INTRODUCTION
Gyrodactylids are ubiquitous ectoparasites on the skin and gills of teleost fish both in marine and freshwater ecosystems (Bakke et al. Reference Bakke, Cable and Harris2007). The most recent species compilation lists 409 gyrodactylid species (Harris et al. Reference Harris, Shinn, Cable and Bakke2004). Until the mid-1990s, most Gyrodactylus species were identified by comparing the morphology of the opisthaptoral hard parts. Over recent years, the application of molecular markers in the taxonomy and systematics of Gyrodactylus species has increased, e.g., the sequencing of the internal transcribed spacers (ITS-1 and ITS-2) of nuclear ribosomal DNA (rDNA). Ziętara and Lumme (Reference Ziętara and Lumme2002) showed that many Gyrodactylus species can be discriminated by internal transcribed spacer (ITS) sequences, and subsequent phylogenetic analyses yielded tree topologies that were basically consistent with those of classical taxonomy. However, G. salaris Malmberg, Reference Malmberg1957 cannot be differentiated from its closest relative G. thymalli Žitňan, Reference Žitňan1960 by means of ITS-1 and ITS-2 sequences (Cunningham, Reference Cunningham1997), even considering parasites from a wide geographical range (Ziętara and Lumme, Reference Ziętara and Lumme2002).
G. salaris has seriously hampered the natural juvenile production of Atlantic salmon, Salmo salar L., in Norway over the past 3 decades and has caused great harm both in ecological and economic terms (Johnsen et al. Reference Johnsen, Møkkelgjerd and Jensen1999; Mo et al. Reference Mo, Norheim and Hellesnes2004), whereas G. thymalli is considered harmless to any of its known potential hosts. However, Jørgensen et al. (Reference Jørgensen, Larsen, Jørgensen, Bresciani, Kania and Buchmann2007) and Olstad et al. (Reference Olstad, Robertsen, Bachmann and Bakke2007) have recently described 2 strains of G. salaris with restricted infectivity and reproduction on Atlantic salmon.
Several approaches using different markers have addressed the taxonomy and systematics of G. salaris and G. thymalli. To date, the mitochondrial cytochrome c oxidase subunit 1 gene (cox1) is the only marker allowing for a genetic discrimination of populations of G. salaris and G. thymalli (see e.g., Hansen et al. Reference Hansen, Bachmann and Bakke2003, Reference Hansen, Martinsen, Bakke and Bachmann2006; Meinilä et al. Reference Meinilä, Kuusela, Ziętara and Lumme2004). However, the 2 species differ in host preference (Soleng and Bakke, Reference Soleng and Bakke2001; Bakke et al. Reference Bakke, Harris and Cable2002; Sterud et al. Reference Sterud, Mo, Collins and Cunningham2002), which is an important characteristic for considering G. salaris and G. thymalli different species. Discrimination of the 2 species based on multivariate statistics on morphometric measurements have proved possible but only a few populations (McHugh et al. Reference McHugh, Shinn and Kay2000; Shinn et al. Reference Shinn, Hansen, Olstad, Bachmann and Bakke2004) were included in these studies.
In the present study, we explored whether there is a consistent pattern of morphometrical variation between G. salaris and G. thymalli over a broad range of natural populations by using a set of ‘traditional’ point-to-point morphometric measurements. Such morphometric measurements are known to be affected by environmental parameters, such as temperature (see Malmberg, Reference Malmberg1970; Mo, Reference Mo1991a, Reference Mob, Reference Moc; Dmitrieva and Dimitrov, Reference Dmitrieva and Dimitrov2002; Davidova et al. Reference Dávidová, Jarkovský, Matĕjusová and Gelnar2005) that may impede an objective analysis of opisthaptor morphology. However, we were interested to investigate the ‘diagnostic realism’ in studying natural populations rather than analysing populations of gyrodactylids raised experimentally in a constant environment. To test whether G. salaris and G. thymalli are 1, 2, or more than 2 species, sample locations were chosen to comprehensively reflect the known variation in cox1 genealogy.
