Introduction

For over 20 years, a threat to the esthetic values of horse-chestnut has been posed by the inconspicuous moth horse-chestnut leaf miner (Cameraria ohridella Deschka & Dimić, Lepidoptera: Gracillariidae), which was first recorded in Macedonia in the 1970s and rapidly spread throughout Central and Western Europe (Valade et al. 2009). The economic cost arising from the protection of Aesculus trees against this pest is substantial. For example, in Germany current costs of five major urban centers range from 10,020,000 to 33,800,000 € annually, but if trees in green spaces died, replacement costs would run up to 10.7 billion € (Reinhardt et al. 2003).

The species of the genus Aesculus are characterized by diverse susceptibility to C. ohridella. The main host of this pest is the white horse-chestnut (Aesculus hippocastanum L.), which is native to the Balkans and since the end of the sixteenth century has been introduced throughout Europe as an attractive ornamental tree (Prada et al. 2011). However, considerable variation in susceptibility to C. ohridella has been observed between Ae. hippocastanum individual trees (Straw and Tilbury 2006; Irzykowska et al. 2013). Also susceptible to C. ohridella is the Japanese horse-chestnut (Aesculus turbinata Blume). Resistant species include Aesculus indica (Wall. ex Camb.) Hook, Aesculus californica (Spach) Nutt., Aesculus parviflora Walter, Aesculus assamica Griff., and Aesculus wilsoni Rehder, which are of Asian and/or of North American origin (Straw and Tilbury 2006; D’Costa et al. 2013). Likewise, some species of North American origin (Aesculus sylvatica Bartram, Aesculus pavia L., Aesculus flava Aiton, Aesculus glabra Wild) are colonized by females that deposit eggs on the leaf surface but most larvae die during development (Kenis et al. 2003; Ferracini et al. 2010). Also resistant to C. ohridella is the hybrid of European Aesculus hippocastanum L. and American Aesculus pavia L. named red horse-chestnut (Aesculus × caenea Hayne), a popular tree in Europe (Irzykowska et al. 2013).

Resistance or susceptibility to pests is the result of plant–pest interaction. There are three mechanisms of resistance to herbivores, which are based on the plant properties. Antixenosis includes all plant traits which prevent oviposition and feeding and makes the plant an unsuitable host. Antibiosis involves harmful effects of plant on insect and tolerance that is the ability of plant to withstand or repair damage caused by the herbivore. The essential aspect of the relationships between host plants and insects is the chemical composition of leaves, which is decisive for the occurrence of interactions that facilitate or hamper feeding (Mithöfer and Boland 2012).

Plant–pest communication can take place through visual stimuli (Lev-Yadun and Gould 2009; Kocíková et al. 2012). The green leaf blades of host plants are willingly attacked by herbivores whose caterpillars feed on their tissues (Karageorgou and Manetas 2006). Chloroplast pigments are necessary for proper function of photosynthesis and are indirectly responsible for the synthesis of carbohydrates, which are a valuable nutrient for phytophagous insects (Jonas and Joern 2013). Female horse-chestnut leaf miners also react to the appearance of leaves and deposit their eggs on leaves with a large proportion of green area, which are not already occupied by mines, because such leaves have an abundance of food for larval growth and development (Johne et al. 2006a). The presence of anthocyanin pigments in leaves may also be a visual signal to insects (Archetti and Brown 2004). It was observed that red cabbage having a higher anthocyanin concentration than green cabbage was more attractive for caterpillars of lepidopteran pests (Coelho 2004). On the other hand, red coloration of leaves may also act as a warning signal for plant-eating insects, indicating that plants are well defended or have poor nutritional quality (Lev-Yadun and Gould 2009; Karageorgou and Manetas 2006; Maskato et al. 2014). High anthocyanin level may provide a visual cue to insect herbivores about a relatively high level of leaf total phenolic compounds which inhibit feeding of larvae and lead to their death (Karageorgou et al. 2008; Bouzouina et al. 2012; Oszmiański et al. 2014). In addition, feeding of herbivorous insect may induce accumulation of phenolics in leaves of host plants (Bernards and Båstrup-Spohr 2008). Chemical compounds that may act as defensive agents as well as alimentary deterrents are flavonols, which inhibit the growth of larvae, causing an extension of the larval period and increased larval mortality (Adeyemi et al. 2010; Martemyanov et al. 2012).

