Upload
independent
View
0
Download
0
Embed Size (px)
Citation preview
ORI GIN AL PA PER
Independent hybrid populations of Formica polyctena Xrufa wood ants (Hymenoptera: Formicidae) aboundunder conditions of forest fragmentation
Bernhard Seifert • Jonna Kulmuni • Pekka Pamilo
Received: 8 October 2009 / Accepted: 15 February 2010 / Published online: 2 March 2010� Springer Science+Business Media B.V. 2010
Abstract Combined genetic and morphological data indicate frequent hybridisation
between the wood ants Formica polyctena Forster 1850 and F. rufa Linnaeus 1761 in
Central Europe. The genetic and morphological traits give a concordant picture of
hybridisation with a strong correlation between the genotypic admixture proportions at 19
microsatellite loci and the first vectors of a principal component analysis (P \ 0.001) and
of a 3-class discriminant analysis (P \ 0.001) of 15 quantitative morphological characters.
This integrative approach enabled a grouping into F. polyctena, the hybrid and rufa.
Genetic differentiation between the hybrid and F. rufa is significantly larger than between
the hybrid and polyctena, indicating gene flow mainly between the latter entities. A sug-
gested gene flow bias towards F. polyctena agrees with differential queen acceptance and
mating behaviour. Both genetic and phenotypic colony parameters indicate predominance
of monogyny in F. rufa but of polygyny in polyctena and the hybrid. Hybrids are inter-
mediate between the parental species in body size, diagnostic morphological characters,
monogyny frequency, size of nest population, nest diameter and infestation rate with
epizootic fungi. The three entities respond differently to woodland fragmentation. Hybrids
are significantly more abundant in forests with a coherent area\300 ha than in woodland
above this size. Regions with high hybrid frequency in Germany—the Eastern Oberlausitz
(23%) and the Baltic Sea islands Darss, Hiddensee and Rugen (28%)—are characterised by
a fragmented woodland structure whereas regions with low hybrid frequency—Branden-
burg and the lower Erzgebirge (3.4%)—have clearly larger and more coherent forest
Electronic supplementary material The online version of this article (doi:10.1007/s10682-010-9371-8)contains supplementary material, which is available to authorized users.
B. Seifert (&)Senckenberg Museum of Natural History Gorlitz, Am Museum 1, 02826 Gorlitz, Germanye-mail: [email protected]
J. KulmuniDepartment of Biology and Biocenter, University of Oulu, Box 3000, 90014 Oulu, Finland
P. PamiloDepartment of Biosciences, University of Helsinki, Box 65, 00014 Helsinki, Finland
123
Evol Ecol (2010) 24:1219–1237DOI 10.1007/s10682-010-9371-8
systems. Data from other European countries indicate habitat fragmentation to be a
facilitating factor but no essential precondition for interspecific hybridisation in these ants.
Hybrids are hypothesised to have selective advantage in fragmented systems because of
combining the main reproductive and dispersal strategies of the parental species.
Keywords Interspecific hybrids � Habitat fragmentation � Microsatellites �Morphometry � Integrative taxonomy
Introduction
The wood ant species Formica polyctena Forster 1850 and Formica rufa Linnaeus 1761
are important elements of temperate forest ecosystems of the West Palaearctic. They are
considered to give protection against a number of pest insects in natural and secondary,
managed forests (reviewed by Otto 1967) and are a symbol and main target of nature
conservation in many countries of Europe. Yarrow (1955) and Betrem (1960) considered
F. polyctena and F. rufa as clearly different species and this view had been generally
adopted for 35 years. The situation changed when Seifert (1991) published a compre-
hensive study on external morphology and biological parameters of 430 nests collected in
different regions of Central, East and North Europe. In addition to the typical F. polyctenaand F. rufa, he found a third entity which was intermediate in each investigated phenotypic
or biological character: body size, eight size-corrected pilosity characters, monogyny
frequency, size of nest populations, diameter of nest mounds and infestation rate with
epizootic fungi. He concluded that the third entity was a fertile hybrid between
F. polyctena and F. rufa.In Germany, the hybrid was particularly abundant in a landscape not poor in woodland
but with a high degree of forest fragmentation: as many as 27.4% of 212 investigated nests
were identified as hybrids in eastern Oberlausitz. In contrast, the hybrid was rather rare
(only 6.6% of 218 investigated nests) in regions with large, coherent forests. Attempting to
explain this finding, Hofener et al. (1996) simulated wood ant populations in fragmented
versus coherent woodland systems—parameters considered were distribution of Servifor-mica host nests, dispersal of colonies and queens, number and growth of worker and queen
population, mating and colony founding behaviour of queens, intranidal queen dominance
and acceptance, mortality of colonies and alates, exchange of nest populations within
polycalic systems and territorial fight. They argued that the hybrid can display a selective
advantage in fragmented woodland systems by combining the potency of F. rufa for
dispersal flight and single-queen socially parasitic colony foundation with the potency of
F. polyctena for shifting to polygyny, propagation by colony-fission and building up
locally dominant polydomous colonies.
Eighteen years later, the existence of a hybrid population remains to be the most
probable explanation for eastern Oberlausitz, but the alternative hypothesis that the
intermediate entity represents a morph of F. rufa with no hybridisation cannot be fully
excluded—the more as there has been no genetic analysis so far. In this paper, we want to
overcome this deficiency and present an investigation of nuclear DNA markers (micro-
satellites) in combination with a refined analysis of external morphology. We focus this
combined genetic and morphological approach on the ‘‘hot spot’’ of hybridisation in the
eastern Oberlausitz but will also apply the morphology-based method to describe the
situation in Europe as a whole. Furthermore we will reinvestigate the data of Seifert (1991)
1220 Evol Ecol (2010) 24:1219–1237
123
with more advanced data processing methods and integrate these and the current data in the
analysis of forest fragmentation with the aim to confirm Seifert’s hypothesis of higher
hybrid abundance in fragmented forest systems.
Materials and methods
Nest samples
Fifty nest samples (17 of Formica polyctena Forster 1850, 20 of F. rufa Linnaeus 1761, 12
of F. polyctena x rufa; one F. polyctena nest sampled in two years), collected in the eastern
Oberlausitz, Germany within an elliptic area of about 420 km2 were investigated both
morphologically and genetically (SI 1, online appendix).
A total of 618 Formica polyctena et rufa complex samples collected during the years
1962–2008 were evaluated by morphology alone (Table 1). Of these 148 were investigated
with the current system (see below) and 470 samples with the earlier investigation system
of Seifert (1991).
