Molecular Ecology (2010) 19, 1675–1690
doi: 10.1111/j.1365-294X.2010.04585.x
Hybridization and polyploidy as drivers of continuing
evolution and speciation in Sorbus
ASHLEY ROBERTSON,* TIMOTHY C. G. RICH,† ALEXANDRA M. ALLEN,* LIBBY HOUSTON,*
C A T R O B E R T S , * J O N R . B R I D L E , * S T E P H E N A . H A R R I S ‡ and S I M O N J . H I S C O C K *
*School of Biological Sciences, University of Bristol, Woodland Road, Bristol, BS8 1UG, UK, †Department of Biodiversity &
Systematic Biology, National Museum of Wales, Cathays Park, Cardiff, CF10 3NP, UK, ‡Department of Plant Sciences,
University of Oxford, South Parks Road, Oxford, OX1 3RB, UK
Abstract
Interspecific hybridization and polyploidy are pivotal processes in plant evolution and
speciation. The fate of new hybrid and polyploid taxa is determined by their ability to
reproduce either sexually or asexually. Hybrids and allopolyploids with odd chromosome numbers are frequently sterile but some establish themselves through asexual
reproduction (vegetative or apomixis). This allows novel genotypes to become established by isolating them from gene flow and leads to complex patterns of variation. The
genus Sorbus is a good example of taxonomic complexity arising from the combined
effects of hybridization, polyploidy and apomixis. The Avon Gorge in South-west Britain
contains the greatest diversity of Sorbus in Europe, with three endemic species and four
putative endemic novel hybrids among its 15 native Sorbus taxa. We used a combination
of nuclear microsatellite and chloroplast DNA markers to investigate the evolutionary
relationships among these Sorbus taxa within the Avon Gorge. We confirm the genetic
identity of putative novel taxa and show that hybridization involving sexual diploid
species, primarily S. aria and S. torminalis and polyploid facultative apomictic species
from subgenus Aria, has been responsible for generating this biodiversity. Importantly
our data show that this creative evolutionary process is ongoing within the Avon Gorge.
Conservation strategies for the rare endemic Sorbus taxa should therefore consider all
Sorbus taxa within the Gorge and must strive to preserve this evolutionary process rather
than simply the individual rare taxa that it produces.
Keywords: apomixis, biodiversity, conservation, hybridization, polyploidy, self-incompatibility,
Sorbus
Received 9 July 2009; revision received 25 January 2010; accepted 29 January 2010
Introduction
Interspecific hybridization and genome multiplication
(polyploidy) are now accepted as pivotal processes in
plant diversification and speciation, both separately and
more commonly in combination, through the formation
of allopolyploids (Stebbins 1950; Grant 1981; Arnold
1997; Rieseberg & Willis 2007). Indeed, hybridization
and polyploidy offer the clearest routes to sympatric
speciation in plants. Many examples of plant speciation
through hybridization and ⁄ or polyploidy have now
Correspondence: Simon J. Hiscock, Fax: +44 (0) 117 3317985;
E-mail: simon.hiscock@bristol.ac.uk
2010 Blackwell Publishing Ltd
been described (see: Arnold 1997; Rieseberg 1997; Soltis
& Soltis 1999; Abbott 2003; Rieseberg & Willis 2007).
Some (e.g. Helianthus, Arabidopsis, Brassica, Tragopogon,
Spartina and Senecio) have proved to be valuable model
systems for investigating genetic and genomic processes
involved in plant speciation (Adams & Wendel 2005;
Hegarty & Hiscock 2005, 2008; Salmon et al. 2005; Chen
2007).
The breeding systems associated with plants involved
in hybridization and polyploidy play a critical role in
determining the likelihood of hybrid and ⁄ or polyploid
formation and in determining the likelihood of their
long-term survival (Grant 1981; Rieseberg 1997; Hörandl
2006). In sexually reproducing plants, outbreeding often
1676 A . R O B E R T S O N E T A L .
encourages hybridization and the formation of new homoploid and allopolyploid offspring, while inbreeding
(selfing) will perpetuate new homoploids and allopolyploids and reinforce reproductive isolation between them
and their parental taxa (Grant 1981; Rieseberg 1997;
Coyne & Orr 2004). Asexual reproduction, through
apomixis (hereafter synonymous with agamospermy; the
production of ‘maternal’ clonal seeds), offers a more
secure means of perpetuating new hybrid and allopolyploid genotypes than selfing and affords immediate,
often total, reproductive isolation from parental taxa. Not
surprisingly therefore, apomixis is frequently associated
with hybridization and polyploidy where it has been critical for the establishment of many novel taxa arising by
these processes (Grant 1981; Briggs & Walters 1997;
Hörandl 2006; Hörandl & Paun 2007; Whitton et al.
2008). The association between polyploidy and apomixis
has been widely noted and it has been suggested that
polyploidy either acts as a stimulus for apomixis or a prerequisite for its maintenance or both (Grimanelli et al.
2001; Whitton et al. 2008). Because apomixis essentially
‘freezes’ the new hybrid or polyploid genotype, it can
facilitate the emergence of highly complex patterns of
phenotypic diversity when hybridization and polyploidy
occur recurrently among predominantly sexual parental
taxa (Hörandl 2006; Hörandl & Paun 2007; Whitton et al.
2008). The result is taxonomic complexity in genera
where apomixis is associated with hybridization and
polyploidy, for example, Alchemilla (Sepp et al. 2000),
Hieracium (Bicknell et al. 2000; Fehrer et al. 2007), Arabis
(Koch et al. 2003), Rosa (Nybom et al. 2004), Craetegus (Lo
et al. 2009), and Sorbus (Nelson-Jones et al. 2002; Robertson et al. 2004a, b). Recent reviews have considered the
role of hybridization and polyploidy in divergence and
speciation within sexually reproducing plants (Rieseberg
& Willis 2007; Hegarty & Hiscock 2008), yet these processes in asexual (agamic) complexes, where hybrids
and ⁄ or polyploids reproduce via obligate or facultative
apomixis, have been less well studied (but see: Richards
2003; Hörandl 2006; Hörandl & Paun 2007).
In the genus Sorbus (whitebeams, service trees, and
rowans), hybridization, allopolyploidy, autopolyploidy
and apomixis (both obligate and facultative) have contributed to the complex patterns of variation seen
among the ‘intermediate’ European taxa (Liljefors 1953,
1955; Warburg 1962; Challice & Kovanda 1978; Proctor
& Groenhof 1992; Rich & Jermy 1998; Nelson-Jones
et al. 2002; Robertson et al. 2004a, b; Robertson & Sydes
2006; Chester et al. 2007). Embryological studies have
shown that apomixis in Sorbus is gametophytic apospory with development of the embryo dependent upon
pollination to initiate an endosperm (pseudogamy)
(Liljefors 1953, 1955; Robertson et al. 2004b). Interest in
the genus Sorbus lies with both the vulnerability of the
taxa per se and the evolutionary processes that have
generated these taxa (Hollingsworth 2003; Ennos et al.
2005). Taxonomically complex groups, such as Sorbus,
pose important evolutionary questions about the interactions between hybridization, genome multiplication
and breeding systems among parental and derivative
taxa. Furthermore, generation of taxonomic novelty has
direct consequences for taxon conservation.
In Europe, there are three major centres (‘hot-spots’) of
on-going Sorbus diversification and speciation: Scandinavia, South-east Europe and Britain (Hedlund 1901; Kovanda 1961; Warburg & Kárpáti 1968). In Britain, Sorbus is
represented by 39 named taxa and at least five unnamed
taxa (Stace 1997; Rich & Jermy 1998; Rich & Proctor 2009;
Rich et al. 2009). Of the 19 species (i.e. excluding hybrids
and recently described taxa) listed in the British Red List
(Cheffings & Farrell 2005), three are recognized as
‘critically endangered’, four as ‘endangered’ and four as
‘vulnerable’ according to IUCN (2001) criteria, making
Sorbus a priority genus for conservation in the UK.
The Avon Gorge in South-west Britain (Bristol) is the
richest and most important site for Sorbus diversity in
Europe. There are four sexual diploid species [Sorbus
aucuparia (rowan), S. aria (whitebeam), S. torminalis
(wild service-tree) and S. domestica (service tree); the latter is not involved in hybridization and is not discussed
further], one primary diploid hybrid (S. · thuringiaca)
and seven polyploid taxa (S. anglica, S. bristoliensis,
S. eminens, S. leighensis, S. porrigentiformis, S. whiteana,
and S. wilmottiana). The sexual species are outcrossing
and self-incompatible (Raspé & Kohn 2007; Robertson
et al. unpublished) whereas the polyploids are believed
to be apomictic. In addition to these native Sorbus taxa,
there are five introduced species (S. croceocarpa, S. decipiens, S. intermedia, S. latifolia and S. glabruiscula) all of
which are polyploid and believed to be apomictic.
