THE MORPHOLOGICAL VARIATION OF MYZUS PERSICAE ( HEMIPTERA : APHIDIDAE ) FROM PEACH AND TOBACCO IN SERBIA AND MONTENEGRO 1

Multivariate morphometric analysis was used to compare nine characteristics of 47 populations of Myzus persicae (Sulzer) originating from two host-plants, peach and tobacco, from 13 localities in 2004 and 34 localities in 2005, in Serbia and Montenegro. Multivariant discriminant analysis showed there to be a distinct discrimination between the populations from the peach and tobacco host-plants. The most important discrimination characteristics are the ultimate rostral segment length and processus terminalis length, which are greater in the aphids from tobacco than in those from peach. This is the first indication that in this part of Europe there are two subspecies: M. persicae (Sulzer) and M. persicae nicotianae Blackman.


INTRODUCTION
Myzus persicae (Sulzer) (Hemiptera: Aphididae) is an extremely polyphagous aphid species that feeds on over 400 plant species from 40 different families, and it is a major pest to many crops (Blackman and Eastop, 2000).In Serbia, M. persicae was found on more than 40 herbaceous plants (Petrović, 1998).It has two overwintering strategies: holocyclic with Prunus persicae (L.) as the primary host, and anholocyclic on many secondary hosts (Blackman and Eastop, 2000;Petrović-Obradović, 2003).For very long time populations of M. persicae on tobacco, Nicotiana tabacum L., have been considered to be different from populations on other plants (Blackman and Eastop, 2007).Multivariate morphometric analysis has revealed that populations of M. persicae feeding on tobacco are morphologically distinct from those on other host-plants, and the tobaccofeeding form has been given the name M. nicotianae Blackman (Blackman, 1987).However, several studies using molecular methods have provided evidence that the aphids on tobacco are not distinct at the species level from other populations of M. persicae (Clements et al., 2000a;Clements et al., 2000b;Field et al., 1994;Margaritopoulos et al., 1998).Also, it was considered that M. nicotianae populations were permanently parthenogenetic until holocyclic populations were found in Greece on peach (Margaritopoulos et al., 2002).These data suggest that tobacco-feeding aphids cannot be considered a valid species, and Eastop and Blackman (2005) proposed that the tobacco-adapted form should be called Myzus persicae ssp.nicotianae.
The morphological differences between tobacco-adapted and non-tobacco-adapted forms were investigated mostly in Greece, Italy and Japan (Margaritopoulos et al., 2003;Margaritopoulos et al., 2007).One sample from Serbia was included in studies of the morphological variation within and between the populations of the group of M. persicae (Blackman, 1987).But there has not been detailed research into the morphology of populations of M. persicae from peach and tobacco in this part of Europe (Serbia and Montenegro).For these reasons it was deemed important to investigate.

MATERIAL AND METHODS
The aphid samples were collected from tobacco fields and peach orchards in Serbia and Montenegro (Fig. 1) in the spring and summer of 2004 and 2005.The infested leaves were placed in plastic bags.The bags were placed in isolated plastic containers containing ice packs and transferred to the laboratory.Apterous adult females from each sample were collected and preserved in tubes filled with ethyl alcohol (75%) until slide preparation.The aphids were mounted on slides according to the method of Blackman and Eastop (1984).Aphids from a total of 47 samples were measured, 17 of which were collected from peach (140 specimens) and 30 from tobacco (278 specimens).A minimum of four and maximum of 12 apterous adult parthenogenetic females were measured from each sample.
Nine variables (characteristics, traits) were measured: the length of the third antennal segment (ant III), the length of the base of the sixth antennal segment (base VI), the length of the terminal process of the sixth antennal segment (pt), the length of the ultimate rostral segment (urs), the length of hind femur (hf), the length of second segment of the hind tarsus (ht2), the length of siphunculus (ls), the maximal width of the distal swollen part of the siphunculus (mws) and the length of cauda (lc).The methods of measurement are illustrated by Ilharco and van Harten (1987).All measurements were carried out with a phase contrast microscope (LEICA DMLS).
All variables that were used in the analyses followed normal distribution with homogeneity of variance.The data were tested for normality using Kolmogorov-Smirnov and Shapiro-Wilk tests.
When there are several dependent variables, multiple tests of (likely) correlated dependent variables (Tabachnik and Fidell, 1996) suffer from inflated Type I errors (Zar, 1999).
A multivariate analysis of variance (MANOVA) allows for the comparison of the population means of all variables of interest at the same time (multivariate response), rather than considering multiple responses as a suite of univariate responses (Zar, 1999).This reduces the magnitude of Type I errors.The statistical significance of MANOVA can be determined in a variety of ways.The most often used statistic test of Wilks' lambda was applied (Zar, 1999).
One-way MANOVA was used to examine the effects of the host-plant on M. persicae morphological variation.The fixed factors were the two host-plants (peach and tobacoo).When the MANOVA was statistically significant, subsequent univariate ANOVAs were performed to elucidate which responses contributed to the significant multivariate response.For this purpose a Unequal N HSD post-hoc test was used.Additionally, to describe and interpret the effects from the MANOVA, a multivariate discriminant analysis (DA) was used following MANOVA.Discriminant analysis was employed on all the data in order to determine the relative importance of characteristics as discriminators between a priori groups and the relative positions of the centroids of those groups (Manly, 1986).In addition, canonical variables were computed.All statistical analyses were conducted using the Statistica 6 software package (StatSoft, Inc 2001).
All the aphid samples from tobacco are from peach-growing regions and were collected during summer, while samples 4P, 5P, 6P, 7P, 8P, 9P, 10P and 11P were collected in tobacco-free areas from spring to early summer.