Table 1 Nest samples of the F. polyctena et rufa complex investigated in the exclusively morphology-based approach and determinations derived from this method
Country New data system Old data system
All poly pXr rufa All poly pXr rufa
Austria 1 1
Belgium 5 1 4
Czech Republic 1 1
England and Wales 19 1 16 2 3 2 1
Finland 14 9 1 4 2 1 1
France 1 1
Germany 84 33 17 34 418 154 80 184
Greece 3 3
Poland 4 1 2 1
Russia, Moscow 4 3 1 15 4 9 2
Russia, Voronesh 12 6 5 1
Spain 1
Sweden 2 1 1 19 11 2 6
Switzerland 10 3 5 2
All countries 148 51 45 52 470 176 96 199
‘‘New data system’’ applies to samples collected in 1981–2008 and subject to the current investigationsystem and ‘‘old data system’’ to samples taken in 1962–1990 and subject to the former investigation system(Seifert 1991). Because of the need to increase the number of hybrid samples, hybrid frequencies are muchlarger than those expected for random sampling—with the exception of realistic ratios in Finland andBritain. Acronyms: poly—F. polyctena, pXr—hybrid, rufa—F. rufa. Detailed collecting data are availableon request from the primary author
Evol Ecol (2010) 24:1219–1237 1221
123
Rationale of hybrid identification using the combined approach
The main rationale of hybrid identification included the following steps:
(1) The 3-class discriminant analysis (DA) of quantitative morphological data of the new
data system (see below) was used to form an initial hypothesis on whether the sample
could belong to the F. polyctena or F. rufa species.
(2) The gene pools of the parental species for estimating admixture frequencies were
defined by combining the morphological hypothesis and the Bayesian mixture
analysis of genotypes.
(3) We then compared genetic classifications with morphological data using both
hypothesis-driven (DA) and unsupervised (PCA) approaches to improve hybrid
identification and to re-evaluate conflicting determinations.
DNA extraction and amplification
DNA was extracted from five workers per nest with DNAeasy Tissue Kit (Qiagen) using
the manufacturer’s protocol designed for insects. We genotyped a total of 19 microsatellite
loci per individual. The microsatellite primers used were designed for F. exsecta(Gyllenstrand et al. 2004) FE7, FE13, FE17, FE19, FE37, FE38, FE42, for F. paralugubris(Chapuisat 1996) FL12, FL20, FL29 and for F. yessensis (Hasegawa and Imai 2004) FY3,
FY4, FY5, FY7, FY9, FY10, FY12, FY13, and FY15. Genotypes were assayed by poly-
merase chain reaction with fluorescent labelling using Peltier Thermal Cycler-200-PCR
equipment (MJ Research). Microsatellite loci were amplified by initial round of denaturing
at 94�C for 2 min, followed by 34 cycles of 94�C for 30 s, 48–65�C (depending on primer
pair; SI 2 of online appendix) for 30 s, 72�C for 30 s, with a final step of 72�C for 30 min.
PCR reactions were performed in a reaction volume of 10 ll containing 10 9 Mg2? -free
dyNAzymeTM buffer (Finnzymes), 250 lM of each dNTP, 0.5 lM labelled primer,
0.5 lM unlabelled complementary primer, 0.5 U dyNAzymeTM II DNA polymerase
(Finnzymes),1–3.5 lM MgCl2 (Finnzymes; depending on primer pair; SI 2 of online
appendix),1 ll template (genomic DNA of unknown concentration) and H2O up to final
volume. Genotypes were resolved by capillary electrophoresis (3730 DNA Analyzer,
Applied Biosystems) using 500 LIZ size standard (Applied Biosystems) and scored with
Genemapper v.4.0 (Applied Biosystems).
Analysis of the genotype data
Admixture proportions were estimated for each individual on the basis of their
microsatellite genotypes by using a Bayesian admixture model in the program BAPS
(Corander and Marttinen 2006). We first used the mixture model of BAPS to delineate
genetically homogeneous groups by forcing the number of groups to be three. The
groups represented well morphologically identified F. polyctena, F. rufa and the hybrids.
One nest which was morphologically identified as hybrid but genetically clustered with
F. polyctena was removed from the F. polyctena cluster in order to minimize the
presence of possible hybrids in the two parental gene pools which were then used in the
admixture analysis. Hybridisation was assessed also by using a Bayesian estimation of
parental and hybrid genotypes without any information on the morphological clustering.
This was done using the program NewHybrids (Anderson and Thompson 2002) which
aims to infer recent hybrids (F1, F2, backcrosses) from the genotypic data (the number
1222 Evol Ecol (2010) 24:1219–1237
123
of sweeps was 20,000 for burnIn and 100,000 for the data gathering). The power of the
program NewHybrids to detect possible hybrids was evaluated by computer simulations,
following the recommendations by Vaha and Primmer (2006). The simulations were
carried out by assuming that the morphologically identified F. polyctena and F. rufarepresent the parental gene pools. Simulated genotypes including the two parental species
(30% each), F1 and F2 hybrids, and both types of backcrosses (10% of each category)
were formed and then analysed by NewHybrids. Simulations were repeated ten times.
The simulated data showed that NewHybrids finds F. rufa with a high precision (effi-
ciency and accuracy [90%) but tends to mix F. polyctena and the hybrids and does not
well separate different types of hybrids from each other. We present the NewHybrids
results by pooling the four hybrid classes and calculating for each individual the overall
probability that it is a hybrid. Furthermore, principal coordinate analysis (program PCO,
Andersson 2003) based on pairwise genetic distances between pairs of nests was used to
examine the genetic relationships among the nests. Genic differentiation between groups
was estimated as FST by using the algorithm of Weir and Cockerham (1984). The
significance of the pairwise estimates was tested by permutation of nests (1,000 times).
Genetic relatedness among worker nest mates was estimated using the program GENREL
(Pamilo 1984).
Morphological data recording
A detailed description of morphological data recording allowing reproducibility is given in
SI 3 of the online appendix. Here we only inform on the basic principles. Analysis of the
individual composition of nest samples was performed to check if intermediate sample
means (=hybrid indication) might be caused by intranidal coexistence of pure parent
phenotypes. The basic unit for the identification of the three entities was nest sample means
because only the same nests but not the same individuals were subject to a parallel
morphological and genetic investigation. An average of six fully dried workers per nest
mounted on a pinned cardboard were evaluated for fifteen numerically scored phenotypic
characters. In order to save working time, sample size was occasionally reduced to three
when prior visual inspection showed a monomorphic worker population but it was
extended to nine to have more accurate sample means in case of polymorphic nest
populations. All measurements and counts were made using a high-performance stereo-
microscope Leica Wild M10 equipped with a 1.6 9 planapochromatic front lens at
160–320 9 magnification. Standard positioning of the specimens was achieved under use
of a pin-holding stage permitting full 3-dimensional rotations.
Seifert (1991) described the characters and data recording used in the earlier morpho-
logical investigation system. The main difference of the 1991 investigation to the current
system was the use of unmounted, ethanol-stored specimens, a smaller and partially
deviating character set but higher average numbers of investigated specimens per nest. This
did not allow pooling the 1991 and current data in a combined morphological analysis.
However, the determination of species with these separate analyses was used in the
analysis of the effects of forest fragmentation.
Removal of allometric variance
In most species groups of Formica, morphological characters are strongly influenced by
allometric growth. In order to make characters directly comparable in synoptic tables, a
Evol Ecol (2010) 24:1219–1237 1223
123
removal of allometric variance (RAV) was performed with the procedure described by
Seifert (2008). RAV was calculated assuming that all individuals have an identical
cephalic size of 1.75 mm. We applied RAV functions in which the collective parameters
were calculated as the arithmetic mean of the species-specific functions of twelve Palae-
arctic Formica rufa group species. Evaluation of scatter plots indicated the use of mon-
ophasic linear RAV functions (online appendix SI 4).
