Characterization of Salix nigra floral insect community and activity of three native Andrena bees

Abstract Salix nigra (black willow) is a widespread tree that hosts many species of polylectic hymenopterans and oligolectic bees of the genus Andrena. The early flowering of S. nigra makes it an important nutritive resource for arthropods emerging from hibernation. However, since S. nigra is dioecious, not all insect visits will lead to successful pollination. Using both visual observation and pan‐trapping, we characterized the community of arthropods that visited S. nigra flowers and assessed differences among male and female trees as well as the chemical and visual drivers that influenced community composition across 3 years. We found that male trees consistently supported higher diversity of insects than female trees and only three insect species, all Andrena spp., consistently visited both sexes. Additionally, Andrena nigrae, which was the only insect that occurred more on female than male flowers, correlated strongly to volatile cues. This suggests that cross‐pollinators cue into specific aspects of floral scent, but diversity of floral visitors is driven strongly by visual cues of yellow male pollen. Through time, the floral activity of two Andrena species remained stable, but A. nigrae visited less in 2017 when flowers bloomed earlier than other years. When native bee emergence does not synchronize with bloom, activity appears to be diminished which could threaten species that subsist on a single host. Despite the community diversity of S. nigra flowers, its productivity depends on a small fraction of species that are not threatened by competition, but rather rapidly changing conditions that lead to host‐insect asynchrony.

provisioned for development throughout the remainder of the year, all within the bloom time of their specific host (Danforth, 2007;Linsley, 1958;Stevens, 1949). In addition to being a valuable resource for early-emerging generalist floral insects, willow species belonging to the genus Salix are the primary hosts of many oligolectic bees, especially those belonging to one of the largest bee genera, Andrena (Ostaff et al., 2015;Stevens, 1949).
Salix encompasses between 300 and 400 species of shrubs and trees with a dual pollination system, which for any one Salix species occurs on a continuum between wind (anemophilous) and insects (entomophilous) (Argus, 2011;Karrenberg et al., 2002;Tamura & Kudo, 2000). Salix biology creates a unique environment for insect reward collection due to its dioecious nature. In order for sexual reproduction to occur, insects must locate male plants to collect pollen and then carry it to a separate female plant in the population (Dötterl et al., 2014). Host location typically occurs through a combination of visual, olfactory, and reward cues.
Salix species have a nonshowy inflorescence arranged in a catkin form where male flowers are often yellow (due to pollen color) and female flowers tend to be green (Füssel et al., 2007;Karrenberg et al., 2002). Additionally, Salix species emit a complex mixture of volatile organic compounds that are important as olfactory signals to insects, and both male and female plants offer nectar rewards (Füssel et al., 2007;Tollsten & Knudsen, 1992). However, upon locating a host plant, some insects may rob flowers of their resources and not carry pollen between male and female individuals (Galen & Butchart, 2003). Salix nigra, known by its common name black willow, is a treeform, entomophilous willow that grows throughout the Eastern United States north to Maine, west to North Dakota and south to Georgia (Burns & Honkala, 1990). The extensive range and productivity of S. nigra as well as its early bloom, typically February in its southern range through late June in more northern states, makes it an ideal resource for early-emerging insects (Burns & Honkala, 1990;Ostaff et al., 2015). Studying the mechanisms that S. nigra employs to attract insects as well as the influence of sex of tree on floral insect community through time is important in determining the competition for and potential stability of catkin resources, native oligolectic bee activity, and S. nigra reproductive success.
The goal of this study was to characterize the community of insects that visit S. nigra catkins and examine how the total floral community responded to tree sex, geographic position, volatile organic compound (VOC) profiles, and secondary metabolites in catkins and leaves. For comparison, we also evaluated the community of floral insects captured using visual survey techniques and pan traps placed in tree canopies. Finally, we examined the effect of tree sex, VOCs, and survey year on the activity of three native Andrena bee species, including the willow oligolectic bees Andrena macoupinense and Andrena nigrae.

| Population and site description
The target population of S. nigra was located in the West Virginia University Core Arboretum in Morgantown, West Virginia (39.6462°N, 79.9811°W). The Core Arboretum is an old-growth forest that contains 91 acres of native shrubs, trees, and herbaceous plants. It is located on a hillside that stretches between Monongahela Boulevard and the Monongahela River and contains riparian and floodplain sites with a small grove of S. nigra. The population contained thirty-two trees of which twelve were identified as female and twenty were male (Figure 1). Across all years of survey, 15 individual trees were observed, seven of which were female and eight males. Of these individuals, two were female and four were male trees in common among all years. Insect specimens were carefully hand collected from trees throughout the survey time for family-/species-level identification. Additionally, survey month and day were recorded for each observation to use as covariates in models analyzing individual insect activities as well as flower phenology. Finally, early-season herbivore activity, which was made up of observations of insects that feed on trees during and after catkins were no longer present, was collected in 2019 at the end date of tree flowering by counting number of herbivore occurrences on visual survey branches.

