Introduction

Most plant-feeding insects feed on only one or a few host species (Jermy, 1984; Bernays and Graham, 1988), but the exact proportion of true specialists is difficult to determine. In fact, it is very likely that the number of monophagous insect species is currently underestimated, because speciation and ecological specialization in insects is frequently accompanied by little or no change in morphology (Mayr, 1963; Diehl and Bush, 1984; Drès and Mallet, 2002). Thus, an adaptive radiation to multiple hosts may result in a complex of morphologically indistinguishable host races or sibling species, which are then erroneously interpreted as one polyphagous species (Diehl and Bush, 1984; Drès and Mallet, 2002). However, cryptic species or host races can sometimes be distinguished by using molecular methods, such as allozyme electrophoresis (Feder et al, 1988; McPheron et al, 1988; Roininen et al, 1993; Emelianov et al, 1995; Condon and Steck, 1997; Nyman, 2002; Aguin-Pombo, 2002), random amplified polymorphic DNA (De Barro et al, 1995; Pornkulwat et al, 1998), or DNA sequencing (Scheffer, 2000; Scheffer and Wiegmann, 2000).

The aim of this study was to use allozyme electrophoretic data to clarify species identification in the agromyzid genus Phytobia Lioy. The genus consists of species that are separated from the rest of the Agromyzidae not only morphologically but, more importantly, based on the unique feeding habit of their larvae (von Tschirnhaus, 1971; Spencer, 1976; Spencer and Steyskal, 1986): the growing larvae mine within the cambial cylinder of growing trees and bushes, tunnelling several meters downwards from tree crowns towards the base (Barnes, 1933; Kangas, 1935; Hanson and Benjamin, 1967; Ylioja et al, 1998). Host trees fill the larval tunnels with parenchyma tissue that keep the tunnels visible within the annual rings of trees (Greene, 1914; Kangas, 1935; Gregory and Wallner, 1979; Ylioja et al, 1998). In trees with high commercial value used for furnishing and other specialty purposes, the dark larval tunnels are often considered defects, and can decrease the value of raw material considerably (Kangas, 1935; Hanson and Benjamin, 1967; Spencer, 1976; Spencer and Steyskal, 1986; Ylioja et al, 1998).

In addition to being a pest for the forestry industry, Phytobia is a promising group for evolutionary studies. It is apparently one of the most primitive genera in the Agromyzidae (Nowakowski, 1962; Spencer, 1990), and larval tunnels have been found in fossil wood dating back to the Eocene, ca. 37–58 million years ago (Süss and Müller-Stoll, 1980; but see Spencer and Steyskal, 1986, for a cautionary note). The globally distributed genus comprises over 50 currently recognized species, all of which are apparently mono- or oligophagous (Spencer and Steyskal, 1986); the host plant species belong to numerous genera in about 60 families (Süss and Müller-Stoll, 1980; Spencer, 1990). Thus, Phytobia can be used to study how the utilization of different host species has changed during the evolutionary history of insects.

Studying the evolution of host-plant associations in Phytobia would require knowledge on the phylogenetic relationships between the species within the genus, but even the basic taxonomy of the group is unclear. Species descriptions in Phytobia are based on morphological characters of adults, which can be very similar between some species (Spencer, 1976; Spencer and Steyskal, 1986). In many cases, only small differences in male genitalia may be used to differentiate the species (Spencer, 1976; Spencer and Steyskal, 1986). The adults are difficult to catch and rear, and the white, slender larvae, which can be easily collected from young trees, offer few useful morphological characters for taxonomic studies (Barnes, 1933; Kangas, 1935; Spencer, 1976; Süss and Müller-Stoll, 1980). Consequently, it is not surprising that the taxonomy of the genus has been changing constantly, and the host ranges of many species are unknown (Spencer, 1976, 1990; Spencer and Steyskal, 1986).

The taxonomy of Phytobia has been most extensively studied in Europe and North America, but even in these areas there is considerable confusion about the number of known species. During the first half of the last century, Kangas (1935, 1949) collected and described cambium miners in southern Finland. The first two new species that he described were P. betulae Kangas on Betula spp., and P. aucupariae Kangas on Sorbus aucuparia L. Kangas (1949) considered that European Salicaceae hosts support three Phytobia species: P. cambii Hendel and P. barnesi Hendel on Populus and Salix species, and P. tremulae Kangas on Populus tremula L. However, these three Salicaceae-feeding species have later been synonymized under P. cambii (Spencer, 1976, 1990; Martinez et al, 1985). Recently, even P. betulae has been synonymized with P. cambii (von Tschirnhaus, 2000).

