CONTRASTING POLLINATION EFFICIENCY AND EFFECTIVENESS AMONG FLOWER VISITORS OF MALVA SYLVESTRIS, BORAGO

Biotic pollination is an important factor for ecosystem functioning and provides a substantial ecosystem service to human food security. Not all flower visitors are pollinators, however, and pollinators differ in their pollination performances. In this study, we determined the efficiencies of flower visitors to the plant species Malva sylvestris, Borago officinalis and Onobrychis viciifolia by analysing stigmatic pollen deposition. We further calculated pollinator effectiveness by scaling up single-visit pollen deposition using visitation frequency. Flowervisitor groups differed in their efficiencies at the single-visit level and not all of them deposited more pollen compared to unvisited stigmas. Bumblebees tended to be most efficient in depositing pollen per single visit across the three plant species. Due to the by far highest visitation frequencies, Apis mellifera showed the highest effectiveness in depositing pollen per hour for M. sylvestris and B. officinalis, but not for O. viciifolia, for which the Bombus lapidarius complex was both the most frequent and the most effective pollinator group. Hence, the most frequent flower visitors were most effective in our study. For non-dominant pollinator groups, however, visitation frequencies contributed disproportionally to pollinator effectiveness. Thus, combining pollen deposition per singlevisit with visitation frequency is necessary to reveal true pollinator performance and to better understand flowervisitor interactions.


INTRODUCTION
Pollen transfer is an essential process for seed and fruit production in sexually reproducing plants and around 90% of all angiosperm species are animal pollinated ). Biotic pollination is therefore an important factor for wild plant reproduction  and an important ecosystem service to human food security (Gallai et al. 2009;Calderone et al. 2012). Despite the importance of biotic pollination, many obscurities about plant-pollinator interactions still need to be solved (Mayer et al. 2011). For example, the terms "flower visitor" and "pollinator" are often used synonymously, without actual proof of pollen transfer (Ne'eman et al. 2010). In addition, performances as pollinators have been shown to differ among flower visitors (Ne'eman et al. 2010;Jędrzejewska-Szmek & Zych 2013;Popic et al. 2013;Ballantyne et al. 2015). This general gap in knowledge complicates finding appropriate pollinators for agricultural systems (Slaa et al. 2006), interpreting specialisation and generalisation (Jędrzejewska-Szmek & Zych 2013;Popic et al. 2013;Ballantyne et al. 2015) and predicting ecological and economic consequences of pollinator loss or invasions (Goulson 2003).
Insects, and above all bees, are considered the most important pollinators (Buchmann & Nabhan 1996). Pollen is transferred passively while insects utilise flowers for foraging, as shelter or mating site, among other reasons (Inouye et al. 1994). Not all flower visitors, however, are involved in pollination. Reasons are mismatches in morphology or behaviour or illegitimate exploitations of floral resources (i.e. nectar and pollen thieves or robbers) (Inouye 1980). Therefore, pollination services of a given flower visitor may differ among plant species and pollination success for a given plant species differs among pollinators (Fenster et al. 2004;King et al. 2013;Popic et al. 2013;Ballantyne et al. 2015).
Both indirect (e.g. visit duration, visitor frequency, pollen removal, stigma receptivity) and direct measurements (e.g. stigmatic pollen deposition, seed set) have been used to analyse pollination performances of flower visitors (summarised by Ne'eman et al. 2010), but previous research suggests that direct measurements are more reliable (Johnson & Steiner 2000;Adler & Irwin 2006;King et al. 2013;Ballantyne et al. 2015). Therefore, to analyse pollination performances, distinguish flower visitors from true pollinators and determine the strength of plant-visitor interactions, it seems crucial to investigate the transfer of pollen by flower visitors more closely for multiple plant species.
In this study, we analysed the performance of flower visitors to the three plant species Malva sylvestris, Borago officinalis and Onobrychis viciifolia directly by determining stigmatic pollen deposition. The three plant species are visited by a variety of insect species. Yet, it is not known, whether all flower visitors are pollinators and whether pollinators differ in their pollination performances when direct measurements are used. We first compared pollen loads on unvisited and visited stigmas to evaluate pollination performance of abiotic factors versus flower visitors overall.
Then, pollination performances of flower visitors to M. sylvestris, B. officinalis and O. viciifolia were analysed in a hierarchical approach: (1) A flower-visitor group was identified as efficient pollinator when more pollen grains were deposited on stigmas per single visit compared to pollen loads on unvisited stigmas. (2) Pollinators were ranked regarding their pollination efficiencies by comparing single-visit pollen depositions. (3) Pollinators were ranked regarding their pollination effectiveness by comparing their pollen depositions per hour, including visitation frequencies (Fumero-Cabán & Meléndez-Ackerman 2007;Madjidian et al. 2008;Rader et al. 2009).

