Abstract
Anthropogenic activities are leading to changes in the environment at global scales, and understanding these changes requires rapid, high-throughput methods of assessment. Pollen DNA metabarcoding and related methods provide advantages in throughput and efficiency over traditional methods, such as microscopic identification of pollen and visual observation of plant-pollinator interactions. Pollen DNA metabarcoding is currently being applied to assessments of plant-pollinator interactions and their responses to land-use change such as increased agricultural intensity and urbanisation, surveillance of ecosystem change, and monitoring of spatiotemporal distribution of allergenic pollen. In combination with historical specimens, pollen DNA metabarcoding can compare contemporary and past ecosystems. Current technical challenges with pollen DNA metabarcoding include the need to understand the relationship between sequence read and species abundance, develop methods for determining confidence limits for detection and taxonomic classification, increase method standardisation, and improve of gaps in reference databases. Future research expanding the method to intraspecific identification, analysis of DNA in ancient pollen samples, and increased use of museum and herbarium specimens could open further avenues for research. Ongoing developments in sequencing technologies can accelerate progress towards these goals. Global ecological change is happening rapidly, and we anticipate that high-throughput methods such as pollen DNA metabarcoding are critical for assessing these changes and providing timely management recommendations to preserve biodiversity and the evolutionary and ecological processes that support it.
Keywords: pollen; pollination; DNA metabarcoding; metagenomics; environmental DNA; global change ecology; ecosystem change
Introduction
Anthropogenic activities are leading to global changes in the environment, including, habitat loss (Ellis, Klein Goldewijk, Siebert, Lightman, & Ramankutty, 2010), climate change (Hansen, Reudy, Sato, & Lo, 2010), biodiversity decline (e.g., Bowler et al., 2020; Butchart et al., 2010), and the spread of invasive species and diseases (Hulme, 2009; Pyšek et al., 2020). Such global changes can act additively or interactively (Didham, Tylianakis, Gemmell, Rand, & Ewers, 2007; Peters et al., 2019) to alter species composition through events such as local introductions and extinctions (Mathiasson & Rehan, 2020; Portman, Tepedino, Tripodi, Szalanski, & Durham, 2018), shifts in phenology (Bartomeus et al., 2011; Forrest, 2015), and changes in the dispersal and connectivity of populations (Damschen et al., 2019). These impacts can subsequently affect the spatiotemporal overlap of species and their behaviour, which can alter species interactions, restructure food webs (Dunn et al., 2018; Kortsch, Primicerio, Fossheim, Dolgov, & Aschan, 2015; Richardson et al., 2021), and create network instability (Brosi & Briggs, 2013; Revilla, Encinas-Viso, & Loreau, 2015). Ultimately, these changes can lead to negative impacts on ecosystem services (e.g., pollination; Burkle, Marlin, & Knight, 2013; Potts et al., 2010), economic productivity (e.g., decreasing agricultural production; Lark, Spawn, Bougie, & Gibbs, 2020; Reilly et al., 2020) and human health and quality of life (e.g., changing distribution and phenology of allergenic pollen; Anderegg et al., 2021; or loss of pollinator-dependent crops; M. R. Smith, Singh, Mozaffarian, & Myers, 2015). Understanding and managing these changes requires rapid, high-throughput assessments and responses.
Pollen is a powerful biomarker for detecting spatial and temporal variation in species assemblages and interspecific interactions, making it ideal for a high-throughput assessment of global ecological change (Hornick et al., 2021). For example, the presence of pollen in environmental samples can be used to determine plant species composition, which can be helpful for surveying changes in biodiversity (Leontidou et al., 2021; Matthias et al., 2015), comparing current ecosystems to historical samples (Gous, Swanevelder, Eardley, & Willows-Munro, 2019; Simanonok et al., 2021), the early detection of biological invasions (Tremblay et al., 2019), and monitoring airborne allergenic pollen impacting human health (Suanno, Aloisi, Fernández-González, & Del Duca, 2021). Pollen can also be used to assess changes in phenology (Burkle et al., 2013), detect plant-pollinator interactions (Bänsch et al., 2020; Gresty et al., 2018; Kaluza et al., 2017; Lucas, Bodger, Brosi, Ford, Forman, Greig, Hegarty, Jones, et al., 2018; Richardson et al., 2021; Sponsler, Shump, Richardson, & Grozinger, 2020), reconstruct pollen transport networks (Tur, Vigalondo, Trojelsgaard, Olesen, & Traveset, 2014), and reconstruct past vegetation and, from this, climate (Courtin et al., 2021; Liu et al., 2021; Niemeyer, Epp, Stoof-Leichsenring, Pestryakova, & Herzschuh, 2017; Laura Parducci et al., 2019). The identification of pollen from the bodies of animals is particularly useful for the reconstruction of plant-pollinator interaction networks because it increases the temporal scale of information obtained, leading to more connected networks than those reconstructed through observations of flower visitation (Arstingstall et al., 2021; Bosch, Gonzalez, Rodrigo, & Navarro, 2009; de Manincor et al., 2020).
