Annals of Botany 113: 861–872, 2014 doi:10.1093/aob/mct314, available online at www.aob.oxfordjournals.org

Maintenance of strong morphological differentiation despite ongoing natural hybridization between sympatric species of ()

Emma J. McIntosh1,2,*, Maurizio Rossetto1, Peter H. Weston1 and Glenda M. Wardle2 1The Royal Botanic Gardens and Domain Trust, Sydney, New South Wales, Australia and 2School of Biological Sciences, the University of Sydney, Sydney, New South Wales, Australia * For correspondence. E-mail [email protected]

Received: 21 July 2013 Returned for revision: 29 October 2013 Accepted: 13 December 2013 Published electronically: 31 January 2014

† Background and Aims When species cohesion is maintained despite ongoing natural hybridization, many questions

are raised about the evolutionary processes operating in the species complex. This study examined the extensive Downloaded from natural hybridization between the Australian native shrubs Lomatia myricoides and L. silaifolia (Proteaceae). These species exhibit striking differences in morphology and ecological preferences, exceeding those found in most studies of hybridization to date. † Methods Nuclear microsatellite markers (nSSRs), genotyping methods and morphometric analyses were used to uncover patterns of hybridization and the role of gene flow in morphological differentiation between sympatric

species. http://aob.oxfordjournals.org/ † Key Results The complexity of hybridization patterns differed markedly between sites, however, signals of intro- gression were present at all sites. One site provided evidence of a large hybrid swarm and the likely presence of mul- tiple hybrid generations and backcrosses, another site a handful of early generational hybrids and a third site only traces of admixture from a past hybridization event. The presence of cryptic hybrids and a pattern of morphological bimodality amongst hybrids often disguised the extent of underlying genetic admixture. † Conclusions Distinct parental habitats and phenotypes are expected to form barriers that contribute to the rapid re- version of hybrid populations to their parental character state, due to limited opportunities for hybrid/intermediate advantage. Furthermore, strong genomic filters may facilitate continued gene flow between species without the danger of assimilation. Stochastic fire events facilitate temporal phenological isolation between species and may at Griffith University on March 29, 2014 partly explain the bi-directional and site-specific patterns of hybridization observed. Furthermore, the findings suggest that F1 hybrids are rare, and backcrosses may occur rapidly following these initial hybridization events.

Key words: Hybridization, Lomatia myricoides, , Proteaceae, microsatellite, morphometrics, admixture, hybrid zone, introgression, species cohesion, Australia, fire.

INTRODUCTION Evidence of extensive hybridization has been detected in the genus Lomatia (Weston and Crisp, 1994; Milner et al., 2012), The maintenance of species cohesion despite ongoing gene ex- making it an ideal system for exploring the role of natural hybrid- change via natural hybridization in is a phenomenon attract- ization in speciation and adaptation across the genus. This study ing increasing research attention (e.g. Moccia et al.,2007;Sambatti focused on hybridization patterns between Lomatia myricoides et al.,2012). Natural hybridization can create bridges for gene flow, (C.F. Gaertn.) Domin and L. silaifolia 21 (Sm.) R.Br. offering a platform for adaptive evolution by introducing variation (Proteaceae), two Australian native shrub species that are and novel traits into populations, potentially resulting in introgres- highly differentiated in morphology and habitat preferences sion and admixture of genotypes (Rieseberg and Wendel, 1993; (Pellow et al., 2009). Although closely related, they are not Ellstrand et al., 1996; Arnold, 1997; Broyles, 2002). sister species, and a geographical pattern of shared chlorotypes With advances in genetic technologies, researchers have been provides further evidence for hybridization (Milner et al.,2012). able to uncover greater complexity within hybrid populations Lomatia myricoides (river lomatia) is a tall, woody shrub (to (Lai et al., 2012; Twyford and Ennos, 2012) and we can now 7 m) with simple leaves up to 20 cm long (dentate in seedlings delve deeper into how interspecific gene exchange can be and young adults, becoming entire in large adults) and lateral ongoing despite the presence of strong reproductive barriers inflorescences (Weston and Crisp, 1994; Wilson et al., 1995). (Sambatti et al., 2012), and pinpoint evidence of adaptive intro- In contrast, Lomatia silaifolia (crinkle bush) is a shrub to 2 m gression (Heliconius Genome Consortium, 2012). There is sub- in height, with highly dissected leaves [mostly pinnately stantial evidence that evolution can occur at the level of loci divided twice or three times (Fig. 1)] and terminal inflorescences (Wu, 2001; Nosil and Schluter, 2011) and that the hybrid form (Wilson et al., 1995; Pellow et al., 2009). The species have wide is often a transitional phase in a much larger dynamic exchange distributions, broadly overlapping at the boundaries of their of genetic material between parental lines, via backcrossing and ranges in the Sydney region (Wilson et al., 1995), where their introgression (Salvini et al., 2008; Gompert et al., 2010). flowering times overlap in summer (Beadle et al., 1972). # The Author 2014. Published by Oxford University Press on behalf of the Annals of Botany Company. All rights reserved. For Permissions, please email: [email protected] 862 McIntosh et al. — Species cohesion maintained despite ongoing hybridization in Lomatia

Brisbane

Sydney Sydney Canberra

N Downloaded from

Melbourne

0 200 400 km 0 20 40 km 10 cm

F IG. 1. Leaf outlines and corresponding distributions of Lomatia myricoides (lower leaf outline, light grey circles on map), L. silaifolia (top, black circles) and reported http://aob.oxfordjournals.org/ hybrids (middle, dark grey circles) along the east coast of Australia based on herbarium records (www.chah.gov.au/avh, last accessed April 18, 2011). (Inset) Study sites: NSF, 33825′29.9′′S, 150812′14.4′′E, altitude 1160 m; RNP, 3488′59.9′′S, 15181′0.1′′E, 120 m; BH, 33840′59.9′′S, 15189′0′′E, 60 m. Darker areas represent higher altitudes. We have not confirmed the identities of the ten putative records of L. silaifolia from south of Canberra and these should be regarded as dubious.