The availability of high quality and quantity of nutritional compounds (proteins, carbohydrates) is also an essential factor in the feeding of many herbivorous insects (Zhang et al. 2011). Good nutrition is necessary for the longevity of oviposition period, lifetime fecundity and oviposition rate in Lepidoptera (Cahenzli and Erhardt 2013; Marchioro and Foerster 2013). Plant amino acids are a key source of nitrogen, which is a significant factor in egg production (Cahenzli and Erhardt 2013). Carbohydrates, which are also an important component in the diet of many Lepidoptera, cover energy requirements and provide carbon skeletons for somatic development, fitness and reproductive output (Glendinning et al. 2007; Marchioro and Foerster 2013). High activity of amylase, maltase and saccharase which was found in the midgut of C. ohridella suggests that starch, maltose and saccharose are primary carbohydrate components of their diet (Stygar et al. 2010).

The aim of the study was to examine the content of chemical compounds which may act as attractants (chloroplast pigments, anthocyanins), defensive substances (anthocyanins, phenols and flavonols) and nutrients (free α-amino acids and carbohydrates) in four Aesculus trees differing in susceptibility to C. ohridella. The research was focused on Japanese horse-chestnut (Aesculus turbinata), which is susceptible to this pest, resistant painted buckeye horse-chestnut (Aesculus × neglecta var. lanceolata Sargent, also known as Aesculus sylvatica Bartram), and two white horse-chestnut (Aesculus hippocastanum) trees that are growing close to each other and are not equally susceptible to this pest. One of them was colonized by the pest earlier, and the leaves are damaged to a greater extent than on the other tree.

These studies were undertaken in order to clarify the differences in resistance of the examined horse-chestnut trees to C. ohridella that we have observed for several years. According to the literature data, we assume that horse-chestnut trees not infested or those on which the leaf miner does not continue feeding contain compounds acting harmfully on the pest population and do not contain nutritional compounds for this pest (Johne et al. 2006b; Kropczyńska-Linkiewicz et al. 2008). On the other hand, trees inhabited by C. ohridella are likely to contain substances working as attractants that make it easier for the insect to find the host plant and may have adequate quantity and quality of nutritional compounds (Johne et al. 2006a; Kropczyńska-Linkiewicz et al. 2008; Svatos et al. 2009; Marchioro and Foerster 2013). The differences in the chemical composition of the leaves of these Aesculus trees will allow the identification of substances which may be important in the interaction between C. ohridella and these trees. The obtained results can be used for further phytochemical studies on larger populations of horse-chestnut species.

Materials and methods

Weather data

Data on weather conditions in the successive years of study were obtained from the website of Poznań University of Life Sciences (http://www.au.poznan.pl/wogr/hobo/index.html). The measurements were performed by the meteorological Onset HOBO Weather Station installed in the Experimental Station of Poznań University of Life Sciences. Results are presented as means for each month in the period from March to October.

Plant material and sample collection

Four trees of the genus Aesculus were included in this study: two Aesculus hippocastanum trees growing at the campus of Poznań University of Life Sciences, one tree of Ae. turbinata and one of Ae. × neglecta, both located in the Botanical Garden of the Adam Mickiewicz University. Fully developed compound leaves were randomly chosen from the same location within the tree crown (similarly lighted place at the same height of each tree crown) every 14 days, at the same time of day (between 8 and 9 am) from May to September 2011 and 2012. Collected material was put into plastic bags and transferred to the laboratory within 30–60 min. Sample collection and laboratory assessment were carried out by two persons.

Estimation of damage of leaves by horse-chestnut leaf miner

The degree of leaf damage was evaluated visually by the same person by estimating the size of leaf blade occupied by mines in relation to the total leaf area and was expressed in percentages. Each time the assessment was performed on 15 compound leaves from each tree. After assessment the leaves were used for chemical analyses.

