Pollen report: quantitative review of pollen crude protein concentrations offered by bee pollinated flowers in agricultural and non-agricultural landscapes

Introduction

The green agricultural revolution during the mid-20th century drastically increased productivity in agriculture and changed land use on a global scale (Evenson & Gollin, 2003). The combination of high-yield crop varieties, chemical fertilizer, plant protection products and intensified mechanization have amplified crop biomass production, which in turn has enabled the support of an ever-growing human population (Evenson & Gollin, 2003; Pingali, 2012). At the same time, the associated reduction in plant diversity in intensified agricultural habitats (e.g., large scale mass flowering crop-cultures) has been suggested to adversely affect pollinator populations which provide essential ecosystem services (Goulson et al., 2015; Potts et al., 2010; Roulston & Goodell, 2011). In particular, bees (Hymenoptera: Apoidea), a diverse group of primarily phytophagous insects present in most major habitats around the globe (Michener, 2000; Wcislo & Cane, 1996), have recently come into focus because they provide a large share of pollination services in agricultural habitats and some populations are apparently in decline (Goulson et al., 2015). Bees rely solely on plant derived resources (pollen and nectar) to satisfy their nutritional needs (Brodschneider & Crailsheim, 2010; Michener, 2000; Roulston & Goodell, 2011), which can be problematic in agricultural landscapes because bee species foraging outside the restricted flowering period of pure culture dominated agricultural settings might be deprived of adequate alternative food sources. This phenomenon has been termed nutritional mismatch and has been suggested as one potential direct driver for the apparent decline in some bee populations (Vaudo et al., 2015).

In order to ease nutritional stress on bee populations in modern agricultural settings the establishment of alternative foraging habitats has been incentivized via agro-environmental management schemes in the EU and elsewhere (Goulson et al., 2015; Lye et al., 2009; Phillips & Lowe, 2005; Potts et al., 2015; Vaughan & Skinner, 2008). Such schemes seek to provide bees with flower resources outside the mass flowering periods of commercial crops in additional supplement to potential alternative resources (e.g., flowering plants in the field margins), but have traditionally focused solely on providing plants to attract and sustain social bees, particularly bumblebees (Vaudo et al., 2015; Wood, Holland & Goulson, 2017). Only recently has the important role of solitary bees in this context been recognized (Scheper et al., 2015). However, in all cases increasing the quantity of flowering plants and the lengthening timing of flowering alone is likely insufficient to maintain healthy bee populations. The quality of floral resources (including sugar concentration in nectar and protein content of pollen) and its natural variability also plays a major role in bee health with direct consequences, at least for the fitness of social bees (Brodschneider & Crailsheim, 2010; Di Pasquale et al., 2013; Roulston, Cane & Buchmann, 2000; Tasei & Aupinel, 2008; Vaudo et al., 2016; Vaudo et al., 2015). Such qualitative aspects of bee nutrition should be taken into consideration to develop a complementary and nutritionally optimized resource base for bee populations in agricultural landscapes (Vaudo et al., 2015).

To facilitate the integration of flower resource quality in pollinator management schemes we compiled a geographically explicit data base of pollen quality (measured as crude protein content) offered by bee-visited flowers in an agricultural setting. Given that pollen is the main protein source for bees’ offspring, crude protein concentration in pollen is directly linked to the amount of protein bees can extract from their habitat and has traditionally served as a proxy for pollen quality (Roulston, Cane & Buchmann, 2000; Roulston & Goodell, 2011; Vaudo et al., 2015). However, most of the evidence supporting the importance of crude protein concentration for the fitness of bees (Brodschneider & Crailsheim, 2010) and their ability to adjust their individual (Ruedenauer, Spaethe & Leonhardt, 2015; Ruedenauer et al., 2018) or collective response according to their protein requirements (Fewell & Bertram, 1999; Pernal & Currie, 2001) stems from social bees. In the case of solitary bees the picture is less clear (Eckhardt et al., 2014; Haider, Dorn & Müller, 2013; Nicholls & Hempel de Ibarra, 2017; Vanderplanck et al., 2014), but in many cases solitary bees will also likely benefit from increased protein availability. Only in recent years additional factors, such as amino acid composition, lipids and potentially secondary plant metabolites, have emerged as variables potentially shaping pollen quality and consequently foraging decisions for bees in particular for solitary ones (Cook et al., 2003; Nicholls & Hempel de Ibarra, 2017; Palmer-Young et al., 2019; Sedivy, Müller & Dorn, 2011; Vaudo et al., 2016). In contrast to the extensive literature on pollen crude protein content (Roulston, Cane & Buchmann, 2000), data on these emerging quality factors are still scarce and often ambiguous and were consequently not included in this first analysis. However, it would be a logical and important next step to merge these data with information on secondary quality characteristics and bee ecology to utilize their full potential.

To give an overview of the collected data and potential applications we used the generated database to compare the crude protein concentration of pollen resources bees can encounter in agricultural landscapes in- (crop and weeds) and off-field (wild plants) around the globe. In a second step we identified crops, which provide bees with low, likely sub-optimal pollen and protein and suggest plants which could serve as high-quality alternative protein sources during times of need.

