Collectively, this dataset contains the data and code required to replicate analyses in Comerford, Carroll and Egan, testing the hypothesis that red shouldered soapberry bugs (J. haematoloma) are consuming nectar and providing a pollination service for their host plants. However, the pollination benefit to the host is later reduced by seed predation from the pollinator’s offspring. Data cover 5 laboratory- and field-based experiments conducted in 2018 and 2019 at Rice University, Houston TX. In a test of insect pollen capture data (Test_of_Pollen_Capture.csv) shows that insects collected on either natal host plants capture pollen on their proboscis when feeding on nectar. We then show that this pollen capture contributes to pollination of their host plants demonstrated in a greenhouse experiment on host plant Cardiospermum halicacabum (Test_of_Pollination_Success_in_a_Greenhouse.csv and Greenhouse_time.csv), and in the field on host plant Koelreuteria elegans (Pollination_Success_in_the_Field.csv). We then demonstrate that nectar consumption increases insect longevity of the insect (Test_of_Insect_Longevity_with_Nectar_Feeding.csv). Lastly, data (Cost_of_a_Pollinating_Seed_Predator_to_the_Host_Plant.csv) shows that nymphs feeding on seeds have a direct cost to the plant via reduced seed viability.

Study System

The red-shouldered soapberry bug’s ecology and evolutionary history is well-described in terms of diet (Carroll and Loye 1987; Comerford et al. 2022), species introductions (Tsai et al. 2013), communication (Zych et al. 2012), dispersal (Winchell et al. 2000; Comerford et al. 2023), and developmental genetics (Carroll 2008; Fawcett et al. 2018; Angelini et al. 2022). The species’ native range is throughout the United States southward to northern South America. It is reported to be an obligate seed predator where it feeds exclusively on the seeds within the soapberry family, Sapindaceae, for growth and reproduction (e.g., Carroll and Loye 1987, 2012). This host plant association is an important aspect of J. haematoloma’s biology as both feeding and mating occur on, or around, their host (Carroll 1988). Gravid females generally oviposit into the soil at the base of the host plant or directly into the host’s seedpod via the feeding holes created by Lycaenid caterpillars (Carroll 1988) These insects possess tube-like mouthparts that are colloquially referred to as “beaks”. As a seed predator, adult, and late instar nymphs feed by utilizing their beaks to penetrate the seed pods of their host plants. The fruit of each host plant species varies in seed depth, resulting in divergent selection on beak length by host plant-association (Carroll 1988; Carroll and Boyd 1992). In southeastern Texas, where this study was conducted, J. haematoloma specialize in three host plants in the family Sapindaceae: the western soapberry tree (Sapindus saponaria var. drummondii), balloon-vine (Cardiospermum halicacabum), and the goldenrain tree (Koelreuteria elegans). (See Supplemental Natural History for further details.)

Natural History Observations

In the field, we observed both female and male J. haematoloma commonly nectar feeding on their host plants as well as nearby flowering non-host plants. Moreover, this behavior was most common when the primary seed food source was not available. This observation was repeated across sites (N = 13), U.S. states (Texas, Louisiana, Florida, and Virginia), years (2017-2021), and all regional host-plant-associated populations (Sapindus, Cardiospermum, and Koelreuteria), which led to our hypothesis that this was not an incidental association, but something more important to the biology of the insect and its host plants.

To assess the generality of nectar feeding by J. haematoloma, we searched museum records, community science reports, and the primary literature. First, we reviewed museum collections for non-host associations at the Smithsonian National Museum of Natural History (N = 317 specimens) and the Texas A&M University Insect Collection (N = 346 specimens). Second, we conducted a digital search for images of J. haematoloma interacting and feeding off of flowers on the community science platforms iNaturalist (www.inaturalist.org) and Project Noah (https://www.projectnoah.org), as well as a general Google (www.google.com)  image search that was performed in June of 2020 using the combination of search terms: “Jadera haematoloma* OR red-shouldered bug* OR soapberry bug* OR red-shouldered soapberry bug* AND flowers* OR nectar* OR plant* OR host* OR incidental”. We examine 5,798 resulting images on iNaturalist, 36 images on Project Noah, and 918 images in our Google image search. Collectively, between museum, community science, and digital image searches, we surveyed 7,415 records of J. haematoloma. Third, we surveyed the primary scientific literature for any discussion of nectar feeding and/or non-host-plant associations. We performed our search on Google Scholar in June 2020 using the same combinations of search terms. This generated 47 articles and 5 books that we searched for additional nectar feeding and incidental host records.

