Food searching behaviour of a Lepidoptera pest species is modulated by the foraging gene polymorphism

The level of damage inflicted on a crop by an agricultural pest species depends on the pest's ability to colonize the field area. Colonization ability, in turn, is related to pest density and to individual propensity to oviposit or feed on multiple plants within a field. Identifying molecular mechanisms involved in plant–arthropod relationships is a prerequisite for developing specific and environmentally friendly control measures (Santiago et al., 2013). It is also a major step toward understanding evolutionary mechanisms in plant–arthropod relationships. So far, most investigations in these fields have focused on sensory mechanisms involved in plant recognition (Briscoe et al., 2013; Glaser et al., 2013; Hallem et al., 2006) or on physiological mechanisms of susceptibility or resistance to plant compounds, plant varieties or insecticidal molecules (Hogenhout and Bos, 2011; Kloth et al., 2012). Genetic and molecular bases of foraging activity are particularly well studied in Drosophila (Sokolowski 2001) and in the honeybee (Guan et al., 2013; Page et al., 2012), but little is known in phytophagous pest insects.

Here we present the first study of the role of the foraging (for) gene in food searching behaviour in a cereal pest lepidopteran. The order Lepidoptera includes a number of major agricultural pest species. Their larvae feed on plants, and typically have a several-week development time, during which many species move from plant to plant to complete their development. Sesamia nonagrioides Lefèbvre (Lepidoptera, Noctuidae) is a stem borer found on a variety of monocots belonging to the Poaceae, Cyperaceae and Typhaceae families (Le Ru et al., 2006). It is also called the Mediterranean maize borer or pink stem borer, as it is the major insect pest of maize and sorghum in several Mediterranean countries (Cordero et al., 1998; Eizaguirre and Fatinou, 2012). It can also cause economic damage to cultivated cereals in sub-Saharan West Africa (Kfir et al., 2002).

In this species, caterpillars typically move from plant to plant. In humid and mild areas, often 100% of maize plants have been attacked by the time they are harvested (Cordero et al., 1998). In addition to direct plant damage, this larval behavioural activity can seriously alter maize eatability, because Fusarium fungi often develop in the larval tunnels (Santiago et al., 2013). These fungi produce fumonisin mycotoxins that can accumulate in the maize kernel, and cause several pathologies in humans and other mammals, including esophageal cancer under some circumstances (Bryla et al., 2013).

The gene for, which encodes a cGMP-dependent protein kinase (PKG I), has been described in the literature as having a conserved evolutionary relationship with food-related behaviours. This is because it is involved in modulating foraging strategies in a variety of invertebrate species (Ben-Shahar, 2005; Fitzpatrick and Sokolowski, 2004; Fitzpatrick et al., 2005; Kaun and Sokolowski, 2009; Lucas et al., 2010; Thamm and Scheiner, 2014). The phylogenetic analysis of this gene among metazoan taxa shows high conservation of its structure and suggests a strong evolutionarily conserved link between PKG I and food-related behaviours (Fitzpatrick et al., 2005). In the nematodes Caenorhabditis elegans (Fujiwara et al., 2002) and Pristionchus pacificus (Kroetz et al., 2012), and in the fruit fly Drosophila melanogaster (Sokolowski, 2001), induced or natural mutations of the gene, respectively, are associated with the expression of different behavioural and physiological phenotypes. In D. melanogaster, flies and larvae of

the ‘rover’ genotype move farther than do individuals with the ‘sitter’ genotype when they forage. When food is scarce, rover larvae have higher survivorship and faster development than sitters, which is explained by their higher glucose absorption efficiency (Kaun et al., 2007). In other species, and more particularly in eusocial hymenoptera such as honeybees, Apis mellifera (Ben-Shahar, 2005; Thamm and Scheiner, 2014), and the ants Pheidole pallidula (Lucas and Sokolowski, 2009) and Pogonomyrmex barbatus (Ingram et al., 2005), modulation of food searching behaviours results from variation in for activity throughout the development of behavioural tasks. In the wasp Vespula vulgaris (Tobback et al., 2008) and in two Bombus species [B. ignites (Kodaira et al., 2009) and B. terrestris (Tobback et al., 2011)], results support the idea that for expression levels are related to behavioral caste differentiation and thus to for activity. In the locust Schistocerca gregaria, higher expression of the for orthologue characterizes the gregarious phase that typically precedes group migration (Lucas et al., 2010), and transcript levels change when the availability of food is altered and in crowd- versus isolated-reared insects (Tobback et al., 2013). In the aphid, change in for expression has also been observed in response to crowding, usually followed by dispersion (Tarès et al., 2013).

