Host plants of the non‐swarming edible bush cricket Ruspolia differens

Abstract The edible Ruspolia differens (Orthoptera: Tettigoniidae) is a widely‐consumed insect in East Africa but surprisingly little is known of its host plant use in the field. We studied host plants used by non‐swarming R. differens for 15 months, in central Uganda. In particular, we assessed the use of host plant species with respect to host cover in the field and host parts used by R. differens, also recording their sex, developmental stages, and colour morph. Ruspolia differens were found on 19 grass and two sedge species and they were observed predominantly (99% of 20,915 observations) on seven grasses (namely, Panicum maximum, Brachiaria ruziziensis, Chloris gayana, Hyparrhenia rufa, Cynodon dactylon, Sporobolus pyramidalis, and Pennisetum purpureum). Ruspolia differens was most frequently observed on the most common grass of each study site but P. maximum, and S. pyramidalis were used more frequently than expected from their cover in the field. Furthermore, R. differens were observed predominantly on inflorescences (97% of feeding observations) and much less frequently on the leaves (3.0%), stems (0.1%), and inflorescence stalks (0.1%) of grasses and sedges. Host use was not independent of sex, developmental stage, or colour morph. Panicum maximum was the preferred host of the youngest nymphs of R. differens. Overall, our findings indicate that a continuous supply of diverse grass resources with inflorescences is necessary for the management and conservation of wild populations of R. differens.

the Indian Ocean (Bailey & McCrae, 1978;Massa, 2015). The typical colour morphs are green and brown, while rarer colour morphs include purple superimposed on either green or brown (Bailey & McCrae, 1978;McCrae, 1982). In non-swarming populations, all developmental stages can be found throughout the year but the population densities are higher in the rainy seasons and lower in the dry seasons . The individuals are mainly nocturnal, but in the swarming phase the activity-level can remain high during the daytime, when they can opportunistically be predated by birds (Bailey & McCrae, 1978). Swarming usually occurs during and after the peak of the rainy seasons, for example, in Uganda, around May and in November-December (Bailey & McCrae, 1978). Wild R. differens are harvested during swarming periods with increasingly efficient light traps (Okia et al., 2017), yet the specific reproductive areas of the swarming R. differens have not been identified, and the long-term consequences of harvesting wild populations are completely unknown.
Surprisingly little is known about host plants used by R. differens. In the field, individuals have been observed feeding mainly on flowers and the young grains of grasses (Bailey & McCrae, 1978;Swaine, 1964). When reared in the laboratory they readily accept artificial foods and the leaves, flowers, and grains of many grasses, including cultivated cereals (Brits & Thornton, 1981;Hartley, 1967;Malinga et al., 2018aMalinga et al., , 2018bValtonen et al., 2018). Most accounts of R. differens' requirements are based on sporadic field observations, fragmentary data, or are deduced from experience in the laboratory. Empirical information on the host plants used by R. differens in the wild is needed to understand the importance of different plant species for the nutritional ecology of R. differens. Such information could be also used in future for the management and conservation of R. differens populations in the wild.
In this study, we report a long-term assessment of the host plants used by non-swarming R. differens in two semi-natural grasslands of central Uganda. We addressed the following questions: (a) What plant species do R. differens use in the field, and are the plant species used in the same proportions as they occur in the field? (b) What plant parts are used by R. differens in the field? (c) Do females and males, or different developmental stages, or different colour morphs of R. differens differ in their host use?
Overall, are (d) females and males, and (e) the green and brown colour morphs represented in equal proportions in the nonswarming population? If R. differens is a facultatively oligophagous grass-specialist, as predicted by laboratory assessments, with a clear preference for certain grass or sedge species , we predicted that certain host species are used more frequently than their abundance in the field alone would predict. We also hypothesised that host use by R. differens may vary depending on sex or developmental stage, due to the different physiological state or nutritional requirements of individuals (Behmer & Joern, 1994;Boys, 1978;Unsicker, Oswald, Köhler, & Weisser, 2008).
Finally, we hypothesised that host use is different among colour morphs since different grasses potentially provide different levels of camouflage for green and brown colour morphs (Karpestam & Forsman, 2011).

