Occasional hybridization between a native and invasive Senecio species in Australia is unlikely to contribute to invasive success

Study species

Senecio madagascariensis is a diploid plant, initially introduced to south-eastern Australia from the KwaZulu-Natal province of South Africa in the early part of the 20th century (Radford et al., 2000). Molecular analysis of contemporary and historical field collections has pointed to at least two separate introductions (Dormontt et al., 2014). Senecio pinnatifolius (previously S. lautus) is a tetraploid plant native to Australia. There have been multiple taxonomic treatments of the species complex (Ali, 1969; Radford, 1997; Roda et al., 2013; Thompson, 2005), with each agreeing on distinction of ‘dune’, ‘headland’ and ‘tableland ecotypes. Senecio pinnatifolius var. pinnatifolius (dune ecotype) occurs on coastal sands along the east coast of Australia and is the only S. pinnatifolius ecotype analysed in the current study, hereafter referred to only as Senecio pinnatifolius. Initially included in the S. pinnatifolius complex, S. madagascariensis was recognised as a separate species after Hilliard’s (1977) treatment of Asteraceae in Natal (Sindel et al., 1998). This separation has been supported by morphological comparisons (Thompson, 2005) and cytological studies (Radford, Liu & Michael, 1995) finding 2n = 20 for S. madagascariensis and 2n = 40 for S. pinnatifolius.

Both species are considered annuals or short lived perennials (Radford & Cousens, 2000), look superficially identical, growing to approximately 0.6 m and with bright green leaves and yellow inflorescences that are heterogamous and radiate. The species can be reliably distinguished by the number of involucral bracts present, 18–21 in S. madagascariensis, 11–14 in S. pinnatifolius. Both species are outcrossing, self-incompatible (Ali, 1966) and insect pollinated, predominantly by the introduced European honey bee Apis mellifera and various species of Syrphidae (White, 2008); seeds are wind dispersed. In comparisons of life history traits between the species, S. madagascariensis was found to perform better than S. pinnatifolius with respect to seedling, growth and fecundity measures but S pinnatifolius maintained a stronger soil seed bank (Radford & Cousens, 2000).

Sample collection and seed germination

One allopatric population of each species was sampled along with two sites where the two species occurred together (Table 1). At each site, twenty individuals of each species were sampled with fresh leaf material preserved in silica beads for DNA extraction. Additionally, at shared sites, multiple mature seed heads were collected from sampled plants where available and stored for later germination. Plants were sampled in a systematic fashion (across the contact zone at shared sites) and the location of all plants recorded with GPS (with the exception of the allopatric S. madagascariensis population where coordinates were not recorded). Plants that were not sampled for DNA were identified in the field and GPS coordinates recorded. The GPS recorded relative position to within 1 m accuracy. Two years after initial sampling, we revisited one of the shared sites to survey the changes in abundance of S. pinnatifolius and S. madagascariensis.

In the laboratory, mature achenes were detached from their pappus and the seed coat nicked with a scalpel. Seeds were grown on moist filter paper with gibberellic acid (GA₃) in a 12 h photoperiod at 25 °C to stimulate germination. All germinated seedlings, up to a maximum of ten per parent plant, were frozen at −80 °C prior to DNA extraction, except in one case where 20 seedlings were used (Table 1, see results for further explanation). The number of seeds per parent plant that successfully germinated were classified as ‘ten or more’, or ‘less than ten’. Germination rates were compared between sites using a permutation approach in Resampling Stats Add-In for Excel v4.0 (https://www.statistics.com/). Seeds per parent plant that successfully germinated (‘ten or more’, or ‘less than ten’) were sampled without replacement to simulate the same number of parent plants per site as the empirical data. The proportion of ‘less than ten’ seedlings per site were compared in 10,000 simulations to the empirical data to obtain estimated P values.

