Relationships, gene flow and species boundaries among New Zealand Fuscospora (Nothofagaceae: southern beech)

Abstract

We analysed DNA profiles at eight simple sequence repeat (SSR) loci for 449 individuals of Fuscospora species from New Zealand. The two species with serrated leaves (F. fusca and F. truncata) are clearly distinct genetically despite the occurrence of wild hybrids. Further, from our results, we infer that they are no more closely related to each other than they are to either of the entire-leaved species (F. cliffortioides and F. solandri). Although genetic groups corresponding with F. cliffortioides and F. solandri can be distinguished, our data suggest that considerable admixture may occur between them in some areas. Nonetheless, the broadly sympatric distribution of these two genetic groups, and their consistency with morphological and ecological groups, support their recognition at species rank rather than their treatment as varieties of one variable species.

Keywords:

• black beech
• hard beech
• hybrid
• hybridisation
• microsatellites
• mountain beech
• New Zealand
• Nothofagus subgenus Fuscospora
• red beech
• SSR

Introduction

Southern beech (Nothofagaceae Kuprian.) has been central to a number of debates about the origins and evolution of plants in New Zealand, particularly the relative importance of ancient Gondwanan lineages and vicariance or later dispersals in the composition of the flora (Knapp et al. 2005), and the importance or otherwise of hybridisation in the adaptation and evolution of plant species (Wardle et al. 1988; Ogden 1989). They have also been studied with regard to the impact of past glacial periods on the present-day distribution of species in New Zealand (Wardle 1963; Thomsen 2002; Leschen et al. 2008), and elsewhere in the Southern Hemisphere (Marchelli & Gallo 2004; Worth et al. 2009; Vergara et al. 2014). Following historical clearance of lowland conifer–broadleaved forests, southern beech species are now collectively the most common and widespread trees in New Zealand (McGlone et al. 1996) and their ecology has been extensively studied (Wardle 1984; Leathwick 1998; Hurst et al. 2012). All of the southern beech species that are native to New Zealand are endemic.

Phylogenetic studies have shown that four distinct clades of southern beech are extant (Martin & Dowd 1993; Manos 1997; Setoguchi et al. 1997; Sauquet et al. 2012; Heenan & Smissen 2013). At a time when they were often included alongside Northern Hemisphere beech (Fagus L.) and oak (Quercus L.) within Fagaceae Dumort., these clades were recognised at sub-generic rank within Nothofagus Blume (Hill & Read 1991). Subsequently, southern beech species have been accepted as belonging to a distinct family, Nothofagaceae, recognising that they are the sister group to the remainder of the order Fagales (Manos et al. 1993; Manos & Steele 1997; The Angiosperm Phylogeny Group 1998) and therefore relatively distantly related to Fagaceae. Given this change in family placement, and after assessment of the levels of genetic and morphological diversity in Nothofagaceae in comparison with other families of Fagales, Heenan & Smissen (2013) recognised the four extant southern beech clades as four genera: Nothofagus Blume (in southern South America), Fuscospora (R.S.Hill & J.Read) Heenan & Smissen (in Australasia and southern South America), Lophozonia Turcz. (in Australasia and southern South America) and Trisyngyne Baill. (in New Caledonia, New Guinea, and nearby islands).

Two of these genera are present in New Zealand: Lophozonia and Fuscospora. The genus Lophozonia is represented in New Zealand by L. menziesii (Hook.f.) Heenan & Smissen (silver beech) and Fuscospora by F. cliffortioides (Hook.f.) Heenan & Smissen (mountain beech), F. fusca (Hook.f.) Heenan & Smissen (red beech), F. solandri (Hook.f.) Heenan & Smissen (black beech) and F. truncata (Colenso) Heenan & Smissen (hard beech). The New Zealand Fuscospora species have either entire leaves (F. cliffortioides and F. solandri) or serrated leaves (F. fusca and F. truncata). Mountain beech was historically considered a species distinct from F. solandri (as Fagus cliffortioides Hook.f. –Hooker 1844, or as Nothofagus cliffortioides (Hook.f.) Oerst. –Cockayne 1926; Cockayne & Atkinson 1926) but Poole (1958) reduced it to a variety of F. solandri because the two forms were linked by numerous populations displaying intermediate morphology. However, some authors have continued to regard mountain beech as a distinct species (Molloy et al. 1999; Meurk & Hall 2006; McGlone et al. 2011) and this was the position and rank adopted by Heenan and Smissen (2013) when segregating Fuscospora from Nothofagus. Fuscospora cliffortioides and F. solandri have a wide area of sympatry in the North and South Islands, have differentiated habitats, can be distinguished by their morphology and were therefore considered by them to be better recognised at species rank. According to Wardle (1984), black beech (F. solandri) is a tall tree averaging 20–25 m, and sometimes reaching 30 m, while mountain beech (F. cliffortioides) is smaller, usually 12–15 m, sometimes reduced to a prostrate shrub of only 0.5 m, but on fertile lowland sites sometimes reaching 25 m. The elevational limit of F. solandri is usually 600–700 m, but it can grow up to 800 m in the north of its range. Fuscospora cliffortioides frequently ascends to 1300–1400 m and sometimes higher. The two taxa are often distinguished by differences in leaf shape, but both are variable and intermediate forms are common. The species with serrated leaves, F. fusca and F. truncata, can also be difficult to distinguish and are considered by some to be more closely related to each other than to the entire-leaved species, or to be dubiously distinct (e.g. van Steenis 1953; Ogden 1989; Govaerts & Frodin 1999).

