The origin of a novel form of Senecio (Asteraceae) restricted to sand dunes in southern Sicily

Authors

  • Mark A. Chapman,

    Corresponding author
    1. Department of Biological Sciences, Vanderbilt University, VU Station B 351634, Nashville, TN 37235-1634 USA;
    2. Sir Harold Mitchell Building, School of Biology, University of St Andrews, St Andrews, Fife KY16 9TH, UK
    Search for more papers by this author
  • Richard J. Abbott

    1. Sir Harold Mitchell Building, School of Biology, University of St Andrews, St Andrews, Fife KY16 9TH, UK
    Search for more papers by this author

Author for correspondence: Mark A. Chapman Tel: +1 615 9363893 Fax:+1 615 3436707 Email: mark.a.chapman@vanderbilt.edu

Summary

  • • The taxonomy of diploid Mediterranean Senecio sect. Senecio (Asteraceae) is complex, owing to a recent species radiation, high morphological plasticity and occasional interspecific hybridization.
  • • A study was conducted to resolve the origin of a novel form of Senecio restricted to sand dunes in southern Sicily, Italy. This has been described previously as morphologically intermediate to Senecio gallicus and Senecio glaucus ssp. coronopifolius, indicating a possible hybrid origin, or as a variant of Senecio leucanthemifolius.
  • • Plants of this form grown in a glasshouse were morphologically intermediate to S. glaucus and S. leucanthemifolius, but were also similar to some cultivated individuals of S. gallicus. No evidence for a hybrid origin was obtained from a survey of random amplified polymorphic DNA variation; instead the plants surveyed were most closely allied to Tunisian S. glaucus. They were also polymorphic for the same set of cpDNA haplotypes present in Tunisian S. glaucus.
  • • We conclude that the Sicilian Senecio is a variant form of North African S. glaucus ssp. coronopifolius, which most probably dispersed to sand dunes in southern Sicily in the relatively recent past. The presence of several cpDNA haplotypes in this material indicates that there have been multiple introductions of the species to Sicily.

Introduction

In rapidly evolving species groups, taxonomic boundaries can be blurred because (1) there are few diagnostic characters to distinguish taxa, (2) high levels of phenotypic plasticity make species identification difficult and (3) occasional interspecific hybridization and gene flow may erode species differences in certain areas. Thus, within such complexes, some populations may be difficult to assign with certainty to particular taxa. Under these circumstances, taxonomic resolution might be aided by comparing material grown in a common environment and/or by examining material at the molecular level. This approach can also resolve the origins of taxonomically ambiguous material.

Mediterranean Senecio sect. Senecio (Asteraceae) is taxonomically complex and species limits are often poorly defined (Alexander, 1979; Abbott et al., 1995; Coleman, 2002) especially among the diploids. This is partly due to most diploids being products of a recent radiation in the group (Comes & Abbott, 2001), and also to high levels of phenotypic plasticity and species interfertility (Alexander, 1979; Abbott et al., 1995). Examples of recent hybridization and gene flow have been documented in the section (Abbott, 1992; Abbott & Lowe, 1996; Comes & Abbott, 1999; James, 1999) and have resulted in the origin of the diploid hybrid species Senecio squalidus (Abbott et al. 2000, 2002), several allopolyploid species (Ashton & Abbott, 1992; Lowe & Abbott, 1996, 2000; Coleman et al., 2003; Chapman, 2004) and an introgressant form of Senecio vulgaris (Abbott et al., 1992).

The purpose of this investigation was to determine the taxonomic status and origin of a type of Senecio found on sand dunes in the region of Licata and Pozallo in southern Sicily, Italy. In the most recent revision of Mediterranean Senecio section Senecio, Alexander (1979) described the plant as intermediate in morphology to Seneciogallicus Vill. and Senecio glaucus L. ssp. coronopifolius (Maire) Alexander, indicating a possible hybrid origin. A study of genetic variation in the widespread S. glaucus (Coleman, 2002) showed that the plant is closely allied to Tunisian material of S. glaucus ssp. coronopifolius, although no other species were examined and therefore a hybrid origin cannot be ruled out. To add to the confusion, Comes & Abbott (2001) referred to the material as Senecio leucanthemifolius Poiret, based on its morphological affinity to this species in Sicily. Thus it is feasible that Licata/Pozallo material is a variant form either of S. gallicus, S. glaucus ssp. coronopifolius or S. leucanthemifolius. Alternatively it could be a hybrid between two or all three of these species, each of which is diploid (2n = 20) and widespread throughout parts of the Mediterranean region (Alexander, 1979). In Sicily, S. leucanthemifolius is common (Alexander, 1979, and personal observations), whereas S. gallicus (although listed from Sicily by Alexander, 1979) has not been found there by one of us (R. J. A.) despite several searches. Alexander (1979) also listed a specimen of S. glaucus ssp. coronopifolius from Licata, Sicily, but referred to this material as intermediate to S. gallicus and S. glaucus ssp. coronopifolius in his revision. Hybrids between S. gallicus and S. glaucus ssp. coronopifolius showing 35% stainable pollen have been produced artificially, whereas crosses between S. gallicus and S. leucanthemifolius have not been successful (Alexander, 1979). Crosses between Moroccan material that was intermediate to S. glaucus ssp. coronopifolius and S. leucanthemifolius var. fradinii, and each of its putative parent species yielded hybrids with high pollen fertility (Alexander, 1979).

