Bonferroni correction), thus in squares with many clones exhibiting high numbers of ramets, seed set is expected to be reduced compared with squares with few clonal genets. This can create a differentiation within the population, where marginal patches develop into sites of plants focusing mainly on vegetative propagation, while central patches have plants putting more energy into seed production. The lack of correlation between census size and proportion of genets in the squares indicates that clonality may not be density dependent. Thus, a lack of sites for establishment does not necessarily cause plants to switch from sexual reproduction to vegetative propagation through rhizome growth.

In the previous study addressing population structure in Scandinavian A. lyrata populations (Gaudeul et al., 2007), the proportion of genets in Norwegian populations was found to vary from 0·55 to 1, with an average ± s.d. of 0·84 ± 0·13. In the Spiterstulen population, however, no clonality was shown due to the absence of identical genets. This contrasts with the present study, where clonality was found to be quite high, resulting in G/N ratios within some squares to be as low as 0·22. The main reason for this deviation compared with the previous study has to do with the sampling scheme. Whereas Gaudeul et al. (2007) in most of the populations sampled plants with at least 50-cm intervals, plants were sampled down to 0·5 cm apart in order to discern in more detail the local genetic structure. For most squares, 50 cm was above the maximum clone size; hence, it is no surprise that the sampling scheme of Gaudeul et al. (2007) failed to discover any clonality in Spiterstulen. In five squares (C, J, N, P and Q), G/N was <0·5 as a result of either many small or several large genets, all containing two or more ramets. If the three Norwegian populations (N3, N7 and N14) in Gaudeul et al. (2007), where distance between sampled individuals were only 20 cm, is considered, the proportion of genets rises to 0·71. This is quite close to the present estimate for the Spiterstulen population (0·66), indicating that when sampling A. lyrata, a distance of 20 cm between samples is sufficient to detect most fine-scale SGS. Another effect of the high proportion of clonal individuals in the Spiterstulen population was the relatively high levels of linkage disequilibrium, similar to what was found in populations exhibiting high levels of clonality in Gaudeul et al. (2007), but considerably higher than for 14 Norwegian A. lyrata populations where close to no clonality was evident (P_D = 0·08; M Falahati-Anbaran et al., NTNU, Norway, unpubl. res.).

Population genetic structure

The lack of any isolation by distance pattern on the population scale illustrates the fact that the SGS is prevalent only at very fine spatial scales, i.e. within squares, and is caused by clonality. By considering larger distance classes, i.e. larger distances between plants (>15–25 cm), very little evidence of spatial structure is found in neutral loci. This could be compared with results from A. lyrata in Iceland, where there are indications of isolation by distance on very local scales (10–20 m; Schierup et al., 2006), but not on regional scales (200–300 km; Schierup et al., 2008). However, Schierup et al. (2008) looked at self-incompatible haplotypes, which are subject to selection in populations across a large region, while in Norway, Gaudeul et al. (2007) found significant isolation by distance on the same scale (hundreds of kilometres) using neutral loci. The reason for this difference is not clear; the Icelandic populations could have been sampled in localities which are more interconnected with each other through an open landscape, or, more likely, as the self-incompatible alleles are subject to selection, regional genetic structure based on such markers will be different than for microsatellites. It could also be that the Icelandic population of A. lyrata has dispersed and reached its current distribution at a much later time compared with Norway, not giving enough time to create an IBD pattern (Schierup et al., 2008).

Contrary to what was expected, genetic structure on the population level was not found. Only two microsatellite loci deviated from HWE across the whole population, and a test for isolation by distance did not reveal any significant pattern. Thus it seems that despite the topographic barriers, there is a substantial gene flow throughout the large population. This is in contrast to what has been found in other studies considering SGS on such local scales, where indications have been found of restricted gene flow and subdivision of the population (Schaal, 1980; Loiselle et al., 1995; Clauss and Mitchell-Olds, 2006; Schierup et al., 2006; Vaughan et al., 2007). It is unknown whether this is due to seed or pollen transport. Pollen transport should be considered more efficient in long-distance transportation, but as this A. lyrata population is mainly situated along streams, water transport of seeds can contribute as well. It was anticipated that the river might limit insect movement and result in differentiation between squares on each side of the river. No such differentiation was found, but direct measures on pollen transport and the effect of the river as a vector of gene flow will be necessary. Clauss and Mitchell-Olds (2006) found that much of the genetic variation within the central European Plech population was due to substructure, likely a consequence of the specialized habitat of the plants (rock outcrops). In Norwegian populations, A. lyrata typically inhabits river banks and has more continuous populations. This may be the main reason for the lack of any obvious structure in the Spiterstulen population.

Conclusions

It has been shown that vegetative spread through rhizomes is a common way of propagation in A. lyrata at Spiterstulen, shaping the spatial population structure at very fine scales in this Norwegian high altitude population. The lack of a clear structure across the population as a whole shows, however, that A. lyrata can act as a more or less random mating unit over considerable distances on a local scale. Contrary to what should be expected, levels of clonality are not related to plant density, and other factors such as soil texture and nutrient availability may play a more crucial role for this pattern. Seed transport can be facilitated by water flow, especially during flooding events, but the relationship between pollen transport, seed transport and population structure is still unclear. Finally, as neutral microsatellite markers in this study show that the overall spatial genetic pattern in this population is random, it may be easier to detect possible patterns of differentiation due to local microsite variation against this background.

SUPPLEMENTARY DATA

Supplementary data are available online at www.aob.oxfordjournals.org and show the distribution of A. lyrata clones in two squares that were subjected to exhaustive sampling.