Given how common observations of nectar feeding were within J. haematoloma, we also searched iNaturalist for records of nectar feeding for the other seven species of Jadera listed in iNaturalist  (N= 753 records), and across the entire soapberry bug subfamily Serinethinae, including the three species in the genera Boisea (N= 7,732 records), and six species of Leptocoris (N=2,627 records) that were included in the iNaturalist database.

Test of Nectar Consumption

To test qualitatively if adult J. haematoloma consume nectar, we conducted nectar feeding trials using flowers from the host plant, K. elegans, where a UV pigmented dye was added to the nectar. Each inflorescence included in the trial was collected in October 2020 from local trees in Houston, TX. To keep trial conditions similar, extra flowers were stripped from the inflorescence’s branch to leave three flowers. A non-toxic, UV sensitive fluorescent purple pigment (Techno Glow Products, Ennis, TX, USA) was added to a clear commercially available nectar (Kaytee ElectroNectar®, Chilton, WI, USA) to create a 1:10 solution. Approximately 5uL of pigmented solution was injected into the base of each flower corolla proximate to the nectar lobes.

Individuals used in this feeding trial originated from the same K. elegans trees that provided the flowers. To encourage feeding, newly molted adults of both sexes and fourth instar nymphs were not fed for 48-hours but were given access to water only from 1.5ml plastic vials with a cotton stopper. Flowers with UV dye and insects were placed together inside a 34.6 x 21 x 12.1cm plastic feeding arena (Home Products International Inc, Chicago, IL) and observed until nectar feeding commenced (32% fed on nectar within 1 hour). Individuals observed nectar feeding were allowed to continue until they initiated a break in contact with the flower (typically, ~15 minutes). Those insects were then frozen in a -80°C ultra-freezer and subsequently dissected with two cross-sectional cuts. The first cut exposed the insects’ salivary ducts and esophagus by transversely separating the clypeolabral region of the head anterior to the eyes but posterior to the clypeus and beak. The second cut exposed the foregut by transversely separating the pronotum and the mesonotum sections of the thorax. The cut areas were exposed to UV-A emitting LED black lights (51 LED Blacklight, Vansky, Maharashtra, India) to assess for the presence of nectar. Dissections and observations for fluorescence were performed and imaged under 40X magnification using a Dino-Lite digital camera (Dino-Lite Edge AM7915MZT, Dino-Lite Premier; AnMo Electronics Corp, Taiwan) with complementary Software Studio Version 2.1.0.9. Twelve adults and four fourth instar nymphs were dissected to confirm nectar consumption (UV fluorescence = nectar in the salivary ducts and/or foregut).

Test of Pollen Capture

To evaluate J. haematoloma’s capacity to collect pollen on the integument, we conducted an experiment that compared the number of pollen grains captured by an insect’s beak after nectar feeding on two of its local host plants, C. halicacabum and K. elegans. Female insects were collected as third instar nymphs from each of the host plants in nature and reared to adulthood on their respective host’s seeds in the lab. As adults, ten individuals from each host were allowed to ingest nectar from their natal host plant’s flowers (e.g., C. halicacabum-associated insects nectar feeding on C. halicacabum). To evaluate for potential local adaptation in pollination having arisen between populations, wherein pollinators have evolved structures or behaviors that improve the pollen capture and transfer of their local host association, we transferred an additional ten insects from each host and allowed them to nectar on the alternative host (e.g., Cardiospermum-associated insects nectar on K. elegans flowers).  At the completion of the first nectar feeding event insects were frozen and the number of pollen granules on each insect’s beak was counted with a Dino-Light digital microscope at 200X magnification. Flowers from C. halicacabum used in this assay were grown from wild seeds in the greenhouse; flowers from K. elegans were collected from branches in nature the day of experiments. To evaluate pollen capture, we constructed a generalized linear model (GLM) assuming a log-linked Poisson distribution. We used ‘pollen counts’ as our response variable and for predictor variables, we included the ‘host’ plant species, the ‘host-association’ of the insect, and an interaction term.

