Plant Syst. Evol. 237: 137–151 (2003) DOI 10.1007/s00606-002-0248-3

A brief evolutionary excursion comes to an end: the genetic relationship of British species of sect. Gentianella ()

M. O. Winfield1, P. J. Wilson2, M. Labra3, and J. S. Parker1

1Cory Laboratory, Department of Sciences, University of Cambridge, Cambridge, UK 2Wessex Environmental Associates, Redlynch, Salisbury, UK 3Dipartimento di Biologia, Sezione di Botanica Generale, Universita` di Milano, Milano, Italy

Received May 20, 2002; accepted October 28, 2002 Published online: March 31, 2003 Springer-Verlag 2003

Abstract. There is much taxonomic confusion in species were genetically more similar to each other the genus Gentianella section Gentianella, five than they were to individuals of the same species species of which are found in Britain. Gentianella from other populations. It is proposed that G. anglica is a British endemic restricted in its distri- anglica may be an early flowering form of G. ama- bution to the chalk downland of southern England. rella that has been created and maintained as a It is considered to be threatened because of habitat consequence of former grassland management loss, and due to hybridisation with the closely practices. related, widely distributed G. amarella. The Red Data Book species, G. uliginosa, shares morpho- Key words: AFLP, conservation, extinction, logical similarities with the other two species but genetic diversity, Gentianella anglica, Gentianella has a very restricted distribution. Individual amarella, Gentianella uliginosa, hybridisation, from populations across the range of these three universal cpSSRs. species were analysed using AFLPs and universal cpSSRs to determine the degree of genetic variabil- Five species of the genus Gentianella section ity within and between them, and to look for Gentianella are recorded in the British Isles. evidence of hybridisation. Samples of the two other According to the Flora Europaea, these species Gentianella section Gentianella species found in fall into three main groups that are essentially Britain, G. campestris and G. germanica, were also well defined morphologically and create few included in the study. The analysis of chloroplast taxonomic problems: the amarella group con- SSRs was not informative for G. anglica, G. ama- tains G. amarella (L.) Bo¨rner, G. anglica rella and G. uliginosa, while G. campestris and G. (Pugsley) E. F. Warb. and G. uliginosa (Willd.) germanica differed from each other and the other Bo¨rner while the other two species, G. campes- three species at a single locus. Principal co-ordinate analysis of the AFLP data revealed only three tris (L.) Bo¨rner and G. germanica (Willd.) distinct groups: one group contained G. campestris Bo¨rner give names to their own groups (Tutin samples, another contained all samples of G. ger- et al. 1972). However, in general, the manica, and a third contained all individuals from of section Gentianella is difficult and confused the other three species. In mixed populations of G. in Britain and in the rest of Europe. Problems anglica and G. amarella, individuals of the two arise principally from phenology; variation in 138 M. O. Winfield et al.: Genetic structure of Gentianella spp. in Britain plant size and flowering time may be correlated develops in the absence of regular grazing and with germination time (autumn, winter or cutting: it requires the sward to be kept to a spring). It is also proposed that hybridisation height of about 2–3 cm (Wilson 1999). and backcrossing between species occur fre- G. amarella, the autumn gentian, is a quently so that hybrid swarms are presumed circumpolar species with its centre of distribu- to result. Finally, the characters which differ- tion in southern Europe. It has a wide distri- entiate species and subspecies are mainly bution in Britain and is common and quantitative rather than qualitative, so that widespread in calcareous pastures and on differences such as node number and pedicel dunes; it is found on both limestone and chalk. proportions underlie identification. It is unclear With the exception of a small number of how much of the variation detected in natural Cornish sites, all populations of G. anglica populations of these species is environmental in occur nearby or intermixed with populations origin and how much is due to genetic differ- of G. amarella (Rich et al. 1997). However, the ences since cultivation from seed under con- two species are thought to be completely trolled conditions has proved problematic. sexually isolated because of their distinct and However, it is clear that these Gentianella non-overlapping flowering times; G. anglica species show considerable phenotypic plasticity flowers from May to June, while G. amarella which adds to the taxonomic confusion. flowers from July to October. In recent years, The early gentian, Gentianella anglica, one however, there have been reports of popula- of Britain’s few endemic species, is included on tions containing hybrids (Margetts and David the UK Government Biodiversity Action Plan 1981, Rich et al. 1997). Indeed, G. anglica list of priority species and receives protection subsp. cornubiensis N. M. Pritch, which until under Schedule 8 of the Wildlife and Coun- 1996 was considered a distinct taxonomic unit, tryside Act 1981. It is also protected under is now thought to be the hybrid G. angli- Appendix I of the Bern Convention and ca · G. amarella (G. · davidiana). This has led Annexes IIb and IVb of the European Com- to the additional concern that G. anglica might munity Habitats and Species Directive (Wigg- become extinct by hybridisation with its more ington 1999). This small annual is found abundant relative (Rich et al. 1997). exclusively in closely grazed, calcareous grass- The dune gentian, G. uliginosa, is a species lands and is principally restricted to the chalk with a very restricted range in Britain, being downlands of south eastern and southern found principally at four or five sites in Wales England. It is preferentially found on south where it occurs in dune slacks or in wet and west facing slopes. The species is consid- hollows (Wigginton 1999). There has also been ered of conservation importance because much report that it has been found on Colonsay in of its natural habitat, nutrient poor calcareous Scotland (Gulliver 1998). It is a Red Data grassland, is being destroyed or altered, and in Book species that is considered vulnerable in many places the species has suffered a sharp both Britain and Europe (Holyoak 1999, decline (Darby 2000, Davis and Duckworth Wigginton 1999). The species, which flowers 2000). It is being lost from many of its former from May to October, is always found in close sites following the ploughing of downland, and proximity to populations of G. amarella. the application of nitrogen fertilisers and In order to understand genetic differentia- herbicides to increase productivity – atmo- tion within Gentianella sect. Gentianella in spheric pollution has aggravated this problem Britain, and more specifically the status of because it results in rainfall with a high the three species belonging to the amarella nitrogen content (Bobbink 1998). At other group, a molecular analysis, using AFLPs and sites, G. anglica has been lost as a result of universal cpSSRs, to determine the level of changes in grassland management, since it DNA variation was carried out. This paper cannot survive in tall, dense vegetation that describes research that was aimed to resolve M. O. Winfield et al.: Genetic structure of Gentianella spp. in Britain 139 more clearly the degree of genetic differentia- from plants that were in flower (i.e., late May– tion between the three species G. anglica, June) and from sites where there had been no G. amarella and G. uliginosa, and to look for previous reports of hybridisation with G. amarella. evidence of hybridisation between them that A similar rationale was used for the collection of could be threatening their integrity. We pre- G. amarella samples; collections were made in late sent results obtained from a study of 115 August/September from six populations. Samples were also collected from four hybrid populations; individual plants from 18 separate populations i.e., populations in which G. anglica, G. amarella representing the range of the three species in and their hybrid, G. · davidiana, had been report- Britain. In addition, samples of G. campestris ed. From two of these populations, Eastbourne and G. germanica were included in order to (population 13, Fig. 1) and Stackpole Warren provide outgroups against which to compare (population 16, Fig. 1), samples were collected genetic variability in the other three species. early in the year as representative of G. anglica. From the other two hybrid populations, Bindon Materials and methods Hill (population 14, Fig. 1) and Gear Sands (population 15, Fig. 1), samples were collected at Plant material. Samples of the five species used in three separate times of year (four individuals from the study were collected from a total of 134 each site at each visit): the first collection was made individual plants representing 23 populations to correspond with the flowering time of G. anglica (Fig. 1). Samples of pure G. anglica were collected (late May/early June); the third collection was from six populations covering the range of the made to correspond with that of G. amarella (late species in the U.K. (Fig. 1). Leaves were collected August/early September); the second collection was