Sesamia nonagrioides is a relevant candidate with which to study conservation of and polymorphism in the for gene in a lepidopteran, and its role in a behavioural trait directly associated with plant injuries. We conducted this study with an integrative experimental approach that included analyses of sequence conservation and polymorphism, allelic frequencies, and the link between foraging activity, observed in an adapted actimeter (Chardonnet et al., 2012), and pharmacological activation of PKG, allelic variation and gene expression. By linking phenotype to genotype, this study shows how the for gene can modulate a behavioural function linked to pest character in a lepidopteran species.

We documented sequence conservation and polymorphism in S. nonagrioides. Primers based on for sequences of other lepidopterans allowed us to sequence the full-length transcript in S. nonagrioides. The gene cDNA was 2235 bp long, encoding 745 amino acids. We did not find evidence of several transcripts. We compared the protein sequence with the homologous Drosophila sequence for which the protein domains are mapped (Fig. 1). The gene was thus identified as an orthologue of the Drosophila dg2 gene (i.e. foraging) coding for PKG I (Osborne et al., 1997). We labelled the gene Snfor.

The secondary structure of the protein was similar to that described in other species (e.g. Osborne et al., 2011). The regulatory domain is encoded by the first 185 amino acid residues, followed by cGMP1 (186–302 residues) and cGMP2 (303–418 residues) binding domains. The catalytic domain is located between the 434 and 693 residues.

A phylogenetic tree was constructed to check the position of the S. nonagrioides PKGI protein among other species' orthologues. It was based on 34 PKG I sequences (Fig. 2), which confirms the different clades corresponding to insect orders already evidenced by Lucas et al. (Lucas et al., 2010), and places the S. nonagrioides sequences in a new sub-clade gathering noctuid species within the well-supported Lepidoptera clade.

By comparing the Snfor protein sequence with for sequences from other lepidopteran species (Bombyx mori, AF465600; Lobesia botrana, DQ666642; Sesamia calamistis, obtained at our laboratory), we showed a high genetic identity to other orthologous genes in this order (≥94%).

We compared all Snfor sequences obtained from cDNA of 15 individuals from a maize field collection in Haute-Garonne (France). There were several synonymous single-nucleotide polymorphisms (SNPs). We found only one non-synonymous SNP in the complete Snfor gene sequence. This mutation was found at the 1699 base position: an A/G nucleotide change that induces a codon change. When adenine is at the first position, the codon codes for serine in the protein sequence; when guanosine is in the first position, it codes for glycine. Mapping with the Drosophila sequence indicated that this substitution was located in the catalytic domain of the protein (Fig. 1). By comparison with for sequences from other lepidopteran species, it appeared that these other sequences had a serine at the polymorphic Snfor position. The GenBank accession numbers of the two variants are KC310497 (SnFor^Gly) and KC310498 (SnFor^Ser).

The next step was to quantify allelic and genotypic frequencies in a natural population from maize fields (Haute-Garonne, France). We genotyped 143 field-collected insects using the amplification refractory mutation system (ARMS) technique (Newton et al., 1989). In the samples collected yearly from 2012 to 2014 (N=94, 49 and 93, respectively), approximately 40% were SnFor^Gly/Gly homozygotes (38 to 39%), 40–50% were SnFor^Gly/Ser heterozygotes (51, 43 and 40%, respectively) and less than 20% were SnFor^Ser/Ser homozygotes (10, 18 and 19%, respectively) There was no significant difference in the genotypic frequencies between the three years (Chi-square test, d.f.=2, P>0.05).

