All colonies of the Helicoverpa caterpillars were maintained in the laboratory at 75% ± 5% relative humidity and temperature (27 ± 1°C) under a controlled photoperiod (L16:D8). Both larvae of H. armigera and H. assulta were obtained from established laboratory colonies, which were reared on an artificial diet prepared from the following ingredients: wheat bran (150 g), soybean powder (80 g), yeast powder (25 g), casein (40 g), sorbic acid (3 g), ascorbic acid (3 g), sucrose (10 g), agar (20 g), vitamin composite powders (8 g), acetic acid (4 ml), and distilled water (1,500 ml) (Wu et al., 1990; Wu and Gong, 1997; Jiang et al., 2010). Adults were supplied with a 10% v/v solution of sucrose in water.

D-(-)-Fructose (Cas:57-48-7), D-(+)-glucose (Cas:50-99-7), L-proline (Cas:147-85-3), gossypol (Cas:303-45-7), capsaicin (Cas:2444-86-4), and tomatine (Cas:17406-45-0) were obtained from Beijing Solarbio Science & Technology Co., Ltd. Nicotine (Cas:54-11-5) was from Alfa Aesar. Ethanol absolute (Cas:64-17-5) and methanol (Cas:67-56-1) were from Tianjin De-En Chemical Reagent Co., Ltd. PVP (Cas:9003-39-8) was obtained from Tianjin Guangfu Fine Chemical Research Institute.

The dual-choice plant leaf disc bioassay was used to test the feeding preference of 5th instar larvae of the two Helicoverpa species as described by Wang et al. (2017). In general, leaf discs (10 mm diameter, about 156 mm²) were punched from fresh leaves of pepper Capsicum frutescens L., “Yu-Yi” (Solanaceae), which then were immersed in control or treatment solutions for 30 min. The plant primary metabolites D-fructose (1.0, 10, 30, 50 mM), D-glucose (1.0, 10, 30, 50 mM), and L-proline (0.1, 1.0, 10, 50 mM) were dissolved in water. The plant secondary metabolites gossypol, tomatine, and capsaicin were dissolved in solvent I (0.25% methanol, 5% ethanol, and 0.32% polyvinylpyrrolidone (PVP) in water) at 0.001, 0.01, 0.1, and 1.0 mM. Nicotine was dissolved in solvent II (0.16% PVP in water) at the concentrations of 0.001 mM, 0.01 mM, 0.1 mM, and 1.0 mM. The solvents were used as control.

Before the test, the fifth-instar caterpillars had been starved for about 8 h. A single caterpillar was placed in the center of a Petri dish (12 cm diameter) with a moist filter paper (Φ11 cm, Jiaojie^®, China). Four solvent-treated leaf discs and four plant metabolite-treated leaf discs were arranged in an ABABABAB fashion around the dish. All Petri dishes were put under evenly distributed LED strip lights (8,000 Lm) at a temperature of 27 ± 1°C. Areas of all remnants of leaf discs were measured by using a transparency film (PP2910, 3M Corp.) when two of the four disks of either plant (A or B) had been consumed. Each caterpillar was tested only once. For the feeding preference assays, at least 90 replicates were conducted.

The feeding preference index was calculated as follows:

The electrophysiological sensitivity of gustatory neurons in the styloconic sensilla on the maxillary galea of caterpillars to the plant metabolites was investigated using the single sensillum recording technique (van Loon, 1990; Roessingh et al., 1999). In brief, a head of an excised 5th instar caterpillar was mounted on a silver wire electrode which was connected to the input of a pre-amplifier (Syntech Taste Probe DTP-1, Hilversum, The Netherlands). The lateral or medial styloconic sensillum was recorded for the sensitivity to a stimulus at different concentrations. D-fructose, D-glucose, and L-proline were used as stimuli of primary metabolites with concentrations varying from 0.01, 0.1, 1.0 to 10 mM in 2 mM KCl. The previous work has shown that 2 mM KCl was an adequate electrolyte solvent for Helicoverpa caterpillars (Tang et al., 2015; Ma et al., 2016). The concentrations of gossypol, capsaicin, tomatine, and nicotine were from 0.001, 0.01, 0.1 to 1.0 mM. The first three stimuli were dissolved in solvent I, and nicotine was in solvent II. Both solvents for electrophysiological tests consist of 2 mM KCl. In case of synergistic interactions of mixed metabolites to GRNs, only a single sensillum in one caterpillar was tested for the responses to one kind of stimulus from low to high concentration. The electrolyte solvent was also tested as the control. For a single test, a glass microelectrode (tip diameter ca. 30 μm) filled with a stimulating solution was moved to contact with the tip of the lateral or medial sensillum with the aid of a micro-manipulator. The duration of a single stimulation was 2 s with a time interval of at least 3 min. Amplified signals were digitized by an A/D interface (IDAC-4, Syntech) and sampled into a personal computer. For each given concentration of a stimulus, the electrophysiological responses of at least 10 larvae were recorded.