Materials and Methods

Results

In total, we found 316 measurements of percent crude protein content, of which 302 could be unambiguously attributed to one plant category (see materials and methods) and used for analysis. Protein concentrations ranged from 10–61%, with the majority of these measurements from wild flowers (N = 127, crop N = 94, weed N = 81). In general, the data are evenly spread across regions and plant categories (Table 1), except only limited data were available for wild plants in Africa (N = 14). Overall, we find small differences in the median protein levels between communities with crops having slightly lower concentrations (median_crop = 25.2%) than wild species (median_wild = 28.5% see Fig. 1). When looking at the regional scale we find this variation reflected in the New World communities (North America and South America see Fig. 1). In all other regions such variation seems less pronounced or absent. When looking at genus level differences based on all observations we find pronounced variation between genera (Fig. 2), with two genera providing particularly high-quality pollen (Solanum and Senna) and two genera representing globally important crops (Zea and Helianthus) offering pollen with low protein content (Fig. 2). When comparing these results with the pattern based on species/variety means we find only minor deviations between the two analysis and no change in the overall genus level patterns (Table S1).

Discussion

Overall crude pollen protein content in bee-visited plants is around 26% with similar values for all categories (Fig. 1, Table 1). The only apparent deviation from these values appear in the wild plant communities of the New World (median North America = 38.3% and median South America = 40.4%), while plant communities in the other regions have comparable pollen protein concentrations (Fig. 1, Table 1). We found that only six genera offer pollen with low protein concentration (crude protein < 20%), including two main crop species: sunflowers (Helianthus: median around 15%) and maize (Zea: median around 16%). Most other plant genera offer pollen of comparable protein concentration (Fig. 2).

Interestingly, protein content in wild plants of the new world appear somewhat elevated while crop and weed species exhibit similar protein levels worldwide. In the latter case this is expected as these plant communities are more homogenous in all regions as a result of their intended (crop) or involuntary introduction (weeds) around the globe. In contrast, most wild plants in this study are geographically restricted (non-global distribution), suggesting that this potential pattern could be caused by differences in the community composition between regions. However, caution needs to be taken as this pattern could also be the result of incomplete sampling (e.g., in Africa) or could represent sampling bias because the results are based on peer-reviewed publications and could simply reflect the focus of the study authors.

Two genera (Solanum and Senna) were identified with higher pollen protein concentration than the majority of plant genera visited by bees (Fig. 2). Interestingly both genera possess poricidal anthers, which naturally restrict pollen access by non-specialized pollinators and potential pollen robbers (De Luca & Vallejo-Marín, 2013). The resulting reduced risk of pollen loss might have contributed to the evolution of increased pollen quality. Most genera have similar pollen protein concentrations, which are at a level likely suitable to support bee populations (Day et al., 1990; Di Pasquale et al., 2013; Roulston, Cane & Buchmann, 2000; Tasei & Aupinel, 2008; Wcislo & Cane, 1996). In addition, our findings support the conclusion that maize and sunflower have low pollen quality in terms of protein content (Hass et al., 2019; Maurizio & Grafl, 1980). While sunflower can offer high quality nectar as an alternative reward (Maurizio & Grafl, 1980; Tepedino & Parker, 1982) to attract bees, maize is primarily wind pollinated and does not offer nectar rewards. Therefore, sunflower, and in particular maize, is considered less attractive for bees and are likely only occasionally visited in the absence of alternative pollen sources (McGregor, 1976). These results suggest that it might be of particular interest to supply bees in sunflower- and maize-dominant agricultural landscapes with high quality pollen species, adjusted for the season and region (Hass et al., 2019).

Given that pollen is the primary protein source for the majority of bees it is important to ensure adequate protein supply when planning alternative foraging areas (Vaudo et al., 2015). While it is clear that increased protein supply can be beneficial to developing larvae and insufficient protein supply can result in larval malnutrition with clear adverse effects (Brodschneider & Crailsheim, 2010; Di Pasquale et al., 2013; Hass et al., 2019; Tasei & Aupinel, 2008), there is only limited support for a simple relationship between crude protein concentration and bee fitness (Babendreier et al., 2004; Brodschneider & Crailsheim, 2010; Di Pasquale et al., 2013; Tasei & Aupinel, 2008). It is likely that in addition to protein content, other quality markers such as lipid content, amino acid composition and secondary plant metabolites might play an important role in determining pollen quality for bees (Day et al., 1990; Hass et al., 2019; Maurizio & Grafl, 1980; Nicholls & Hempel de Ibarra, 2017; Nicolson & Human, 2013; Pernal & Currie, 2000; Ruedenauer et al., 2018; Tasei & Aupinel, 2008). In contrast to nectar quality (sugar concentration), there is only some evidence that bees can reliably separate high from low quality pollen and adjust their collecting behavior according to their needs (Nicholls & Hempel de Ibarra, 2017). However, recent work suggest that bumblebees are able to do this impressive feat on an individual level (Ruedenauer, Spaethe & Leonhardt, 2015) and that honeybees can on a collective level (likely using feedback from their larvae; Pernal & Currie, 2001; Ruedenauer et al., 2018) if quality differences are sufficiently large. These promising findings suggest that by adding selected high quality pollen resources to agricultural landscapes social bees would likely benefit directly, while at least some solitary bees will likely be able to utilize such additional resources (Hass et al., 2019; Scheper et al., 2015).

Conclusions

This paper is a first step to collect the available data on pollen quality, in terms of protein content, offered to bee visited flowers in agricultural habitats and make them easily accessible in an electronic format. In the future this dataset could be combined with more detailed information on both pollen quality (e g. lipids, amino acid composition and secondary plant metabolites) as well as plant traits (e.g., flowering period and local geographic distribution), which could enable improved bee management practices and potentially more realistic landscape level modelling approaches to facilitate bee conservation in modern agricultural habitats.

Supplemental Information