The aim of the present study was to determine if Senecio material that occurs on sand dunes near Licata and Pozallo in southern Sicily is a hybrid derivative or a variant form of S. gallicus, S. glaucus ssp. coronopifolius or S. leucanthemifolius. To this end, a detailed morphometric analysis was conducted on cultivated material from Licata and Pozallo, and material of the three potential parent species. In addition, material was surveyed for random amplified polymorphic DNA (RAPD) and chloroplast (cp) DNA variation. Previous studies by Comes & Abbott (1998, 1999, 2001) have analysed cpDNA variation within and among Mediterranean Senecio sect. Senecio providing background for further comparative studies of cpDNA variation in the section.

Materials and Methods

Plant material

Seeds were collected from approx. 30 different plants from each of the two populations of uncertain taxonomic status that occur on sand dunes near Licata and Pozallo in southern Sicily. Both populations were well established, containing over 200 plants. In addition, seed was obtained from a similar number of plants from each of several populations of S. gallicus, S. glaucus ssp. coronopifolius and S. leucanthemifolius var. leucanthemifolius that occur in the Mediterranean region (Table 1; Fig. 1). Following germination, seedlings were transferred to 9-cm diameter pots containing 3 : 1 Levington's M2 compost: gravel and raised in a glasshouse with day/night temperatures of 20°C/12°C and a 16-h daylength supplemented by 400 W metal halide lamps. Plants were arranged in a randomized block design.

Table 1. Senecio material examined. n refers to number of individuals surveyed in the morphometric (morph) and random amplified polymorphic DNA (RAPD) analyses
SpeciesCodePopulation locationLatitudeLongitudeCollector1 and daten (morph)n (RAPD)
  • 1

    Collections were made by R. J. Abbott (RJA), M. Coleman (MC), H. P. Comes (HPC), C. Petít (CP) and C. Thébaud (CT). NA, not analysed.

UnknownLILicata, Sicily37°07′ N13°57′ EHPC and RJA (1995)610
UnknownPOPozallo, Sicily36°44′ N14°15′ EHPC and RJA (1995)6 9
S. glaucus ssp. coronopifoliusENEnfida Plage, Tunisia36°03′ N10°27′ EMC and RJA (2000)4 6
S. glaucus ssp. coronopifoliusEJEl Jem, Tunisia35°15′ N10°15′ EMC and RJA (2000)5 5
S. glaucus ssp. coronopifoliusSBSbeitla, Tunisia35°13′ N09°30′ EMC and RJA (2000)5 7
S. glaucus ssp. coronopifoliusHSHoumt-Souk, Tunisia33°55′ N10°52′ EMC and RJA (2000)7 8
S. l. var. leucanthemifoliusMAMakthar, Tunisia35°50′ N09°12′ EMC and RJA (2000)NA 5
S. l. var. leucanthemifoliusTATabarka, Tunisia36°55′ N08°45′ EMC and RJA (2000)7 6
S. l. var. leucanthemifoliusBIBizerte, Tunisia37°18′ N09°52′ EMC and RJA (2000)5 8
S. l. var. leucanthemifoliusHAEl Haouaria, Tunisia37°03′ N11°00′ EMC and RJA (2000)5 8
S. l. var. leucanthemifoliusHMHammamet, Tunisia36°25′ N10°40′ EMC and RJA (2000)5 5
S. l. var. leucanthemifoliusSLSan Leone, Sicily37°21′ N13°35′ EHPC and RJA (1996)5 6
S. l. var. leucanthemifoliusCTCastellammare, Sicily38°02′ N12°52′ EHPC and RJA (1996)NA 5
S. l. var. leucanthemifoliusGTGioia Tauro, Italy38°26′ N15°55′ EHPC and RJA (1996)4 6
S. l. var. leucanthemifoliusSETorre San Emiliano, Italy40°06′ N18°20′ EHPC and RJA (1996)NA 6
S. l. var. leucanthemifoliusGALGalipoli, Italy40°03′ N17°59′ EHPC and RJA (1996)3 6
S. l. var. leucanthemifoliusTPTour de la Parata, Corsica41°54′ N08°36′ ECT (2001)5 6
S. l. var. leucanthemifoliusPROPropriano, Corsica41°40′ N08°54′ ECT (2001)7 6
S. gallicusMVVMont Ventoux, France44°09′ N05°16′ ECP (1997)6 6
S. gallicusVNMVilla Nova, Portugal37°43′ N08°47′ WHPC and RJA (1995)6 3
S. gallicusSRSa Riera, Spain41°57 N03°12′ ERJA (1997)6 6
Figure 1.

Sampling locations for Senecio material analysed in this study. Population codes are given in Table 1.

Morphometric analysis

Three to seven plants per population were measured for 13 morphological traits at the stage of full anthesis of the apical capitulum. Seven characters were descriptors of the apical inflorescence (i.e. inflorescence length, capitulum length and width, mean ray floret length and width, number of ray florets and number of calyculus bracts) and six were descriptors of the midleaf (i.e. length, area, perimeter, standardized area (square root of leaf area divided by leaf length), leaf dissection (leaf perimeter divided by square root of leaf area) and standardized leaf perimeter (leaf perimeter divided by leaf length)).