G. uliginosa 17. Whitford (5) 18. Oxwich (5)

G. campestris 19. Holmsley airfield (5) 20. Kit Hill (3) 21. Colonsay (3) G. germanica 12 22. Sands Bank (3) G. anglica 21 23. Turville (5) 1. Elsea Pit (10) 2. Ivinghoe (5) 7 3. Afton Down (5) 4. Winklebury Hill (5) 5. Prescombe Down (5) 6. Aucombe Bottom (9)

G. amarella 7. Cow Green Reservoir (5) 1 8. Therfield Heath (7) 11 9. Grangelands (5) 10. Selsley (5) 2 8 11. Llanymyrech (5) 19 18 12. Colonsay (5) 17 22 Fig. 1. Map of Britain 16 6 23 Hybrid populations 5 9 showing the 23 sites from 13. Beachy Head (5) 10 4 which samples were collect- 14. Bindon Hill (12) 13 15 ed. The numbers in brackets 3 15. Gear Sands (12) 14 refer to the number of indi- 16. Stackpole Warren (5) 20 viduals collected at each site 140 M. O. Winfield et al.: Genetic structure of Gentianella spp. in Britain between these two times (mid-July). In each case, Weising and Gardner (1998). Of the 10 primer pairs samples were taken only from plants in flower used in Weising and Gardner (1998) only seven exhibiting the characteristics of G. anglica, G. ama- (ccmp2, ccmp3, ccmp4 ccmp6, ccmp 7, ccmp8 and rella or G. · davidiana, respectively. Material was ccmp10) were used in this study. Analysis was also collected from two populations of G. uliginosa, performed on five individuals, selected at random, three populations of G. campestris and two popu- from each of the five species used in the study. lations of G. germanica. Amplification products were separated as described N.B. Since, in Britain, Gentianella is protected, above and visualised using silver staining. permission was obtained from the relevant local Cluster analysis. Bands from AFLP analysis authorities before material was collected. were scored as present (1) or absent (0) and the DNA extraction and AFLP analysis. Genomic resultant data matrix analysed using the GEN- DNA was extracted according to manufacturer’s STAT 5 statistical programme (Payne et al. 1993). instructions using Nucleon PhytoPure DNA Ex- Pairwise similarities between samples were estimat- traction Kits (Amersham Pharmacia). AFLP anal- ed on the basis of the number of shared fragments ysis was performed as described by Winfield et al. using the Jaccard coefficient. A similarity matrix (1998). Genomic DNA was restriction digested was constructed for the combined data. Cluster using the enzyme combination PstI and MseI, and analysis was performed using the group average then ligated to PstI (5¢-Biotin-CTC GTA GAC method (UPGMA) and a dendrogram constructed. TGC GTA CAT GCA-3¢;3¢-CAT CTG ACG CAT Bootstrap analysis (Felsenstein 1985) using the GT-5¢) and MseI (5¢-G ACG ATG AGT CCT programme WinBoot (www.irri.org/winboot.htm) GAG-3¢;3¢-TAC TCA GGA CTC AT-5¢) adapters: was performed with 1000 repetitions to provide Biotinylated fragments were selected using pre- confidence levels for the dendrogram. The com- washed, streptavidin beads (Dynal). Selective PCR bined similarity matrix was also used in principal was initially performed using several different co-ordinate analysis and the resultant first two primer sets on four samples of DNA. On the basis components were plotted. All data analysis was of band number and clarity, two primer pairs, carried out on the data set including monomorphic MseI.1/PstI.1 and MseI.2/PstI.1, were chosen for bands (N.B. polymorphic bands that appeared in the analysis: MseI.1, 5¢-GAT GAG TCC TGA only one individual were not included in the data GTA Aga a-3¢; MseI.2, 5¢-GAT GAG TCC TGA set, and bands that were absent in only one GTA Aac a-3¢, PstI.1, 5¢-GAC TGC GTA CAT individual were considered to be monomorphic). GCA Gac-3¢. The MseI primer was end-labelled Measures of genetic diversity. Nei’s genetic using [c-33P]ATP (Amersham). The polymerase distance (Nei 1978) was calculated for all five chain reaction was carried out in a 20 ll volume species, and for the populations within the three containing: 2 ll of DNA template bound to beads; species G. anglica, G. amarella and G. uliginosa 2 ll of 10X PCR buffer (200 mM Tris, 5 mM KCl, using the programme Popgen32: in the latter case, 33 15 mM MgCl2, pH 8.4); 30 ng P-labelled MseI analysis was carried out on the data set with and primer; 30 ng of the PstI primer; 1 unit of Taq without monomorphic bands included in order to DNA polymerase (BRC). The thermal profile was: allow the widest possible comparison with other an initial 5 min denaturation at 94 C; 12 cycles of studies. Mantel tests were carried out on the 94 C for 30 s; 65 C for 30 s ()0.7 C each cycle); populations from the species groups to determine 72 C for 1 min. The reaction was then continued whether genetic distance was correlated with geo- for an additional 23 cycles with a constant anneal- graphic distance. ing temperature of 56 C. The reaction was com- Analysis of cpSSRs. The results of the chlo- pleted with a final extension at 65 C for 30 min. roplast microsatellite analysis were sufficiently Amplification products were separated on 6% simple that they could be interpreted directly. No denaturing, acrylamide gels. After transfer to statistical analysis was carried out on these data. 3MM paper (Whatman) and drying, gels were exposed to Kodak Biomax film for from 1–4 days. Results cpSSRs. Amplification of chloroplast simple sequence repeats was performed according to the AFLP analysis. The total number of clear, protocol using radioactively labelled primers of scoreable bands produced by each primer pair M. O. Winfield et al.: Genetic structure of Gentianella spp. in Britain 141