To test the role of the Snfor gene on foraging activity in S. nonagrioides, and establish a causal link between the gene and the behaviour, we used a pharmacological approach that allowed us to artificially manipulate the activity level of cGMP-dependent protein kinase. It is a common technique to use a pharmacological agent to modulate specific characters, especially in non-model species where transgenic studies are not yet possible. In the literature, 8-bromo-cyclic guanosine monophosphate (8-Br-cGMP; an analogue of endogenous cGMP, the second messenger of the protein) has been described as a specific activator of PKG of D. melanogaster (Dawson-Scully et al., 2010), A. mellifera (Ben-Shahar et al., 2002) and P. pallidula (Lucas and Sokolowski, 2009). Here, we gave larvae a non-invasive oral chronic treatment of the PKG activator.

At the end of the treatment, they were individually tested in the ‘actimeter’ (Fig. 3). The ‘actimeter’ system was designed by Chardonnet et al. (Chardonnet et al., 2012) from a device initially developed for the fruit fly, the DAM system (see Materials and methods). This original tubular system allowed us to automatically record the number of times a larva switched between two food sources (patches).

Foraging activity in the actimeter was significantly influenced by treatment over the first day of recording (ANOVA, F_3,124=5.138, P=0.002; Fig. 4A). The number of patch changes was higher in the group treated at 500 μmol l⁻¹ (multiple comparison test, 500 versus 1000 μmol l⁻¹, P=0.018, 500 versus 0 μmol l⁻¹, P=0.012; Fisher test, P>0.05). A smaller activation was seen in the 250 μmol l⁻¹ group. There was no effect at 1000 μmol l⁻¹, potentially because of toxic effects of the treatment under higher-concentration conditions (Ben-Shahar et al., 2002). We saw no effect of the treatment on the second day, suggesting a rapid decrease in efficiency of the drug once it is no longer being administered (ANOVA, F_3,124=1.011, P>0.05). The treatment had no effect on larval growth (each caterpillar was weighed each day during the treatment and at the end of the behavioural test), excluding an indirect size-related effect on behaviour. Together these data show that PKG activation increases the foraging activity observed in S. nonagrioides larvae.

Although PKGs are the principal target of cGMP (Hofmann et al., 2009), the great similarity between PKA (cAMP-dependent protein kinase) and PKG makes it possible to also activate PKA by cGMP, particularly when cGMP is present in high concentrations (Hofmann et al., 2009; Pilz and Castell, 2003). We therefore checked whether PKA activation influenced the foraging behaviour of S. nonagrioides. We substituted 8-Br-cGMP with 8-Br-cAMP, an activator of PKA (Ben-Shahar et al., 2003), using the same protocol of non-invasive oral chronic treatment (n=40 per concentration tested). At the end of the treatment, we tested larvae individually in the actimeter. The foraging activity was not significantly influenced by the treatment over the whole experiment (ANOVA, day 1: F_3,158=0.348, P>0.05; day 2: F_3,158=0.582, P>0.05; Fig. 4B). The treatment had no effect on larval size, excluding an indirect size-related effect on behaviour. These results show that PKA activation does not significantly modulate the foraging activity of S. nonagrioides larvae in the actimeter.

The presence of an Snfor polymorphism in the studied population allowed us to test the correlation between genotypes and foraging phenotypes. We thus tested the foraging activity of larvae and genotyped them afterward.

One-hundred and seventy-nine third-instar larvae were tested in the actimeter for 3 days (72 h). We identified 59 SnFor^Gly/Glyhomozygotes, 46 SnFor^Ser/Ser homozygotes and 74 SnFor^Gly/Ser heterozygotes in this laboratory-reared sample. We observed a significant effect of genotype on foraging activity (Kruskal–Wallis test, K=13.387, d.f.=2, P=0.001; Fig. 5), revealing that SnFor^Ser/Ser larvae left a food patch more frequently than SnFor^Gly/Gly individuals, while SnFor^Gly/Ser individuals were intermediate between the two homozygotes (post hoc Steel–Dwass–Critchlow–Fligner test: Ser versus Gly, P=0.001; Ser versus Het and Het versus Gly, P>0.05).