The analysis of electrophysiological responses of styloconic sensilla to different stimuli was performed with the aid of AutoSpike v. 3.7 software (Syntech, Hilversum, The Netherlands). Briefly, in the case of the identification of GRNs, by measuring the amplitude, shape, and phasic temporal pattern, three impulse spikes were generally identified and labeled as small (S), intermediate (M), and large (L), which best responded to water, metabolites, and salt, respectively (Sollai et al., 2014; Ma et al., 2016). For distinguishing M-type spikes induced by primary metabolites and secondary metabolites, the intermediate 1 (M1) and intermediate 2 (M2) were assigned based on the spike amplitudes, correspondingly. The mean impulse frequency of each GRN in the first second (spk.s^–1) was calculated.

For the comparison of feeding preferences of caterpillars between control and treatment, the value of the preference index was arcsine transformed and then subjected to the paired-sample t-test (P < 0.05).

All the values of the impulse frequency (spk.s^–1) were square-root transformed before analysis. One-way ANOVA followed by the Student–Newman–Keuls (SNK) post-hoc test (P < 0.05) was used to compare the difference of the firing frequency of one type of GNR to one stimulus at different concentrations. The independent t-test was used to compare the mean impulse frequency of the same type of GRN between species. Finally, the GLM-Univariate was used to analyze the order of the mean impulse frequency of one type of GRNs to different compounds within species followed by the SNK post-hoc test for multiple comparisons (P < 0.05). All data were analyzed using SPSS software version 16.0.

In most recordings, three types of GRNs were identified from both medial and lateral sensilla of two Helicoverpa species in response to three plant primary metabolites, labeled as the “S” GRNs, “M1” GRNs, and “L” GRNs which best responded to water, primary metabolites, and salt, respectively (e.g., see identified representative GRNs in Figure 1). In the medial sensillum, the responses of “M1” GRNs of H. armigera caterpillars to each primary metabolite increased with the concentration increasing from 0, 0.01 mM, 0.1 mM, 1.0 mM to 10 mM [Figure 2A, one-way ANOVA of fructose: F_{(4, 45)} = 37.393, P < 0.0001; Figure 2B, glucose: F_{(4, 35)} = 51.272, P < 0.0001; Figure 2C, proline: F_{(4, 40)} = 29.965, P < 0.0001]. The mean response frequencies of “M1” GRNs of H. armigera induced by 10 mM fructose, 10 mM glucose, and 10 mM proline were 51.70 ± 3.490 spk.s^–1, 37.44 ± 4.378 spk.s^–1, and 55.62 ± 7.161 spk.s^–1, respectively.

Similarly, “M1” GRNs in the medial sensillum of H. assulta also showed increasing responses to each primary metabolite with increasing concentrations [H. assulta in Figure 2A’; one-way ANOVA of fructose: F_{(4, 56)} = 99567, P < 0.0001; Figure 2B’, glucose: F_{(4, 46)} = 55.164, P < 0.0001; Figure 2C’, proline: F_{(4, 48)} = 93.889, P < 0.0001]. The mean response frequency of “M1” GRNs in the medial sensillum of H. assulta to 10 mM fructose, 10 mM glucose, and 10 mM proline were 58.44 ± 5.430 spk.s^–1, 44.44 ± 4.045 spk.s^–1, and 61.0 ± 6.881 spk.s^–1, respectively. The responses of “M1” GRNs in the medial sensillum to one stimulus with the same concentration were always not significantly different between the two Helicoverpa species (Figures 2A–C, all comparisons: P > 0.05) except fructose at 0.01 mM which induced a significantly higher response of “M1” GRNs in H. armigera than that in H. assulta (Figure 2A, independent-sample t-test: df = 24, t = 2.411, P = 0.024). In the lateral sensillum, in contrast, the responses of “M1” GRNs to the three primary metabolites were low and the responses were similar between caterpillars of the two species (Figures 2A’–C’).

