Recently, progress has been made in our understanding of speciation in herbivorous insects, and mathematical models demonstrate that adaptive speciation driven by selection on ecological traits is theoretically plausible (Kondrashov and Kondrashov 1999; Dickman and Doebeli 1999; Doebeli 2005). However, as yet not much work has been done on the actual nature of these traits. The theoretical results make clear that host shifts and concomitant changes in host acceptance behaviour are important factors in adaptive speciation, but the underlying changes in the neural system are poorly studied and understood. Only recently work addressing this type of question is starting to appear (cf. Olsson et al. 2005a, b) .

Although host acceptance in phytophagous insects is usually viewed as solely governed by the adult (“mother knows best”), it should be noted that in many species larval host acceptance is equally important. Lepidopterous larvae in particular often display striking food preferences, which are based on a small set of chemoreceptors (Schoonhoven and van Loon 2002), thus making caterpillars promising models to study the neural basis of host plant recognition, host shifts, and the evolution of insect–plant relations. However, such an approach requires detailed knowledge about the responses of the involved receptors, information that is almost completely lacking.

External chemoreceptors in caterpillars are located on the antennae, on the maxillae, and in the epipharynx. The gustatory sensilla styloconica on the galea are important in host acceptance behaviour, and a large body of data is available about their sensitivity (see Schoonhoven and van Loon 2002; Schoonhoven et al. 2005 for review). The other sets of receptors, the epipharyngeal organs on the inside of the labrum, and the sensilla on the antennae and the maxillary palps, are more difficult to study and as a consequence less well known. The maxillary palp receptors in particular have received very little attention. Except for the pioneering studies of Schoonhoven and Dethier (1966), recordings from the maxillary nerve of the silkworm by Hirao et al. (1976, and a recent paper on gustatory receptors in the palp of Choristoneura fumiferana (Albert 2003), no electrophysiological data on the maxillary palps of Lepidoptera appears to be available. However, behavioural work indicates that the influence of the palps on feeding can be as significant as that of the taste sensilla on the galea (Hanson and Dethier 1973; Glendinning et al. 1998; de Boer 1993, 2006).

The small ermine moth genus Yponomeuta Latreille (Lepidoptera: Yponomeutidae) consists of some 75 species (Gershenson and Ulenberg 1998). Most are specialised on Celastraceae, and phylogenetic analysis (Menken 1996; Turner H, Lieshout N, van Ginkel W, Menken SBJ, submitted; Ulenberg, personal communication) strongly suggest that this association is the ancestral state in the genus. Moreover, mapping host use upon the phylogeny indicates an intriguing evolutionary scenario: the genus most likely originated in the Far East and dispersed to the western Palearctic concomitant with a (probably single) shift to Rosaceae and further to Salicaceae (Menken 1996). Interestingly, one lineage within this clade, Y. cagnagellus (Hübner), shifted back to the ancestral host, so this species is secondarily associated with Celastraceae. Yponomeuta plumbellus on the other hand is present in Western Europe, but this species is more closely related to the clades in the Far East, and retained the ancestral relation with Celastraceae.

The feeding of the Western European Yponomeuta species, and in particular Y. cagnagellus, is relatively well studied. Both the behaviour (Gerrits-Heybroek et al. 1978; van Drongelen 1980; Kooi and van de Water 1988; Peterson et al. 1990) and the underlying sensory physiology (van Drongelen 1979; Roessingh et al. 1999) have been addressed. The results demonstrate a clade of predominantly monophagous species with well-adapted sensory systems displaying high sensitivity for sugar alcohols characteristic for the host–plant families, as well as reduced sensitivities for host-specific deterrents. However, this whole body of work is focused on gustatory responses of styloconic sensilla on the galea that may provide only part of the relevant sensory inputs (de Boer 2006).

Here we present a morphological, behavioural, and electrophysiological survey of chemoreceptors on the maxillary palps in the larvae of Yponomeuta cagnagellus. In addition, we performed a comparative study into the effects of benzaldehyde on feeding behaviour of three related Yponomeuta species. Benzaldehyde strongly stimulated sensory cells in the maxillary palp of Y. cagnagellus, but its role in feeding behaviour is unknown. This compound is of particular interest because it is prominent in Rosaceae such as Prunus spinosa (Humpf and Schreier 1992), the host of Y. padellus (L), but absent from Euonymus europaeus, the host of Y. cagnagellus and Y. plumbellus (Denis and Schiffermüller) (Fig. 1). In this comparison we take advantage of the fact that the phylogenetic tree indicates that Y. plumbellus, although present in Western Europe, belongs to a clade that still displays the original ancestral association with Celastraceae. Yponomeuta padellus is in the middle of the clade that shifted to the benzaldehyde containing Rosaceae. Yponomeuta cagnagellus on the other hand is now feeding on the Celastraceous E. europaeus, that does not contain benzaldehyde, but this species most likely originated from an ancestor that was feeding on Rosaceae. An analysis of the phylogenetic pattern of the response to benzaldehyde in these three species will therefore contribute to our understanding of the flexibility of sensory systems and the role of sensory stimuli in host shifts and adaptive speciation.