The four New Zealand species of Fuscospora overlap considerably in their ranges, and mixed forests are common; yet each species has distinctive ecological requirements (Wardle 1984; Leathwick 1995). Published chromosome counts for F. cliffortioides and F. fusca are all n = 13 (Armstrong & Wyile 1965; Dawson 2000), in agreement with an unpublished count for F. truncata of 2n = 26 (P. de Lange, New Zealand Department of Conservation, pers. comm. 30 January 2014) and it seems likely this number is also to be found in F. solandri. All the species are wind pollinated. Hybrids between species of Fuscospora have been reported from sites where more than one species occurs. Wild southern beech hybrids from New Zealand were brought to the attention of science by Leonard Cockayne and their occurrence and character variation are described in Cockayne and Atkinson (1926). The morphology of probable hybrids in the wild varies considerably (Cockayne 1926; Cockayne & Atkinson 1926), suggesting that second- and later-generation hybrids may occur. Artificial hybrids have been generated between F. solandri and F. fusca, and the fertility of seed from wild hybrid F. cliffortioides × F. fusca has been demonstrated (Poole 1951).

Given their dominance in some landscapes over wide areas of the country, and their past and potential future value as a timber resource, surprisingly little work has been published on the genetic variation within and among southern beech species in New Zealand. Exceptions include the important provenance trial reported by Wilcox and Ledgard (1983) and the isozyme studies of Haase (1992, 1993a,b, 1994). Haase (1992) reported little genetic variation between populations of Fuscospora species in New Zealand, with most isozyme variation occurring within populations. However, levels of isozyme polymorphism in these species were low. Genetic variation and hybridisation in species of Lophozonia (as Nothofagus subgenus Lophozonia) and Nothofagus in Australia and South America have been studied using chloroplast DNA, allozymes and simple sequence repeat markers (Marchelli & Gallo 2001; Quiroga et al. 2005; Worth et al. 2009; Acosta & Premoli 2010; Acosta et al. 2012; Premoli et al. 2012; Vergara et al. 2014).

In this study, we report analysis of eight simple-sequence repeat (SSR) loci (also called microsatellites) for 449 individuals of Fuscospora including plants identified as F. cliffortioides, F. fusca, F. truncata, F. solandri and possible hybrids among them. These data were assessed for evidence of phylogenetic relationships among the species, hybridisation and recent gene flow among them, and for phylogeographic structure within the species. A particular focus is to examine whether the SSR loci provide any support for the recognition of mountain (F. cliffortioides) and black beech (F. solandri) as distinct species.

Materials and methods

Details of sample locations are shown in Appendix 1 (Table A1.1). Between 1 and 12 plants were sampled at each location. Species identifications were made with reference to Wardle (1984) and other sources. In general, we did not attempt to diagnose hybrid plants a priori because hybrids can be cryptic (resembling closely one parental species or the other, or exhibiting leaf dimorphism between crown and epicormic branches: Wardle 1984), particularly those between the serrate-leaved species F. fusca and F. truncata, and between the two entire-leaved species F. cliffortioides and F. solandri. However, two plants from Tom Creek had leaf morphology strongly suggesting that they were F. cliffortioides × F. fusca and one plant from MacFarlane Mound had leaf morphology strongly suggesting F. cliffortioides × F. truncata (see Table A1.1). These were initially grouped with F. fusca and F. truncata respectively for Structure analysis (see below). Leafy twigs were sampled in the field and frozen on return to the laboratory until DNA was extracted. DNA extractions were conducted using the Maxwell® DNA extraction system (Promega, Madison, WI, USA) from leaf, twig or leaf-bud (preferred if present) then further purified by phenol–chloroform–isoamyl alcohol extraction and recovered using a Zymo spin column (Zymo Research Corp., Irvine, CA, USA). The seven SSR loci presented in Smissen et al. (2012) were amplified as described in that reference. We also used the primer pair ncutas13 of Jones et al. (2004), designed for Lophozonia cunninghamii (Hook.f.) Heenan & Smissen (as Nothofagus cunninghamii (Hook.f.) Oerst.) with an annealing temperature in polymerase chain reactions of 60 °C. Markers were sized using an ABI3730 genetic analyser. Allele scoring was undertaken manually using GeneMarker 1.51 (SoftGenetics, State College, PA, USA) to visualise fragment size analysis files. Occasional reactions failing to produce a detectable fragment were repeated. Samples for which reactions for more than two loci failed to produce a detectable fragment twice were omitted as low-quality DNA extracts. All scores were checked by repeat scoring. For six of the loci, alleles differed in size by amounts corresponding to the length of their characterised repeat unit and presented no difficulty in scoring. Two loci (notssr2 and notssr8) presented more continuous variation in allele size. For these two loci, fragments were binned according to the size classes presented in Appendix 2 (Tables A2.1 and A2.2).