Processing of morphological data
Canonical discriminant analyses (DA) and principal component analyses (PCA) of abso-
lute head size CS and the fourteen RAV-corrected characters were run using the SPSS 10.0
statistical package. A parallel run of an ordinary DA and of a ‘‘leave-one-out cross-
validation’’ DA (LOOCV-DA, Lachenbruch and Mickey 1968, Lesaffre et al. 1989) was
performed to realistically estimate the error rate. The data presented by Seifert and Schultz
(2008) show that the mean of the pessimistic error indication by the LOOCV-DA and of
the optimistic error indication by the ordinary DA—here called integrated error rate—is
close to the true error.
The starting point was a supervised analysis by a DA assuming three classes:
F. polyctena, F. rufa and putative F. polyctena x rufa. If a run of a DA rejected an a priorihypothesis of a sample, the new hypothesis was taken up and iterative runs were performed
until the indicated error rate achieved a minimum. This partially self-organising approach,
in which only the starting situation is set by a hypothesis of the investigator, was applied to
reduce the subjective component and led to changed determinations in a number of
samples. All 148 European samples were included to increase the mean sample size per
entity to a number required for a reasonable application of a DA.
The data of Seifert (1991) were reinvestigated by the same analysis systems but without
RAV corrections.
Estimation of monogyny frequency, nest diameter, nest-population size
and fungal infestation
Monogyny of specific nests was determined by a modification (Seifert 1991) of the
function of Otto (1960) which uses arithmetic mean, skewness and kurtosis of the head
width distribution of 30–170 workers randomly collected from the mound surface.
Nest diameter refers to the perimeter described by the most peripheral nest entrances.
These are either at mound base or within the outer ring of soil ejections. In nests with an
elliptic basal area, the arithmetic mean of the large and small diameter is recorded.
The nest-population size figure A is a dm2-estimate of nest surface covered by ants
during the main activity period from late April to late September under exclusion of
climatic situations with activity depression. As a rule the estimate was performed at air
temperatures at the nest site between 15� and 22�C. A is the estimated per cent ratio of nest
surface covered by ants multiplied with the whole nest surface calculated by geometric
formulae.
The infection with epizootic fungi was recorded during morphological inspection. There
was no species determination by mycologists but it is generally assumed that ants of the
Formica rufa group are parasitised by Aegeritella superficialis Balazy et Wisniewski 1974
and closely related species and to a lesser degree by Erynia sp. (Wisniewski 1976, 1977;
Espadaler and Wisniewski 1987; Espadaler and Monteserın 2003).
1224 Evol Ecol (2010) 24:1219–1237
123
Estimation of woodland fragmentation
Woodland fragmentation is inversely proportional to the mean size of coherent woodland
areas in a given area. Area was considered coherent if the woodland patches were linked by
woodland stripes of at least 15 m width whereas lines of single trees along streets and
highways were not regarded as linking elements. Any woodless agricultural, natural, set-
tlement or industrial area separating forest patches by more than 30 m was considered to be
an isolating structure. Motor highways, railway lines and rivers wider than 15 m were also
considered as fragmenting elements. This method pays attention to structures which rep-
resent likely barriers for ground movements of wood ant nests but not to spots which mated
gynes can reach by dispersal flight. Forest area was estimated by summing up areas of
approximated geometric forms (triangles, parallelograms, ellipses etc.) under use of the air
photographs and ruler tool provided by the Google-Earth� 2009 Tele Atlas.
The analysis of woodland fragmentation was only done for German sites with exact
geographic information. We pooled the old data of Seifert (1991), including 411 samples
collected between 1979 and 1989, with data of the new investigation giving a total of
495 samples. We further added 58 samples which were determined by subjective assess-
ment. Since we morphotyped only few nests from phenotypically unambiguous
F. polyctena supercolonies in larger forests but usually all nests in very small forest
patches, the overall abundance of this species in larger forests is strongly underestimated
by the morphometric data set in Table 1.
The analysis included a total of 127 German forest areas differing in size between 0.01
and 65,000 ha. Three patches within forest areas\100 ha were oversampled because of a
detailed local study: Liebstein West (n = 44 samples), Grossmachnow (n = 21) and
Deutsch-Paulsdorf (n = 17). Considering the fact that the maximum sample number in
patches without oversampling was 15 and in order to reduce a bias of data in regional or
overall analyses, the data from these three patches were given a weighting factor of
w = 15/n (i.e., 0.34, 0.71 and 0.88). The weighting factor in the 124 remaining patches
was 1.0. The chi-square test of independence (X2 test) to check the frequency of the hybrid
against those of the parental species was performed as recommended in Sokal and Rohlf
(1995).
Results
Initial definition of morphospecies
Using the current data pool, a self-organizing, iterative 3-class DA stabilised at an inte-
grated error rate of 1.3% and discriminated 93.2% of the 148 European samples with
posterior probabilities of P [ 0.95 (Fig. 1). The rather high 6.8% proportion of cases with
P \ 0.95 was caused by unclear determinations within the hybrid/parental species tran-
sition zones which are unavoidable if hybrids are fertile and backcross with both parental
species.
Using the data pool of Seifert (1991), an iterative 3-class DA stabilised at an integrated
error rate of 0.7% and discriminated 87.2% of the 470 European samples with posterior
probabilities of P [ 0.95. Compared to the determinations by Seifert (1991), the new type
of analysis resulted in the following changes: 1.5% reduction in F.polyctena, 3.6% increase
in the hybrid and 2.1% reduction in F. rufa. All changes occurred in the transition zone
between the hybrid and the parental species.
Evol Ecol (2010) 24:1219–1237 1225
123
Bayesian clustering of microsatellite genotypes
The nineteen loci had a total of 120 alleles (95 in F. polyctena and 88 in F. rufa). The mean
heterozygosity per locus was 0.53 in F. polyctena, 0.49 in F. rufa and 0.56 in the hybrids.
The parent species had no clear diagnostic allele differences, and the amount of differ-
entiation between them was FST = 0.13 ± 0.033 (P \ 0.001). The distance of putative
hybrids to F. polyctena was FST = 0.05 ± 0.012 (P \ 0.001) and to F. rufa FST = 0.10 ±
0.021 (P \ 0.001).
Genetic clustering of the nests with a Bayesian mixture analysis agreed well with the
morphological identification (Table 2). Sample No 128 was genetically clustered together
with F. polyctena but because it was morphologically identified as hybrid we moved it to
the hybrid group in further analyses. Otherwise, the group which was genetically con-
sidered to be hybrids included also some nests which were morphologically identified as
either F. polyctena or rufa (Table 2). We calculated for each individual two estimates for a
hybrid probability, the admixture proportion with BAPS and the overall hybrid probability
with NewHybrids. The two estimates show a significant correlation (Spearman rank cor-
relation rs = 0.69, df = 245, P \ 0.001). Using the morphological classification of nests,
the median admixture proportions differed significantly from each other between the
groups (Mann–Whitney’s test t?[ 8.3, P \ 0.001).
We compared morphological identification and genetic admixture proportions on the
level of nest means (Table 2; Fig. 2). This bears the risk of a wrong hybrid indication in
case of coexistence of both parent species in the same nest. This risk was found to be very
low. According to morphotyping, there were only 1.3% of 148 nests with mixtures of
parental species in the whole W Palaearctic range and within the 50 genetically typed
samples from the Oberlausitz only nest No 125 matched this condition. It contained 3
F. polyctena, 1 hybrid and 2 F. rufa workers according to morphotyping and 1 F.polyctena, 2 hybrids and 2 F. rufa according to genotyping (note that morpho- and
genotyping used same nests but different individuals!).