These allelic frequencies were not significantly different from those observed in the analysis of field-collected larvae (Chi-square test, d.f.=4, P>0.05).

Variation in the level of for expression is known to have a major effect on the behaviour of several insect species. In social Hymenoptera, changes in the expression rates accompany the shift between nursing and foraging behaviours (Ben-Shahar et al., 2003; Heylen et al., 2008; Ingram et al., 2005; Ingram et al., 2011; Tobback et al., 2011), and constitute one of the explicative elements of the division of labour within the colony. To generate a wider view of a potential link between the behavioural phenotypes and the Snfor gene, we compared the gene expression level in the whole body and in the brain of the SnFor^Ser/Ser and SnFor^Gly/Gly homozygote strains founded from the polymorphic study population.

We first compared Snfor whole-body expression between SnFor^Ser/Ser and SnFor^Gly/Gly larvae. Individuals were taken from the two reared strains without activity testing, to avoid a situation where variation in the gene expression would result from behavioural testing (we chose the fourth-instar because larvae usually reached that stage during actimeter experiments). Snfor expression was significantly higher in the whole body of SnFor^Ser/Ser larvae than in SnFor^Gly/Gly larvae (n=30 per strain, P<0.001; Fig. 6A) (Pfaffl, 2001).

We then compared the Snfor brain expression between the SnFor^Ser/Ser and SnFor^Gly/Gly homozygote strains of S. nonagrioides, as above. Snfor expression was also higher in the brain of SnFor^Ser/Ser larvae than in SnFor^Gly/Gly larvae (eight pools of eight brains per population, P<0.01; Fig. 6B).

The results of these molecular, pharmacological and behavioural analyses show that the for gene is directly involved in modulating a food searching trait in S. nonagrioides, a non-model species and an important pest of cultivated cereals in Europe.

This work showed clear sequence and function conservation of the for gene in S. nonagrioides. We concluded that Snfor was an orthologue of the homologous Drosophila dg2 gene (i.e. for) coding for the PKG I (Osborne et al., 1997), which corresponds to the cGK I of mammals, encoded by the prkg-1 gene (Hofmann et al., 2009). We based this conclusion on the strong conservation of sequence of the whole coding region and of the domain structure of the protein. The Snfor gene also has a high genetic identity to the other orthologous genes sequenced in a few Lepidoptera genera, such as Bombyx, Lobesia and Sesamia.

Next, by modulating foraging behaviour induced by artificially manipulating PKG activity, we identified a causal link between the PKG enzyme and the behaviour. PKG I, the for-encoded kinase, was very likely activated because we also showed consistent positive correlation between Snfor expression and foraging activity, assuming that gene expression correlates with the enzyme activity. These findings showed that the foraging function of the for gene is also conserved in a lepidopteran species.

In S. nonagrioides, we found that higher expression of Snfor in the brain was correlated with higher frequency of patch changes in the actimeter. There are indications in the literature of some downstream effectors, signalling pathways and sensory effects that begin to explain how PKG I activity may influence foraging traits. In the nematode C. elegans, EGL-4, the orthologue of for, has been reported to modulate gene expression of two chemosensory receptors (van der Linden et al., 2008). EGL-4 nuclear localization and activity in olfactory neurons was necessary to promote plasticity in odour responsiveness (He and O'Halloran, 2013; Juang et al., 2013). In mammals, PKG I is present in numerous tissues, and it phosphorylates more than 10 substrates known to modulate neuronal signalling pathways, including ion channels and G-protein (Hofmann et al., 2009). One of these substrates is the transporter of the neurohormone serotonin (5-hydroxytryptamine) (Zhang et al., 2007), which is involved in the regulation of feeding and digestion in many animals from worms to mammals (Gillette, 2006), including insects such as ants (Falibene et al., 2012), the honeybee (French et al., 2014) and D. melanogaster (Neckameyer and Bhatt, 2012). It is thus possible that variation in the for gene in S. nonagrioides may affect larval foraging behaviour via a serotonergic pathway.