Study site and plant species
The study was conducted at the research farm "Oberer Hardthof" of the University of Giessen, Germany, which is surrounded by farmland and a small area of woodland. Four flower mixtures ("Lebensraum1", "Odin1", "Odin2", and "Veitshöchheimer Bienenweide") were sown on former arable land on 28 May 2013 in a block design, consisting of a row of eight blocks with four plots (4 m × 4 m) per block. Within each block, plots of the four flower mixtures were randomised and separated from each other by 3.5 m-wide strips of grassland. For determining pollination performances of flower visitors, we chose the three plant species M. sylvestris, B. officinalis and O. viciifolia, which were present in all (M. sylvestris and O. viciifolia) or in three (B. officinalis) of the four flower mixtures sown at the study site. Differences in pollinator performance between seed mixtures were not tested, because plant-community composition data did not allow for meaningful hypotheses.
Malva sylvestris, B. officinalis and O. viciifolia are frequently applied in flower mixtures, sown as flower strips in agricultural landscapes, and are cultivated as crops or for pharmaceutical purposes (Janick et al. 1989;Gasparetto et al. 2012;Hayot Carbonero et al. 2011). Malva sylvestris is native to Germany and the other two species are considered to be neophytes. The common mallow M. sylvestris (Malvaceae) generates actinomorphic, dish-shaped and upturned pinkish-purple flowers with at least 11 ovules per flower (Kumar et al. 2014). The central style divides into numerous filamentous stigmas and is surrounded by several anthers at its base (Kumar et al. 2014; Appendix I). Flowers of borage B. officinalis (Boraginaceae) are actinomorphic, dish-shaped and downward directed and change from pink to blue as they age. There are four ovules per flower (De Haro-Bailón & Del Rio 1998). The style ends in a terminal stigma and is surrounded by cone-like anthers at its base (Ghorbel & Nabli 1998; Appendix I). The common sainfoin O. viciifolia (Fabaceae) produces inflorescences with up to 120 white to pink, papilionaceous flowers (Goplen et al. 1991) with a single ovule per flower (Galloni et al. 2007). The stigma and the anthers are enclosed within the two fused and boat-shaped keel petals (Appendix I).
The three plant species produce pollen and large amounts of nectar. Nectaries are located at the base of the corollas and are indicated by prominent nectar guides in M. sylvestris and O. viciifolia (Westrich 1989). The plant species are considered self-compatible, but flowers show features to avoid self-pollination or self-fertilisation, i.e. spatial and temporal separation of mature stigmas and anthers (M. sylvestris, B. officinalis) (Montaner et al. 2000;Kumar et al. 2014) or presence of a stigmatic cuticle (O. viciifolia) (Galloni et al. 2007). Given these floral traits and observational data on flower visitors, the three plant species are assumed to rely mainly on insects such as bees (M. sylvestris, B. officinalis, O. viciifolia), hoverflies and butterflies (M. sylvestris) for pollen transfer between flowers of the same (geitonogamy) or other conspecific plants (allogamy) (Richards & Edwards 1988;Corbet et al. 1991;Goplen et al. 1991;Comba et al. 1999;Kumar et al. 2014).

Pollination efficiencies of flower visitors
We analysed the pollination efficiencies of flower visitors to M. sylvestris, B. officinalis and O. viciifolia directly by determining stigmatic pollen deposition per single visit. Sampling took place between 7:30 a.m. and 6:30 p.m. on warm and sunny days with low wind speeds from 3 June to 4 July 2014 (on 19 days for M. sylvestris, 15 days for B. officinalis and 13 days for O. viciifolia). Stigmas were sampled from all 32 plots (M. sylvestris, 1055 stigmas), from 16 plots representing all flower mixtures (O. viciifolia, 411 stigmas) and from 11 plots representing two flower mixtures (B. officinalis, 252 stigmas). Flowers of the three plant species were bagged before blooming (perforated polypropylene bags, holes 2 mm in diameter) and uncovered after flowers had fully opened and stigmas had turned receptive. We photographed the first flower visitor from an appropriate distance to prevent disturbances and removed the stigmas after the first visitor had left. Only flower-visitor groups with at least eight visited stigmas were further analysed. At the same time, receptive stigmas from bagged flowers were sampled as controls before any visit took place to account for pollen transfer due to self-or windpollination, animals smaller than 2 mm, which could enter the bags, or the experimental handling of the flowers. Sampled stigmas were stored in individual wells of plastic 96 cell-culture arrays in a freezer (-20 °C) until pollen analysis.
For counting attached pollen grains, visited and unvisited stigmas were unfrozen and mounted in glycerin-jelly slides stained with basic fuchsine following the protocol of Kearns J Poll Ecol 21 (1) & Inouye (1993) to increase the contrast (Appendix I).
Morphologically conspecific pollen grains attached to stigmas were counted by light microscopy (between ×40 and ×400). We identified the photographed flower visitors and grouped them according to size, coloration and/or taxonomy. Flower-visitor groups were considered to be efficient pollinators when pollen loads on visited stigmas exceeded those on unvisited stigmas. Pollinator efficiencies (i.e. mean numbers of pollen grains deposited per single visit) were calculated by subtracting the mean numbers of pollen grains found on unvisited stigmas, serving as controls, from mean numbers of pollen grains on stigmas visited by the flower-visitor groups for each plant species.