Fig. 1 Phylogeny of Yponomeuta, based on the internal transcribed spacer region (ITS-1) and on 16S rDNA (16S) and cytochrome oxidase (COII) mitochondrial genes, using maximum parsimony (Turner H, Lieshout N, van Ginkel W, Menken SBJ, submitted). Host plant use (more ...)

Fig. 2 Schematic diagram of the electrophysiological set-up used to measure responses to both gustatory and olfactory stimuli. Open arrows indicate airflow, black arrows electrical signals. Note that in contrast to standard tip recording the stimulus pipette (more ...)

The preparation was placed in a continuous stream of clean, moistened air (80 ml/s, 25 cm/s). A second (control) air stream (2 ml/s) was injected in the main stream through an empty pasteur pipette. Odour stimuli were delivered by switching this second stream for 1 s through a pasteur pipette containing the stimulus. For stimulation with CO₂, a single injected stream of commercial available bottled CO₂ (Hoekloos, the Netherlands) was switched on. This injected stream could be adjusted to produce CO₂ concentrations in the main stream between 0.1 and 10%. Olfactory stimuli were diluted in paraffin oil (10⁻³, 10⁻², and 10⁻¹ v/v). Twenty-five microlitres of these solutions was applied to filter paper strips (6.0 × 0.5 cm) and was placed in pasteur pipettes. The following compounds (99% pure) were used: 1-hexanol and benzaldehyde (Fluka, Switzerland); (Z)-3-hexenyl acetate and (E)-2-hexenal (Z)-3-hexen-1-ol (Roth, West Germany); hexanal (98% pure) (Merck, USA); limonene, citral (97% pure), and geraniol (98% pure) (Aldrich, Germany). All stimuli were applied in random order with at least 30 s between them. Longer intervals were used after strong stimuli. Preparations were discarded after 60 min and were changed after a successful recording. Odour stimuli were stored at 4°C between recording sessions, and replaced every few days.

The AC amplified action potentials from the sensory cells were recorded on tape (Racal FM taperecorder). To facilitate computer-assisted analysis, the valve signal and a pre-pulse 1 s earlier were recorded on an additional channel.

Fig. 3 Feeding assay air flow diagram. Both the control group and treatment group consisted of 18 Petri dishes

A leaf disc with a diameter of 11 mm made from fresh host leaves was placed in each Petri dish on top of a piece of filter paper slightly wetted with distilled water. To reduce baseline palatability and detect feeding stimulation and/or inhibition, the leaf discs were extracted in alcohol and then washed in distilled water for 10 min. The duration of extraction needed to induce a weak feeding response depending upon both the host plants and leaf quality. For Y. cagnagellus, E. europaeus leaf discs were soaked in 96% EtOH for 24 h. Prunus spinosa leaf discs for Y. padellus were soaked for 2 h in 70% EtOH. Yponomeuta plumbellus appears later in the season than the other two species and feeds on slightly older E. europaeus foliage. For these older leaves extraction of leaf discs for 1 h in EtOH was sufficient to reduce palatability. The feeding experiments lasted 4–5 h. Usually 1 or 2 h after the start of the experiment, caterpillars started eating. The experiment was terminated after at least two third of the insects had been eating for more than 3 h.

To quantify the amount of feeding, the remainders of the leaf discs were scanned with an EPSON perfection 1250 scanner. The surface area was measured on a Macintosh computer using the open source program NIH Image version 1.62 (developed at the U.S. National Institutes of Health and available on the Internet at http://www.rsb.info.nih.gov/nih-image/). The differences in consumption between treatment and control groups were evaluated using a Mann–Whitney test in SPSS (SPSS Inc. 2001). For graphical display a feeding index (C − T)/(C + T) × 100% (ranging from −100 to 100%), was calculated, in which C and T denote, respectively, the mean of the remaining leaf surface in the control group and in the treatment group.

Fig. 4 Cryo-SEM micrographs of Yponomeuta cagnagellus.a Ventral view of the head. b Maxillary palp showing the digitiform sensillum (leftmost arrow) and a large campaniform sensillum (rightmost arrow). c Overview of the tip of the most distal segment. Note the (more ...)

On the apex seven sensilla basiconica and one sensillum styloconicum can be distinguished. On the side of the segment, a digitiform sensillum and at least one campaniform sensillum are visible (4b). With the exception of the styloconic sensillum, which has a prominent base, the sensilla on the apex are simple blunt cones. The lateral and medial sensilla, for which a gustatory function has been suggested, seem to be slightly more tapered (Fig. 4e, f) but the difference is only slight. Structural details like grooved surfaces or apical pores were not observed.