We also compared the general responding patterns of “M1” GRNs in one sensillum within the same species to the three primary metabolites using the GLM-Univariate with compounds and concentration as the fixed factors. It shows that the responses of “M1” GRNs in the medial sensillum of H. armigera caterpillars to the three compounds were significantly affected by both compounds and concentration (GLM-Univariate: compounds, df = 2, F = 4.199, P = 0.017; concentration, df = 4, F = 107.877, P < 0.0001). Analysis of the SNK post-hoc test showed that the responses of “M1” GRNs in the medial sensillum of H. armigera to glucose were significantly lower than those to fructose and proline (SNK post-hoc test: P < 0.05). However, for H. assulta caterpillars, the responses of “M1” GRNs in medial sensillum of H. assulta to the three compounds were not significantly affected by compound (GLM-Univariate: compounds, df = 2, F = 1.040, P = 0.356; concentration, df = 4, F = 234.979, P < 0.0001). Similarly, the responses of “M1” GRNs in lateral sensillum in both Helicoverpa species to the three compounds were also not significantly affected by compounds but affected significantly by concentrations (GLM-Univariate of H. armigera: compounds, df = 2, F = 0.563, P = 0.571; concentration, df = 4, F = 88.709, P < 0.0001; GLM-Univariate of H. assulta: compounds, df = 2, F = 1.630, P = 0.199; concentration, df = 4, F = 22.90, P < 0.0001).

The three primary metabolites also induced responses of “S” GRNs and “L” GRNs in both sensilla of the two Helicoverpa species. While the responses of the two types of GRNs to each compound were low with a non-significant change among different concentrations (SNK test after ANOVA for each compound: P > 0.05) (Figure 3).

The high concentration of fructose drove obvious appetitive feedings of both H. armigera caterpillars [Figure 4A; paired-sample t-test: 30 mM, t(162) = −1.999, P = 0.047; 50 mM, t(110) = −2.88, P = 0.005] and H. assulta caterpillars [Figure 4A’; paired-sample t-test: 10 mM, t(139) = −3.329, P = 0.002; 30 mM, t(94) = −5.704, P < 0.0001; 50 mM, t(104) = −7.116, P < 0.0001]. However, glucose showed no obvious effect on the feeding of H. armigera at the given concentrations [Figure 4B; paired-sample t-test: 1 mM, t(126) = 0.700, P = 0.485; 10 mM, t(116) = −0.218, P = 0.828; 30 mM, t(108) = 1.358, P = 0.177; 50 mM, t(117) = 0.522, P = 0.602], while it drove appetitive feedings of H. assulta caterpillars at high concentrations [Figure 4B’; paired-sample t-test: 30 mM, t(104) = −2.308, P = 0.023; 50 mM, t(103) = −2.865, P = 0.004].

Similarly, proline had no significant effect on the feeding of H. armigera caterpillars [Figure 4C; paired-sample t-test: 0.1 mM, t(136) = 0.400, P = 0.690; 1.0 mM, t(199) = −0.803, P = 0.423; 10 mM, t(107) = −1.020, P = 0.310; 50 mM, t(235) = −1.223, P = 0.223], while feeding preferences of H. assulta were significantly elicited at 1.0, 10, and 50 mM [Figure 4C’; paired-sample t-test: 1.0 mM, t(125) = −2.541, P = 0.012; 10 mM, t(141) = −2.252, P = 0.026; 50 mM, t(112) = −4.276, P < 0.0001].