DNA extraction

DNA extraction was conducted using a slightly modified form of the cetyltrimethylammonium bromide (CTAB) method of Doyle and Doyle (1990) (for details see Chapman, 2004). Large-scale extractions (approx. 2 g of leaf material) were conducted for cpDNA restriction fragment length polymorphism (RFLP) analysis, while small-scale extractions (approx. 200 mg leaf material) were used for RAPD analysis.

RAPD analysis

Initially, seven DNA samples (two from each of S. gallicus, S. glaucus and S. leucanthemifolius, and one from a Licata plant) were amplified using 60 random decamer oligonucleotides (Operon Technologies sets A, B and C). Each 25 µl polymerase chain reaction (PCR) contained 2.5 µl 10× reaction buffer (supplied with enzyme; 160 mm (NH4)2SO4; 670 mm Tris-HCl (pH 8.8) and 0.1% Tween-20), 2.5 mm MgCl2, 0.2 mm of each dNTP, 0.2 µm primer, 1 unit of BioTaq DNA polymerase (BioLine, London, UK) and 5 ng genomic DNA. The thermal cycling profile began with a denaturation at 94°C for 3 min, followed by 44 cycles of 30 s at 94°C, 45 s at 35°C and 90 s at 72°C, and a final elongation step of 4 min at 72°C. Amplification products were resolved in 1.2% agarose gels containing 0.33 µg ml−1 ethidium bromide. Gels were run for c. 3 h at 100 V and visualized using ultraviolet (UV) transillumination.

From these 60 primers, 17 were chosen based on the production of clear banding patterns, and a repeat PCR conducted to ensure consistent amplification. Bands that amplified in only one of the repeated PCR reactions were not scored. Bands that amplified in more than one species were tested for homology as follows. First the bands were excised from gels and purified using a gel extraction kit (Qiagen, Crawley, UK). A repeat PCR was then conducted on the gel-extracted DNA (as described earlier) and digested with three 4 bp-cutter restriction enzymes following the manufacturer's protocol (Promega, Southampton, UK). Digested PCR products were separated on 8% polyacrylamide gels and stained with ethidium bromide. Products showing identical restriction patterns for at least two of the enzymes were considered homologous between taxa (Rieseberg, 1996). Following this screen, eight primers were chosen for the full analysis of 133 samples, based on their ability to produce reproducible bands that were easy to score and were homologous between species (Table 2). Of the 40 fragments examined, nine were found in only one or other of the three parent species and were assumed to be homologous between individuals within the particular species. Of the remaining 31 bands scored, 19 were successfully tested for homology. Ideally, all bands shared between taxa should have been tested; however, for the remaining 12 bands, the product of interest was of a similar size to another amplified by the same primer and could not be excised cleanly from the gel.

Table 2.  Details of random amplified polymorphic DNA (RAPD) primers used in the analysis
PrimerPrimer sequence (5′−3′)Number of bands scoresBand size (bp)
A04AATCGGGCTG5400, 500, 800, 1050, 1500
A10GTGATCGCAG6200, 300, 350, 450, 600, 800
A19CAAACGTCGG5300, 400, 500, 600, 650
B04GGACTGGAGT5300, 500, 550, 600, 700
B07GGTGACGCAG4350, 400, 500, 1000
C15GACGGATCAG6200, 300, 400, 500, 600, 750
C16CACACTCCAG6200, 300, 550, 600, 750, 1150
C17TTCCCCCCAG3550, 750, 800

Because all 133 DNA samples could not be amplified in one PCR plate, four control DNAs and two negative controls (no DNA) were included in each plate to ensure that repeatable banding patterns were produced by different PCR amplifications with the same primer. Bands were scored as present (1) or absent (0) and recorded in a binary data matrix for statistical analysis. Fragment names were specified to reflect the primer used and approximate size of the amplification product (e.g. A41050 refers to a band produced by primer A4 that was approximately 1050 bp long).

cpDNA RFLP

A previous study of cpDNA variation in Mediterranean species of Senecio identified 16 different cpDNA haplotypes based on 27 restriction fragment length polymorphisms (Comes & Abbott, 2001). Samples in the present study were analysed for seven haplotypes (A, B, C, F, G, H and J) recorded previously by Comes & Abbott (1998, 1999, 2001) in S. gallicus, S. glaucus and S. leucanthemifolius. This involved the use of four enzyme–probe combinations (for details see Comes & Abbott, 2001).

Fifty-eight individuals were analysed for cpDNA RFLP haplotype. This included 12 individuals from Licata, 12 from Pozallo, 14 Tunisian S. leucanthemifolius individuals (from five different populations) and 20 individuals of Tunisian S. glaucus ssp. coronopifolius (from four different populations). Restriction digestion, agarose gel electrophoresis and Southern blotting followed standard protocols with minor modification (Chapman, 2004).