Fig. 5 I Responses of a (E)-2 hexenal cell in the maxillary palp of Yponomeuta cagnagellus to plant odour compounds. A Hexanal 10−1, B Hexanal 10−2. C (Z)-3-hexen-1-ol 10−1. D (Z)-3-hexen-1-ol 10−2. The stimulus period (1 s) (more ...)

Fig. 6 Response spectra for 19 cells sensitive to plant odour compounds in the maxillary palps of Yponomeuta cagnagellus at a concentration of 10−2. Response intensity is indicated by the diameter of the circles. Empty spots are missing observations

Dose response relations for green odours in the two main cell groups are given in Fig. 7. In Fig. 7a, the responses of the (E)-2-hexenal sensitive cells to the other green odours are plotted. The same is done in Fig. 7b for the (Z)-3-hexen-1-ol cells. In the alcohol-sensitive group, several cells responded to both (Z)-3-hexen-1-ol and 1-hexanol (cells 6–12) but cells that reacted mainly to one of the compounds were also found (cells 13–19).

Fig. 7 Dose–response relations of olfactory receptor cells in the cells of Yponomeuta cagnagellus. a (E)-2-hexenal sensitive cells (data from cell 1 to 5 in Fig. 6, plus three additional cells). b (Z)-3-hexen-1-ol sensitive cells (data from cell (more ...)

The (Z)-3-hexen-1-ol and 1-hexanol cells were often also highly sensitive to benzaldehyde (Fig. 7c). Evidence indicating that this sensitivity could be attributed to another cell (e.g. superpositions or irregular spike interval histograms) was not observed. The “aldehyde” cells in our sample did not show this sensitivity and only two cells responded weakly to benzaldehyde.

Although the sensitivity to benzaldehyde was often associated with that for the leaf alcohols, examples of both benzaldehyde insensitive cells (cell 14–15) and apparent benzaldehyde specialists (cells 18–19) were found.

Fourteen cells responded selectively to CO₂ (Fig. 8). The dose–response curve (Fig. 9) is almost linear from 0.2 to 2.5% CO₂. No significant response to other stimuli was found for these cells. All tested C6 compounds sometimes inhibited spike activity in CO₂ cells, but only at high concentrations (10⁻¹ v/v). The terpenoids limonene, citral, and geraniol never inhibited the CO₂ cells (n = 14).

Fig. 8 Response of a CO2 sensitive cell in the maxillary palp of Yponomeuta cagnagellus. a to e are responses to, respectively, 0.26, 1.0, 1.5, 2.5, and 10% CO2. The time marker indicates 1 s

Fig. 9 Dose–response relationship of 12 CO2 sensitive cells in the palp of Yponomeuta cagnagellus stimulated with CO2 or clean air. Vertical bars indicate 95% confidence interval

Three cells did not react to chemical stimuli but responded to changes in temperature induced by a hot or cold metal rod in the vicinity of the palp. A drop in temperature evoked a strong response. An increase in temperature inhibited the cells (Fig. 10), i.e. these cells functioned as cold receptors.

Fig. 10 Response of a cold receptor in the maxillary palp of Yponomeuta cagnagellus. a Stimulation with a cold metal rod at an upwind distance of about 10 mm. b Stimulation with a warm metal rod at an upwind distance of about 10 mm. Stimulus (more ...)

Fig. 11 Feeding index of three Yponomeuta species. FI = (C − T)/(C + T) (ranging from −100 to 100%). C and T denote, respectively, the mean of the remaining leaf surface in the control group and (more ...)

The feeding response of Y. padellus was tested at four concentrations, i.e. 0.5 × 10⁻⁴, 0.5 × 10⁻², 1 × 10⁻², and 0.5 × 10⁻¹ (Fig. 11). Benzaldehyde significantly stimulated the feeding of caterpillars at all concentrations except at the lowest concentration (percentage leaf disk left on benzaldehyde and control at 0.5 × 10⁻² was, respectively, 76.5 ± 3.7 and 84.5 ± 3.4%, Mann–Whitney U 375, P = 0.003, At 1 × 10⁻² 76.6 ± 3.8 and 90.4 ± 1.7%, Mann–Whitney U = 321, P = 0.001 and at 0.5 × 10⁻¹ 70 ± 3.8% and 82.8 ± 3.3% Mann–Whitney U = 408, P < 0.007). In Y. plumbellus, no effect of benzaldehyde on food intake was found, and the insects appeared to be fully indifferent to the compound (Fig. 11).