The four investigated plant secondary metabolites induced high responses of the medial sensillum (e.g., see representative traces in Figures 5A,A’, 6A,A’) compared to the relatively low responses of the lateral sensillum of the two Helicoverpa species (e.g., see traces in Figures 5B,B’, 6B,B’). Three types of GRNs, in most traces, were identified in the responses of both sensilla to the four compounds, including the “S” GRNs, the “M2” GRNs, and the “L” GRNs, which best responded to water, the secondary metabolites, and salt, respectively (e.g., see representative identified GRNs in Figure 5).

In general, the responses of “M2” GRNs in both sensilla of the two Helicoverpa species induced by four secondary metabolites were high, while the responses of “S” GRNs and “L” GRNs in both sensilla induced by four secondary metabolites were relatively low. The responses of “M2” GRNs in the medial sensillum to each of the four plant secondary metabolites were different between the two species. Gossypol induced higher levels of response of “M2” GRNs in medial sensillum of H. armigera caterpillars than that of H. assulta (Figure 7A, independent-sample t-test of 0.001 mM: df = 18, t = 2.79, P = 0.0121; 0.01 mM: df = 31, t = 3.19, P = 0.0033; 0.1 mM: df = 34, t = 3.70, P = 0.0001; 1.0 mM: df = 38, t = 4.45, P = 0.0001). Tomatine at 0.001 mM and 0.01 mM induced lower responses of “M2” GRNs in H. armigera than those in H. assulta caterpillars (Figure 7B, independent-sample t-test of 0.001 mM: df = 18, t = −7.65, P < 0.0001; 0.01 mM: df = 41, t = −4.06, P = 0.0002) but elicited higher levels of response at high concentration in H. armigera than that of H. assulta (Figure 7B, independent-sample t-test of 0.1 mM: df = 33, t = 3.36, P = 0.002; 1.0 mM: df = 33, t = 1.26, P = 0.2128).

Different from tomatine, nicotine at 0.001 mM and 0.01 mM induced higher levels of responses of “M2” GRNs in the medial sensillum of H. armigera than those of H. assulta (Figure 7C, independent-sample t-test of 0.001 mM: df = 19, t = 8.69, P < 0.0001; 0.01 mM: df = 28, t = 6.91, P < 0.0001) but elicited lower levels of response at 1.0 mM in H. armigera than that of H. assulta caterpillars(Figure 7C, 1.0 mM: df = 17, t = −2.88, P = 0.0105). Capsaicin elicited relatively lower levels of responses of “M2” GRNs in the medial sensillum of H. armigera caterpillars than those of H. assulta caterpillars (Figure 7D, independent-sample t-test of 0.01 mM: df = 20, t = −2.35, P = 0.0271; 0.1 mM: df = 31, t = −4.05, P = 0.0003; 1.0 mM: df = 43, t = −2.33, P = 0.0245).

For responses of “M2” GRNs in the lateral sensillum, it showed that gossypol and tomatine induced low and similar responses of “M2” GRNs between the two Helicoverpa species (Figure 7A’ 0.001 mM gossypol: df = 13, t = 0.01, P = 0.99; 0.01 mM gossypol: df = 17, t = −0.02, P = 0.98; 0.1 mM gossypol: df = 19, t = −0.70, P = 0.49; 1.0 mM gossypol: df = 22, t = 0.54, P = 0.60; Figure 7B’, 0.001 mM tomatine: df = 15, t = −0.92, P = 0.37; 0.01 mM tomatine: df = 22, t = −0.92, P = 0.37; 0.1 mM tomatine: df = 20, t = −0.40, P = 0.69; 1.0 mM tomatine: df = 19, t = 0.18, P = 0.86). However, the responses of “M2” GRNs in the lateral sensillum to both nicotine and capsaicin at 0.1 mM and 1.0 mM were higher in H. assulta caterpillars than those of H. armigera (Figure 7C’, 0.1 mM nicotine: df = 23, t = −3.30, P = 0.0031; 1.0 mM nicotine: df = 29, t = −4.13, P = 0.0004; Figure 7D’, 0.1 mM capsaicin: df = 18, t = −3.23, P = 0.0047; 1.0 mM capsaicin: df = 16, t = −9.27, P < 0.0001).