Lomatia myricoides and L. silaifolia are both capable of differentiation between parental species and their hybrids (Minder resprouting from subterranean lignotubers after fire, but their sus- et al.,2007; Oberprieler et al., 2010). ceptibility to fire and their post-fire growth are distinctively differ- Prior to detailed genomic studies, natural hybridization events ent (Benson and McDougall, 2000). Lomatia myricoides is found must be confirmed, using affordable, rapid screening methods at Griffith University on March 29, 2014 growing on the banks of watercourses in fire-resistant moist forest such as morphological analyses and genotyping techniques or fire-prone eucalypt forest or at the interface of these communi- (Gompert et al., 2010; Marques et al., 2010). In this study we ties. In contrast, L. silaifolia occurs in fire-prone open eucalypt employed morphometric and genetic methods to investigate woodlands and forests (Benson and McDougall, 2000; Pellow the magnitude and the potential causes and consequences et al.,2009), raising questions about the role of selection for diver- of natural hybridization events between L. myricoides and gent habitats in the maintenance of species cohesion. L. silaifolia. In particular we sought to: (1) determine the Fire plays an important role in the reproductive biology of extent of genetic admixture present in hybridizing populations L. silaifolia, as prolific flowering occurs in the first and second and whether it was accurately reflected by morphological vari- seasons after a fire and rarely thereafter (Denham and Whelan, ation; (2) determine whether the patterns of hybridization were 2000). Flowering of L. myricoides, on the other hand, is not fire- uniform between geographically separated hybrid populations dependent and plants are more susceptible to fire damage. seemingly presenting different degrees of in situ sympatry; and Consequently, these species are likely to flower concurrently at (3) analyse the directionality of interspecific gene flow in view a site only if some L. myricoides plants remain unaffected in of the fire dependency of the reproductive biology of Lomatia. the years immediately following a fire. Under such circum- stances, flowers of L. myricoides are likely to outnumber those of L. silaifolia. This suggests the potential for asymmetrical MATERIALS AND METHODS gene flow from L. myricoides into L. silaifolia and their hybrids where the two species co-occur, and for a large number Study species of backcrossed hybrids towards L. myricoides. Lomatia myricoides occurs along the south-east coast of Australia, Intriguingly, the frequency of morphological intermediates south-west of the Great Dividing Range to Melbourne (Fig. 1). [inferred as putative hybrids (Weston and Crisp, 1994)] varies Lomatia silaifolia also occurs along the east coast and Dividing between sympatric populations of the parental species, raising Range, but extends northward to the Tropic of Capricorn. The questions about the degree of intrapopulation gene flow at differ- species are broadly sympatric in the Sydney Basin and the Blue ent sites. The contrasting phenologies, habitat preferences and Mountains region. susceptibilities to fire of the parental species are likely to form Conflorescences in Lomatia are racemes or panicles of lateral strong selective pressures against the persistence of the inter- flower pairs, and flowers appear to be protandrous (Weston, mediate hybrid form. Furthermore, the evolutionary potential 2007). Following fertilization, fruits are dehiscent follicles con- of a population restricted to F1 (first-generation hybrid) indivi- taining numerous, flattened, winged seeds that are wind- duals (such as in Milne et al., 2003) may be significantly less dispersed (Pellow et al., 2009). Lomatia myricoides flowers than a population with a high frequency of gene flow and weak between December and January, whilst L. silaifolia flowers McIntosh et al. — Species cohesion maintained despite ongoing hybridization in Lomatia 863 between November and February (Beadle et al., 1972), and the nSSRlociwereamplifiedin10mlreactionscontaining0.5ngof species share insect pollen visitors (E. J. McIntosh, pers. genomic DNA following the PCR protocol in Milner et al. (2013). observ.). Preliminary seed germination trials using genetically PCRs were repeated for 20 % of samples across all primers and determined hybrids and pure species revealed that hybrids PCR products were visualized by agarose gel electrophoresis were fertile, because flowering produced seeds with germina- before genotyping on an ABI 3730 Genetic Analyzer (Applied tion rates comparable to those of the parental species (results Biosystems, Macrogen, Korea) in a diluted 20 mL multiplex not presented). sample containing 1 mL from each primer pair. Genotyping was also repeated for 20 % of samples to ensure consistent and reliable allele calling. A GeneScanTM 500 LIZw size standard was applied Study sites and nSSR profiles were examined using GeneMapper v4.0 This study focused on populations within the broader sympat- (Applied Biosystems, 2005). Lomatia species and their hybrids ric zone between L. myricoides and L. silaifolia. Herbarium are diploid and allele calls consistently yielded no more than records (www.chah.gov.au/avh) Last accessed April 18, 2011 two peaks, hence no evidence of polyploidy was encountered in and field surveys were used to select three sites representing dif- Lomatia hybrids. The presence of null alleles was tested using ferent degrees of in situ sympatry and hybridization between Micro-Checker v2.2.3 (van Oosterhout et al., 2004).

these species. The upland site at Newnes State Forest (NSF) Downloaded from appeared to be a hybrid swarm of multiple interbreeding Assignment methods and backcrossing hybrid generations, L. silaifolia and L. myri- In order to understand the genotypic structure of each popula- coides growing alongside many plants of intermediate morph- tion and the evolutionary consequences of interspecific gene ology. Both species occurred in close proximity along a creek flow in Lomatia, model-based Bayesian clustering analyses were at the southern coastal site in the Royal National Park (RNP), . conducted in NewHybrids v1 1(Anderson and Thompson, http://aob.oxfordjournals.org/ with a small number of morphological intermediates. At the 2002). This software allocates individuals to one of six genotypic northern coastal site, Bobbin Head (BH), the two species were classes (pure L. myricoides,pureL. silaifolia, F , F ,backcrossto separated by a steep altitudinal gradient along the edge of a 1 2 L. myricoides or backcross to L. silaifolia). No a priori knowledge gully and morphological intermediates were not detected of an individual’s genetic background is required using this despite such a herbarium specimen having been collected in method, meaning that admixture can be detected without the the area in 1969 (accession number NSW130756). need for diagnostic alleles (Anderson and Thompson, 2002). Preliminary tests were conducted using random combinations of Sampling strategy priors, to ensure consistent partitioning of genotypic classes. Bayesian estimates were stable and did not vary greatly depending at Griffith University on March 29, 2014 Transects were constructed to sample representative indivi- on the length of burn-in period, run length or the application of duals of L. myricoides, L. silaifolia and putative hybrids at priors, hence no priors were applied in final NewHybrids analyses. 10 m intervals (Cornelissen et al., 2003). Analyses were conducted bystudysite. Mean posterior probabilities For morphological analysis, 72 Lomatia plants (26 at NSF, 25 were computed from five runs each of a burn-in period of 30 000 at BH and 21 at RNP) were sampled from across the study sites in runs (to allow the Markov chain Monte Carlo simulations to con- October 2010. From each , five mature leaves from the most verge) and a run time of 300 000 sweeps, using the uniform prior. recent growth season were sampled consecutively from the base The original intention was to assign each specimen to one of of the shoot, excluding thosewith evidence of damage or disease. the six genotypic classes when its posterior probability of Herbarium specimens are lodged at the National Herbarium of belonging to that genotypic class equalled or exceeded P ¼ New South Wales. 0.90 (cf. Viscosi et al., 2009; Pinheiro et al., 2010). For genetic analysis, an additional two transects were sampled No specimen met the condition of P ≥ 0.90 for being in the NSF population (total n ¼ 101) as the large number of allocated to the F , F or backcross genotypes as 12 nuclear morphological intermediates suggested the potential for a 1 2 microsatellite markers provided insufficient genetic resolution hybrid swarm to exist, which was worthy of closer examination. (cf. Vaha and Primmer, 2006; Field et al., 2009). As a result, Fresh leaves were silica-dried and stored at –20 8C prior to ex- individuals with P , 0.90 of belonging to a pure L. myricoides traction. Total DNA was extracted from 50 mg of leaf tissue or L. silaifolia genotypic class were considered admixed, of using the DNeasyw 96 Protocol (Qiagenw, Hilden, Germany) hybrid origin, and were lumped into a single class. Three geno- and quantified by agarose gel electrophoresis. typic classes were used for further analysis: pure L. myricoides, pure L. silaifolia and the hybrid pool. It appeared that a substan- Molecular analyses tial genotyping effort, involving more than 48 loci, would be required to separate backcrosses from pure parental individuals Nuclear microsatellites (nSSRs) were used to investigate in future studies (Vaha and Primmer, 2006). intrapopulation level gene flow and diversity between Hereafter, the term ‘hybrid’ refers to individuals genetically L. myricoides, L. silaifolia and their hybrids, due to their high assignedas admixed,andspecies assignments arebasedongenotypic heterozygosity, specificity at low taxonomic levels and highly analysis. polymorphic and co-dominant traits (Goldstein and Pollock, 1997). A total of 147 plants were genotyped at 12 polymorphic Genetic diversity nSSR loci using primers developed for L. silaifolia (LS013, LS046, LS050, LS059, LS064a, LS122, LS129, LS136, The mean number of alleles per polymorphic locus (AP), LS163, LS167, LS178 and LS189a) (Milner et al., 2013). allelic richness (RS)(Mousadik and Petit, 1996), and observed 864 McIntosh et al. — Species cohesion maintained despite ongoing hybridization in Lomatia and expected heterozygosity (Ho and He, respectively) were cal- Correlating genetics and morphology culated by pure species and hybrid cohorts, within sites. Pairwise We sought to determine whether genotypic and morphologic- genetic distances between species and hybrids within sites, and al variations in hybridizing populations were correlated. Lomatia between pure species across sites, were estimated using Weir individuals that had been both successfully genotyped and and Cockeham’s F [Weir and Cockerham, 1984; calculated ST included in morphometric analyses from three study sites (24 in FSTAT version 2.9.3 (Goudet, 2001) following 10,000 permu- from BH, 20 from RNP and 24 from NSF) were included in tations]. Jost’s D (Jost, 2008) was also calculated to account for Mantel tests (Mantel, 1967), conducted in GenAlEx following the dependency of F measures on within-population diversity ST 999 permutations. Mantel tests check for a statistical relationship (Meirmans and Hendrick, 2011). between the elements of two distance matrices with correspond- ing entries. Genetic and morphometric matrices had previously Genetic differentiation between parental species and hybrids been calculated (Codom—genotypic distance in GenAlEx and Principal coordinate analyses (PCoAs) were used to produce Euclidean distance in Primer v6 respectively). comparative graphical representations of genotypic similarity based on a matrix of Codom—genotypic distances, in RESULTS GenAlEx (Peakall and Smouse, 2006). Assignment results Downloaded from Morphometric analysis NewHybrids results revealed the presence of pure parental species Variation in leaf morphology and total plant height between and admixed individuals with varying degrees of introgression at L. myricoides, L. silaifolia and their putative hybrids was examined all three sites. The majority of genotyped plants in the large NSF in an attempt to detect phenetic groupings of individuals and to quan- population were of hybrid origin (78 %, n ¼ 79); few were pure tify variation present within and between these groups. Reproductive parental species (L. myricoides n ¼ 15, L. silaifolia n ¼ 7) http://aob.oxfordjournals.org/ material was not included due to limited availability during the study (Figs 2 and 3). Hybrids ranged from highly to partially admixed season. Leaf images were scanned in Adobe Photoshop version 7.0 at this site (Fig. 2B), a pattern also demonstrated by the gradation with 400 dpi resolution and saved as jpeg files. or ‘V’ shape of admixture inFig. 3. Hybrids were detected at lower Seven single-parameter shape descriptors were selected for numbers in RNP (n ¼ 5), alongside pure L. myricoides (n ¼ 11) morphometric analysis: degree of leaf division; number of leaf and L. silaifolia (n ¼ 5) (Table 1, Fig. 3). lobes; leaf dissection index (LDI); petiole:lamina length ratio; At RNP the composition of hybrid genotypic classes leaf length:width ratio; lamina circularity; and plant height. (Supplementary Data Fig. S1) suggested that several hybrid indi-