Traditionally, taxonomic identification of pollen is based on the visual observation of pollen morphology. Pollen grains are stained and mounted on microscope slides and identified to the lowest possible taxonomic group using morphological characteristics visualised with light microscopy. While counting pollen with this method is possible and can be used to assign species abundance, it requires highly trained specialists (of whom there is a small and dwindling number) and is time-consuming, leading to most identifications taking place on small subsets of pollen samples (Stillman & Flenley, 1996). Moreover, the lack of variation in distinctive morphological characteristics limits the microscopic identification of pollen typically to the level of genus and often only family (Lau et al., 2019; Mander & Punyasena, 2014; Richardson et al., 2018). Automated taxonomic identification has been suggested (Stillman & Flenley, 1996) and implemented to overcome these limitations in three ways. First, image analysis of morphological characteristics employs methods such as texture analysis (Marcos et al., 2015), and multiple convolutional neural networks (Bourel et al., 2020; Olsson et al., 2021; M. Polling et al., 2021; Sevillano, Holt, & Aznarte, 2020). Second, methods based on chemical characteristics include Raman spectroscopy (Pereira, Guedes, Abreu, & Ribeiro, 2021), magnetic resonance spectroscopy (MRS) (Klimczak, Ebner von Eschenbach, Thompson, Buters, & Meuller, 2020), or Fourier-transform infrared (FTIR) spectroscopy (Muthreich, Zimmermann, Birks, Vila‐Viçosa, & Seddon, 2020; Zimmermann, 2018). Spectroscopy has been applied to single pollen grains (Diehn et al., 2020) and, as far as we know, it has not been used for mixed-species pollen samples and may be limited in this capacity. Machine learning (Gonçalves et al., 2016) or deep learning (Dunker et al., 2021) methods can be used to improve the identification of taxa from the data obtained with either morphological or chemical methods. In addition, flow cytometry has been successfully used to count pollen and sort grains for downstream analyses (Kron, Loureiro, Castro, & Čertner, 2021). Combining chemical and image analyses with flow cytometry and deep learning methods can yield fast and accurate taxon identification and quantification (Dunker et al., 2021). However, although these methods work in some contexts, only a small number of different pollen types are often included, and most studies have not tested multi-species pollen samples. Moreover, most advanced techniques require specialised equipment and extensive training datasets to calibrate the taxonomic assignments. Research on a third method for pollen identification, molecular genetics (i.e., the use of DNA sequences), has recently gained attention for its high-throughput capabilities and ubiquitous techniques and equipment and is the focus of this paper.
The DNA present in the cells of pollen can be used for the taxonomic identification of plant species by using standard DNA barcoding (for identifying a single-species sample), metabarcoding (mixed-species samples), or single-pollen genotyping (direct amplification and sequencing of individual pollen grains; Isagi & Suyama, 2011). Standard DNA barcoding approaches for plants typically use chloroplast DNA (cpDNA; e.g., rbc L, mat K, trn L; CBOL Plant Working Group et al., 2009) and/or nuclear ribosomal DNA (e.g., ITS1, ITS2; Chen et al., 2010). This approach has been extended to the identification of pollen, where DNA metabarcoding and high-throughput sequencing (HTS) are used to sequence PCR-amplified DNA of all the species in a mixed-species pollen sample (Bell et al., 2016; Cristescu, 2014; Taberlet, Coissac, Pompanon, Brochmann, & Willerslev, 2012). This approach can be applied in any standard molecular laboratory (assuming PCR product is subsequently sent off for HTS sequencing) and has been used extensively on bulk pollen extract for detecting plant-pollinator interactions through freshly collected pollinator specimens (e.g., Keller et al., 2015; Lucas et al., 2018) and museum specimens (e.g., Gous et al., 2019), identifying the floral composition of honey (e.g., Hawkins et al., 2015), and monitoring allergenic pollen (e.g., Kraaijeveld et al., 2015).