Chemical analyses

The content of chloroplast pigments in leaves was determined immediately after transfer to the laboratory. Plant material for the determination of other compounds was frozen in liquid nitrogen and stored at −20 °C until analysis. Only green parts of leaves (not occupied by mines) were taken for all evaluations. Chemical analyses were done during the entire growing season in the resistant Ae. × neglecta tree and the relatively resistant Ae. hippocastanum individual in both 2011 and 2012 and in the susceptible Ae. turbinata tree in 2012. However, on the sampling dates in which more than 60% of the leaf blade was occupied by mines chemical analyses were not carried out (Ae. turbinata tree from 14 July 2011, relatively susceptible Ae. hippocastanum tree from 14 July 2011 and 13 August 2012).

Each parameter measured for each tree, date and year was estimated in five independent replications. Each replication was a sample of plant material which derived from the middle leaflet of different compound leaves collected from the same tree.

Chloroplast pigments

Total chlorophyll and carotenoids content was estimated according to the method of Hiscox and Israelstam (1979). Leaf samples (100 mg) were cut into pieces and pigments were extracted at 65 °C using 5 cm3 of dimethyl sulfoxide (DMSO). Optical density of the extract was measured at 480, 649 and 663 nm. The content of total chlorophyll and carotenoids was calculated following the modified Arnon equations (Wellburn 1994) and was expressed in milligrams per gram of fresh weight (mg g−1 FW).

Anthocyanins

Plant material (500 mg) was homogenized with 3 cm3 of 0.5 N HCl, and centrifuged at 6000g for 10 min. Absorbance of the supernatant was measured at 530 nm (Wang et al. 2000). Anthocyanin content in leaf tissue was calculated using a calibration curve of cyanine chloride and was expressed in micrograms per gram of fresh weight (μg g−1 FW).

Total phenolic compounds

Leaf samples (250 mg) were cut into small pieces and extracted at 60 °C first with 5 cm3 of 96% ethanol for 3 min and then with 5 cm3 of 65% ethanol for 5 min. Each extract was poured into a test tube and it was filled up to the final volume of 10 cm3. To 0.5 cm3 of extract 0.1 cm3 of Folin-Denis reagent was added and exactly 3 min later 1 cm3 of 10% Na2CO3. After 20 min the absorbance was measured at 660 nm (Swain and Hillis 1959). The level of phenolics was calculated from the standard curve for coumaric acid and expressed in micrograms per gram of fresh weight (μg g−1 FW).

Flavonols

Leaf samples (500 mg) were cut into small pieces and homogenized with 5 cm3 of methanol, HCl and distilled water (90:1:1, v/v/v). The solution was stirred and heated (60 °C) for 10 min, cooled at room temperature for 15 min and centrifuged at 6000g for 30 min (Day 1993). The absorbance of the supernatant was measured at 254 nm with a UV/visible spectrophotometer (Jasco V-530 UV–VIS Spectrophotometer). Flavonol content in leaf tissue was calculated using the calibration curves of quercetin (Stefova et al. 2001) and was expressed in micrograms per gram of fresh weight (μg g−1 FW).

Free α-amino acids

The ninhydrin method was applied to estimate the level of amino acids (Baily 1962). Leaf samples (250 mg) were cut into small pieces and homogenized with 3 cm3 of ethanol. The homogenate was centrifuged at 6000g for 30 min. 0.1 cm3 of the supernatant was mixed with 0.9 cm3 of 0.4 M acetic acid buffer, pH 5 and 1 cm3 of ninhydric reagent (2 g of ninhydrin + 50 cm3 of 50% ethanol). The mixture was incubated at 100 °C for 15 min and then cooled. The absorbance of samples was determined at 570 nm. Amino acid content in leaf tissue was calculated using the calibration curves of alanine and was expressed in micrograms per gram of fresh weight (µg g−1 FW).

Total soluble carbohydrates

The anthrone colorimetric method was applied to measure carbohydrate content (Luo and Huang 2011). Leaf samples (250 mg) were cut into small pieces and extracted for 30 min at 60 °C using 4 cm3 of 80% ethanol. 0.5 cm3 of supernatant was mixed with 1 cm3 of anthrone reagent (25 cm3 H2SO4 + 5 mg anthrone reagent). After 15 min the absorbance of the colored reaction product was read at 620 nm. The amount of carbohydrates was calculated using a standard curve for glucose and was expressed in milligrams per gram of fresh weight (mg g−1 FW).