Fig. 1 Canonical discriminant analysis of 15 phenotypic characters of worker nest sample means ofFormica polyctena (black rhombs), F. polyctena x rufa (grey triangles) and F. rufa (white squares) of 148samples from the entire European range
1226 Evol Ecol (2010) 24:1219–1237
123
Table 2 Initial morphological grouping by a 3-class DA, genetic grouping (k = 3) and percentage ofF.polyctena alleles determined by Bayesian clustering with BAPS, within-nest relatedness according to theGENREL programme and hybrid probability estimated by NewHybrids in 49 samples of the Formicapolyctena et rufa complex in the eastern Oberlausitz, Saxony, Germany. The percentage of predicted F. rufaalleles is complementary and not given. Data based upon 5 workers per nest sample
No. Site Morphol.group
Geneticgroup
Admixture(% polyctena,BAPS)
Related-ness
Hybrid probability(NewHybrids)
32 Deutsch-Paulsdorf (1996) poly poly 97.0 0.15 0.03
35 Deutsch-Paulsdorf (1996) poly poly 90.2 0.21 0.14
36 Deutsch-Paulsdorf (1996) poly poly 89.2 0.29 0.13
37 Deutsch-Paulsdorf (1996) poly poly 97.6 0.48 0.05
198 Diehsa (2005) poly poly 99.0 0.34 0.03
199 Diehsa (2005) poly poly 99.8 0.31 0.04
200 Diehsa (2005) poly p X r 76.2 0.26
201 Diehsa (2005) poly poly 73.6 0.22 0.71
19 Ebersbach (1996) poly poly 99.0 0.26 0.02
129 Ebersbach (1997) poly p X r 88.6 0.08
1 Girbigsdorf (1996) poly p X r 76.8 0.11
1 Girbigsdorf (2005) poly p X r 69.8 0.30
17 Gorlitz, cemetery (1996) poly poly 92.6 0.21 0.19
18 Gorlitz, cemetery (1996) poly poly 94.8 0.47 0.09
3 Konigshain (1996) poly poly 88.0 0.31 0.45
5 Konigshain (1996) poly poly 92.8 0.22 0.32
6 Konigshain (1996) poly poly 94.4 0.15 0.23
21 Ludwigsdorf (1996) poly poly 97.0 0.68 0.03
31 Deutsch-Paulsdorf (1996) p X r p X r 80.4 0.02
33 Deutsch-Paulsdorf (1996) p X r p X r 69.4 0.18
34 Deutsch-Paulsdorf (1996) p X r p X r 62.8 0.25
125 Ebersbach (1997) p X r p X r 36.0 0.48
128 Ebersbach (1997) p X r p X r 73.0 0.20
9 Liebstein West (1996) p X r p X r 59.4 0.36
10 Liebstein West (1996) p X r p X r 53.0 0.21
11 Liebstein West (1996) p X r p X r 82.4 0.00
26 Liebstein West (1996) p X r p X r 72.6 0.06
27 Liebstein West (1996) p X r p X r 56.2 0.32
28 Liebstein West (1996) p X r p X r 73.8 0.02
29 Liebstein West (1996) p X r p X r 74.0 0.17
196 Diehsa (2005) rufa rufa 0.0 0.67 0.00
197 Diehsa (2005) rufa rufa 8.4 0.68 0.02
202 Diehsa (2005) rufa rufa 0.6 0.76 0.11
203 Diehsa (2005) rufa rufa 6.4 0.18 0.28
204 Diehsa (2005) rufa rufa 0.0 0.79 0.00
20 Ebersbach (1996) rufa rufa 4.6 0.59 0.03
22 Ebersbach (1996) rufa rufa 5.6 0.31 0.23
23 Ebersbach (1996) rufa rufa 3.4 0.24 0.11
126 Ebersbach (1997) rufa rufa 0.0 0.56 0.00
Evol Ecol (2010) 24:1219–1237 1227
123
Regarding intranidal relatedness, we refrained from presenting estimates for the hybrid
nests because it is difficult to define the proper reference population. Within-nest relat-
edness was estimated as r = 0.60 (SE = 0.04) in 19 F. rufa nests and r = 0.31
(SE = 0.04) in 14 F. polyctena nests (Table 2)—after removing the effects of inbreeding,
the estimates changed to r* = 0.45 in F. rufa and r* = 0.00 in F. polyctena. The results
indicate that F. rufa nests in the Oberlausitz are largely monogynous or have only a small
number of queens, whereas F. polyctena is almost completely polygynous. This is in line
with monogyny frequencies determined by combining Otto-function estimates with data
from field observations (Seifert 1991). Extracting only the Oberlausitz subsample from
Seifert’s data pool, monogyny frequencies in this region are 78.6% in 70 F. rufa nests,
13.7% in 51 hybrid nests and 4.0% in 75 F. polyctena nests.
Matching of genetic and morphologic data
Genetic clustering based on admixture proportions was highly correlated with morpho-
logical clustering performed with both the hypothesis-driven and explorative approaches.
The 1st canonical vector of a DA of the all-European morphological data set is highly
correlated with the proportion of F. polyctena alleles (r = -0.900, P \ 0.0001, Fig. 2) as
is the 1st factor of a PCA (r = -0.845, P \ 0.0001, data not shown). The picture of strong
correlation of morphology and genetics is only disturbed by sample No 130 which has
44.6% F. polyctena alleles (Fig. 2). This sample is genetically a hybrid both in the mean
and in individual data of all five workers; in contrast all five morphotyped workers are very
clear F. rufa. This mismatch is difficult to explain; we cannot exclude a confusion of
samples. Anyway, a mismatch ratio of only 2% between morphology and genetics is a very
good result for a hybrid scenario with parent species not having private alleles. We decided
to use the morphological determination of sample No 130 in any of the results presented
below.
A principal coordinate analysis (PCO) based on pairwise FST estimates between the
nests treated the genetic data in a very different way and is thus an independent approach.
The first two axes of PCO explained somewhat over 50% of all the genetic variation. The
Table 2 continued
No. Site Morphol.group
Geneticgroup
Admixture(% polyctena,BAPS)
Related-ness
Hybrid probability(NewHybrids)
127 Ebersbach (1997) rufa rufa 7.7 0.76 0.05
205 Groditz E (2005) rufa rufa 0.2 0.50 0.01
206 Groditz E (2005) rufa rufa 9.2 0.76 0.04
2 Konigshain (1996) rufa rufa 6.0 0.45 0.02
4 Konigshain (1996) rufa rufa 4.4 0.82 0.01
7 Konigshain (1996) rufa rufa 0.2 0.61 0.02
124 Liebstein South (1997) rufa rufa 0.0 0.72 0.00
131 Liebstein South (1997) rufa rufa 0.0 0.80 0.00
132 Liebstein South (1997) rufa rufa 0.8 0.75 0.01
218 Liebstein South (2005) rufa rufa 0.0 0.45 0.00
130 Liebstein South West (1997) rufa p X r 44.6 0.83
1228 Evol Ecol (2010) 24:1219–1237
123
morphotypes of F. polyctena and F. rufa are well separated along the first axis, and the
hybrid morphotypes are partly intermediate and partly overlap with F. polyctena (Fig. 3).