We used the software Structure 2.3.3 (Pritchard et al. 2000) to conduct admixture analysis of the total dataset. Settings were as follows: length of burnin period 200,000, number of MCMC reps after burnin 800,000, use admixture model, infer alpha, initial value of alpha 1.0, separate alpha for each population, uniform prior for alpha, maximum value for alpha 10.0, SD of proposal for updating alpha 0.025, allele frequencies independent among populations, infer lambda, separate lambda for each population, starting value of lambda 1.0. The value of k, the number of populations, was estimated as 4 based on examination of mean likelihood values and Δk (Evanno et al. 2005) at k from 1 to 10 using 10 replicate runs with a burnin of 20,000 and 80,000 generations after burnin. Structure output was examined using the Structure Harvester website (Earl & vonHoldt 2012).

We also conducted a Structure analysis with k = 3 (which had the effect of considering F. cliffortioides and F. solandri as a single species) to identify hybrids between the serrated-leaved species or between either of the entire-leaved species and the serrated-leaved species. Individuals for which 90% probability intervals from Structure excluded pure ancestry were removed from the dataset and the remaining samples were grouped by locality and species for the purposes of estimating genetic distances among populations, using Powermarker 3.25 (Liu & Muse 2005). In the case of three plants coming from sites at which both F. fusca and F. truncata grow, identifications were altered after structure analysis prompted re-examination of leaf characters. Distances were estimated using the model of Nei (1972) and a UPGMA tree generated in Powermarker and drawn using MEGA 5.1 (Tamura et al. 2011).

A second Structure analysis was conducted for the F. cliffortioides and F. solandri samples, excluding those suggested by the k = 3 analysis to be hybrids with F. fusca or F. truncata (i.e. those for which 90% confidence intervals excluded pure ancestry from one species), using the same parameters as above except that k was set to 2 (as indicated by Δk) and allele frequencies were modelled as correlated between populations (Falush et al. 2003). The correlated allele frequencies model produced results in which individuals were generally inferred to be less admixed and likelihood and posterior probability scores for the data were marginally better than for the allele frequencies independent model. For each site we calculated the mean proportion of ancestry from each hypothetical population.

Correlation between Nei's (1972) population genetic distances and geographic distances among populations of F. solandri and F. cliffortioides were assessed using the Mantel test implemented in R.

Results

Summary statistics for each species are reported in Table 1. Across the 8 loci and 449 samples in our total dataset there were 86 missing observations (no alleles scored for a sample at a locus). Many of these were for locus notssr8 from F. fusca for which 40 of 91 samples did not produce a fragment. Only a single sample each of F. cliffortioides and F. truncata did not produce a detectable fragment for this locus. The other missing observations were distributed over all the species and loci. In general, observed heterozygosity values were lower than the expected heterozygosity. Locus notssr13 was monomorphic in F. truncata, whereas locus ncutas13 was monomorphic in F. fusca.

Table 1 Summary statistics for Fuscospora species.

Marker	Major allele frequency	No. genotypes	No. samples	No. observations	No. alleles	Avail.^1	HE	HO	PIC^2
F. cliffortioides
notssr6	0.43	12	140	140	6	1	0.68	0.65	0.62
notssr2	0.62	13	140	140	8	1	0.55	0.36	0.50
notssr22	0.90	4	140	136	3	0.97	0.18	0.14	0.17
notssr8	0.39	22	140	140	11	1	0.69	0.61	0.64
notssr18	0.41	18	140	139	8	0.99	0.72	0.62	0.68
BC9	0.24	42	140	138	13	0.96	0.86	0.83	0.85
notssr13	0.74	7	140	140	5	1	0.42	0.39	0.37
Ncutas13	0.92	2	140	139	2	0.99	0.15	0.16	0.14
Mean	0.58	15	140	139	7	0.99	0.53	0.47	0.49
F. fusca
notssr6	0.96	3	91	91	3	1	0.06	0.07	0.06
notssr2	0.43	14	91	91	7	1	0.69	0.60	0.64
notssr22	0.45	6	91	91	3	1	0.63	0.54	0.56
notssr8	0.86	5	91	51	4	0.56	0.26	0.25	0.23
notssr18	1.00	1	91	91	1	1	0.00	0.00	0.00
BC9	0.31	21	91	90	7	0.99	0.79	0.73	0.76
notssr13	0.64	3	91	91	2	1	0.46	0.48	0.36
ncutas13	1	1	91	89	1	0.98	0	0	0
Mean	0.71	6.75	91	85.63	3.50	0.94	0.36	0.34	0.33
F. solandri
notssr6	0.82	7	113	112	5	0.99	0.32	0.33	0.30
notssr2	0.86	7	113	110	5	0.97	0.26	0.23	0.25
notssr22	0.93	3	113	105	3	0.93	0.13	0.13	0.12
notssr8	0.65	14	113	112	8	0.99	0.53	0.46	0.48
notssr18	0.72	9	113	107	5	0.95	0.46	0.41	0.43
BC9	0.20	42	113	112	12	0.99	0.88	0.88	0.87
notssr13	0.64	14	113	110	5	0.97	0.56	0.52	0.53
Ncutas13	0.96	3	113	113	2	1	0.07	0.05	0.07
Mean	0.72	12.38	113	110.13	5.63	0.97	0.40	0.38	0.38
F. truncata
notssr6	0.36	15	85	84	6	0.99	0.75	0.73	0.71
notssr2	0.61	7	85	83	4	0.98	0.51	0.53	0.43
notssr22	0.57	4	85	84	3	0.99	0.49	0.49	0.38
notssr8	0.63	6	85	84	3	0.99	0.54	0.52	0.47
notssr18	0.52	9	85	85	4	1	0.60	0.51	0.53
BC9	0.61	21	85	84	9	0.99	0.59	0.46	0.57
notssr13	1	1	85	85	1	1	0	0	0
ncutas13	0.74	3	85	84	2	0.99	0.38	0.46	0.31
Mean	0.63	8.25	85	84.13	4	0.99	0.48	0.46	0.42

¹ Avail., availability (no. observations divided by no. samples).