The selected variables are important taxonomic traits separating viduals were early generational backcrossestowards L. silaifolia. at Griffith University on March 29, 2014 these species (Pellow et al., 2009), and were screened to avoid a For example, RS1 had a posterior probability of P ¼ 0.242 of high degree of collinearity with other predictor variables (Sokal belonging to the backcross to L. silaifolia genotypic class. and Rohlf, 1995). The variables were also required to displaya dif- Three individuals had a high, although not significant, posterior ferenceinmedianvaluesbetweenL.myricoidesand L.silaifoliaas probability of belonging to the F1 generation, including RH13 inferred from box and whisker plots. (P ¼ 0.743). RH13 was also one of the most morphologically Manual counts of leaf lobes and leaf divisions were conducted intermediate plants, with numbers of leaf divisions and lobes on colour image files and morphometric measures were computed intermediate between, but outside, the ranges of either parental using ImageJ software (version 1.43u, W. S. Rasband, National species (Fig. 2C). Institutes of Health, Bethesda, MD, USA). The degree of leaf div- Only one hybrid was uncovered at BH, with a moderately high ision and number of leaf lobes were defined according to Hickey posterior probability of being of pure L. silaifolia background (1973) and Huff et al. (2003). Many L. silaifolia leaves had over- (P ¼ 0.859) and morphologically grouping amongst individuals lappingleaflets,hence measures ofleafarea andperimeterandcal- of L. silaifolia. culations of LDI (McLellan and Endler, 1998) are acknowledged Comparisons between initial in-field morphological assign- as under-estimations; however, they remain valuable measures ment notes and genotyping assignment results revealed that for separating the species. The petiole:lamina length ratio was large numbers of plants had initially been misidentified, particu- selected as a means of standardizing for petiole length, and larlyat NSF. At this site, 51 % (n ¼ 52) of genotyped plants were leaf length:width ratios and lamina circularity (Huff et al., initially misidentified: 31 plants suspected to be L. silaifoliawere 2003) were included as alternative measures of leaf shape. in fact hybrids, eight of which had a low posterior probability Plant height was measured according to Cornelissen et al. (P , 0.50) of being pure L. silaifolia. A further 17 plants sus- (2003). pected to be L. myricoides were also hybrids, three of which In order to examine morphological relationships between had P , 0.50 of being pure L. myricoides. A further four plants species and hybrids within sites, principal component analyses were misidentified as hybrids, despite the genetics revealing (PCAs) based on covariance matrices were used to produce ordi- that three were pure L. myricoides (P . 0.90) and one was nations from morphometric data using Primer v6 software, and pure L. silaifolia (P . 0.90). methods detailed in Clarke and Gorley (2006) were followed. Specimen data were normalized by trait, and leaf length:width Genetic diversity ratio, petiole:lamina length ratio and LDI were log10-transformed to reduce skew. The degree of leaf division was excluded from the All 12 nSSR loci were found to be polymorphic and highly PCAduetothenon-normaldistributionofthevariableandthecor- variable across both species at all three sites, amplifying a total relation with number of leaf lobes. of 229 alleles across 147 Lomatia plants. The most variable McIntosh et al. — Species cohesion maintained despite ongoing hybridization in Lomatia 865

Morphological Molecular 4 1·2 ABL. myricoides 3 Hybrids 0·8 L. silaifolia 2 0·4 1 # lobes circ· NSF 0 0 log L:W logpet:

PC2 (4·1 %) –1 log LDI height Iam –0·4 PC2 (17·8%) –2 –0·8 –3

–4 –1·2

–4 –3 –2 –1 0 1 2 3 4 –1·2 –0·8 –0·4 0 0·4 0·8 1·2 Downloaded from PC1 (88·1 %) PC1 (26·0%) 4 1·2 C D 3 0·8 2 http://aob.oxfordjournals.org/ 0·4 1 circ· # lobes RNP 0 Site 0 log L:W logpet:Iam log LDI height