Statistical analysis

For evaluation of statistical differences between the analyzed individual trees of two species (Ae. turbinata vs. Ae. × neglecta) and two trees of the same species (Ae. hippocastanum) we used analysis of variance (ANOVA) with repeated measurements (n = 5), where the dependent variables were the data of successive sampling dates and independent variables were individual trees. All statistical analyses were conducted for the same number of sampling dates. The significance level was set at α = 0.05. The differences were considered as statistically significant when p values were less than or equal to α. All numerical analyses were performed using the STATISTICA 10 package.

Results

Weather conditions

The weather conditions were slightly different in 2011 and 2012. The average precipitation was higher in 2012, mainly due to higher precipitation in the period from May to August (Table 1). The average humidity was at similar levels for most months in 2011 and 2012, except for June, when it was much lower in 2011 than in 2012. The insolation measured on the basis of the number of sunny hours in 2011 was higher than in 2012 in five months (March, June, August, September and October), but in May it was similar in these two years. The average air temperature was lower in March and July, and higher in June, in 2011 in comparison to the same months in 2012, being at similar levels in the other months in both years.

Table 1 Weather conditions from March to October (monthly means) during 2011 and 2012

Damage of leaves by the horse-chestnut leaf miner

In both years, greater damage of leaves was found in the relatively susceptible individual of Ae. hippocastanum. Similarly, the susceptible Ae. turbinata was characterized by greater damage than the resistant Ae. × neglecta (Fig. 1). Moreover, in the relatively susceptible Ae. hippocastanum tree and susceptible Ae. turbinata earlier and greater damage of leaf blades was found in 2011 than in 2012. This effect was probably caused by 2-week-earlier start of feeding by the pest in 2011 than in 2012. Already in mid-July 2011, almost 90% of the leaf blade area was covered with mines in the relatively susceptible tree of Ae. hippocastanum, and 60% in Ae. turbinata. In 2012 at the beginning of September 90% damage of the leaf blades was observed in the relatively susceptible white horse-chestnut tree and only 30% in Ae. turbinata. The leaf blades of resistant Ae. × neglecta were inhabited by single females but symptoms of feeding were never observed.

Fig. 1
figure 1

The damage of leaf blades in Japanese horse-chestnut (Aesculus turbinata), painted buckeye (Aesculus × neglecta) and two trees of white horse-chestnut (Aesculus hippocastanum) in years 2011 and 2012

Results of chemical analyses

Statistically significant differences in the level of all estimated compounds throughout both years of experiments were observed between resistant Ae. × neglecta and susceptible Ae. turbinata (Table 2). In leaves of Ae. × neglecta higher levels of chlorophyll, carotenoids and α-amino acids were observed, while in Ae. turbinata there were higher levels of anthocyanins, phenols, flavonols and carbohydrates. Moreover, there were significant differences in time-dependent changes in the level of almost all examined compounds in these species, with the exception of chlorophyll in 2011 and carbohydrate and amino acids in 2012. In the two white horse-chestnut trees statistically significant differences were found only in the level of chlorophyll, anthocyanins, phenolics and flavonols, which were higher in the tree relatively susceptible to this pest (Table 3). The changes in the levels of chlorophyll, carotenoids and anthocyanins in these trees proceeded differently in time, in 2011. However, in 2012 different time-dependent changes were revealed in the level of almost all examined compounds with the exception of carotenoids and anthocyanins. The contents of particular compounds in the leaves of examined trees on successive sampling dates in 2011 and 2012 are shown in Figs. 2, 3, 4, 5, 6, 7, and 8.