One highlight of the present paper is the discovery of an SNP that correlates with differential Snfor expression and with differential behavioural activity. The SNP may itself have an influence on PKG activity because it is located in the catalytic site of the kinase, and may therefore affect its affinity for particular substrates. Functional consequences of single nucleotide changes have already been reported (Lenormand et al., 1999; Rottschaefer et al., 2011). It might be possible to predict the effect of the amino acid change after the 3D structure of this part of the protein is described (Osborne et al., 2011). However, a linkage disequilibrium between the SNP and possible mutations in the regulatory sequences cannot be discarded.

Our findings in S. nonagrioides present analogies with D. melanogaster. In this fly species, the for gene has two natural alleles that are associated with two specific foraging strategies (phenotypes) (Sokolowski 2001): rovers and sitters. Increased gene expression or PKG activity is correlated with increased foraging behaviour. Individuals of the genotype SnFor^Ser/Ser, which have a greater propensity to leave a food patch and a higher level of gene expression, could then be seen as ‘rover-like’, while SnFor^Gly/Gly, which are more sedentary and have a lower level of gene expression, would be ‘sitter-like’. However, in Drosophila there are several non-synonymous mutations between the two variants and these are present on several transcripts of Dmelfor (GenBank accession numbers FJ200471–FJ200478), each coding for a different PKG I isoform that may have specific substrate. This may explain why so many functions (sensory traits, locomotion, learning traits, glucose homeostasis, etc.) differ between rovers and sitters. Our results did not reveal more than one transcript, but more specific experimental approaches (northern blot) are still needed to confirm this point.

Our S. nonagrioides study population was composed of the three genotypes SnFor^Gly/Gly, SnFor^Ser/Ser and SnFor^Gly/Ser, with a higher frequency of the SnFor^Gly allele. In this population collected in a maize field in Haute-Garonne, the non-significant variation of the genotype ratio between larvae collected from 2012 to 2014 in the same area suggests that there is a mechanism that maintains the polymorphism. There are several possible mechanisms of polymorphism maintenance in natural populations. For example, polymorphism can be understood as a mechanism that allows species to maintain their fitness in a changing environment, as was described for the for gene polymorphism in the Drosophila model (Sokolowski et al., 1997). When food is scarce, rover flies and larvae have an adaptive advantage because they are better at finding new food patches and they assimilate nutrients more efficiently (Kaun et al., 2007). However, the high energetic cost of the rover strategy means that sitters – which use less energy for foraging – have higher fitness when food is abundant, thus maintaining a natural polymorphism (Kent et al., 2009; Sokolowski et al., 1997). This explains how the fitness of each genotype can change depending on environmental factors [gene–environment interaction (Burns et al., 2012)].

PKG I has pleiotropic effects and is also known to influence food odour learning (Mery et al., 2007) and olfactory and taste sensitivity – key components in evaluating food quality (Scheiner et al., 2004; Juang et al., 2013). As S. nonagrioides is able to feed and develop on several wild and cultivated host plants, our results open up new research avenues to understand some of the molecular mechanisms that contribute to the adaptation of this species to plant resources or other environmental components. The polymorphism shown here may also have an effect on the pest character of the insect, because more active variants are expected to attack more plants and favour the transmission of plant pathogens. In this context, analysis of for genotype frequencies in natural populations of S. nonagrioides, in the whole area of its distribution and on its different host plant species, opens new perspectives for predicting its damage levels in cereal crops, and for developing specific and environmentally friendly control measures.

To conclude, here we have shown conservation of Snfor for orthologue function for the first time in a lepidopteran, and discovered a polymorphism of the Snfor orthologue as a genetic mechanism that may play a role in the expression of specific foraging strategies. We illustrated how variation in this single gene could predict phenotype and discussed potential issues in the context of insect adaptation to plants and of pest management.