Table 1. Total number of amplified bands, poly- Table 2. The band differences present in the three morphic bands and percentage polymorphic bands species G. anglica (A), G. amarella (B), G. uliginosa from two both primer pairs combined (D) and in the individuals from the hybrid popu- lations (B). The figures in the clear boxes represent No. of No. of Percentage of the frequency of the band found in that particular bands polym. bands polym. bands species; for example, bands 1, 2, 4, 6, 9, 25, 31 and G. anglica 126 20 15.9% 32 were found in some samples of G. anglica but Hybrid 122 15 12.3% were absent in the other three groups pops. G. amarella 120 18 15.0% A B CD G. uliginosa 114 0 0.0% 1 0.06 G. campestris 101 17 16.8% 2 0.06 G. germanica 104 23 22.1% 3 0.19 0.11 0.15 4 0.08 5 0.89 0.46 1.00 6 0.06 was 83 and 86 (MseI.1/PstI.1 and MseI.2/ 7 0.92 1.00 1.00 1.00 PstI.1), respectively. Of these, 63 (75.9%) and 8 1.00 1.00 0.81 1.00 60 (69.7%) were polymorphic. In the com- 9 0.14 bined data set for all five species there were 169 10 0.08 11 0.83 0.81 0.85 1.00 bands of which 123 (72.8%) were polymor- 12 0.17 0.08 phic. The number of bands in each of the five 13 0.72 1.00 0.17 1.00 groups and in the hybrid populations, and 14 1.00 0.94 0.58 1.00 the percentage of polymorphic bands is shown 15 0.47 0.64 1.00 1.00 in Table 1. Among the 115 samples from the 16 1.00 0.86 1.00 1.00 17 0.28 three species G. anglica, G. amarella, G. uligi- 18 0.11 0.23 nosa and the hybrid populations, the number 19 1.00 1.00 0.54 of bands scored was 134, of which only 34 20 0.11 (25.4%) were polymorphic (Table 2). In the 21 0.81 0.92 1.00 22 0.78 1.00 0.54 1.00 samples of G. anglica, there were eight poly- 23 0.14 0.19 0.38 1.00 morphic bands that were not present in 24 0.14 G. amarella, and, conversely, that were two 25 0.08 bands in samples of G. amarella that were 26 1.00 1.00 0.88 1.00 27 0.28 0.06 0.19 1.00 absent in G. anglica: in all cases these bands 28 0.11 0.50 0.73 were of low frequency in the respective species 29 0.28 0.44 0.92 and defined populations rather than the species 30 1.00 1.00 0.92 1.00 groups (Table 2). There were no species diag- 31 0.19 32 0.06 nostic bands. 33 0.35 Principal co-ordinate analysis (the first two 34 1.00 1.00 0.92 1.00 components accounting for 68% of the varia- tion) based on Jaccard similarities permitted Band absent the identification of three main clusters of 1.00 Band monomorphic individuals (Fig. 2a): individuals of G. campes- 0.14 Band polymorphic tris formed one cluster, individuals of G. ger- manica formed a second cluster, whilst a third cluster contained all samples of G. amarella, support for the G. campestris and G. germanica G. anglica, G. uliginosa and those from hybrid groups – the population groupings within these populations. A UPMGA dendrogram of band species were also well supported (Fig. 2b). The sharing similarity showed the same three individuals of G. anglica, G. amarella, G. uligi- groups (Fig. 2b). There was strong bootstrap nosa and those from the hybrid populations 142 M. O. Winfield et al.: Genetic structure of Gentianella spp. in Britain

a 0.4 B 0.2

0.0 -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2

-0.2 A

-0.4 2nd co-ordinate (14.2%) 2nd co-ordinate C

-0.6

1st co-ordinate (53.8%)

