INTRODUCTION
Discovering how the interactions of plants and pollinators play out is critical to understanding how they mutually rely on each other, and molecular methods are an increasingly common toolkit in this endeavor. In particular, identifying the source of DNA in mixed-DNA samples has become increasingly popular with the development of a range of metabarcoding methods. These approaches rely on single-locus PCR amplicons, leveraging the power of high-throughput DNA sequencing to compare reads to a database of putative DNA sources. The applications of metabarcoding are manifold, including analyses of microbiomes and diets (Sousa et al. 2019), as well as environmental DNA analysis to quantify community composition (Deiner et al. 2017, Sousa et al. 2019). Such molecular approaches have also recently been applied to plant identification from the pollen found on bees to determine which species of flowers they have visited (Wilson et al. 2010; Galimberti et al. 2014; Sickel et al. 2015; Bell et al. 2016, Bell et al. 2017). Historically, pollen has been identified using microscopic morphological differences to distinguish different plant species (Martin and Harvey 2017). However, for some plant taxa, morphological similarity of pollen between distinct species makes this method impossible. In these situations, pollen grains taken from wild bees can be molecularly interrogated to identify the species of plant they are from (Mitchell et al. 2009, Galliot et al. 2017, Lucas et al. 2018).
Molecular methods for identifying pollen origin have been generally restricted to making inferences about the presence or absence of a source plant species from pollen (via metabarcoding techniques; Bell et al. 2017). This is in part because previous methods have been unable to reliably quantify relative abundance of mixed pollen samples for several reasons, discussed in detail in Bell et al. (2019). One reason is that when using plastid DNA in metabarcoding (e.g. Galimberti et al. 2014), it is unclear how the abundance of chloroplast DNA (cpDNA) is related to overall pollen abundance; if the ratio of cpDNA and pollen grains is not one-to-one, this could bias estimates of relative abundance (Richardson et al. 2015, Bell 2019). Another bias of pollen metabarcoding approaches has to do with how polymerase chain reaction (PCR) amplifies target markers (Bell et al. 2017). In PCR, the final concentration of amplicon DNA after a full PCR protocol is not necessarily directly correlated to input DNA concentration. This is because most PCR protocols will take the amplification process into the “plateau phase”, usually after approximately 30 thermal cycles. At the plateau stage, amplicon concentration may be a function of exhausted reaction reagents rather than original concentration of input DNA - the desired information. Because of this, studies using the “plateau abundance” of amplicons could be subject to PCR bias, especially in cases of low abundance (rarity) in the sample.
Though the potential weakness of using PCR to quantify amplicons has been raised by previous authors (e.g. Richardson et al. 2015; Bell et al 2017; Bell et al. 2019), it is still of interest to use molecular methods to estimate the relative abundance of different pollen sources found in bees’ scopae or corbiculae (that is, in their pollen-carrying morphological structures). This is in part because understanding resource use among mutualists such as plants and pollinators is fundamental to determining the extent to which they may rely on each other for population persistence (Roulsten and Gooddell 2011), support ecosystem functioning (Lucas et al. 2018), and mediate interspecific competition and coexistence (Johnson 2019). In fact, as Bell et al. (2019) noted, the ability to molecularly determine species abundances in mixed-pollen samples may be ‘groundbreaking’ for understanding plant-pollinator communities, because successful pollen transport between plants determines plant seed production.
Plant-pollinator interactions can provide a predictive framework for how pollinators introduce reproductive interference and/or fitness benefits to co-flowering plants (James 2020). The effects of pollinator visitation on plant seed set are determined by the extent to which pollinators (1) carry a mix of pollen on their bodies while foraging (constancy) and (2) carry rare versus common species in their pollen balls (preference). During a single pollen collection bout, bees can visit multiple flower species, a behavior known as inconstancy, or visit the same flower species, known as constancy (Kunin and Iwasa 1996). Pollinator inconstancy exposes plants to reproductive interference via heterospecific pollen transfer, which has been shown to drive lower seed production and fitness in plants (Mitchell et al. 2009; Carvalheiro et al. 2014; Arceo-Gómez et al. 2019). Pollinator preference is a measure of flower choice by pollinators. If a strongly competitive plant is preferred by pollinators, pollinator preference might exacerbate competitive exclusion, but if a weakly competitive plant is preferred by pollinators, preference could mitigate competitive exclusion. Though pollinator-mediated plant interactions are most often understood through the lens of pollinator behavior, ample evidence suggests that linking pollinator visitation to plant seed set is problematic: not all plant-pollinator contacts result in pollen transfer (Mayfield 2001; Popic et al. 2013; Ballantyne et al. 2015, 2017; Barrios et al. 2016). Because of this, the relative abundances of pollen on bees could provide valuable information about how pollinators mediate plant interactions.