There are also four taxa that show unusual ‘intermediate’ morphologies which have been described recently
as hybrids (Rich et al. 2009); these are S. · avonensis
(= S. aria · S. porrigentiformis), S. · houstoniae (= S. aria ·
S. bristoliensis), S. · proctoriana (= S. aucuparia · S. scalaris) and S. · robertsonii (= S. aria · S. eminens). Existing
hypotheses to account for the origin of the Avon Gorge
polyploid Sorbus taxa are based primarily on morphological, cytological and peroxidase isozyme studies
(Warburg 1962; Richards 1975; McAllister 1986; Proctor
et al. 1989; Proctor & Groenhof 1992; Rich & Jermy
1998; Fig. 1). Molecular evidence supporting these
hypotheses has come from two studies investigating
relationships among all UK Sorbus taxa. Nelson-Jones
et al. (2002) used restriction fragment length polymorphism (RFLP) variation in nuclear and mitochondrial
DNA to ascribe the majority of the polyploid ‘intermediate’ Sorbus taxa to one of three major aggregates:
2010 Blackwell Publishing Ltd
EVOLUTION AND SPECIATION IN SORBUS 1677
Nothosubgenus
Tormaria
Subgenus Aria
S. aria
Subgenus
Torminaria
S. bristoliensis
2n = 3x = 51 AAT
S. decipiens
S. torminalis
2n = 4x = 68
2n = 2x = 34 TT
S. croceocarpa
2n = 4x = 68
S. latifolia
2n = 4x = 68
AATT
2n = 2x = 34 AA
Nothosubgenus
Soraria
S. porrigentiformis
Subgenus
Sorbus
2n = 4x = 68 AAAA
S. anglica
S. eminens
2n = 4x = 68 AABB
S. aucuparia
2n = 4x = 68 AAAA
S. x thuringiaca
2n = 2x = 34 BB
S. whiteana
2n = 2x = 34 AB
2n = 3x = 51 AAA
S. wilmottiana
2n = 3x = 51 AAA
S. intermedia
2n = 4x = 68 AABT
Fig. 1 The current hypotheses for the origins of Sorbus taxa in the Avon Gorge, according to Rich & Jeremy (1998), and Nelson-Jones
et al. (2002) excluding recently described taxa. Members of the nothosubgenus Tormaria and nothosubgenus Soraria are thought to
represent allopolyploids, with Tormaria taxa having genomes inherited from both S. torminalis and a member of the subgenus Aria
and Soraria taxa having genomes inherited from both S. aucuparia and a member of subgenus Aria. Each member of subgenus Aria is
thought to comprise solely of genomes inherited directly or indirectly from S. aria. The proposed genome composition (Liljefors 1955;
Richards 1975) is indicated with capital letters: A: subgenus Aria genome, B: S. aucuparia genome, T: S. torminalis genome. For chromosome count references see Bailey et al. (2008).
subgenus Aria (formerly ‘Aria group’), nothosubgenus
Soraria (formerly ‘Anglica group’), and nothosubgenus
Tormaria (formerly ‘Latifolia group’), derived from the
diploid female parents S. aria, S. aucuparia and S. torminalis, respectively. More recently, Chester et al. (2007)
used plastid DNA microsatellite markers to determine
the female parentage of endemic UK Sorbus taxa and
produced results consistent with those of Nelson-Jones
et al. (2002). Neither of these studies included all of the
Avon Gorge Sorbus taxa and neither could ascertain the
paternal parents of hybrid taxa.
Here we describe an extensive molecular analysis of
227 individuals from 18 Sorbus taxa in the Avon Gorge
and from two S. rupicola individuals using seven
nuclear microsatellite markers and a chloroplast
(cp)DNA marker. For each of the 15 intermediate (polyploid ⁄ putative apomictic) Sorbus taxa co-existing in the
Avon Gorge we have assessed: (i) genetic diversity and
whether reproduction is sexual or apomictic (obligate or
facultative); (ii) the number of independent origins;
(iii) maternal and paternal affinity; and (iv) evidence
for backcrossing. This has allowed us to formulate an
extensive picture of the evolutionary relationships
between the Avon Gorge Sorbus taxa and to assess the
extent of ongoing evolutionary processes occurring
within and among the Sorbus taxa that comprise this
unique agamic complex.
material. Ten native species, four introduced species
and four intermediate putative hybrids were sampled
(see Table 2). Species identifications were based on
morphological characteristics of leaves and berries,
previously identified as robust diagnostic characters
defining the different Sorbus species (Rich & Jermy
1998).
Where possible leaf buds were collected from
between 20 and 30 individuals for each taxon. The only
material collected from outside of the Avon Gorge was
from two S. rupicola trees, one from the Wye Valley
and the other from South Wales. Sorbus rupicola was
included in this study as it is the only polyploid Sorbus
that is widespread in Northern Europe and it has been
previously demonstrated to be involved in the origin
of other intermediate Sorbus taxa (Robertson et al.
2004a, b). Even though S. rupicola is not currently
present in the Avon Gorge, it may have been present
post-glacially because populations can be found within
20 km of the Avon Gorge. Herbarium voucher material
for each sampled individual was deposited at the
National Museum of Wales (NMW), Cardiff. Samples
were collected in a non-random method (selectively
sampled from each patch in the Gorge) in order to
maximize the representation of potential genotypes
(Bayer 1990). GPS readings of tree locations and
herbarium voucher details are included in Table S1
(Supporting Information).
Materials and methods
DNA extraction
Plant material
Field surveys of the Avon Gorge were undertaken
during autumn 2004 to identify and collect Sorbus
2010 Blackwell Publishing Ltd
DNA was extracted from frozen leaf buds using the
Qiagen DNAeasy plant kit following the manufacturer’s
instructions.
1678 A . R O B E R T S O N E T A L .
Chloroplast DNA markers
A single non-coding region of the Sorbus chloroplast
genome, trnH–psbA was amplified using primers
designed by Hamilton (1998). Polymerase chain reaction
(PCR) mixtures (20 lL total volume) contained: one unit
of Hotstart Taq DNA polymerase (Qiagen), 2 lL of 10·
reaction buffer (Qiagen), 2 mM of each dNTP (Bioline),
with 0.2 lM of each forward and reverse primer and
20–50 ng of genomic DNA. PCR reactions were performed in an MJ-Research PTC-200 Peltier Thermal
Cycler using the following programme: 1 step at 95 C
for 15 min followed by 35 cycles of 1 min at 94 C,
1 min at 55 C, 1 min at 72 C and a 10 min extension
step of 72 C. PCR products from 12 of the Avon Gorge
taxa were purified using QIAquick spin columns (Qiagen). These purified PCR products were sequenced
using the Big Dye terminator cycle sequence kit (ABI)
and an ABI 3730xl sequencer (Perkin-Elmer). Sequences
were analysed using ChromasPro sequence analysis
software (Technelysium Pty Ltd).
For RFLP analysis of PCR products, digestions were
conducted with the restriction enzyme Tsp5901 (New
England Biolabs), according to the manufacturer’s
instructions. Digestion products were separated and
analysed by horizontal gel electrophoresis using 3.0%
Metaphor agarose gels. Gels were prepared in 1 · TBE
buffer. Running times were varied to ensure that the
DNA products migrated three-quarters of the way
down the gel. Gels were stained with ethidium bromide
for 15 min and de-stained for 15 min before visualization under UV light. Gels were photographed with the
Bio Image system to retain a permanent record.
Nuclear microsatellite markers
Seven nuclear microsatellite loci were amplified using
three sets of primers developed for Malus · domestica
(Gianfranceschi et al. 1998) and four pairs developed
for Sorbus torminalis (Oddou-Muratorio et al. 2001). Four
of these primer-pairs were redesigned, from sequences
deposited in Genbank, to facilitate amplification of a
wide range of Sorbus taxa (Table 1). Five of the forward
primers were labelled with fluorescent dyes. Reverse
primers were tailed with a non-target-specific sequence
to overcome allele sizing problems associated with the
primer sequence-dependant adenylation of PCR products by most Taq polymerases (Brownstein et al. 1996).
PCR reaction mixtures (20 lL total volume) contained
the following components ⁄ concentrations: one unit of
Hotstart Taq DNA polymerase (Qiagen), 2 lL of 10·
reaction buffer (components here) (Qiagen), 2 mM of
each dNTP (Bioline), with 0.2 lM of each forward and
reverse primer and 20–50 ng of genomic DNA. PCR
reactions were performed in a MJ-Research PTC-200
Peltier Thermal Cycler. The following programme was
used: 1 step at 95 C for 15 min followed by 35 cycles
of 1 min at 94 C, 1 min at 55 C, 1 min and 30 s at
Table 1 Nucleotide sequences of nuclear and cpDNA primers used in this study
Locus
Dye
Primers
Repeat Na Ha
*CH01F02†
6-Fam
(GA)22 11
*CH02D11†
Vic
MSS5‡
6-Fam
MSS13‡
6-Fam
MSS16‡
Vic
*CH01F09†
No dye
MSS6‡
No dye
trnH§
psbA§
No dye
Forward CCACATTAGAGCAGTTGAGGATGA
Reverse ATAGGGTAGCAGCAGATGGTTGT
Forward AAATAAGCGTCCAGAGCAACAG
Reverse GGGACAAAATCTCACAAACAGA
Forward CCCCAACAACATTTTTCTCC
Reverse CCTCTCGCTCTTTGCCTCT
Forward GAAAATTCCTTCCCGAACTTCAT
Reverse AACTCACTCGGATTTTGGAACCT
Forward ATGTCACATCTCTCCCCTTGTGT
Reverse TTTTGCCCTCAAAGAATGCCTTA
Forward ATGTACATCAAAGTGTGGATTG
Reverse GGCGCTTTCCAACACATC
Forward CGAAACTCAAAAACGAAATCAA
Reverse ACGGGAGAGAAACTCAAGACC
ACTGCCTTGATCCACTTGGC
CGAAGCTCCATCTACAAATGG
(AG)21 26
(GA)19 15
(GA)12 11
(GA)28 23
(AG)22
(CA)14
4
*CH01F02, CH02D11, and CH01F09 are also referred to in this paper by their abbreviated names F02, D11, and F09 respectively.