² PIC, polymorphism information content (Botstein et al. 1980).

Analysis of the complete dataset using Structure and comparing the Δk statistic according to Evanno et al. (2005) suggested that four populations best explain the data. The ancestry plot for individual samples is shown in Fig. 1. The four groups generally correspond very well with the four named taxa (F. cliffortioides, F. fusca, F. solandri and F. truncata), but F. cliffortioides and F. solandri are less well separated and some populations appear to be a mixture of both species and possibly hybrids between them. The analysis suggested that the sampled populations of F. fusca and F. truncata are essentially ‘pure’, but a few individuals appeared to be hybrids either between these two species or with F. cliffortioides or F. solandri. Likewise, a few F. cliffortioides and F. solandri individuals appeared to be hybrids with F. fusca or F. truncata. The 90% confidence intervals for individual ancestry estimates were very broad (supplementary information) and therefore these estimates should be treated with considerable caution, particularly where small proportions of ancestry from one population are indicated. However, 90% probability intervals from Structure excluded pure ancestry from any one population for 23 samples (including those vouchers listed as hybrids in Table A1.1). With k = 3 this decreased to 20 samples.

Figure 1 Ancestry plot for structure analysis with k = 4. Blue, Fuscospora cliffortioides; black, F. solandri; red, F. fusca; yellow, F. truncata.

A UPGMA tree generated from Nei's (1972) genetic distances among populations grouped them into clusters corresponding with F. fusca, F. truncata and a group combining F. cliffortioides and F. solandri together (Fig. 2). Within this last group there is some separation into groups corresponding with the two species, but a few populations were misplaced compared with their field ID. The Piano Flat and Tom Creek F. cliffortioides populations cluster with F. solandri populations, whereas the Kidd's Bush and Heron F. cliffortioides populations cluster together separately from the remaining F. solandri and F. cliffortioides populations. Correlation between genetic and geographic distances for populations of the entire-leaved species was weak (Pearson's r = 0.1036) and non-significant (P = 0.134).

Figure 2 UPGMA dendrogram using Nei's (1972) genetic distance treating each collection site as a population.

Analysis of F. solandri and F. cliffortioides genotypes without F. fusca, F. truncata or hybrids with those two species (as suggested by the k = 3 Structure analysis) using k = 2 returned clusters more or less consistent with the two species but with a number of mixed populations and considerable admixture (not shown). Again, probability intervals for individual ancestry estimates were very broad. To summarise this analysis we have plotted the mean group 1 ancestry estimates for each locality (Fig. 3). In Fig. 3 populations are arranged from the most F. cliffortioides-like at the top to the most F. solandri-like at the bottom.

Figure 3 A, Mean ancestry estimates for Fuscospora solandri and F. cliffortioides populations from structure analysis of these species only, with k = 2. White, group 1 membership > 0.6 – F. cliffortioides. Grey, group 1 membership 0.40–0.60 – admixed. Black, group 1 membership < 0.40 – F. solandri. B, Map showing locations of populations included in the k = 2 Structure analysis (for site numbers see graph in A).

Discussion

Our small set of SSR markers has proved sufficient to distinguish populations of the species from each other, even when sympatric, with the exception of F. cliffortioides and F. solandri. Structure analyses of our data are also able to distinguish at least some hybrids among the taxa. This analysis provides the first published evidence we are aware of that F. truncata and F. fusca hybridise. Although it has long been expected that they would hybridise given the occurrence of hybrids between each of them and F. solandri or F. cliffortioides and between species in other genera of Nothofagaceae (Heenan & Smissen 2013 and references therein) the morphological similarity between F. fusca and F. truncata makes hybrids between them cryptic in the field.

With one exception, at least one allele could be amplified for each locus from the majority of samples (Table 1), indicating that null alleles are, at worst, rare for these loci in all species. The exception was notssr8, for which we could not amplify any fragment from 40 of 91 F. fusca samples. This indicates the presence of a high-frequency null allele or alleles at this locus in this species, as was suggested from a smaller sample by Smissen et al. (2012). Comparison of observed heterozygosity (H_O) and expected heterozygosity (H_E) in each species (Table 1) shows that the former is generally lower than the latter. This is to be expected because panmixis is an unreasonable assumption across the sampled range of each species. Our samples at individual sites are too small to allow meaningful comparison, treating each as discrete populations.

Ogden (1989) described the New Zealand species of Fuscospora (treated as Nothofagus subgenus Fuscospora) as a coenospecies, the taxa of which exchanged genes relatively freely so that different morphologies came and went with time and changes in selective pressures within a common gene pool. He also stressed the similarity between F. fusca and F. truncata writing ‘… the difference between these two species (degree of leaf serration, presence of domatia, tree size and wood properties) are hardly sufficient to warrant specific status, especially in view of the observation that, where ever they grow together, intermediates are frequent. They differ much less in general appearance than do the extreme forms of the N[othofagus]. solandri cline.’

