Introduction

The sustainable intensification of agriculture is essential to achieve Sustainable Development Goal (SDG) 2, zero hunger, and reduce hidden costs of meeting requirements of other SDGs such as 13, climate action (FAO et al., 2020). This is particularly important in the developing Global South where larger yield gaps are observed compared to the developed countries with more financial resources (David Tilman et al., 2011). Intercropping may offer sustainable solutions with potential for strong positive effects on pest and disease control as well as associated biodiversity (Beillouin et al., 2021).

The concept of “push-pull” mixed cropping was first reported in 1987 as a “stimulo-deterrence” strategy for the reduction of pest damage in cotton, combining a pest repellent intercrop (“push“) interspersed with the main crop, and an attractant border crop (“pull“) to trap pests at the field perimeter (Bruce et al., 1987). As most widely practiced, current push-pull systems benefit from the attractive and repellent stimuli of perennial inter- and border crops (referred to as companion crops), thereby reducing or eliminating the need for pesticides (Pickett et al., 2014). In Kenya, the first push-pull system, introduced in 1997, was designed to reduce yield losses caused by stemborer species and comprised a combination of the main crop maize (Zea mays L), the intercrop molasses grass (Melinis minutiflora), and the border crop Napier grass (Pennisetum purpureum) or Sudan grass (Sorghum vulgare sudanense) (Khan et al., 1997b).

Several volatile substances emitted from molasses grass repelled female stem borers and simultaneously attracted their parasitoid Cotesia sesamiae (Khan et al., 1997a). Following the initial success of this system, there was interest in replacing molasses grass with a leguminous repellent intercrop that could improve soil quality by fixing nitrogen, and provide high-quality fodder. The legumes silverleaf Desmodium (Desmodium uncinatum) and its congeneric (Desmodium intortum) were shown to repel ovipositing stem borers and furthermore to suppress parasitic Striga weeds. (Khan et al., 2000, 2002). The first Desmodium-intercropped version of push-pull used D. uncinatum in place of M. minutiflora together with P. purpureum. As a result of screening for more drought-tolerant companion plants, a version using D. intortum combined with the border crop Brachiaria cv Mulato II was developed as “climate-adapted push-pull“, which has been demonstrated to increase maize yield by a factor of 2.5 (Midega et al., 2015). Recently, a “third generation” push-pull system was evaluated using the intercrop D. incanum to replace D. intortum – which does not flower and set seed in many parts of tropical Africa and thus limits adoption and spread of push-pull by limiting seed supply for the intercrop – with Brachiaria brizantha cv Xaraes as border crop, which is more resistant to herbivorous mites that attack the mulato cultivar (Cheruiyot et al., 2021).

The invasive fall armyworm (FAW), Spodoptera frugiperda, originating from a maize-specialized strain from northern and central America, has become a major threat to African maize crops since 2016 (Goergen et al., 2016; Day et al., 2017; Hailu et al., 2021; Zhang et al., 2023). It spread rapidly through East Africa, with most of the farmers in Ethiopia and Kenya encountering the FAW after the long rains in the first half of 2017 (Kumela et al., 2018). The invasiveness of the moth is fueled by its relatively short life cycle of about four weeks and the capability of the adult females to lay hundreds of eggs (Sparks, 1979). Other traits favoring the FAW’s spread in Africa and beyond are its strong flight capacity, lack of diapause, reported survival in diverse habitats, rapid development of resistance to insecticides / viruses, and the polyphagous nature (Wan et al., 2021). Although its wide host range of at least 353 host species across 76 plant families (Montezano et al., 2018), the high financial and yield losses on the main crop maize due to the pest are particularly devastating (Eschen et al., 2021). The FAW is reported to outcompete resident stemborers (Hailu et al., 2021; Sokame et al., 2021; Mutyambai et al., 2022). The use of pesticides is a popular approach to control the FAW, posing risks to environmental and human health, which includes acute pesticide-related illnesses as many smallholder farmers do not use personal protective equipment while spraying (Tambo et al., 2020). Therefore, sole reliance on pesticides is not sufficient to manage FAW sustainably, and Integrated Pest Management (IPM) strategies, such as the promotion of natural enemies, are desirable (Nyamutukwa et al., 2022; Van den Berg and du Plessis, 2022). Climate-adapted and third-generation push-pull systems have been reported to reduce plant damage and yield loss caused by the FAW (Midega et al., 2018; Hailu et al., 2018; Cheruiyot et al., 2021; Yeboah et al., 2021; Mutyambai et al., 2022).