Statistical analysis

Morphometric analysis  For the purpose of morphometric analysis the widespread species S. leucanthemifolius was subdivided into three groups on the basis of geographic origin. Hence, six groups were compared: (1) Licata/Pozallo Senecio, (2) S. gallicus, (3) S. glaucus ssp. coronopifolius, (4) S. leucanthemifolius (Italy/Sicily), (5) S. leucanthemifolius (Corsica) and (6) S. leucanthemifolius (Tunisia). Two-way analysis of variance (anova) was carried out on each measured trait using the GLM option of sas 8.2 (SAS Institute, Cary, NC, USA) with ‘Groups’ and ‘Populations within groups’ as main effects. Tukey tests were employed to determine the significance of differences between trait means of each pair of groups. Standardized midleaf area was not normally distributed and was transformed into natural logarithms before analysis. A logarithmic transformation did not normalize the distributions of ray floret and calyculus bract number, and these characters were therefore not subjected to anova. Instead, mean number of calyculus bracts was calculated for each of the six groups to allow intergroup comparison. Ray floret numbers clustered around 8 and 13 in the Fibonacci series −8, 13, 21, etc., and therefore ray number for a given individual that was not a Fibonacci number was assigned to the closest Fibonacci number before making comparisons between frequencies in each group. Before principal component analysis (PCA) using minitab (version 13.31, 2000; Minitab Inc., Coventry, UK), untransformed data were standardized by subtracting the trait mean from each value and dividing by the trait standard deviation.

RAPD analysis Lynch & Milligan (1994) recommend that bands absent in < 3/n individuals (where n = sample size) are discarded from genetic analysis to prevent biased estimates of genetic parameters. All bands were present in > 3/n individuals (n = 133) and hence were subjected to analysis. The complete data matrix was analysed to estimate the percentage of polymorphic loci per population (i.e. the number of polymorphic bands in the population divided by the total number of bands scored) and the Shannon–Weaver index (Shannon & Weaver, 1949) of within-population phenotypic diversity (I). The index is calculated using the formula I = –Σpi log2pi, where pi is the frequency of the ith band (King & Schaal, 1989).

Pairwise genetic distances were estimated as Euclidean distance (Excoffier et al., 1992), defined for RAPDs by Huff et al. (1993). Principal coordinate analysis (PCO) was conducted on the distance matrix using genalex (Peakall & Smouse, 2002). An assignment test was conducted following the method of Paetkau et al. (1995) using assignment calculator (Brzustowski, 2002). This was conducted to assign Licata and Pozallo individuals to the remaining 19 populations examined in the RAPD study. Analysis of molecular variance (amova; Excoffier et al., 1992) was performed using arlequin 2.001 (Schneider et al., 2000) on the same pairwise squared Euclidean distance matrix. Initially, amova was conducted on five predefined groups based on clusters observed in the PCO analysis (i.e. (1) Licata/Pozallo; (2) S. gallicus; (3) S. glaucus; (4) S. leucanthemifolius (Tunisia/Italy/Sicily); and (5) S. leucanthemifolius (Corsica)). Subsequently, amovas were carried out after combining Licata and Pozallo populations with S. glaucus, and also for comparing Licata/Pozallo material with S. glaucus. For the latter amova, two RAPD bands (A10200 and C16750) were recorded as absent in < 3/n individuals (n = 45 individuals) and hence were excluded from analysis. Pairwise ΦST values (analogous to Fisher's FST) were calculated, first to measure differentiation between populations and then between groups. Significance of differences were calculated based on permutation procedures (10 000 replicates).

cpDNA haplotype analysis  The cpDNA haplotypes detected in the present study were combined for analysis with appropriate data recorded previously by Comes & Abbott (1998, 1999, 2001).

Results

Morphometric variation

There were significant differences between the means of the six Senecio groups examined for all 11 characters analysed (Table 3). Significant variation was also observed among populations within taxa, except for capitulum length and mean ray floret length. Based on a comparison of means, Licata and Pozallo material most closely resembles S. leucanthemifolius in leaf morphology, but is broadly similar to each of the species examined (S. gallicus, S. glaucus ssp. coronopifolius and S. leucanthemifolius var. leucanthemifolius) in capitulum type. For only one character (mean ray floret length) was Licata/Pozallo material significantly different from these three species. For this character it was intermediate to S. glaucus on one hand and S. gallicus and S. leucanthemifolius on the other. Senecio glaucus and S. gallicus produced the least number of calyculus bracts (2.95 ± 0.39 (SE) and 1.17 ± 0.38 per individual, respectively), whereas Italian/Sicilian and Tunisian S. leucanthemifolius had the greatest means for this trait (14.50 ± 1.67 and 14.77 ± 1.15, respectively). Corsican S. leucanthemifolius produced, on average, 5.75 ± 0.39 calyculus bracts per capitulum, which was similar to the mean number produced by Licata/Pozallo material (5.17 ± 0.58). The number of ray florets per capitulum clustered around the value 13 for 11 of 12 Licata/Pozallo individuals and for most individuals of S. gallicus, S. glaucus and Corsican S. leucanthemifolius examined. However, for over half of Italian and Tunisian S. leucanthemifolius plants surveyed (seven of 12, and 15 of 22, respectively) number of ray florets clustered around the figure eight.

Table 3.  Mean (± SE) of 13 morphological characters measured on Senecio material
CharacterLicata/Pozallo n = 12S. gallicusn = 18S. glaucusn = 21S. leucanthemifolius
Italy n = 12Corsica n = 12Tunisia n = 22Among populations within taxaP-value Among taxa
  • Senecio leucanthemifolius was divided into three groups based on geographic origin. n is the number of individuals sampled per group. Shared letters in superscript implies a nonsignificant (P > 0.05) difference according to Tukey's test.

  • 1

    First value is number of individuals with 8-ray florets, second value is number of individuals with 13-ray florets.