MATERIALS AND METHODS
Sampling of Gyrodactylus
In total, 10 populations have been sampled and 168 specimens measured. Four Gyrodactylus populations were sampled from wild infected Atlantic salmon populations, namely, the rivers Skibotnelva (northern Norway), Rauma (north-western Norway), Drammenselva (southern Norway), and Göta älv (south-western Sweden). In the 3 Norwegian rivers, serious gyrodactylosis caused by G. salaris has been documented (Johnsen et al. Reference Johnsen, Møkkelgjerd and Jensen1999). One sample was collected from rainbow trout (Oncorhynchus mykiss Walbaum) (from a fish-farm in lake Bullaren, south-western Sweden), and one sample was collected from Arctic charr (Salvelinus alpinus L.) in lake Pålsbufjord (central South-Norway). The population of Gyrodactylus in Lake Bullaren was identified as G. salaris by the OIE Reference Laboratory for Gyrodactylus (G. salaris) in Norway (Mo et al. Reference Mo, Norheim and Hellesnes2004). The sample from Lake Pålsbufjord was confirmed as G. salaris based on ITS rDNA and cox1 sequences (Robertsen et al. Reference Robertsen, Hansen, Bachmann and Bakke2007). Four Gyrodactylus populations were sampled from wild infected grayling (Thymallus thymallus L.) populations, namely Lake Lesjaskogsvatn (central South-Norway), rivers Trysilelva (south-eastern Norway), Rena (south-eastern Norway) and Hnilec (Slovakia). The river Hnilec is one of the 2 type localities for G. thymalli according to Žitňan (Reference Žitňan1960) (see below). For the purpose of clarity, in this manuscript populations are referred to as G. salaris if recovered from Atlantic salmon and G. thymalli if recovered from grayling. Accordingly, the samples from Lake Bullaren and Lake Pålsbufjord will be discussed independently. Wild fish were either captured by electro-fishing, gill netting or by rod-fishing. Parasite specimens were isolated from fins taken from the host and stored in 96% ethanol during transportation to the laboratory. The cox1 sequences of the presently analysed samples are presented in Hansen et al. (Reference Hansen, Bachmann and Bakke2003, Reference Hansen, Martinsen, Bakke and Bachmann2006). Further details are listed in Table 1.
Table 1. The Gyrodactylus populations, place of analyses and designation of the mitochondrial haplotype clades
(UoO=University of Oslo; UoS=University of Stirling.)
1 Samples of O. mykiss originated from a fish farm in Lake Bullaren.
2 Sampling in Pålsbufjord performed during specified period.
Type-material
Gyrodactylus salaris was first described by Malmberg (Reference Malmberg1957) from Atlantic salmon in the Hölle Laboratory on the Baltic river Indalsälv, Sweden. The original description was based on 1 individual from a sample collected in 1952Footnote 1. Nevertheless, Malmberg prepared a total of 26 individuals from the same sample. The material is deposited in the Swedish Museum of Natural History, Stockholm. For the present study, 12 specimens from ‘the type-sample’ were photographed at the University of Stockholm using a Leica DC 300 camera mounted on a Leitz (Dialux) stereomicroscope under a×63 dry and a×100 immersion-oil objective. For scaling of the pictures, a stage micrometer from Albert Sass, Berlin (2 mm in 200 parts) was used. Measurements were performed on the Oslo Leica IM1000 v.4.0 software. (G. salaris ID.: 7089–7100; no. 7089 being registered as lectotype and the others as paralectotypes (Sven Boström, personal communication)).
Gyrodactylus thymalli was first described by Žitňan (Reference Žitňan1960) from grayling from the rivers Hnilec and Hron, Slovak RepublicFootnote 2. The material is deposited in the East Slovak Museum, Koŝice, the Slovak Republic. A total of 12 individuals were included in the description. Of these, 6 were available for analysis in the present work (ID. no. Z-11229/1-6). The material was photographed and analysed at the Department of Zoology, Natural History Museum, University of Oslo, Norway according to the procedures described below.