A systematic review on the chemical ecology of push-pull systems by Lang et al., (2022)) found only thirty publications (seven primary sources, and 23 publications reporting on results from these primary sources) on the chemistry of push-pull and related mixed-cropping systems, from which 206 compounds were reported to be potentially associated with push and pull effects. Of these, 101 were categorized as plant volatiles and reported by studies with a focus on plant-insect interactions (as opposed to studies with a focus on Striga control). However, none of these publications reported volatiles sampled under field conditions. Two recent papers, both published after the literature review conducted by Lang et al., (2022)), reported additional potentially bioactive compounds from companion plants in maize-Desmodium-grass push-pull systems, and both publications supported the hypothesis that volatiles from the companion crop D. intortum repel the FAW and attract parasitic wasps (Sobhy et al., 2022; Peter et al., 2023). In contrast, a third recent publication by Erdei et al., (2024) detected low levels of volatiles from D. intortum plants and found no consistent preference of FAW for maize over either D. intortum or maize together with D. intortum, and concluded that the protective effects of the intercrop were more likely to result from physical trapping of FAW larvae.

Thus, further investigations are needed to understand the mechanisms by which the maize-Desmodium-grass push-pull system can protect maize from FAW. A summary of the studies to date on push-pull effects on the FAW can be found in Figure 1. To our knowledge, no data on plant volatiles from the more drought-resistant D. incanum have been published so far. Therefore, this paper focuses on the volatiles of D. intortum and D. incanum and the effects of exposure to the headspaces of these plants, versus direct exposure to the plants, on the invasive pest FAW. We collected volatiles from plants in farmers’ fields as well as in bioassay setups under semi-field con-ditions, and conducted oviposition and flight tunnel bioassays to assess FAW moth preferences.

Figure 1.

Results

Desmodium Volatile Profiles

A total of 25 substances were measured in at least 2/3 of field-collected samples from each of the Desmodium species, of which 11 occurred in both species. These substances include (E)-β- ocimene, (Z)-3-hexen-1-ol, (Z)-3-hexen-1-ol, acetate, 1-octen-3-ol, 3-octanone, caryophyllene, (3E)-4,8-dimethyl-1,3,7-nonatriene (DMNT), germacrene d, (3E,7E)-4,8,12-Trimethyl-1,3,7,11-tridecatetraene (TMTT), and five other peaks that were not fully identifiable. The unidentified peaks were cate-gorized based on mass spectra and relative retention times as a benzene derivative, a naphthalene derivative, a monoterpene, and two sequiterpenes (for feature information, see Table 1. The sesquiterpene (E)-α-bergamotene only occurred in D. incanum, while another unidentified sesquiterpene only occurred in D. intortum (see Figure 2 and for a per-sample heatmap Figure 2—figure Supplement 1). Of these substances, (Z)-1,5-octadien-3-ol, 1,2,3-trimethylbenzene, 3-pentanol, and hexyl acetate are here for the first time reported in connection with the chemical ecology of pushpull cropping systems (Lang et al., 2022; Sobhy et al., 2022; Peter et al., 2023). Germacrene D and 1-hexanol were not yet reported from Desmodium species, but were reported in relation to mixed cropping for pest control. Germacrene D was detected in elevated levels in maize plants grown in soil from push-pull fields in comparison with those grown in soil from maize monocultures (Mutyambai et al., 2019), while germacrene D and 1-hexanol both were reported to be emitted from bean plants, Vicia faba, and used by the black bean aphid, Aphis fabae, for host detection (Pickett and Khan, 2016; Webster et al., 2008).

Table 1.

Figure 2.

Figure 2—figure supplement 1.

Oviposition Bioassays

Oviposition choice bioassays were conducted to determine the influence of direct or indirect contact with Desmodium plants on the oviposition behavior of FAW moths. A fraction of the potted Desmodium plants used in the oviposition assays was sampled for volatiles the night before the start of trials, and the composition of these samples was compared with the substances and features detected from Desmodium plants sampled in farmers’ fields. In the comparison between field Desmodium plants, and potted bioassay plants that were sampled in a smaller number, the majority of substances showed a higher relative abundance in potted D. incanum (bioassay conditions), whereas in D. intortum, some substances occurred in higher concentrations in the potted plants, while others were not detected. Sixteen of the volatiles detected in Desmodium plants in farmers’ fields were also present in either uninfested or infested maize plants, with 14 substances overlapping (see Figure 3).

Figure 3.

The count of eggs and egg batches showed that fewer eggs were laid directly on Desmodium plants, with the egg count on maize plants being 7.9 or 6.8 times higher than on D. incanum or D. intortum, respectively (see Figure 4 and Figure 4—figure Supplement 1). A mixed model with egg counts as the observed variable and accounting for plant position detected a significant difference between the control and all other treatments (F(1) = 9.21, p = 0.003), as well as for the comparison of the indirect and the direct treatments (F(1) = 9.32, p = 0.003). However, no significant difference was found between the two Desmodium species (F(1) = 1.01, p = 0.32) (see Table 3).