Table 2 ANOVA results for the content of estimated compounds in the leaves of painted buckeye (Aesculus × neglecta) and Japanese horse-chestnut (Aesculus turbinata) in years 2011 and 2012
Table 3 ANOVA results for the content of estimated compounds in the leaves of two trees of white horse-chestnut (Aesculus hippocastanum) in years 2011 and 2012
Fig. 2
figure 2

Level of total chlorophyll in leaf blades of Japanese horse-chestnut (Aesculus turbinata), painted buckeye (Aesculus × neglecta) and two trees of white horse-chestnut (Aesculus hippocastanum) in years 2011 and 2012. Data are the means of five replications ± standard deviation marked as vertical bars. Significant differences between genotypes at the sampling dates are marked as asterisk

Fig. 3
figure 3

Level of carotenoids in leaf blades of Japanese horse-chestnut (Aesculus turbinata), painted buckeye (Aesculus × neglecta) and two trees of white horse-chestnut (Aesculus hippocastanum) in 2011 and 2012. Data are the means of five replications ± standard deviation marked as vertical bars. Significant differences between genotypes at the sampling dates are marked as asterisk

Fig. 4
figure 4

Level of anthocyanins in leaf blades of Japanese horse-chestnut (Aesculus turbinata), painted buckeye (Aesculus × neglecta) and two trees of white horse-chestnut (Aesculus hippocastanum) in 2011 and 2012. Data are the means of five replications ± standard deviation marked as vertical bars. Significant differences between genotypes at the sampling dates are marked as asterisk

Fig. 5
figure 5

Level of phenols in leaf blades of Japanese horse-chestnut (Aesculus turbinata), painted buckeye (Aesculus × neglecta) and two trees of white horse-chestnut (Aesculus hippocastanum) in 2011 and 2012. Data are the means of five replications ± standard deviation marked as vertical bars. Significant differences between genotypes at the sampling dates are marked as asterisk

Fig. 6
figure 6

Level of flavonols in leaf blades of Japanese horse-chestnut (Aesculus turbinata), painted buckeye (Aesculus × neglecta) and two trees of white horse-chestnut (Aesculus hippocastanum) in 2011 and 2012. Data are the means of five replications ± standard deviation marked as vertical bars. Significant differences between genotypes at the sampling dates are marked as asterisk

Fig. 7
figure 7

Level of α-amino acids in leaf blades of Japanese horse-chestnut (Aesculus turbinata), painted buckeye (Aesculus × neglecta) and two trees of white horse-chestnut (Aesculus hippocastanum) in 2011 and 2012. Data are the means of five replications ± standard deviation marked as vertical bars. Significant differences between genotypes at the sampling dates are marked as asterisk

Fig. 8
figure 8

Level of carbohydrates in leaf blades of Japanese horse-chestnut (Aesculus turbinata), painted buckeye (Aesculus × neglecta) and two trees of white horse-chestnut (Aesculus hippocastanum) in 2011 and 2012. Data are the means of five replications ± standard deviation marked as vertical bars. Significant differences between genotypes at the sampling dates are marked as asterisk

Both in 2011 and 2012, the pest-resistant Ae. × neglecta was characterized by significantly higher total chlorophyll content in the leaf blades compared to susceptible Japanese horse-chestnut (Fig. 2). Likewise, in the leaves of the relatively resistant Ae. hippocastanum total chlorophyll content in both 2011 and 2012 was significantly higher than in the relatively susceptible one. However, these differences occurred only from mid-May to mid-June (Fig. 2).

Throughout the 2011 growing season the carotenoid content in the leaf blades of Ae. × neglecta was significantly higher than in susceptible Ae. turbinata (Table 2; Fig. 3). By contrast, although in 2012 the differences between these Aesculus trees in the level of carotenoid pigments were statistically significant (Table 2), they were lower than in 2011 (Fig. 3). The level of carotenoids in leaves of Ae. × neglecta was higher than Ae. turbinata but only on four sampling dates (21 May, 4 June, 30 July and 13 August). In the leaves of the resistant Ae. hippocastanum tree in 2011 a slightly higher carotenoid level was observed than in the relatively susceptible one. However, in 2012 there was no difference in the level of these compounds between examined white horse-chestnut trees (Fig. 3).

