Manduca sexta caterpillars parasitized by the wasp Cotesia congregata stop chewing despite an intact motor system

Parasitic manipulators are parasites that alter the behaviour of their hosts to enhance their own fitness (Moore, 2002). Examples of parasitic manipulators range from viruses to parasitoid insects (Moore, 2002). The specific mechanisms through which any of them manipulate host behaviour remains poorly understood, especially at the neurobiological level (Hughes and Libersat, 2018). Parasites can produce long-term changes in host behaviour, often using what appear to be novel methods of control (Adamo, 2013). Studying these novel methods could lead to important discoveries about the neural regulation of behaviour.

How parasites alter host neurobiology has been studied in only a few systems. The system that is perhaps the most thoroughly investigated is that of the jewel wasp Ampulex compressa and its host, the cockroach Periplaneta americana. In one of the few examples of a parasitic manipulator targeting a specific brain area, the female wasp uses her stinger to deliver venom into the central complex of the brain of her host (Gal et al., 2014). The wasp injects a cocktail of substances, altering activity in a number of neural circuits (Hughes and Libersat, 2018). After being stung, the cockroach host falls into a long-term quiescent (i.e. a sleep-like) state both behaviourally and neurophysiologically (Emanuel and Libersat, 2017). The current hypothesis is that the venom injected by the female wasp activates an already existing neural circuit in the cockroach that produces quiescence (Hughes and Libersat, 2018), allowing the wasp's offspring to develop undisturbed. It also suppresses activity in the central complex, reducing the ability of sensory stimuli to induce a motor response (Rana et al., 2022).

Most parasites, however, do not have direct access to the brain in the way that the jewel wasp does (Hughes and Libersat, 2018). For example, the parasitic wasp Cotesia congregata (Say) (Braconidae), does not physically contact the brain of its host, but can still induce profound behavioural changes. Cotesia congregata is able to suppress the feeding behaviour of its host, the caterpillar stage of the hawkmoth, Manduca sexta L. (Sphingidae). This suppression occurs despite the robustness of feeding behaviour in M. sexta caterpillars. Normally, M. sexta caterpillars in their later instars feed in frequent bouts throughout a 24 h period in both the laboratory (Reynolds et al., 1986) and the field (Bernays and Woods, 2000). Indeed, the reliability of feeding in this caterpillar has facilitated a number of studies on its feeding behaviour (e.g. Bowdan, 1988; Timmins and Reynolds, 1992) and its neural control (e.g. Griss, 1990; Griss et al., 1991). However, at a specific stage in the development of the parasitoid, the caterpillars' normal feeding behaviour abruptly stops. The cessation of host feeding is critical for wasp survival; it prevents the host from eating the wasps (Adamo et al., 1997; Adamo, 1998).

Cotesia congregata injects eggs, venom and polydnavirus into the haemocoel of its caterpillar host (Beckage and Gelman, 2002). The wasp eggs hatch and their larvae develop within the host haemocoel for about 2 weeks (Fulton, 1940; Beckage and Riddiford, 1978). Despite significant endocrinological changes in the caterpillar host during wasp larval development (Beckage and Gelman, 2002), the host's behaviour, including feeding, appears normal (Adamo et al., 1997). This continues until the end of the wasps' second larval instar, when the wasp larvae undergo their final larval moult as they scrape and digest their way out of the host cuticle (Adamo et al., 2016). After exiting the host, the wasps spin cocoons and pupate on the outside of the caterpillar (Fulton, 1940). At this time, the caterpillar host exhibits a profound decline in self-generated behaviours such as locomotion and feeding (Adamo, 2013). Caterpillars do not eat while the wasps emerge and their feeding behaviour never returns to baseline even after the wasps eclose as adults 4–5 days later (Adamo et al., 1997; Beckage and Templeton, 1986). The caterpillar ultimately starves to death (Kester and Jackson, 1996).

Host caterpillar feeding has been reported to slow about a day before the wasps emerge (Adamo et al., 1997; Miles and Booker, 2000). During wasp emergence, there are substantial changes in the levels of blood-borne neuroactive compounds. For example, neurohormonal octopamine levels increase (Adamo et al., 1997), as does the expression of the gene for the cytokine plasmatocyte-spreading peptide (PSP) in the host's fat body (Adamo et al., 2016). A similar cytokine, growth-blocking peptide (GBP), is also elevated in the armyworm Pseudaletia separata, when it is parasitized by the wasp Cotesia kariyai. This has been shown to alter neural function by increasing dopamine production in nervous system tissue (Noguchi et al., 2003). There is evidence that the increases in cytokines and hormones that occur as the wasps emerge are causally linked to the change in host behaviour (Adamo, 2019). For example, evidence suggests that the elevated titres of neurohormonal octopamine are at least partly responsible for the loss of rhythmic output from the central pattern generator(s) (CPG) in the host's frontal ganglion that regulates swallowing and foregut peristalsis (Miles and Booker, 2000).

Although normal late instar M. sexta caterpillars feed frequently, they exhibit short non-feeding periods when satiated. They also stop feeding just prior to moulting, a state referred to as ‘moult-sleep’ (MacWilliam et al., 2015). Additionally, caterpillars exhibit illness-induced anorexia (Adamo, 2005; Adamo et al., 2006). Caterpillars clearly have non-feeding periods, triggered by various internal conditions. The wasp could take advantage of these pre-existing mechanisms for suppressing feeding in the caterpillar by activating one of them according to its own needs. To determine whether wasps suppress host feeding by activating one of these pre-existing mechanisms, we examined the microstructure of parasitized caterpillar feeding behaviour before and after wasp emergence. We compared this with measurements of the microstructure of feeding in unparasitized caterpillars as they approached moult-sleep and during illness-induced anorexia. Because feeding is a complex behaviour, we predicted that feeding cessation would follow a particular pattern, i.e. the slowing of feeding components in a specific sequence, depending on the physiological context. If the microstructural changes that occurred during wasp-induced feeding suppression resembled the changes shown during one of the other natural examples of feeding cessation, then this would suggest a possible mechanism that the wasp could be exploiting.