Figure 4.

Figure 4—figure supplement 1.

Figure 4—figure supplement 2.

Figure 4—figure supplement 3.

Figure 4—figure supplement 4.

Figure 4—figure supplement 5.

Table 2.

Table 3.

Wind Tunnel Bioassays

Wind tunnel bioassays were conducted to determine short-term effects of plant volatiles on the flight behavior of FAW moths.

Two-thirds of the moths tested showed ten or fewer segment changes during the five minutes of each experiment. Across all treatments, at least 77% of the moths showed no movement in the last 2 min of the experiment or only little activity, with a maximum of two segment changes in the last 3 min of the experiment, which was interpreted to mean that they had made a decision within the five-minute timeframe of the experiment. Overall, moths showed a reduction in segment changes over time, particularly for the comparison of maize vs. maize + D. intortum (see Figure 5). Across all treatments, a few individuals displayed high flight activity of up to 76 segment changes with no indication of decline over time.

Figure 5.

Figure 5—figure supplement 1.

No significant effects were detected when comparing the stays in the maize segments versus the Desmodium segments across the treatments using a mixed model (see Table 4). However, a preference for the side with maize is apparent for the comparison of maize vs. maize + D.intortum, especially when comparing the mean stay duration as well as the final settling location of the moth, whereas moths showed a slight preference for maize + D. incanum instead of maize alone (see Figure 5). Fourteen of nineteen moths exposed to airflow from maize vs. maize + D.intortum settled in a segment of the wind tunnel within the first three minutes of the experiment, a larger proportion than in the maize vs. maize control or the maize vs. maize + D. incanum treatment.

Table 4.

Discussion

The fall armyworm (FAW) Spodoptera frugiperda invaded sub-Saharan Africa in 2016 and is responsible for major yield losses in maize crops (Day et al., 2017; Goergen et al., 2016). Push-pull cropping of maize with Desmodium and border grasses is reported to reduce FAW damage (Midega et al., 2015; Cheruiyot et al., 2021; Mutyambai et al., 2022; Yeboah et al., 2021; Hailu et al., 2018) and to increase maize yield (Midega et al., 2018; Cheruiyot et al., 2021; Yeboah et al., 2021). Push-pull cropping systems benefit from the attractive and repellent stimuli of perennial companion crops (Pickett et al., 2014), with more than one hundred plant volatiles from different classes, especially terpenoids and fatty acid-derived “green leaf” volatiles, reported to be associated with push-pull effects (Lang et al., 2022). Nevertheless, the repellent role of volatiles from greenleaf Desmodium (D. intortum) has been investigated in only three studies and almost exclusively in potted plants, with conflicting reports of whether these volatiles repel FAW moths, and the only study reporting field measurements recovering few volatiles in relatively low abundance (Sobhy et al., 2022; Peter et al., 2023; Erdei et al., 2024). Here, we characterized volatiles sampled from both D. intortum (greenleaf Desmodium) and D. incanum growing in farmer fields or in pots, and conducted bioassays with FAW moths to determine whether the presence of Desmodium plants or their volatiles would alter moth preferences for maize. We identified up to 17 - 19 substances in headspace samples from Desmodium plants, often in abundance similar to or greater than volatiles measured from maize plants under comparable conditions, and we observed that FAW moths preferred to oviposit on maize rather than on Desmodium plants of either species. However, proximity to a Desmodium plant sharing the airspace did not influence FAW oviposition preferences for maize in a greenhouse cage choice assay, nor did FAW moths show a significant preference for maize volatiles over volatiles from a combination of maize and Desmodium plants in a wind tunnel choice assay. We note that these assays were performed in a risk area for malaria and other mosquito-borne illnesses, and so experimenters were protected with DEET, which has low volatility but is active in the vapor phase, where it can cause host avoidance by binding to insect olfactory receptors (although this has been mostly studied in the dipterans Aedes aegypti and Drosophila melanogaster) (DeGennaro, 2015). We reduced any potential influence on the lepidopteran in our experiments by using minimal amounts of DEET, avoiding direct contact of any DEET with experimental materials, and conducting behavioral assays without the insects being exposed to volatiles from the experimenter (experimenters were generally not present during oviposition assays, and the insects were in an isolated airspace during wind tunnel bioassays). We conclude that Desmodium plants in push-pull cultivation emit known bioactive volatiles, but cannot confirm that these volatiles alone repel FAW moths based on our bioassays, and therefore suggest that other features of the maize-Desmodium-border grass push-pull system, including the influence of border grasses as well as belowground interactions and other repellent traits of Desmodium, may be important for protecting maize from FAW moths and their larvae.

Emission of Potentially Bioactive Volatiles by Desmodium Intercrops

The headspace of Desmodium plants was dynamically sampled on Tenax TA adsorbent directly in push-pull fields, as well as from potted plants used for oviposition bioassays, overnight.