Our results do not support this view. Rather, both admixture analysis using Structure (Fig. 1) and clustering of population genetic distances (Fig. 2) show that the entire-leaved species are well differentiated from the serrated-leaved F. fusca and F. truncata, which are also well differentiated genetically from each other. Although hybrids do occur when they grow together, there is no evidence in our data for gene flow between the two serrated-leaved species, or between either of them and the entire-leaved species extending beyond contact zones. In most cases, the plants indicated as having mixed ancestry from entire- and serrated-leaved species or from both serrated-leaved species are from populations where they co-occur. An exception is the Kidd's Bush population of entire-leaved beech. Although there is no F. fusca currently at this exact site, it does occur in the general area (Wardle 1984, p. 103), so this is not a surprising result. Similarly, although we did not collect either of the entire-leaved species from Ohura Road, entire-leaved beech is present in the area and since several samples from this site were taken from seedlings without careful assessment of their morphology, it is not surprising that some hybridisation is suggested by the Structure analysis. Conversely, the 90% probability intervals from our Structure analysis suggests that one entire-leaved beech sample from Lewis Pass has some proportion of its ancestry from F. truncata despite the absence of that species from that area. Since F. truncata is not (to our knowledge) found at or near this site, the results must be rejected as false or attributed to hybridisation following long-distance pollen dispersal. We favour the former explanation because there is no aspect of the leaf morphology of these samples that suggests a hybrid origin, but cannot exclude the latter entirely.

Although F. truncata and F. fusca are morphologically similar in sharing leaves larger than those of F. solandri and in having leaves with serrate margins, our data do not support the contention that they are especially closely related. Indeed, comparison among the genera of Nothofagaceae would suggest these character states are symplesiomorphic. A study of genetic relationships among New Zealand Nothofagaceae, using isozyme analysis, also cast doubt on a particularly close relationship between F. fusca and F. truncata (Haase 1994).

However, F. cliffortioides and F. solandri show much less differentiation in our data (Figs 1 and 2). The Structure analysis in particular suggests that some populations can be interpreted as pure or nearly pure F. solandri (black in Fig. 1) or F. cliffortioides (blue in Fig. 1), whereas others appear to be a mixture of both. Even if these populations are accepted to be admixed, because confidence intervals for individual ancestry estimates are so broad (supplementary data), it is not clear whether these mixed populations are each composed of relatively uniform hybrid derivatives, or complex hybrid swarms, and individual assignments are probably not accurate. To allow for a simpler comparison of ancestry at a population level Fig. 3 displays the mean ancestry estimates at each site from the F. cliffortioides and F. solandri structure analysis with k = 2. These results are largely in accord with morphological identifications of the plants, with F. solandri populations mostly having a mean proportion of group 1 membership < 0.4 and F. cliffortioides populations > 0.6. Plants from Piano Flat were collected as F. cliffortioides, and are described as such by the New Zealand Department of Conservation (undated), while Structure analysis of our data yielded a mean group 1 membership of 0.53. Leaves from trees at this site are variable; some are consistent with F. solandri and this location is in an area where Wardle (1984) indicates that forms intermediate between the taxa ‘outside the range of black beech proper’ are present.

Some F. solandri populations from the Canterbury foothills (Alford Reserve, Ashley River, Awahou Reserve, Lees Valley, Mt Thomas and Sharplin Falls in our sampling) have been considered morphologically intermediate between that species and F. cliffortioides (Wardle 1984). Our data suggest they are genetically admixed (Fig. 3, 0.40–0.58 proportion group 1), consistent with a hybrid origin or mixed composition for these populations. The Mt Grey population is geographically proximate to these, but has been considered to represent more typical F. solandri (Wardle 1984), and our data are consistent with this (0.32 proportion group 1). Fuscospora solandri was extensively logged in Canterbury in the late 19th and early 20th centuries (Wardle 1984, p. 331) resulting in its relative rarity today. Presumably the higher altitude admixed populations sampled in this study were both less desirable as timber and less accessible to commercial logging. After Piano Flat, the Tongariro population has the lowest proportion of group 1 membership (0.62) for plants collected as F. cliffortioides, despite samples from there having uniformly the leaf morphology of that species. Fuscospora solandri is also present in the central North Island so it is possible that some genetic influence of this species is also present in our Tongariro samples, which were collected from the lower edge of a forest (at c. 900 m) in an area disturbed by human activity. More sampling of populations of both species from this area will be necessary to determine if ‘pure’ F. cliffortioides is present and the nature of any admixture.

A further complication in assessing morphological variation within the entire-leaved species is that both Poole (1958) and Wardle (1984) suggest that introgressive hybridisation with F. fusca might have played a role. Our data do not provide support for this hypothesis (except in confirming that hybridisation occurs), but it remains highly plausible.