G. anglicaHybrid pops. G. amarella G. uliginosa

b

G. anglica G. amarella Hybrid pops. G. uliginosa

A

100

97 100 98 B 95 94 100 99 C 50 60 70 80 90 100 Percentage similarity

Fig. 2. a) PCO plot showing the relationship between the five species included in the study, and b)UPGMA dendrogram showing the same relationship. The bootstrap values placed above the forks of the dendrogram show the percentage of times the group to the right of that fork occurred – only percentages greater than 75% are shown. A ¼ G. anglica, G. amarella, G. uliginosa and hybrid populations; B ¼ G. campestris; C ¼ G. germanica) M. O. Winfield et al.: Genetic structure of Gentianella spp. in Britain 143 formed a single, well supported cluster. In Nei’s unbiased genetic distance (GD) was almost all cases, samples taken from the same estimated for the five species and the hybrid population grouped together in the dendro- populations (Table 4). The genetic distance gram. However, there was only weak bootstrap between the two outgroups and the other three support for these groupings: no group, be it species was high. Between G. anglica/G. ama- based on population or species category, had rella, G. anglica/G. uliginosa and G. amarella/ more than 50% bootstrap support, and in most G. uliginosa, however, the distance was low cases support was much less than 50%. Species (GD ¼ 0.014, 0.031 and 0.030, respectively). A groups showed some separation on the dendro- dendrogram of this relationship (not shown) gram, but this was not bootstrap supported. placed the hybrid populations between G. ang- Since the G. anglica samples collected from lica and G. amarella, and G. uliginosa slightly the known hybrid populations may be hybrids apart. and so confounding the net separation of the Genetic distance was also calculated for two species, these populations were removed the individual populations of the three species from the data set and the PCO analysis G. anglica, G. amarella and G. uliginosa. re-performed. The first three components, Results are shown for the data set with and accounting for 44.3% of the variability in without the monomorphic bands in order to the data set, were not able to clearly separate allow comparison with other studies (Table the species although G. anglica and G. amarella 5). This relationship is also shown as a were not completely superimposed (Fig. 3). dendrogram (Fig. 4). Although most of the The samples of G. uliginosa grouped with those populations of the species G. anglica and of G. amarella at a position most distant from G. amarella group together, there is some the G. anglica samples. cross-over. The G. anglica samples from the Measures of genetic diversity. The mean population at Elsea Pit group with G. ama- band sharing similarity within species groups rella samples from the geographically closest was high, ranging from 88.8% in G. germanica populations at Therfield Heath and Grange- to 100% in G. uliginosa (Table 3). Mean band lands. Samples of G. amarella from Colonsay sharing similarities between the three species in the Scotish Isles group with G. anglica G. anglica, G. amarella and G. uliginosa were populations. The populations of G. uliginosa also high and only slightly lower than those group with G. amarella. within the species. The mean similarity of the Mantel test. The mantel test was only individuals of the G. anglica, G. amarella, performed on the populations of the three G. uliginosa and the hybrid populations com- species G. anglica, G. amarella, G. uliginosa bined was 94.3%. The outgroups, G. campes- with and without the hybrid populations. tris and G. germanica, shared 55.6% similarity There was no obvious correlation between with each other, and only 44.9% and 48.3% genetic distance of these populations and similarity, respectively, with the individuals of geographic distance. the three other species combined. Relationship between G. anglica and G. The maximum difference between any amarella in the hybrid pops. The most strongly individual sample of G. amarella and G. anglica supported topology for the data from the two was only 9%; that is almost the same level of hybrid populations, Gear Sands – the popula- difference that one finds between any two tion from which the first report of hybrids was individuals of the same taxon. Indeed, the made in 1966 (Margets and David 1981) – and samples of G. anglica from Winklebury Hill in Bindon Hill, regardless of their morphological Wiltshire were more dissimilar, one from and phenological similarity to one or the other another (data not shown), than were any of the two parental species, separate the sample of G. anglica from any sample of individuals on the basis of population not G. amarella. species (Fig. 5). Although there was an appar- 144 M. O. Winfield et al.: Genetic structure of Gentianella spp. in Britain

0.8 a 0.6

0.4

0.2

0.0 -0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 -0.2 2nd. co-ordinate = 15.7% 2nd.

-0.4

-0.6 1st. co-ordinate = 18.0%

G. anglica G. amarella G. uliginosa

0.5 b 0.4

0.3

0.2

0.1

0.0 -0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 -0.1 3rd co-ordinate = 10.6% -0.2

-0.3

-0.4 1st. co-ordinate = 18.0%

G. anglica G. amarella G. uliginosa

Fig. 3. PCO plots showing the relationship of individuals of G. anglica, G. amarella,andG. uliginosa having removed from the data set the individuals from the known hybrid populations – Beachy Head, Bindon Hill, Gear Sands and Stackpole Warren. a) First vs second component, b) first vs third component

Table 3. Mean between and within group band sharing similarities. Hybrid populations are those in which both G. anglica and G. amarella and the hybrid taxon, G. · davidiana, are known to found G. anglica Hybrid pops. G. amarella G. uliginosa G. campestris G. germanica G. anglica 95.3% Hybrid pops. 94.1% 96.2% G. amarella 93.5% 93.2% 94.5% G. uliginosa 94.3% 94.5% 94.6% 100.0% G. campestris 45.3% 44.7% 44.8% 44.5% 93.6% G. germanica 48.6% 48.5% 47.9% 48.0% 55.6% 88.8% M. O. Winfield et al.: Genetic structure of Gentianella spp. in Britain 145