All substances occurring in fewer than 2/3 of the field samples were excluded from further analysis, which represents an arbitrary limit. Quantitative comparisons within single substances are possible, but must be handled with care, as normalization based on the biomass and leaf area of the plants was not possible. The chosen adsorbent, Tenax TA, is known to be effective for lipophilic to medium-polarity organic compounds of intermediate molecular weight (ca. C7-C26) (Dettmer and Engewald, 2002; Tholl et al., 2020). It is possible that headspace samples changed during storage (up to 3 months tightly closed at room temperature) as degradation or thermal rearrangements on Tenax TA material have been reported before (Dettmer and Engewald, 2002; Alborn et al., 2021). We extracted headspace samples using thermal desorption, which is simple and sensitive, but not appropriate for some compounds, such as the sesquiterpene germacrene A, which are susceptible to thermal rearrangement or degradation (Tholl et al., 2006).

Despite these caveats, our samples yielded a variety of volatiles with high signal:noise ratios, of which 13 had been previously reported from sampling the headspace of potted D. intortum plants onto Porapak Q for sampling durations of 24 h or 48 h followed by solvent extraction and GC-MS analysis (Sobhy et al., 2022; Peter et al., 2023). In contrast, Erdei et al., (2024) detected fewer volatiles in lower relative abundance from D. intortum using solid-phase microextraction (SPME) and a saturation time of 18 - 24h to sample intact, mechanically damaged or herbivoryinduced potted plants or intact plants in Tanzanian and Ugandan fields. Our study is the only one of these that limited the window of sampling strictly to the nighttime hours corresponding to the reported activity window for the FAW(Sparks, 1979). In pretests no shorter activity window could be determined as the moths still showed mating behaviour and potential oviposition at around midnight (see Appendix 1). Differences might be explained due to the different use of sampling techniques, as the sensitivity of SPME is lower in comparison to dynamic headspace collection (Tholl et al., 2006). To our knowledge, the volatile profile of D. incanum was not previously reported. In conclusion, the hypothesis that D. intortum and D. incanum emit potentially bioactive volatiles is supported.

Oviposition Choice Bioassay

FAW oviposition behaviour was observed over three nights in two treatments per Desmodium species. In one treatment, FAW moths had direct contact with Desmodium plants and in another, they had only indirect contact through aerial exposure to plants outside of, but directly next to a mesh cage. Moths always had direct contact with maize plants.

Critical consideration must be given to differences in the volatile profiles between the potted plants and those measured in farmers’ fields, which led to a separation of potted and field D. intortum in the second dimension of a PCA, and a near-separation of D. incanum in the first dimension. As the untargeted analysis was conducted based on the field Desmodium samples, any volatile substances that only occur in the potted plants were not identified or analyzed further (full results of GC-MS analysis are available with the source data for this publication). Oviposition on maize can alter the maize volatile profile and affect volatile emission in the following days, which is why the maize samples are divided into plants before and after (infested) oviposition exposure. Of the substances occurring in field Desmodium plants, approximately half were abundant in healthy and infested potted maize plants, with only one substance, (E)-α-Bergamotene, being present solely in infested maize plants. The sampling of potted plants was conducted using the same procedure as in the field, but in a greenhouse with netted walls on two sides that allowed air circulation as for the oviposition bioassays. Overall, this highlights the importance of studying oviposition cues and moth preferences directly in push-pull fields in the future.

In all treatments, the number of eggs laid per batch showed no difference depending on the position, which indicates that FAW moths did not adjust their behavior once they started laying eggs. This differs slightly from the outcomes reported by Peter et al., (2023), where a reduction in egg number per batch was observed and the number of eggs, but not the number of egg batches, differed between maize vs. maize + D. intortum. However, a clear preference for maize compared to the two Desmodium species was observed in the treatments where FAW moths had direct contact with both plant species, with a large effect size of 7 – 8 times more eggs on maize than on Desmodium. The preference of maize over D. intortum in oviposition assays with direct contact to both plants is consistent with various publications using similarly controlled setups (Sobhy et al., 2022; Peter et al., 2023; Erdei et al., 2024). There was little oviposition on the cage in our bioassays, which indicates that moths primarily chose between plants in our setup, but differs from the strong preference found by Sobhy et al., (2022) of FAW moths to oviposit directly on the mesh of a bioassay cage rather than on the plants when offered maize co-planted with D. intortum. Desmodium and maize plants were planted in the same pots by Sobhy et al., (2022), and therefore maize volatiles may have been affected by belowground interactions or through priming by Desmodium volatiles prior to or during oviposition bioassays. This is in fact closer to the configuration in push-pull fields, but does not isolate the influence of the Desmodium headspace, as we sought to do here.