Genetic analysis

DNA extractions were carried out using the Machery-Nagel Nucleospin Plant II Kit with the PL2/PL3 buffer system. Two published microsatellite loci (Le Roux & Wieczorek, 2007) originally developed for S. madagascariensis and found to be cross compatible with S. pinnatifolius, were used to screen all adults and seedlings from both species (Table 2). PCR reactions were prepared with ∼20 ng of template DNA, 1× reaction buffer, 0.2 mM of each dNTP, 2.5 mM MgCl₂, 0.4 µM of each primer, and 0.02 U Amplitaq Gold^® (Applied Biosystems, Foster City, CA, USA) to give a final PCR reaction volume of 10 µL. Reactions involved an initial denaturation step of 94  °C for 2 min, followed by 35 cycles at 94 °C for 1 min, the loci specific annealing temperature for 1 min (Table 2), 72 °C for 1 min and 30 s, and a final extension at 72 °C for 30 min. One published chloroplast microsatellite locus (Weising & Gardner, 1999) (Table 2) was found to produce bands mutually exclusive to S. pinnatifolius and S. madagascariensis and so was included to allow identification of the maternal parent of any hybrid adults detected in the field. Reactions were prepared with ∼20 ng of template DNA, 1x reaction buffer, 0.2 mM of each dNTP, 2.5 mM MgCl₂, 0.5 µM of each primer, and 1 U IMMOLASE™ DNA polymerase (Bioline, London, UK) to give a final PCR reaction volume of 10 µL. Reactions involved an initial denaturation step of 94 °C for 5 min, 30 cycles of 94 °C for 20 s, 50 °C for 20 s, 72 °C for 20 seconds, and a final extension at 72 °C for 30 min. Products were separated using the ABI 3730 DNA analyzer (Applied Biosystems, Foster City, CA, USA) with the GeneScan™—500 LIZ^® size standard. Genemapper^® Software v4.0 (Applied Biosystems, Foster City, CA, USA) was used to score fragments. Scoring was recorded in a binary matrix with presence or absence of particular alleles indicated by a 1 or 0. This method allowed for polyploidy and diploid data to be directly compared and analysed together. DNA from thirty one individuals (9% of samples) were amplified twice for microsatellite analysis to enable estimation of error rates, calculated according to DeWoody, Nason & Hipkins (2006).

Amplified fragment length polymorphisms (AFLPs) were assessed according to the method of Vos et al. (1995) with modifications. Restriction digests were performed in 20 µl reactions with ∼200 ng of DNA, 1× restriction digest buffer 2, 10 U MseI (New England Biolabs, Ipswich, MA, USA), 10 U EcoRI (New England Biolabs, Ipswich, MA, USA), and 1×BSA. Reactions were incubated for 3 h at 37 °C, followed by 20 min at 65 °C to denature the enzymes. Adapters were ligated to the digested fragments in reactions containing 20 µl of digested DNA, 1×T4 ligase buffer, 2.5 µM EcoRI adapter, 0.25 µM MseI adapter and 3 U of T4 DNA ligase (New England Biolabs, Ipswich, MA, USA). Reactions were incubated overnight at 16 °C.

Pre-selective amplifications contained 2 µl of digested and ligated DNA, 1×Optimised DyNAzyme™ EXT buffer (including 1.5 mM Mg²⁺), 0.2 mM of each dNTP, 0.5 μM MseI (+C), 0.5 μM EcoRI (+A) primers and 0.25 U DyNAzyme™ EXT DNA polymerase to give a final PCR reaction volume of 25 µL. Reactions involved an initial denaturation step of 75 °C for 2 min, then 20 cycles of 94 °C for 30 s, 56  °C for 30 s, 75 °C for 2 min, and a final extension at 60 °C for 30 min. PCR products were run on agarose gel to check for successful amplification.

Selective amplifications contained 1 µl of 1 in 30 diluted pre-selective PCR product, 1×TaqGold buffer (Applied Biosystems, Foster City, CA, USA), 2 mM MgCl₂, 0.2 mM of each dNTP, 0.3 µM MseI + 3bp primers, 0.3 µM EcoRI +3 bp primers and 0.75 U TaqGold (Applied Biosystems, Foster City, CA, USA) in a final PCR reaction volume of 15 µL. Reactions involved an initial denaturation step of 94 °C for 2 min, then 10 cycles of 94 °C for 30 s, 65−56 °C for 30 s (reduce by 1  °C per cycle), 72 °C for 2 min, then 26 cycles of 94 °C for 30 s, 56 °C for 30 s, 72 °C for 2 min and a final extension at 60°C for 5 min. Twelve selective amplifications were trialled using a range of +3 bp primer combinations on four individuals of each species. Products were run on 5% acrylamide gels using a Gelscan GS2000 (Corbet Research) and the three most suitable combinations (based on appropriate number and strength of bands, polymorphisms and ease of scoring) were chosen for selective amplification of all samples (Table 2). Products were separated using the ABI 3730 DNA analyzer (Applied Biosystems) with the GeneScan™—500 LIZ^® size standard. Forty one adult individuals (12% of total individuals) were re-extracted for DNA and the AFLP process repeated to allow loci validation and error rate calculations. Vegetative material from seedlings was too small to allow for repeated extractions, so only adults were used. A negative control was included throughout the extraction/AFLP process to enable exclusion of non-specific bands.