†Microsatellite primers derived from Malus · domestica DNA (Gianfranceschi et al. 1998).
‡Microsatellite primers derived from S. torminalis (Oddou-Muratorio et al. 2001).
§Chloroplast primers from Hamilton (1998). Na = number of alleles amplified across all Avon gorge taxa. Ha = number of
chloroplast haplotypes detected.
2010 Blackwell Publishing Ltd
EVOLUTION AND SPECIATION IN SORBUS 1679
72 C and a 10 min extension step of 72 C. PCR amplification products from the labelled primers were separated on a laser-based capillary electrophoresis
instrument, the ABI 3730xl. Alleles were sized relative
to an internal size standard and resulting electropherograms were analysed using GeneMarker genotyping
software. Each individual electropherogram was scored
manually. The PCR amplification products from the
two unlabelled primers were separated by horizontal
agarose gel electrophoresis (2%). Gels were stained
with ethidium bromide and visualized under UV light.
Data analysis
Models used to detect genetic diversity within and
among species and those that determine population
structure and hybrid origins, rely on the assumptions
that allele frequencies can be readily determined and
that source populations are in Hardy–Weinberg equilibrium. Apomictic taxa are generally polyploid and contain genomes where individual loci are not free to
recombine, thus violating the principal assumption of
most genetic models. To complicate matters further,
allele frequencies are difficult to determine for polyploid taxa. The number of different alleles detected in a
polyploid individual is often lower than its ploidy level.
For example, a diploid individual carrying alleles A
and B at a locus must have one copy of each, while a
triploid individual carrying alleles A and B may have
two copies of either A or B and one of two different
genotypes, AAB or ABB.
The most appropriate measure to study genetic diversity for apomictic taxa is to compile and compare multilocus genetic phenotypes. Therefore, for each
individual analysed in this study, microsatellite allele
phenotypes were determined for each of the seven
amplified loci. These data were combined to construct
the multilocus phenotypes of each sample. The numbers of individuals belonging to each multilocus phenotype were determined for each species. To estimate
clonal diversity the following parameters were calculated:
1 the proportion of distinguishable genotypes (Ellstrand & Roose 1987) was measured as Ng ⁄ Ni, where
Ng is the number of genotypes and Ni is the total number of individuals sampled.
2 Simpson’s diversity index (D) modified for finite
sample size by Pielou (1969) measures the probability
that two individuals selected at random from a population of N plants will have different genotypes. D can
range from 0 to 1, with 1 being the maximum diversity.
P
Each index value was calculated as: D = 1 ) {ni(ni )
1) ⁄ N(N ) 1)} where ni is the number of individuals of
genotype i and N is the sample size.
2010 Blackwell Publishing Ltd
3 Genotypic evenness (Fager 1972) was measured as:
E = D ) Dmin ⁄ Dmax ) Dmin, where Dmin = (G ) 1)(2N )
G) ⁄ N(N ) 1) and Dmax = (G ) 1)(N) ⁄ G(N ) 1) where G
is the number of clones and N is the sample size. Index
values can range from 0 for a population dominated by
one genotype, to 1 for a population in which all genotypes are represented equally. Two simple measures of
allele diversity, the total number of different alleles seen
across all loci and the mean number of different alleles
carried by each individual, averaged across loci, were
calculated using F-Dash (Obbard et al. 2007). Ploidy
level estimates were based on the maximum number of
displayed alleles at a single locus. For example, at two
loci, MSS5 and MSS16, three alleles were displayed for
S. bristoliensis suggesting a triploid genome.
Each multilocus phenotype was transformed into binary code and metric distances were computed between
them using Dice similarity coefficients (Dice 1945) with
NTSYSPC version 2.11a software (Rohlf 2002). The relationships among the multilocus phenotypes were visualised using principal coordinate analysis (PCO) with
NTSYSPC software (Rohlf 2002) and neighbour-joining
analysis (NJ) using the NEIGHBOUR programme in PHYLIP
3.6 (Felsenstein 2004).
To determine the origins of the putative apomictic
taxa simulated matings (crosses and selfs) were performed among and within the Sorbus taxa to determine whether any of the various mating possibilities
(self-fertilization, interspecific cross-fertilization, or
intraspecific cross-fertilization) could generate a specified polyploid multilocus phenotype. A table of results
for each polyploid was compiled listing the number of
allele mutations required for any mating to generate a
phenotypic match. For any mating where the number
of allele mutations required was zero, the taxa were
considered potential progenitors. For phenotypes that
could not be matched, the next best mating and the
number of allele mutations required to make the match
were recorded. All putative apomictic multilocus phenotypes were considered as potential parents of each
other; there were no prior assumptions regarding taxonomic hierarchy used in this study. For the sexual taxa,
all possible multilocus phenotypes that could be created
from the observed population genepool were considered potential parents. Table S5a and S5b (Supporting
Information) shows how this process was carried out
for S. eminens. The only cross that matches the S. eminens multilocus phenotype is S. aria · S. porrigentiformis
(Table S5b). It is also possible to derive from this table
the number of alleles that are missing from each cross
in order to match the multilocus phenotype of S. eminens. For example, five specific alleles are missing from
the S. aria genepool for S. aria to be the sole parent of
S. eminens (Table S5b) and three alleles are missing
1680 A . R O B E R T S O N E T A L .
from the combined genepools of S. aria and S. rupicola
for the origin of S. eminens to be S. aria · S. rupicola.
Character compatibility analysis (Mes 1998) was used
to determine if any of the observed differences seen
among polyploid multilocus phenotypes was best
explained by the accumulation of mutations or whether
the character patterns observed are more parsimoniously accounted for by genetic recombination through
any of the various mating possibilities outlined above.
An incompatibility is illustrated by a comparison of
potential patterns observed in pair wise, binary characters; if mutation is the sole mechanism for the generation of variation, only three of the four possible
combinations of characters can be achieved in a lineage,
excluding the presence of back mutations. If all four
possible character combinations are present, this is
deemed an incompatibility and is most parsimoniously
explained by genetic exchange. Matrix incompatibilities
were calculated, using the Jactax.exe function in PICA
version 4.0 (Wilkinson 2001), for three taxonomic
groups based on their cpDNA haplotypes, subgenus
Aria, nothosubgenus Tormaria and S. torminalis and
nothosubgenus Soraria and S. aucuparia. Multilocus
phenotypes responsible for the greatest number of
incompatibilities were successively deleted from the
data set until all incompatibility was resolved or the
number of genotypes exhausted. If removal of virtually
all the genotypes is required to achieve matrix compatibility, it can be concluded that the relationships among
these genotypes are essentially the result of recombination. The binary coded matrix, derived from the
multilocus phenotypes, was used as the input for the
Jactax programme, with the modification that the interdependence of markers was specified. This prevents the
calculation of character incompatibility between alleles
of the same microsatellite locus.
Results
Chloroplast (cp)DNA analysis
Sorbus aria, S. aucuparia and S. torminalis each had
unique trnH–psbA sequences. Two insertion ⁄ deletion
polymorphisms (indels) were found that distinguished
S. aria from S. torminalis and three nucleotide substitutions were found that distinguished S. aucuparia from
both S. aria and S. torminalis (Fig. 2). Sorbus eminens,
S. porrigentiformis, S. wilmottiana, S. whiteana and
Fig. 2 Chloroplast DNA sequences for the intergenic region trnH–psbA. Nucleotide differences are shaded and the three nucleotide
substitutions that distinguish S. aucuparia from S. aria ⁄ S. torminalis are highlighted (*). Regions A and B represent sequence
deletions ⁄ insertions that distinguish S. aria from S. torminalis. NNN: region not clearly sequenced due to polymerase slippage.
GenBank accession numbers are given in brackets.
2010 Blackwell Publishing Ltd
EVOLUTION AND SPECIATION IN SORBUS 1681
S. latifolia (with the exception of a single nucleotide
substitution) each had sequences that matched S. aria
(Fig. 2). Two taxa, S. bristoliensis and S. croceocarpa, had
sequences identical to S. torminalis (Fig. 2). Sorbus aucuparia-type sequences were found in S. anglica, S. · thuringiaca and S. intermedia (Fig. 2). One of the nucleotide
substitutions that differentiated both S. aria and S. torminalis from S. aucuparia was situated within a Tsp5901
restriction digestion site (AATT). By digesting the PCR
products with Tsp5901, S. aria, S. aucuparia and S. torminalis each had a unique RFLP pattern (data not shown).
This diagnostic PCR-RFLP assay provided a convenient
and efficient method for determining the chloroplast
haplotypes of large numbers of Avon Gorge Sorbus taxa.
When applied in this manner, the test revealed that
there was no variation in restriction site pattern within
any of the sexual taxa, S. aria, S. torminalis or S. aucuparia
or, as expected, within any of the polyploid taxa. The
S. aria type restriction pattern was also found in S. ·
robertsonii, S. leighensis and S. · avonensis, and the
S. torminalis type was also found in S. decipiens, and S. ·
houstoniae. These haplotypes are consistent with those
reported by Lemche (1999) and Chester et al. (2007).