Four plant secondary metabolites also induced responses of “S” GRNs and “L” GRNs in both sensilla of the two Helicoverpa species. While the responses of the two GRNs to each compound were low with non-significant change among different concentrations (SNK test after ANOVA for each compound: P > 0.05) (gossypol: Figures 8A,A’, B,B’; tomatine: Figures 8C,C’, D,D’; nicotine: Figures 9A,A’, B,B’; capsaicin: Figures 9C,C’, D,D’).

By comparing the responses of “M2” GRNs within one sensillum to the four secondary metabolites, it shows that the responses were significantly affected by both compounds and concentrations in either Helicoverpa species (GLM-univariate analysis of medial sensillum of H. armigera: compounds, df = 3, F = 39.814, P < 0.0001; concentrations, df = 4, F = 188.576, P < 0.0001; compounds × concentrations, df = 12, F = 7.659, P < 0.0001; medial sensillum of H. assulta: compounds, df = 3, F = 19.4448, P < 0.0001; concentrations, df = 4, F = 151.172, P < 0.0001; compounds × concentrations, df = 12, F = 17.406, P < 0.0001) (Table 2). However, the ranks of the general responding frequency between the two species were different. The response of “M2” GRNs in medial sensillum of H. armigera was the strongest to nicotine, followed by gossypol and tomatine, then low response to capsaicin (Table 3). However, for H. assulta, tomatine induced the strongest response of “M2” GRNs in the medial sensillum, followed by nicotine, capsaicin, and gossypol (Table 3). For the “M2” GRNs in the lateral sensillum between the two species, tomatine induced relatively stronger responses than those induced by gossypol, nicotine, and capsaicin in H. armigera, whereas gossypol induced the lowest responses compared to those by other three compounds in H. assulta (Table 3).

Gossypol at 0.1 and 1.0 mM drove appetitive feedings in H. armigera caterpillars [Figure 10A; paired-sample t-test: 0.1 mM, t(135) = −4.403, P < 0.0001; 1.0 mM, t(97) = −3.415, P = 0.001], but 0.1 mM and 1.0 mM gossypol drove aversive feedings in H. assulta caterpillars [Figure 10A’; paired-sample t-test: 0.1 mM, t(99) = 3.268, P = 0.001; 1.0 mM, t(137) = 2.179, P = 0.031]. Tomatine at concentrations of 0.01 and 0.1 mM drove appetitive feedings in H. armigera caterpillars [Figure 10B; paired-sample t-test: 0.01 mM, t(103) = −2.371, P = 0.02; 0.1 mM, t(114) = −3.324, P = 0.001], while 0.1 mM and 1.0 mM tomatine significantly deterred feedings of H. assulta caterpillars [Figure 10B’; paired-sample t-test: 0.1 mM, t(118) = 6.941, P < 0.0001; 1.0 mM, t(170) = 9.369, P < 0.0001].

Nicotine at the concentration of 1.0 mM deterred feedings of H. armigera caterpillars [Figure 10C; paired-sample t-test: 1.0 mM, t(98) = 6.471, P < 0.0001] but drove appetitive feedings of H. assulta caterpillars at concentrations of 0.1 and 1.0 mM [Figure 10C’; paired-sample t-test: 0.1 mM, t(101) = −7.569, P < 0.0001; 1.0 mM, t(110) = −2.916, P = 0.004]. Capsaicin at the concentration of 1.0 mM significantly drove aversive feedings of H. armigera caterpillars [Figure 10D; one-sample t-test: 1.0 mM, t(100) = 2.972, P = 0.004), while 0.01 and 0.1 mM capsaicin significantly drove appetitive feedings of H. assulta caterpillars [Figure 10D’; one-sample t-test: 0.01 mM, t(90) = −5.727, P < 0.0001; 0.1 mM, t(100) = −2.412, P = 0.018].