This confirms the indication by the Bayesian analysis that more than 50% of the hybrid
genomes are from F. polyctena. There are several possible factors which could explain this
asymmetry—first of all selection for genotypes and dilution of F. rufa alleles by
Fig. 2 First canonical vector of a discriminant analysis of 15 quantitative morphological characters (largelyseparating the parental species) plotted against the percentage of F. polyctena alleles obtained from aBayesian admixture analysis of 19 microsatellite loci. The percentage of F. rufa alleles is complementary
Fig. 3 Principal coordinate analysis of pairwise genetic distances between nest samples. The two shownaxes explain a little over 50% of variation found in 19 microsatellite loci. The first axis indicates differencesbetween the species and the 2nd axis differences between the nests within a species. The much largervariation along the second axis in F. rufa is probably resulting from the much higher level of monogyny
Evol Ecol (2010) 24:1219–1237 1229
123
differential mating behaviour. The second axis separates mainly F. rufa nests, probably
because there are stronger genetic differences due to monogyny.
Intermediate position of hybrids in colony parameters
The intermediate position of hybrids is also evident in colony parameters (Table 3). The
hybrids differ significantly from the parental species in the size of nest population, nest
diameter and infestation ratio with epizootic fungi. These characters are largely a conse-
quence of colony structure: growing queen number increases the size of nest populations
and nest diameter while propagation by nest-splitting favours spreading of fungal
infestations.
Increased hybrid frequency in regions with woodland fragmentation
The distributions of F. polyctena, F. rufa and the hybrids in forests smaller or larger than
300 ha differ from each other (Table 4). The hybrid is found in Germany relatively more
often in small forest patches than F. polyctena (X2 13.82, P \ 0.001) and F. rufa (X2 14.82,
P \ 0.001). The distribution of the parental species is not different under this classification
(X2 0.05, P = 0.459). When classifying the forest patches into ‘‘very small’’ (\5 ha),
‘‘small’’ (5–299 ha) and ‘‘large’’ (300–65000 ha), the distribution of the hybrid differs
Table 3 Biological data of F. polyctena et rufa complex species from entire Germany. Significance testingof per cent ratios by X2 test, of nest diameter and population size by a two-tailed t test; n = number ofinvestigated nests
Monogyny Nest diameter(cm)
Population sizefigure (dm2)
Ascomyceteinfestation
F. polyctena 3.0% (n = 135) 201 ± 123 (n = 95) 66.8 ± 88.6 (n = 102) 26.1% (n = 153)
F. polyctena vs.hybrid
P \ 0.039 P \ 0.001 P \ 0.001 P \ 0.056
Hybrid 11.7% (n = 77) 131 ± 68 (n = 70) 24.2 ± 25.0 (n = 69) 14.9% (n = 74)
Hybrid vs. F. rufa P \ 0.001 P \ 0.001 P \ 0.001 P \ 0.001
F. rufa 81.0% (n = 153) 91 ± 40 (n = 136) 7.1 ± 8.7 (n = 144) 1.7% (n = 177)
Table 4 Frequency of the three entities of the F. polyctena et rufa complex on sampling spots embeddedwithin different areas of coherent woodland given as absolute numbers corrected for oversampling and aswithin-species percentages
Area interval of coherentwoodland (ha)
Mean areasize (ha)
Number ofareas
Hybrid F.polyctena F. rufa
Two forest area classes
[0,300) 60.0 74 58.6 (= 72.7%) 109.1 (= 48.70%) 100.9 (= 47.6%)
[300,65000] 11538 58 22.0 (= 27.3%) 115.0 (= 51.3%) 111.0 (= 52.4%)
Three forest area classes
(0,5) 1.5 27 16.0 (= 19.9%) 19.0 (= 8.5%) 37.0 (= 17.5%)
[5,300) 77.9 47 42.6 (= 52.9%) 90.0 (= 40.2%) 63.9 (= 30.2%)
[300,65000] 1,1538 58 22.0 (= 27.3%) 115.0 (= 51.3%) 111.0 (= 52.4%)
1230 Evol Ecol (2010) 24:1219–1237
123
clearly from both F. polyctena (X2 16.47, P \ 0.001) and F. rufa (X2 16.53, P \ 0.001) but
also the parent species differ (X2 9.95, P = 0.007).
Discussion
The intermediate position of hybrids and their relation to the parental species
Different lines of evidence provide a concordant and convincing indication for hybridisation
between F. rufa and F. polyctena. The clearest argument is the high correlation of mor-
phological clustering by DA or PCA with genetic clustering based on the Bayesian analysis
(Fig. 1). The intermediate genetic position of morphological hybrids is also supported by the
FST values, the PCO (Fig. 3) and by intermediate biological features (Table 3). Furthermore
the morphological hybrids are genetically intermediate at the individual level (significant
differences of individual admixture proportions in Mann–Whitney test). It should be noted
that the parental pools of F. polyctena and F. rufa shared allelic polymorphisms without clear
diagnostic alleles—the genetic differences were thus frequency differences. If there are
different levels of introgression over many generations, it may become difficult to draw clear
boundaries between the hybrid and the parental species. Considering these limitations, the
clear separation of the three entities (Fig. 1) is remarkable.
Hybrids are intermediate when considering all morphological characters but they
approach F. rufa in the two characters nGU and GuHL which most strongly contribute to
the separation of the parent species in a DA (online appendix, SI 5). This has led to the
alternative hypothesis that the hybrid phenotype could represent a morph of F. rufa (one of
the reasons to conduct this study). This hypothesis is clearly rejected by the genetic
analysis. Genetically, hybrids are closer to F. polyctena as indicated by their mean
admixture proportion of 66.3% F. polyctena alleles (different from 50% with t = 4.2,
df = 9, P \ 0.01), by the FST values and by the significant overlap of the hybrid and
F. polyctena cluster in the PCO (Fig. 3). From these data it seems likely that backcrossing
of the hybrid and introgression occurs mainly with the F. polyctena parent. Biased gene
flow towards F. polyctena could be expected on the basis of differential queen acceptance
and mating behaviour of the species (Gosswald 1942, 1981; Seifert 1991) and is also
directly suggested by a microsatellite analysis from Sweden (Gyllenstrand et al. 2004). The
following points seem important in this context:
(a) There is a trend in Formica and in ants in general that workers in polygynous nests
are less aggressive against invading mated queens (e.g., Gosswald 1942; Buschinger
1970; Pisarski 1982; Stuart et al. 1993; Sundstrom 1993, Lecat et al. 2008).
According to Seifert (1991), the frequency of polygynous nests increases in Germany
as a whole from F. rufa (14.1%) to the hybrid (85.7%) and to F. polyctena (97.6%).
Consequently, queen acceptance (and gene flow) should be easier from F. rufa to the
hybrid and to F. polyctena and from the hybrid to F. polyctena.