We also examined the neurobiological underpinnings of the changes in feeding behaviour in parasitized larvae. Suppression of feeding could be accomplished by interfering with the ability of an organism to perceive food (sensory), to be able to physically respond to food (motor), or to reduce feeding drive (motivation). We focused on whether the motor programme for chewing was intact in parasitized caterpillars. Previous work has shown that at least one CPG, the CPG for swallowing found in the frontal ganglion, is compromised (Miles and Booker, 2000). The CPG for chewing is located in the suboesophageal ganglion (SEG; Griss et al., 1991; Rohrbacher, 1994a). As long as the SEG is connected to the mandibular muscles and to peripheral nerve inputs, the caterpillar shows a normal chewing rhythm, even in vitro (Griss et al., 1991). The chewing CPG is regulated by interneurons, such as IN 101, which receives both inhibitory and excitatory projections from the caterpillar's brain (i.e. supraoesophageal ganglion) (Rohrbacher, 1994b). Mechanical stimulation of the mouthparts can activate a few cycles of the chewing CPG, but the combination of mechanical and chemical stimulation from food is required to activate and sustain chewing and ingestion (Griss et al., 1991). Therefore, to suppress the feeding motor system, a parasitic manipulator could: (a) prevent sensory inputs from reaching the SEG and/or brain, (b) suppress neuromuscular connections, thereby preventing neural signals from reaching the mandibular muscles, (c) damage muscles, (d) interfere with the chewing CPG within the SEG, and/or (e) enhance inhibitory signals to the SEG.

To determine which of these possibilities may be used in the C. congregata–M. sexta parasite–host relationship, we recorded electromyograms from the chewing muscles of parasitized and unparasitized caterpillars as they fed. We made recordings of evoked excitatory junction potentials (EJPs) from the chewing muscle fibres, and measured the activity induced in the mandibular nerve (MdN) by stimulation of the mouthparts with a piece of plant material applied to the labrum. To determine whether ascending neural information might be the wasp's target, we severed the connectives between the suboesophageal and thoracic ganglia. Rowell and Simpson (1992) showed that cutting these connectives removed an inhibitory input and activated chewing in normal, unparasitized caterpillars. A similar result in parasitized caterpillars would indicate that the circuitry underlying chewing was likely intact but was being suppressed. Such a result would suggest that the wasp alters higher order feeding processes (e.g. feeding motivation) of its host, as opposed to disrupting the chewing CPG.

Colonies of M. sexta were reared in individual containers at Binghamton University as described in del Campo and Miles (2003) and at Dalhousie University as described in Adamo et al. (2016). Wasps (C. congregata) were reared at Binghamton University as in Miles and Booker (2000) or at Virginia Commonwealth University as in Bredlau and Kester (2015) and shipped to Dalhousie University as cocoons. Work at Dalhousie University was approved by the University Committee on Laboratory Animals (I20-08).

For experiments on parasitized fourth-instar caterpillars, caterpillars in their second instar were exposed to a cage of adult wasps. For experiments on fifth-instar caterpillars, caterpillars in their third instar were exposed to the wasps. No more than two wasps were allowed to parasitize a single caterpillar. Because C. congregata is a gregarious species, there are typically 80–200 larvae in a single host (Beckage and Gelman, 2002).

Because parasitized caterpillars do not develop at the same rate as unparasitized caterpillars (Beckage and Gelman, 2002), we could not directly compare them with control caterpillars of the same age. Comparisons within electrophysiological and feeding microstructure recordings were made as trends over time, and for parasitized caterpillars, time was presented as days before or after the parasite larvae emerged. For electromyogram and feeding behaviour studies, caterpillars were followed until the parasites emerged, and the number of days before this event was therefore known. For electrophysiological studies that required dissection, the days before parasite emergence were estimated, based on the average of data from undissected caterpillars. Pre-5 indicates 5 days before emergence, Pre-3 is 3 days before emergence, and so on. Em+1 indicates 1 day after emergence. For unparasitized caterpillars, 5th0 represents the day they moulted to the fifth instar, 5th1 represents 1 day later, and so on.

We examined the microstructure of feeding cessation due to different causes: illness-induced anorexia, moult-sleep and parasitism of both fourth- and fifth-instar caterpillars (Fig. 1). Fourth-instar caterpillars were used in some of the feeding microstructure studies because they exhibit moult-sleep and fifth-instar caterpillars do not. Parasitized fourth-instar caterpillars approaching parasite emergence were compared with unparasitized caterpillars that were approaching moult-sleep and with immune-challenged caterpillars. Parasitized fifth-instar caterpillars were compared with unparasitized fifth-instar caterpillars.