Here, we develop and use a novel high throughput amplicon sequencing method to quantify the relative abundance of different pollen sources on bee pollinators visiting a group of sympatric winter annual plants in the genus Clarkia (Onagraceae). This group of plants – C. cylindrica ssp. clavicarpa (Jeps.) Lewis & Lewis, C. speciosa ssp. polyantha Lewis & Lewis, C. unguiculataLindl., C. xantiana ssp. xantiana A. Gray — are sympatric in the woodland-chaparral areas of the southern foothills of the Sierra Nevada mountain range (from here, we do not use their subspecies epithets). The four species of Clarkia rely on a small group of bee pollinators specialized on the genus Clarkia rather than any one species (MacSwain et al. 1973; Moeller 2005). Though theseClarkia have distinct adult phenotypes, their pollen grains are morphologically indistinguishable. The Clarkia also co-occur with each other more often than they occur alone in plant communities in their range of sympatry, and assemblages can contain one to four species of Clarkia (Eisen and Geber 2018). Finally, Clarkia bloom much later in the growing season than the vast majority of co-occurring flowering annual plants, and as such are often the only flowering plants where they occur.
A previous study of the bee visitors to one of the four species of interest to this study, C. xantiana , showed that though 49 species of bee visit C. xantiana, there were only 12 likely “core pollinators” of the species, nine of which carry almost exclusively Clarkia pollen (Moeller 2005). Studies including the other three species of Clarkia in their range of sympatry withC. xantiana have found that there are three consistently common pollinator taxa in multi-species Clarkia assemblages (Singh 2013, James 2020). The most common pollinator, Hesperapis regularis(Melittidae) has been shown to preferentially visit C. xantiana . Preferences of bees in the Lasioglossum genus (Halictidae), the second most common pollinator taxon, is unclear: they have been shown to visit all Clarkia species at relatively the same rates (Singh 2014) or preferentially visit C. xantiana and C. cylindrica (James, 2020). The unresolved nature of Lasioglossumpreferences are at least partially explained by the fact that it is difficult to identify different Lasioglossum species when observing them on the wing, as in (James, 2020). Despite preferences,Hesperapis regularis and Lasioglossum species visitC. cylindrica, C. unguiculata, and C. xantiana regularly, and are inconstant when foraging in diverse arrays and thus likely to transfer incompatible pollen between plants (James 2020). The final most-common bee pollinator in the system, Diadasia angusticeps(Apidae), is behaviorally more specialized on one Clarkiaspecies, C. speciosa , and rarely visits the other species ofClarkia (Singh 2013, James 2020).
Critically, experimental evidence in this system has linked the behavioral inconstancy and preference of pollinators with Clarkiaseed production (James 2020). Clarkia speciosa exhibits low pollen limitation to reproduction, which may be explained by the constancy and preference behavior of Diadasia angusticeps. The other three species exhibit higher pollen limitation to reproduction, which may be due to the inconstancy of Hesperapis regularis andLasioglossum sp . However, because pollinator visitation does not equate with pollen transfer, it remains unknown if pollinator preference and constancy in the Clarkia system in fact determine plant interactions.
The wealth of natural history knowledge and the morphological similarity of Clarkia pollens make the Clarkia system ideal for developing a method that can both identify and quantify different species in pollen samples. In this paper, we develop a method that we call “quantitative amplicon sequencing” or “qAMPseq” to quantify the relative abundance of Clarkia pollen in pollen balls from wild bees. Quantitative amplicon sequencing uses the amplification curve of PCR as a backbone for quantification: it targets single nucleotide polymorphisms private to each species, and then uses PCR to amplify these regions with the goal of post-amplification sequencing, as in metabarcoding. Critically, PCR amplification is stopped before saturation at four different times so one can estimate when each species’ amplification curve crosses a critical threshold (as in quantitative PCR or qPCR; Figure 2). The estimate of when the curve crosses the critical threshold point is then used to estimate the relative abundance of each species in each sample. If previously-observed trends in bee behavior match what they carry in their pollen balls (James 2020), then we predict the inconstant bees,Hesperapis regularis and Lasioglossum sp. , will carry multiple species of pollen at once. In addition, because these particular bee taxa have an established behavioral preference forC. xantiana , we expect that they will carry more C. xantiana than other Clarkia species. We also predict that the pollen on Diadasia angusticeps will contain only C. speciosa pollen. Finally, because Hesperapis andLasioglossum sp. are the most common taxa in this region of Clarkia sympatry, we predict an overall pollinator preference for C. xantiana pollen.