RESULTS
Descriptive statistics of the quantitative traits of M. persicae are given in Table 1.The one-way MANOVA of M. persicae from the two host-plants revealed the significant effects of host-plant interaction on M. persicae morphological variation (p<0.001,Table 2).The univariate results for each of the traits are shown in Table 2. Results from the Unequal N HSD post-hoc test revealed that only the variable mws (maximal width of distal swollen part of siphunculus) is nonsignificant.
The result of the discriminant analysis showed there to be distinct discrimination between the populations from tobacco and peach based on the first canonical axis (CVA1) (Fig. 2).
The total correct percentage of the classification matrix of all the groups was high (97.368%).
From the standardized coefficients for the canonical variables it is evident that the first canonical variable (CVA1) describes 63.43 % of the total variability; the first and second together (CVA1+CVA2), 76.88 %; the first, second, and third, 84.31 %.The length of the urs (ultimate rostral segment) and length of the pt (processus terminalis) contribute most to this discrimination.The following characteristics contribute, but to a lesser extent: length of hf (hind femur) and length of ls (siphunculus) (Table 3).
The result of DA showed the most important and distinct discrimination to be between the populations from tobacco and peach based on the first canonical variable (Fig. 2).
Since the first and second discriminatory axes describe the bulk of variability (76.88%), discrimination is clearly evident in Fig. 2 between the populations of aphids from peach (left-hand grouping of centroids, numbers colored in blue) along the first discriminatory axis, and the populations from tobacco (right-hand grouping of centroids, numbers colored in black).
This function is a positive coefficient for the variables urs, pt, hf, and ls, but has negative weight for the variables mws and ht2 (Table 3).Both subspecies are found in two basic color forms, green and red.The green form was dominant in the peach, while the red form was most numerous in tobacco.The aphid from peach is most often light-green, while the subspecies from tobacco is dark-red.The green form of aphid from tobacco is considerably darker than the green form from peach.
The symptoms caused by these aphids on the given hosts are also different.Both subspecies form dense colonies, most often on the reverse side of the leaf, but the peach aphid curls the peach leaves, whereas the tobacco aphid causes no deformations to tobacco leaves.Both subspecies are economically very significant pests.

DISCUSSION
In the present study, morphological variation was examined in M. persicae populations from peach and tobacco host-plants.As previous results have shown that morphological characteristics are stable after many years of feeding on the same secondary host (Blackman, 1987;Margaritopoulos et al., 2000), such characteristics were measured in the aphids collected from the field.Nine characteristics were measured, and multivariate morphometric analysis revealed that aphids feeding on peach are morphologically distinct from those feeding on tobacco.The most important discriminative characteristics are the lengths of the ultimate rostral segment length and processus terminalis, which are greater in populations from tobacco.
Aphids with a longer last rostral segment, processus terminalis, hind femur, and siphunculus are tobacco aphids and on the basis of these characteristics it is easy to separate the aphids from peach and tobacco host-plants.
Canonical discriminant analysis clearly indicates that aphids from peach differ morphologically from those from tobacco, and that the differences are great enough for us to regard the populations from these two hosts in Serbia and Montenegro as separate subspecies: M. persicae (Sulzer) and M. persicae nicotianae Blackman.
All aphid samples from tobacco are from peachgrowing regions, meaning that both subspecies exist at the same locations.The clones originating from peach were not found on tobacco plants.In southern Italy, tobacco-adapted and non-tobacco-adapted forms co-exist in the same region but they have different reproductive strategies.The tobacco-adapted form has no ability to reproduce sexually (Margaritopoulos et al., 2003).Holocyclic clones of the subspecies from tobacco were not found in Spain, Germany and France (Kephalogianni et al., 2002) But, in northern Greece and Japan, in the main peach-growing regions, the tobacco-adapted form has the ability to overwinter as diapause eggs on peach (Margaritopoulos et al., 2002, Margaritopoulos et al., 2007).In Serbia and Montenegro the life cycle is not known at the moment.
The present investigations indicate the existence of two subspecies that clearly differ morphologically.However, in order to determine the degree of separation of these subspecies and the possibility of their imminent divergence into two species (a question considered by Eastop and Blackman (2005)), further efforts need to be devoted to the study of the life cycle and research at the molecular level.

Fig. 2 .
Fig. 2. Plots of scores (means of canonical variables) of the first two canonical axes (CVA1 and CVA2) of 47 populations of Myzus persicae from two different host plants and geographically distant areas in Serbia and Montenegro (numbers = population means = centroids) (blue numbers -samples from peach populations and black numbers -samples from tobaco populations).
Mean values for nine morphological traits of 47 Myzus persicae populations from peach and tobacco (the measurement unit is μm).