| Study area and sites
The study was conducted from November 2015 to January 2017 in grazing lands at the Makerere University Agricultural Research Institute, Kabanyolo (MUARIK), Uganda. This mixed farm is located at 0°28′N, 32°37′E, about 20 km north of Kampala, at an elevation of approximately 1,150 m a.s.l. Rainfall patterns are bimodal, most rain occurs between March and May, and between September and November, with mean annual rainfall of 1,170 mm (Nsubuga, Olwoch, & Rautenbach, 2011). The mean daily minimum temperature in the Kampala region is 17.6°C, and the mean daily maximum temperature is 27.8°C (WMO, 2018).
Two study sites were selected at MUARIK (Figure 2

| Use of host plants
Data on the host plants used by R. differens was collected by walking along a pre-defined trail, established at both study sites, and starting at a random point on the trail at the beginning of each census. On each sampling night the study site was censused for 3.5 hr between 7:30 and 11:00 p.m. All observations were made by the same observer (R.O.) using binoculars and an overhead light. Each study site was visited between one and six nights per month (in 83% of the cases between three and five nights per month) for over 15 months between November 2015 and January 2017. Whenever R. differens was spotted, the plant species, and the sex, developmental stage, and colour morph of the insect were recorded. For the individuals which were observed feeding (67% of total observations), the plant part (inflorescence, leaf, stem, or inflorescence stalk) being eaten was also recorded. Species identification of host plants was ensured by collecting samples of plants and identifying them at the herbarium of Makerere University following Clayton, Phillips, and Renvoize (1974) and Katende, Birnie, and Tengnäs (1995). The sex of individuals was only determined starting from 4th instar because these stages have developed sex characteristics (the presence of ovipositor can be used to identify females). Developmental stages were categorised as 1st to 6th instars for males and 1st to 7th instars for females (only females have a 7th instar) depending on body length (Brits & Thornton, 1981).

| Host plant cover and number of inflorescences
To determine the availability of host plants in the field, the leaf cover and number of inflorescence of grasses and sedges (the potential host plants) were measured at both study sites during the peak of the first (April-May) and second (November-December) annual rainy seasons in 2016. We established parallel, 150 m long transects at distances of 50 m at both study sites. Along each transect line, 1.8 m radius circular plots were established every 50 m. Whenever the plot extended into trees, shrubs, or herbs only, the plot was relocated by moving it 10-15 m forward into grasslands (thickets of shrubs and trees were excluded because R. differens are mostly grassland dwellers; Bailey & McCrae, 1978). This method produced a total of 58 plots (24 plots at Site 1 and 34 plots at Site 2).
Each plot was divided, through its centre, into four equal sections. In each section, the percentage leaf cover of every grass and sedge species encountered (16 species) and herbs (pooled) and trees (pooled) was visually scored using the scale: 0%, 1%, 2%, 3%, 4%, 5%, 6%, 7%, 8%, 9%, 10%, 15%, 20%, 25%, …, 95%, 100%. Furthermore, the number of inflorescences of all grass and sedge species was counted. To minimise the risk of subjective errors when estimating leaf cover, the same person (R.O.) conducted the estimates in both the study sites. For each plot, we calculated the average of leaf cover (or number of inflorescences) across the four sections and across studied times and used these values in statistical analysis.

| Statistical analysis
We decided to use only leaf cover as an indicator of host plant availability in the statistical analysis, because we considered it to be a F I G U R E 2 Location of the study. Site 1 (S1) and Site 2 (S2)  Chi-square goodness-of-fit tests (for the two study sites separately) were conducted in order to test if the grass and sedge species are used by R. differens in the same proportions as they occur in the field. For this, we calculated the observed frequencies for each of the seven most commonly used grass and sedge species, as well as for the rest of the grass and sedge species (categorised as "Rare combined"). The observed frequencies were compared to the proportions of these species in leaf cover data (summing to 1); assuming these species (and not herbs and trees) form the available resource pool for R. differens. We decided to keep the analysis of the two sites separate because they had different plant community compositions.
We fitted a generalized linear model (  were most frequently observed on the most common grass at the site ( Figure 3). However, the grasses were not used in the same proportions as they occurred in the field (Chi-square goodnessof-fit tests; Site 1: χ 2 = 13,284, df = 7, p < 0.001; Site 2: χ 2 = 4,335, df = 7, p < 0.001). At both study sites, P. maximum, and S. pyramidalis were used more frequently than expected from their leaf cover in the field (Figure 3).

| Host plants and their parts
The R. differens were predominantly observed feeding on inflorescences (97% of 14,193 feeding observations) and much less frequently on leaves (3.0%), stems (0.1%), and inflorescence stalks (0.1%) of grasses and sedges.  At both Site 1 (59% of observations) and Site 2 (56%), females were observed more commonly than males (Chi-square goodnessof-fit tests; Site 1, χ 2 = 332.8, df = 1, p < 0.001; Site 2, χ 2 = 100.1, df = 1, p < 0.001). Of those developmental stages where it was possible to determine sex, females represented 49.4% of medium nymphs, 77.4% of large nymphs and 55.7% of adults (data pooled for the two study sites). The result of the generalized linear model also suggests that different colour morphs used hosts differently, but this depended on the study site ( Figure 6). At both sites, proportionally more brown individuals were observed on B. ruziziensis than green individuals.