Inflorescence length (mm)28.37 ± 3.26a,b 33.92 ± 1.58b50.69 ± 3.86c21.19 ± 1.39a,b60.74 ± 7.20c17.09 ± 0.78a  0.023  0.011
Capitulum length (mm) 9.95 ± 0.31d,e  9.34 ± 0.28c,d10.52 ± 0.14e 8.32 ± 0.29a,b 7.62 ± 0.16a 8.67 ± 0.21b,c  0.189  0.021
Capitulum Base Width (mm) 5.22 ± 0.17b,c  3.90 ± 0.14a 5.39 ± 0.10c 4.19 ± 0.14a 4.89 ± 0.16b 4.04 ± 0.11a  0.027  0.001
Mean ray floret length (mm)11.54 ± 0.32b  8.46 ± 0.43a13.13 ± 0.42c 7.06 ± 0.30a 7.23 ± 0.43a 8.06 ± 0.29a  0.117  0.001
Mean ray floret width (mm) 3.35 ± 0.18b  2.25 ± 0.08a 2.82 ± 0.08a,b 2.58 ± 0.08a,b 2.54 ± 0.11a,b 2.55 ± 0.09a,b  0.024  0.049
Midleaf length (mm)57.38 ± 2.83a 84.27 ± 6.51b74.40 ± 1.89b71.01 ± 3.99a,b55.78 ± 2.67a61.75 ± 2.94a< 0.001  0.004
Midleaf area (mm2)575.6 ± 28.6b,c1439.0 ± 266.0e556.9 ± 50.0a,b,c798.9 ± 32.9b,c,d331.0 ± 28.3a,b889.4 ± 88.9b,c,d< 0.001< 0.001
Midleaf perimeter (mm)277.3 ± 12.6a 684.4 ± 52.2c434.6 ± 26.7b193.3 ± 14.7a165.3 ± 9.0a202.0 ± 18.3a  0.013< 0.001
Standardized midleaf area 0.43 ± 0.02b  0.43 ± 0.02b 0.31 ± 0.01a 0.41 ± 0.02b 0.32 ± 0.01a 0.49 ± 0.01c< 0.001< 0.001
Midleaf dissection11.72 ± 0.63b 21.05 ± 1.98d18.55 ± 0.74c 6.83 ± 0.45a 9.40 ± 0.65a,b 7.06 ± 0.56a< 0.001< 0.001
Standardized leaf perimeter 4.99 ± 0.37b  8.56 ± 0.69c 5.85 ± 0.33b 2.70 ± 0.09a 3.00 ± 0.17a 3.34 ± 0.21a< 0.001  0.001
Mean number calyculus bracts 5.17 ± 0.58  1.17 ± 0.38 2.95 ± 0.3914.50 ± 1.67 5.75 ± 0.3914.77 ± 1.15  
Number of ray florets1 1, 11  3, 15 3, 18 7, 5 1, 1115, 7  

The first two axes of the PCA plot (Fig. 2) accounted for 34.0% and 24.1%, respectively, of the total variation. Principal component axis 1 separated S. glaucus and S. gallicus from S. leucanthemifolius, while PCA2 separated S. gallicus from S. glaucus. Midleaf perimeter, midleaf dissection and standardized leaf perimeter contributed most to the first principal component with eigenvector values of −0.439, −0.466 and −0.415, respectively. Characters that contributed most to the second principal component included capitulum base width (eigenvector value 0.450), mean ray floret width (0.361) and midleaf area (0.433). No additional structuring was observed when PCA3 was plotted (not shown). Licata/Pozallo material was intermediate in morphology to S. glaucus and S. leucanthemifolius, showing some overlap with certain individuals of S. glaucus, Corsican S. leucanthemifolius and some individuals of S. gallicus. Corsican S. leucanthemifolius tended to be morphologically distinct from Tunisian and Italian representatives of the species.

Figure 2.

Principal components analysis (PCA) plot of the first two axes based on 11 morphological variables in Senecio glaucus (squares), Senecio leucanthemifolius (multiplication signs, Tunisia; open circles, Corsica; plus signs, Italy/Sicily), Senecio gallicus (closed circles) and the populations of taxonomic uncertainty (triangles).

RAPD variation

The RAPD diversity, measured in terms of percentage polymorphic loci and the Shannon–Weaver diversity index, tended to be lower in Licata/Pozallo material than in Tunisian S. glaucus, but was similar to that recorded in the majority of populations of S. leucanthemifolius from Italy and Tunisia, and greater than that resolved in Corsican S. leucanthemifolius and S. gallicus populations (Table 4).

Table 4.  Percentage polymorphic loci (% PL) and Shannon's index of diversity (I) calculated among 21 populations of Senecio
TaxonPopulation1% PLI
‘Licata/Pozallo’LI32.50.180
PO35.00.161
S. glaucusEN42.50.244
EJ52.50.297
SB35.00.188
HS37.50.201
S. leucanthemifolius (Tunisia)MA30.00.180
TA32.50.183
BI35.00.184
HA45.00.220
HM25.00.147
S. leucanthemifolius (Italy/Sicily)SL35.00.193
CT32.50.184
GT30.00.166
SE45.00.238
GAL37.50.212
S. leucanthemifolius (Corsica)TP15.00.098
PRO15.00.097
S. gallicusMVV20.00.110
SR20.00.110
VNM 7.50.047

Seven of the 40 RAPD fragments surveyed in S. gallicus, S. glaucus and S. leucanthemifolius were either private (i.e. found in only one taxon at a frequency of 0.1–0.4) or specific (i.e. had a frequency of > 0.4 in one taxon only; Table 5). Both markers specific to S. glaucus were fixed in Licata/Pozallo material, while one of two private bands in S. glaucus occurred at low frequency among Licata/Pozallo plants. By contrast, neither the band specific to S. gallicus nor the two private bands in S. leucanthemifolius occurred in Licata/Pozallo material. Licata/Pozallo material contained no unique bands.