Morphometry
For the morphometric analyses, individual Gyrodactylus specimens (n=12–30) from each population were removed from fins and fixed in 96% ethanol. The opisthaptors were cut off and the opisthaptoral hard parts were subsequently prepared using a method slightly modified after Harris et al. (Reference Harris, Cable, Tinsley and Lazarus1999). The haptors were digested in 75 mm Tris, 10 mm EDTA, 5% SDS and 100 mg/ml proteinase K, pH 8·0. After washing twice in H2O, the released hard parts were fixed in ammonium picrate glycerine (Malmberg's fixative). The digital image analyses were performed by only 1 person (K. Olstad), but at 2 different laboratories; the Institute of Aquaculture, University of Stirling, UK, and at the Department of Zoology, Natural History Museum, University of Oslo, Norway (for details, see Table 1). At the University of Stirling, specimens were photographed using a JVC KY-F30B 3CCD camera with an interfacing×2·5 top lens fitted to an Olympus BH2 compound microscope under a×100 oil objective, and measured using the Zeiss KS300 iC/Windows Release v.3.0 (1997) (Carl Zeiss Vision 5 GmbH, Munich, Germany/Imaging Associates Ltd, Thame, Oxfordshire, UK) software. At the Natural History Museum, University of Oslo, specimens were photographed using a Leica DC 500 camera mounted on a Leica DM 6000B stereomicroscope under a×100 oil objective, and measured using the Leica IM1000 v.4.0 software.
A total of 27 point-to-point measurements were taken on each specimen. Of these, 23 were identical to the measurements described by Shinn et al. (Reference Shinn, Hansen, Olstad, Bachmann and Bakke2004), and with the same annotations. (i) Marginal hook (MH): total length (MHTL), shaft length (MHSHL), sickle length (MHSL), sickle proximal width (MHSPW), sickle toe length (MHSTL), sickle distal width (MHSDW), sickle aperture (MHAD), instep height (MHIH). (ii) Hamulus (H): proximal shaft width (HPSW), aperture distance (HAD), point length (HPL), distal shaft width (HDSW), hamulus inner angle (HIA), inner curve length (HICL), shaft length (HSL), root length (HRL), total length (HTL). (iii) Ventral bar (VB): total length (VBTL), process to mid length (VBPML), median length (VBML), membrane length (VBMBL), centre length (VBCL), lateral length (VBLL), total width (VBTW), width (VBW), membrane maximal width (VBMMW) and process length (VBPL).
The 2 hamulus measurements, the aperture angle (HAA) and the point curve angle (HHPCA) used by Shinn et al. (Reference Shinn, Hansen, Olstad, Bachmann and Bakke2004), were not considered in this study. The 4 measurements VBCL, VBLL, VBW, and VBMMW represent new morphometric variables and are depicted in Fig. 1.
Fig. 1. Scanning electron micrograph (SEM) of the ventral bar (VB) from a Gyrodactylus specimen parasitizing grayling in the river Hnilec, Slovakia, illustrating the four new measurements. (A) Ventral bar centre length (VBCL): The distance between the top of the median bar and the distal part of the membrane, along a line following the centre of the structure. (B) Ventral bar lateral length (VBLL): The distance between an orthogonal line between ventral bar bases towards the membrane (measurement D) to the distal top of the ventral bar (excluding the process). (C) Ventral bar width (VBW): The maximum width of the median portion of the ventral bar, excluding the processes. (D) Ventral bar membrane maximal width (VBMMW): the distance between the ventral bar bases towards the membrane.