The purpose of our study was to investigate whether allozyme electrophoretic markers can be used to differentiate between Phytobia species occurring on eight host tree species belonging to five different genera in three plant families. An allozyme-based identification method would circumvent the need to rear the larvae to adults, facilitate rapid identification, and provide tools for research aiming at management of Phytobia.

Furthermore, we wanted to study whether the host ranges of the species can be explained by the phylogenetic relationships of the host plants, because it has been suggested that endophagous (mining) larval habits may enhance the probability of parallel cladogenesis (= cospeciation) between insects and their hosts (Farrell and Mitter, 1990, 1998).

Materials and methods

Life history of Phytobia

All Phytobia species that have been studied are univoltine (Kangas, 1935, 1949; Spencer, 1976). In the spring, the ca. 0.5 cm long, black females oviposit into growing shoots of their host trees (Barnes, 1933; Kangas, 1949; Ylioja et al, 1998; Ylioja, 2000). The larva starts mining down the shoot, and when it reaches the trunk, it continues downward, mining in the differentiating xylem just below the cambium (Gregory and Wallner, 1979; Ylioja et al, 1998). The feeding tunnel can be seen as a brownish 1–2 mm wide streak under the bark of the host tree (Spencer, 1976; Ylioja et al, 1998). The larvae may reverse their direction and mine up and down if they reach the base of the tree (Greene, 1914; Grossenbacher, 1915; Kangas, 1935; Hanson and Benjamin, 1967; Ylioja et al, 1998). In autumn, the ca. 2 cm long larvae of most Phytobia species exit the tree via a vertical slit that they cut in the bark; pupation and overwintering occurs in the ground, and the adults emerge the next spring (Barnes, 1933; Kangas, 1935, 1949; Hanson and Benjamin, 1967; Spencer, 1976). However, in some species (eg, P. aucupariae), the larvae overwinter under the bark and resume feeding the next spring (Grossenbacher, 1915; Kangas, 1949, 1955; Spencer, 1976; Spencer and Steyskal, 1986). In these species, the larvae exit the tree and pupate on the ground in late spring, and the adults emerge, mate, and oviposit later in the summer (Grossenbacher, 1915; Kangas, 1949, 1955; Spencer, 1976).

Sample collection and electrophoresis

Following Kangas (1935, 1949), we collected larvae of Phytobia from the surroundings of the Punkaharju Research Station in southeastern Finland (61°48’N, 29°20’E). Larvae were collected by peeling the bark off young 2–5 m tall trees representing eight host species: Betula pendula Roth (40 larvae from 12 trees), B. pubescens Ehrh. (60 larvae from 19 trees), Alnus incana (L.) Moench (25 larvae from 21 trees), A. glutinosa (L.) Gaertn. (20 larvae from 10 trees), Populus tremula L. (15 larvae from 11 trees), Salix phylicifolia L. (35 larvae from 25 trees), S. caprea L. (30 larvae from 19 trees), and Sorbus aucuparia L. (30 larvae from 23 trees). Of these, Betula and Alnus belong to the plant family Betulaceae, Populus and Salix to Salicaceae, and Sorbus to Rosaceae. Most collections were made between 28 July and 20 August in 1997; the larvae infesting Sorbus aucuparia were, however, collected from the same location in July 1998, because the larvae collected in the previous autumn were too small for electrophoresis. All larvae were stored in Eppendorf tubes at −80°C.

In all, 255 larvae were used for electrophoresis (Table 1). The samples were homogenized in 60 μl sample buffer (Ferguson, 1980), after which the tubes were centrifuged at 10 000 rpm for 4 min; 45 μl of the supernatant was used in standard starch gel electrophoresis according to the protocol outlined in Vuorinen (1984). The horizontal 12% starch gels were cut into 2 mm thick slices and stained for specific enzymes. Buffer systems were as in Roininen et al (1993) and Nyman et al (1998), except that Fumh-1 and Fumh-2 were scored from buffer B gels (see Nyman et al, 1998).