Test of Pollination Success in a Greenhouse

We conducted a greenhouse experiment to test J. haematoloma’s contribution as a pollinator of one of its host plants, C. halicacabum. We grew 48 plants from seed (Outsidepride, Independence, OR) using a commercial potting soil (Miracle–Gro, Scotts Miracle-Gro Company, Marysville, OH) in a greenhouse under shaded natural light conditions and a constant 24˚C. Each plant was grown in its own 15.25 cm pot for eight weeks. All growing plants were kept isolated from invertebrates in 0.61m x 0.61m x 0.92m mesh exclusion cages (Caterpillar Castle, Live Monarch, Blairsville, GA) to prevent pollination and herbivory prior to the experiment. Plants were randomly assigned to a treatment: ‘no pollinators’, ‘J. haematoloma only’, and ‘ambient background pollinators’. Plants in each treatment group were divided into replicate mesh exclusion cages, each containing three plants to allow for cross-pollination. The ‘no pollinator’ group (6 cages) was isolated from all potential pollinators. The ‘J. haematoloma only’ group (6 cages) was isolated from all pollinators except was stocked with five adult unmated female J. haematoloma (deceased insects were replaced to maintain five active pollinators throughout). These two treatment groups were kept in the greenhouse for the entire experiment. Our third treatment of ‘ambient background pollinators’ (4 cages) were kept outdoors under shade in mesh cages, but with the cages open to allow all potential natural pollinators full access. To account for the possibility that plants could be wind pollinated, a rotating fan produced a low level of continuous simulated wind across all greenhouse plants. The cumulative number of flowers and seed pods per plant was recorded each week for seven weeks. Successful pollination of a flower was assessed as the development of a seed pod. New seed pods were removed within the first week of their development to prevent them from becoming an alternative food source, so as to encourage continued nectar feeding.

To analyze the role of J. haematoloma in pollination, we used a beta binomial linear regression model (Geissinger et al. 2022). Our response variable was the proportion of flowers that produced seed pods. We transformed the response variable by adding a nominal value of 0.001 to our proportional data (x+0.001) to remove zeros and bound the variable between zero and one. The predictor variables included in the model was the ‘pollinator’ treatment (no pollinator, J. haematoloma only, and ambient background pollinators).

Test of Pollination Success in the Field

We performed a pollinator exclusion experiment on free-living K. elegans, a medium sized dichogamous tree capable of self-pollination (Avalos et al. 2019). Over a two-year period (2018-2019), we isolated 51 K. elegans inflorescence branches with Nytex screen cages (1.8 m × 3.66 m × 1.83 m) that slip over the branch and tie at both ends to prevent pollinator visitation. Each isolated inflorescence was paired with a replicate inflorescence from the same branch. These replicates were visually matched for similarity in both their overall size and the number of flower buds. Bags were secured over the inflorescence prior to flower bloom to block pollination before the experiment. Due to wind damage and/or human interference during the experiment final sample sizes/replicates were not equal. In 2018, treatment (N=10) and control (N=11) branches were isolated from four trees located in southwest Houston, TX (29°41’03.7” N, 95°28’37.0” W). The pollinator treatment bagged inflorescences were each stocked with three adult females collected directly from the tree where the experiment was conducted, while the control bags were maintained as a complete pollinator exclusion. In 2019, we repeated this experiment with 13 treatment and 17 controls bags placed across five trees located on the Rice University campus (29°42’39.1” N 95°24’11.2” W).  In both years, the experiment started on the 25^th of September, commensurate with the peak flowering period of the host plant. The experiment was conducted over a 45-day period that ended with the abscission of the majority of the flowers on each host plant and all developing seed pods were counted and recorded as a proxy for pollination success.

We assessed J. haematoloma’s influence on pollination success using a GLMM with a Poisson distribution. For our response variable, we used the number of seed pods produced. To eliminate zeros from the distribution, we transformed seed pod values by adding one to every value. We included our isolation treatment (‘no pollinators’ versus ‘J. haematoloma’) as our predictor variable. We allowed for random intercepts by replicate. We initially included year as a random effect but found that the variance associated with this effect was approaching zero, so it was removed from the final model.