Table 4. Nei’s unbiased genetic distance calculated from the data set containing all bands, both poly- morphic and monomorphic G. anglica Hybrid pops. G. amarella G. uliginosa G. campestris G. germanica G. anglica **** Hybrid pops. 0.0114 **** G. amarella 0.0141 0.0211 **** G. uliginosa 0.0311 0.0293 0.0299 **** G. campestris 0.5599 0.5645 0.5492 0.6086 **** G. germanica 0.4865 0.4913 0.4821 0.5304 0.3159 **** ent separation of the early and late flowering either AFLPs or cpSSRs. Interspecific vari- types, this was only weakly supported. ability between G. amarella and G. anglica, G. amarella and G. uliginosa and G. anglica and G. uliginosa, as expressed by both band Chloroplast SSRs sharing similarity and genetic distance, was The seven chloroplast microsatellite loci were very low. In a study of Rumex acetosa using not particularly informative. Although all the same technique, primer pairs and statistical seven primer pairs successfully amplified a analysis, no individual shared more than 89% single band, they revealed no polymorphisms of bands, and average band sharing between within or between the three taxa G. anglica, samples was 80% (unpublished data). Similar- G. amarella and G. uliginosa. Samples of ly, interspecific similarity in Ranunculus omio- G. campestris had a fragment longer than phyllus and R. tripartitus was 83% and 76% those from the other species by one base pair respectively, while that between the two species at the locus amplified by the primer pair was only 40% (upublished data). In an AFLP ccmp2, while those of G. germanica had a study of the rare species, Pedicularis palustris, fragment one base pair longer at locus ccpm4. Schmidt and Jensen (2000) reported mean There were no other differences. genetic similarities (Dice rather than Jaccard) of 82%, while Cole and Kuchenreuther (2001), using RAPDs, reported within and between Discussion population similarities in species of Aconitum The principal aim of this study was to evaluate of c.90% and c.80%, respectively. In an AFLP the level of genetic diversity in the British study of wild carrot in Denmark, genetic endemic G. anglica, and to determine its distance between populations, (only polymor- distinctiveness from the two closely related phic bands used in the analysis), ranged from species G. amarella and G. uliginosa. A small 0.0721 to 0.1538 (Shim and Jorgensen 2000) number of samples of G. campestris and which is not disimilar to the distances observed G. germanica, members of the the same section between G. anglica and G. amarella (Table 4). of the genus Gentianella, were included in the Chloroplast microsatellites, which, in several study to provide outgroups to the other three studies, have been shown to reveal intraspecific species. Although five species, then, were variability (Provan et al. 2001) also failed to included in the analysis, only three clearly distinguish between the three species under identifiable groups were revealed. Two groups, consideration. In conclusion, the degree of those containing G. campestris and G. germa- genetic variability between the three Gentia- nica, correspond with recognised species nella species was remarkably low, and, in all groupings based on morphological and phe- cases, was less than, or equal to, that between nological characters. The other three species, populations of the two outgroups. Indeed, the however, were not clearly separable using degree of variability of the three species 146 M. O. Winfield et al.: Genetic structure of Gentianella spp. in Britain

considered together was of the same order of that in G. campestris, and was less than the

G. uli. infraspecific variability in G. germanica al- though a smaller number from a very restricted range of these latter were included in the analysis. On the basis of these two marker systems, then, only three unambiguous taxa rather than five can be recognised: G. campes- tris, G. germanica and a G. anglica/G. amarella/ G. uliginosa aggregate. Thus, while one should be cautious of making taxonomic conclusions based solely on the results from molecular markers, it is worthy of note that the three species share levels of similarity that are characteristic of within species variation in other taxa (Cole and Kuchenreuther 2001, Diaz Lifante and Aguinagalde 1996). Results from the hybrid populations Bin-

(2 pops. combined) populations calculated from the data don Hill and Gear Sands are particularly interesting since they indicate that an early and a late gentian from the same site are more closely related than are two early gentians or

G. uliginosa two late gentians from different populations. This suggests that the entities recognised as and G. anglica and G. amarella are not related by evolutionary descent. If this is the case, the two should not be considered separate species, but, rather, early and late flowering forms of the G. amarella

, same taxon. This would give the same taxo- nomic treatment to Gentianella as that given to other grassland species in which early and late

G. anglica flowering forms are recognised but not affor- ded the taxonomic level of species (Zopfi 1995, 1998). It would also be in agreement with the taxonomic treatment of Gentianella in conti- nental Europe (Zopfi 1990, Lid and Lid 1994, Lennartsson 1997), and early British accounts (Townsend 1883). Given the level of taxonomic confusion in this species complex, it is, perhaps, not sur- prising that G. amarella, G. anglica and G. uli- ginosa are genetically so similar. Populations are reported to show great variability in 0.1277 0.1681 0.18600.1561 0.2294 0.2257 0.2225 0.1812 0.2873 0.1234 0.2397 0.2138 0.1802 0.2364 0.2801 0.0918 0.2316 **** 0.1165 0.1046 0.0408 0.0235 0.2273 **** Elsea Pit Ivinghoe Afton Winklebury Prescombe Aucombe Therfield Grangelands Selsley Llanymynech Colonsay morphology, and the characters that are used