Compared to morphology-based identifications, DNA-based identifications of pollen provide several advantages: there are more people with the required laboratory and bioinformatics expertise; greater taxonomic resolution can be obtained (Lau et al., 2019); there is potential for higher throughput (Sickel et al., 2015); and it is possible to identify the entire pollen composition of a sample (e.g., pollen load carried by an individual pollinator), rather than being limited to a subsample that is tractable by microscopic identification. DNA metabarcoding still has some challenges, mainly in respect to assigning abundances (Bell et al., 2019) and species resolution, which depends on the gene region used for metabarcoding and the quality and completeness of the relevant reference database (Jones, Twyford, et al., 2021). In most cases, however, pollen DNA metabarcoding resolves taxonomy equally well or better than traditional morphological based methods (Keller et al., 2015; Leontidou et al., 2018; Macgregor et al., 2019). Pollen DNA metabarcoding data is generally considered to be semi-quantitative (i.e., sequence read counts are correlated with pollen grain numbers in a sample) (Baksay et al., 2020; Hawkins et al., 2015; Alexander Keller et al., 2015; Kraaijeveld et al., 2014; Marcel Polling et al., 2022; Richardson et al., 2018; Richardson et al., 2021), particularly for the most abundant taxa in a sample (Bänsch et al., 2020), however, quantification is dependent on factors such as study system and choice of gene region, as well as laboratory and bioinformatic methods (Berry, Mahfoudh, Wagner, & Loy, 2011; O’Donnell, Kelly, Lowell, & Port, 2016; Piñol, Senar, & Symondson, 2019; Richardson et al., 2018; Richardson et al., 2015). Ongoing method development in pollen DNA metabarcoding is likely to improve taxonomic resolution and quantification accuracy, and could eventually include intraspecific identifications (i.e., distinct subspecies or populations within a species). These developments could include using more of the genome (Bell, Petit, et al., 2021; Lang, Tang, Hu, & Zhou, 2019), using longer reads (Peel et al., 2019), and correcting for copy number of barcode gene regions (L. Garrido-Sanz, Senar, & Pinol, 2021).
While methods for pollen DNA metabarcoding are still evolving, it is evident that a molecular approach to pollen identification represents an important tool for understanding and monitoring ecosystems under global change. This paper will review the history of pollen DNA metabarcoding and related methods, discuss current applications of these methods, outline the basic requirements for a DNA-based pollen identification study, and provide a current assessment of progress on technical issues and future research directions.
History of pollen DNA metabarcoding
Research identifying species or genotypes of plants using DNA from pollen began in the 1990s and was based on Sanger sequencing of individual pollen grains. Given that conifers typically have male inheritance of plastids, early studies indicated that it should be easier to genotype conifer pollen using the ‘standard’ DNA barcoding gene regions (rbc L and mat K; CBOL Plant Working Group et al., 2009) given that they are on the plastid genome. However, in most angiosperm species, plastids are typically inherited maternally, and plastid DNA is less abundant after pollen maturation (Nagata, 1996). Thus, in the early days of pollen barcoding, researchers expected this method to work effectively on gymnosperm pollen, but not necessarily on angiosperm pollen. Suyama et al. (1996) were the first to amplify and sequence cpDNA of Abies (fir, a gymnosperm) pollen collected from Quaternary peat at Kurota Lowland, Fukui, Japan. Petersen, Johansen, and Seberg (1996) were the first to amplify short regions of cpDNA from single pollen grains of angiosperms (Hordeum and Secalegrasses). The technique of Suyama et al. (1996) was later used to analyse cpDNA from ancient pollen of conifers (Pinus sylvestris; L. Parducci, Suyama, Lascoux, & Bennett, 2005) and angiosperms (Fagus orientalis ; Paffetti et al., 2007) extracted from ancient sediments, Pinus pollen grains collected from a glacier (Nakazawa et al., 2013), and airborne pollen grains of Pinus (Ito, Suyama, Ohsawa, & Watano, 2008).