In the wall of the most distal segment, a digitiform sensillum is present which is normally innervated by one neuron (Schoonhoven and Dethier 1966; Albert 1980; Devitt and Smith 1982; Keil 1996). In E. messoria and Heliotis amigera this sensillum appears to possess a laminated outer dendritic segment. Such laminated dendrites have been associated with cold receptors (see Altner and Loftus 1985), as well as with CO₂ receptors (Chu-wang et al. 1975; McIver and Siemicki 1984; Lee et al. 1985; Bogner et al. 1986). In Y. cagnagellus both cold receptors and CO₂ receptors were found. The digitiform sensillum of Y. cagnagellus might therefore house a CO₂ receptor as suggested by Keil (1996) but without further electrophysiological evidence cold reception cannot be excluded.

From Fig. 6 is clear that although at least two groups of cells with comparable spectral responses can be distinguished (cells 1–5, cells 6–19), however, there are no sharp boundaries between these classes (cf. cell 5), and considerable variation does exist. Since care was taken to record only from cells that were reliable and reproducibly responding, it is likely that this variation does reflect natural variability in the receptor cell population. The separation in aldehyde- and alcohol sensitive cells should therefore at present merely be considered as a convenient way to summarise the data. Until more recordings are made, it cannot be excluded that a continuum from more specialised to, more generalised cells exists.

In the present study, no cells with a primary response to (Z)-3-hexenyl acetate were found in the maxillary palps of Y. cagnagellus, but receptors sensitive to this compound have been found on the antennae of this species (Roessingh, unpublished results) as well as on the antennae of Pieris brassicae larvae (Visser and de Jong 1988). In this respect it is interesting to note that in Manduca sexta fibers from the maxillary nerve travel into a core of neuropile in the suboesophageal ganglion that also receives antennal axons (Kent and Hildebrand 1987).

The maxillary palps of caterpillars contain 15–30 chemoreceptors (Schoonhoven and Dethier 1966; Albert 1980; Devitt and Smith 1982). If the 85 cells in the present study represent a random sample from the population in the palps, then the proportion of neurons sensitive to each of the tested stimuli could be estimated. However, it must be stressed that the position of the electrode was not reproducable between recordings. As a consequence, the recordings by no means represent a random sample. In spite of this limitation it can be concluded that receptors for the tested terpenoids and for (Z)-3-hexen-1-yl acetate are relatively scarce, since no cells primarily sensitive to these stimuli were encountered.

From a phylogenetic perspective a different hypothesis can thus be formulated. Yponomeuta cagnagellus is located within the European clade that feeds mostly on Rosaceae, which do contain benzaldehyde. Considering that the most recent common ancestor of this clade was presumably a Rosaceae feeder, the sensitivity of Y. cagnagellus to benzaldehyde could be viewed as a maintained sensitivity to a feeding stimulant from the ancestral host.

Under this hypothesis both Y. cagnagellus and Y. padellus are expected to be stimulated by benzaldehyde. Yponomeuta plumbellus, a Celastraceae feeder outside the European clade, is expected to be unresponsive because it has never been associated with Rosaceae nor have its ancestors. The results from the feeding experiment obtained from the three species are consistent with the latter scenario. Benzaldehyde stimulated feeding in Y. cagnagellus and Y. padellus and increased food intake by 6–12%. Yponomeuta plumbellus did not respond to benzaldehyde at all. The sensory response to benzaldehyde in Y. cagnagellus is therefore probably an evolutionary relict of the association with Rosaceae of the ancestral lineage of this species. In a previous study (van Drongelen 1979), dulcitol sensitivity was found in Y. evonymellus and Y. padellus, species feeding on Rosaceae that do contain very little or nodulcitol. In this case it was similarly concluded that the dulcitol sensitivity might be an evolutionary relict from the ancestral host association with Celastraceae.

Although feeding in both Y. cagnagellus and Y. padellus was stimulated by benzaldehyde, Y. padellus (associated with a benzaldehyde-containing hosts) responded over a much larger concentration range than Y. cagnagellus (a species that is using a benzaldehyde-free host). It should be noted that care must be taken when interpreting these results. It is not clear what the natural concentrations of host plant odours that are experienced by the larvae, since a steep gradient in the boundary layer exists, and concentrations near the leaf surface can be much higher than those obtained from headspace samples (Schoonhoven et al. 2005). However, since the sensory responses show good dose–response relations, and the behaviour yields a clear optimum, we feel that the used concentrations are covering the natural range. It will be interesting to investigate the actual genetics underlying the observed sensitivity differences between the species.

We wish to thank Felix Thiel (TFDL, Wageningen) for invaluable help with the Cryo SEM, Teris van Beek and J. Cozijnsen for assistance and advice concerning the headspace analysis, Maarten Posthumus for the GC/MS work, Louis Schoonhoven for critical reading of the manuscript, and Birgit Nübel for rearing Yponomeuta larvae. Special thanks to Erich Städler and the Forschungsanstalt Wädenswil for providing time and facilities to complete an early version of this manuscript.