| Canopy pan-trap technique
In the year 2019, pan traps were constructed by painting three-ounce plastic cups with either fluorescent blue, fluorescent yellow, or white paint (Guerra Paint and Pigment). Fluorescent paints were a mixture of 16 ounces of the fluorescent dispersion to 1 gallon of silica flat paint.
Three cups, one of each color, were affixed with velcro to a bucket lid as a platform to be raised into tree canopies below flowering branches ( Figure S1) on five female and six male trees. A soapy trap solution was prepared by addition of approximately 5 ml of organic unscented dish soap into one gallon of water. Cups were filled 3/4 of the way full with soap solution. Traps were raised to the bottom of each tree canopies (2-12 m; average 4 m) at 9:00 a.m. in the morning and emptied daily at 5:00 p.m. Visual surveys were performed on the same days. Captured insects from each cup were transferred to separate vials containing 70% ethanol for later identification.

| Insect identification
Abundant insects were identified to species level while rare insects were identified to family. Native bee species identifications were

| Flower volatile and tissue collection
Dormant branches were collected from the field in early spring for six female and nine male Salix nigra trees along the river. Branches were allowed to root and flower in the Department of Biology greenhouse in buckets of water. Volatile organic compounds (VOCs) were collected using a dynamic headspace method (Keefover-Ring, 2013).
Nylon oven bags were placed over the flowering branches and bags were secured with thin gauge wire around cotton pads that had been

| VOC and metabolite characterization
Floral VOC samples were analyzed with a Thermo Trace 1310 GC coupled to a Thermo ISQ MS with electron ionization (EI) at 70.0 eV at 250°C, using helium as the carrier gas at 1.0 ml/min with the injector temperature set at 250°C. Oven conditions included an initial temperature of 40°C followed by an immediate ramp of 3°C min −1 to 200°C. Available standards, samples, and a continuous series of nalkanes (C 8 -C 20 ; Sigma-Aldrich) were injected (1 μl) in the split mode onto a TR-5MS capillary column (30 m × 0.25 mm I.D., film thickness 0.25 μm; Thermo Fisher Scientific). Compounds were identified with retention time matches to pure standards, mass spectra, and/or linear retention indexes calculated with the alkane series (Adams, 2001;El-Sayed, 2021;NIST, 2008). Standard curves of available compounds were used to calculate final VOC results, which were expressed as ng compound g −1 DW hr −1 .

| Sample collection for chemical characterization
Catkins and leaves collected from 24 trees (14 males and 10 females) in the field and in the greenhouse were characterized for five different secondary metabolites, all phenolic glycosides; salicin, isosalicin, salicortin, tremuloidin, and termulacin, as well as total metabolites.
Leaves were flash-frozen in the field and later shipped on dry ice to the University of Wisconsin-Madison, WI. The flash-frozen catkins and leaves were lyophilized, counted (catkins) and weighed (catkins and leaves), and then ground with steel balls in plastic scintillation vials in a ball mill. Accurately weighed portions (~15 mg) of powered leaf tissue were extracted with cold (4°C) methanol (1.00 ml) containing salicylic acid-d6 (Sigma-Aldrich) as an internal standard, sonicated in an ice bath (15 min), and then centrifuged to obtain a clear supernatant for analysis.

| Chemical analyses
Two μL of all standard and sample solutions was injected onto the UHPLC [Waters Acquity I-Class UPLC with a photodiode array detector (PDA) and a 3100 SQ mass spectrometer (MS), Milford, MA, USA] and separated peaks with a Waters Acquity CSH C-18 column (2.1 × 100 mm, 1.7 μm) at 40°C with a flow rate of 0.5 ml/min, using a gradient of water (solvent A) and acetonitrile (solvent B), both containing 0.1% formic acid. The PDA was configured to scan from 210-400 nm, with 1.2-nm resolution and a sampling rate of 20 points/s. The MS operating parameters for were as follows: cone potential, 30 V; capillary potential, 2,500 V; extractor potential, 3 V; RF lens potential, 0.1 V; source temperature, 120°C; desolvation temperature, 250°C; desolvation gas flow, 500 L/h; cone gas flow, 10 L/h; infusion rate, 5 μl/min; dwell time, 0.025 s.
Standard curves of methanol solutions, also containing the salicylic acid-d6 internal standard, of various purified compounds were used to calculate the concentrations in the extracted leaves, which were then normalized by sample dry weight and expressed as mg compound g −1 . Commercially available standards of salicin (Sigma-Aldrich), and salicortin, tremuloidin, and tremulacin were used that had been previously isolated and purified from aspen foliage (Lindroth et al., 1986).