The studied population was initiated from larvae collected from maize stems in fields located in a 10 km² area in Haute-Garonne, France. GPS coordinates of the central point of the collection area are 43°22.171′N, 1°11.413′E. Approximately 300 larvae were collected yearly and added to the laboratory population to maintain genetic diversity and ensure that insects were as similar as possible to those from the field. Insects were reared on an artificial diet for one to four generations before being included in the experiments, at the Laboratory Evolution, Génomes et Spéciation (LEGS, France), in incubators (26°C, 70% relative humidity, and a 16 h:8 h light:dark cycle). The diet was composed of dried maize leaf powder (13.2% dry matter), soybean powder (45.9% dry matter), saccharose (18.2% dry matter), dead pulverulent baker's yeast (11.9% dry matter), and a mix of vitamins, antibiotics, antifungal agents and preservatives (4.3% dry matter) (adapted from Onyango and Ochieng'-Odero, 1994).

In order to sequence the orthologue Snfor, total RNA of S. nonagrioides was extracted from whole larvae with TRI Reagent (Sigma-Aldrich, St Louis, MO, USA). Reverse transcription was performed using the iScript cDNA Synthesis Kit (Bio-Rad, Hercules, CA, USA). Primers were designed based on for sequences of other Lepidoptera [B. mori, AF465600; L. botrana, DQ666642; Mythimna separata, GQ844298; Spodoptera littoralis, Poivet et al. (Poivet et al., 2013)]. These primers are D14B (ACTCTCAGTGCTCGGTAGCC), D2 (GGCAATCYTGTACAACTSCAAGCG), R1 (GGCCCAWAGYCKGCAGTCMG), R3 (GCGCTAGRCCTTCCCAGTTRAARCC) and R13 (AWSACWCCGAWMAMCAMT). D14B was designed in the 5′ UTR region and R13 in the 3′ UTR region. These primers allowed us to sequence the full-length transcript in S. nonagrioides (Fig. 1).

In order to construct the phylogeny, we used the same PKG I sequences as Lucas et al. (Lucas et al., 2010, as well as sequences of several other species: Sesamia calamistis (KM017967), Spodoptera exigua (FJ606838), Spodoptera frugiperda (KM017966), Mus musculus (XM_006526768) and C. elegans (NM_067740). In total, 34 sequences were used for the tree. As the N-terminal domain is very divergent and cannot be aligned across all species, only the last 323 amino acids of the C-terminal domain of the protein were kept for the phylogeny. We used the website www.phylogeny.fr (Dereeper et al., 2008, Dereeper et al., 2010) to construct our phylogeny. The sequences were first aligned with MUSCLE 3.7 (Edgar, 2004) and then a maximum likelihood tree was constructed with PhyML 3.0 (Guindon et al., 2010). Five hundred bootstrap replications were run to check the consistency of the nodes.

Once the complete cDNA sequence of the Snfor gene had been obtained, new primers were designed in order to sequence the whole gene in 15 individuals, directly from field collections. cDNA from each individual was amplified using three different primer combinations: D14B (see above)/R11 (GAAGAAATCGCCTTTTGTGAG); D10 (CAAGCCGAATACACCGACT)/R10 (TCCGGTAAGCAACTCAAACA) and D12 (GGGATCTCAAACCTGAGAACTT)/R15 (AATTACAAATGGCGTTAAATCT). Verification by agarose gel electrophoresis showed that only one product was amplified by each primer pair. The sequences were clean and did not show the presence of multiple transcripts.

To detect heterozygotes and to target the observed SNP, large-scale genotyping of field-collected individuals was performed on DNA using the ARMS technique (Kim et al., 2011). Genomic DNA was extracted from larvae or from adult thoraxes using a DNeasy Blood and Tissue Kit (Qiagen, Venlo, The Netherlands). Two PCRs were performed on each individual. Each PCR contained two primers that bind to either side of the SNP and amplify a fragment that is always present (Arms-GenF: CGACGATGCCACAACTAGA and R12: AGTCTGCACTAATATCATGACCTC). Each PCR also contained a third allele-specific primer that allowed the amplification of a smaller fragment, provided the individual concerned had a copy of the corresponding allele. These primers are Arms-Gly-RT (TAGTTTCACATAGCCTTTTCC) and Arms-Ser-RC (TAGTTTCACATAGCCTTTCCT). According to the presence or absence of the smaller fragment in the two different PCRs, we could deduce whether the individuals were homozygotes for the SnFor^Gly allele, homozygotes for the SnFor^Ser allele, or heterozygotes. In 2 years of collection (2012 and 2013), 143 individuals were genotyped.