The pest-susceptible Ae. turbinata was characterized by a higher level of anthocyanins than resistant Ae. × neglecta, both in 2011 and 2012 (Fig. 4). A slight differences in the level of these pigments between the two Ae. hippocastanum trees differing in susceptibility to this pest were found in 2011 and 2012, but only on two or three dates (Fig. 4).

The content of phenolic compounds in 2011 and 2012 was significantly higher in the leaves of susceptible Ae. turbinata than resistant Ae. × neglecta (Fig. 5). Similarly, a higher level of these compounds was observed in the leaves of the pest-susceptible Ae. hippocastanum tree compared to the relatively resistant one (Fig. 5). Furthermore, the content of phenolic compounds in resistant Ae. × neglecta remained at a similar level at successive dates of determination. In Ae. turbinata and both specimens of Ae. hippocastanum their level increased in the course of feeding. It should be noted that the Ae. hippocastanum tree relatively resistant to horse-chestnut leaf miner was characterized by a twofold higher level of phenolic compounds in 2011 than in 2012. This could be associated with the more rapid development of the pest in 2012.

In 2011 a slightly higher level of flavonols but only on two dates was observed in the leaves of Ae. turbinata compared to Ae. × neglecta, whereas in 2012 these differences were larger, found on four dates (Fig. 6). Similarly, the pest-susceptible Ae. hippocastanum was characterized by a higher level of flavonols than the relatively resistant tree of this species and the observed differences were more pronounced in 2012 than in 2011 (Fig. 6).

In both years of experiments α-amino acid content was slightly higher in the resistant Ae. × neglecta than the susceptible Ae. turbinata (Fig. 7). There were no significant differences in the levels of these compounds between the two examined Ae. hippocastanum trees in 2011. However, in 2012 a slightly higher level of α-amino acids was observed in the relatively resistant Ae. hippocastanum tree (Fig. 7).

In 2011 as well as in 2012 the level of carbohydrates was significantly higher in susceptible Ae. turbinata than resistant Ae. × neglecta (Fig. 8). Leaf carbohydrate content in both Ae. hippocastanum trees was at the same level in 2011 but at most dates throughout 2012 was significantly higher in the relatively susceptible tree (Fig. 8).

Discussion

Weather conditions (air humidity, sunlight and temperature) are basic factors that influence phytophagous insects directly and indirectly (Jaworski and Hiszczanski 2013). Temperature exerts a significant effect on the development of C. ohridella (Dimić et al. 2000). Depending on climate conditions, the horse-chestnut leaf miner produces two to four generations per year. Usually in a warm dry summer C. ohridella is able to complete three generations (Tilbury et al. 2004; Girardoz et al. 2007). The flight period of the adults emerging from overwintering pupae (first generation) begins at the end of April and the beginning of May. The second generation usually appears in June-July and the third in August–September (Kukula-Mlynarczyk and Hurej 2007; Pál et al. 2014) The final generation overwinters in the stage of pupae, which are known to be extremely frost tolerant and survive in the fallen leaves (Tomiczek and Krehan 1998; Pál et al. 2014). It was reported that the density of the first and second generation of C. ohridella moths was greater during a warm than during a cool summer, which resulted in greater damage of horse-chestnut leaves (Kukuła-Młynarczyk and Hurej 2007). However, during a cool summer only two generations were developed and less damage of leaves was observed (Girardoz et al. 2007). Our research has shown that horse-chestnut individuals more susceptible to the pest were characterized by greater damage of leaves in 2011 than in 2012. This could be a result of slightly higher air temperature in April and June of 2011 than 2012, this being the period of larval development and emergence of the adults of the first and second generation.

A substantial influence on the behavior of insects is exerted by chemical composition of the host plants (Johne et al. 2006c; Johne et al. 2008; D’Costa et al. 2014). The reciprocal ratio of plant pigments contained in the leaf blades determines the final color of these organs and constitutes a visual signal which facilitates adult Lepidoptera to locate the host plant (Kočíková et al. 2012). The present results revealed that total chlorophyll content in the leaves of Ae. × neglecta completely resistant to C. ohridella was significantly higher than in Ae. turbinata, which is occupied by this pest. Similarly, higher leaf chlorophyll level in the relatively resistant tree of Ae. hippocastanum was coincident with the period in which it was less colonized by the horse-chestnut leaf miner than the susceptible one. This could mean that the intense green color of the leaves does not correspond to the visual pattern responsible for the location of host plants by the pest moths. On the other hand, Johne et al. (2006a) observed that the females of horse-chestnut leaf miner deposit more eggs on green leaf blades than on those which are covered by mines, because the oviposition and larval mining cause changes in the profile of leaf volatile compounds and odor pattern of leaves which inhibit other individuals from oviposition on the same leaves.