Nei’s unbiased genetic distance for the to define the taxa (corolla size, number of internodes, height of main stem, length of terminal internode) are non-discrete and tend nech ColonsayG. uliginosa 0.1816 0.1971 0.1670 0.2202 0.2110 0.1426 0.2247 0.2895 0.1973 0.1917 **** 0.0484 ElseaIvinghoeAfton 0.1407Winklebury **** 0.1951Prescombe **** 0.1840Aucombe 0.0305 0.0274 0.1528Therfield 0.0284 0.1330 0.0384 0.1837 0.0793 0.0956Grangelands 0.0062 0.0401 **** 0.1548 **** 0.1316 0.1171 0.0407Selsley 0.1217 0.2056Llanymy- 0.0369 0.1505 0.1209 0.0174 0.0326 0.0941 0.3144 0.0254 0.2008 0.1722 0.2575 **** 0.1849 0.0248 0.0093 0.0213 0.2748 0.0205 0.0498 0.0321 0.0283 0.3143 0.2678 0.0262 0.2489 0.0112 0.0376 0.3033 **** 0.0422 0.0325 0.0410 0.0505 0.2336 0.0497 0.1794 0.2881 0.0602 0.0382 0.0610 0.0360 0.1448 0.0278 **** 0.0568 0.0463 0.0370 0.0479 **** 0.0599 0.2111 0.0387 0.0381 0.0300 0.0543 0.0586 0.1961 0.0271 0.0417 0.0387 0.0451 0.0478 0.0446 0.2008 0.0444 0.0482 0.0415 0.0343 0.0413 0.0350 0.0439 0.0581 **** 0.0467 0.0311 0.0504 0.0573 0.0203 0.0464 0.0395 0.0481 0.0579 0.0416 0.0259 Table 5. set containing all bandsHead, (above Bindon the Hill, horizontal), Gear and Sands only and polymorphic Stackpole bands Warren (below – the are horizontal). not included The in known this hybrid table populations – Beachy to overlap (Pritchard 1959; Rich et al. 1997; M. O. Winfield et al.: Genetic structure of Gentianella spp. in Britain 147

0.02 0.01 0.00 Sweden the species is considered to be taxo- nomically doubtful (Wigginton 1999). Elsea Pit There are, however, morphological and phenological differences between the taxa Therfield Heath regardless of the fact that apparently there is Grangelands no, or very little, genetic difference. This Selsley apparent incongruity may find explanation in the fact that AFLPs and SSRs essentially Llanymynech provide information for neutral loci: that is, Oxwich coding sequences may be poorly represented Whitford by such markers. Several studies have shown that large differences in morphology can be Ivinghoe governed by small changes at a limited number Winklebury Hill of genes (Bradshaw et al. 1995, Doebley et al. 1997, Andersson 2001). If only a small number Prescombe Down of genes, or indeed, a single gene separate these Aucombe Bottom Genetianella spp. one wouldn’t necessarily Afton Down expect to be able to distinguish the taxa based on the analysis of neutral AFLP and cpSSRs Colonsay markers.

0.12 0.10 0.08 0.06 0.04 0.02 0.00 What might account for the morphological and phenological differences between G. ang- Nei’s genetic distance lica, G. amarella and G. uliginosa? Explana- tions that have been offered are based on Fig. 4. Dendrogram of Nei’s unbiased genetic dis- environmental factors, on germination time tance between the populations of G. anglica, G. ama- and on life cycle (annual vs. biennial). Wilson rella (names of populations in italics )andG. uliginosa (names of populations underscored). The scale above (2000) observed microsite differences between the dendrogram is calculated from the data set G. anglica and G. amarella. G. anglica grows at containing all bands, while the scale below the sites with a level of disturbance consistent with dendrogram is based on polymorphic bands only cattle grazing and where the vegetation is short. G. amarella grows at less disturbed sites Rich and Jermy 1998; Wilson 1999, 2000). with taller vegetation. The morphological Taxa are described on the basis of population characters that distinguish G. uliginosa may means (Clapham et al. 1962, Rich et al. 1997, also be a reflection of plastic responses to Rich and Jermy 1998) where both taxa co-exist environmental factors (Briggs and Walters and hybrids occur. The only clear and reliable 1997); i.e., the phenotype may be a response characteristic for defining the taxa, then, is to growth in a dune habitat. flowering time; G. anglica flowers from March According to Wilson (2000), G. anglica to early July and G. amarella from late July to may be an annual that germinates in spring, early October (Lennartsson 1997, Rich and rapidly grows to maturity and flowers (a life Jermy 1998). The dune gentian, G. uliginosa, cycle consumed in only 14 weeks), while the also flowers in late summer, but is smaller than autumn gentian, G. amarella, is a biennial that G. amarella and is found exclusively in dune germinates in spring and early summer of one slacks. However, morphologically, it is very year and flowers in the autumn of the similar to both G. anglica and G. amarella. following year. Others, however, have reported Ross-Craig (1963) could ‘‘find no difference that both species are biennial (Pritchard 1959). between the British G. anglica and German These differences in phenology might give rise specimens of G. uliginosa’’, and in Norway and to the various morphological differences that 148 M. O. Winfield et al.: Genetic structure of Gentianella spp. in Britain