Today, we know cpDNA is present in angiosperm pollen since cpDNA gene regions (e.g., rbc L, trn L) have been successfully amplified from single pollen grains and bulk pollen samples in many studies. Single-pollen genotyping (MATSUKI, ISAGI, & SUYAMA, 2007; Suyama, 2011) remains labour intensive and a challenge for the research community. However, Isagi and Suyama (2011) have successfully used multiplex PCR and a single-pollen genotyping method on fresh pollen to conduct paternity analysis and to infer the pattern and distance of pollen dispersal in modern plant populations. The same technique was successively used in several studies by their groups (Hasegawa, Suyama, & Seiwa, 2009; Hirota et al., 2013; MATSUKI et al., 2007; Matsuki, Tateno, Shibata, & Isagi, 2008). Sanger sequencing has also been used with the cloning of PCR products (amplicons) to identify pollen from bulk samples (Bruni et al., 2015; Galimberti et al., 2014).
With the advent of HTS technology, DNA-based pollen identification is no longer dependent on the time-consuming isolation and sequencing of individual pollen grains (Aziz & Sauve, 2008; MATSUKI et al., 2007) or the cloning of amplicons prior to Sanger sequencing (Bruni et al., 2015; Galimberti et al., 2014). Instead, with HTS, researchers have been able to sequence pollen from bulk samples using DNA metabarcoding. This breakthrough has allowed for rapid, large-scale species identification of species within mixtures. Early proof-of-concept papers on pollen DNA metabarcoding demonstrated the feasibility of the method (e.g., Cornman et al., 2015; Hawkins et al., 2015; Alexander Keller et al., 2015; Kraaijeveld et al., 2014; Richardson et al., 2015) and it has since been applied to a range of applications. These include understanding the foraging behaviour of honeybees (e.g., Jones, Brennan, et al., 2021; Alexander Keller et al., 2015; Richardson et al., 2021; Richardson et al., 2015) and other pollinators (e.g., Bell, Batchelor, et al., 2021; Kratschmer, Petrović, Curto, Meimberg, & Pachinger, 2020; Lucas, Bodger, Brosi, Ford, Forman, Greig, Hegarty, Neyland, et al., 2018; MacGregor et al., 2019), examining historical flower visitation (Gous et al., 2019; Simanonok et al., 2021), monitoring allergenic pollen (Brennan et al., 2019; Kraaijeveld et al., 2014), biodiversity assessments (Leontidou et al., 2021; Johnson et al. 2021), determining the floral origin of honey (Hawkins et al., 2015; Milla et al., 2021; Khansaritoreh et al., 2020), and monitoring invasive species (Tremblay et al., 2019).
Advances and cost reductions in HTS and the advent of third generation sequencing technologies may further improve pollen-based DNA identifications. As the costs of HTS decrease, researchers are moving from traditional DNA metabarcoding, based only on a small number of gene regions, to metagenomics, which is based on whole genomes or reduced-representation genomes. Methods based on whole-genome shotgun sequencing of pollen mixtures, either using the plastid reads only (Lang et al., 2019) or all reads (Bell, Petit, et al., 2021), have shown improved taxonomic resolution and quantification over DNA metabarcoding, but still require the presence of suitable reference databases for identification. This is discussed in more detail in section 5.
Current and potential applications of pollen DNA metabarcoding and related methods in global change ecology
Since the early pollen DNA metabarcoding papers, several papers have published methodological improvements and proof-of-concept for a range of sample types. More recently, these methods have begun to be applied to ecological questions, including papers addressing questions related to global ecological change. To assess the current application of pollen DNA metabarcoding to questions of global change, we completed a Web of Science search (accessed 11/30/2021) with the terms: “pollen” and “metabarcoding” or “pollen” and “meta-barcoding”. From a list of 134 results, we excluded irrelevant papers, reviews, and those papers which focused solely on methods development. We also added several papers (n=28) to this list based on previous knowledge. Following these alterations, we examined a reduced list of 80 papers from 2014-2021 (Supplementary Table S1). Generally, we found increasing numbers of papers with time, and a shift in 2017 from predominantly proof-of-concept papers to those exclusively focused on answering an ecological question of interest (Figure 1). Research to date has been concentrated on pollinator foraging behaviour, with an early focus on honeybee foraging, likely due to the ease with which mass amounts of pollen can be collected from hives with pollen traps or from honey samples. However, applications of pollen DNA metabarcoding are varied and address, among many other topics, lepidopteran migration, historic foraging reconstruction, and airborne pollen monitoring.