Table 5.  Frequencies of seven random amplified polymorphic DNA (RAPD) bands that were private (P) or taxon-specific (TS) to either Senecio gallicus, Senecio leucanthemifolius or Senecio glaucus.
RAPD product1CategoryS. gallicusS. glaucusS. leucLicata/ Pozallo
  • 1

    Band sizes are given as subscripts.

A41050TS0.6670.0000.0000.000
A10200TS0.0000.9230.0001.000
A10300P0.0000.2690.0000.000
A19400P0.0000.1920.0000.158
C15750TS0.0000.6920.0001.000
A19300P0.0000.0000.2050.000
C15300P0.0000.0000.2880.000

A plot of the first two principal coordinates (Fig. 3) showed that PCO1 separated S. glaucus and Licata/Pozallo material from S. gallicus and S. leucanthemifolius, and S. gallicus from Corsican S. leucanthemifolius. Genetic substructuring within S. leucanthemifolius was evident, with individuals from Italy/Sicily and Tunisia separated from Corsican material along PCO2. Both coordinates, in combination, separated most individuals of Tunisian from Italian/Sicilian S. leucanthemifolius. Licata/Pozallo material formed a distinct cluster that was most similar to Tunisian S. glaucus.

Figure 3.

Principal coordinates analysis (PCO) plot based on 40 RAPD products in Senecio glaucus (squares), Senecio leucanthemifolius (multiplication signs, Tunisia; open circles, Corsica; plus signs, Italy/Sicily), Senecio gallicus (closed circles) and the populations of taxonomic uncertainty (triangles).

The assignment test conducted on Licata/Pozallo individuals assigned 17 of 19 individuals from this source to the inland population of S. glaucus at Sbeitla (SB), Tunisia, and the two remaining individuals to other S. glaucus populations.

All pairwise ΦST values between populations were significant except for those between the BI and HA, and BI and HM populations of Tunisian S. leucanthemifolius (Table 6). These particular population pairs were separated by approximately 100 km. A comparison of ΦST values between groups shows that Licata/Pozallo material was more similar to S. glaucus ssp. coronopifoliusST = 0.183) than to S. gallicusST = 0.557) or S. leucanthemifoliusST = 0.401–0.565).

Table 6.  Interpopulation ΦST values (below diagonal) calculated among 21 populations of Senecio
Taxon PopulationUnknownS. glaucusS. leucanthemifolius
TunisiaSicily/ItalyCorsicaS. gallicus
LIPOENEJSBHSMATABIHAHMSLCTGTSEGALTPPROMVVSRVNM
  1. All ΦST values were significant (P < 0.05) except those underlined. Values above diagonal are among-group ΦST values.

LI0.000  0.183    0.468    0.401  0.565  0.557 
PO0.1870.000                   
EN0.3590.3860.000     0.232    0.296  0.472  0.399 
EJ0.3360.4170.1970.000                 
SB0.1900.2730.2390.2220.000                
HS0.2740.3470.1410.2250.2620.000               
MA0.5710.5890.2610.3790.4700.3740.000      0.225  0.468  0.428 
TA0.5420.5940.3180.3700.5060.4050.1770.000             
BI0.5130.5470.2460.3400.4060.3030.2410.1940.000            
HA0.5260.5480.2390.3530.4220.3450.2460.1870.0150.000           
HM0.5090.5460.2470.2700.3720.3150.1910.2680.0260.1780.000          
SL0.5770.6060.4930.4860.5310.5570.4240.3800.4380.4490.4220.000    0.343  0.365 
CT0.4630.5250.3700.4250.4810.3890.4580.3960.3950.4130.3610.3470.000        
GT0.5560.6110.4340.4480.5380.4790.4450.4850.4040.4420.4330.3510.4590.000       
SE0.6140.6730.4900.4200.6100.5670.5210.4060.4990.4290.4990.4870.4580.5210.000      
GAL0.5650.6220.4520.3950.5390.5490.4940.4210.4630.4450.4360.4030.4090.4250.2330.000     
TP0.7020.7310.6270.6390.7070.6960.6850.6510.6480.6230.7060.6940.7260.6360.6270.5390.000  0.498 
PRO0.6880.7040.6060.6120.6880.6640.5660.6510.6580.6540.6660.6360.6940.6070.6710.6440.6980.000   
MVV0.6420.6840.5040.5710.6270.5420.5890.6210.5520.5480.6120.6760.6630.5440.6920.6320.7140.7450.000  
SR0.6350.6780.5550.5460.6270.5570.6280.6360.5690.5770.6120.5750.6130.5470.6670.6020.7820.7280.4560.000 
VNM0.6020.6630.4650.4570.6020.4870.5600.5010.4430.4530.5800.5980.6120.5220.5830.5250.7090.7550.4950.5970.000

amova conducted on groups corresponding to the main clusters observed in the PCO plot (Fig. 3) (i.e. Licata/Pozallo material, S. gallicus, S. glaucus, Corsican S. leucanthemifolius and other S. leucanthemifolius; Table 7) showed that there was significant variation between groups (28.7% of total), among populations within groups (26.2%) and within populations (45.1%). A similar partitioning of variation was evident when Licata/Pozallo material was grouped with S. glaucus in the amova. When Licata/Pozallo material was compared only with S. glaucus, the variation between groups (16.0%) was not quite significant (P = 0.068).