Immune activation leads to illness-induced anorexia in M. sexta, a behavioural state that has some similarities to parasitized caterpillars after wasps have emerged (Adamo, 2005). Fourth-instar day 2 (4th2) M. sexta caterpillars were weighed and assigned to one of three groups (n=9 per group): unparasitized (no treatment), immune challenged (injected with heat-killed bacteria and fungus) and sham (cuticle pierced with a sterile needle). There were no differences in the average initial mass among the three groups (one-way ANOVA, F_2,24=0.033, P=0.968). Illness-induced anorexia caterpillars were immune challenged with injection of a 10 µl mixture of heat-killed Bacillus cereus (Gram-positive bacterium, Microkwik culture, Carolina Biological, Burlington, NC, USA), Serratia marcescens (Gram-negative bacterium, Microkwik culture, Carolina Biological), and Beauveria bassiana (strain GHA, fungus, BotaniGard 22WP, Laverlam, Butte, MT, USA), as described in a previous study (McMillan et al., 2018). Immune-challenge injections were performed with a disinfected 10 µl Hamilton syringe (Model 701, Hamilton Company, Reno, NV, USA). The sham caterpillars were pierced with the sterile tip of the same model of syringe, in the same location as the immune challenge injection. All three groups of caterpillars were then returned to their 16 fluid ounce (∼0.5 l) clear polypropylene containers. They were allowed to feed ad libitum for 2 h and then were deprived of food for 1 h before the feeding trial to normalize gut contents across caterpillars. For these experiments, food deprivation was limited to only 1 h because any longer has previously been shown to affect feeding microstructure (Bowdan, 1988). In the feeding trial, caterpillar feeding behaviours were recorded until two full meals had occurred, or until an hour without feeding was observed. We used Bowdan's (1988) definition of a meal as a period of continuous feeding with no pauses longer than 2 min.

In the hours approaching the moult into their next instar, M. sexta caterpillars exhibit a period of reduced feeding and activity (MacWilliam et al., 2015). Fourth-instar moult-sleep caterpillars were selected (n=9) by estimating the time before they would begin the moult to their fifth instar. By measuring the extent of cuticle apolysis (separation of the cuticle from the body) at the seventh abdominal spiracle, the time until head capsule slippage, and the start of the moult, can be predicted (Langelan et al., 2000). Caterpillars were selected for this group if cuticle apolysis predicted head capsule slippage occurring within the next 8 h. Selected caterpillars were food deprived for 1 h to normalize their gut contents before the feeding trial and the trial followed the previously described protocol.

These caterpillars were parasitized on the first day of their second instar. Starting from the third day of their fourth instar (4th3), until 3 full days after the wasps emerged from the caterpillar, the feeding microstructure for parasitized caterpillars was recorded daily. Before each recording, the caterpillar was food deprived for 1 h to normalize gut contents. Each recording lasted until two meals had been recorded or until the caterpillars failed to feed for 1 h. We obtained the following number of recordings: fourth-instar unparasitized (controls) n=14, 2 days prior to wasp emergence (Pre-2) n=11, 1 day prior to wasp emergence (Pre-1) n=14, 1 day after wasp emergence (Em+1) n=14, 3 days after wasp emergence (Em+3) n=14.

We recorded the microstructure of feeding in parasitized and unparasitized fifth-instar caterpillars because these were the stages used for the electrophysiological studies. Healthy caterpillars were exposed to female wasps on the first day after the moult to the third instar. Once 10 days had passed and the caterpillars were in their fifth instar, parasitized caterpillars had their feeding microstructure recorded daily, continuing until 3 days after the emergence of larval wasps from the caterpillar. Recordings for fifth-instar caterpillars followed the same procedure as used for the fourth-instar caterpillars. Pre-2 and Pre-1 parasitized caterpillars were compared with healthy unparasitized caterpillars at days 5th2 and 5th3, respectively. The ages for unparasitized caterpillars were chosen because the average mass best matched those of unparasitized caterpillars on their respective days (Pre-2: 3.33±1.28 g, Pre-1: 3.37±1.33 g, 5th2: 2.72±0.86 g, 5th3: 3.49±1.14 g; means±95% confidence intervals, CI, n=21 for each group).

Data for feeding microstructure comparisons in all but the moult-sleep group were derived from the second meal recorded in a session to avert the impact of food deprivation (Bowdan and Wyse, 1997). Data for the moult-sleep caterpillars were derived from the second-to-last meal recorded before the caterpillars were confirmed as entering moult-sleep. We used the penultimate rather than the last meal because there was a high degree of variability in the structure of the final meal before moult-sleep, making it difficult to compare with other groups. Caterpillars still showed both the physiological and behavioural signs of approaching moult-sleep during the time of the penultimate meal (Langelan et al., 2000; MacWilliam et al., 2015).

For all feeding microstructure experiments, a custom-built automated mechanism was used to non-invasively record the microstructure of meals eaten by larvae feeding on a laboratory diet (Great Lakes Hornworm, Romeo, MI, USA). Caterpillars were placed on a plasticine ramp leading up to food resting on a force transducer (described below). The plasticine walls and a clear plastic ceiling ensured that the caterpillar stayed in position and reduced motion artifacts, while not restraining the caterpillar.

Force measurements were made using an 8 mm³ block of Manduca wheat germ diet placed on a force transducer (MLT050/A, ADInstruments, Grand Junction, CO, USA) connected via a bridge pod amplifier (ML110, ADInstruments) to a digital data acquisition system (ML760, ADInstruments). The force transducer's range was 0–50 g with a post-amplification sensitivity of 6.2 μV and LabChart range of −20 to +20 mV.

Data were acquired from behavioural recordings using LabChart (Chart5 v5.5.6, ADInstruments). This software produced values for the time between bites (in s) and the peak force amplitude (in mV) for each bite in a meal. Bites were then grouped into bouts, which were periods of repeated biting with pauses between bites not exceeding 1.15 s, as defined in other studies (Slater and Lester, 1982; Slater, 1974; Bowdan, 1988). To account for potential variation between the food cubes' physical quality as a force conductor and variation in individual caterpillar mass, the bite amplitude data were standardized within the individual meal, giving the lowest amplitude bite an arbitrary value of 10, and the highest amplitude bite a value of 100, and giving all other bite amplitudes a new value based on this scale.

Feeding microstructure comparison between unparasitized and parasitized fourth- and fifth-instar caterpillars was done via a linear mixed effects model performed in SPSS (version 28). This assessed the influence of parasitism, age and time point in a meal on each feeding microstructure variable, using caterpillar mass as a covariate. The model's fixed effects were parasitized status, age and time in meal, and all their possible interactions, with individual caterpillar intercepts as a random effect.