The UPGMA tree generated for population genetic distances also showed some clustering into groups corresponding with F. cliffortioides and F. solandri. Where clustering deviates from taxonomic assignment some explanations can be advanced. The Kidd's Bush population included several samples that the Structure analysis suggested are hybrids with F. fusca, but we did not exclude them because 90% probability intervals encompassed pure F. cliffortioides ancestry. The inclusion of probable hybrids is sufficient to explain the position of this population in the tree. Indeed, the relationships of populations within F. truncata indicated by the UPGMA analysis also appear to be influenced by the inclusion of hybrids in some populations. In this case, the McFarlane Mound and Ohura Road populations appear in a basal position within the species. The Piano Flat population is discussed above. The Tom Creek population was identified as F. cliffortioides when collected and appears to be predominantly this species in Structure analyses (Figs 1 and 3), but clusters with populations of F. solandri in the UPGMA tree: reappraisal of herbarium specimens (Table A1.1) suggests some leaves are rather rounded, approaching F. solandri. Like the Piano Flat population, the Tom Creek population may be of admixed origin and it occurs in the southern South Island in the area where Wardle (1984) reports intermediate forms. This site is also one at which hybridisation with F. fusca has taken place and this may be affecting both the genetic analysis and the leaf morphology. The remaining deviation is the Heron F. cliffortioides population, which appears as the most cliffortioides-like in the structure analysis but apart from the other populations of the species in the UPGMA tree. Unlike the other three cases, no immediate explanation for this discrepancy is apparent.

Although the ranges of F. solandri and F. cliffortioides overlap, the former is more common in the north and the latter more common in the south. However, genetic distances among populations of the entire-leaved species treated together are poorly predicted by geographic distance, suggesting that isolation by distance does not account for the genetic structure detected among them.

In reducing F. cliffortioides to a variety of F. solandri (both as Nothofagus), Poole (1958) stressed that many trees and whole populations could not be assigned with confidence to one or the other variety, but that there are important ecological and timber property differences between the ‘extreme’ ends of what he perceived as a more or less continuous spectrum of character variation. This he attributed to ‘geographical or clinal’ variation. Here, we interpret this view as hypothesising a common gene pool, out of which ‘extreme’ morphological forms have arisen through natural selection acting locally on populations and metapopulations to independently produce the forms represented by typical mountain (F. cliffortioides) or black beech (F. solandri).

In our view, Poole's position was sound at the time it was published, and his interpretation was quite sufficient to explain the morphological variation he documented. Genetic data consistent with this sort of geographic variation within a species have been reported for a few New Zealand plants (e.g. Helichrysum lanceolatum, Smissen et al. 2006; Cordyline australis, Beever et al. 2013) and plants from other regions (e.g. Helichrysum italicum, Galbany-Casals et al. 2011) and is understood from a theoretical population genetic perspective (Latta 2003). However, the data we report here are more consistent with a scenario in which the two forms diverged from each other in allopatry but are now in secondary contact and undergoing hybridisation and probably some introgression. This is illustrated by the greater similarity shown between some geographically distant but morphologically similar populations (e.g. Mangateitei and Mt Grey) than between geographically proximate but morphologically dissimilar populations (e.g. Mangateitei and Tongariro). Under a scenario where the morphological differences between mountain and black beech were maintained by habitat selection, neutral genetic markers such as those examined in this study would be more similar between geographically proximate populations than between geographically distant populations. However, other scenarios might be generated and the data currently available do not allow a conclusive test. Because F. solandri and F. cliffortioides are maintained in apparently pure form, at least over much of their area of sympatry, we suggest that they are reproductively isolated and appropriate for recognition at species rank under the biological species concept (Mayr 1942) if not applied too strictly so as to imply whole-genome reproductive isolation (see Rieseberg et al. 2004). Invocation of any particular reproductive isolating mechanism would be speculative at this time but phenological differences are noted by Wardle (1984, p. 254, citing unpublished data of A. L. Poole). Although further work is required to clarify the boundaries between these two species and the extent of gene flow between them, our results are consistent with their continued recognition. Future work should include additional samples of both species and additional genetic markers.

The practical issue highlighted by Poole, the problem of assigning trees or populations of intermediate morphology to one or the other taxon, remains an issue regardless of the rank at which the taxa are named. Moreover, this problem is far from restricted to the taxa at issue here. New Zealand examples of species which hybridise often, and which form self-sustaining hybrid swarms or introgressed populations, include Phormium cookianum and P. tenax (Smissen et al. 2008), and Pseudopanax lessonii and P. crassifolius (Morgan-Richards et al. 2009).

We suggest that the two entire-leaved southern beech species should be treated pragmatically such that populations and individuals should be identified with one or the other species even if there is some evidence for gene flow from the other species (or indeed from F. fusca or F. truncata). However, some populations and individuals are probably better treated as hybrids, using the hybrid formula F. cliffortioides × F. solandri. Most populations from the Canterbury foothills are probably best referred to F. solandri, as proposed by Wardle (1984) despite our data suggesting some component of the F. cliffortioides genome is present. None of the populations we sampled from the south of the South Island appear to be pure black beech (F. solandri), but the Piano Flat population may represent what Wardle (1984) called ‘intermediate forms between black beech and N. solandri var. cliffortioides (mountain beech)’ and might best be treated as F. cliffortioides × F. solandri. Further work will be required to establish whether and to what extent pure F. solandri is present in this part of the country. In the North Island, where the relative abundances of the two species are reversed, work is also required to examine the extent of true F. cliffortioides and how much admixture with F. solandri has taken place. In particular, further sampling of F. cliffortioides, including from higher altitudes should be undertaken.