Many applications of pollen DNA metabarcoding that address ecological change take advantage of the improved resolution and efficiency that these methods can provide. One specific area of ecological research is in the reconstruction of plant-pollinator interaction networks from the pollinator perspective. An important consideration when creating plant-pollinator interaction networks is the perspective that a given methodology provides. Historically, visitation networks have been plant-focused, with pollinators counted or collected from selected flowers. These observations can provide a complete interaction network for the flowers and are thus a good representation of pollinator visitation. However, such observations incompletely represent the dietary intake of the flower-visiting animal (Arstingstall et al., 2021; Popic, Wardle, & Davila, 2013). On the other hand, pollen-based methods, such as DNA metabarcoding or light microscopy, provide the animal perspective and allow exploration of the dietary input obtained from these interactions (Piko et al., 2021; Pornon, Andalo, Burrus, & Escaravage, 2017; Zhao et al., 2018). Pollen-based methods may also enable the assessment of plant-pollinator networks in hard-to-observe places, thus avoiding sampling biases. For example, observation-based detections are impractical in the forest canopy, but pollen analyses can capture bee foraging patterns (C. Smith, Weinman, Gibbs, & Winfree, 2019).
Another advantage of pollen-based methods, and an essential factor to be considered in the study design, is the scalability of the samples that can be used to infer and contrast preferences of individual foragers (Casanelles‐Abella et al., 2021; Elliott et al., 2020; Piko et al., 2021), and hive, colony, nest or species level assessments (Danner, Keller, Härtel, & Steffan-Dewenter, 2017; Nürnberger, Keller, Härtel, & Steffan‐Dewenter, 2019; Sickel et al., 2015). Individual-level assessments allow researchers to address potential intraspecific variation by having snapshots of foraging and immediate responses to spatiotemporal or anthropogenic changes (Piko et al., 2021). Longer-term samples provide comprehensive insights into the complete foraging spectrum and species’ dietary niche, co-evolution and long-term responses to changes (Kaluza et al., 2017; Vaudo, Biddinger, Sickel, Keller, & López-Uribe, 2020; Wilson et al., 2021). Resource partitioning and specialisation can be analysed throughout the entire network from the community to the individual (Brosi, 2016; Elliott et al., 2020; Lucas, Bodger, Brosi, Ford, Forman, Greig, Hegarty, Jones, et al., 2018). In addition, historical samples can be used as an input (e.g., from museum specimens or honey samples), which allows a direct comparison of foraging changes through extended time periods (Gous et al., 2019; Jones, Brennan, et al., 2021).
This “animal perspective” also directly links to the functional and nutritional components of pollen provided to animals and their offspring and can thus, for example, be used to relate impacts of changes in plant resource diversity to nutritional needs, development, and health (Donkersley et al., 2017; Trinkl et al., 2020). Recently, the importance of transmitting microbes between plants and pollinators by hitchhiking pollen grains, nectar and animal bodies have been implied (A. Keller et al., 2021; McFrederick & Rehan, 2016; Zemenick, Vannette, & Rosenheim, 2021). This accounts both for microbes with beneficial (Dharampal, Carlson, Currie, & Steffan, 2019; Vuong & McFrederick, 2019) and detrimental (A. Keller et al., 2018; Voulgari-Kokota, Steffan-Dewenter, & Keller, 2020) effects on host ecology, nutrition and health (Engel et al., 2016; Vannette, 2020; Voulgari-Kokota, Grimmer, Steffan-Dewenter, & Keller, 2018). Given these advantages of pollen-based networks and the high-throughput capabilities of genetic methods, a wide range of broad and fine-scale ecological questions may be answered. Particularly urgent are studies investigating how interactions change throughout space (across land-use gradients, resource availability) and time (short-term, e.g., seasonal, or long-term via the use of historical museum specimens) and elucidating consequences that accompany climate change and habitat disruption and other anthropogenic influences.