Nuclear microsatellite analysis
In total, 86 alleles from five labelled microsatellite loci
were amplified from 227 adult Sorbus individuals. The
number of alleles amplified at each locus across all taxa
ranged from 11 at locus F02 to 26 at locus D11
(Table 1). The total number of different alleles displayed by each taxon ranged from 10 for S. leighensis,
and S. · avonensis, to 38 for S. torminalis (Tables 2, 3
and Table S2, Supporting Information). The mean number of different alleles carried by each individual averaged across all loci ranged from 1.6 per locus in S. aria
to 3.4 per locus in S. porrigentiformis (Table 2). Amplification products from the two unlabelled microsatellite
loci, CH01F09 and MSS6, proved to be genome-specific,
with CH01F09 only being amplified in S. aria and associated hybrids and MSS6 only amplified in S. torminalis
and associated hybrids. These two loci were scored on
a presence-absence basis for each of the 227 adult individuals (Table 3, Table S2). When data from all seven
loci were combined, the total number of detectable
multilocus phenotypes scored across all taxa was 86.
With the exception of S. eminens (with two multilocus
Table 2 Genetic diversity measures for the Avon Gorge Sorbus taxa
Taxon
Native sexual taxa
S. aria
S. aucuparia
S. torminalis
S. · thuringiaca
Native polyploids
S. eminens
S. · robertsonii
S. porrigentiformis
S. leighensis
S. · avonensis
S. wilmottiana
S. whiteana
S. bristoliensis
S. · houstonii
S. anglica
Introduced polyploids
S. croceocarpa
S. decipiens
S. latifolia
S. intermedia
Taxon outside of Avon Gorge
S. rupicola
N
Ni
Xo
Xe
A’
H’
Ng
Ng ⁄ Ni
D
E
>1000
20
50
2
34
14
20
2
2
2
2
2
2
2
2
2
26
29
38
14
1.6
1.6
1.7
1.7
34
14
19
2
—
—
—
—
—
—
—
—
—
—
—
—
80
1
80
80
2
40
20
300
1
40
27
1
22
4
2
26
5
32
1
20
4
3
4
3
3
3
3
3
4
3
4
na
4
na
na
3
3
3
na
3, 4
15
13
17
10
10
11
12
11
15
15
3
2.6
3.4
2
2
2.2
2.4
2.2
3
3
2
1
1
1
1
1
1
1
1
1
0.07
—
0.04
—
—
0.03
0.2
0.03
—
0.05
0.07
—
—
—
—
—
—
—
—
—
0.00
—
—
—
—
—
—
—
—
—
6
3
10
6
4
1
6
4
4
3
4
4
4
4
4
4
16
12
16
16
3.2
2.4
3.2
3.2
1
1
1
1
—
—
0.16
—
na
2
4
4
16
3.2
1
—
—
—
—
—
—
—
—
—
—
—
—
—
N = estimated population size in the Avon Gorge; Ni = number of individuals sampled; Xo = observed ploidy, based on the
maximum number of displayed alleles at a single locus; Xe = expected ploidy, based on published cytological work; A¢ = total
number of different alleles seen across all loci; H¢ = mean number of different alleles carried by each individual averaged across all
loci; Ng = number of multilocus phenotypes detected; Ng ⁄ Ni = proportion of clones detected; D = estimate of multilocus phenotype
diversity (multiclonal spp. only); E = multilocus phenotype diversity (multiclonal spp. only).
2010 Blackwell Publishing Ltd
1682 A . R O B E R T S O N E T A L .
Table 3 Multilocus phenotypes for the polyploid Sorbus taxa
Alleles at locus:
Taxon
F02
Subgenus Aria
S. eminens
196,202,206
S. eminens 154
196,202,206
S. x robertsonii
196,200,202
S. porrigentiformis
196,202,206,208
S. leighensis
196,202
S. x avonensis
202,206
S. wilmottiana
200,208
S. whiteana
200,208
S. rupicola
196,204,206,214
Nothosubgenus Tormaria
S. bristoliensis
192,202
S. x houstonii
192,196,202
S. croceocarpa
192,198,202
S. decipiens
194,198,208
S. latifolia
192,194,200,206
Nothosubgenus Soraria
S. anglica
192,202,206
S. intermedia
194,196,206
MSS5
D11
MSS13
120,126,134,142
120,126
120,126,142
120,132,136,142
120,140,142
120,126,140
126,132,140
126,134,140
124,132,136,180
159,163,179
163,179
163,179
159,205
159,193
159,179
159,171
159,189,193
159,169,175
199,201,205
199,201,205
203,205
199,201,203,209
201,205
199,205
199,205
199,205
199,201,203
120,134,144
120,134,140,144
126,132,140
126,130,134
128,132,142
159
159,179
159,179,196
159,163
159,169
126,136,142
124,128,132
phenotypes), each of putative apomictic taxa displayed
just one multilocus phenotype (Table 3). In contrast,
every S. aria and every S. aucuparia individual and 19
out of 20 S. torminalis individuals had a unique multilocus genotype. All individuals genotyped for the sexual
taxa displayed either one or two alleles at each locus,
confirming them to be diploid. In contrast, each individual genotyped among the putative apomictic taxa displayed more than two alleles, usually three or four, at a
minimum of one locus, confirming that they were all
polyploid (Table 3).
Of the 86 microsatellite alleles, distributed among the
19 Sorbus taxa investigated, 43 were shared between the
sexual taxa and the polyploids, 28 were found only
among the sexual taxa and 15 were restricted to the
polyploids. Excluding apomicts the sexual taxa had
high numbers of characteristic alleles, 13 for S. aria, 18
for S. aucuparia and 22 for S. torminalis (Table S2).
Genetic similarity and dissimilarity, among the 86 multilocus phenotypes was calculated and the subsequent
matrices were used as the input for principal coordinate
analysis (PCO) and neighbour-joining (NJ) analysis
respectively. The PCO showed complete differentiation
of the three sexual taxa S. aria, S. torminalis and S. aucuparia (Fig. 3). Members of subgenus Aria formed a distinct group but with a clear subdivision, one sub-group
consisting solely of S. aria individuals and the other
sub-group consisting of a mix of S. aria and the subgenus Aria polyploids. All members of nothosubgenus
Tormaria, with the exception of S. decipiens, formed a
MSS16
F09
MSS6
166,178
166
166,172,178
166,170,178
166
166
166,172
166,172
166,170
1
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0
0
0
199,201
199,201,205
187,199,201,205
195
187,199,207
166,178,198
166,178,198
166,168,198
166,192,200
166,168,196,234
1
1
1
1
1
1
1
1
0
1
157,159,205
199,201,203
162,166,170
1
0
153,159,165
199,201,203
162,166,170,182
1
1
single group lying between subgenus Aria and S. torminalis. The slightly anomalous position of the central
European S. decipiens may represent further hybridization between a member of the nothosubgenus Tormaria
and a member of the subgenus Aria or be a consequence of its parent taxa not being included in the
genetic analysis. Sorbus · thuringiaca individuals fell
between the S. aria subgroup, of the subgenus Aria, and
S. aucuparia. Sorbus anglica sat between the polyploid
subgroup of the subgenus Aria and S. aucuparia. The
position of S. intermedia was central among the three
main groups. This result is consistent with the findings
of Nelson-Jones et al. (2002) who found that S. intermedia had inherited a proportion of its genome from each
of S. aria sensu lato, S. torminalis and S. aucuparia.
The results of the NJ analysis concur with those of
the PCO analysis with both showing a clear differentiation of the three sexual taxa S. aria, S. torminalis and
S. aucuparia and also the intermediate polyploid taxa
forming clusters near or between the proposed progenitor taxa (Fig. 4). What the NJ tree does show, which is
not apparent in the PCO, is the polyphyletic nature of
the polyploids within both the subgenus Aria and the
nothosubgenus Tormaria. Where clearly individuals
within each microspecies are monophyletic, as they are
genetically identical (with the exception of S. eminens),
each polyploid group as a whole consists of a number
of monophyletic clusters, four within the subgenus Aria
and three within the nothosubgenus Tormaria (Fig. 4). It
must be noted however that without statistical support
2010 Blackwell Publishing Ltd
EVOLUTION AND SPECIATION IN SORBUS 1683
Au
0.64
Au
Au
Au
Au
Au
Au
Au
Th3
Au
Au
0.38
Au
Au
Au Au
Th5
Dim–2
Aria group 1
0.11
Ar
Ar Ar
Ar
ArArArAr
Ar
Ar Ar
Ar
Ar
Ar
Ar ArAr
ArAr
Ar
Ar Ar
Ar
Ar
Ebc
Ar ArWil
ArAr
Ar
Ar ArAr
Whi
EmB
PoB
ArPoC
S. aucuparia
Dec
Fig. 3 Principal coordinate analysis of
the similarity relationships among the
86 Avon Gorge Sorbus multilocus phenotypes, showing the first and second
axes. The major groupings are circled
and labelled. Subgenus Aria is divided
into two; Aria group 1 consists solely of
S. aria and Aria group 2 consists of both
S. aria and subgenus Aria polyploids.
Taxon abbreviations as per Table 4. The
proportion of the variation accounted
for by these two axes is 31%.
