Field sites and phenology in southern Ontario, Canada

We studied several aggregations of eastern carpenter bees nesting in six wooden bridges at the Glenridge Quarry Naturalization Site, in St. Catharines, Ontario, Canada (43.1° N, 79.2° W). Each bridge was approximately 4.5 m long and 2.2 m wide and contained from 3 to 34 nests. Bridges were 30–60 cm above the ground, and nest entrances were located on the underside of each bridge. Three bridges with smaller aggregations were located within 50 m of one another, whereas the other three bridges, which contained larger aggregations, were located at least 300 m apart.

At our study sites, the flight season is divided into three phases. The first phase begins in mid-April to mid-May, depending on local weather conditions. Males emerge first from hibernation in their natal nests; they guard territories directly in front of nest entrances, pouncing on females as they depart from or arrive at their nest entrances. Matings also occur at other sites, such as in flower patches, but we observed matings only near nest entrances. Some males also establish territories near other landmarks and in floral patches where females may forage.

The second phase of the flight season begins several days to a couple weeks after most males have emerged, when females also emerge from their nests. Phase 2 is referred to as the nestmate provisioning phase because foragers feed adult female nestmates during this time, rather than provisioning brood. Phase 3, the brood provisioning phase, generally begins in late May or early June, when females begin provisioning brood cells.

Observation methods

Cup traps were used to catch bees leaving and entering nests in order to mark bees and track foraging behaviour (Peso and Richards Reference Peso and Richards2010). Traps consisted of plastic cups with a 1- to 2-cm hole cut in the base and with cellophane held in place with an elastic band to cover the top (Fig. 1). The cups were then placed over nest entrances and secured in place with velcro, and modelling clay was used to seal gaps around the base of each cup to prevent bees from escaping. Traps were set up upon arrival at a study site each morning (09:00 hours, local time) and were removed at the end of the day (17:00 hours). Cup traps were attached to the nest entrances at five aggregations in 2012 and 2013, at two in 2016, at five in 2017, and at six from 2018 to 2021.

Fig. 1. Cup trap used to capture bees arriving and departing from nest entrances: A, each cup has a hole cut in the base, two strips of Velcro on the base, and an elastic band and cellophane to cover the open side of the cup; B, cups were attached over each nest entrance.

When a bee was first captured, its nest of origin and sex were recorded. Bees were also caught and marked flying near the aggregations, where their nests of origin were unknown. Collected bees were placed on crushed ice in a cooler to arrest their movements. After cooling, the left mesothoracic tarsus of each bee was removed and immersed in 0.5 mL of distilled 100% ethanol in a 1.5-mL microfuge tube. Tarsi were stored on ice until the end of each observation day before being transferred to a –80 °C freezer for eventual DNA extraction and microsatellite genotyping. Each bee was then marked with a unique tri-colour combination using enamel model paints. Paints were deposited on the bee’s thorax between the tegulae and on the head between the ocelli using a plastic pipette. After a bee was marked, it was placed back into the nest entrance from which it had departed, or it was released near the aggregation if it was caught by net. If paint wore off a bee, it was recaptured and remarked. The numbers of bees captured and marked each year are presented in Table 1.

Table 1. Sample sizes for inferring copulation frequencies using field observations and microsatellite genotyping. Visual observations of individually marked bees were based on field work carried out from 2016 to 2021. Genotype analyses were carried out for females collected in 2012, 2013, and 2016–2020.

Observations of mating behaviour

Attempted and successful copulations were recorded as part of more intensive observations of female foraging effort and male territoriality from 2016 to 2021. Observations at one aggregation per day were carried out by one person from 09:00 to 17:00 hours on sunny days without rain. Whenever copulations were noticed over the course of daily behavioural observations, they were recorded. Based on visual observations, we calculated mating frequency as the harmonic mean number of observed copulations (both complete and incomplete) per bee. Visual observations of matings around nesting aggregations underestimate actual mating frequencies because not all sites were observed each day, and copulations likely also occur away from nest sites.