Statistical analyses
All statistical analyses as well as ordination were performed using the software PAST v.1.66 (Hammer et al. Reference Hammer, Harper and Ryan2001). The hamulus inner angle (HIA) was transformed to the cosine in order to obtain linear functions prior to data analyses. Since marginal hook total length (MHTL) and shaft length (MHSHL) were not possible to obtain from the type materials, these measurements were omitted in the analyses.
Since photographing of the specimens and measuring of the sclerites were performed at 2 different localities with 2 different microscopes and software, the same specimens (n=10) from a randomly chosen population (Skibotnelva) first measured at the University of Stirling, were re-measured at the Natural History Museum, Oslo. The non-parametric Bray-Curtis MANOVA was used to test whether the multivariate means of the 2 data-samples were equal.
Based on the hypothesis that the sampled individuals belong to 2 distinct species, k-means clustering preset for 2 groups was run on the total data-matrix. The cluster assignments in k-means clustering are initially random which implies that the results may differ from run to run. The procedure was therefore repeated >10 times. Clustering into 3 and 4 groups was also run in order to test alternative hypotheses regarding number of groups (e.g., groupings according to host-species).
In order to test the hypothesis that any 2 populations had equal multi-dimensional means, Wilk's lambda MANOVA and Hotelling's pairwise post-hoc comparisons were run. The Hotelling's pairwise post-hoc comparisons were run both with and without Bonferroni corrections. The low number of individuals available for study in this analysis in combination with the high number of variables made a dimensionality-reduction of the data-matrix necessary. Principal Component Analysis (PCA) scores on the variance-covariance matrix were therefore obtained, and the statistical tests run on a limited number of the PCA axis-scores (but maintaining >95% of the total variation). For a numerical comparison of distances between population multivariate distributions, pair-wise Mahalanobis distances were calculated using S-PLUS 6.0, Professional Release 1.
RESULTS
The morphometric differences of 10 populations of the monogenean G. salaris on Atlantic salmon, rainbow trout and Arctic charr and G. thymalli on grayling, were studied. These particular Gyrodactylus populations were selected as they comprehensively represent mitochondrial haplogroups as described by Hansen et al. (Reference Hansen, Bachmann and Bakke2003, Reference Hansen, Martinsen, Bakke and Bachmann2006) (Table 1), and are thus expected to represent the range of intra- and interspecific variation of G. salaris and G. thymalli. The assignment of mitochondrial haplotypes to parasite populations is in some instances based on the cytochrome oxidase I sequence of only 1 individual, hence intra-populational mitochondrial sequence variation cannot be excluded. Individual cytochrome oxidase I sequence data were not determined and intra-populational sequence variation was thus not considered in the present work.
The data obtained in the different labs were analysed together, as all measurements were performed by the same person and no significant differences were observed between any of the different instrumental set-ups used in the present investigations (non-parametric Bray-Curtis MANOVA P=0·297).
The type-material of G. salaris and G. thymalli constitutes whole-mounted worms that have been pre-treated with formalin. Such treatment is not favourable for light-microscope studies of the hard parts. The variance of the measurements from these individuals might thus be increased. Some individuals could not be reliably measured due to oblique positions of the hard parts and were thus discarded from the analyses (G. salaris measured: ID no. 7090, 7093, 7094, 7096, 7097 – G. thymalli measured: ID no. Z-11229/2-6). A summary of the morphometric measurements of the hard parts of the G. salaris and G. thymalli type-material is given in Table 2. Although the averages for the measurements of the marginal hook total length (MHTL), marginal hook shaft length (MHSAL), and marginal hook sickle length (MHSL) might indicate a clear grouping in 2 species in accordance with the a priori species delineation, the ranges between the groups substantially overlap. MHTL and MSHAL for the G. salaris and G. thymalli type-material were unfortunately impossible to obtain. Thus, the averages and ranges did not reveal a single measurement unambiguously grouping the populations in accordance with an a priori species delineation.