| D ISCUSS I ON
Our results show that, in the field, non-swarming R. differens occur on a wide range of grass species, but obviously selectively. Availability of host plants seems to determine the pattern of host use in the field to some degree, since at both study sites the majority of R. differens were observed on the most common grass of the site. Yet, some host plants turned out to be more used than their abundance alone would predict. Selectivity in host plant use by R. differens has also been demonstrated experimentally . Among the common host plants found in our study, R. differens occurred more frequently on P. maximum than its cover alone would predict. Based A wide range of potential host plants for R. differens provides a good possibility for diet mixing (Bernays & Bright, 1993;Malinga et al., 2018b). The wide range of host plants utilized by R. differens has been observed earlier in wild populations (Bailey & McCrae, 1978;Swaine, 1964) and multiple-choice experiments in the laboratory have shown that R. differens individuals feed on several host grasses when offered together . Diet mixing is associated with better performance in R. differens, including shorter developmental time and higher adult fresh weight and female fecundity (Malinga et al., 2018b). Diet mixing is considered to be beneficial for Orthopterans due to improvement in the balance of nutrients ingested or the dilution of toxins (Bernays & Minkenberg, 1997;Miura & Ohsaki, 2004;Unsicker et al., 2008).
It is evident that non-swarming population of R. differens predominantly use grass inflorescences in the field (97% of feeding observations). This is also consistent with early observations made in East Africa (Bailey & McCrae, 1978, and references therein).
Swarming R. differens are also occasionally recorded as pests on the developing seeds (at the milk stage) of several cereal crops (Swaine, 1964). In the laboratory, R. differens prefers grass inflorescences or seeds over stems or leaves (Hartley, 1967;Valtonen et al., 2018).
Ruspolia differens' nearly exclusive use of grass inflorescences and developing seeds as food possibly helps them to meet the energy requirements for their exceptional swarming behaviour (Bailey & McCrae, 1978). Flowers and grains are richer in protein, when compared to the leaves of grasses, and therefore generally more nutritious for herbivores (Bernays & Chapman, 1994;Roulston, Cane, & Buchmann, 2000). Thus, inflorescence and grain feeding may not only be necessary for survival, growth, and reproduction (Joern & Behmer, 1997) in non-swarming populations but also for amassing the fat reserves required to maintain flight over considerable distances during swarming (Bailey & McCrae, 1978;Karuhize, 1972).
Female and male R. differens differ to some degree in their host choice (females were found more frequently on B. ruziziensis than males), possibly due to differences in nutritional requirements   (Behmer & Joern, 1994;Boys, 1978;Unsicker et al., 2008), or this may reflect differential use of host plants for protection from predation. The sex ratio of R. differens was female-biased, which could be mainly explained by the slower development of female nymphs (77% of large nymphs were females, as only females have the final 7th instar; Brits & Thornton, 1981). Furthermore, there could also be higher mortality of males in the large-nymph stage. The slightly higher proportion of females among adults (56%) could be at least partly explained by higher recruitment rate of males to swarms, as swarms are typically male-biased (Bailey & McCrae, 1978 (Lawton, 1983). The different developmental stages might also vary in their nutritional requirements or physical capacity to use hosts (Hochuli, 2001;Werner & Gilliam, 1984). For example, young nymphs with small mandibles might be unable to consume tough hosts that adults can cope with (Hochuli, 2001). Also, feeding on tough grasses has been associated with large heads among Orthopterans (Bernays & Hamai, 1987). camouflage (Bailey & McCrae, 1978;Valverde & Schielzeth, 2015).
However, for the strictly nocturnal nymphs, the role of colour morph in improving their survival in dim-light conditions should be explored in further studies (see e.g., Meyer-Rochow & Teh, 1991). The proportions of colour morphs are not static, as green morphs of R. differens increase in proportion after rainy seasons, when vegetation greens up (Matojo & Yarro, 2010;Opoke et al., 2018). The colour morph dynamic is a two-way process, green nymphs are known to change to brown nymphs and vice versa during moulting (Robinson & Hartley, 1978), possibly induced by the changes in the environment.
Overall, the results of this study provide useful insights for the management and conservation of R. differens, which have hitherto  should be a priority target for conservation in the face of current grassland degradation.

ACK N OWLED G EM ENTS
This research was funded by Academy of Finland (Project no 14956 to HR) and UniPID (FinnCEAL+ grant to AV). We thank U. Nsereko for assistance with data collection, the two anonymous referees for their valuable comments, the Office of the Ugandan President and the National Council for Science and Technology for permissions to conduct this study, and the Makerere University Agricultural Research Institute, Kabanyolo, Uganda for hosting the project.

CO N FLI C T O F I NTE R E S T S
None declared.  HR and AV wrote the manuscript. HR secured funding. All authors read and approved the manuscript.

DATA ACCE SS I B I LIT Y
Data is available from the Dryad Digital Repository: https://doi. org/10.5061/dryad.f845h1n.