(b) Gynes of monogynous F. rufa and hybrid nests are larger and physically stronger
than those of polygynous societies (Seifert 1991) and they are supposed to have a
stronger influence on workers in having a more attracting or appeasing pheromonal
system = ‘‘grossere duftliche Dominanz’’ (Gosswald 1981). Both factors give them
a higher fitness during socially parasitic colony foundation in Serviformica but could
also facilitate their adoption in polygynous nests of the F. polyctena et rufa complex
in which weaker gynes are present.
Evol Ecol (2010) 24:1219–1237 1231
123
(c) Gynes and males of the polygynous societies of the hybrid and F. polyctena show a
high frequency of intranidal mating and a reduced tendency to perform a swarming
flight to external mating places (Seifert 1991). In contrast, males and gynes of
monogynous F. rufa nests have a mating flight (Seifert 1991).
(d) Once a F. rufa queen has been accepted in a polygynous nest of a hybrid or of
F. polyctena, her daughters can mate intranidally in the next year and stay as
reproductives in the nest. Such a scenario seems likely since males of polygynous
societies of the hybrid and F. polyctena have regularly been seen to show excessive
and unselective mating behaviour trying to copulate with any female ant on the
mound including their sterile worker nest mates.
The whole data set presented above is best explained when assuming fertility of hybrids
and their enduring existence independent from the parent species. The strongest direct
indication of this is given by their very strong dominance in some isolated woodland
patches. A good example is the locality Liebstein West. According to the data of Seifert
(1991), hybrids accounted for 85% of the worker population and 76% of the nests within
34 wood ant nests in this isolated 7.5 ha woodland patch in 1987/1988—the remaining ants
belonged to F. rufa while F. polyctena was completely absent. The current morphologic
and genetic analyses of 9 and 7 nest samples, respectively found only hybrids in Liebstein
West (Table 2)—undoubtedly hybrids could not have persisted with such a dominance
over 20 years unless fertile and independent from parent species. From an evolutionary
point of view it would be most interesting to check if the whole population in this patch
consists of hybrids and to estimate the genetic contributions of parental species in the
present gene pool.
Hybrid frequency and habitat fragmentation
The underestimation of F. polyctena abundance in large forests (see section Estimation of
woodland fragmentation) increases our estimates of relative hybrid abundance in unfrag-
mented woodland areas. Nevertheless, the data show a significant increase of hybrid fre-
quency with growing woodland fragmentation (P \ 0.001 both against F. polyctena and
F. rufa) when subdividing the forests in two size classes (below and above 300 ha,
Table 4). A subdivision into three size classes—very small, small and large—also showed
significant differences in the overall hybrid distribution compared to those of the parent
species (P \ 0.001 both against F. polyctena and F. rufa). However, the details should be
more complicated: if isolated forest patches become very small, such patches can only be
colonised by flight-dispersal of queens establishing monogynous nests after accidental
extinction of the resident wood ant population (Seifert 1991; Maki-Petays et al. 2005).
Considering the known monogyny frequencies of the three entities (Table 3), the fre-
quency in very small forest patches should be low for F. polyctena, moderate for the hybrid
and largest for F. rufa. The data in Table 4 support this idea: there is a significant
advantage of the hybrid over F. polyctena in very small compared to large patches
(X2 14.24, P \ 0.001). A significant advantage of F. rufa over F. polyctena exists in very
small compared to small patches (X2 = 9.91, P = 0.002) and in very small compared to
large patches (X2 = 5.17, P = 0.023). However, there is no indication that the hybrid is
less abundant than F. rufa in very small patches. This can be seen as an indication of a
well-developed colonising ability of the hybrids in highly fragmented systems as supposed
by Seifert (1991) and Hofener et al. (1996).
1232 Evol Ecol (2010) 24:1219–1237
123
A very high hybrid abundance (23% out of 291 weighted samples) was found in the
Oberlausitz whereas this proportion was only 6% in 227 weighted samples from outside
this area. The main hybridisation area in the Oberlausitz is characterised by a very patchy
and highly differentiated landscape: there is a big number of rocky hills of very different
size within large areas of fertile agricultural land. As a rule, only these hills are covered by
woodland which causes a high degree of woodland fragmentation (Fig. 4). Another region
in Germany with an increased hybrid frequency (28% within 25 samples) is the landscape
complex of the Baltic Sea islands Darss, Hiddensee and Rugen where forest fragmentation
is comparable to Oberlausitz. In contrast, only 3.4% of hybrids were found within 145
samples from Brandenburg and the lower Erzgebirge where larger forests predominate.
Other regions or localities in Europe with high hybrid frequencies were: England and
Wales (82% within 22 samples, collected between 1975 and 2005), the region of Schlatt/
Kanton Zurich (50% within 10 samples, collected 2005), the forest at the Zvenigorod
Biological Station 56 km W of the centre of Moscow (47% within 19 samples, collected
1985) and the Voronesh Nature Reserve in Russia (42% within 12 samples, collected in
1962 by G.M. Dlussky). England and Wales provide an excellent example of strong
woodland fragmentation over several hundred years. The landscape structure near Schlatt
in Switzerland, however, is ambiguous in respect to our hypothesis: forest area is here
strongly structured (with many protrusions and peninsular areas) but according to our
definition it is contiguous for about 1500 ha and the basic woodland distribution has been
stable for more than 100 years. The situation near Zvenigorod does not support our
hypothesis: there are about 2,400 ha of contiguous, compact forest area and this situation
had been stable during the twentieth century (G. Dlussky, pers. comm. 2009). The exact
position of the collecting site in the Voronesh Nature Reserve is unclear, but the area is
remarkable in that it is the only big forest within a huge area of agricultural steppe. The
Fig. 4 Detail of the woodland structure in the hybridisation zone in the Eastern Oberlausitz. Image widthequals 15.3 km. Woodland area given in blackish grey (= dark green in the online version), settlement areain medium grey, agricultural and grassland area in lighter greys and white. Black (= red in the onlineversion) sections in circle diagrams indicate hybrid frequencies for four patches with hybridfrequencies [ 35%
Evol Ecol (2010) 24:1219–1237 1233
123
forest was heavily logged before (and after) it became protected in 1927 (UNESCO—
MAB Bioreserves Directory 2009).
Habitat fragmentation and anthropogenous impact on landscape structure have long
been considered to promote interspecific hybridisation in plants and animals (e.g.,
Anderson 1948; Stebbins 1959, 1980; Pearson 1983; Levin et al. 1996; Seifert 1999;
Simberloff 2008). The conclusion from our wood ant data is that forest fragmentation
facilitates hybridisation but is not a necessary precondition as suggested by the Zvenigorod
example. Hofener et al. (1996) explored plausible selective forces behind this process in
computer simulations but they also stated that better data on swarming, dispersal flights
and intranidal dominance behaviour of queens are urgently needed.
Evolutionary and taxonomic implications
Fertile hybridisation between concordantly recognised species has been reported in many
groups of animals (reviewed by Mallet 2006). Ducks and geese (Anatidae) offer a phe-
notypically and numerically most impressing example. Here, postzygotic isolation is often
completely lacking, even intergeneric hybridisation between dramatically different species
is no exception and 76% of 21 British duck species hybridise in the wild (Millais 1902,
1913; Phillips 1915, 1921; Gillham and Gillham 1996). According to Price and Bouvier
(2002), bird hybrids become infertile after average divergence times of 5 myr (passerines)
and 17 myr (nonpasserines).