Table 2. Summary of the morphometric measurements for the 10 Gyrodactylus populations analysed in this study
(A detailed explanation of abbreviations is provided in the Materials and Methods section. Each measured feature is given in micrometres (μm) except HIA, for which the measurements are given in degrees, followed by the standard deviation and the range (in parentheses). n – Number of specimens measured. * – Rainbow trout. ** – Arctic charr. *** – Grayling. Others hosted by Atlantic salmon.)
In the k-means clustering, both for 2 and 3 groups, the G. salaris type-material and the G. thymalli type-material were categorized in separate groups (Table 3). When clustering into 2 groups, the Bullaren population (from rainbow trout) most closely resembled the ‘G. salaris group’ whereas the Pålsbufjord population (from Arctic charr) represented an intermediate group positioned between the G. salaris and G. thymalli groups (Table 3). The grayling parasites from Trysilelva resembled gyrodactylids collected from Atlantic salmon. The remaining populations of Gyrodactylus resembled type-material representing their respective species in accordance with the a priori species designations based on host species. When clustering into 3 groups, parasites sampled from Atlantic salmon and grayling basically followed the same pattern as when clustering into 2 groups, although some individuals from all populations but Hnilec were classified into Group 3 (Table 3). Individuals from both Bullaren (from rainbow trout) and Pålsbufjorden (from Arctic charr) were mainly classified into Group 3. With few exceptions, the pattern when clustering into 4 groups was consistent with the pattern when clustering into 3 groups (Exceptions: 1, 1 and 3 individuals from G. salaris type-material, Drammen and Skibotnelva, respectively, were classified into Group 4. Data not shown in table).
Table 3. Individual classifications based on k-means clustering into 2 and 3 groups of the morphometric measurements from the 10 populations of Gyrodactylus analysed as well as the G. salaris and G. thymalli type-material
(Number of specimens classified into respective groups is given. * – Rainbow trout. ** – Arctic charr. *** – Grayling. Others hosted by Atlantic salmon.)
In a PCA-plot (run on the variance-covariance matrix) based on the measurements from all 10 populations and the type-material of both species (Fig. 2), the 2 populations representing the name-bearing types separated well from each other. Furthermore, the populations Drammenselva, Skibotnelva, Rauma, Göta älv, Bullaren and Trysilelva apparently grouped with the G. salaris type-material, and the populations Hnilec, Lesjaskogsvatn and Rena with the G. thymalli type-material. Gyrodactylids collected from Arctic charr from Lake Pålsbufjord, on the other hand, appear to represent an intermediate group positioned between the main G. salaris and G. thymalli clusters. The graphical presentation of the PCA analysis did not reveal any groupings in the material related to cox1 phylogeny or to temperature/season of sampling. In principal components higher than 2, the PCA plot did not reveal any clear groupings of the data.
Fig. 2. Principal component analyses of 27 morphometric variables taken from10 populations of Gyrodactylus salaris and G. thymalli. Open triangles – G. salaris (hosted by Atlantic salmon) from Drammenselva, Skibotnelva, Rauma, and Göta älv, as well as the G. salaris type-material. Black triangles – G. salaris from Lake Bullaren rainbow trout. Shaded triangles – G. salaris from Lake Pålsbufjord Arctic charr. Open circles – G. thymalli (hosted by grayling) from Hnilec, Lesjaskogsvatn and Rena, as well as the G. thymalli type-material. Shaded circles – G. thymalli from River Trysilelva grayling. (A) A distribution map of each Gyrodactylus population with the first principal component axis (PC1) plotted against the second (PC2). (B) The distribution of each Gyrodactylus population when the second principal component axis (PC2) is plotted against the third (PC3).