Fifth-instar caterpillars were used for these measurements. The area of a leaf of the host plant Nicotiana alata (Link & Otto) eaten by parasitized and unparasitized caterpillars was determined as follows. Larvae were deprived of food for 2 h, then placed on a leaf of N. alata held between two dowels, with the cut end submerged in water. Caterpillars were left undisturbed and observed for feeding activity. Only bites that took off pieces of the leaf were counted. The larvae were permitted to feed until they had taken 100 bites. Then, they were removed from the leaf and the leaf was scanned on a flatbed scanner. The number of pixels for the image was determined using Adobe Photoshop 7. Pixel counts from leaf images scanned before and after the 100 bites were converted to cm² and the amount consumed was calculated from the difference.

Fifth-instar caterpillars were used for all electrophysiological studies.

Caterpillars were deprived of food 2–4 h before recordings were made. Stainless steel wire electrodes (20 µm diameter) were placed in the closer muscles, which occupy most of the dorsal and lateral head capsule. Activity was recorded using an AC-coupled differential amplifier (Model 1600, A-M Systems Inc., Sequim, WA, USA). The recording electrode was implanted in the mandibular closer muscle through a hole pricked in the head capsule. The indifferent electrode was placed in the abdominal dorsal horn by snipping off the tip of the horn and inserting the wire through the open tip. Both wires were secured in place with cyanoacrylate glue (Krazy glue, Borden Inc.). The caterpillars were placed vertically along the edge of a leaf of N. alata that was held upright between two wooden dowels and the leaf's base was placed in water. Caterpillars were left undisturbed until feeding commenced, at which time muscle activity was recorded. The following number of recordings were made: unparasitized caterpillars 5th0 n=11, 5th2 n=10 and 5th3 n=11; parasitized caterpillars Pre-5 n=9, Pre-3 n=10 and Pre-2 n=10. Because post-emergence larvae feed so rarely (Beckage and Riddiford, 1982, 1983; see also this paper), electromyogram data was not taken for this group.

Recordings were sent to a digital data acquisition system (Digidata 1320, Axon Instruments) then recorded, displayed and analysed using Axoscope software (Axon Instruments). Electromyograms were analysed by averaging 10 consecutive cycles within a feeding bout from each of the caterpillars.

Recordings of the EJPs in the mandibular closer muscle were carried out for both parasitized and unparasitized caterpillars. For unparasitized caterpillars, recordings were successful for (n=10) individuals on 5th0, 5th2 and 5th3. For parasitized caterpillars, individuals were successfully recorded on Pre-5 (n=4), Pre-3 (n=11) and Pre-2 (n=8). In addition, recordings were made of parasitized caterpillars 24 h after emergence of the parasites (Em+1, n=13). Ten consecutive EJPs were averaged for each individual. Caterpillars were anaesthetized on ice, cut at the first abdominal segment, discarding the abdomen, and then dissected along the dorsal midline anteriorly through the head capsule. The gut was removed to expose the mandibular musculature and nervous system. Preparations were pinned out on a plastic Petri dish lined with silicone elastomer (Sylgard, Dow Corning, Midland, MI, USA) and bathed in physiological saline (Bestman and Booker, 2006). Recording and stimulating glass pipettes were prepared from thin (1 mm) walled borosilicate glass capillaries and pulled with a Flaming/Brown micropipette puller (Sutter Instrument Co., Novato, CA, USA). Recording electrodes had 30–60 MΩ resistance when filled with 2 mol l⁻¹ potassium acetate. To stimulate the MdN, a blunt micropipette was used as a suction electrode and the cut end of the MdN was suctioned into the electrode. Stimuli consisted of single pulses of 500 ms duration delivered to the MdN from a stimulator (MODEL 2100, A-M Systems Inc.). Muscle recordings were digitized and recorded as described for the electromyograms above. Traces were analysed off-line with Axoscope. To determine whether the effect of parasitism in the muscle fibres was restricted to the chewing motor system, or whether it was simply due to a general decline in all muscle function in the parasitized caterpillars, we also looked at EJP amplitudes in the ventral internal medial muscle of the abdominal body wall (VIM; Levine and Truman, 1985) in parasitized and unparasitized caterpillars. The dorsal nerve was stimulated and EJP amplitudes recorded as described for the mandibular closer muscle.

Extracellular recordings of the activity of the mandibular muscle motorneurons were made from the MdN, which exits the SEG. The chewing CPG was activated by application of a fragment of N. alata leaf to the taste receptors on the mouthparts. This technique was effective even in post-emergence parasitized caterpillars that did not feed and allowed us to compare the viability of this circuit in parasitized caterpillars that were still feeding with those of caterpillars that no longer initiated feeding bouts after parasite emergence. Caterpillars were anaesthetized on ice. The dorsal side of the caterpillar was glued onto a wooden applicator stick (Rubber Cement, Elmer's Products, Westerville, OH, USA), positioning the head capsule so that it extended beyond the tip of the stick. Rubber bands were tied around the caterpillar and stick to constrict the gut to prevent regurgitation and contamination of the thoracic region. The first rubber band was tied at the second abdominal segment and the second was tied at the eighth segment to secure the animal in place. The animal on its stick was then placed on a wax block. The sides of the thorax were pinned to the wax and stretched forward to tilt the head capsule backwards, exposing the SEG. A window was cut on the ventral side of the animal to expose part of the MdN projecting from the SEG.