Int
Ang
Rup
S. torminalis
Por
Em
–0.15
Aria group 2
Cro
Bbc
Nothosubgenus
Tormaria
–0.42
–0.54
–0.26
0.02
To
Br
Lat
To
ToTo
To To
ToTo
To
To
To
To
To
ToTo
To
To
To
To
0.30
0.58
Dim–1
values phylogenetic conclusions from NJ trees must be
interpreted with caution.
Mating simulations
Simulated matings (crosses and selfs) were performed
among and within the 19 Sorbus taxa to determine
whether any of the various mating possibilities (self-fertilization, interspecific cross-fertilization or intraspecific
cross-fertilization) could generate a specified polyploid
multilocus phenotype. A table of results for each polyploid was compiled listing the number of allele mutations required for any mating to generate a phenotypic
match (Table S3a–17b, Supporting Information). For
any mating where the number of allele mutations
required was zero, the taxa were considered potential
progenitors (Table 4). For phenotypes that could not be
matched, the next best mating and the number of allele
mutations required to make the match were recorded
(Table 4). Within the subgenus Aria single genotype
matches were found for S. eminens (S. aria · S. porrigentiformis) and for S. · robertsonii (S. aria · S. eminens).
Multiple genotype matches were found for S. wilmottiana (two), S. leighensis (two) and S. · avonensis (12). The
genotypes of S. porrigentiformis, S. rupicola and S. whiteana could not be matched (Table 4). No single genotype
matches were found for any of the nothosubgenus Tormaria taxa. Two taxa, S. bristoliensis (eight) and S. ·
2010 Blackwell Publishing Ltd
houstoniae (three), had multiple phenotype matches
(Table 4). The genotypes of the remaining three taxa,
S. croceocarpa, S. decipiens and S. latifolia could not be
matched, which is consistent with their treatment as
introduced species in the British flora. Single genotype
matches were found for S. anglica (S. aucuparia · S. porrigentiformis) and for S. · thuringiaca genotype 252 (S. aucuparia · S. aria). The genotype of S. intermedia and S. ·
thuringiaca genotype 232 could not be matched
(Table 4), the former consistent with it being an introduced species.
Incompatibility analysis
Based on cpDNA haplotypes each of the three taxonomic groups had high incompatibility counts, subgenus Aria 266, nothosubgenus Tormaria and S. torminalis
285 and nothosubgenus Soraria and S. aucuparia 165
(Fig. 5). Within each of these groups the majority of
genotypes had to be removed to allow complete compatibility of the data set, subgenus Aria 89%, nothosubgenus Tormaria and S. torminalis 76% and
nothosubgenus Soraria and S. aucuparia 66%. The only
compatibility found among the polyploid taxa was
between the two multilocus phenotypes of S. eminens.
Each of the remaining polyploids, including S. eminens,
(the subgenus Aria was reanalysed without the single
S. eminens phenotype 154) was removed from the
1684 A . R O B E R T S O N E T A L .
0.1
To19
To18
To14
To11
To3 To7 To10To12
To16
To15 To13
To5 To2
To9
To8
To6
To20
To4
To1
To17
Au6
Au1
Au5
Au2 Au4
Au3
Au8
Au10Au11
Au14
Au13
Au7
Au12
Lat
Au9
Th3
Br
Cro
Bbc
Rup
Int
Th5
Por
Ang
Whi
Wil
Dec
Ar7
Ar10
PoB
Ar23
Ar9
Ar6
Ar33
Ar4
Ar15
Ar27
Ar21
Ar1Ar26
Ar24
Ar20
Ar22 Ar17
Ar5
Ar14
Ar28 Ar11
Ar16
Ar34 Ar2
Ar32
Ar12
Ar19
Ar8 Ar3
Ar31
Ar29 Ar25
EmB
Ar30
Ar13 PoC Em
Ar18
Ebc
Fig. 4 Neighbour-joining tree of 86 Avon Gorge Sorbus multilocus phenotypes. The tree was constructed using a dissimilarity
matrix, 1-Dice similarity estimates, based on microsatellite analysis. Taxon abbreviations are as Table 4.
analysis before compatibility had been reached. It can
be concluded that relationships among the majority
of the multilocus phenotypes within each group are
essentially the result of recombination, as opposed to
mutation. The origin of each of the polyploids, with
the exception of the single S. eminens 154 genotype,
was therefore found to be incompatible with clonal
reproduction.
Discussion
This study has shown conclusively that the generation
of genetic novelty within the agamic Sorbus complex of
the Avon Gorge has been driven primarily by a series
of interspecific hybridisations and backcrosses among
closely related taxa, with each new genotype being
fixed and propagated via apomictic reproduction. The
result is a series of new closely related taxa, microspecies and hybrids (as defined by Rich et al. 2009), that
are reproductively isolated from each other, but which
occasionally participate in further sexual hybridisation
events leading to a complex pattern of ongoing reticulate evolution of Sorbus within the Avon Gorge.
The distribution of nuclear microsatellite alleles
within and among the various Sorbus taxa identified
those which reproduce sexually and those which reproduce primarily by apomixis. For each of the sexually
reproducing taxa (S. aria, S. aucuparia and S. torminalis)
each individual (with the exception of two S. torminalis
individuals) had a unique multilocus genotype, as
would be predicted for obligate (self-incompatible)
outcrossing taxa and, with the exception of S. eminens,
each of the putative apomictic taxa had a single multilocus genotype, as would be expected for asexually reproducing, apomictic, taxa. The microsatellite allele data
also allowed confirmation of ploidy levels among the
different taxa. Sexual taxa were diploid and apomictic
taxa polyploid, which is largely consistent with other
2010 Blackwell Publishing Ltd
EVOLUTION AND SPECIATION IN SORBUS 1685
Table 4 Origins of the Avon Gorge Sorbus taxa compiled from the results of the cpDNA and nuclear DNA analysis
Mating simulation
matches
Taxa (Key)
CpDNA type
PCO position
S. eminens (Em)
S. · robertsonii (EbC)
S. porrigentiformis (Po)
S. aria
S. aria
S. aria
Aria group
Aria group
Aria group
S. aria (Ar) · Po
Ar · Em
S. leighensis (PoB)
S. · avonensis (PoC)
S. aria
S. aria
Aria group
Aria group
S. wilmottiana (Wi)
S. whiteana (Wh)
S. aria
S. aria
Aria group
Aria group
Ar · Em, Ar · Po
12 possible crosses see
supplementary data table
Ar · Po, Ar · Wh
S. rupicola (Ru)
S. bristoliensis (Br)
S. aria
S. torminalis
Aria group
Nothosubgenus Tormaria
S. croceocarpa (Cr)
S. torminalis
Nothosubgenus Tormaria
S. decipiens (De)
S. torminalis
S. latifolia (La)
S. aria
Groups with S. intermedia
and S. anglica
Nothosubgenus Tormaria
S. · houstonii (BrB)
S. torminalis
Nothosubgenus Tormaria
S. anglica (An)
S. aucuparia
S. intermedia (In)
S. aucuparia
S. thuringiaca 232 (Th3)
S. aucuparia
Intermediate between
S. aria, S. aucuparia
Intermediate between
S. aria, S. aucuparia and
S. torminalis
Intermediate between
S. aria and S. aucuparia
S. thuringiaca 252 (Th5)
S. aucuparia
Intermediate between
S. aria and S. aucuparia
estimates made from cytological studies (Bailey et al.
2008). The total number of alleles recorded across all
loci ranged from 10 to 17 for the polyploid taxa and
from 26 to 38 for the sexual taxa. The lower values of
A¢ (the total number of different alleles scored across
loci, Table 2) for the polyploid taxa suggest monophyletic origins (single hybridization events) and
subsequent reproductive isolation through apomixis. In
general sexual taxa have higher levels of A¢ due to their
2010 Blackwell Publishing Ltd
Next best mating
simulation match,
(n) = number of
allele mutations
required to make a match
Em · Ru (3) EbC ·
Ru (3) BrB · An (3),
Em · An (3), Int ·
Em (3)
Ar · An, Ar · Br,
Ar · BrB, Ar · Cr,
Ar · De, Ar · Em,
Ar · Po, Ar · PoC,
Ar · Wi, all (2)
Em · S. torminalis (Tor),
Po · Tor, BrB · BrB,
BrB · Em, BrB · Po,
BrB · Pob, BrB · PoC,
BrB · Tor
Ar · Tor (2),
Ar · Brb (2)
PoC · Tor (2),
Cr · PoC (2),
Cr · Wi(2)
Ar · Br, Em · Tor,
EmC · Tor
S. aucuparia (Au) ·Po
Au · Po (2),
Au · Ru (2)
Ar · Au (3),
Au · Au (3),
Au · In (3)
Ar · Au
ability to recombine genotypes among individuals. The
average number of alleles at each locus, carried by each
individual, ranged from 2.0 to 3.4 for the polyploid taxa
and from 1.6 to 1.7 for the diploid taxa. High values of
H¢ (the average number of alleles at each locus carried
by each individual, Table 2) for polyploid taxa are consistent with these taxa being the progeny of genetically
diverse parents with this genetic diversity subsequently
being fixed in them by apomictic reproduction.
Matrix incompatibility
1686 A . R O B E R T S O N E T A L .
300
250
T
A
S
200
150
100
50
0
1
3
5
7
9 11 13 15 17 19 21 23 25 27 29 31 33 35
Number of phenotypes deleted
Fig. 5 Reduction of matrix incompatibility upon successive
deletion of multilocus phenotypes in each of the three taxonomic groups based on cpDNA haplotypes. Subgenus
Aria = A, Nothosubgenus Tormaria and S. torminalis = T and
Nothosubgenus Soraria and S. aucuparia = S.