In contrast to ducks and geese, wood ants of the genus Formica are morphologically
difficult to identify and hybrid phenotypes are likely to escape our subjective perception.
It is therefore likely that the frequency of hybridisation in these ants may be higher than
reported. In addition to the case reported here, credible examples for interspecific
hybridisation detected by genetic or combined genetic and morphological evidence come
from Formica aquilonia Yarrow 1955 9 F. polyctena (Goropashnaya et al. 2004,
Saapunki et al. 2008), F. paralugubris Seifert 1996 9 F. aquilonia (Bernasconi 2009)
and F. pratensis Retzius 1783 9 F. lugubris Zetterstedt 1838 (Seifert and Goropashnaya
2004). In addition, combined morphological and chorological evidence (unpublished
results of B.Seifert) indicates at least another two cases: hybridisation between
F. lugubris and F. rufa in southern Finland and one case of F. truncorum Fabricius 1804
crossing with an indeterminate species (probably F. rufa) in southern Bavaria. These data
document hybridisation in seven out of the nine (78%) currently recognised European
species.
Even if so many Formica species can hybridise, the frequency of hybrids is generally
small in natural populations. Within 250 carefully examined nest samples of F. pratensisand another 150 samples of F. truncorum from Europe (unpublished results of B.Seifert)
less than 1% were suspected to be F1 hybrids with another species. This figure possibly
gives us an idea of the ‘‘normal’’ hybrid frequency between wood ant species at the
individual level and tells us that the European average of 6–8% in F. polyctena X rufa is
exceptional and a potential taxonomic problem. As fertile hybridisation between
F. polyctena and F. rufa is widespread, we have to consider whether they are good species
or genetically determined ecological races or morphs of the same species with different
social and ecological strategies. The latter solution was chosen by Seifert (1991).
Gyllenstrand et al. (2004) showed that in central Sweden the genetic clusters follow
morphological rather than geographical boundaries, strongly supporting the existence of
two separate gene pools, one of F. polyctena and the other of F. rufa. Therefore, we follow
the practice also adopted by Seifert (1996, 2007) and consider them as separate species.
1234 Evol Ecol (2010) 24:1219–1237
123
Both F. polyctena and F. rufa are geographically widespread from western Europe to
western Siberia, and they are genetically very similar. We found no good diagnostic
ncDNA markers and the mtDNA lineages are incompletely sorted with very low nucleotide
diversity (Goropashnaya et al. 2004). It seems reasonable to conclude that the species have
separated not earlier than during the last glaciation, and it is difficult to predict whether
they continue to become more differentiated or will gradually fuse together through
hybridisation. The distribution of individual admixture estimates in the German
F. polyctena x rufa hybrids and their independent persistence in the form of locally
dominant, isolated populations indicate a high fertility and fitness.
Hybridisation is increasingly understood as an important factor of evolution also in
animals (Mallet 2007). Many instances in ants refer to strategic hybridisation with hybrid
workers who need to be viable but not fertile (Seifert 1999, 2006; Umphrey 2006).
Hybridogenous formation of a new species with maintenance of the parental species is
apparently rare in animals. It seems rewarding to consider the F. polyctena et rufa case also
within this context. The hybrid nests in Germany were spatially clustered and all (or most)
workers showed admixed genomes. These observations suggest that the hybrid population
can be an independently operating and evolving unit. One possible example of such an
evolution is provided by Formica paralugubris Seifert 1996 from the Alps. The external
morphology of F. paralugubris gynes and workers is intermediate between F. lugubris and
F. aquilonia (unpublished data of B.S) while allozymes and mtDNA sequences are very
close to F. aquilonia (Pamilo et al. 1992; Goropashnaya et al. 2004; Bernasconi 2009). It is
tempting to hypothesise that F. paralugubris has rapidly evolved from a single hybrid
supercolony having been isolated during the last Pleistocene in a Nunatak or somewhere in
an Alpine valley. We are currently examining two other cases in which hybridisation
between two Formica species may have led to a formation of a new evolutionary lineage,
putatively a new species between F. lugubris and F. aquilonia and another one between
F. aquilonia and F. polyctena (Bernasconi 2009; Saapunki et al. 2008). It may turn out that
hybridisation could generally have a significant role in ant evolution with supercolonies
acting as large, separately evolving kin groups (Helantera et al. 2009).
Acknowledgments We wish to thank Philip Attewell, Wouter Dekoninck, Dieter Bretz, Gennady Dlussky,Katrin Moller, Rainer Neumeyer and Roland Schultz for providing samples and Riitta Jokela for the laboratoryanalyses. Jaqueline Gitschmann helped with the transformation of air-photographs into a graphics. The workwas supported by grants from the Academy of Finland (1122210 to P.P.).
References
Anderson E (1948) Hybridisation of the habitat. Evolution 2:1–9Andersson MJ (2003) PCO: a FORTRAN computer program for principal coordinate analysis. Department
of Statistics, University of Auckland, New ZealandAnderson EC, Thompson EA (2002) A model-based method for identifying species hybrids using multilocus
genetic data. Genetics 160:1217–1229Bernasconi C (2009) Integrative taxonomy of the Formica rufa group (Hymenoptera: Formicidae). PhD
Thesis, University of LausanneBetrem JG (1960) Uber die Systematik der Formica-rufa-Gruppe. Tijdschr Entomol 104:51–81Buschinger A (1970) Neue Vorstellungen zur Evolution des Sozialparasitismus und der Dulosis bei Ameisen
(Hym., Formicidae). Biol Zbl 88:273–299Chapuisat M (1996) Characterization of microsatellite loci in Formica lugubris B and their variability in
other ant species. Mol Ecol 5:599–601
Evol Ecol (2010) 24:1219–1237 1235
123
Corander J, Marttinen P (2006) Bayesian identification of admixture events using multi-locus molecularmarkers. Mol Ecol 15:2833–2843
Espadaler X, Monteserın S (2003) Aegeritella (Deuteromycetes) on Formica (Hymenoptera, Formicidae) inSpain. Orsis 18:13–17
Espadaler X, Wisniewski J (1987) Aegeritella superficialis Bal. et Wisn. and A. tuberculata Bal. et Wisn.(Deuteromycetes), epizootic fungi on two Formica species in the Iberian Peninsula. Butlletı de laInstitucio Catalana d’Historia Natural 54(6):31–35
Gillham E, Gillham BL (1996) Hybrid ducks. A contribution towards an inventory. BL Gillham, Wallington,Surrey, England, UK, 88 pp