Extracellular recordings were obtained from in vivo preparations of the MdN. The MdN was gently lifted on a silver wire hook electrode. The surrounding saline level was lowered and petroleum jelly was injected around and beneath the SEG to insulate it electrically from the rest of the body. Taste receptors on the mouthparts were stimulated by holding a leaf of N. alata to them. In most cases, this elicited the chewing motor pattern, and in some preparations these minimally dissected caterpillars would bite off pieces of the leaf. To verify that the neural activity was from the mandibular closer muscle motorneurons, we simultaneously recorded electromyograms from the mandibular closer muscle. Spikes recorded in the MdN that correlated 1:1 with spikes in the muscle were counted as activity of the closer muscle motorneuron. The number of successful recordings was: unparasitized caterpillars 5th0 n=7, 5th2 n=8 and 5th3 n=7; parasitized caterpillars, Pre-5 n=8, Pre-3 n=8, Pre-2 n=5; post-emergence parasitized caterpillars Em+1 n=5.

MdN recordings were also made in parasitized and unparasitized caterpillars to test the effects of cutting the connectives on the motor pattern. Recordings were made from: unparasitized caterpillars 5th2 and 5th3 n=19; parasitized caterpillars Pre-2 and Pre-3 n=12; and after emergence of the parasites Em+1 n=7. Caterpillars were prepared as described above. MdN activity upon stimulation of the mouthparts with a leaf was recorded to obtain pre-cut data, then the connectives between the SEG and first thoracic ganglion were cut. The MdN activity upon leaf application was recorded again post-cut after a 15 min waiting period to reduce damage artifacts.

All recordings were analysed using six successive bites in a trace. Traces were analysed off-line with Axoscope (version 10.3).

Prior to conducting parametric analyses (e.g. analysis of variance, ANOVA), data were tested for a normal distribution (Shapiro–Wilks), equality of error variances (Levene's test) and F-test for heteroscedasticity. Data were analysed using SPSS (versions 25, 26 and 27) and Prism (version 9.2).

Unparasitized healthy control, immune challenged and sham-injected fourth-instar, day 2 caterpillars all showed similar patterns in feeding microstructure (Table S1A), with feeding parameters changing over the duration of a meal but not differing across groups (Fig. S1, Table S1B). Similarly, unparasitized caterpillars approaching moult-sleep showed no differences in their microstructure compared with unparasitized fourth-instar controls ending a meal (Fig. S2, Table S1C,D). Both parasitized and unparasitized fourth-instar caterpillars showed longer interbout intervals and a decrease in mean bite amplitude over time in a meal, and there were fewer bites per bout with age (Fig. S3, Table S2). However, neither parasitism nor any interaction of factors influenced the feeding microstructure of fourth-instar caterpillars (Fig. S3, Table S2).

Although both control and parasitized caterpillars gained weight over the fifth instar, parasitized caterpillars (n=22) had lower mass than unparasitized caterpillars (n=27) for the entire fifth instar (Fig. 2A, repeated measures two-way ANOVA, significant effect of parasitism, F_1,47=237, P<0.001; change with time, F_1.7,78=1105, P<0.001; significant interaction, F_1.7,77=428.7, P<0.001). After the wasps emerged and the cocoons were removed from the caterpillar, caterpillars (n=23) continued to lose mass (0.13±0.12 g, mean±95% CI) over the next 3 days (paired t-test, t₂₂=5.15, P<0.001). Average mass 1 day after parasite emergence was 2.51±0.29 g, and 3 days after emergence it was 2.38±0.25 g (data not shown).

Freely feeding parasitized caterpillars, 3 days before parasite emergence (Pre-3), consumed less leaf area in 100 bites than 5th2 controls (Šidák's multiple comparison's test, P=0.005), but there was no difference between the leaf consumption of parasitized caterpillars (Pre-5) and 5th0 controls (P=0.98) or between Pre-2 and 5th3 controls (P=0.38) (Fig. 2B). The number of caterpillars in each group was: unparasitized 5th0 n=19, 5th2 n=10, 5th3 n=16; parasitized Pre-5 n=18, Pre-3 n=13, Pre-2 n=11.

Feeding microstructure differed between parasitized and unparasitized caterpillars, with parasitized caterpillars having shorter feeding bouts, longer interbout intervals and more bites per second (Figs 3 and 4, Tables 1 and 2; Table S3). Feeding microstructure also changed with the age of a caterpillar, with both parasitized and unparasitized caterpillars having more bites per second and longer interbout intervals on their second day of testing (Pre-1 and 5th3) compared with the first (Pre-2 and 5th-2) (Figs 3 and 4, Tables 1 and 2; Table S3). In both parasitized and unparasitized caterpillars on both days of testing, caterpillars showed a decline in bite amplitude from the beginning and middle to the end of their meals (Figs 3 and 4, Tables 1 and 2; Table S3). This was consistent with the decline in bite amplitude over the course of a meal in all of the groups of fourth-instar caterpillars (Figs S1–S3, Tables S1–S3). However, there were no significant interactions between factors, even though parasitism, day of testing and time in meal each individually and independently affected the measured feeding components (Table S3).

On the day the parasites emerged, Em+0, none of the 21 caterpillars tested fed. Only 9.5% of host caterpillars (2/21) were observed engaging in a full meal during the 1 h recording period the day after parasite emergence, and 3 days after parasite emergence, 5/21 caterpillars tested ate a full meal. Because of the small number of feeding caterpillars at these stages of parasitism, it was not possible to statistically compare the data with unparasitized controls. However, the few caterpillars that did eat trended toward shorter feeding bouts, longer interbout intervals and lower bite amplitudes, compared with the parasitized caterpillars prior to parasite emergence (Fig. 5, Table 2).