The larvae were treated with the PKG activator 8-Br-cGMP (Fluka Chemie AG, Switzerland), by ingestion. This molecule crosses cellular membranes and is relatively resistant to deterioration by the enzymes responsible for the deterioration of the endogenous cGMP (phosphodiesterases) (Ben-Shahar et al., 2003). Second-instar larvae were collected from several egg masses to reduce the probability of siblings, and individuals of homogeneous size were chosen preferentially (caterpillars were individually weighed before and after the experiment). They were isolated in 2 ml Eppendorf tubes. Every tube contained rearing diet supplemented with 8-Br-cGMP at a given concentration, and was changed every day. A range of concentrations was chosen from the literature [honeybees (Ben-Shahar et al., 2002); ants (Lucas and Sokolowski, 2009)] to estimate a possible dose effect: 0 μmol l⁻¹ (control, n=40), 250 μmol l⁻¹ (n=24), 500 μmol l⁻¹ (n=40) and 1000 μmol l⁻¹ (n=24). The treatment was applied for 3 days, and caterpillars were weighed every 24±1 h in order to control for a possible effect on growth. At the end of the treatment, the locomotor activity linked to foraging behaviour of the insects (then third-instar) was individually tested in the ‘actimeter’ (description below) for 48 h.

The larvae were treated with the PKA activator 8-bromo-cyclic adenosine monophosphate (Sigma-Aldrich) by ingestion. We applied the same procedure of non-invasive oral chronic treatment as described above. The following concentrations were tested: 0 (control), 250, 500 and 1000 μmol l⁻¹; n=40 per group (Ben-Shahar et al., 2002). At the end of the treatment, the larvae were individually tested in the ‘actimeter’ for 48 h.

This system was designed by Chardonnet et al. (Chardonnet et al., 2012) from a device initially developed for the fruit fly, the DAM system (Drosophila Activity Monitoring System, DAM2®; Trikinetics, Waltham, MA, USA) (Freeman et al., 2010; Hirsch et al., 2009; Huang et al., 2009). The actimeter apparatus is made of tubular units (200×7 mm i.d.), each provided with three infrared photoelectric detectors that automatically count and record larval movement between two food patches (Fig. 3). The tubular configuration reproduces a stem-like confined structure, and the artificial diet inserted at both tube ends creates two food patches. The food patches were made of rearing diet; one patch is enough to feed a caterpillar during the whole experiment (i.e. the food patch was not entirely consumed when an immobile larva stayed in a unique patch for the 3 days of the experiment). To compare locomotor activity between the third-instar larvae from the different treatment groups, a ‘leaving the patch’ event was recorded each time a larvae crossed at least two of the three consecutive infrared beams. The device allows us to test up to 32 larvae simultaneously. The order of the larvae from the different groups in the actimeter was systematically alternated to avoid a position effect. The tests were performed in the same environmental conditions as for rearing (26°C, 70% relative humidity, and a 14 h:10 h light:dark cycle). The data were collected using DAMSystem collection® software (TriKinetics, Waltham, MA, USA).

In order to test for a correlation between Snfor genotypes and foraging phenotypes, we tested the foraging activity of 179 third-instar larvae from the studied polymorphic population of S. nonagrioides. The third-instar was chosen as it corresponds to the instar when the insects begin to leave the hatching plant and forage for new stems in their natural environment (Jepson, 1954). It corresponds to the middle of larval development, which includes six to seven larval instars until feeding cessation and pupation. The behavioural analyses were conducted for 72 h according to the ‘actimeter’ protocol described above. At the end of the behavioural test, the larvae were frozen in liquid nitrogen and then sequenced using the ARMS technique (they were then at the fourth-instar).

Two strains homozygous for Snfor, SnFor^Gly/Gly and SnFor^Ser/Ser, were each initiated from 15 homozygote founder couples taken from the laboratory-reared polymorphic population and genotyped after reproduction. For the whole-body analysis, 30 fourth-instar larvae per strain were tested. For the brain analysis, 64 fourth-instar larvae per strain were tested. In this condition, the brains (deutocerebrum) were pooled in groups of eight (i.e. eight replicates for each population). The live larvae were retrieved from the artificial diet at the fourth-instar stadium, a stage usually reached during behavioural testing protocol, and immediately frozen in liquid nitrogen and stored at −80°C until RNA extraction.