Hybrid origins of the Avon Gorge Sorbus
microspecies
Comparisons of cpDNA type and multilocus microsatellite genotypes among the 19 Sorbus taxa of the Avon
Gorge allowed us to make predictions about the origins
of the polyploid apomictic taxa. The cpDNA type of
S. eminens, S. porrigentiformis, S. wilmottiana and S. whiteana matched that of S. aria and the PCO and NJ analyses placed these four taxa within the subgenus Aria
cluster. These results are consistent with previous
morphological, cytological and molecular studies that
proposed S. aria was at least one of the parents (directly
or indirectly) of these taxa and the other parent, if not
S. aria, was a member of the subgenus Aria (Warburg
1952; Proctor & Groenhof 1992; Nelson-Jones et al. 2002;
Rich & Houston 2006; Chester et al. 2007). Nuclear
microsatellite markers used in the present study were
variable in their ability to determine taxon origins so
we were unable to determine the precise origins of all
the taxa. However, they were able to show, through
character compatibility analysis, that the origin of each
polyploid (with the exception of the single S. eminens
genotype 154) was incompatible with clonal reproduction and so was explained most parsimoniously by
hybridization and genetic recombination.
The origin of S. eminens was conclusive: only by
crossing S. aria with S. porrigentiformis could the multilocus phenotype of S. eminens be matched. All other
parental combinations (intraspecific and interspecific)
and selfing scenarios were excluded. These results are
consistent with the findings of Nelson-Jones et al. (2002)
which placed S. eminens and S. porrigentiformis in the
same subgenus and with those of Proctor & Groenhof
(1992), who found that S. porrigentiformis, type D, found
on Seven Sisters Rocks in the Wye Valley, shared a similar peroxidase phenotype with S. eminens. Nelson-Jones
et al. (2002) went on to propose an autopolyploid origin
for S. eminens from S. aria. However, our data show
conclusively that S. eminens is the product of an interspecific cross (S. aria · S. porrigentiformis), albeit from
two species within the S. aria group.
The precise origin of S. porrigentiformis could not be
determined because its multilocus phenotype could not
be matched by any parental recombination of microsatellite alleles from within the Avon Gorge, even though
S. aria is clearly one of its parents on morphological
grounds. The most likely explanation for this result is
that S. porrigentiformis originated outside the Avon
Gorge. Previous studies of S. porrigentiformis revealed
varying degrees of genetic variation over wide geographical scales and concluded that this is best
explained by inferring numerous independent origins
(Proctor & Groenhof 1992; Nelson-Jones et al. 2002;
Chester et al. 2007). Such a conclusion however must be
treated with caution because there are other equally
likely sources of genetic variation in apomictic species,
such as mutation and facultative sexuality (reviewed in
Hörandl & Paun 2007). Indeed the results of our study
suggest that genetic variation among S. porrigentiformislike clones could be the consequence of S. porrigentiformis backcrossing with one of its parents. Two
S. porrigentiformis-like clones were found in the Avon
Gorge, S. leighensis and S. · avonensis, and it was shown
that the most likely origin of these taxa was from the
cross S. aria · S. porrigentiformis. The PCO analysis
placed S. leighensis and S. · avonensis in an intermediate
position between S. porrigentiformis and S. aria and the
results of the mating simulations supported these placements. Two potential crosses matched the multilocus
phenotype of S. leighensis: S. eminens · S. aria and S. aria
· S. porrigentiformis. As S. eminens has been shown to be
the progeny of S. porrigentiformis, it is clear that S. leighensis originated from a S. porrigentiformis backcross,
either directly or indirectly, with S. aria. The mating
simulation results for S. · avonensis were not so conclusive as there were 12 possible cross matches. The
crosses S. eminens · S. aria and S. aria · S. porrigentiformis are among the possibilities and S. porrigentiformis
was involved either directly or indirectly in the remaining crosses. These results clearly show that S. leighensis
and S. · avonensis had taxonomically different parents
to S. porrigentiformis so can be ascribed independent
taxonomic status (Rich et al. 2009). For similar reasons,
S. · robertsonii (Rich et al. 2009) was not considered a
clone of S. eminens since the PCO and mating simulations showed conclusively that it originated from the
cross S. aria · S. eminens.
Two potential crosses matched the multilocus phenotype of S. wilmottiana: S. aria · S. porrigentiformis and
S. aria · S. whiteana. Support for the cross S. aria ·
S. porrigentiformis being responsible for the origin of
2010 Blackwell Publishing Ltd
EVOLUTION AND SPECIATION IN SORBUS 1687
S. wilmottiana was also found by Chester et al. (2007).
Morphologically similar to S. wilmotianna, S. whiteana is
a recently described polyploid (Rich & Houston 2006)
also placed in subgenus Aria by its authors and this
designation is supported by our data. The cpDNA type
of S. whiteana matches that of S. aria and the PCO analysis places it clearly within the subgenus Aria cluster.
The results of the mating simulations were however
inconclusive as no match was found for the multilocus
phenotype of S. whiteana. It is likely that S. whiteana
originated outside the Avon Gorge, presumably in the
Wye Valley, which is the only other area where it is
found (Rich & Houston 2006).
The cpDNA type of S. bristoliensis matched that
of S. torminalis and the PCO and NJ analyses placed
S. bristoliensis in an intermediate position between subgenus Aria and S. torminalis. These results are consistent
with previous morphological, cytological and molecular studies (Wilmott 1934; Warburg 1952; Proctor &
Groenhof 1992; Nelson-Jones et al. 2002; Chester et al.
2007), which suggested that S. torminalis was the female
parent of S. bristoliensis and the paternal parent was
from subgenus Aria. The results of our mating simulations for S. bristoliensis were not conclusive and it is
possible that either S. eminens or S. porrigentiformis (from
subgenus Aria) is the paternal parent. Nevertheless, six
other simulated crosses also gave the multilocus phenotype of S. bristoliensis, but all of these included progeny
of either S. porrigentiformis or S. bristoliensis, so were not
considered potential paternal parents of S. bristoliensis.
Closely related to S. bristoliensis was the new taxon
S. · houstoniae, which closely resembled (phenotypically) S. bristoliensis and could easily be mistaken for
another S. bristoliensis clone. However, S. · houstoniae
was found to be a tetraploid, rather than a triploid, and
upon closer inspection appeared morphologically closer
to S. aria than S. bristoliensis (Rich et al. 2009). The
cpDNA type of S. · houstoniae matched that of S. torminalis and the PCO analysis placed it in an intermediate
position between S. bristoliensis and subgenus Aria.
The mating simulations were also consistent with these
results and gave three possible crosses that matched the
origin of S. · houstoniae: S. aria · S. bristoliensis, S. eminens · S. torminalis and S. · robertsonii · S. torminalis. The
most likely cross was concluded to be S. aria · S. bristoliensis, which would require that an unreduced triploid
S. bristoliensis egg be fertilized by a reduced haploid
S. aria sperm. It is not uncommon for an unreduced
triploid egg to be fertilized by a reduced haploid sperm
in the genus Sorbus (Liljefors 1953; Robertson et al.
2004b), making this seemingly unlikely sexual scenario
the likely source of S. · houstoniae B. The S. eminens ·
S. torminalis cross would require either S. torminalis to
produce an unreduced egg in a cross with a reduced
2010 Blackwell Publishing Ltd
diploid S. eminens male gamete or S. torminalis producing a normal haploid egg crossed with an unbalanced
triploid S. eminens male gamete. The S. · robertsonii ·
S. torminalis cross would require similar gamete formations to S. aria · S. bristoliensis except that on this occasion the subgenus Aria member, S. eminens, would need
to produce an unbalanced triploid male gamete that
crossed with a reduced haploid S. torminalis egg. Unreduced S. torminalis gametes and unbalanced subgenus
Aria male gametes have not yet been recorded in Sorbus
(Liljefors 1953; Robertson et al. 2004b).