Genemapper^® Software v4.0 (Applied Biosystems, Foster City, CA, USA) was used to manually allocate bins to appropriate loci, all duplicated samples were visualised and where consistent banding was apparent between samples, this was assigned as a specific locus. Once manual binning was complete, the full dataset was automatically scored using Genemapper^® Software v4.0 (Applied Biosystems, Foster City, CA, USA) and raw peak height data obtained. The raw peak height data were then used with AFLPScore v1.4 (Whitlock et al., 2008) to minimise error whilst maximising number of retained loci. AFLPScore allows the user to select a range of loci selection thresholds (the average intensity of bands at a specific locus, above which a locus is retained in the dataset) and phenotype calling thresholds (the intensity of a given band, either in absolute terms, or as a percentage of the average for that locus, above which band presence will be called). By comparing combinations of different locus selection and phenotype calling thresholds, the user can select thresholds which result in reduced error and maximised retained loci. After error reduction via AFLPscore, a phenotype matrix was exported and loci with the highest error rates systematically removed to create 11 separate datasets with error rates of 0%, 1%, 2%, 3% etc. up to 10% and additionally one with a 17% error rate (the output from AFLPscore with no loci removed). To assess the effects of each error rate on overall information content, the data from the allopatric populations of each species were analysed using the program STRUCTURE (Pritchard, Stephens & Donnelly, 2000) with RECESSIVEALLELES set to 1 to account for dominant data (Falush, Stephens & Pritchard, 2007). Number of predefined populations (K) was set from 1 to 5. Each run consisted of a burn-in period of 100,000 Markov Chain Monte Carlo (MCMC) repetitions, followed by 1,000,000 MCMC repetitions, the program was run five times to allow averaging of results in CLUMPP (Jakobsson & Rosenberg, 2007). Plots were displayed in DISTRUCT (Rosenberg, 2004). The final dataset was chosen based on how well it could detect the expected structure (designation of K = 2, highest probability of individuals belonging to the appropriate species cluster) and how robust it was to the negative impacts of higher error (such as the signal from plate effects) see Zhang & Hare (2012) for an in-depth discussion and analysis of this approach.

Data analysis

To assess hybridisation, the AFLP and microsatellite data were combined into one data matrix, in the case of the microsatellites, each allele was either designated as present or absent. Assignment of an individual as either a pure parental species or a hybrid was based on a consensus between two different analysis methods, with the most conservative (i.e., non-hybrid) designation accepted if results were inconsistent between methods. The first method used the allocation procedure in the program AFLPOP (Duchesne & Bernatchez, 2002). The allopatric populations of each species were set as sources, and the remaining samples allocated to either one of the pure species or hybrid origin by the program. Zero frequencies were corrected as 1∕n + 1, where n is the sample size. The allocation minimal log-likelihood difference (MLD) was initially set to 1 (meaning allocation only occurred when designation was 10 times more likely than any other possible origin). Samples that could not be allocated in this way were re-run with MLD set to 0 (allocating to highest likelihood source regardless of the magnitude of difference between alternate likelihoods).

The second method used the program STRUCTURE (Pritchard, Stephens & Donnelly, 2000) with extensions implemented by Falush, Stephens & Pritchard (2007) to account for genotypic ambiguity that is inherent in dominant markers; RECESSIVEALLELES was set to 1. STRUCTURE has been used successfully to assess datasets comprised of individuals with different ploidy levels (De Hert et al., 2012; Pinheiro et al., 2010; Zalapa et al., 2011). Number of predefined populations (K) was set to 2. Each run consisted of a burn-in period of 100,000 Markov Chain Monte Carlo (MCMC) repetitions, followed by 1,000,000 MCMC repetitions, the program was run five times to allow averaging of results in CLUMPP (Jakobsson & Rosenberg, 2007). Plots were displayed in DISTRUCT (Rosenberg, 2004). Clustering of adult and seedling genotypes of each species at both allopatric and shared sites were visualised with a principal coordinate analysis (PCoA) in GENALEX v6.4 (Peakall & Smouse, 2006; Peakall & Smouse, 2012). Hybrid zone mapping was completed using ArcGIS v9.2 (ESRI, Redlands, CA, USA).