Electromyograms from the mandibular closer muscle in spontaneously feeding caterpillars showed that the number of spikes per burst increased in older unparasitized caterpillars (Fig. 6A; two-way ANOVA, significant effect of parasitism, F_1,54=14.9, P=0.0003; significant effect of time, F_2,54=17.5, P<0.0001; significant interaction, F_2,54=4.32, P=0.018). In contrast, parasitized caterpillars showed no increase in the number of spikes per burst with age (Šidák's multiple comparison test, Pre-2 versus Pre-5, P=0.75; Fig. 6A). Therefore, by day 3 after moulting to fifth instar (5th3), unparasitized caterpillars had more spikes per burst than parasitized caterpillars (Šidák's multiple comparison test, P=0.0003). Burst duration did not differ between parasitized and unparasitized caterpillars (Fig. 6B; Šidák’s multiple comparison test, non-significant effect of parasitism, F_1,53=1.13, P=0.29; no effect of time, F_2,53=0.14, P=0.86; and no significant interaction effect, F_2,53=2.2, P=0.12). Interburst interval did not differ significantly between parasitized and unparasitized caterpillars (Fig. 6C; two-way ANOVA, F_1,55=0.32, P=0.58; no significant effect of time, F_2,55=1.73, P=0.19; and no interaction, F_2,55=2.93, P=0.07).

The EJP amplitude was smaller in parasitized caterpillars than in unparasitized caterpillars. With time, the EJP amplitude declined in unparasitized caterpillars; however, parasitized caterpillars showed no trends in EJP amplitude with time (Fig. 6D; two-way ANOVA, changes due to parasitism, F_1,47=171; changes due to time, F_2,47=7.28, P=0.0018; interaction, F_2,47=6.9, P=0.0024). Interestingly, the parasitized caterpillars did not show any significant change in EJP amplitude after the parasites emerged, a time when the caterpillars largely cease feeding (Beckage and Riddiford, 1982, 1983) (Fig. 6D; one-way ANOVA, F_3,32=5.1, P=0.005, corrected using Benjamini–Hochberg, P=0.89, pre- versus post-emergence, Dunnett's multiple comparisons test, P>0.05). Resting potentials and EJP durations did not differ over time or between parasitized and unparasitized caterpillars (data not shown).

In order to assess whether the EJP changes we observed in the chewing muscle were confined to these muscles, or whether they were part of a general change in the physiology of other muscles in parasitized caterpillars, we looked also at a non-feeding muscle, the VIM. We found that while EJP amplitude changed in both sets of caterpillars over time, there were no statistically significant differences between parasitized and unparasitized caterpillars (data not shown; EJP VIM amplitude: two-way ANOVA, parasitized versus unparasitized, F_1,31=3.44, P=0.07; time, F_2,31=4.11, P=0.026; interaction, F_2,31=2.19, P=0.13). The number of successful recordings was: unparasitized caterpillars, 5th0 n=6, 5th 2 n=8 and 5th3 n=9; parasitized caterpillars, Pre-5 n=4, Pre-3 n=5 and Pre-2 n=5. There was also no statistical difference in the EJP amplitude of parasitized caterpillars before and after the wasps emerged (one-way ANOVA, pre- versus post-emergence, F_3,17=2.24, P=0.12, n=7).

In parasitized caterpillars, recordings from the MdN showed that the chewing motor programme could be activated by stimulating the mouthparts with a leaf from its host plant as in unparasitized caterpillars (Fig. 7). This was true even for the post-parasite emergence caterpillars, indicating that the taste receptors and mechanoreceptors on the mouthparts were still functional at this time. The chewing pattern in parasitized caterpillars was weaker than in unparasitized caterpillars, as the number of spikes per burst was significantly greater in unparasitized caterpillars (Fig. 7A; two-way ANOVA, parasitized versus unparasitized, F_1,36=108.4, P<0.0001; time, F_2,36=0.22, P=0.80; interaction, F_2,36=2.5, P=0.1). For parasitized caterpillars, post-emergence non-feeding caterpillars showed the same range of spikes per burst as pre-emergence parasitized caterpillars that were still feeding (pre- versus post-emergence, one-way ANOVA, F_3,21=0.89, P=0.46). Burst durations were also greater in unparasitized caterpillars compared with parasitized caterpillars, and both changed over time (Fig. 7B; two-way ANOVA parasitized versus unparasitized, F_1,36=58.35, P<0.0001; time, F_2,36=9.0, P=0.0007; interaction, F_2,36=4.56, P=0.017). There was no significant change in burst duration after the wasp larvae emerged from their host (pre- versus post-emergence, one-way ANOVA, F_1,21=1.48, P=0.25). Interburst intervals for parasitized and unparasitized caterpillars did not differ significantly and did not show significant changes over time, including 24 h after parasite emergence (Fig. 7C; two-way ANOVA, parasitized versus unparasitized, F_1,36=1.71, P=0.20; time, F_2,36=0.44, P=0.65; interaction, F_2,36=0.77, P=0.47; pre- versus post-emergence, one-way ANOVA, F_3,21=0.66, P=0.59).

Rowell and Simpson (1992) reported that cutting the connectives posterior to the SEG caused the release of a fictive chewing motor pattern. We tested whether this manipulation could also activate chewing in non-feeding parasitized caterpillars. After sectioning of the connectives posterior to the SEG, chewing increased dramatically in previously non-feeding post-emergence caterpillars when a section of leaf was applied to the mouthparts (Fig. 8A; paired t-test, t₉=6.87, P<0.0001, n=10 each group).

Cutting the connective altered the burst activity of the MdN. The procedure led to a dramatic decrease in interburst interval in all groups (Fig. 8B; Wilcoxon matched-pairs signed rank test, unparasitized controls, W=−136, n=19, P=0.005; pre-emergence caterpillars, W=−166, n=12, P=0.001; post-emergence caterpillars, W=−28, n=7, P=0.016), but had no effect on the number of spikes per burst (Fig. 8C; all Wilcoxon matched-pairs signed rank tests, P>0.1). Burst duration decreased following connective cuts in control caterpillars (Fig. 8D; Wilcoxon matched-pairs signed rank test W=−170, n=19, P=0.0002) and pre-emergence parasitized caterpillars (W=−58, n=12, P=0.02), but not in post-emergence caterpillars (W=−15, n=6, P=0.16).