The cpDNA type of S. anglica matched that of S. aucuparia and the PCO analysis placed S. anglica in an
intermediate position between subgenus Aria and
S. aucuparia, consistent with previous studies (Warburg
1952; Proctor & Groenhof 1992; Nelson-Jones et al. 2002;
Chester et al. 2007). The results of the mating simulations
were conclusive with only one simulation matching
the multilocus phenotype of S. anglica: S. aucuparia ·
S. porrigentiformis. This result is consistent with the
findings of Proctor & Groenhof (1992), who suggested,
based on isozyme studies, that S. porrigentiformis was
likely to be the subgenus Aria member that crossed with
S. aucuparia.
Continuing evolution of Sorbus in the Avon Gorge
Hybridization and backcrossing appear largely responsible for ongoing evolution of new apomictic polyploid
Sorbus taxa within the Avon Gorge. Indeed, similar evolutionary dynamics have been described for Sorbus in
other UK and Scandinavian populations (Liljefors 1955;
Bolstad & Salvesen 1999; Nelson-Jones et al. 2002; Robertson et al. 2004a, b; Houston et al. 2009) and such processes appear common among agamic complexes as
mechanisms for generating genetic and phenotypic novelty that is then maintained by natural selection (Hörandl et al. 2001; Hörandl 2006; Hörandl & Paun 2007;
Whitton et al. 2008). The frozen niche variation model
(Vrijenhoek 1984) proposes that such recurrently
formed apomictic lineages ‘freeze’ subsets of parental
genetic variation that allows for niche partitioning
among the apomictic lineages. Such apomictic lineages
have narrower ecological niches than sexual taxa but
collectively they use the resource space of their environment more effectively than sexuals (Vrijenhoek 1984;
Hörandl 2006). The model predicts that apomicts will
be better able to colonize spatially heterogeneous environments than sexuals (Hörandl 2006). The Avon Gorge
represents a particularly heterogeneous environment
with extreme local niche variation associated with, for
instance: altitude, gradient, soil depth, shade and moisture and anecdotal observation suggests that apomictic
polyploids are more likely to be found in the more
1688 A . R O B E R T S O N E T A L .
‘extreme’ niches than their sexual counterparts (Hiscock
et al. unpublished observations). These observations are
therefore consistent with the frozen niche variation
model even though the sexual diploid taxa S. aria,
S. torminalis and S. aucuparia coexist with the polyploid
apomicts within the Gorge, albeit perhaps within distinct niches. Further ecological investigations will be
needed to shed light on the niche preferences of sexual
and apomictic Sorbus taxa within the Avon Gorge.
The general lack of genetic diversity found within the
Avon Gorge Sorbus polyploids was interesting given
that, with the possible exception of S. porrigentiformis
and S. whiteana, both parents of each taxon are still
found in the Gorge and so could provide opportunities
for additional hybridisations and backcrosses. This lack
of genetic variation within the microspecies could indicate relatively recent evolutionary origins of the apomictic polyploids, probably since the last ice-age when
the Avon Gorge is presumed to have acted as an important refugium for native plants during the forest maxima (Pigott & Walters 1954; Pring 1961). Alternatively,
sexual reproduction may be an extremely rare event
among Sorbus apomicts in the Avon Gorge, because of
the need for generation of unreduced or unbalanced
gametes. For instance, in the origin of S. eminens,
and probably many other Sorbus polyploids including
S. porrigentiformis and S. rupicola, one or both of their
parents will have needed to produce unreduced or
unbalanced gametes to attain their formation. Sorbus
eminens has inherited two copies of its genome from
S. aria and two from S. porrigentiformis. The two copies
inherited from S. porrigentiformis should not have posed
any significant problems as it represents a normal meiotic reduction for a tetraploid. The two copies inherited
from S. aria require the production of an unreduced
gamete. Unreduced gamete formation has been shown
to occur in Sorbus by Liljefors (1953) and Aas et al.
(1994) where it has been proposed to be involved in the
origin of a number of taxa such as S. bohemica in central
Europe (Jankun & Kovanda 1988). If the production of
such unreduced gametes is rare, then the lack of multiple origins for S. eminens and possibly other polyploid
taxa, is clear.
This molecular marker analysis of the origins of the
Avon Gorge Sorbus complex has shown that, as with
the Sorbus rupicola–S. arranensis–S. pseudofennica-S. pseudomeinichii sequence on Arran (Robertson et al. 2004a,
b; Robertson & Sydes 2006) and its equivalent in Scandinavia, the S. meinichii complex (Bolstad & Salvesen
1999), repeated backcrossing of apomictic derivatives
with parental taxa is responsible for the generation of
new apomictic taxa. As with Arran, evolution of Sorbus
in the Avon Gorge is therefore ongoing and the
dynamic nature of this evolutionary process indicates
that conservation strategies for the rare endemic Sorbus
species in the Avon Gorge should not be simply species-based, but rather should aim to conserve the evolutionary processes that have been responsible for their
formation by conserving all Sorbus taxa within the Avon
Gorge equally (see Ennos et al. 2005). At a time of
unprecedented speeds of climate change the identification and preservation of such dynamic evolutionary
processes will be essential for maintaining biodiversity.
Acknowledgements
We thank Matthew Hegarty and Christopher Thorogood for
helpful comments on earlier versions of this manuscript and
Christian Lexer and two anonymous referees for constructive
suggestions on the final improvements. This work was funded
by a research grant from the Leverhulme Trust.
References
Aas G, Maier J, Baltisberger M, Metzger S (1994) Morphology,
isozyme variation, cytology and reproduction of hybrids
between Sorbus aria (L.) Crantz and S. torminalis (L.) Crantz.
Botanica Helvetica, 104, 195–214.
Abbott RJ (2003) Sex, sunflowers, and speciation. Science, 301,
1189–1190.
Adams KL, Wendel JF (2005) Polyploidy and genome
evolution in plants. Current Opinion in Plant Biology, 8, 135–
141.
Arnold ML (1997) Natural Hybridization and Evolution. Oxford
University Press, New York.
Bailey JP, Kay QON, McAllister H, Rich TCG (2008)
Chromosome numbers in Sorbus L. (Rosaceae) in the British
Isles. Watsonia, 27, 69–72.
Bayer RJ (1990) Patterns of clonal diversity in the Antennaria
rosea (Asteraceae) polyploid agamic complex. American
Journal of Botany, 77, 1313–1319.
Bicknell RA, Borst NK, Koltunow AM (2000) Monogenic
inheritance of apomixis in two Hieracium species with
distinct developmental mechanisms. Heredity, 84, 228–237.
Bolstad AM, Salvesen PH (1999) Biosystematic studies of
Sorbus meinichii (Rosaceae) at Moster, S. Norway. Nordic
Journal of Botany, 19, 547–559.
Briggs D, Walters SM (1997) Plant Variation and Evolution, 3rd
edn. Cambridge University Press, Cambridge.
Brownstein MJ, Carpten JD, Smith JR (1996) Modulation of
non-templated nucleotide addition by Taq DNA polymerase:
primer
modifications
that
facilitate
genotyping.
BioTechniques, 20, 1004–1010.
Challice J, Kovanda M (1978) Flavonoids as markers of
taxonomic relationships in the genus Sorbus in Europe.
Preslia, 50, 305–320.
Cheffings CM, Farrell L (2005) Species Status No. 7. The vascular
plant Red Data List for Great Britain, JNCC, Peterborough.
Chen ZJ (2007) Genetic and epigenetic mechanisms for gene
expression and phenotypic variation in plant polyploids.
Annual Review of Plant Biology, 58, 377–406.
Chester M, Cowan RS, Fay MF, Rich TCG (2007) Parentage of
endemic Sorbus L. (Rosaceae) species in the British
2010 Blackwell Publishing Ltd
EVOLUTION AND SPECIATION IN SORBUS 1689
Isles—evidence from plastid DNA. Botanical Journal of the
Linnean Society, 154, 291–304.
Coyne JA, Orr HA (2004) Speciation. Sinauer Associates,
Massachusetts.
Dice LR (1945) Measures of the amount of ecologic association
between species. Ecology, 26, 297–302.
Ellstrand NC, Roose ML (1987) Patterns of genotypic diversity
in clonal plant species. American Journal of Botany, 74, 123–131.
Ennos RA, French GC, Hollingsworth PM (2005) Conserving
taxonomic complexity. Trends in Ecology and Evolution, 20,
164–168.
Fager EW (1972) Diversity: a sampling study. American
Naturalist, 106, 293–310.
Fehrer J, Gemeinholzer B, Chrtek J, Bräutigam S. (2007)
Incongruent plastid and nuclear DNA phylogenies reveal
ancient intergeneric hybridization in Pilosella hawkweeds
(Hieracium, Cichorieae, Asteraceae). Molecular Phylogenetics
and Evolution, 42, 347–361.
Felsenstein J (2004) Inferring Phylogenies. Sinauer Associates,
Sunderland, Massachusetts.
Gianfranceschi L, Seglias N, Tarchini R, Komjanc M, Gessler C
(1998) Simple sequence repeats for the genetic analysis of
apple. Theoretical and Applied Genetics, 96, 1069–1076.
Grant V (1981) Plant Speciation, 2nd edn. Columbia University
Press, New York.
Grimanelli D, Leblanc O, Perotti E, Grossniklaus U (2001)
Developmental genetics of gametophytic apomixis. Trends in
Genetics, 17, 597–604.
Hamilton MB (1999) Four primer pairs for the amplification of
chloroplast intergene regions with intraspecific variation.
Molecular Ecology, 8, 521–523.
Hedlund T (1901) Monographie der Gattung Sorbus. Kungliga
Svenska Vetenskaps-Akademiens Handlingar, 35, 1–147.
Stockholm.
Hegarty MJ, Hiscock SJ (2005) Hybrid speciation in plants:
new insights from molecular studies. New Phytologist, 165,
411–423.
Hegarty MJ, Hiscock SJ (2008) Genomic clues to the
evolutionary success of polyploid plants. Current Biology, 18,
R435–R444.
Hollingsworth PM (2003) Taxonomic complexity, population
genetics and plant conservation in Scotland. Botanical Journal
of Scotland, 55, 55–63.
Hörandl E (2006) The complex causality of geographical
parthenogenesis. New Phytologist, 171, 525–538.
Hörandl E, Paun O (2007) Patterns and sources of genetic
diversity in apomictic plants: implications for evolutionary
potentials and ecology. In: Apomixis: Evolution, Mechanisms and
Perspective (eds Hörandl E, Grossniklaus U, Van Dijk PJ,
Sharbel T), pp. 169–194. ARG-Gantner, Ruggell, Liechtenstein.