To evaluate female receptivity, we distinguished between complete and incomplete copulations, presuming that copulations were less likely to be completed when females were less receptive. A complete copulation was recorded when a male mounted a female by grabbing her thorax from above with his legs, followed by abdominal thrusting with the male’s front legs on the female’s thorax. Mating pairs either landed on the ground or flew high into the air while in copula, often out of the observer’s view. An incomplete copulation was recorded when a female dislodged a male before abdominal thrusting occurred. Incomplete copulations either ended quickly or were followed by a sequence of male–female interactions for an extended period of time. For example, after dislodging a male, a female sometimes turned in flight and extended her legs out to the side, then flew towards the male, moving from side to side, while the male flew backwards and matched her movements. Males usually flew slightly higher off the ground than females during this type of interaction, maintaining their distance until females turned away, after which males either pursued females again or flew away. After some complete or incomplete copulations, females landed on the ground, and males hovered nearby. Males that hovered near landed females rarely attempted to mount females on the ground. An example of this type of mate-guarding behaviour can be seen at https://youtu.be/4TBJOBfwjuA.

DNA extraction and microsatellite genotyping

Chelex 100 Resin (Bio-Rad Laboratories, Hercules, California, United States of America) was used to extract DNA from tarsal samples following protocols described by Casquet et al. (Reference Casquet, Thebaud and Gillespie2012). Ethanol was removed from each tube by drying the template for 4–6 hours in a 45 °C incubator. Following the incubation step, we added 150 µL of 10% Chelex Resin solution (1.5 g Chelex 100 Resin and 15 mL Milli-Q water) and 10 µL of Proteinase K (Qiagen^®, Hilden, Germany) to each microfuge tube. The tubes were then placed in a thermoshaker at 55 °C and 400 rpm for 6–8 hours. Following incubation, DNA templates were either stored at 4 °C for samples to be amplified the same week or at –80 °C for samples to be amplified later.

Genomic DNA was amplified at 16 previously described microsatellite loci (Vickruck Reference Vickruck2015). Bees previously collected in 2012 and 2013 were genotyped at nine loci: XV3, XV7, XV23, XV24, XV27, XV30, XV39, XV42, and XV43 (Vickruck Reference Vickruck2015; Vickruck and Richards Reference Vickruck and Richards2021). Bees from 2016 were genotyped at eight of these same loci, excluding XV23. Bees from 2017 to 2020 were sequenced at 16 loci, the nine listed above, plus XV1, XV9, XV10, XV12, XV14, XV28, and XV29 (Vickruck Reference Vickruck2015). Each DNA sample was amplified with a 15-µL polymerase chain reaction using 40–70 ng of DNA, 0.5 units of Taq DNA polymerase (New England Biolabs, Ipswich, Massachusetts, United States of America), 1× Thermobuffer (New England Biolabs), 0.2 mM dNTPs, 0.2 µM forward primer, and 0.2 µM reverse primer. Each forward primer contained a poly-A fluorophore, 56-FAM, or HEX. The polymerase chain reaction conditions were 95 °C for 3 minutes, 40 cycles of 94 °C for 30 seconds, 60 °C for 30 seconds, and 72 °C for 30 seconds, and then a final extension of 72 °C for 300 seconds. Electrophoresis of each polymerase chain reaction product was conducted using an ABI 3730×l analyser (Applied Biosystems Corporation, Thermo Fisher Scientific, Waltham, Massachusetts) at the Peter Gilgan Centre for Research and Learning Genetic Analysis Facility (Hospital for Sick Children, Toronto, Ontario, Canada). Each run contained positive and no-template controls for each locus to account for variation between reactions. Alleles were called manually by A. Proulx using the free software, Applied Biosystems Microsatellite Analysis Software (MSA by Thermo Fisher Scientific). The numbers of female bees genotyped each year are presented in Table 1.

Inferring female mating frequencies from microsatellite genotypes

A singly mated (monandrous) female produces daughters that are all full sisters, whereas a multiply mated (polyandrous) female produces daughters that can be a mix of full and half sisters. Therefore, the proportions of monandrous and polyandrous females in the population each year can be inferred from the proportions of full sibships and maternal half sibships among genotyped females.

We used COLONY, version 2.0.6.4 (Jones and Wang Reference Jones and Wang2010), to identify probable full and half sisters from three different data subsets, based on the 9, 8, or 16-locus genotypes described above (Table 2). Because female eastern carpenter bees can overwinter twice (Richards and Course Reference Richards and Course2015), sisters may be collected in consecutive years. We therefore included females collected in multiple years in the input data for COLONY but then excluded predicted relationships that were three or more years apart. Each data subset was run five times using different starting seeds in order to detect all possible sibling relationships. Sibships were inferred using the full-likelihood method in COLONY, based on the following assumptions: random mating, possibly polygamous male and female parents, haploid males, and diploid females. We used a medium-length run with medium likelihood precision, with run specifications and sibship priors set to default values, an initial allele mistyping rate of 0.0001, and an initial allelic dropout rate set to 0.0001 for each locus.