| Floral insect community and sex effect on composition
To determine whether there were any sex differences in multivariate floral visitor community, the R package vegan (Oksanen et al., 2019) was used ordinate the data using nonmetric multidimensional scaling

| Sex differences in tree chemistry
Non-metric multidimensional scaling and ANOSIM were also utilized to determine if there were any sex differences in multivariate VOC and metabolite compositions by testing point grouping by sex of tree. Total monoterpenes, sesquiterpenes, VOC emissions, and metabolites were tested using a one-way analysis of variance (ANOVA) in SAS software version 9.4 to determine if there were any differences among sexes.

| Tree chemistry and insect community relationship
Pairwise geographic distances were calculated among trees from GPS coordinates. VOC production per catkin was scaled to branchlevel production using average catkin counts per branch for each individual tree to correlate to insect activity. Bray-Curtis dissimilarity matrices were generated for all datasets, including floral community, early-season herbivore community, pairwise geographic distances, VOCs, and catkin/leaf metabolites. A Mantel test was utilized to determine whether differences in floral insect community were a function of pairwise geographic distances, or differences in catkin VOCs, catkin metabolites, or leaf metabolites. This analysis was also repeated for the tree early-season herbivore community and catkin/ leaf metabolite dissimilarities. Insects that were abundant in surveys, including A. macoupinense, Andrena morrisonella, A. nigrae, Lasioglossum coeruleum, and parasitic wasps belonging to the Braconidae family, were extracted from datasets. Additionally, species richness was calculated from the dataset as total number of species to visit each tree, and Shannon-Weaver diversity was calculated using the R package vegan. A nested ANCOVA was used to analyze differences in transformed insect counts, richness, and Shannon-Weaver diversity with the following model: where ( ) indicates the independent variable is nested in another variable.
Additionally, an environmental fit was conducted using the R package vegan with floral volatile compounds that were overlaid on the floral visitor community. The resulting patterns were then evaluated to select specific volatile compounds to look for linear relationships to insect activity using Pearson correlations with Bonferroni p-value corrections to account for multiple testing. Insects chosen to test included Bombus sp., A. macoupinense, A. morrisonella, A. nigrae, Miridae, Chalcosyrphus nemorum, and Sarcophagidae. VOCs of interest included acetophenone, cisβ-terpineol, ethyl-1-hexanol, germacrene D, hexenyl acetate, octanal, octen-2-ol, and trans-3-pinanone.
2.9.5 | Floral community through time Among all 3 years, after accounting for mortality and branch loss, a total of six trees along the river overlapped among surveys, including two female and four male trees for 77 observation. NMDS and ANOSIM were utilized to determine the effect of survey year on floral insect community composition. Independent numeric variables associated with survey day, including year, Julian date, military time, and temperature, and these variables were correlated with the NMDS configuration using the environmental fit vector analysis in vegan (envfit function). Variables that were found to significantly correlate to the community dataset were added as additional covariates in all nested ANCOVA models.
Insects that were abundant in surveys, including combined counts of A. macoupinense, A. morrisonella, and A. nigrae were extracted from the community dataset and a nested ANCOVA to analyze differences in transformed insect counts, richness, and Shannon-Weaver diversity with the following model: where ( ) indicates the independent variable is nested in another variable.

| Floral insect community and sex effect on composition
For visual surveys, across all 3 years of data, 3,160 insects were observed to visit flowers. Of those observations, 88.9% were hymenopteran, 9.6% were dipteran, 1.3% were hemipteran, and 0.2% were coleopteran. Additionally, of all insects observed, bees belong-  Table 1).

| Tree chemistry and insect community relationship
Mantel tests (Table 3)

| 2019 survey method comparison and floral community
An NMDS with results from both survey methods indicated the appropriate number of dimensions for analysis was three (stress = 0.12; Figure 4a). An ANOSIM indicated that the floral community composition was dependent upon survey method (R = 0.666, pvalue = 0.001). Pan traps captured 284 total insects with the majority of insects captured belonging to the orders Diptera (33%) and Coleoptera (41%). There were 1,531 insect observations made during visual surveys with the majority (87%) belonging to the order Hymenoptera ( Figure 4b).
The NMDS vector analysis indicated that only Julian date was significantly correlated with the NMDS configuration (Vector Max R = 0.304; p-value = 0.001; Table S1). Julian date was then selected for use as a covariate in all nested ANCOVA models.