PC2 (13·2 %) –1 –0·4 PC2 (22·4%) –2 –0·8 –3 at Griffith University on March 29, 2014 –4 –1·2 –1–2–3–4 43210 1·20·80·40–0·4–0·8–1·2 PC1 (78·9 %) PC1 (33·1%) 4 1·2 EF 3 0·8 2 0·4 1 BH circ· 0 log L:W logpet:Iam 0 # lobes log LDI

PC2 (5·8 %) –1 height PC2 (14·5%) –0·4 –2 –0·8 –3

–4 –1·2 –3–4 –1–2 210 43 0·80·40–0·4–0·8–1·2 1·2 PC1 (90·6 %) PC1 (43·2%)

F IG. 2. Variation in sympatric hybridizing populations of Lomatia by site (NSF, RNP and BH). Lomatia myricoides, L. silaifolia and their hybrids are as indicated in the key in (A). (Left) Morphometric PCAs for sites NSF (A), RNP (C) and BH (E) calculated using six single-parameter shape descriptors[vectors: numberof leaf lobes (# lobes), leaf dissection index (LDI), petiole:lamina length ratio (pet:lam), leaf length:width ratio (L:W), lamina circularity (circ.) and height], across a total of 72 Lomatia plants generated using Primer v6 (Clarke and Gorley, 2006). (Right) Molecular PCoAs of 148 Lomatia plants at NSF (B), RNP (D) and BH (F) computed via a matrix of Codom—genotypic distances in GenAlEx (Peakall and Smouse, 2006). Species assignments are based on genotypic analysis at P ≥ 0.90. Note that some lines for shape descriptors overlay each other. marker across all sites was LS178, with a mean of 30 alleles, Hybrids at NSF had the highest allelic diversity of all samples whereas LS136 and LS064A were the least variable, with across all sites, with a mean of 16.333 alleles per locus (AP), and seven and eight alleles, respectively. allelic richness of 15.957 (RS) (Table 1). Differentiation between 866 McIntosh et al. — Species cohesion maintained despite ongoing hybridization in Lomatia

1·0

0·8

0·6

0·4

Posterior probability 0·2

0·0 BH RNP NSF

F IG. 3. Bar plots depicting Bayesian admixture proportionsestimated using NewHybrids (Anderson and Thompson,2002) with samples from the three studysites (25 from BH, 21 fromRNPand 101 from NSF). The genotypeof each individual is represented bya bar, broken into the posteriorprobabilities of the individual belongingto a Lomatia myricoides (white), L. silaifolia (black) or admixed (grey) background. Downloaded from

TABLE 1. Population diversity indices for Lomatia at three sites where the two parental species occur in sympatry

† Site Species* n AP RS Ho He

NSF L. myricoides 15 7.750 + 0.914 7.592 + 0.884 0.731 + 0.066 0.719 + 0.064 http://aob.oxfordjournals.org/ L. silaifolia 76.167 + 0.683 5.842 + 0.633 0.748 + 0.063 0.717 + 0.054 Hybrids 79 16.333 + 1.982 15.957 + 1.939 0.741 + 0.038 0.819 + 0.044 RNP L. myricoides 11 4.917 + 0.583 4.703 + 0.545 0.698 + 0.069 0.633 + 0.051 L. silaifolia 55.250 + 0.552 4.700 + 0.444 0.763 + 0.064 0.707 + 0.047 Hybrids 5 4.333 + 0.414 3.372 + 0.334 0.796 + 0.078 0.647 + 0.053 BH L. myricoides 12 6.083 + 0.839 5.638 + 0.764 0.612 + 0.071 0.649 + 0.074 L. silaifolia 12 8.667 + 0.907 8.667 + 0.907 0.729 + 0.057 0.744 + 0.048 Hybrids 1 1.833 + 0.112 1.833 + 0.112 0.833 + 0.112 n/a

Abbreviations: n, number of samples; AP, mean number of alleles per polymorphic locus; RS, allelic richness; Ho, observed heterozygosity; He, expected at Griffith University on March 29, 2014 heterozygosity (mean + s.e.); n/a, not applicable (there was only one hybrid at BH). *Species assignments are based on genotypic analysis at P ≥ 0.90. † Values of He for hybrids should be treated with caution as they were calculated in violation of the assumptions of Hardy–Weinberg equilibrium.

parental species was lowat NSF (pairwise FST ¼ 0.146 + 0.043, TABLE 2. Pairwise FST between parental species and between P , 0.001, Table 2; D ¼ 0.533) and observed heterozygosity in parents and the hybrid cohort within sites (mean across 12 nSSR both parental species exceeded that expected (Table 1). This, loci + s.e.) combined with high overall heterozygosity measures (He ¼ Site 0.752 + 0.032, Ho ¼ 0.740 + 0.032) added to evidence of a hybrid swarm at this site. Speciesa NSF RNP BH The distinction between parental species was also low at RNP . . . . (pairwise FST ¼ 0 214 + 0 048, P , 0 01, Table 2; D ¼ 0 679), L. myricoides 0.146 + 0.043*** 0.214 + 0.048** 0.215 + 0.053*** although greater than between parental species at NSF. Gene and L. silaifolia flow at RNP between L. silaifolia and hybrids was slightly L. myricoides 0.035 + 0.011*** 0.126 + 0.022** 0.162 + 0.039 higher than that between L. myricoides and hybrids, as inferred and hybrids L. silaifolia and 0.032 + 0.014** 0.113 + 0.036* 0.009 + 0.037 from FST values (Table 2) and the overall heterozygosity was hybrids high (He ¼ 0.662 + 0.029, Ho ¼ 0.752 + 0.040). Gene flow between the parental species at BH (pairwise There was only one hybrid at BH. a FST ¼ 0.215 + 0.053, P , 0.001, Table 2) did not differ Species assignments are based on genotypic analysis at P ≥ 0.90. greatly from that at RNP, which was interesting given the lack *P , 0.05; **P , 0.01; ***P , 0.001. of recent hybrids found at this site (Fig. 2E, F). However, the two species were most differentiated at BH according to Jost’s Genetic differentiation between parental species and hybrids D (D ¼ 0.741). Intriguingly, gene flow between L. myricoides from each site The PCoAs using 12 nSSRs proved effective means of visu- (Table 3) was lower than, or comparable to, interspecific gene alizing the degree of genetic differentiation and population flow between L. myricoides and L. silaifolia at any one site structures at each site. At NSF, a PCoA showed a cloud of (Table 2). L. silaifolia plants in comparison displayed a high genetic variation spanning the parental extremes and indicated level of gene flow between sites, particularly between RNP and the presence of a hybrid swarm and extensive interspecies gene BH (FST ¼ 0.009 + 0.013) (Table 3). flow (Fig. 2B). In the RNP genetic PCoA three clusters were McIntosh et al. — Species cohesion maintained despite ongoing hybridization in Lomatia 867

TABLE 3. Pairwise FST within species between sites (mean across 12 nSSR loci + s.e.)