Table 1 Allele frequencies and sample sizes (N) at ten polymorphic loci in Phytobia larvae collected from eight different host species
Full size table

Ten variable enzyme loci were scored: phosphoglucomutase (Pgm, EC 5.4.2.2.), malic enzyme (Me, EC 1.1.1.40), malate dehydrogenase (Mdh, EC 1.1.1.37), isocitrate dehydrogenase (Idh, EC 1.1.1.42), aspartate aminotransferase (Aat-1 and Aat-2, EC 2.6.1.1), triose-phosphate isomerase (Tpi, EC 5.3.1.1), fumarase hydratase (Fumh-1 and Fumh-2, EC 4.2.1.2), and glycerol-3-phosphate dehydrogenase (G3pdh, EC 1.1.1.8).

Data analysis

Allele frequencies and sample sizes were calculated by using BIOSYS-1 version 1.7 (Swofford and Selander, 1989). Deviations from Hardy-Weinberg equilibrium (HWE) in individual loci were tested by Fisher’s exact test in the Genetic Data Analysis program (GDA; Lewis and Zaykin, 2001). Ten thousand permutations were used in the tests, and the obtained significance levels were adjusted using a sequential Bonferroni correction (Rice, 1989). Linkage disequilibrium between pairs of polymorphic loci was tested similarly, but the genotypes in individual loci were preserved in the permutation procedure.

Clustering analyses were performed using the TFPGA version 1.3 program (Miller, 1997). UPGMA clustering analyses were based on Nei’s (1978) unbiased genetic distances, and cluster support was inferred by bootstrapping 1000 times over loci. Because the samples collected from S. aucuparia and the Alnus species had no alleles in common (Table 1), genetic distances between these two groups could not be calculated. Consequently, two separate clustering analyses were performed: when the S. aucuparia sample was included, the Alnus samples were excluded, and vice versa.

Results

Allele frequencies and sample sizes in individual loci are presented in Table 1. Seven statistically significant deviations from HWE were found in the samples. With the exception of the Idh locus in the B. pubescens sample, all deviations occur in the Pgm and Aat-2 loci in all Populus and Salix samples (all Bonferroni corrected P:s < 0.05), caused by a deficiency of heterozygotes. In addition, these two loci are in statistically significant linkage disequilibria in all samples collected from these three host species (all uncorrected P < 0.02). In the other samples, no statistically significant linkage disequilibria were found.

Based on these observations, the individuals collected from P. tremula, S. phylicifolia and S. caprea were sorted according to their genotype at the Pgm locus, at which heterozygotes between some alleles were totally missing (only Pgm51/66 heterozygotes were found); the resulting ‘sorted’ samples are shown in Table 1 as three putative species (‘Salicaceae 1-3’), and their corresponding genotypes and hosts are shown in Table 2. Because of the linkage disequilibrium between the Pgm and Aat-2 loci, the sorting also leads to clear differences in allele frequencies at the Aat-2 locus (Tables 1 and 2). Three individuals having missing data in the Pgm locus were excluded from the sorted data set, as was one individual having a typical Betula/Alnus genotype. This last individual may represent an oviposition error (see Discussion). In the sorted data set, most deviations from HWE are no longer statistically significant in the ‘Salicaceae 1-3’ samples, but there is still a statistically significant deviation in ‘Salicaceae 3’ at the Aat-2 locus (P < 0.001). This may be caused by the fact that weakly stained Aat-244/156 heterozygotes may have been inadvertently scored as Aat-2100/100 homozygotes, because the heterodimer is stained more strongly than the homodimers. However, excluding these individuals from the sorted data set has no meaningful effect on the results and conclusions in this study. The only statistically significant linkage disequilibrium in the sorted data set is between Fumh-1 and Fumh-2 in ‘Salicaceae 3’, but the disequilibrium is not significant if a Bonferroni correction is made (uncorrected P = 0.027).

Table 2 The three putative Salicaceae-feeding species that were obtained by sorting the individuals collected from Populus tremula, Salix phylicifolia, and S. caprea according to their genotype at the Pgm locus (see Results), and the host plants on which these individuals were found. Because of the linkage disequilibrium between Pgm and Aat-2, the sorting also results in clear differences in allele frequencies at the Aat-2 locus (see Table 1)
Full size table

The UPGMA clustering analysis using the original data set (Figure 1a) shows that the samples collected from the two Betula species and the two Alnus species form a distinct and well supported cluster. The samples collected from Populus tremula and the two Salix species are grouped together in a more heterogeneous cluster. The sample collected from S. aucuparia is extremely different from the other samples. When the clustering analyses are redone using the sorted data set, the results are otherwise similar to the original data, except that the heterogeneity within the Salicaceae-feeding cluster is increased (Figure 1b).