In our bioassays, no significant effect could be seen in the treatments where moths had only indirect contact with Desmodium volatiles coming from a Desmodium plant proximate to one maize plant, but outside of the cage. This is not consistent with the hypothesis that Desmodium volatiles repel FAW moths as proposed (Pickett et al., 2014; Sobhy et al., 2022). It is possible that the distance of about 60 cm between the plants in the cages was not suitable to create a gradient of volatiles sufficient to allow moths to choose between maize plants closer to versus further from Desmodium, in which case neighboring treatments with distances of 50 – 100 cm could have also affected each other. To our knowledge, the only comparable setup was reported by Erdei et al., (2024), in which two maize plants were exposed to volatiles of either D. intortum or an artificial plant in a modified wind tunnel, and moths did not show a significant preference for either maize plant.

In conclusion, we found a preference of FAW for maize over D. intortum and D. incanum, but this preference could not be solely attributed to Desmodium volatiles in our experiments. It is not to be expected that herbivores use all potential host plants equally, and preferences among host plants or for host over non-host plants can be based on different plant traits. FAW moths may also prefer maize over Desmodium due to other traits such as the surface texture, as proposed in the findings from Erdei et al., (2024) that silicate trichomes on D. intortum pose a danger to FAW larvae; or chemical repellents directly on the leaf surface. Belowground interactions or priming between Desmodium and maize plants co-planted in push-pull fields may also change moths’ preference for maize, an effect which could not have been observed with our experimental design, but might explain inconsistencies between our findings and those from Erdei et al., (2024) versus those from Sobhy et al., (2022).

Wind Tunnel Bioassay

The short-term flight behavior of FAW moths was tested in a two-choice wind tunnel setup where volatiles of a maize plant vs. maize + Desmodium were compared against each other. Flight behavior showed large variation among the individual moths, but three-quarters of the moths had settled after at least three minutes or showed fewer than three segment changes in the last two minutes of the experiment. This experimental setup has limited space and duration, and moths may behave differently over longer times or when given more space to maneuver, as under normal field conditions. Furthermore, the quarter of the moths that showed high activity indicate the difficulty of quantifying the responses of such active fliers. Interestingly, a larger proportion of moths settled after 2 minutes when presented with maize versus maize + D. intortum, rather than continuing to fly. The moths showed a weak preference for the volatiles of the single maize plant over the volatiles of maize + D. intortum, whereby they slightly preferred volatiles from D. incanum + maize over maize alone but did not distinguish between volatiles from two maize plants (negative control). However, none of the treatments showed any significant preference. In contrast, repulsion of moths by D. intortum volatiles with or without the addition of maize was observed in no-choice assays when the total distance flown upwind at the end of the experiment was used as a metric (Sobhy et al., 2022; Peter et al., 2023). Based on these wind tunnel results, it cannot be excluded that the exchange of D. intortum with D. incanum in third-generation push-pull fields may reduce protection against FAW moths. However, in direct oviposition choice assays (in which moths were enclosed in a cage with one maize and one Desmodium plant), maize was preferred equally over either Desmodium species.

Conclusion

Several field studies have described positive effects of push-pull systems with the companion crops D. intortum or D. incanum on maize yield (Midega et al., 2015, 2018; Cheruiyot et al., 2021). Here, we show that 17 - 19 volatile substances that have been previously shown to affect pest host choice were found in the headspaces of both companion crops D. intortum and D. incanum within the activity window of FAW moths. However, we did not observe a clear repellent effect of Desmodium volatiles on the FAW in our bioassays. FAW moths clearly preferred to oviposit on maize over D. intortum or D. incanum plants. Our results thus indicate that moths prefer maize over Desmodium and that this may be influenced by short-range mechanisms such as an unfavorable leaf surface of Desmodium. However, the main interest in the context of push-pull is how Desmodium contributes to protecting co-planted maize. A reduction of FAW damage on maize under push-pull cultivation has been demonstrated in multiple studies (Midega et al., 2018; Hailu et al., 2018; Cheruiyot et al., 2021; Yeboah et al., 2021; Mutyambai et al., 2022). Based on the literature, the pull effect of border crop volatiles may be important for reducing FAW damage in push-pull fields, in combination with the attraction of parasitoids by Desmodium volatiles and unfavorable host qualities of Desmodium plants for FAW (Figure 1). Furthermore, the Desmodium intercrop may be equally or more important for weed suppression, soil fertility, and possibly improving maize plant health and rigor as for volatile-mediated repellence of FAW in the push-pull system.