In summary, our SSR data readily distinguish the serrated-leaved New Zealand Fuscospora species F. fusca and F. truncata, but suggest considerable admixture between the entire-leaved F. cliffortioides and F. solandri. However, our results also suggest that variation within entire-leaved beech is better reflected by the recognition of distinct species, with morphological intermediates interpreted as the result of secondary contact and hybridisation than as a single variable species. Although with the number of markers currently available ancestry estimates for individual plants lack the precision necessary to distinguish between F₁ and later-generation hybrids, the latter seem likely to occur at least rarely between most combinations of species. Although hybrids are common, gene flow among the species appears to be limited and localised (except perhaps between F. cliffortioides and F. solandri), but again the sampling of more loci and populations will be required to rigorously test this.

Supplementary file

Supplementary file: Structure results file with ancestry estimates for individual plants.

Acknowledgements

This work was funded by the New Zealand Ministry of Business, Innovation and Employment (Capability Fund). Rowan Buxton and Jen Hurst contributed to sample collection. Caroline Mitchell and Dagmar Goeke (Landcare Research) provided technical support. Matt McGlone (Landcare Research) provided stimulating discussion. Peter Haase, Gary Houliston (Landcare Research), Leon Perrie (Museum of New Zealand Te Papa Tongarewa) and two anonymous reviewers provided constructive comments on a draft manuscript and it was edited by Christine Bezar (Landcare Research).

Appendix 1 Sample sites and vouchers

Table A1.1 Fuscospora sample locations and voucher details.

