Multivoltine insects can produce multiple generations in one year. Favorable conditions support more generations, leading to serious outbreaks. For herbivores, plant nutrient availability is a major environmental factor affecting fitness and it can shift substantially throughout seasons. In a stochastic environment, organisms can adopt several strategies to regulate their nutrient intake and maximize performance. However, data regarding nutrient regulation of wild herbivores are scarce, and even more so regarding potential intergenerational plasticity. To bridge this gap, we measured nutritional regulation and performance of an outbreaking multivoltine herbivore – one of the most serious agricultural pests in the Sahel: Oedaleus senegalensis. We surveyed a field population in Senegal and measured its nutritional preference and regulation across two generations (G1 and G3) using artificial diets and plant choice experiments. In the field, G1 locusts were five to ten times more abundant than G3 locusts. We found that G1 and G3 locusts selected different protein: carbohydrate ratios but also that the strength of regulation was different. G1 locusts regulated their nutrient target more tightly than G3 locusts. In contrast, studies with laboratory populations demonstrate strong regulation for grasshoppers, appearing less plastic than field populations. Both generations selected a carbohydrate-biased nutrient ratio, although it was more carbohydrate-biased for G3 locusts. In both cases, plant nutrient contents in the field were more protein-biased than their preferred diet. Therefore, choices by locusts were likely influenced by other ecological variables such as leaf toughness or plant defenses. G1 females were heavier and laid more eggs than G3 females. However, G3 locusts survived longer during the experiment than G1 locusts, suggesting a potential generational trade-off between reproduction and survival. Our data highlight the importance of studying nutritional regulation in situ and incorporating field and lab data to better understand foraging decisions and nutritional trade-offs.

I Oedaleus senegalensis: a multivoltine and migratory species

The Senegalese grasshopper O. senegalensis is a grass‐feeder and a major pest of millet and other cereal crops of subsistence agriculture in the Sahel zone of West Africa. O. senegalensis typically produces three generations throughout the rainy season, although when conditions are unfavorable, only two generations are produced (Batten 1969). This species is considered a non-model locust (Song 2011) and locust phase polyphenism (Pener and Simpson 2009, Cullen et al. 2017) is typically poorly understood in non‐model species (Song 2011). However, Senegalese grasshoppers, like other locust species, are migratory. They are nocturnal fliers that can travel hundreds of kilometers in one night (Riley and Reynolds 1983, Cheke 1990, Maiga et al. 2008). G1 and G2 travel from South to North along the Intertropical Convergence Front, ahead of the heavy rains. G3 migrates back South at the end of the rainy season (Maiga et al. 2008).

II Field site, host plants, and locust abundance

We conducted these studies in the summer of 2018 in the Kaffrine region of Senegal. The regional woody shrubland savanna landscape is topographically flat and marked by agricultural expansion that has replaced native dry forests (Mbow et al. 2008). Precipitation ranges from an average of 2 mm in the dry season (November–April) to an average of 737 mm during the rainy season (May–October). Dry season temperatures average ~ 27 °C compared to rainy season average of ~29 °C (D’Alessandro et al. 2015). The early rainy season is marked by quick onset of sprouting vegetation and concomitant emergence or increased activity of associated animals, fungi, and bacteria. By the end of the rainy season, annual plants including crops are reaching maturity with thick leaves and seeds (Fig. 1), and natural enemies of grasshoppers are well established and ubiquitous.

The Kaffrine region, along with Kaolack and Fatick regions, is known as the ‘West Central Agricultural Region’, or Peanut Basin (Tappan et al. 2004) and produces most of the country’s millet and peanut. The two crops are typically grown in rotation. Pearl millet (Pennisetum glaucum) is a rain‐fed crop with excellent tolerance to drought, sandy soil, low nutrient availability and high temperatures. While O. senegalensis is considered the main pest of millet, these grasshoppers are typically found to be more abundant in fallow fields containing weed species (Toure et al. 2013, Word et al. 2019, Le Gall et al. 2020a). We decided to contrast millet with one of those weed grasses. We selected Paspalum scrobiculatum, also called kodo millet or dugubupicc locally (“dugub” is millet and “picc” is bird in Wolof), a wild native grass because it is readily eaten by O. senegalensis and is preferred over other known host grass species like Cenchrus biflorus (Maiga et al. 2008) as shown by a cafeteria experiment that we ran in 2016 (SI, Table 1).

We measured locust abundance and collected locusts for our experiments in the village of Gniby in fallow and millet fields. We selected Gniby because it consistently supports high density O. senegalensis populations (Word et al. 2019, Le Gall et al. 2020a). We collected adult G1 on August 4th and 5th 2018 at three locations 1) 14°43N, 15°67W; 2) 14°42N, 15°67W; 3) 14°25N, 15°40W. On September 20th and 22nd we recorded abundance and collected adult G3 at two locations: 1) 14°41N,15°66W and 2) 14°41N,15°67W. To measure abundance, for G1 we counted locusts in twenty plots of one square meter each. Each plot was separated by at least ten meters. For G3, the millet reached above eight feet high, so we used a transect method and recorded abundance along ten transects that were ten meters long, one meters wide, and separated by five meters each. Results were standardized to locusts per square meter. At both time points, locusts were transported to the Direction de la Protection des Végétaux (DPV) field station in Nganda and kept in wire mesh cages with ambient local vegetation before being used for subsequent experiments.