Fig. 1 summarizes the three ways a feeding caterpillar may cease feeding that were examined in this study. We found that only parasitism significantly altered feeding behaviour. However, our results also show that fifth-instar parasitized caterpillars are clearly capable of chewing, both before and after parasite emergence.

Parasitized caterpillars showed an increase in mass over the fifth instar, demonstrating that they could bite, swallow and digest (Fig. 2A). However, parasitism did affect feeding behaviour, even during this active feeding stage. In the days before the wasps emerged, parasitized caterpillars showed shorter feeding bouts, longer intervals between bouts and increased bites per second during meals compared with unparasitized caterpillars (Table 1, Figs 3 and 4). We found significant changes in these parameters as early as 2 days prior to wasp emergence, rather than just 1 day before as reported previously (Adamo et al., 1997; Miles and Booker, 2000). While the increase in bites per second prior to wasp emergence seems counterintuitive, it may be an effect of the hormonal changes (e.g. elevated juvenile hormone, JH) induced by the wasps prior to emergence, which could lead to paradoxical effects on physiology and behaviour (Beckage and Gelman, 2002). As well, in the day before emergence, the wasp larvae could be a source of physiological stress for the caterpillar as they prepare to scrape out through its cuticle and cause stress-induced changes in feeding behaviour (Kupfermann and Weiss, 1981). We also found that parasitized caterpillars consumed less leaf area in 100 bites than unparasitized caterpillars 3 days before the parasites emerged, but not 2 days before. This may be the result of the unparasitized caterpillars consuming less on 5th3 than they did on 5th2, due to some of the unparasitized caterpillars approaching the wandering stage, when feeding ceases (Fig. 2B).

While feeding, the number of spikes per burst in parasitized caterpillars' mandibular closer muscle remained fairly consistent with the values shown on 5th0 as the fifth instar progressed (Fig. 6). In contrast, the unparasitized caterpillars tended to show increases in muscle activity and bite size as they grew through the instar. During the fifth instar, unparasitized caterpillars dramatically increase their muscle mass (Lin et al., 2011). This increase would be expected to result in an increase in motorneuron and muscle electrical activity as the fifth instar progressed. The lack of change in muscle activity through this instar in parasitized caterpillars could explain why they consumed less of the leaf with every 100 bites compared with unparasitized controls (Fig. 2B), and thus gained less mass during the fifth instar than unparasitized caterpillars (Fig. 2A). Parasitized caterpillars appeared to maintain the feeding dynamics they had on the day they moulted to the fifth instar (Table 2, Figs 2, 6 and 7).

Evoked EJPs in the mandibular closer muscle were smaller in amplitude in parasitized caterpillars than in unparasitized caterpillars throughout the fifth instar (Fig. 6D). This was not found for body wall muscle, suggesting that the feeding motor system may be targeted. Nevertheless, the smaller EJP amplitudes in the mandibular closer muscle do not prevent muscle activity. Perhaps more important to the reduced muscular activity in parasitized caterpillars is the lower spike activity of the mandibular closer muscle motorneuron (Fig. 7A,B). Similar to the results of Griss et al. (1991) for unparasitized caterpillars, stimulation of the mouthparts with a host plant leaf also activated the CPG for chewing in parasitized individuals. However, the mandibular closer muscle motorneuron showed fewer spikes per burst and shorter duration bursts than shown by unparasitized caterpillars (Fig. 7A,B). This difference increased as the time to wasp emergence approached, with the parasitized caterpillars maintaining values similar to those on day 5th0. It thus appears that the differences between parasitized and unparasitized caterpillars through the fifth instar are not due to a deterioration of the parasitized individuals' motor function. Instead, parasitized caterpillars show little change through their fifth instar in bite size, muscle activity and neural input. It is as though the parasitized individuals remain arrested at the early fifth instar, and fail to progress beyond it. This may be related to the maintenance of a larval endocrinological profile in parasitized fifth-instar caterpillars (Beckage and Gelman, 2002). In unparasitized caterpillars, JH levels decline to undetectable levels during the fifth instar, while parasitized fifth-instar caterpillars continue to show elevated levels of JH (Beckage and Riddiford, 1982).

After the wasps emerge from the host, there is a profound suppression of the caterpillars' feeding (Beckage and Riddiford, 1982, 1983; Adamo et al., 1997; Miles and Booker, 2000). We observed this in the present study as well, with 0% (n=21) of the parasitized caterpillars feeding on the day of wasp emergence. While some caterpillars (5/21) showed chewing behaviour 3 days after parasite emergence, they all continued to lose weight after wasp emergence, probably because chewing behaviour was infrequent, and because they were unable to swallow the food they chewed because of a lack of foregut peristalsis (Miles and Booker, 2000).

When considering the feeding microstructure for an entire meal, both parasitism status and age showed an effect on the fifth instar caterpillar. However, while unparasitized caterpillars can suspend feeding under different conditions, a comparison of feeding microstructures during illness-induced anorexia, moult-sleep and parasitism showed that all reached the end of their meal in a similar manner. All caterpillars regardless of group or instar showed similar changes in feeding microstructure and bite force as feeding declined toward its termination. This suggests meal termination occurs in the same way, regardless of context, as there was no impact on any of the microstructure variables from the interaction of time and any other condition. Possibly there is a common final pathway to stop chewing, but this would need to be determined with further study. Unfortunately, this result means that the feeding microstructure data cannot provide guidance as to how the wasp may exploit a specific existing mechanism to suppress host feeding.