NSF RNP BH

NSF – 0.041 + 0.013** 0.020 + 0.012** RNP 0.200 + 0.045*** – 0.009 + 0.013 BH 0.167 + 0.035*** 0.144 + 0.034*** –

Values for L. silaifolia are at the top right, values for L. myricoides are at the bottom left. Species assignments are based on genotypic analysis at P ≥ 0.90. **P , 0.01; ***P , 0.001. revealed, with hybrids intermediate between the parental At RNP, several hybrids separated towards the centre of the or- species clusters, as expected for early generational hybrids dination space, phenotypically more similar to L. silaifolia than (Fig. 2D). A PCoA of genetic distance at BH supported geno- L. myricoides (Fig. 2C). This was primarily due to the effects of

typic assignment methods by confirming that the two morpho- leaf outline complexity (LDI and number of lobes) and petiole Downloaded from logically distinct species clusters were also genetically distinct length (vectors in Fig. 2C). One hybrid shared the ordination (Figs 2Fand3). space of L. silaifolia and was misidentified as L. silaifolia in the field, but genotypic assignment methods suggested it was a Morphometry late-generation backcross. Another hybrid was exceptionally tall, apparently due to its old age, separating from all other indi- Pure L. myricoides and L. silaifolia could be clearly distin- viduals in the ordination space. http://aob.oxfordjournals.org/ guished using morphometric methods. The species did not Phenetic clustering was strongest at BH, where a bimodal PCA overlap even at the extremes of variation in each character state of morphometric variation separated parental species due to (degree of leaf division, number of leaf lobes, LDI, petiole:la- marked differences in leaf shape and plant height (Fig. 2E), mina length ratio, leaf length:width ratio and lamina circularity) and the single late-generation backcross individual was located (Fig. 4), except for plant height, which is highly variable. within the L. silaifolia cluster. Lomatia silaifolia was characterized as a low shrub with a Five individuals at NSF were morphologically cryptic, clus- complex leaf outline; a mode of three leaf divisions, many tering tightly with L. silaifolia in the morphometric ordination

lobes and a high LDI (Figs 2A, C, E and 4). Low leaf length:width despite being genetically more similar to pure L. myricoides. at Griffith University on March 29, 2014 ratios and lamina circularity indices confirmed the comparable Two clustered particularly tightly amongst L. silaifolia indivi- length and width of L. silaifolia laminas, unlike the narrow, duals and both were within the range of variation expected for linear laminas of L. myricoides. The latter was characterized that species for six out of the seven character traits analysed, by its tall stature, simple leaf outlines, no leaf divisions and a despite genetic assignment methods placing them closer to low LDI (Figs 2A, C, E and 4). The low petiole:lamina ratio of L. myricoides. Another plant at NSF, misidentified as L. myricoides indicated a short petiole, in contrast to the long L. silaifolia on the basis of morphology, had a posterior probabil- petioles of L. silaifolia (Fig. 4). ity of P ¼ 0.501 of being from a pure L. myricoides background In a PCA of morphometric data across all three sites (not and only P , 0.002 of being from a pure L. silaifolia back- shown), L. silaifolia clustered tightly when the first and second ground. principal components were visualized, but spread in the third due to variation in the number of leaf lobes, LDI and the petiole: lamina length ratio. Lomatia myricoides was tightly clustered Correlating genetics and morphology when the first and third principal components were visualized Two-way Mantel tests applied to Codom—genotypic distance due to low variation in lamina circularity and the lamina and Euclidean distance matrices across all sites revealed a sig- length:width ratio, but was spread in the second component nificant, low correlation between genetic and morphometric due to wide variation in plant height. datasets (r ¼ 0.442, P , 0.001). When Mantel tests were per- Hybrid individuals were characterized by extreme morpho- formed site by site, a significant, high correlation was detected logical variability, spanning the parental extremes (Fig. 4). The between morphometric and genetic distances amongst Lomatia pooled hybrid cohort was similar to L. silaifolia in height, LDI plants at BH (r ¼ 0.807, P , 0.001); a lesser, but still significant and lamina circularity but closer to L. myricoides in number of correlation was found at RNP (r ¼ 0.616, P , 0.001) and a sig- leaf lobes (Fig. 4). nificant, albeit low, correlation was found at NSF (r ¼ 0.241, The PCAs of morphometric data by site showed distinct paren- P ¼ 0.001). A Mantel test for hybrids only at NSF revealed a tal species clusters, but hybrids were not evenly distributed across non-significant, low correlation (r ¼ 0.073, P ¼ 0.183). the continuum between the two and most fell towards the margins of a particular parental species cluster (Fig. 2A, C, E). At each site, more than 95 % of the variance was explained by the first three DISCUSSION principal components and between 79 and 91 % by the first Extent of genetic admixture in hybridizing Lomatia populations alone. At NSF the PCAwas slightly bimodal, with few intermedi- ate hybrids and a large numberof hybrids morphologicallysimilar Despite their distinct ecological and morphometric character to one or other parental species (Fig. 2A). traits, molecular evidence of admixture was detected at all 868 McIntosh et al. — Species cohesion maintained despite ongoing hybridization in Lomatia

4 250

200 * 3

150 * 2 * 100

1 50 Degree of leaf division Degree

0 0

0·20 1·6

0·15 1·4 Downloaded from 0·10 1·2

0·05 Lamina circularity 1·0 Log leaf dissection index 0 http://aob.oxfordjournals.org/

0 1·5

–0·2 1·0 –0·4

–0·6 0·5 –0·8 at Griffith University on March 29, 2014 0 –1·0

–1·2 width ratio width ratioLog leaf length : Log leaf length : Number of leaf lobes

Log petiole : lamina length ratio Log petiole : –0·5 L. myricoidesHybrids L. silaifolia 6

5

4

3

2 Plant height (m) 1 *

0 L. myricoides Hybrids L. silaifolia

F IG. 4. Box and whisker plots of morphometric traits used to distinguish between between L. myricoides (n ¼ 26), L. silaifolia (n ¼ 19) and hybrids (n ¼ 23), cal- culated in IBMw SPSSw v19 (Sun Microsystems, USA). Shaded boxes span the interquartile range and the solid black line is the median. Individuals not successfully genotyped were excluded. Outliers and extreme outliers are represented by small circles and stars respectively. study sites where L. myricoides and L. silaifolia occurred in sym- glimpse into the different stages of natural hybridization in situ patry. However, the patterns of hybridization differed markedly (Lamont et al., 2003). between sites. Suggestions of multiple hybrid generations and extensive introgression characterized the hybrid swarm at NSF, whereas hybrids were uncommon and likely of early generations Morphological variation and cryptic hybrids at RNP. Despite no morphological signs of recent hybridization The strong morphometric and ecological distinction between at BH, genotyping revealed a signature of admixture in one L. myricoides and L. silaifolia exceeds that found in most studies individual. Such site-specific patterns of introgression offer a of naturally hybridizing plant species (Maguire et al., 2002; McIntosh et al. — Species cohesion maintained despite ongoing hybridization in Lomatia 869