The scores from the PCA (run on the variance-covariance matrix) were used to identify a reduced number of linear combinations explaining a sufficient amount of the total variation in the data set. The first 9 PCs explained 95·4% (PC1 – 62.70%; PC2 – 9·1%; PC3 – 6·5%) of the total variation in the data set, and were thus considered to capture a sufficient amount of the variation in the 27 variables to be the only ones included in the subsequent multivariate analyses. According to established procedures in morphometric analyses, most of the variation explained in PC1 is due to size, particularly when the first PC explains a substantial percentage of the total variation in the data set (Jolicoeur and Mosiman, Reference Jolicoeur and Mosimann1960; Reyment et al. Reference Reyment, Blackith and Campell1984). Accordingly, the following multivariate analyses were both run on a data-matrix comprised of data from the first 9 PCs (PC1–PC9) and on a data matrix in which the data for PC1 had been removed (PC2–PC9; explaining 32·7% of the total variation in the data set). The latter is expected to be an analysis in which a major part of variation due to size is removed, thereby leaving variation principally attributable to changes in shape.
The MANOVA rejected the hypothesis of equal multidimensional means (Wilks-lambda, P≪0·05) for all populations. This was observed both when including and excluding PC1. The Hotelling's T squared post-hoc tests gave significant differences (P≪0·05) between all populations when not Bonferroni corrected. Again, this was observed both with and without PC1. The Hotelling's T squared post-hoc tests when Bonferroni corrected, with few exceptions (Table 4) also gave significant results, despite very low statistical power. When analysing PC1–PC9, cases where the hypothesis of equality could not be rejected involved the salmon parasites from Drammenselva, Skibotnelva, Rauma, and Göta älv as well as the rainbow trout parasites from Bullaren. If removing the first PC, cases where the hypothesis of equality could not be rejected also involved the salmon parasites from Drammenselva, Skibotnelva, Rauma, and Göta älv. Additionally, equality could not be rejected between the grayling parasites from Lesjaskogsvatn and the salmon parasites from Drammenselva, Skibotnelva, and Rauma or between the grayling parasites from Hnilec and the salmon parasites from Skibotnelva and Rauma.
Table 4. Hotelling's pairwise post-hoc comparisons of PC1–PC9 (below the diagonal) and PC2–PC9 (above the diagonal) for 10 populations of Gyrodactylus salaris and G. thymalli, infecting four different host species (see text)
(† – non-significant results (P⩾0·05). * – Rainbow trout. ** – Arctic charr. *** – Grayling. Others hosted by Atlantic salmon.)
The Mahalanobis distances between the populations are listed in Table 5. For PC1–9 the magnitude of the distances indicates a grouping of the G. salaris populations on Atlantic salmon from Drammenselva, Skibotnelva, Rauma and Göta älv. Compared to these pairwise distances, the Mahalanobis distances between the G. salaris populations on rainbow trout (Bullaren) and the other G. salaris populations, are slightly higher. One exception is the population from Drammenselva, with which gyrodactylids from Bullaren share the same mitochondrial haplotype. The most striking difference when analysing Mahalanobis distances from PC2–9 is that G. thymalli from Hnilec resembles more the G. salaris populations (less Pålsbufjord) than the other G. thymalli populations.
Table 5. Mahalanobis distances for the principal components PC1–PC9 (above the diagonal) and PC2–PC9 (below the diagonal) for 10 populations of Gyrodactylus salaris and G. thymalli, infecting four different host species
(* – Rainbow trout. ** – Arctic charr. *** – Grayling. Others hosted by Atlantic salmon.)
DISCUSSION
In the current paper, we analysed the morphological differentiation of 10 populations of G. salaris and G. thymalli that have mitochondrial haplotypes belonging to 6 well-supported haplogroups (Hansen et al. Reference Hansen, Martinsen, Bakke and Bachmann2006). We noticed a substantial interpopulation and intrapopulation variation in hook morphometry. Such variation can either be due to phenotypic plasticity in response to varying environmental factors, or it could reflect true genotypic differentiation (or a combination of both). Given the high level of mitochondrial DNA differentiation in the Gyrodactylus populations, morphometrical variations may be attributed to genotypic variation due to adaptation to local micro- and macroenvironmental constraints. However, if caused by phenotypic plasticity, the observed variation is expected to blur and weaken any gentotypic and phylogenetic signal in the morphometric analyses.