As the parasites emerge from their caterpillar host, they physically damage the host body wall. Given the large immune response being generated at this time (e.g. Adamo et al., 2016), we wondered whether the nerves innervating the mandibular muscle might be damaged by immunopathology. We looked at the amplitude of evoked EJPs of the mandibular closer muscle and found no changes after the wasps emerge (Pre-2 to Em+1; Fig. 6D). We also looked at the activity of the MdN upon sensory stimulation of the mouthparts with a host plant leaf and found no significant differences in spikes per burst, interburst interval or burst duration over this time period (Fig. 7). This would suggest that the behavioural changes shown after parasite emergence are not due to a decline in the viability of the MdN motorneurons. It also indicates that the sensory inputs to activate the chewing motorneurons remain intact.

We tested the functionality of the feeding motor pattern circuitry in parasitized individuals both before and after parasite emergence by cutting the ascending input to the SEG. Cutting the ascending input ‘rescued’ parasitized caterpillars in that they began to show regular bursts of activity in the MdN and vigorous chewing movements upon application of a leaf to the mouthparts, similar to the responses of unparasitized caterpillars given the same treatment (Fig. 8). This result indicates that the neural circuitry for chewing was intact and capable of producing, as far as we could tell, a normal chewing pattern. One interesting implication of cutting the connective suggests that changes in input from the ventral nerve cord input play a role in producing non-feeding periods.

In post-emergence parasitized caterpillars, there does not appear to be damage to the sensory inputs, the CPG for chewing, motor nerves or mandibular muscles. Yet, these caterpillars fail to feed. Taken together, our results point toward a suppression of the initiation of the feeding bout. This suppression could be by way of the ventral nerve cord. However, we cannot rule out that changes in the brain are also involved. Feeding is known to be regulated by structures in the brain (Griss et al., 1991) and C. congregata induces changes in the M. sexta brain (e.g. Zitnan et al., 1995a,b). Additionally, removing ascending input releases only the chewing motor programme; it does not produce the full sequence of feeding behaviour such as re-activating swallowing (Miles and Booker, 2000). These observations suggest that the wasp manipulates multiple components of feeding behaviour when it emerges from its host (Adamo et al., 2016), but does not directly attack the feeding CPG.

Parasitic manipulators frequently alter their host's behaviour by hijacking immune–neural connections (Adamo, 2013). Wasp emergence produces a cytokine storm (Adamo et al., 2016), meaning that immunomodulatory molecules are present at high concentrations and could activate cytokine receptors found on muscle (e.g. Yang and Hultmark, 2017). In Drosophila, muscle participates in the immune response (Yang and Hultmark, 2017), and muscle in M. sexta appears to do the same (Adamo et al., 2023). In M. sexta, an immune challenge reduces the force of the defensive strike, suggesting that muscle is compromised (Adamo et al., 2023). Despite the plausible connection between the massive immune response induced by the exiting wasps and the simultaneous decline in host feeding, our results suggest that cytokines are not depressing mandibular muscle function. Cytokines may be involved in the reduction in host feeding (Adamo et al., 2016), but our results suggest that their major effect is on the CNS circuits involved in feeding initiation.

Parasitic manipulators should be selected to spare host muscle because changes in host behaviour require host muscle activity. Moreover, damaging host muscle will compromise host defensive behaviour, increasing the risk of death for both host and parasite. Therefore, it may be unsurprising that the skeletal muscular system does not appear to be compromised in parasitized M. sexta caterpillars. Parasitized M. sexta maintain robust defence behaviours before and after wasp emergence (Adamo et al., 1997), and we show that mandibular muscles remain effective. However, not all parasitic manipulators spare host muscle. The manipulative fungus Ophiocordyceps unilateralis induces its ant host (Camponotus castaneus) to bite into a leaf in a location that benefits fungal dispersal (Fredericksen et al., 2017). Despite the need for mandibular muscles for this behaviour, these muscles are invaded by fungal cells and suffer substantial atrophy (Fredericksen et al., 2017). However, both opener and closer muscles are affected, resulting in a ‘lock-jaw’ phenotype (Hughes et al., 2011). Other fungal parasitic manipulators also induce muscular atrophy, but they then use their own fungal cells to ensure that the host remains attached to the leaf (de Bekker et al., 2021). These examples demonstrate different ways that parasitic manipulators can resolve a fundamental trade-off (Adamo, 2023), i.e. if parasitic manipulators maximally extract their host's resources, then the host will be unable to produce the behaviour required for parasitic transmission. Parasitic manipulators must forego some host resources (Maure et al., 2011). Cotesia congregata larvae leave considerable host resources behind, allowing the host to function as a bodyguard for the wasp larvae during their metamorphosis (Adamo, 2023). The wasp's survival depends on the robustness of the caterpillar's bodyguard behaviour (Adamo, 1998), providing wasps with a selective advantage for leaving host muscles intact.

The lower body mass, smaller bite sizes when feeding and lower chewing muscle activity of parasitized caterpillars before parasite emergence suggest that parasitized caterpillars do not progress in their development much beyond the first day of the fifth instar. Whether this is related to altered endocrinological profiles (Beckage and Riddiford, 1982; Beckage and Gelman, 2002) remains to be determined. The suppression of feeding after emergence of the parasites is caused, in part, by a decline in the activity of the chewing CPG. This could be induced by ascending and/or descending inputs and/or an increase in CPG activation threshold (e.g. a neuromodulatory state that increases the stimulation required to activate the chewing CPG). Clearly, C. congregata uses a multipronged approach toward manipulating the behaviour and physiology of its host, both before and after parasite emergence.

We thank T. Quezada for collecting data on bite size, A. Miles and S. Ahmed for carrying out some of the electromyogram studies, and Sungwoo Song for his help in processing the feeding microstructure data.