Viscosi et al., 2009). Genetically identified pure species could be regarding the direction of introgression and the evolution of clearly distinguished by morphometric studies; however, when reproductive barriers in this hybrid zone. such approaches were used to interpret hybridizing populations, they masked the true extent of genetic diversity, particularly in Fire dependency of reproductive isolation between L. myricoides the NSF hybrid swarm. and L. silaifolia Lomatia hybrids had highly variable morphologies, likely the results of random segregation, novelty and genetic recombin- We hypothesize that hybridization between L. myricoides and ation amongst multiple hybrid generations, as well as the L. silaifolia is restricted to certain stochastic events as F1 hybrids gradual decay of linkage disequilibrium over time (Rieseberg are rare. Herbarium records suggest that, despite their evident re- and Ellstrand, 1993; Arnold and Martin, 2010). As in an increas- productive compatibility, broadly overlapping flowering seasons ing number of studies, the majority of Lomatia hybrids expressed and substantial area of sympatry, morphological evidence for hy- greater phenetic similarity to either L. myricoides or L. silaifolia bridization is lacking at most sites where L. myricoides and rather than being strictly intermediate in all characters L. silaifolia co-occur. (cf. Lihova´ et al., 2007; Field et al., 2009). Leaf shape in One plausible explanation for the low incidence of F1 hybrids Lomatia may be influenced by diversifying selection towards istemporal reproductive isolation imposed by the timing of flow-

the two parental character states (Oberprieler et al., 2010), ering after fire. Both Lomatia species resprout from lignotubers Downloaded from explaining the contrast between the slightly bimodal morpho- after fire but L. silaifolia concentrates its reproductive effort in metric dataset and the continuum of genetic diversity in hybrid the first year or two following a fire (Denham and Whelan, populations. 2000), whereas L. myricoides is more sensitive and is unlikely For the purposes of identifying hybrids in the field, petiole to flower in the years immediately following fire unless plants length relative to overall leaf length and the number of leaf are not burnt. Fire records at the study sites corroborate this lobes are likely to be the most reliable morphological characters, theory. Early-generation hybrids at RNP may be the result of a http://aob.oxfordjournals.org/ although it should be noted that all species of Lomatia have recent fire event, whereas opportunities for F1 formation at NSF highly variable morphology, L. silaifolia potentially most of arelikelytohavebeenfewersincenofireshadbeenthroughthe all (Weston and Crisp, 1994). areaforat least a decade. As expected, no L. silaifoliawere observed The parental extremes of morphometric variation masked the to flower at NSF in the year of this study (E. J. McIntosh, pers. hybrid origin of many Lomatia plants, particularly in the observ.). complex hybrid swarm at NSF. Hybrids that were misidentified Our findings of bi-directional gene flow between the parental in the field as pure species exhibited high levels of genetic admix- species at NSF do not support our expectation that gene flow is

ture and were likely to be later-generation backcross hybrids re- primarily asymmetrical from L. myricoides into hybrids. at Griffith University on March 29, 2014 sembling the parental species to which they were genetically Therefore, the hypothesis that fire is a mechanism for producing most similar. This explains the low correlation between morpho- asymmetry has been retracted, but the way in which fire shapes metric and molecular patterns of variation within the hybrid patterns of hybridization between Lomatia requires further study. swarm. Interestingly, where species connectivity was not appar- Stochastic fire events are an important feature in the occur- ent at BH, intermediate morphotypes had disappeared, masking rence of hybridization of other species in Australian sclerophyl- the introgressed genes that remained in the population. lous environments, such as between robur and The large proportion of plants incorrectly assigned in the field B. oblongifolia (Usher et al., 2010) and Kunzea rupestris and and the presence of cryptics genetically close to one parent but K. capitata (Tierney and Wardle, 2008). In the latter study, morphologically resembling the other pointed to complex pat- Kunzea hybrids did not resprout after fire, and this lower fitness terns of introgression between L. myricoides × L. silaifolia relative to parental plants appeared to restrict the evolutionary hybrids. While the examination of morphological character potential of hybrids, a trait that should be further investigated traits will always be an important component of hybridization in Lomatia hybrids. The requirement for fire to stimulate flower- studies, these findings highlight the value of molecular tools ing of L. silaifolia is likely to lead to the re-establishment of re- when identifying gene flow among closely related but strongly productive isolating barriers between pure species following differentiated lineages (Minder et al., 2007; Marques et al., F1 formation, and our findings suggest backcrosses may occur 2010). readily following these initial hybridization events (Broyles, There is great potential with next-generation sequencing 2002). As initial genetic bridges, F1 hybrids allow introgressive (NGS) technologies to trace the extent of genome-wide intro- hybridization to subsequently become an important evolutionary gression (Lai et al., 2012; Twyford and Ennos, 2012), identify force in a population (Ellstrand et al., 2007). introgressed genes and determine the evolutionary adaptive effects of introgressed genomic regions (Song et al., 2011; Transient hybrid swarms or ongoing gene exchange via Heliconius Genome Consortium, 2012). The decreasing costs, introgression? high throughput and extensive coverage of NGS technologies make them valuable for non-model species and those with Hybridization between L. myricoides and L. silaifolia is un- little available genomic data, such as Lomatia (Twyford and likely to be a recent phenomenon where these species are sympat- Ennos, 2012), where traditional approaches involve significant ric. Porous species barriers characterize the entire genus; six of time and costs to obtain the large numbers of markers required the nine Australian native Lomatia species have been reported in population genetic studies. The analysis of plastid markers to hybridize, and both L. myricoides and L. silaifolia hybridize alongside the genotyped nuclear markers in this study is a with other Lomatia species elsewhere over their broad geograph- priority for future research and will greatly improve inferences ic ranges (Milner et al., 2012). Furthermore, the high level of 870 McIntosh et al. — Species cohesion maintained despite ongoing hybridization in Lomatia gene flow detected between parental species at NSF, as well the cohesion, such as stabilizing selection or the spread of advanta- presence of introgression at BH despite no morphologically geous alleles (Martinsen et al., 2001; Palma-Silva et al., 2011). identifiable hybrids, hints at a pattern of recurrent introgression This ‘porous view’ of species boundaries (Muir et al., 2012), (potentially over many generations) between L. myricoides involves the maintenance of conserved genomic regions, pre- and L. silaifolia across their shared geographic region. Such find- serving intraspecific cohesion despite the presence of introgressed ings add to growing research suggesting that even strong barriers genes at unlinked or more distantly linked chromosomal regions may not be sufficient to prevent gene flow between species (Mallet, 2005). This is evident between sympatric Pitcairnia (Stecconi et al., 2004; Sambatti et al., 2012). species, in which asymmetrical introgression rates of single loci Genetic distinction between L. myricoides at separate sites was suggested that Bateson–Dobzhansky–Muller incompatibilities unexpectedly revealed in this study. The pattern was not present contribute to the maintenance of species integrity (Palma-Silva amongst sampled L. silaifolia and could be a consequence of the et al.,2011). greater habitat specificity of L. myricoides. Site-by-site differ- ences in L. myricoides match the findings of Milner et al. (2013) that cpDNA gene phylogenies of Australian Lomatia Conclusions species are more congruent with geography than taxonomy, con- Morphological and ecological distinctions between plant

firming that geography has been important in the evolution of the Downloaded from genus. Exploration of within-species genetic structurewas not an species can be maintained despite ongoing gene flow via natural objective of this study but will be the focus of in-depth future hybridization. Localized gene flow and introgression are expected studies, including both nuclear and chloroplast markers, as in to be ongoing between L. myricoides and L. silaifolia and their recent investigations of the closely related Telopea species hybrids wherever they occur in sympatry, due to the permeability (Rossetto et al., 2011, 2012). of this species barrier. Future studies will include the use of NGS

technologies and detailed analyses of plastid data to better charac- http://aob.oxfordjournals.org/ terize theextentand direction of introgression inL.myricoides and Species cohesion in the face of permeable species boundaries L. silaifolia, and to gain insight into how less porous regions of Our findings indicate a need to consider the processes by the genome contribute to maintaining the integrity of parental Lomatia which species maintain distinct morphologies, genotypes, eco- species. logical preferences, phenologies and broader geographic distri- butions, despite ongoing interspecific exchanges of genetic material. Species cohesion in Lomatia is expected to be main- SUPPLEMENTARY DATA