Methods and materials

All volatile sampling and bioassays were performed with two Desmodium species, Desmodium intortum (greenleaf Desmodium) and Desmodium incanum (often referred to as African Desmodium by the push-pull farmers) and in the method section referred to as Desmodium for simplicity. All raw data and code used for the statistical analysis can be found on Zenodo (CERN, Geneva, Switzerland, https://doi.org/10.5281/zenodo.11633890) and Github (github.com/ DariaMOdermatt/DesmodiumVolatilesinPush-Pull).

Pretest Moth Activity Window

The determination of the activity window of the FAW was relevant for the choice of the duration of the volatile sampling. The FAW is reported to be nocturnal (Sparks, 1979), without report of a more precise activity window. For the first cycle, two single maize leaves of the landrace ‘Jowi white’ were placed in wooden cages with meshed side walls (100 x 60 x 60 cm) in high cylindrical glasses filled with water, but sealed with a piece of cotton to prevent moths of entering the glasses. Moths of the age of 3 - 4 days were released in pairs of one female and one male in 10 different cages (resulting in 20 moths in total) and observed from 7pm - 11.30pm in two consecutive nights. The moths were recollected in falcon tubes in the morning after the first night, sealed with a piece of cotton to allow air circulation and stored at a temperature of approx. 25 °C until they were released in the same combinations on the following evening. For the second night fresh maize leaves were provided for the moths to oviposit. For the second cycle, potted maize plants with multiple leaves of the landrace ‘Jowi white’ were placed inside the wooden cages with a piece of cotton soaked with water. Three female and two male moths of different ages were placed each in three cages (resulting in 15 moths in total) and observed between 6pm - 1am in two consecutive nights. During day time the moths and the plant were left unaltered in the cage.

Volatile Sampling

Volatiles of Desmodium were collected from plants in push-pull fields of local smallholder farmers near Mbita in the counties Homa bay and Siaya (coordinates determined with WGS84 coordinate system (latitude, longitude): D. incanum = (-0.382096, 34.175487), (-0.4298279, 34.207891), (0.189957, 34.36072), D. intortum = (-0.382096, 34.175487), (-0.6302736, 34.494595), (-0.551479, 34.314673)). Desmodium volatiles were collected on three different push-pull fields in May and June 2023 during the long rains, selecting four representative plants per field. As it is becoming more common, some of the push-pull fields were mixed with vegetables such as kale or cowpeas. Volatile collections were also performed from potted Desmodium plants, as well as from healthy and infested potted maize plants on the research campus. Five three-week-old maize plants, later labeled as“infested”, were exposed in two cages, each with each three female and two male moths, for two nights prior to the volatile sampling, which resulted in 1 to 10 batches per plant. Potted plants were placed in a greenhouse with a glass roof and two glass walls, in which air was able to circulate through netted walls on the opposing longer walls of the greenhouse.

As an adsorbent, commercially available Tenax TA glass tubes (containing 150 mg adsorbent, 35 – 60 mesh, Markes International Ltd, England or 186 mg adsorbent, 60 – 80 mesh, Merck, Germany) were used. A handful of leaves were enclosed in a preheated roasting bag (Sainsbury’s, London, UK). The inlet tubing pushed 500 mL/min charcoal-filtered air into the bags from the lower bag rim, while the Tenax TA tube was connected to an air outlet tubing through a small hole at the top of the bag. Air flows were regulated by PYE Volatile Collection Kits (B. J. Pye, Kings Walden, UK) as described in Steen et al. (2019). Volatiles were accumulated on the adsorbent by aspirating the air at 200 mL/min (+/- 40 mL/min) for 11.75 h (+/- 30 min) overnight from 7 pm until 6.45 am of the following day. Under the same conditions, four ambient controls were sampled with an empty roasting bag on two fields with each Desmodium species. Additionally, a Tenax TA storage control was stored with all tubes, but never taken to the fields.

TD-GC-MS Measurement

All samples were measured on a TD-GC-MS instrument (Thermal Desorption – Gas Chromatography – Mass Spectrometer) from Shimadzu (TD30-QP2020NX Kyoto, Japan). The Tenax TA tubes were desorbed for 15 min at 220 °C under 80 mL/min nitrogen flow. Volatiles were trapped on a Tenax cooling trap at -20 °C and after desorption was completed, the cooling trap was heated rapidly to 230 °C. The sample was injected with a split ratio of 1:5 and separated on a Rtx-Wax column (30 m, 0.25 mmID, 0.25 μm, Restek Corporation, USA) with the following 35 min oven program: holding at 40 °C for 2 min, heating to 150 °C with 10 °C/min, holding at 150 °C for 2 min, heating to 190 °C with 3 °C/min, heating to 230 °C with 10 °C/min and holding at 230 °C for 3 min. Mass fragments from 33 m/z to 400 m/z resulting from electron ionization at 70 eV were recorded with a scan rate of 5 Hz at an ion source temperature of 230 °C from 3 – 30 min.