Table 7. amova results
 Source of variationdfSum of squaresVariance components% VariationP1
  • (1) amova including five groups (1, Licata/Pozallo; 2, Senecio glaucus; 3, Senecio gallicus; 4, Senecio leucanthemifolius (Tunisia and Italy/Sicily); 5, S. leucanthemifolius (Corsica)). (ii) amova including four groups, as (1) except Licata/Pozallo and S. glaucus combined. (3) amova including two groups, Licata/Pozallo and S. glaucus

  • 1

    Level of significance based on 10 000 permutations.

(1)Among groups  4243.4341.9386628.70< 0.001
Among populations within groups 16224.1541.7698226.20< 0.001
Within populations112341.3213.0475145.11< 0.001
Total132808.9106.75599  
(2)Among groups  3213.2121.9362028.19< 0.001
Among populations within groups 17254.3761.8855427.45< 0.001
Within populations112341.3213.0475144.36< 0.001
Total132808.1906.86925  
(3)Among groups  1 30.2220.8613216.04  0.068
Among populations within groups  4 40.9980.9360916.80< 0.001
Within populations 39137.0463.5140167.15< 0.001
Total 44208.2675.31142  

Chloroplast DNA variation

The results of previous surveys of cpDNA variation in S. gallicus, S. leucanthemifolius and S. glaucus conducted by Comes & Abbott (1998, 1999, 2001) showed that: (1) S. gallicus contained five haplotypes (A, B, F, G, H) with haplotypes B and G being the most common (Comes & Abbott, 1998, 2001); (2) Italian and Sicilian S. leucanthemifolius contained three haplotypes (A, C, F) with haplotype A being most frequent (Comes & Abbott, 2001); and (3) S. glaucus ssp. coronopifolius from Israel and Egypt contained five haplotypes (A, B, C, F, J) with haplotype C being most common (Comes & Abbott, 1999) (Table 8). The present survey established that Licata/Pozallo material contains haplotypes A, B, C, F, and J, all of which are found in Tunisian S. glaucus (Table 8). It also showed that haplotypes B and C were absent from the Tunisian S. leucanthemifolius examined.

Table 8.  Chloroplast DNA haplotypes present in senecio species examined
 nChloroplast DNA haplotypes
ABCFGHJ
  1. (1) Number of each haplotype recorded by Comes & Abbott (1999, 2001) in S. gallicus, S. glaucus and S. leucanthemifolius; (2) number of each haplotype recorded in Licata/Pozallo material and Tunisian S. glaucus and S. leucanthemifolius. See Table 1 for codes.

(1)
S. gallicus (France/Spain/Portugal)1151256 06383 0
S. glaucus (Israel/Egypt) 86 2 8723 00 1
S. leucanthemifolius
 (mainland Italy) 1010 0 00 00 0
 (Sicily)  4 2 0 11 00 0
(2)
Licata/Pozallo Senecio (Sicily) 24 2 7 31 0011
S. glaucus (Tunisia)
 EN  3 2 00 00 0
 EJ  4 2 00 00 0
 SB  2 0 10 00 2
 HS  2 0 01 02 0
 Total 2011 3 11 02 2
S. leucanthemifolius (Tunisia)
 MA  1 0 00 00 1
 TA  1 0 01 02 0
 BI  3 0 00 00 0
 HA  2 0 01 00 0
 HM  2 0 00 00 0
 Total 14 9 0 02 02 1

Discussion

Morphometric analysis demonstrated that Licata/Pozallo Senecio material is morphologically intermediate to S. glaucus ssp. coronopifolius and S. leucanthemifolius var. leucanthemifolius. The material raised in a glasshouse showed some morphological overlap with certain individuals of S. glaucus, Corsican S. leucanthemifolius and some individuals of S. gallicus. Thus, the material is not easy to assign to an established taxon, even when compared with related material raised under similar conditions. It is not surprising, therefore, that there has been taxonomic confusion in the past with regard to these plants (Alexander, 1979; Comes & Abbott, 2001; Coleman, 2002). A further finding to emerge from the morphometric analysis was that Corsican material of S. leucanthemifolius var. leucanthemifolius is morphologically highly divergent from Tunisian and Italian/Sicilian material and might therefore be regarded as a distinctive island form of the species.

Although Licata/Pozallo material is not easily assigned to a particular taxon based on morphology, a survey of RAPD variation showed it to be most similar to Tunisian S. glaucus. Licata/Pozallo material clustered closely with individuals of Tunisian S. glaucus in a principal coordinate analysis, and was fixed for two RAPD bands that were specific to Tunisian S. glaucus and contained one of two private alleles recorded in the species. By contrast, it did not contain the one band identified as specific to S. gallicus or two private bands present in S. leucanthemifolius. No RAPD bands were recorded that were unique to Licata/Pozallo material indicating that this material is possibly of recent origin and has not had sufficient time to accumulate new RAPD alleles. Assignment tests indicated that all 19 Licata/Pozallo plants could be assigned to populations of Tunisian S. glaucus. Overall, the RAPD results strongly indicate that Licata/Pozallo material is a variant form of S. glaucus derived from material introduced to Sicily from North Africa. The possibility that the material is a relict of what was previously a widespread variant in Sicily (and possibly elsewhere in the northern Mediterranean) seems unlikely given its lack of unique RAPD bands.