tained despite the likelihood of continued gene flow and intro- Supplementary data are available online at www.aob.oxford at Griffith University on March 29, 2014 gression wherever the species co-occur (Usher et al., 2010). journals.org and consist of Supplementary Data Figure S1: bar Potential explanations include selective pressures maintaining plots depicting Bayesian admixture proportions estimated species-specific character traits, genomic filters reducing intro- using NewHybrids with samples from three study sites. gression at key genomic sites, and adaptive introgression being triggered only when environmental conditions change (Reed et al., 2011). ACKNOWLEDGEMENTS Gene flow can be ongoing even in cases where selective pres- sures and isolating barriers are strong (Ito et al., 2008; Sambatti We thank the National Herbarium of New South Wales, et al., 2012). The distinct environmental, phenotypic and geno- Australia, M. Milner, K. Thurlby, P. Rymer, C. Connelly, typic characteristics of pure L. myricoides and L. silaifolia may A. Denham, C. Allen, field volunteers and two anonymous constitute two adaptive peaks maintained by selection. We reviewers. This work was supported by the Herman Slade observed very restricted intermediate habitats compared with Foundation (HSF 07-1 to M.R. and P.H.W.). the preferred habitats of the two parental species (unlike Rieseberg et al., 1998; Lihova´ et al., 2007). Microclimatic tem- perature measurements proved insufficient to quantify this LITERATURE CITED habitat transition (data not shown); however, our observations Anderson EC, Thompson EA. 2002. A model-based method for identifying suggested the transition was sharp, and at BH in particular species hybrids using multilocus genetic data. Genetics 160: 1217–1229. L. myricoides was tightly restricted to a steep-sided, wet creek- Arnold ML. 1997. Natural hybridization and evolution. New York: Oxford University Press. line in an otherwise dry sclerophyll vegetation type, where Arnold ML, Martin NH. 2010. Hybrid fitness acrosstime and habitats. Trends in L. silaifolia inhabited the higher, drier slopes. Hybrids were Ecology & Evolution 25: 530–536. found in both habitats but this warrants further investigation, par- Beadle NCW, Evans OD, Carolin RC. 1972. Flora of the Sydney region, revised ticularly if later-generation backcrosses are more likely to be edn. Sydney: Reed. restricted to the characteristic parental habitat types than earlier- Benson D, McDougall L. 2000. Ecology of Sydney plant species: Part 7b: Dicotyledon families Proteaceae to Rubiaceae. Cunninghamia 6: generation hybrids. This ecological barrier is suspected to con- 1017–1202. tribute to the rapid reversion of hybrid populations to their paren- Broyles SB. 2002. Hybrid bridges to gene flow: a case study in milkweeds tal character state due to limited opportunities for hybrid/ (Asclepias). Evolution 56: 1943–1953. intermediate advantage (Arnold, 1997). Clarke K, Gorley R. 2006. PRIMER v6: user manual/tutorial. Plymouth: PRIMER-E. Strong genomic filters may facilitate continued gene flow Cornelissen JHC, Lavorel S, Garnier E, et al. 2003. A handbook of protocols between species without the danger of assimilation, due to the for standardised and easy measurement of plant functional traits worldwide. powerful effect of selective forces acting to maintain species Australian Journal of Botany 51: 335–380. McIntosh et al. — Species cohesion maintained despite ongoing hybridization in Lomatia 871

Denham AJ, Whelan RJ. 2000. Reproductive ecology and breeding system of Minder AM, Rothenbuehler C, Widmer A. 2007. Genetic structure of Lomatia silaifolia (Proteaceae) following a fire. Australian Journal of hybrid zones between Silene latifolia and Silene dioica (Caryophyl Botany 48: 261–269. laceae): evidence for introgressive hybridization. Molecular Ecology Ellstrand NC, Whitkus R, Rieseberg LH. 1996. Distribution of spontaneous 16: 2504–2516. plant hybrids. Proceedings of the National Academy of Sciences of the Moccia MD, Widmer A, Cozzolino S. 2007. The strength of reproductive isola- USA 93: 5090–5093. tion in two hybridizing food-deceptive orchid species. Molecular Ecology Ellstrand NC, Gardner LC, Hegde S, Guadagnuolo R, Blancas L. 2007. 16: 2855–2866. Spontaneous hybridization between maize and teosinte. Journal of Mousadik A, Petit RJ. 1996. High level of genetic differentiation for allelic Heredity 98: 183–187. richness among populations of the argan tree [Argania spinosa (L.) Field DL, Ayre DJ, Whelan RJ, Young AG. 2009. Molecularand morphologic- Skeels] endemic to Morocco. Theoretical and Applied Genetics 92: al evidence of natural interspecific hybridization between uncommon 832–839. Eucalyptus aggregata and the widespread E. rubida and E. viminalis. Muir G, Dixon CJ, Harper AL, Filatov DA. 2012. Dynamics of drift, gene flow, Conservation Genetics 10: 881–896. and selection during speciation in Silene. Evolution 66: 1447–1458. Goldstein DB, Pollock DD. 1997. Launching microsatellites: a review of muta- Nosil P, Schluter D. 2011. The genes underlying the process of speciation. tion processes and methods of phylogenetic inference. Journal of Heredity Trends in Ecology & Evolution 26: 160–167. 88: 335–342. Oberprieler C, Barth A, Schwarz S, Heilmann J. 2010. Morphological and Gompert Z, Lucas LK, Fordyce JA, Forister ML, Nice CC. 2010. Secondary phytochemical variation, genetic structure and phenology in an introgres- contact between Lycaeides idas and L. melissa in the Rocky Mountains: ex- sive hybrid swarm of Senecio hercynicus and S. ovatus (Compositae, tensive admixture and a patchy hybrid zone. Molecular Ecology 19: Senecioneae). Plant Systematics and Evolution 286: 153–166. Downloaded from 3171–3192. Palma-Silva C, Wendt T, Pinheiro F, et al. 2011. Sympatric bromeliad species Goudet J. 2001. FSTAT, a program to estimate and test gene diversities and fix- (Pitairnia spp.) facilitate tests of mechanisms involved in species cohesion ation indices (version 2.9.3). http://www2.unil.ch/popgen/softwares/ and reproductive isolation in Neotropical inselbergs. Molecular Ecology 20: fstat.htm (last accessed 31 March, 2011). 3185–3201. Heliconius Genome Consortium. 2012. Butterfly genome reveals promiscuous Peakall R, Smouse PE. 2006. GENALEX 6: genetic analysis in Excel. exchange of mimicry adaptations among species. Nature 487: 94–98. Population genetic software for teaching and research. Molecular Ecology Hickey LJ. 1973. Classification of the architecture of dicotyledonous leaves. Notes 6: 288–295. http://aob.oxfordjournals.org/ American Journal of Botany 60: 17–33. Pellow BJ, Henwood MJ, Carolin RC. 2009. Flora of the Sydney region, 5th Huff PM, Wilf P, Azumah EJ. 2003. Digital future for paleoclimate estimation edn. Sydney: Sydney University Press. from fossil leaves? Preliminary results. Palaios 18: 266–274. Pinheiro F, De Barros F, Palma-Silva C, et al. 2010. Hybridization and intro- Ito M, Suyama Y, Ohsawa TA, Watano Y. 2008. Airborne-pollen pool and gression across different ploidy levels in the Neotropical orchids mating pattern in a hybrid zone between Pinus pumila and P. parviflora Epidendrum fulgens and E. puniceoluteum (Orchidaceae). Molecular pentaphylla Molecular Ecology 17 var. . : 5092–5103. Ecology 19: 3981–3994. Jost L. 2008. G and its relatives do not measure differentiation. Molecular ST Reed TE, Schindler DE, Waples RS. 2011. Interacting effects of phenotypic Ecology 17: 4015–4026. plasticity and evolution on population persistence in a changing climate. Lai Z, Kane NC, Kozik A, et al. 2012. Genomics of Compositae weeds: EST li- Conservation Biology 25: 56–63. braries, microarrays, and evidence of introgression. American Journal of