Loci selection

Both nuclear microsatellite loci were polymorphic in both species and retained for further analysis (Table 2). The single chloroplast microsatellite locus was polymorphic in S. pinnatifolius (two alleles) and monomorphic in S. madagascariensis but alleles were not shared between species. All adults and seedlings genotyped conformed to their expected species specific chloroplast haplotypes. The observed error rate per allele and per locus for the nuclear microsatellites was zero. Of the 12 AFLP primer combinations trialled, three were chosen for screening all samples (Table 2).

In AFLPScore v1.4 (Whitlock et al., 2008), mismatch error rates were used to optimise scoring parameters using both absolute and relative phenotype calling thresholds on an initial dataset containing 247 loci. The error rate of the exported data set was 0.17 with 233 retained loci (Data S1), achieved by filtering data using an absolute phenotype-calling threshold of 250 relative fluorescence units (RFU), prior to application of a 50 RFU locus-selection threshold. After STRUCTURE analysis, the data set equating to an average error rate of 6% was chosen, as it correctly identified K = 2, indicated high assignment rates of individuals to their correct species, did not display any significant plate effects at K = 3 (the number of plates) and contained a reasonable number of loci (142) (Fig. S1). An overall error rate of 6% is high compared to the 2–5% reported for most AFLP studies (Bonin et al., 2004) but under the maximum threshold of 10% recommended by Bonin, Ehrich & Manel (2007) (Fig. S2). Systematically evaluating the effects of different error rates on result and selecting that which is most informative and least confounding allows the information content of the dataset to be maximised without limiting the included loci in order to conform to an arbitrary cut off point (Zhang & Hare, 2012).

Hybridisation

No adult hybrids were detected in the field. In total, 17 hybrids were observed from 223 seeds (8% of seeds and 5% of all individuals genotyped including adults). Fourteen of these hybrid seeds were from a single S. madagascariensis mother and three from two S. pinnatifolius mothers (Fig. 1). Hybrid seed set was observed at Lennox Head where 6% and 22% of the total seeds sampled for each species at that site were hybrid for S. pinnatifolius and S. madagascariensis respectively. For each adult with a hybrid seed set, the distance to the nearest congeneric was <15 m (Fig. 1). There was uncertainty in the field about seeds collected from what appeared to be a single plant but may have been two adjacent plants. Twenty seeds were germinated from this sample with the hope that separation of individuals could be made in the lab from the results of the genetic analysis. The microsatellite data confirmed that these seeds did indeed come from the same individual, and so one adult has 20 genotyped offspring instead of the usual 10. A single hybrid seed was detected at Ballina from a S. madagascariensis mother (Fig. 1), as designated by agreement between AFLPop and STRUCTURE, however this individual does closely cluster with other pure S. madagascariensis seedlings in the PCoA analysis (Fig. 2), which may indicate a false positive result. AFLPOP (Duchesne & Bernatchez, 2002) allocated 90% of adults and 68% of seedlings with a minimal log-likelihood difference (MLD) of 1 (indicating that the allocation was at least 10 times more likely than any other). The remaining samples allocated with MLD set to 0. One S. madagascariensis seedling was allocated to S. pinnatifolius with MLD set to 0. The chloroplast haplotype of this individual was consistent with S. madagascariensis maternity and it clustered with the hybrid seedlings in the PCoA (Fig. 2), so has been designated as a hybrid. Hybrid origin was more conservatively allocated in the program STRUCTURE (Pritchard, Stephens & Donnelly, 2000) (Fig. 3) with 92% consensus between the two methods. Final designation used the most conservative (non-hybrid) allocation.

Germination and site revisit

Germination success varied somewhat between individuals with 11–38% of parent plants per site producing less than ten seedlings for DNA extraction. However, the simulation approach used to examine germination rates found no significant differences between germination rates at each site and those expected to arise randomly. The Lennox Head site was revisited in 2009, two years after initial sampling and a morphological survey of plant species identity undertaken. All plants observed were identified as S. madagascariensis.