Total RNA was extracted from whole larvae using TRI Reagent (Sigma-Aldrich). The brains were dissected in the RA1 buffer and total RNA was extracted using the NucleoSpin RNA XS kit (Macherey Nagel). In both conditions, a DNase treatment (DNase I recombinant, Roche) was also performed. The integrity of each RNA sample was verified with the Experion Bio-Analyzer (Bio-Rad). All the samples showed a clear peak, without any sign of RNA degradation. One microgram of RNA extracted from the whole body was used to synthesize 20 μl cDNA for each sample using the iScript cDNA Synthesis Kit (Bio-Rad). In the brain extract analysis, 400 ng were used to synthesize cDNA. A 203 bp fragment of the Snfor gene was amplified with the primers QFORF (CATCGACCGGGAGACCTT) and QFORR (TCCTGCTTGAGTCGCTCTG). Three reference genes were used to normalize the samples: rpL13 (170 bp; QRPL13F3-TACCGCAGAAAACAAAACAGGAT/QRPL13R3-GCGGGGTTAAGACCAGATGC), rpS3 (121 bp; QRPS3F3-GCTGAATCGCTCAGATACAAACT/QRPS3R3-GCTTGCCAGAGACTACCACCT) and rpL17 (171 bp; QRPL17F3-TTTGGCACCACACAAGGTC/QRPL17R3-GGCACGGTATGTGCGTCTG). The PCR efficiency was calculated for each gene using a calibration curve with four dilution points using a pool of cDNA samples. In the whole-body extract analysis, the efficiencies were 92% for Snfor, 96% for rpL13, 99% for rpS3 and 95% for rpL17. In the brain extract analysis, the efficiencies were 107% for Snfor, 99% for rpL13, 96% for rpS3 and 91% for rpL17.

PCR was carried out in 10 μl, with 1X FastStart Universal SYBR Green Master (Roche), 0.3 μmol l⁻¹ of each primer and 2 μl of 10-fold-diluted cDNA for the whole-body extracts. The same mix was used for the PCR on brain extracts, but using a fivefold-diluted cDNA. All samples were amplified in duplicate, in order to check the consistency; a negative control was also included for each gene. The gene-specific primers were designed to span an intron, in order to avoid genomic DNA contamination in the analysis. For Snfor and rpL17, genomic DNA could not be amplified with the primers and amplification protocol we used; for rpS3 and rpL13, the fragment amplified with genomic DNA was approximately 900 bp and so any contamination would be revealed by the melting curve analysis. Therefore, non-reverse transcript controls were not included in our experiment.

The PCRs were run on a CFX96 (Bio-Rad) – the amplification protocol was 95°C for 10 min and then 45 cycles of 95°C for 10 s and 60°C for 30 s – followed by a melting curve. All melting curves showed a single peak, suggesting specificity and homogeneity of the amplifications. The quantification cycle (C_q) values were determined based on the single threshold mode. All negative controls had a C_q of 0. The mean C_q standard deviation observed between duplicates was 0.055.

The stability of the reference genes was checked with Bestkeeper software (Pfaffl et al., 2004). The differences between the Snfor transcript concentrations were determined using the ΔΔC_q method [ABI User Bulletin #2 (11-15)]. The statistical significance of the difference in Snfor expression observed between the two populations was tested using REST 2009 software (Pfaffl et al., 2002).

We thank Frederic Mery for lending us the actimeters for trials; Férial Kaoula and Sylvie Nortier for rearing the insects; Pascaline Dumas for sending samples of Spodoptera frugiperda to sequence the foraging gene; Isabelle Clavereau and David Ogereau for their advice on qPCR; and Jean-Christophe Sandoz and Antoine Couto for their help in the dissection method of the Sesamia nonagrioides larval brain. English language editing was performed by Sharilynn Wardrop.