Understanding pollinator responses to land-use change
In recent decades, there have been broad scale shifts in land use with subsequent changes in resource availability to pollinators. Overall, there has been deforestation in the tropics but widespread reforestation and afforestation in temperate regions, often with monocultures or small numbers of tree species (Song et al., 2018). In the US, there has been considerable reforestation and urbanisation (He et al., 2019; Song et al., 2018), at the expense of grassland and herbaceous areas (Lark, Meghan Salmon, & Gibbs, 2015; Otto, Roth, Carlson, & Smart, 2016). Pollen DNA metabarcoding has been used to understand the flexibility of pollinator species dietary niche in response to such changing environmental conditions (Vaudo et al., 2020) and to evaluate and improve conservation efforts (Gresty et al., 2018; Piko et al., 2021). Land-use change can also alter the diversity of resources available to pollinators, with decreased floral richness in agricultural monocultures and context-specific increases or decreases with urbanisation (da Rocha‐Filho et al., 2021; Jones, Brennan, et al., 2021; Lucek et al., 2019; Richardson et al., 2021; Samuelson, Gill, & Leadbeater, 2020). Pollen DNA metabarcoding has been used to understand the response of pollinators to resource-rich and resource-poor environments (Casanelles‐Abella et al., 2021; Danner et al., 2017; Kaluza et al., 2017; Sponsler et al., 2020; Wilson et al., 2021) and to understand the link between behaviour and pollen intake in resource-poor environments (Nürnberger et al., 2019).
Expansions in agricultural land cover can have a range of impacts on pollinators (Potts et al., 2010), including decreasing floral resource diversity (Grab et al., 2019; Richardson et al., 2021), increasing pesticide exposure risk (Douglas, Sponsler, Lonsdorf, & Grozinger, 2020; Douglas & Tooker, 2015), and increased parasite loads (Cohen et al., 2021). Pollen DNA metabarcoding can show how pollinators respond to this changing agroecosystem and can be used to monitor changes. For instance, a comparison of modern-day honey with samples collected 65 years prior demonstrated shifts in honeybee forage composition and a reduction in white clover, Trifolium repens (Jones, Brennan, et al., 2021). Recent studies, using both molecular and morphological pollen identification, have also found that honeybees situated in modern agricultural landscapes tend to collect a lower diversity of forage relative to nearby non-agricultural landscapes (Richardson et al., 2021; Samuelson et al., 2020). To date it is unclear how this variation in forage diversity corresponds to pollinator health. Pollinator diets can also be affected through the uptake of agricultural environmental schemes (AES) in which supplemental planting can improve forage availability for pollinators on farms. Molecular analysis of bee collected pollen can help evaluate if conservation plantings are successfully supporting bee populations and identify which resources bees forage on (Gresty et al. 2018; McMinn-Sauder, 2020). Overall, this research demonstrates how pollinators can be used to monitor changes associated with agricultural land use and that there is a need for continued study. Future research efforts in this area will require greater spatial replication and comparisons to a wider diversity of alternate landscape types, two challenges to which molecular identification of pollen is uniquely suited.
While both agricultural and urban intensification can negatively impact pollinator communities through habitat loss (Potts et al., 2010), many urban environments also support rich pollinator diversity (Baldock et al., 2015; Hall et al., 2017), especially when landscapes exhibit moderate or intermediate levels of urbanisation (Wenzel, Grass, Belavadi, & Tscharntke, 2020). In part, pollinators can thrive in cities due to the intensive cultivation of flowering resources (Baldock et al., 2015; Sponsler et al., 2020) and spontaneous urban vegetation (i.e., weeds) (Lowenstein, Matteson, & Minor, 2018; Turo & Gardiner, 2019). However, cities also filter pollinator communities and select for specific traits such as cavity-nesting and generalist foraging (Wenzel et al., 2020). Recent studies have focused on the management and improvement of urban green spaces (e.g., gardens, vacant lots, road verges, green roofs) to increase the diversity of urban pollinators, particularly for rare and threatened bee species (Threlfall et al., 2015; Turo & Gardiner, 2019). Improved characterisation of urban foraging networks can inform green space development. However, traditional methods for monitoring plant-pollinator interactions (e.g., flower-visitor observations, hand-netting) can be challenging to use in urban areas due to restrictions on sampling private property and the high floral diversity present in the urban matrix (Sponsler et al., 2020). Recently, pollen DNA metabarcoding has been used to characterise plant-pollinator interactions and pollinator diets in urban environments (Casanelles‐Abella et al., 2021; Potter et al., 2019; Sponsler et al., 2020). The findings of this research suggest that pollen DNA metabarcoding can be a useful tool to investigate how urban insect pollinators partition their diet from available floral resources (especially native and non-native forage). Pollen DNA metabarcoding can also evaluate how urban habitat plantings influence bee foraging and subsequent population growth (Potter et al., 2019).