Table 2. Three different data subsets were used for COLONY assignment analyses. The first and second subsets used different numbers of females using the same number of loci, whereas the second and third subsets consisted of the exact same females with different numbers of loci. Genotyped females with fewer than six loci (subsets 1 and 2) or 12 loci (subset 3) were excluded from analyses.

COLONY produces a list all possible pairs (dyads) of full sisters, maternal half sisters (same mother, two different fathers), and paternal half sisters (same father, two different mothers) with likelihoods of 0.1 or greater. We focused on dyads with a likelihood of 0.7 or greater and of 0.8 or greater. We then assembled the dyads into larger sibships, represented by short networks (Fig. 2). No paternal half-sister dyads reached the likelihood thresholds of 0.7 or 0.8, so we analysed only maternal sibships and could not estimate male mating frequencies. Because all high-likelihood dyads represented maternal siblings, we made the simplifying assumption that sibships were the offspring of a single mother. We also assumed that sibships could extend across two consecutive years because some females overwinter twice, but we excluded three-year sibships from consideration.

Fig. 2. Diagram illustrating how full- and half-sister sibships were counted for one run of COLONY based on data subset 3. Nodes (circles) represent individual females, with colours representing the year in which they were collected. Full sisters are linked by red lines and half sisters are linked by gray lines. In total, there are 19 full-sister and 13 half-sister sibships (the latter may contain a mix of full and half sisters). A is a fully Euclidean, full-sister sibship, B is a fully Euclidean, half-sister sibship, and C is a non-Euclidean half-sister sibship. Numbers beside each sibship indicate the most likely number of fathers, given the assumption that each sibship was produced by a single mother.

To illustrate how dyads were assembled into sibling networks such as those depicted in Figure 2, we considered several examples. In the first example, COLONY predicts three full-sister dyads: females A and B, B and C, and A and C. Because each female is the full sister of the other two females (relatedness among siblings is fully Euclidean), these three dyads can be collapsed into a single, full-sister sibship, such as that illustrated in Figure 2A, representing the daughters of one mother mated to one father. Likewise, if COLONY predicts a full-sister dyad for females D and E and two half-sister dyads for females D and F and females E and F, then these three dyads can be collapsed in a single sibship with two full sisters and one half sister, such as that illustrated in Figure 2B. Some predicted sibships are not fully Euclidean. Consider the situation in which COLONY predicts a full-sister dyad (G and H) and one half-sister dyad (G and I). These would be assembled into a non-Euclidean sibship such as that illustrated in Figure 2C. Inferring the number of fathers in Euclidean networks is straightforward; full-sister networks represent the offspring of a single mother and a single father (Fig. 2A), whereas half sisters represent the offspring of a single mother and two or more fathers (Fig. 2B). Inferring the number of fathers in non-Euclidean sibships is more challenging because we cannot always distinguish multiple maternity from multiple paternity. To solve this conundrum, we assumed that sibships were the offspring of a single mother because COLONY failed to predict any paternal half-sister dyads with a likelihood greater than 0.7 or greater than 0.8 in any data subset. For a sibship such as that shown in Figure 2C, this would suggest a single mother and two fathers. If this assumption were violated, the effect would be to overestimate the numbers of fathers (i.e., the number of times a mother mated). We used the networks to count full-sister versus half-sister networks (half-sister networks comprised a mix of full- and half-sister relationships) and then inferred the most likely number of fathers for each sibship. We then calculated the harmonic mean number of fathers per sibship as our measure of female mating frequency.

Additional analyses

Network visualisations and statistical analyses were performed in R, version 4.1.2, running under the R-Studio shell, version 1.1.383 (R Core Team 2022; RStudio Team 2022). We used the igraph package to assemble dyads into networks by designating individual bees as network nodes and their relationships as network edges. Chi-square tests were used to compare female and male mating frequencies and how the proportions of complete and incomplete copulations differed across foraging phases.