Gear Sands

56.2

Bindon Hill

40.3

Fig. 5. Dendrogram, based on Jaccard, showing the relationship of the samples taken at three different times of year from the two hybrid populations Gear Sands and Bindon 96 97 98 99 100 Hill. Samples were collected from plants in Percentage similarity flower corresponding with the morphological characteristics of the either G. anglica (May), hybrids (July) or G. amarella (Aug.). Num- bers at the branching points are bootstrap Collection: End May End Aug. Mid July values define the taxa. For instance, Stewart et al. atic (Pritchard 1959); but see Pons 1991, and (1999) report that annual forms of G. germa- Milberg 1994. Nevertheless, that there is a nica occur and that these are, in general, genetic component to flowering time in G. ama- significantly smaller in all their parts than the rella has been shown by the common garden typical biennial form. Other authors have experiments of Lennartsson (1997). Zopfi also reported that variation in plant size (1990), suggested that internode length is the and flowering time are correlated with germi- character selected for, while others have sug- nation time (Lloyd and Webb 1977, Ter Borg gested germination time. However, in the final et al. 1980, Boutin and Harper 1991). Again, analysis, even though the three species have G. anglica could be considered an early flow- acquired some morphological differentiation ering form of G. amarella. and niche separation, given the evidence from It is unclear how much of the morpholog- the increasing number of reports of hybrids, ical variation detected in natural populations they have not acquired reproductive isolation. of the investigated Gentianella species is envi- According to Mayr’s biological concept ronmental in origin and how much is due to of species (Mayr 1988), then, G. anglica and genetic differences. Cultivation from seed un- G. uliginosa should not be considered a full der controlled conditions has proved problem- species, but rather ecotypes of G. amarella. M. O. Winfield et al.: Genetic structure of Gentianella spp. in Britain 149

It may be that in Britain, and probably in hybridisation between G. anglica and G. ama- other parts of Europe as well (Zopfi 1990), rella is threatening the integrity of the former there has been selection for alleles at a small taxon, be it a species, subspecies or ecotype. It number of loci (or even a single locus), as a may well be that mixed populations in which result of the traditional management of many hybrids and intermediate types have been calcareous grasslands. The typical practise recorded are those populations in which the of mowing or grazing in July would have former management practices are less in evi- removed all forms with intermediate flowering dence so that there are no longer any artificial times and, therefore, driven selection for the barriers preventing occasional hybridisation two extremes, early and late gentian. Once between the two forms. The resulting hybrids, established, the temporally isolated, seasonal then, increase the opportunities for interspecific ecotypes would have constituted two stable gene flow. Nevertheless, the removal of former sub-populations with very little, if any, poten- management practices has not taken place tial for genetic contact. The continuation of sufficiently long ago to allow complete elimina- the same management scheme over many tion of genetic differences between the extreme centuries would have ensured that the two types. The reintroduction of former manage- taxa were reproductively isolated and, conse- ment regimes would probably ensure the main- quently, able to accumulate a number of tenance of G. anglica as a distinct entity. morphological characteristics controlled by one, or a small number of genes. However, in The authors wish to thank, in particular, Ruth British populations of G. anglica and G. ama- Davis for her enthusiasm and encouragement. We rella, there has been very little time for this would also like to thank Dr. David Briggs and drift to occur. Species of Gentianella can only Prof. Peter Grubb for helpful discussions and have arrived in Britain during the last 10,000 advise. The work was supported in full by funding years (Marren 1999). Thereafter, human influ- from PlantLife (registered charity number: ence was of fundamental importance: the 1059559). clearing of forest and the grazing of domesti- cated animals would have produced the open, References short grassland that most Gentianella species Andersson S. (2001) The genetic basis of floral require. Grazing of domestic animals, partic- variation in Senecio jacobaea (Asteraceae). ularly sheep, may have been one of the J. Hered. 95(5): 409–414. selective pressures that resulted in the separa- Bobbink R. (1998) Impacts of tropospheric ozone tion into early and late flowering forms and airbourne nitrogenous pollutants on natural (Theaker and Briggs 1993, Lennartsson and semi-natural ecosystems. New Phytol. 139: 1997). However, the two species have not 161–168. evolved any form of isolation barrier in the Boutin C., Harper J. L. (1991) A comparative study brief time that they have been separated by the of the population dynamics of five species of influence of man. Changes in land manage- Veronica in natural habitats. J. Ecol. 79: 199–221. ment have effectively removed the barrier to Bradshaw H. D., Wilbert S. M., Otto K. G., hybridisation. Less grazing, therefore removal Schemske D. W. (1995) Genetic mapping of of the selective force, has allowed flowering floral traits associated with reproductive isola- Mimulus times of the two species to overlap. Given that tion in Monkeyflower ( ). Nature 376: 762–765. there are no obvious barriers to hybridisation, Briggs D., Walters S. M. (1997) Plant Variation a brief evolutionary excursion has come to an and Evolution, 3rd edn. Cambridge University end: no profound changes in the gene pools of Press, Cambridge. the putative species have taken place. Clapham A. R., Tutin T. G., Warburg E. F. (1962) In terms of species conservation, results Flora of the British Isles, 2nd edn. Cambridge from the mixed populations indicate that University Press, Cambridge. 150 M. O. Winfield et al.: Genetic structure of Gentianella spp. in Britain