Monitoring and surveillance of ecosystem change
Identifying plant biodiversity from pollen provides a valuable tool for surveillance of changes in the ecosystem, which has been applied to the early detection of invasive species and diseases (Tremblay et al., 2019), and could potentially be used for monitoring for the presence of rare, threatened, or endangered species. Community-level monitoring of plant biodiversity through pollen DNA metabarcoding could track changes in species composition, range, and phenology, providing a high-throughput alternative to botanical surveys (Johnson et al., 2021; Leontidou et al., 2021; Milla, Bovill, Schmidt-Lebuhn, & Encinas-Viso, in press). These studies could be particularly beneficial when combined with other high-throughput methods like remote sensing or unmanned aerial vehicle surveys (Ancin-Murguzur, Munoz, Monz, & Hausner, 2020). There is potential for pollen-based monitoring to be closely linked to management, with unexpected detections or non-detections to trigger a management response. Methods that would enable this technology need to be developed to provide probabilistic confidence estimates for identifications and avoid false positives triggering unnecessary management responses. The development of methods for the surveillance of aquatic invasive species via eDNA has been discussed (Darling, Pochon, Abbott, Inglis, & Zaiko, 2020; Sepulveda, Nelson, Jerde, & Luikart, 2020), and the same issues would apply to pollen-based surveillance methods.
Comparison to past ecosystems
The ability to assess global ecological change often relies on the comparison of contemporary data to historical data. Pollen identification is potentially a particularly useful tool for studying ecological change based on its presence on historical animal specimens and its preservation in ancient sediments. This type of sampling enables the comparison of contemporary to past ecosystems and establishes more accurate baselines for conservation. Preserved pollinating insects in museum collections often have pollen on their bodies or corbiculae, and these pollen grains can be identified through DNA metabarcoding to provide information on historical insect foraging patterns (Gous et al., 2019). For example, pollen DNA metabarcoding of a century of foraging by the endangered rusty patched bumble bee found that decline was unlikely to be driven by changes in forage (Simanonok et al., 2021). Similar studies are expected to give insights on the past foraging behaviour of other pollinating species and inform processes associated with more recent declines.
Pollen preservation in ancient sediments, in combination with ancient sedimentary DNA (Capo et al., 2021), also provides a resource for understanding past ecosystems. Usually, the pollen grains are examined morphologically, while the sediments are analysed through DNA sequencing to provide complementary data sources (Liu et al., 2021; Laura Parducci et al., 2017). This approach has been used to determine conservation baselines for offshore islands in New Zealand (Wilmshurst et al., 2014). Pollen retrieved from lake sediments is theoretically an ideal material for ancient DNA analyses in both conifers and angiosperms: depositional conditions are fast and reduce physical damage of the grains; burial is rapid, reducing the exposure of the grains to biotic degradation (L. Parducci, Nota, & Wood, 2019). By accessing the DNA in ancient pollen, there is potential for further information to be obtained on past ecosystems (Niemeyer et al., 2017). This is an area where further method development could prove useful (see section 5).
Monitoring impacts on human health
Changes in pollen abundance and distribution due to climate change will likely have severe impacts on human health (Anderegg et al., 2021; Kurganskiy et al., 2021). Airborne pollen sampling, combined with DNA metabarcoding, allows allergenic species to be monitored across large spatiotemporal scales (Brennan et al., 2019; Leontidou et al., 2018) and can identify seasonal changes in allergenic species (Campbell et al., 2020; Uetake et al., 2021). Many plant families with highly allergenic pollen are difficult to identify through pollen morphology, however, the increased taxonomic resolution provided by DNA metabarcoding can allow allergenic species to be distinguished from non-allergenic species (e.g., Poaceae; Brennan et al., 2019; Urticaceae; Marcel Polling et al., 2022). The relationship between the presence of specific pollen types and human health responses can be investigated to identify the most harmful species (Rowney et al., 2021). Airborne pollen monitoring has been occurring for many years, and successful amplification of DNA from historical microscope slides shows the potential to monitor long-term ecological changes (Marcel Polling et al., 2022).