Goropashnaya A, Fedorov VB, Pamilo P (2004) Recent speciation in the Formica rufa group ants(Hymenoptera, Formicidae): inference from mitochondrial DNA phylogeny. Mol Phyl Evol 32:198–206
Gosswald K (1942) Rassenstudien an der Roten Waldameise Formica rufa L. auf systematischer,okologischer, physiologischer und biologischer Grundlage. Zeitschr angew Ent 18(1):62–124
Gosswald K (1981) Artunterschiede der Waldameisen in Aussehen, Lebensweise, Organisation, Verhalten,Nest- und Straßenbau, Okologie und Verbreitung. -Merkblatter zur Waldhygiene 1/1981. VerlagWaldhygiene Wurzburg
Gyllenstrand N, Seppa P, Pamilo P (2004) Genetic differentiation in sympatric wood ants Formica rufa andF.polyctena. Ins Soc 51:139–145
Hasegawa E, Imai S (2004) Characterization of microsatellite loci in red wood ants Formica (s.str.) spp. andthe related genus Polyergus. Mol Ecol Notes 4:200–203
Helantera H, Strassmann JE, Carillo J, Queller DC (2009) Unicolonial ants: where do they come from, whatare they and where are they going? Tree 24(6):341–349
Hofener C, Seifert B, Kruger T (1996) A genetic model for disruptive selection on colony social organi-zation, reproduction, and ecotype distribution in wood ants inhabiting different woodland habitats. InsSoc 43:359–373
Lachenbruch P, Mickey M (1968) Estimation of error rates in discriminant analysis. Technometrics 10:1–11Lecat V, Fournier D, Aron S (2008) Influence of social structure and royal status on recognition in the ant
Pheidole pallidula. In: Abstracts of international union for the study of social insects, 4th IUSSIEuropean meeting, Belgium, 30 August–4 September 2008, p 160
Lesaffre E, Willems JL, Albert A (1989) Estimation of error rate in multiple group logistic discrimination.The approximate leaving-one-out method. Commun Stat: Theory Methods 18:2989–3007
Levin DA, Francisco-Ortega J, Jansen RK (1996) Hybridisation and the extinction of rare plant species.Conserv Biol 10(1):10–16
Maki-Petays H, Zakharov A, Viljakainen L, Corander J, Pamilo P (2005) Genetic changes associated todeclining populations of Formica ants in fragmented forest landscape. Mol Ecol 14:733–742
Mallet J (2006) Hybridisation as invasion of the genome. TREE 20:229–237Mallet J (2007) Hybrid speciation. Nature 446:279–283Millais JG (1902) The natural history of the british surface-feeding ducks. Longmans, Green & Co, London
and New York, 107 ppMillais JG (1913) British Diving Ducks. Longmans, Green & Co, London, vol. I: 141 pp., vol. II, 164 ppOtto D (1960) Statistische Untersuchungen uber die Beziehungen zwischen Koniginnenzahl und Ar-
beiterinnengrosse bei den Roten Waldameisen (‘‘engere Formica-rufa-Gruppe’’). Biol Zbl 79:719–739Otto D (1967) Die Bedeutung der Formica-Volker fur die Dezimierung der wichtigsten Schadinsekten–Ein
Literaturbericht. Waldhygiene 7:65–90Pamilo P (1984) Genotypic correlation and regression in social groups: multiple alleles, multiple loci and
subdivided populations. Genetics 107:307–320Pamilo P, Chautems D, Cherix D (1992) Genetic differentiation of disjunct populations of the ants Formica
aquilonia and Formica lugubris in Europe. Ins Soc 39:15–29Pearson B (1983) Hybridisation between Lasius niger and Lasius alienus. Ins Soc 30:402–411Phillips JC (1915) Experimental studies of hybridisation among pheasants and ducks. J Exp Zool 18:69–144Phillips JC (1921) A further report on species crosses in birds. Genetics 6(4):366–383Pisarski B (1982) Structure et organisation des societes de fourmis de l’espece Formica (Coptoformica)
exsecta Nyl. (Hymenoptera, Formicidae). Mem Zool 38:1–281Price TD, Bouvier MM (2002) The evolution of F1 postzygotic incompatibilities in birds. Evolution
56:2083–2089Saapunki J, Pamilo P, Seifert B (2008): Stable coexistence of two genetic lineages in one population. In:
abstracts of international union for the study of social insects, 4th IUSSI European meeting, Belgium,30 August–4 September 2008, p 90
1236 Evol Ecol (2010) 24:1219–1237
123
Seifert B (1991) The phenotypes of the Formica rufa complex in East Germany. Abh Ber NaturkundemusGorlitz 65 1:1–27
Seifert B (1996) Ameisen beobachten, bestimmen. Naturbuch-Verlag Augsburg, 352 ppSeifert B (1999) Interspecific hybridisations in natural populations of ants by example of a regional fauna
(Hymenoptera: Formicidae). Ins Soc 46:45–52Seifert B (2006) Social cleptogamy in the ant subgenus Chthonolasius—survival as a minority. Abh Ber
Naturkundemus Gorlitz 77:251–276Seifert B. (2007) Die Ameisen Mittel- und Nordeuropas. Tauer: lutra-Verlags- und Vertriebsgesellschaft,
368 ppSeifert B (2008) Removal of allometric variance improves species separation in multi-character discriminant
functions when species are strongly allometric and exposes diagnostic characters. MyrmecolNews 11:91–105
Seifert B, Goropashnaya A (2004) Ideal phenotypes and mismatching haplotypes errors of mtDNA treeingin ants (Hymenoptera: Formicidae) detected by standardized morphometry. Org Divers Evol 4(4):295–305
Seifert B, Schultz R (2008) A taxonomic revision of the Formica subpilosa RUZSKY, 1902 group (Hyme-noptera: Formicidae). Myrmecol News 12:67–83
Simberloff D (2008) Habitat fragmentation and population extinction of birds. Ibis 137:37–104Sokal RR, Rohlf FJ (1995) Biometry: the principles and practice of statistics in biological research, 3rd edn
edn. WH Freemann, New York, pp 887Stebbins GL (1959) The role of hybridisation in evolution. Proc Amer Phil Soc 103:231–251Stebbins GL (1980) Evolutionsprozesse. Fischer, Stuttgart and New YorkStuart RJ, Gresham-Bissett L, Alloway TM (1993) Queen adoption in the polygynous and polydomous ant,
Leptothorax curvispinosus. Behav Ecol 4:276–281Sundstrom L (1993) Genetic population structure and sociogenetic organisation in Formica truncorum
(Hymenoptera; Formicidae). Behav Ecol Sociobiol 33:345–354Umphrey GJ (2006) Sperm parasitism in ants: selection for interspecific mating and hybridisation. Ecology
87:2148–2159UNESCO—MAB Bioreserves directory (2009): Vornonezhskiy Zapovednik. http://www.unesco.org/
mabdb/br/brdir/directory/biores.asp?code=RUS?11&mode=all (accessed August 2009)Vaha J-PK, Primmer CR (2006) Detecting hybridization between individuals of closely related popula-
tions—a simulation study to assess the efficiency of model-based Bayesian methods to detect hybridindividuals. Mol Ecol 15:63–72
Weir BS, Cockerham CC (1984) Estimating F-statistics for the analysis of population structure. Evolution38:1358–1370
Wisniewski J (1976) Wystepowanie grzyba Aegeritella superficialis Bal. et Wisn. w Wielkopolskim ParkuNarodowym. Pr Kom Nauk Lesn, PTPN Poznan 42:131–135
Wisniewski J (1977) Occurrence of fungus Aegeritella superficialis Bal. et Wisn.,1974, on Formica lugubrisZett. in Italian Alps Boll Soc Ent Ital 109: 83–84
Yarrow IHH (1955) The British ants allied to Formica rufa L. Trans Soc Brit Entomol 12:1–48
Evol Ecol (2010) 24:1219–1237 1237
123