Hörandl E, Jakubowsky G, Dobeš Ch (2001) Isozyme and
morphological diversity within apomictic and sexual taxa of
the Ranunculus auricomus complex. Plant Systematics and
Evolution, 226, 165–185.
Houston L, Robertson A, Rich TCG (2008) The distribution,
population size and growth of the rare English endemic
Sorbus bristoliensis A. J. Wilmott, Bristol Whitebeam
(Rosaceae). Watsonia, 27, 37–49.
Houston L, Robertson A, Jones K, Smith SCC, Hiscock SJ, Rich
TCG (2009) An account of the Whitebeams (Sorbus L.,
2010 Blackwell Publishing Ltd
Rosaceae) of Cheddar Gorge, England, with description of
three new species. Watsonia, 27, 283–300.
IUCN (2001) IUCN Red List Categories. Version 3.1, The World
Conservation Union, Gland.
Jankun A, Kovanda M (1988) Apomixis and origin of Sorbus
bohemica (Embryological studies in Sorbus no. 2). Preslia, 59,
97–116.
Koch MA, Dobes C, Mitchell-Olds T (2003) Multiple hybrid
formation in natural populations: concerted evolution of the
Internal Transcribed Spacer of nuclear ribosomal DNA (ITS)
in North American Arabis divaricarpa (Brassicaceae). Molecular
Biology and Evolution, 20, 338–350.
Lemche EB (1999) The Origins and Interactions of British Sorbus
Species, PhD thesis. Darwin College, Cambridge.
Liljefors A (1953) Studies on propagation, embryology and
pollination in Sorbus. Acta Horti Bergiani, 16, 277–329.
Liljefors A (1955) Cytological studies in Sorbus. Acta Horti
Bergiani, 17, 47–113.
Lo EY, Stefanović S, Dickinson TA (2009) Population genetic
structure of diploid sexual and polyploid apomictic
hawthorns (Crataegus; Rosaceae) in the Pacific Northwest.
Molecular Ecology, 18, 1145–1160.
McAllister HA (1986) The Rowan and Its Relatives (Sorbus spp.).
Ness Series 1. Ness Botanic Gardens, Liverpool.
Mes T (1998) Character compatibility of molecular markers to
distinguish asexual and sexual reproduction. Molecular
Ecology, 7, 1719–1727.
Nelson-Jones EB, Briggs D, Smith AG (2002) The origin of
intermediate species of the genus Sorbus. Theoretical and
Applied Genetics, 105, 953–963.
Nybom H, Esselink GD, Werlemark G, Vosman B (2004)
Microsatellite DNA marker inheritance indicates preferential
pairing between two highly homologous genomes in
polyploid and hemisexual dog-roses, Rosa L. Sect. Caninae
DC. Heredity, 92, 139–150.
Obbard DJ, Harris SA, Pannell JR (2007) Simple allelicphenotype diversity and differentiation statistics for
allopolyploids. Heredity, 97, 296–303.
Oddou-Muratorio S, Aligon C, Decroocq S, Plomion C, Lamant
T, Mush-Demesure B (2001) Microsatellite primers for Sorbus
torminalis and related species. Molecular Ecology Notes, 1, 297–
299.
Pielou EC (1969) An Introduction to Mathematical Ecology. WileyInterscience, New York.
Pigott CD, Walters SM (1954) On the interpretation of
discontinuous distributions shown by certain British species
of open habitat. Journal of Ecology, 42, 95–99.
Pring ME (1961) Biological Flora of the British Isles no. 78.
Arabis stricta Huds. Journal of Ecology, 49, 431–437.
Proctor MCF, Groenhof AC (1992) Peroxidase isoenzyme and
morphological variation in Sorbus L. in South Wales and
adjacent areas, with particular reference to S. porrigentiformis
E. F. Warb. Watsonia, 19, 21–37.
Proctor MCF, Proctor ME, Groenhof AC (1989) Evidence from
peroxidase polymorphism on the taxonomy and reproduction
of some Sorbus populations in south-west England. New
Phytologist, 112, 569–575.
Raspé O, Kohn JR (2007) Population structure at the S-locus of
Sorbus aucuparia L. (Rosaceae: Maloideae). Molecular Ecology,
16, 1315–1325.
1690 A . R O B E R T S O N E T A L .
Rich TCG, Houston L (2006) Sorbus whiteana (Rosaceae), a new
endemic tree from Britain. Watsonia, 26, 1–7.
Rich TCG, Jermy AC (1998) Plant Crib 1998. BSBI, London.
Rich TCG, Proctor MCF (2009) Some new British and Irish
Sorbus (Rosaceae) taxa. Watsonia, 27, 207–216.
Rich TCG, Harris SA, Hiscock SJ (2009) Five new Sorbus
(Rosaceae) taxa from the Avon Gorge, England. Watsonia, 27,
217–228.
Richards AJ (1975) Sorbus L. In: Hybridisation and the Flora of the
British Isle (ed Stace CA), pp. 233–238. Academic Press,
London.
Richards AJ (2003) Apomixis in flowering plants: an overview.
Philosophical Transactions of the Royal Society Series B, 358,
1085–1093.
Rieseberg LH (1997) Hybrid origins of plant species. Annual
Review of Ecology and Systematics, 28, 359–389.
Rieseberg LH, Willis JH (2007) Plant speciation. Science, 317,
910–914.
Robertson A, Sydes C (2006) Sorbus pseudomeinichii, a new
endemic Sorbus (Rosaceae) microspecies from Arran,
Scotland. Watsonia, 26, 9–14.
Robertson A, Newton AC, Ennos RA (2004a) Multiple hybrid
origins, genetic diversity and population genetic structure of
two endemic Sorbus taxa on the Isle of Arran, Scotland.
Molecular Ecology, 13, 123–143.
Robertson A, Newton AC, Ennos RA (2004b) Breeding systems
and continuing evolution in the endemic Sorbus taxa on
Arran. Heredity, 93, 487–495.
Rohlf FJ (2002) NtSYSpc, Numerical Taxonomy and Multivariate
Analysis System. Version 2.11a, User guide. Exeter Software,
New York.
Salmon A, Ainouche ML, Wendel JF (2005) Genetic and
epigenetic consequences of recent hybridisation and
polyploidy in Spartina (Poaceae). Molecular Ecology, 14, 1163–
1175.
Sepp B, Bobrova VK, Troitsky AK, Glazunova KP (2000)
Genetic polymorphism detected with RAPD analysis and
morphological variability in some microspecies of apomictic
Alchemilla. Annales Botanici Fennici, 37, 105–123.
Soltis DE, Soltis PS (1999) Polyploidy: recurrent formation and
genome evolution. Trends in Ecology and Evolution, 14, 348–352.
Stace CA (1997) New Flora of the British Isles, 2nd edn.
Cambridge University Press, Cambridge.
Stebbins GL (1950) Variation and Evolution in Plants. Columbia
University Press, New York.
Vrijenhoek RC (1984) Ecological differentiation among clones:
the frozen niche variation model. In: Population Biology and
Evolutio (eds Woermann K, Loeschcke V), pp. 217–231,
Springer, Berlin.
Warburg EF (1952) Sorbus L. In: Flora of the British Isle (eds
Clapham AR, Tutin TG, Warburg EF), pp. 539–556,
Cambridge University Press, Cambridge.
Warburg EF (1962) Sorbus L. In: Flora of the British Isle (eds
Clapham AR, Tutin TG, Warburg EF), 2nd edn. Cambridge
University Press, Cambridge.
Warburg EF, Kárpáti ZE (1968) Sorbus L. In: Flora Europaea,
Vol. 2 (eds Tutin TG et al.), pp. 67–71, Cambridge University
Press, Cambridge.
Whitton J, Sears CJ, Baack EJ, Otto SP (2008) The dynamic
nature of apomixis in the angiosperms. International Journal
of Plant Sciences, 169, 169–182.
Wilkinson M (2001) PICA 4.0: Software and Documentation,
Department of Zoology, The Natural History Museum,
London.
Wilmott AJ (1934) Some interesting British Sorbus. Proceedings
of the Linnean Society of London, 146, 73–79.
Ashley Robertson is a population geneticist who uses molecular markers to study genetic diversity and speciation in plants.
Timothy Rich is a plant taxonomist specializing in the taxonomy and evolution of critical plant groups in Britain. Alexandra Allen is a plant molecular geneticist working on plant
mating systems, particularly self-incompatibility. Libby Houston is a plant ecologist and all-round naturalist with expert
knowledge of the flora of the Avon Gorge. Cat Roberts is an
undergraduate student with interests in ecology and genetics.
Jon Bridle is a population geneticist and evolutionary biologist
with an interest in how organisms adapt and evolve in
response to ecological change. Stephen Harris is a population
geneticist working on many aspects of plant systematics.
Simon Hiscock is a plant geneticist with various interests in
plant evolutionary genomics, mating system evolution, hybrid
speciation, and polyploidy.
Supporting Information
Additional supporting information may be found in the online
version of this article.
Table S1–S17b Supporting information tables referenced in the
text (Word document)
Please note: Wiley-Blackwell are not responsible for the content
or functionality of any supporting information supplied by the
authors. Any queries (other than missing material) should be
directed to the corresponding author for the article.
2010 Blackwell Publishing Ltd