It should be noted that Ae. turbinata susceptible to C. ohridella was characterized by a higher level of anthocyanins in leaf blades than Ae. × neglecta on all of the dates in both growing seasons. The content of these pigments was higher at the beginning of the growing season before the appearance of the larvae. Therefore, it can be assumed that a higher level of these red pigments affects leaf coloration, recognized as a visual stimulus by females of horse-chestnut leaf miner looking for a place to deposit eggs. Similarly, the Brassica rapa genotype which was characterized by high concentrations of anthocyanins was more attacked by herbivorous insects than green leaf genotypes (Coelho 2004). On the other hand, there are numerous reports showing that higher levels of these pigments may inhibit infestation of leaves (Karageorgou and Manetas 2006; Karageorgou et al. 2008; Lev-Yadun and Gould 2009). Karageorgou and Manetas (2006) believe that insects avoid leaves rich in anthocyanins because they usually contain a lot of phenolics which act as defensive compounds. In fact, the results show that leaves of Ae. turbinata more abundant in anthocyanins contained significantly more phenolics than Ae. × neglecta, but these compounds did not affect the pests adversely because larvae developed well on this species. These results did not confirm the hypothesis that anthocyanins are compounds which may offer increased protection against phytophages. The results of some other studies also indicate that foliar anthocyanins are not toxic to insect herbivores and do not inhibit feeding (Schaefer and Rolshausen 2006).

Among the secondary metabolites, plant phenolic compounds represent a very important group of defensive compounds which play an essential role in resistance to insect herbivores (Ossipov et al. 2001; War et al. 2012). The enhancement of phenolic acids was observed in Archis hypogaea L. infested with three different Lepidoptera pests (Sambangi and Usha Rani 2013). A rapid induced defense response caused by an increased concentration of simple phenolics and monoterpenes was observed in silver birch infested by Lymantria dispar (Martemyanov et al. 2012). In our study the concentration of phenolics in the leaf blades of Ae. turbinata as well as in both individuals of Ae. hippocastanum increased with the development of the pest, which could be considered as a manifestation of chemical defense against feeding. It seems that these horse-chestnut trees expressly defended themselves but it was ineffective. It probably results from the fact that C. ohridella has an effective defense system which includes antioxidant enzymes and detoxifying enzymes, existing in the larvae midgut, responsible for detoxification of allelochemicals produced by host plants (Żaak et al. 2012). The highest activity of these enzymes was found in the midgut of the second larval generation which correlated with increased toxicity of food caused by the accumulation of secondary metabolites in leaves (Żaak et al. 2012). Our research also revealed that during both growing seasons the concentration of phenolics in leaf blades of susceptible Ae. turbinata was higher than in resistant Ae. × neglecta. These results do not show that phenolic compounds may be one of the factors responsible for resistance of the examined Ae. × neglecta individual to C. ohridella. Likewise, the results obtained by D’Costa et al. (2014) seem to contradict the role of leaf phenolics in resistance to the pest, because they showed that leaves of species susceptible (Ae. hippocastanum, Ae. turbinata) to C. ohridella were characterized by a higher level of phenolic compounds than leaves of resistant species (Ae. chinensis, Ae. indica) but also had a similar level of these compounds compared to the resistant Ae. flava. The results of other research revealed that red horse-chestnut resistant to C. ohridella was characterized by higher content of phenolic compounds, especially (−)-epicatechin and procyanidins, as well as a much higher level of saponins, compared to susceptible white horse-chestnut (Kukula-Mlynarczyk et al. 2006; Oszmiański et al. 2014). It may suggest that resistance to the pest depends on the qualitative rather than the quantitative profile of phenolic compounds.