Rieseberg LH, Ellstrand NC. 1993. What can molecular and morphological at Griffith University on March 29, 2014 Botany 99: 209–218. markers tell us about plant hybridization? Critical Reviews in Plant Lamont BB, He T, Enright NJ, Krauss SL, Miller BP. 2003. Anthropogenic Sciences 12: 213–241. distrubance promotes hybridization between Banksia species by altering Rieseberg LH, Wendel JF. 1993. Introgression and its consequences in plants. their biology. Journal of Evolutionary Biology 16: 551–557. In: Harrison RG. ed. Hybrid zones and the evolutionary process. New York: Lihova´ J, Kucˇera J, Perny´ M, Marhold K. 2007. Hybridization between two Oxford University Press, 70–109. polyploid Cardamine (Brassicaceae) species in north-western Spain: dis- cordance between morphological and genetic variation patterns. Annals of Rieseberg LH, Baird SJE, Desrochers AM. 1998. Patterns of mating in wild Botany 99: 1083–1096. sunflower hybrid zones. Evolution 52: 713–726. Maguire TL, Peakall R, Saenger P. 2002. Comparative analysis of genetic di- RossettoM, ThurlbyKA, OffordCA, Allen CB, Weston PH. 2011. The impact versity in the mangrove species Avicennia marina (Forsk.) Vierh. of distance and a shifting temperature gradient on genetic connectivity (Avicenniaceae) detected by AFLPs and SSRs. Theoretical and Applied across a heterogeneous landscape. BMC Evolutionary Biology 11: e126. Genetics 104: 388–398. doi:10.1186/1471-2148-11-126. Mallet J. 2005. Hybridization as an invasion of the genome. Trends in Ecology & Rossetto M, Allen CB, Thurlby KA, Weston PH, Milner ML. 2012. Genetic Evolution 20: 229–237. structure and bio-climatic modeling support allopatric over parapatric spe- Mantel N. 1967. The detection of disease clustering and a generalized regression ciation along a latitudinal gradient. BMC Evolutionary Biology 12: e149. approach. Cancer Research 27: 209–220. doi:10.1186/1471-2148-12-149. Marques I, Feliner GN, Munt DD, Martins-Louc¸a˜o MA, Aguilar JF. 2010. Salvini D, Bruschi P, Fineschi S, Grossoni P, Kjær ED, Vendramin GG. 2008. Unraveling cryptic reticulate relationships and the origin of orphan hybrid Natural hybridization between Quercus petraea (Matt.) Liebl. and Quercus disjunct populations in Narcissus. Evolution 64: 2353–2368. pubescens Willd. within an Italian stand as revealed by microsatellite finger- Martinsen GD, Whitham TG, Turek RJ, Keim P. 2001. Hybrid populations printing. Plant Biology 11: 758–765. selectively filter gene introgression between species. Evolution 55: Sambatti JBM, Strasburg JL, Ortiz-Barrientos D, Baack EJ, Rieseberg LH. 1325–1335. 2012. Reconciling extremely strong barriers with high levels of gene ex- McLellan T, Endler JA. 1998. The relative success of some methods for meas- change in annual sunflowers. Evolution 66: 1459–1473. uring and describing the shape of complex objects. Systematic Biology 47: Sokal RR, Rohlf FJ. 1995. Biometry: the principles and practice of statistics in 264–281. biological research, 3rd edn. New York: W. H. Freeman. Meirmans PG, Hendrick PW. 2011. Assessing population structure: FST and Song Y, Endepols S, Klemann N, et al. 2011. Adaptive introgression of anti- related measures. Molecular Ecology Resources 11: 5–18. coagulant rodent poison resistance by hybridization between old world Milne RI, Terzioglu S, Abbott RJ. 2003. A hybrid zone dominated by fertile mice. Current Biology 21: 1296–1301. F1s: maintenance of species barriers in Rhododendron. Molecular Stecconi M, Marchelli P, Puntieri J, Picca P, Gallo L. 2004. Natural hybridiza- Ecology 12: 2719–2729. tion between a deciduous (Nothofagus antarctica, Nothofagaceae) and an Milner ML, Rossetto M, Crisp MD, Weston PH. 2012. The impact of multiple evergreen (N. dombeyi) forest tree species: evidence from morphological biogeographical barriers and hybridization on species-level differentiation. and isoenzymatic traits. Annals of Botany 94: 775–786. American Journal of Botany 99: 2045–2057. Tierney DA, Wardle GM. 2008. The relative fitness of parental and hybrid Milner ML, McIntosh EJ, Crisp MD, Weston PH, Rossetto M. 2013. Kunzea (Myrtaceae): the interaction of reproductive traits and ecological se- Microsatellite variation for phylogenetic, phylogeography and population lection. American Journal of Botany 95: 146–155. genetic studies in Lomatia (Proteaceae). Australian Systematic Botany 26: Twyford AD, Ennos RA. 2012. Next-generation hybridization and introgres- 186–195. sion. Heredity 108: 179–89. 872 McIntosh et al. — Species cohesion maintained despite ongoing hybridization in Lomatia

Usher AV, Whelan RJ, Ayre DJ. 2010. Window of opportunity: an episode of Weir BS, Cockerham CC. 1984. Estimating F-statistics for the analysis of recruitment in a Banksia hybrid zone demonstrates continuing hybridization population-structure. Evolution 38: 1358–1370. and phenotypic plasticity. Annals of Botany 105: 419–429. Weston PH. 2007. Proteaceae. In: Kubitzki K. ed. Families and genera of vascu- Vaha JP, Primmer CR. 2006. Efficiency of model-based Bayesian methods for lar plants, IX. Berlin: Springer, 364–404. detecting hybrid individuals under different hybridization scenarios and Weston PH, Crisp MD. 1994. Cladistic biogeography of (Proteaceae: with different numbers of loci. Molecular Ecology 15: 63–72. Embothrieae) and their allies across the Pacific. Australian Systematic van Oosterhout C, Hutchinson WF, Wills DPM, Shipley P. 2004. Botany 7: 225–249. MICRO-CHECKER: software for identifying and correcting genotyping Wilson A, Hewson H, Mowatt J. 1995. Lomatia. In: McCarthy P. ed. Flora of errors in microsatellite data. Molecular Ecology Notes 4: 535–538. Australia, Vol. 16, Elaeagnaceae, Proteaceae 1. Melbourne: CSIRO Viscosi V, Lepais O, Gerber S, Fortini P. 2009. Leaf morphological analyses in Publishing, 374–382. four European oak species (Quercus) and their hybrids: A comparison of trad- Wu CI. 2001. The genic view of the process of speciation. Journal of itional and geometric morphometricmethods. Plant Biosystems 143:564–574. Evolutionary Biology 14: 851–865. Downloaded from http://aob.oxfordjournals.org/ at Griffith University on March 29, 2014