Fructose, glucose, and proline have been widely reported to be phagostimulants for a variety of insect herbivores (Albert et al., 1982; Bernays and Chapman, 2001; Liscia et al., 2004; Jiang et al., 2015; Mang et al., 2016). Our present study also shows that fructose could drive appetitive feedings of caterpillars in both Helicoverpa species. Glucose and proline at the given concentrations drove appetitive feedings of H. assulta caterpillars but had no significant effects on the generalist species H. armigera, suggesting that the generalist is less sensitive to the two compounds. This result is consistent with our previous study that the feeding preference of H. armigera caterpillars is more flexible than that of H. assulta if caterpillars pre-exposed to different diets (Wang et al., 2017). The neural constraint hypothesis predicts that specialist herbivores always make more accurate decisions than generalists in the process of selection plants (Bernays, 1998). Our data provide further evidence that the specialist had better ability to perceive the sugars or essential nutrients than the generalist. However, our result indicated that the responding patterns of GRNs in galeal sensilla to each primary metabolite were similar between the two species, suggesting that the difference of feeding preferences should not be attributed to the firing rate of peripheral GRNs but might be from differences of the processing information within the central nervous system.

Gossypol and tomatine are two major plant secondary metabolites from cotton and tomato, respectively, which are toxic or aversive on herbivorous insects (Vickerman and de Boer, 2002; Mulatu et al., 2006; Arnason and Bernards, 2010; Carriere et al., 2019). Our results also show that the two compounds drove aversive feedings of the specialist H. assulta, but caterpillars of the generalist H. armigera exhibited appetitive feedings for the two secondary metabolites. Such kind of secondary metabolites drove appetitive feedings of the generalist herbivores; to our knowledge, they have not been reported to date. We postulate that it should be attributed to the extraordinary adaptive capacity of caterpillars of H. armigera to the two compounds, for example, the tolerance and detoxifying metabolism (Mao et al., 2007; Zhou et al., 2010; Bretschneider et al., 2016; Krempl et al., 2016), while caterpillars of the specialist H. assulta do not feed on cotton and tomato plants in nature (Mitter et al., 1993) and exhibit aversive responses to the two secondary metabolites.

Nicotine (Szentesi and Bernays, 1984; Shields et al., 2008; Hori et al., 2011; Sollai et al., 2015) and capsaicin (Cowles et al., 1989; Hori et al., 2011; Li et al., 2020) have been generally reported as feeding deterrents for herbivorous insects. However, our results demonstrate that the two solanaceous alkaloids elicited appetitive feedings of the specialist H. assulta, while they drove aversive feedings of the generalist H. armigera. We also postulate that it could be attributed to the specialist H. assulta being more adaptive to the two alkaloids than the generalist H. armigera. Firstly, it is known that tobacco and hot pepper are two limited host plants of the specialist H. assulta (Mitter et al., 1993), while the generalists have to deal with lots of toxic plant metabolites based on the neural-constraint hypothesis (Levins and Macarthur, 1969; Bernays and Wcislo, 1994; Bernays and Funk, 1999). Secondly, the adaptations of specialists to nicotine and tobacco plants have been well reported on caterpillars of the tobacco cutworm Manduca sexta (Snyder et al., 1993; Glendinning, 2002; Wink and Theile, 2002; Govind et al., 2010; Kumar et al., 2014). For capsaicin, it has been found that the larval development of H. assulta could benefit from the dietary capsaicin compared to the negative effects on H. armigera (Ahn et al., 2011; Jia et al., 2012). At the level of metabolism, the capacity of degrading the capsaicinoids in H. assulta was overall higher than that in H. armigera (Zhu et al., 2020). Then, our data provide further evidence of adaptation of the specialist H. assulta to the toxic plant metabolites at the behavioral and chemosensory levels, which is similar to the attractive effects of “token stimuli,” the specific secondary metabolites from host plants, on other investigated specialist herbivores (Renwick and Lopez, 1999; del Campo et al., 2001; Miles et al., 2005; Sollai et al., 2018).

For the response of galeal sensilla to the four secondary metabolites, it also indicates that each of the four secondary metabolites stimulated different responding patterns of GRNs between the two closely related species. Combining the differences of feeding preferences with the taste response of GRNs of the two species, it suggests that the activities of peripheral GRNs to the four alkaloids could contribute to the difference of feeding behaviors between the two Helicoverpa species. Therefore, it seems that the neural coding for behavioral decisions of the investigated secondary metabolites in the two Helicoverpa species is different from that for behavioral decisions of the primary metabolites. The present results suggest that the two Helicoverpa species evaluate the plant primary metabolites differently at the CNS level, while they evaluate the secondary metabolites differently at both peripheral and central levels.