Feature Detection

A feature list was generated in MZmine (Version 3.9.0, (Schmid et al., 2023)) from all compounds that occurred in at least one-third of the field samples of either Desmodium species and in less than 4/5 of the combined control samples (ambient control and storage control). Features were numbered and considered as unidentified if there was no corresponding reference compound. After removal of contaminants, the features were added to an existing target list that was composed with commercially available plant volatiles that can be mainly classified as terpenoids and benzenoids, which as a whole contain a large variety of functional groups such as alcohols, aldehydes, ketones and esters. Additionally, the existing list included four green leaf volatiles ((E)-2-hexen-1-ol acetate, (E)-2-hexenal, (Z)-3-hexen-1-ol, (Z)-3-hexen-1-ol acetate) and a selection of more specific substances reported in push-pull fields such as (E)-4,8–dimethyl–1,3,7-nonatriene (DMNT) and (E,E)-4,8,12-trimethyl-1,3,7,11-tridecatetraene (TMTT). Integration of all substances and features was manually checked for the all control samples (ambient and storage control) using the LabSolutions Insight GCMS software (Shimadzu corporation, Kyoto, Japan). Peaks for all targets that occurred in at least 1/2 of the Desmodium field samples and in no more than 3/5 of controls were manually checked for all samples. Eventually, all targets that occured in at least 2/3 of the field samples of D. intortum or D. incanum and in no more than 3/5 of the controls were considered as present. In Table 1 the electron impact ionization (EI)-MS data of the unidentified features are displayed, and in Table 2 the origin of all reference standards used to identify the target hits can be found.

Oviposition Bioassays

Plants

All plants were planted in plastic pots in black cotton soil in Mbita, Kenya. Maize plants (SC Duma 43, Seed Co Limited, Nairobi, Kenya) were grown from seeds with the addition of fertilizer (3 g DAP 18-46-0, Yara East Africa LTD, Nairobi, Kenya) and used at the age of 2 – 4 weeks with five to seven fully-grown leaves. Desmodium plants were obtained from a push-pull field on the icipe Mbita campus and kept in pots without fertilizer.

Fall Armyworm Moths

To form the FAW colony, wild individuals were collected in the counties Siaya, Kisumu, Migori, and Vihiga in Western Kenya. Larvae were fed on artificial diet based on soy flour, wheat germ, raw linseed oil, mineral mix, sugar, aureomycin, vitamins, agar, methyl parabene, sorbic acid, and calcium propionate (Article No 870-265-3747, Southland Products Inc, Lake Village AR, USA). Moths were fed on water and in rare cases after emergence on 10% honey solution. The colony was occasionally restocked with wild FAW moths. Each day after hatching, moths of both sexes were transferred to a cage to allow mating until the start of the experiment.

Experimental Procedure

Cages (100 x 60 x 60 cm) lined with wooden floors and ceilings and netted walls were placed in greenhouses that allowed air circulation through mesh side walls (for pictures of the real setup, see Figure 4—figure Supplement 2, for exact positioning in the greenhouses, see Figure 4—figure Supplement 3 and Figure 4—figure Supplement 4). Two plants were placed within the cage at the greatest distance possible (approx. 60 cm) in two opposite corners. For the control treatment, two maize plants were placed in the cage, while the direct Desmodium treatment consisted of one maize plant and one Desmodium plant in the cage. In two additional treatments with indirect contact, two maize plants were placed in the cage, while one Desmodium plant was placed outside the cage unreachable for the FAW moths, but in proximity to one maize plant. Three female and two male moths at the age of 4 – 5 days were released approx. 1 h before dusk and allowed to oviposit under natural conditions of L12:D12 for three consecutive nights. Eggs were collected in five different groups depending on the position in which they were laid, either on one of the two plants, close to one of the plants within a maximum of 20 cm distance, or on the cage further than 20 cm from any plant (labeled as “No Decision”). Each treatment was repeated 19 – 21 times in total over five cycles, with up to 5 replicates per treatment per cycle.

Egg Count

Each batch of FAW eggs of was collected with sticky tape to separate the layers and spread all eggs out to one dimension. Each egg batch was taped to white paper and photographed in a UV imaging system (Syngene, Cambridge, England) against UV light coming from underneath the paper. A script for semiautomatic counts in ImageJ (Version 1.54f, National Institutes of Health, USA) was developed and used for counting the exact number of eggs per picture (see Figure 4—figure Supplement 5). The code is available in the data depsitory.

Wind tunnel Bioassays

Plants

All plants were planted from seeds without the addition of fertilizer in plastic pots in standard red soil mixed with manure (ratio of 2:1) at the icipe campus in Nairobi, Kenya. Maize plants (SC Duma 43, Seed Co Limited, Nairobi, Kenya) were used at the age of 4 – 7 weeks with five to eight fullygrown leaves. It was observed that maize plants seemed to grow more slowly in Nairobi than in Mbita, which could be explained by the generally lower temperatures in Nairobi. D. intortum and D. incanum seeds were obtained from a push-pull field on the icipe campus Mbita and were 6 – 8 weeks and 7 months old when used, respectively.