Cole C. T., Kuchenreuther M. A. (2001) Molecular Payne R. W., Lane P. W., Digby P. G. N., et al. markers reveal little genetic differentiation (1993) GenStat 5 Reference Manual, Release 3. among Aconitum novboracense and A. columbia- Oxford University Press. num (Ranunculaceae) populations. Amer. J. Bot. Pons T. L. (1991) Dormancy, germination and 88(2): 337–347. mortality of seeds in a chalk grassland flora. Darby A. (2000) Farming with flowers. Plantlife J. Ecol. 79: 765–780. spring 2000: 9. Pritchard N. M. (1959) Gentianella in Britain. Davis R., Duckworth J. (2000) Back from the brink Watsonia 4: 169–192. 2000: the current status of Britain’s threatened Provan J., Powell W., Hollingsworth P. M. (2001) wild plants within Plantlife’s species recovery Chloroplast microsatellites: new tools for studies programme. Plant Life Report, pp. 9–10, 2000. in plant ecology and evolution. Trends Ecol. Diaz Lifante Z., Aguinagalde I. (1996) The use of Evol. 16(3): 142–147. random amplified polymorphic DNA (RADP) Rich T. C. G., Holyoak D. T., Margetts L. J., markers for the study of taxonomical relation- Murphy R. J. (1997) Hybridisation between ships among species of Asphodelus sect. Verinea Gentianella amarella (L.) Boerner and G. anglica (Asphodelaceae). Amer. J. Bot. 83: 949–953. (Pugsley) E. F. Warb. (Gentianaceae). Watsonia Doebley J., Stec A., Hubbard L. (1997) The 21: 313–325. evolution of apical dominance in maize. Nature Rich T. C. G., Jermy A. C. (1998) Plant Crib. 386: 485–488. Botanical Society of the British Isles. London. Felsenstein J. (1985) Confidence limits on phylog- pp. 228–230. enies: an approach using the bootstrap. Evolu- Ross-Craig (1963) Drawings of British Plants: parts tion 9: 366–369. XIX–XXIII, Lobeliaceae to Verbenaceae. G. Bell Gulliver R. L. (1998) Population sizes of Genetia- and Sons, London. nella uliginosa (Willd.) Bo¨rner, Dune Gentian, Schmidt K., Jensen K. (2000) Genetic structure and on Colonsay (V.C. 102) in 1996. Watsonia 22: AFLP variation of remnant populations in the 111–113. rare plant Pedicularis palustris (Scrophularia- Holyoak D. T. (1999) Gentianella uliginosa (Willd.) ceae) and its relation to population size and Bo¨rner (Gentianaceae) rediscovered in north reproductive components. Amer. J. Bot. 87(5): Devon. Watsonia 22: 428–429. 678–689. Lennartsson T. (1997) Seasonal differentiation - a Shim S. I., Jorgensen R. B. (2000) Genetic structure conservative reproductive barrier in two grass- in cultivated and wild carrots (Daucus carota L.) land Gentianella (Gentianaceae) species. Plant revealed by AFLP analysis. Theor. Appl. Genet. Syst. Evol. 208: 45–69. 101: 227–233. Lid J., Lid D. (1994) Norsk Flora, 6th edn. Det Ter Borg S. J., Janse A., Kwak M. M. (1980) Life Norske Samlaget, Oslo. cycle variation in Pedicularis palustris L. (Scroph- Lloyd D. G., Webb C. J. (1977) Secondary sex ulariaceae). Acta Bot. Neerl. 29: 397–405. characters in plants. Bot. Review 43: 177–216. Theaker A. J., Briggs D. (1993) Genecological Margetts L. J., David R. W. (1981) A Review of the studies of groundsel (Senecio vulgaris L.). IV. Cornish Flora 1980. Institute of Cornish Studies, Rate of development in plants from different Redruth. habitat types. New Phytol. 123: 185–194. Marren P. (1999) Britain’s rare flowers. T & AD Townsend F. (1883) Flora of Hampshire. Lovell, Poyser, London. Reeve and Co., London. Mayr E. (1988) Towards a new philosophy of Tutin T. G., Heywood V. H., Burges N. A., Moore biology: Observations of an evolutionist. Bel- D. M., Valentine D. H., Walters S. M., Webb D. knap press, Camb. Mass. U.S.A. A. (1972) Flora Europaea. Volume 3. Diapensi- Milberg P. (1994) Germination ecology of the aceae to Myoporaceae. Cambridge, Cambridge endangered biennial Gentianella campestris. Biol. University Press. Cons. 70: 287–290. Weising K., Gardner R. C. (1998) A set of Nei M. (1978) Estimation of average heterozygosity conserved PCR primers for the simple sequence and genetic distance from a small number of repeat polymorphisms in chloroplast genomes of individuals. Genetics 89: 583–590. dicotyledonous angiosperms. Genome 42: 9–19. M. O. Winfield et al.: Genetic structure of Gentianella spp. in Britain 151

Wiggington M. J. (1999) British Red Data Books: 1 (Scrophulariaceae). Plant Syst. Evol. 198: 209– Vascular Plants. Joint Nature Conservation 233. Committee, Peterborough. Zopfi H. J. (1998) Life-history variation among Wilson P. J. (1999) The distribution and status of populations of Euphrasia rostkoviana Hayne Gentianella anglica (Pugsley) E. Warb. English (Scrophulariaceae) in relation to grassland man- Nature and PlantLife Report, no. 119, 1999. agement. Biol. J. Linn. Soc. 64: 179–205. Wilson P. J. (2000) Early gentian, Gentianella anglica (Pugsley) E. Warb. Survey and monitor- ing work in 1999. English Nature and PlantLife Report, no. 147, 2000. Addresses of the authors: Dr. Mark O. Winfield Winfield M. O., Arnold G. M., Cooper F., Le Ray (e-mail: Mark.Winfi[email protected]), Via M., White J., Karp A., Edwards K. (1998) A Volta 12, I-20052 Monza (MI) Italy. J. S. Parker, study of genetic diversity in Populus nigra subsp. Cory Laboratory, Department of Plant Sciences, betulifolia in the Upper Severn Area of the UK University of Cambridge, 47 Bateman Street, using AFLP markers. Molec. Ecol. 7: 3–10. Cambridge CB2 1JF, UK. P. J. Wilson, Wessex Zopfi H. J. (1990) Aestival and autumnal vicariads Environmental Associates, 4 Prospect Place, Grove of Gentianella (Gentianaceae): a myth? Plant Lane, Redlynch, Salisbury SP5 2NT, UK. M. Syst. Evol. 174: 139–158. Labra, Dipartimento di Biologia, Sezione di Zopfi H. J. (1995) Life history variation and Botanica Generale, Universita` di Milano, Via infraspecific heterochrony in Rhinanthus glacialis Celoria 26, I-20133 Milano, Italy.