Type-material and species assignment
The samples in the present study were assigned to species a priori based on host species: the samples from Drammenselva, Skibotnelva, Rauma and Göta älv were assigned to G. salaris, and the samples from the Hnilec, Lesjaskogsvatn, Trysilelva and Rena were considered G. thymalli. The Bullaren and Pålsbufjord populations were sampled from rainbow trout and Arctic charr, respectively. Both parasite populations do, however, share identical mitochondrial cox1 sequences with, for example, the Drammenselva population and have thus been regarded as G. salaris (see Hansen et al. Reference Hansen, Bachmann and Bakke2003; Robertsen et al. Reference Robertsen, Hansen, Bachmann and Bakke2007). Variation along the first principal component was continuous with no immediately apparent groupings. A more detailed analysis, however, revealed 2 separate groups but with the population from Pålsbufjord in between. The 2 groups included (i) the G. salaris type-material, the a priori assigned G. salaris populations, Bullaren, and the Trysilelva population parasitizing grayling, and (ii) the G. thymalli type-material, and the a priori assigned G. thymalli populations (except Trysilelva). The results from the k-means clustering were consistent with the PCA: the Bullaren population most closely resembled the ‘G. salaris group’ whereas the Pålsbufjord population represented an intermediate group positioned between the G. salaris and G. thymalli groups. The grayling parasites from Trysilelva resembled gyrodactylids collected from Atlantic salmon. The remaining populations of Gyrodactylus resembled the type-material representing their respective species in accordance with the a priori species designations. From a taxonomic viewpoint, these morphometric data confirm the species status of the a priori assigned populations except for the Trysilelva population.
Intra- and interspecific variation
The Gyrodactylus populations included in this study could be distinguished from each other by morphometry; all populations had significantly different multidimensional means. It is noteworthy that intra-specific pairwise Mahalanobis distances were lower among G. salaris parasitizing salmon than among the populations of G. thymalli parasitizing grayling, this also holds true when excluding the Trysilelva population from the comparison (Mann-Whitney U test: P=0·03).
When PC1, which is usually considered the size-axis (Jolicoeur and Mosiman, Reference Jolicoeur and Mosimann1960; Reyment et al. Reference Reyment, Blackith and Campell1984), was excluded from the analyses, the Mahalanobis distances indicated that G. thymalli specimens from Trysilelva were different from all G. salaris specimens, whereas the Hnilec and Lesjaskogsvatn populations resembled the G. salaris specimens. This pattern was, however, not fully congruent with the Hotelling's pairwise comparisons when relaxing the statistical power; the Trysilelva population was not significantly different from the G. salaris populations when including size. It is noteworthy that for a size-measure such as hamulus total length (HTL) the specimens from Hnilec, with a water temperature of 12°C at the time of sampling, were the biggest (mean 83 μm), whereas the specimens from Skibotnelva, which were collected at 3°C, were the smallest (mean 68 μm). If one assumes that these size differences are due to ambient temperature affecting otherwise undistinguishable traits, the observed pattern would be opposite of previous publications stressing that cold conditions will lead to larger hooks due to extended development-time (suggested by Kulemina, Reference Kulemina and Skarlato1977; Mo, Reference Mo1991a). Whether or not the observed variation is within the range of phenotypic plasticity of size and shape or represents genetic differentiation, remains unknown. For example, it has been shown from laboratory experiments that Gyrodactylus exhibit differences in developmental time and fecundity on different hosts (Cable et al. Reference Cable, Harris and Bakke2000). Since the opisthaptoral hard parts do not grow after birth (Kulemina, Reference Kulemina and Skarlato1977), the particular parasite strain – host species relationship may be expected to constitute a significant factor for variation in size and shape given growth is allometric, or for size only if growth is isometric.