Fall Armyworm Moths

To form the FAW colony, wild individuals were collected in Central Kenya such as in the counties Kiambu, Muranga, and Embu and restocked with wild individuals every three months. Larvae were reared on maize leaves, while moths were only fed with water. The colony was occasionally restocked with wild FAW moths. Each day after hatching, moths of both sexes were transferred to a cage to allow mating until the start of the experiment. At least 2 h prior to the experiment, female moths were separated from males and and placed in the darkened experimental room for adjustment to temperature and humidity.

Experimental Procedure

The five upper leaves of a maize plant were wrapped in a preheated roasting bag (Sainsbury’s, London, UK) with the addition of approximately the same biomass of Desmodium on one side of a two-choice wind tunnel (100 x 30 x 30 cm) and compared to a single maize plant wrapped in the same manner on the other side. Control treatments were conducted with one maize plant wrapped in this way on each side. Air was pushed through activated charcoal filters and into the roasting bag from the lower rim of the bag at a rate of 740 – 820 mL/min and a Teflon tube conveyed the plant volatiles from the top of the roasting bag to one side of the wind tunnel. Inside the wind tunnel, the air was pumped out from the center at the 0 cm mark resulting in an air stream at the air transfer tubing of 480 – 570 mL/min with no greater difference than 4 mL/min between the two sides. One female moth at an age of 4 – 5 days was released through a hole in the center at 0 cm and its flight behavior was observed for 5 minutes. The wind tunnel was separated into 5 sectors of 20 cm each, and a time stamp was set every time the moth changed the segment (See Figure 5—figure Supplement 1). All experiments were conducted between 7 pm and 1:30 am under red light. Each trial was recorded on video using a mobile phone camera (Fairphone 4, 48 MP), which captured the wind tunnel from a front view. The videos were rewatched for the data acquisition and in cases where moths were barely visible, complemented with the notes from the live observation. To simplify the data analysis, the two left sectors (50 – 30 cm left & 30 – 10 cm left) were combined, as were the two right sectors (10 – 30 cm right, 30 – 50 cm right).

Statistical analysis

All statistical analysis was performed in RStudio (R version 4.3.2, RStudio version 2024.04.2). Mixed models were performed to determine significant effects of the positions of FAW eggs in the oviposition bioassays or for the duration of stay in the different sectors in the wind tunnel bioassays. In both bioassays a control with two maize plants was performed. Therefore, the control replications were included twice in the dataset, once with the egg count data or the duration of stay of both sides interchanged. In addition, the replication was included as a random effect into the model equations. For the oviposition bioassay, the egg counts on the maize or the Desmodium plant were compared amongst all treatments, with the inclusion of the greenhouse, the cage (combination of the cage group and the cage number), the start date and the replication number as random effects. For the wind tunnel bioassay, the duration of stay closer to the maize or the Desmodium plants were compared amongst all treatments, with the inclusion of the start date and the replication number as random effects.

Author Contribution

Conceptualization: D.M.O., F.C., and M.C.S.; Data curation: D.M.O.; Formal analysis: D.M.O., B.S.; Funding acquisition: M.C.S., F.C.; Investigation: D.M.O.; Methodology: D.M.O., M.C.S., F.C.; Project administration: M.C.S., F.C.; Resources: M.C.S., F.C., A.T.; Supervision: M.C.S., F.C., D.M.M., A.T., L.A.D.; Validation: D.M.O.; Visualization: D.M.O., L.A.D.; Writing – original draft: D.M.O., M.C.S; Writing – review & editing: M.C.S., F.C., B.S., D.M.M., L.A.D., A.T.

Acknowledgements

This research was supported by the European Union’s Horizon 2020 research and innovation programme under grant agreement No. 861998. We are thankful for experimental assistance from Silas Ouko, Chrispin Onyango, Nashon Opiyo, Michael Machongo, Amos Mwangangi and Collins Onjura.

Appendix 1 Pretests: Moth Activity Window

Pretests were carried out to determine the activity window of the FAW moths, as it was important for deciding the time window of the headspace sampling of volatiles that are relevant for the FAW moths. The FAW moth is reported to be nocturnal (Sparks, 1979) and was checked regularly from dusk until around midnight. In the first hours only little activity was observed, while toward midnight more flight and mating activity was observed. As a result, the moth activity could not be limited to the few hours of the evening and volatile sampling was proceeded from 7pm - 6.45am covering all dark hours.

Appendix 1—figure 1.