Figure 1
figure 1

UPGMA clustering dendrograms of Phytobia samples collected from eight different host tree species, based on pairwise Nei’s (1978) unbiased genetic distances calculated from 10 polymorphic allozyme loci. Host species are shown to the right of the dendrograms, numbers above branches are bootstrap proportions (%) from 1,000 replicates. Results are shown (a) for the original data set, and (b) for a data set in which the Salicaceae-feeding individuals have been sorted into three putative species (Salicaceae 1-3) according to their genotype at the Pgm locus (see Results and Table 2 for explanation, and list of hosts for these species). In both cases, analyses including the S. aucuparia sample were made while excluding the Alnus samples, and vice versa (see Materials and methods). Note the cut in the scale.

Full size image

Thus, the results indicate that there are five Phytobia species that feed on the eight host species included in this study: the Phytobia larvae collected from the Betula and Alnus species represent a single species, as do the larvae from S. aucuparia. However, the larvae from Populus tremula and the two Salix species apparently represent three species with partially overlapping feeding ranges. The distribution of individuals on different host species is statistically significantly different between the Salicaceae-feeding species (χ24 = 40.23, P < 0.001; see Table 2). The assumptions of the χ2-test are violated because some individuals were collected from the same tree, and thus they may be siblings (however, larvae representing different species could be found on single tree individuals). Thus, the test was repeated so that if multiple larvae representing a given species had been collected from a given tree individual, they were treated as only one larva, ie, it was assumed that they represent the offspring of a single female. The result of the analysis remains essentially the same (χ24 = 25.76, P < 0.001).

The aforementioned results are supported by the results obtained in a pilot study, conducted in 1997 using larvae collected from five of the host species in 1996 (n = 9–41 larvae per host species; results not shown). In the pilot survey, allele frequencies were close to the ones observed in this study and, consequently, the larvae collected from the two Betula species and A. incana were very different from the samples collected from the two Salix species. Likewise, all ‘Salicaceae 1–3’ species were observed in the Salix samples in the preliminary study.

Discussion

Although Phytobia has been extensively studied in Europe, the number of species that actually exist in the region is still unclear. As stated above, Kangas (1935, 1949) considered that there are five Phytobia species that feed on the tree species represented in this study: P. betulae on Betula spp., P. aucupariae on S. aucuparia, and three species (P. cambii, P. barnesi, and P. tremulae) on hosts belonging to the Salicaceae. Later, Spencer (1976) considered that the presence/absence of the mid-tibial bristle, which Hendel (in Barnes, 1933) and Kangas (1949) had mainly used to separate the Salicaceae-feeding species from each other, was not a clear enough character for species separation. Consequently, Spencer (1976, see also Spencer, 1990) considered that there exists only one European Salicaceae-feeding species (ie, P. cambii), and this view has later been followed by others (Martinez et al, 1985; von Tschirnhaus, 2000). In the end, von Tschirnhaus (2000) even synonymized P. betulae with P. cambii, which would mean that there would be only two European Phytobia species on the hosts in this study: P. cambii on Betulaceae and Salicaceae hosts, and P. aucupariae on S. aucuparia.

The results from our study clearly support Kangas’ (1935, 1949) interpretation of the species situation in Phytobia. The five species that can be distinguished in our study are characterized by fixed differences in various loci. The clearest one is the S. aucuparia species (= P. aucupariae), and the apparently single species on Betula and Alnus (= P. betulae) is distinguished by the Pgm100 allele (Table 1). One individual having this allele was found on S. caprea, but because the individual was homozygous for the allele, it seems more likely that it represents an oviposition error rather than evidence for hybridization between species. Oviposition errors are probably a common phenomenon in phytophagous insects, although quantitative studies are rare (Larsson and Ekbom, 1995). The three Salicaceae species can be separated on the basis of their genotype at the Pgm locus (Table 2). We note that these Salicaceae-feeding types were also found in the preliminary survey that was conducted with samples collected from the study area in the previous summer (see Results). The three species apparently represent P. cambii, P. barnesi, and P. tremulae, but further study is needed in order to establish which ‘allozyme species’ corresponds to which morphospecies.