Test of Insect Longevity with Nectar Feeding

We qualitatively tested the benefits of nectar feeding on survivorship in adult J. haematoloma. Each individual insect was a newly molted adult, which as nymphs were reared on K. elegans seeds, but were prevented from feeding upon maturation. We included only female insects in this trial to reduce the confounding effect of sex (Carroll 1991). Assays were conducted in the same feeding arenas described above lined with sterile sand and a 15x8 cm section of cardboard egg-carton for climbing. The control groups were placed on a water diet only, while the nectar treatment group was provided an ad libitum diet of ElectroNectar (Kaytee Products Inc., Chilton, WI), a synthetic generalized electrolyte fortified hummingbird food designed to closely mimic the components of flower nectar. Specifically, the synthetic nectar includes the following ingredients: water, sucrose, potassium sorbate, citric acid, sodium bicarbonate, and potassium bicarbonate; with the following core components, moisture (max.) = 80.0%, sucrose (min.) = 20.0%, sodium (min. – max.) = 0.004% - 0.02%, and potassium (min.) = 0.035% (Kaytee Products Inc.).

Water and nectar sources were constructed using 1.5ml microcentrifuge tubes sealed with cotton stoppers. One drinking source was included for every three insects and all drinking sources were replaced daily. This assay was replicated four times per treatment with 36 unmated females included in each replicate (N=144). Mortality was assessed daily with deceased insects immediately removed to prevent cannibalism. The trial ran until complete mortality of all participants occurred. To evaluate the influence of nectar feeding on survival, we conducted a Kaplan-Meier survival analysis with ‘survival’ (# of days before death) as our response variable and ‘diet’ (nectar or water only) as our predictor variable. Building off of this, a Cox proportional hazard model was conducted to estimate hazard ratios associated with diet treatment.     

Cost of a Pollinating Seed Predator to the Host Plant

We performed a germination assay for seeds fed on by J. haematoloma to measure the cost of seed predation on viability. We used commercially purchased C. halicacabum seeds (Outsidepride® Independence, OR) grouped by three and placed in 90 x 15 mm petri dishes (Corning® Gosselin Corning Inc., Twerksbury, MA). Prior to inclusion, seeds were assessed for viability by being placed in water, with seeds that float being discarded. Neonatal nymphs circumvent their host’s thick seedcoats by utilizing feeding holes created by adult insects (Cenzer 2021). To increase juvenile survival in under laboratory rearing conditions, seeds are often manually cracked to simulate adult feeding damage (see Supplemental Methods). To mimic adult feeding damage without also reducing seed viability we bored four (0.31mm diameter) holes in the surface of each seed using a 30-gauge lancet (0.31 mm in diameter) with a softclix Accu-Chek® lancing device (Bio-Dynamics, Boehringer Mannheim, Indianapolis, IN). For each seed predation treatment, 30 seeds were divided evenly between 10 replicates. Five treatment levels were included in the experiment with control (zero nymphs), three (one nymph per seed), six (two nymphs per seed), nine (three nymphs per seed) and twelve (four nymphs per seed) nymphs. Originally, an additional treatment group of five nymphs was performed, however nymph mortality was high due to nymph density and competition. During the trial, first instar nymphs were allowed to develop to adulthood feeding only on the seeds available in the arena. Each arena was provided fresh water daily. Insects were removed as teneral adults, and the seeds were placed independently into Miracle-Gro potting media filled wells of cardboard egg cartons. Planted seeds were watered daily and given 80 days to germinate in the greenhouse under slightly shaded natural light and a constant 24˚C. At the end of the experiment, germination success was evaluated as a proxy for seed viability.

To assess how seed predation influences seed viability, we conducted a GLM with a binomial distribution. Our response variable in the model was seed ‘germination’ and our predictor variable was the number of insects that fed on the seed. Earlier mixed models considered both feeding arena and the area in the greenhouse where the seeds were planted, however, variance was close to zero and their effect was removed from the model.

Additional Details on Statistical Analysis

All statistical analysis and data illustrations were conducted with R v.3.6.1 (R Core Team, 2019). GLM and GLMM’s were performed in the R package “lme4” (Bates et al. 2011), while beta regression models were conducted using “betareg” package (Cribari-Neto and Zeileis 2010). Subsequent p-values were estimated in the “lmerTest” package (Kuznetsova et al. 2017), and Tukey tests were performed in the “emmeans” package (Length 2021). Kaplan-Meier survival curves and Cox proportional hazard model estimates were conducted in the “survival” package (Therneau et al. 2000; Therneau 2024).