The results of the survey of cpDNA haplotype variation were somewhat surprising in that they showed that Licata/Pozallo material was highly polymorphic, containing haplotypes A, B, C, F and J. Haplotypes A, F and J were recorded in at least two of the putative parent species; however, only S. glaucus possessed all five haplotypes present among Licata/Pozallo plants. Haplotype B was absent from S. leucanthemifolius, and C and J were absent from S. gallicus (and rare in S. leucanthemifolius). This finding provides a further pointer that Licata/Pozallo material is derived from S. glaucus ssp. coronopifolius, introduced to Sicily from North Africa.

The high level of cpDNA variation present among Licata/Pozallo plants contrasts with the lower levels of RAPD diversity recorded in the same material relative to that in Tunisian populations of S. glaucus. The presence of five cpDNA haplotypes in Licata/Pozallo material indicates that there have been multiple introductions of the species to Sicily, while reduced levels of RAPD diversity might suggest that a bottleneck occurred during or following colonization. It is unlikely that the cpDNA diversity has arisen in situ in Sicily, because the maximum number of mutations between cpDNA haplotypes in Licata/Pozallo material (i.e. five) is only two less than the maximum number of seven mutations that differentiate the seven cpDNA haplotypes resolved within and among S. glaucus, S. leucanthemifolius and S. gallicus (Comes & Abbott, 2001). The possible bottleneck effect indicated by the RAPD results is unlikely to have been severe, given that RAPD diversity in Licata/Pozallo material is approximately equivalent to that recorded in S. leucanthemifolius populations from Tunisia and Italy, and considerably greater than that found in the two populations of S. leucanthemifolius surveyed from Corsica. It might be of interest to conduct additional surveys of nuclear genetic variation in the future using other marker systems before firm conclusions are reached about the levels of nuclear genetic diversity in Licata/Pozallo material.

Conclusions

We conclude that Licata/Pozallo material is a variant form of Senecio glaucus ssp. coronopifolius. We also suggest that this variant, which for certain morphological characters is not easily distinguished from either S. gallicus or S. leucanthemifolius, was most probably introduced to Sicily from North Africa in the relatively recent past. It is not possible to rule out that the material is long-standing in Sicily and a relict of a wider distribution of S. glaucus in the northern part of the Mediterranean basin. However, the species is currently not known from elsewhere in this part of the Mediterranean and an absence of unique RAPD alleles argues against the possibility of it being long-standing in Sicily and isolated from the distribution of S. glaucus in North Africa. The likelihood of a relatively recent introduction, possibly as a consequence of human activity, is given further weight by the fact that the material is restricted to sand dunes next to the southern ports of Licata and Pozallo. The material has not been reported from elsewhere in Sicily or in mainland Italy; however, a systematic survey of the southern coast of Sicily is required to clarify the distribution of the material in this region. It is feasible that since its introduction the material has been subject to selection for certain characters, which might be the cause of its current intermediate morphology to S. glaucus, S. leucanthemifolius and S. gallicus. Despite the dunes of southern Sicily and Tunisia appearing similar, it is interesting to note that Tunisian S. glaucus is restricted to southern latitudes both on the coast and inland. In northern Tunisia, the species is replaced by S. leucanthemifolius at both coastal and inland locations. It is feasible, therefore, that the environment in the Licata/Pozallo area is different from that under which Tunisian material of S. glaucus normally grows and might therefore have imposed selection for a novel form of the species adapted to local conditions. In the future, it will be necessary to determine whether Licata/Pozallo material is reproductively isolated from Tunisian S. glaucus and, if so, whether it might be regarded as a new species.

In addition to describing Licata/Pozallo material as intermediate to S. glaucus and S. gallicus, Alexander (1979) described material from elsewhere in the Mediterranean as intermediate in morphology to other pairs of Senecio taxa. For example, he described material intermediate to S. glaucus and S. leucanthemifolius from Libya, and to S. gallicus and S. leucanthemifolius from Gibraltar. Whether such material is of hybrid origin or a variant form of one particular species remains to be established. Recently, Coleman & Abbott (2003) examined the possibility that atypical material of S. leucanthemifolius var. casablancae from the southern part of its range in Morocco originated following hybridization with the more southerly distributed S. glaucus ssp. coronopifolius. Although some evidence of past hybridization was provided by cpDNA markers, there was no evidence that nuclear DNA introgression had occurred and, consequently, it was concluded that the atypical morphology of the southern material was most probably the result of divergence caused by selection and/or drift. In groups such as Mediterranean Senecio section Senecio which have undergone a rapid radiation in recent times, it is of interest to determine whether and how often reticulation has occurred in the wild, and whether such events result occasionally in the origin of new hybrid taxa. Clearly, the results of the present study emphasize the point that morphological intermediacy is only an indicator of hybridity. Molecular analysis is required to prove that material is hybrid.

Acknowledgements

We thank Max Coleman for assistance in the field and laboratory, Peter Comes for advice on cpDNA analysis, Tom Meagher for statistical advice, and Christophe Thébaud and Christophe Petít for supplying seed. The research was conducted during the tenure of a NERC research studentship to M. A. C.

Ancillary