Taxonomic implications
One of the alternative taxonomic scenarios proposed by Hansen et al. (Reference Hansen, Bachmann and Bakke2003) is that G. salaris and G. thymalli may represent a complex of more than 2 sibling species. The fact that the samples from Bullaren and Pålsbufjorden grouped together separately from the Atlantic salmon and grayling parasites might bring indications of such a scenario, particularly since they share a common cox1 haplotype (Robertsen et al. Reference Robertsen, Hansen, Bachmann and Bakke2007). However, these 2 populations also share a common cox1 haplotype with the G. salaris from River Drammenselva (Hansen et al. Reference Hansen, Bachmann and Bakke2003), a population that they did not group together with in the present analyses. The present analyses may, in this matter, be biased due to the selection of the populations, and an extended sampling including specimens from Finland and Russia (clades III and IV according to Meinilä et al. Reference Meinilä, Kuusela, Ziętara and Lumme2004) may alter the results. The rejection of the hypothesis of equal multi-dimensional means and the high level of correct assignments for specimens from all studied populations in the classification tests may, on the other hand, at least not contradict the hypothesis of G. salaris and G. thymalli being a complex of more than 2 sibling species. However, the present morphometrical differentiation as presented by the PCA analysis is not suitably sufficient to allow for a clear delineation on a subspecies level and does not, therefore, provide strong support for this hypothesis. According to the sibling species complex hypothesis (Hansen et al. Reference Hansen, Bachmann and Bakke2003), one would expect minor differentiation of the G. salaris populations from Skibotnelva and Rauma representing the same mitochondrial haplogroup and both parasitizing Atlantic salmon. However, the G. salaris populations from these 2 rivers are as differentiated as any other populations under study. When taking into account that some of the morphological characters of the opisthaptoral hard parts in Gyrodactylus are also influenced by temperature (see Malmberg, Reference Malmberg1970; Mo, Reference Mo1991a, Reference Mob, Reference Moc; Dmitrieva and Dimitrov, Reference Dmitrieva and Dimitrov2002; Davidova et al. Reference Dávidová, Jarkovský, Matĕjusová and Gelnar2005), and probably also by host species (Dmitrieva and Dimitrov, Reference Dmitrieva and Dimitrov2002; Huyse and Volkaert, Reference Huyse and Volckaert2002; unpublished findings), we are tempted to conclude that the observed morphometric variation in the measured characters for G. salaris and G. thymalli does not indicate more than 2 sibling species.
In conclusion, our morphometric data did not clearly support the delineation of the G. salaris/G. thymalli species complex into 2 or more groups. Our study also illustrates that there is an urgent need for more detailed knowledge on the environmental factors influencing the phenotype of Gyrodactylus populations. At present, by far most of the 409 described gyrodactylid species (Harris et al. Reference Harris, Shinn, Cable and Bakke2004) are ‘morphological species’, as they are described without including molecular data. More information on the environmental impact on gyrodactylid morphology is, therefore, needed for the interpretation of the taxonomic significance of the morphometrical differences observed between Gyrodactylus populations in general. With particular reference to G. salaris and G. thymalli, our results indicate that the species determination of gyrodactylid species taken from wild populations of fish, when based solely on morphometrics, is more complicated and problematic than initially thought.
This work was supported by the Norwegian Research Council ‘Wild Salmon Programme’ (Project no. 145861/720) and the National Centre for Biosystematics (Project no. 146515/420, co-funded by the NRC and the NHM, University of Oslo, Norway). We thank G. Malmberg and V. Dudinak for providing access to and additional information regarding the G. salaris and G. thymalli type material, respectively; V. Hanzelova for providing samples of G. thymalli from river Hnilec, Slovakia; H. Hansen, D. Gammelsæter, T. Haugen, L. Karlsson, I. Perä, O. Eide, P. Arnkværn, and T. Olstad for help in collecting material in Scandinavia; H. J. Berg and C. Vollelv for assisting in SEM photography; and G. Robertsen for assisting in the preparation of specimens.