Studies on insect host races have shown that host-dependent disruptive selection can cause and maintain large differences in allele frequencies at allozyme loci, even when there is considerable gene flow between the host races (Feder et al, 1997; Filchak et al, 2000; Drès and Mallet, 2002). In the case of Phytobia, host-associated disruptive selection seems an unlikely explanation for the results above, because there are fixed differences at single or multiple loci and, more importantly, the ‘Salicaceae 1–3’ species overlap in their host use. Thus, our results indicate that all five species that can be identified on the basis of allozyme data are true biological species (sensu Mayr, 1963) that do not interbreed.

At the species level, the host associations can clearly be explained by the taxonomic affinities of the hosts. Betula and Alnus, which support a single Phytobia species (= P. betulae), belong to the plant family Betulaceae. In Central Europe, P. betulae has been reported to feed on Corylus and Carpinus, which also belong to the same family (Spencer, 1990). Sorbus aucuparia, which supports a very distinct Phytobia species (= P. aucupariae), belongs to the family Rosaceae, and the three species (’Salicaceae 1–3’) with overlapping host associations all feed on trees belonging to the Salicaceae. Thus, the results conform to the common pattern that when an herbivorous insect is oligo- or polyphagous, the hosts tend to be closely related (Bernays and Graham, 1988).

However, on a larger scale it is evident that the evolution of host use in Phytobia cannot be explained by simple parallel cladogenesis (= cospeciation) between the flies and their host plants. Although UPGMA clustering is sensitive to rate heterogeneity in the lineages that are being studied (Swofford et al, 1996), the species occurring on S. aucuparia is so different from the four others that it is highly likely that it was the first one to diverge. In fact, the genetic distance between this species and the other samples is more typical of insects in separate genera (see Emelianov et al, 1995). This is also supported by the fact that P. aucupariae has a rather distinct morphology and life cycle: whereas the larvae of species that feed on Betulaceae and Salicaceae exit the trees in the autumn to pupate and overwinter on the ground, the larvae of P. aucupariae overwinter in the trees and pupate on the ground in the next spring (Kangas, 1949; Spencer, 1976). Species having a rather similar morphology and life cycle are also known from other tree species belonging to Rosaceae (Grossenbacher, 1915; Kangas, 1955; Spencer and Steyskal, 1986). However, the Rosaceae has probably been colonized independently at least twice, because some of the Rosaceae-feeding species overwinter as pupae on the ground (Spencer, 1976, 1990; Spencer and Steyskal, 1986). In any case, the branching pattern observed in our study, ie, (Rosaceae(Betulaceae, Salicaceae)), clearly contradicts the current hypothesis of the phylogenetic relationships between the host species represented in this study, ie, (Salicaceae(Betulaceae, Rosaceae)) (Angiosperm Phylogeny Group, 1998). A somewhat similar result was found by Scheffer and Wiegmann (2000) in the case of agromyzid Phytomyza leaf miners that mine in the leaves of hollies (Ilex spp.), because they also could rule out an explanation based on cospeciation. Together, these results from two distantly related agromyzid taxa provide further support for the view that strict parallel cladogenesis between insects and plants is a rare phenomenon (Mitter et al, 1991).

It has been suggested that mining larval habits may predispose insects to cospeciation with their hosts, but clearly the situation in Phytobia is very different from the situation in the Phyllobrotica and Tetraopes beetles discussed by Farrell and Mitter (1990, 1998). In contrast to the tissues of the hosts of these beetles (the Lamiales and Asclepiadaceae, respectively), the differentiating xylem in deciduous trees is probably not strongly defended by secondary compounds and, thus, host chemistry may present only weak barriers for colonization by Phytobia. This implies that larval endophagy per se does not affect the likelihood of parallel cladogenesis, but may do so in combination with heavy chemical defence of the hosts (see also Farrell and Mitter, 1990, 1998).

In conclusion, it is evident that in Finland the genus Phytobia is represented by at least five reproductively isolated species, and this is probably also true elsewhere in Europe. Allozyme electrophoresis appears to be a powerful tool for species identification in this genus. More thorough sampling of different tree species is needed, because it is likely that new Phytobia species can be found on unstudied, economically unimportant tree species. The host use of the individual species in this study seems to be explained by the taxonomic affinities of the host plants, but on a larger scale the evolution of host-plant associations cannot be explained by strict cospeciation between Phytobia and their hosts. A phylogenetic study of the whole Phytobia would add to the knowledge of evolution in the Agromyzidae, as well as in phytophagous insects in general.