| Floral community through time
An ANOSIM indicated that the floral community composition dif-   sesquiterpenes, and all VOCs were also found to be similar among sexes. Similarly, differences in floral scent and plant tissue defenses among trees were not strong drivers of the community of insects that visited the flowers. Thus, we found no evidence that the influence that sex of tree on floral insect assemblages was due to the effects of VOCs or metabolite composition, although more subtle relationships might be revealed with more intensive sampling.
Our results are unusual given that dioecious plants, in which both sexes offer rewards, are often dimorphic in both scent TA B L E 4 Test of random effects p-values extracted from nested ANCOVA for most abundant floral visitors as well as calculated species richness and Shannon-Weaver diversity for 2019 survey analysis

F I G U R E 5
Average activity of most common floral visitors and calculated community metrics from 2019 Salix nigra surveys. Letters to the left of boxes indicate significantly different means as determined by a Tukey's HSD (p-value < 0.05) composition and emission levels (Ashman, 2009;Füssel, 2007;Okamoto et al., 2013). Nevertheless, this pattern does not always hold true in the Salix genus. For example, Salix repens, Salix bicolor, Salix caprea, and Salix cinerea all have similar overall volatile composition among male and female flowers (Füssel, 2007;Tollsten & Knudsen, 1992). This may reflect differences in the balance between visual and olfactory cues among Salix species, although this hypothesis remains to be robustly tested.
Although male and female trees did not differ in VOC and leaf metabolite composition, there was a sex effect on the metabolite composition of catkins. We also found that catkin metabolites exerted a strong influence over herbivore composition, suggesting that floral defenses are utilized differently among male and female trees, in turn attracting unique assemblages of herbivores. In the early season, trees may be at risk of herbivores feeding directly on flowers, which act as a large resource sink in plant tissues (Wäckers et al., 2007).
Floral larceny is a threat to female reproduction since females must maintain flowers through seed production, and this relationship may be antagonized in a dioecious system where chance of accidental pollination is rare (Maloof & Inouye, 2000;Richardson, 2004). Thus, it is important that females invest more resources toward defense of catkins, as supported by our finding that female individuals emitted more (3E)-hexenyl acetate and (2E)-hexenyl acetate. Hexenyl acetates have been frequently characterized as common green leaf volatiles emitted upon the crushing of plant tissue (Heil et al., 2008;Wei & Kang, 2011;Whitman & Eller, 1990), and they appear to also be a common component in floral scent (Kaiser, 1994;Messinger, 2006;Tollsten & Knudsen, 1992). Hexenyl acetates are associated with the attraction of insect parasites, which may provide female flowers in S. nigra added protection benefits against floral resource theft (Ngumbi et al., 2009;Whitman & Eller, 1990 to those relying heavily on visual cues (Laubertie et al., 2006;Tuell & Isaacs, 2009;Vrdoljak & Samways, 2012).   (Ribble, 1974;David W Ribble, 1968;Stevens, 1949). Of the three species, A. nigrae emergence is more closely coupled to the bloom of local S. nigra trees. Additionally A. nigrae has been recorded as a primary pollinator of S. nigra in two other states, Illinois and North Dakota, indicating that it may be more tightly coupled to the biology of S. nigra than other observed native bees in our site (Ribble, 1968;Stevens, 1949 (Ribble, 1974).
We have used a combination of survey approaches to characterize factors affecting community of insect floral visitors in S. nigra. our main findings of sex dimorphism were stable across years, suggesting that this may be a general characteristic of the species. However, we detected potential inter-annual variability in the S. nigra-Andrena interaction, which illustrates that shifting seasonal transitions could detrimentally affect plants that depend on early-emerging arthropods for sexual reproduction as well as the arthropods that depend on resources provided by early-flowering plants.

ACK N OWLED G M ENTS
We thank the following individuals for assistance with fieldwork-Jacob Miller, Brandon Sinn, and Mark Burnham. Zachariah Fowler provided logistical support and access to the WVU Core Arboretum

CO N FLI C T O F I NTE R E S T
None declared.

DATA AVA I L A B I L I T Y S TAT E M E N T
All raw phenotypic data have been deposited in Dryad (https://doi. org/10.5061/dryad.nvx0k 6drr), including arthropod survey counts and metabolite profiles.