	F. fusca	F. cliffortioides	F. solandri	F. truncata	Latitude	Longitude	Voucher
Alford Reserve	No	No	Yes	No	43°35′14.20″S	171°30′27.55″E	No voucher
Arthur's Pass	No	Yes	No	No	42°55′12.70″S	171°33′33.49″E	Morse, Richardson, Buxton s. n. 20/03/10 (CHR 635344)
Ashley River	No	No	Yes	No	43°13′47.39″S	172°13′24.37″E	RD Smissen 600 (CHR 623679)
Awaawa Reserve	No	No	Yes	No	43°33′38.15″S	171°31′59.04″E	Unvouchered
Awahou Reserve	No	No	Yes	No	39°09′24.82″S	174°41′40.41″E	RD Smissen 670 (CHR 623686)
Borland Lodge	Yes	No	No	No	45°46′36.41″S	167°32′15.62″E	Unvouchered
Bullocks	No	No	No	Yes	42°06′16.77″S	171°24′01.06″E	Morse, Richardson, Buxton s. n. 18/04/10 (CHR 635307, CHR 635308)
Charleston	Yes	No	No	Yes	41°58′17.51″S	171°25′53.40″E	Morse, Richardson, Buxton s. n. 18/04/10 (CHR 635302, CHR 635303 F. truncata)
Charleston Ditch	No	Yes	No	No	41°58′13.80″S	171°25′54.53″E	Unvouchered
Dart	Yes	No	No	No	44°33′36.69″S	168°22′08.40″E	SJ Richardson & K boot s.n. 25/06/10 (CHR 635350, CHR 635351)
Denniston	No	Yes	No	No	41°17′46.05″S	173°34′13.21″E	Unvouchered
Eglinton	Yes	Yes	No	No	44°53′38.25″S	168°05′18.24″E	Morse, Richardson, Buxton s. n. 05/03/10 (CHR 635336 F. cliffortioides); Morse, Richardson, Buxton s. n. 05/03/10 (CHR 635338, CHR 635339 F. fusca)
Gladstone	No	No	Yes	No	41°07′24.32″S	175°47′03.34″E	RD Smissen 622 (CHR 623690)
Granville	No	No	No	Yes	42°19′23.26″S	171°40′31.59″E	Morse, Richardson, Buxton s. n. 17/04/10 (CHR 635310)
Green Lake	No	Yes	No	No	45°44′25.99″S	167°22′17.44″E	Unvouchered
Hanlon	No	No	No	Yes	41°25′26.47″S	172°06′09.75″E	Morse, Richardson, Buxton s. n. 20/04/10 (CHR 635318)
Heron	No	Yes	No	No	43°19′54.04″S	171°07′19.52″E	Morse, Richardson, Buxton s. n. 10/03/10 (CHR 6353465, CHR 635346)
Huxley	No	Yes	No	No	43°59′47.26″S	169°48′17.89″E	Morse, Richardson, Buxton s. n. 09/03/10 (CHR 635349)
Karamea Bluff	Yes	No	No	Yes	41°30′47.53″S	172°01′13.31″E	Morse, Richardson, Buxton s. n. 19/04/10 (CHR 635315, CHR 635316, CHR 635317 F. truncata); Morse, Richardson, Buxton s. n. 19/04/10 (CHR 635313, CHR 635314 F. fusca)
Kidd's Bush	No	Yes	No	No	44°26′32.72″S	169°16′12.98″E	Morse, Richardson, Buxton s. n. 04/03/10 (CHR 635343)
Lees Valley Road	No	No	Yes	No	43°14′20.94″S	172°12′16.12″E	RD Smissen 598 (CHR 623677), RD Smissen 599 (CHR 623680)
Lewis Pass	No	Yes	No	No	42°22′45.03″S	172°23′54.64″E	Morse, Richardson, Buxton s. n. 16/04/10 (CHR 635301)
Macfarlane Mound	No	Yes	No	Yes	44°01′58.77″S	168°43′55.71″E	Morse, Richardson, Buxton s. n. 03/03/10 (CHR 635334 F. cliffortioides); Morse, Richardson, Buxton s. n. 03/03/10 (CHR 635335, CHR 635356, CHR 635362 possible F. cliffortioides × F. truncata); Morse, Richardson, Buxton s. n. 03/03/10 (CHR 635332, CHR 635333 F. truncata)
Mangapae	Yes	No	Yes	No	38°37′57.31″S	176°52′06.78″E	Unvouchered
Mangateitei	No	No	Yes	No	39°25′08.00″S	175°27′04.14″E	RD Smissen 652 (CHR 623687)
Mirza	No	No	Yes	No	41°52′05.79″S	174°06′17.17″E	Morse, Richardson, Buxton s. n. 23/04/10 (CHR 635294, CHR 635295, CHR 635296, CHR 635297, CHR 635298)
Mt Grey	Yes	No	Yes	No	43°08′17.03″S	172°31′29.29″E	Unvouchered
Mt Thomas	No	No	Yes	No	43°13′48.08″S	172°15′08.11″E	RD Smissen 609 (CHR 623678), RD Smissen 610 (CHR 623681), RD Smissen 608 (CHR 623682), RD Smissen 606 (CHR 623683), RD Smissen 607 (CHR 623684)
Ohura Road	No	No	No	Yes	38°59′17.99″S	174°49′12.51″E	Unvouchered
Oparara	Yes	No	No	No	41°09′11.34″S	172°11′58.14″E	Morse, Richardson, Buxton s. n. 20/04/10 (CHR 635304, CHR 635305)
Otira	Yes	No	No	No	42°47′31.47″S	171°36′24.61″E	SJ Richardson s. n. 30/07/10 (CHR 635354, CHR 635352, CHR 635353)
Peasoup	Yes	Yes	No	Yes	41°54′05.47″S	172°14′49.86″E	Morse, Richardson, Buxton s. n. 21/04/10 (CHR 635300 (F. fusca); Morse, Richardson, Buxton s. n. 21/04/10 (CHR 635322, CHR 635323 F. truncata)
Pelorus	No	No	Yes	No	41°17′46.05″S	173°34′13.21″E	Morse, Richardson, Buxton s. n. 23/04/10 (CHR 635299)
Piano Flat	Yes	Yes	No	No	45°33′03.72″S	169°01′41.21″E	Morse, Richardson, Buxton s. n. 07/03/10 (CHR 635342 F. cliffortioides), Morse, Richardson, Buxton s. n. 07/0310 (CHR 635340, CHR 635341 F. fusca)
Pohangina	No	No	Yes	No	40°07′54.83″S	175°51′02.90″E	RD Smissen 632 (CHR 623689)
Ronga	No	No	No	Yes	41°08′06.22″S	173°36′00.32″E	Morse, Richardson, Buxton s. n. 21/04/10 (CHR 635319)
Sharplin Falls	No	No	Yes	No	43°37′52.02″S	171°25′11.33″E	Unvouchered
Staircase	Yes	No	No	No	42°19′02.15″S	171°53′35.53″E	Morse, Richardson, Buxton s. n. 16/04/10 (CHR 635306)
Te Marua	No	No	Yes	No	41°05′13.92″S	175°09′01.25″E	RD Smissen 617 (CHR 623691)
Tom Creek	Yes	Yes	No	No	45°56′45.50″S	169°28′26.22″E	Morse, Richardson, Buxton s. n. 08/03/10 (CHR 635324, CHR 635325, F. cliffortioides); Morse, Richardson, Buxton s. n. 08/03/10 (CHR 635365, CHR 635367, CHR 635368 F. cliffortioides × F. fusca); Morse, Richardson, Buxton s. n. 08/03/10 (CHR 635326, CHR 635327 F. fusca)
Tongariro River	No	Yes	No	No	39°13′53.66″S	175°46′35.14″E	RD Smissen 642 CHR 623688
Waitotara	No	No	Yes	No	39°45′48.68″S	174°48′39.88″E	RD Smissen 673 CHR 623685
Woolshed Hill	Yes	Yes	No	No	42°59′12.42″S	172°00′01.87″E	Morse, Richardson, Buxton s. n. 01/03/10 (CHR 635328, CHR 635329 F. cliffortioides); Morse, Richardson, Buxton s. n. 01/03/10 (CHR 635331 F. fusca)

Appendix 2 Allele definitions

Table A2.1 Allele definitions – notssr2.

Nominal allele size	Observed size range
212	212.1–212.8
215	214.7
222	221.6–222.3
223	223.1
224	223.9–224.4
225	224.7–225.6
226	225.9–226.2
228	227.9–228.5
231	231.0–231.9
234	234.2–234.8
238	237.4–238.1

Table A2.2 Allele definitions – notssr8.

Nominal allele size	Observed size range
210	209.5–209.7
213	212.6–213.1
216	215.3–216.1
218	217.9–218.6
219	218.9–220.2
221	221.1–221.4
222	222.0–223.6
225	225.4–225.8
229	228.5–229
232	232.1
235	234.9–235.4
241	240.9