III Artificial diet choice experiment

The aim of this experiment was to compare the protein:carbohydrate ratios selected by G1 and G3 adults when provided with artificial diets. We initiated the choice experiments in August after the start of the rainy season (G1) and in September (G3) towards the end of the rainy season within a day or two of locust collection. For this, we weighed and put locusts in individual aerated plastic containers (15 x 10 cm). Each cage contained a water tube, a perch for roosting, and two dishes containing artificial diets. We prepared and dried the food in our laboratory at Arizona State University (United States) following the method developed by Simpson and Abisgold (1985). In total, we made three diets varying in protein to carbohydrate ratios but otherwise isocaloric: p35:c7, p28:c14, and p35:c7. For each diet, “p” stands for percent of protein in the diet, and “c” stands for percent of carbohydrates in the diet, by dry mass. The protein component of all foods was a 3:1:1 mix of casein, peptone and albumen, while the digestible carbohydrate (henceforth carbohydrates) component was a 1:1 mix of sucrose and dextrin. All foods contained similar amounts of Wesson’s salt (2.4%), cholesterol (0.5%), linoleic acid (0.5%), ascorbic acid (0.3%) and vitamin mix (0.2%) (Dadd, 1961). The remainder of the diet was cellulose, a non-nutritive bulking agent.

We gave the locusts one of two treatments of pre-weighed diet pairings: p7:c35 & p35:c7 or p7:c35 & p28:c14 and used 20-25 locusts per treatment (approximately half males and half females). After three days, the diets were removed and dried for 24-36 h at 60ºC (drying oven Memmert and Kowell C1-1) and then re-weighed to record consumption at the nearest 0.1 mg. Locusts were weighed at the beginning and the end of the experiment to calculate mass gain.

IV Plant choice experiment

The aim of this experiment was to contrast nutrient selection results from the artificial diet choice experiment with nutrient selection from a choice experiment conducted with local plants. For G1, fresh plants were collected in Nganda on the day the experiment started and kept bundled with their stems soaked in water. The millet (P. glaucum) was collected at the seedling stage (~ 25 cm tall, 1-4 leaves sprouted) in a local collaborator farmer’s field. The weed (P. scrobiculatum) was collected on the side of the road, also at seedling stage (~15 cm tall). For G3, we followed the same protocol but this time we collected mature weed leaves, mature millet leaves, and millet seeds at the milky stage (Fig. 1) which are also consumed by locusts (Boys 1978). The first plant choice experiment was started on August 5th; the second one involved an additional treatment (millet seeds) and therefore had to be divided in two experimental blocks to be manageable. These experimental blocks took place on September 21st and September 27th.

For each experiment, we put 12 male and 12 female locusts (n=24) in individual plastic cages (14 x 8 x 4cm) fitted with a perch for roosting. The leaf stems were maintained in water in a plastic tube sealed with cotton. We cut the millet ears in pieces of roughly 4-5 cm and they were given as is. The experiments each lasted 24 h. At the end of each experiment, we dried all plant material. We estimated dry consumption from regression equations. For this we weighed 20-32 fresh leaves in August, and again in September; we then dried them at 60ºC for 48 to 72 h and re-weighed them. Regression equations for transforming between plant fresh and dried mass can be found in the supplementary information (SI, Appendix 1).

For G3 locusts, we added millet ears. However, despite cutting the millet ears, each piece was several folds heavier than grasshoppers and thus fluctuation in water content made it impossible to collect an accurate estimate of consumption with the same method used for leaves. Thus, we used visual observation to establish preference, for this an observer recorded which food item (weed leaves or millet seeds) had the most grasshopper feeding damage after 30 minutes.

V Plant no-choice

For the plant no-choice experiments, we used the same setup as for the plant choice experiment except that we gave the grasshoppers one plant item only: young millet or weed leaves in August (start date August 6th), and mature millet leaves, millet kernel, or mature weed leaves in September (start date September 24th). We used 24 grasshoppers per treatment, half males and half females for G1. We had difficulties collecting enough females for G3 so we used 24 grasshoppers per treatment: 9-10 females and 14-15 males. Each time, the experiment lasted a week and we changed millet and weed leaves every other day and every three days for the kernels. For each grasshopper, we recorded consumption and egg production (for females). We reported consumption for day 0-2 when most locusts were still alive. The cages were checked daily for mortality and the presence of parasites at the time of death was recorded (unidentified dipterans made up the vast majority of parasites and about 25% of the grasshoppers were parasitized).

VI Chemical analysis

We ran the chemical assays on the dried plant material in our laboratory at Arizona State University (United States). For this, we ground plant samples for 30 s at 200 rpm using a Retsch MM 400 ball mill. We measured plant protein content with a Bradford assay and non-structural carbohydrate content using the phenol–sulphuric acid method (Deans et al. 2018).

VII Statistical analysis

Locust abundance in the field and locust mass in the plant no-choice experiment were analyzed using ANOVAs. (Although the data were not normally distributed, ANOVA is robust against the violation of normality when group sizes are equal (Donaldson 1968)). Other analyses met the assumptions of parametric tests. The protein:carbohydrate ratios selected in the artificial diet and plant choice experiments were analyzed with MANCOVAs using start mass as a covariate to account for size differences. The dry amount of plant eaten in the choice experiment and plant no-choice experiment, the protein and carbohydrate intakes from the artificial diet and the plant choice experiments were analyzed using ANCOVAs with start mass as a covariate. Sex was included as a cofactor. Visual preference (locusts that did not eat after 30 minutes were not included in the analysis) in the plant choice experiment, egg laying and parasitism in the plant no-choice experiment, were analyzed by nominal logistic fit. Survival in the plant no-choice experiment was analyzed by survival analysis with Weibull distribution.