Flavonols are compounds that have antinutritive activity and inhibit insect development (Adeyemi et al. 2010; Martemyanov et al. 2012). A significant increase in the level of flavonols was observed in leaves of the Ae. turbinata tree in July 2012. After this date the damage of leaf blades did not increase. One can assume that the increase of flavonols could be caused by C. ohridella larval mining and resulted in the reduction of oviposition by conspecific females from the next pest generation. Therefore, the inhibition of further damage may be caused by induced chemical defense in which flavonols play an important role (War et al. 2012).

An important factor which stimulates larvae to feed is the abundance of nutrients in leaves (Bede et al. 2007). Insects more willingly select leaves that are rich in amino acids and carbohydrates (Glendinning et al. 2007). The level of amino acids in the leaves may be an important factor affecting the size of the phytophagous population. Insects that feed on tissues rich in amino acids grow faster, are larger and more fertile, and their progeny have a greater chance of survival in comparison to insects which feed on food containing a low level of these compounds (Cahenzli and Erhardt 2013; Marchioro and Foerster 2013). Our results do not confirm that the level of amino acids is a factor which affects the feeding of the horse-chestnut leaf miner. In both years throughout May and June the pest-susceptible Ae. turbinata had a lower amino acid content in leaf blades than the resistant Ae. × neglecta, whereas in both specimens of Ae. hippocastanum the content of amino acids was at a similar level throughout both growing seasons. Similarly, in other studies amino acid level in leaves of Aesculus species was not a trait which favored oviposition preference of C. ohridella and larval feeding (D’Costa et al. 2014). In the midgut of C. ohridella there are many enzymes catalyzing carbohydrate degradation, which indicates that these compounds are important nutrients for this pest (Stygar et al. 2010). The results of our study confirm the important role of carbohydrates in the stimulation of feeding of the horse-chestnut leaf miner. The pest-susceptible Ae. turbinata tree was found to have, in both growing seasons, a significantly higher level of leaf carbohydrates than the resistant Ae. × neglecta, while the relatively susceptible Ae. hippocastanum tree contained a slightly higher level of carbohydrates than the relatively resistant ones, but only in 2012. Slightly different and ambiguous results were obtained by D’Costa et al. (2014), who found that Ae. hippocastanum and Ae. turbinata susceptible to C. ohridella had similar carbohydrate contents compared to the resistant Ae. chinensis and Ae. indica species, and all these species had lower carbohydrate levels than Ae. flava, resistant to the pest.

In conclusion, the obtained results revealed that traits which may influence oviposition preference and favor the feeding of C. ohridella include anthocyanin and carbohydrate levels. The higher leaf carbohydrate content in the susceptible Ae. turbinata tree compared to resistant Ae. × neglecta suggests that the leaves of the former tree are more valuable food resources for larvae of this pest. Moreover, it may be speculated that the higher anthocyanin level in leaves of the Ae. turbinata specimen affects the leaf color, which may be an attractive visual stimulus for females of C. ohridella. However, the differences in the level of these compounds between the two Ae. hippocastanum trees, although statistically significant, are slight and do not explain their different susceptibility. The current results demonstrate that the examined leaf traits do not constitute a complete description of relationships between C. ohridella and examined trees of the genus Aesculus. The chemical composition of horse-chestnut leaves is probably more complex, and many other biochemical and physiological traits are responsible for the multifaceted interaction between this pest and different genotypes of Aesculus sp. The important results of Irzykowska et al. (2013) showed significant inter-species genetic variation within an Ae. hippocastanum population grown in urban green spaces, and 17 molecular markers strongly associated with their susceptibility to C. ohridella were found. Broader and more extensive investigations are needed in order to determine the traits that make some Aesculus trees more attractive and why they are willingly colonized by this pest while others are not colonized.

Author contribution statement

MP—experimental design, sample collection and chemical analysis in 2011 and 2012, preparation of the manuscript; HB—data analysis, discussion of results and preparation of the manuscript; JW—chemical analysis in 2012; KM and MG—statistical analysis and interpretation of results.