Visitation frequencies and pollinator effectiveness
For estimating pollinator effectiveness in terms of pollen deposition per hour, visitation frequencies of flower visitors to the three plant species were recorded during separate observations. Observations took place between 7:30 a.m. and Per-hour visitation frequencies were calculated by dividing the total number of visits recorded for a flowervisitor group by the total number of observational hours and the number of flowers simultaneously observed. Pollinator effectiveness (i.e. mean numbers of pollen grains deposited per hour) was calculated only for flower-visitor groups being efficient pollinators per single visit. For this, the mean number of pollen grains deposited per single visit by each pollinator group was multiplied with its corresponding visitation frequency.

Statistical analysis
Wilcoxon rank-sum tests were used to test for differences in the numbers of pollen grains on visited and unvisited stigmas for each plant species (function "wilcox.test" in the stats package of R version 3.2.3; R Core Team 2015). To differentiate flower visitors from pollinators and rank flower-visitor groups according to their pollination efficiency, numbers of pollen grains on unvisited stigmas and on stigmas visited by the different flower-visitor groups were compared by Poisson generalized linear models (GLMs) with a correction for overdispersion (i.e. family was set to quasipoisson) for each plant species (function "glm" in the stats package of R version 3.2.3). To provide levels of uncertainty in pollinator effectiveness (a product of the mean pollen deposition and a constant visitation frequency), we calculated standard errors by multiplying the error of pollen deposition efficiencies with the corresponding visitation frequency.

Pollination efficiencies of flower visitors
Visited stigmas had significantly increased pollen grain numbers compared to unvisited stigmas for each plant species, the mean increase ranging from over two-fold (B. officinalis) to over three-fold (O. viciifolia) and over fourfold (M. sylvestris) (Tab. 1). Stigmas sampled from the three plant species were visited mostly by bees, with other insects being less common.
For analysing pollen deposition, enough (eight or more) visited stigmas from M. sylvestris were obtained for four flower-visitor groups: Apis mellifera, the Bombus lapidarius complex (mainly B. lapidarius, but possibly also including the similarly coloured species B. ruderarius, B. pratorum, B. soroeensis), the B. terrestris complex (mainly B. terrestris or B. lucorum, but possibly also including the similarly coloured species B. magnus, B. cryptarum) and Halictidae (including only species of the genera Halictus and Lasioglossum smaller than 10 mm) (Tab. 2; Appendix II). Pollen deposition for each group of flower visitors was always significantly larger than pollen loads on unvisited stigmas (Tab. 2). Mean 2.7 ± 0.2*** * P < 0.05; ** P < 0.01; *** P < 0.001. P-values were calculated using Wilcoxon rank-sum tests and indicate differences between the mean numbers of pollen grains on visited compared to unvisited stigmas.

Pollination efficiencies of flower visitors
Overall, flower visitors deposited more pollen grains than wind or other vectors, supporting the pollinator dependencies of M. sylvestris, B. officinalis and O. viciifolia described in the literature (Richards & Edwards 1988;Corbet et al. 1991;Goplen et al. 1991;Comba et al. 1999;Kumar et al. 2014 assuming that within-flower autogamous pollination is minimised by floral traits of the three plant species, these bees were visiting other conspecific flowers before, thereby taking up pollen that stayed available for the pollination of subsequent flowers (Inouye et al. 1994;Ne'eman et al. 2010  Variation in pollen deposition efficiency occurs due to differences in the degree to which flower and visitor traits match, defining a visitor's ability to take up and deposit pollen grains. Because possible factors determining pollen transfer, such as body size (Willmer & Finlayson 2014), hairiness, tongue length (Hobbs et al. 1961), floral constancy or preference (Waser 1986), pollen-transporting structures and nectar-and pollen-collecting behaviours (Inouye 1980;Thorp 1999;Michener 2007) are highly diverse in bees, generalisation about pollination performances is difficult (Fenster et al. 2004). Nevertheless, generalist bumblebee groups (Goulson & Darvill 2004) were the most efficient (for M. sylvestris and B. officinalis) or among the most efficient pollinators (for O. viciifolia) across the three plant species in our study. This supports that bumblebees can often be more efficient per single visit across plant species of different families than honeybees (Wilson & Thomson 1991;Willmer et al. 1994;Javorek et al. 2002;Ballantyne et al. 2015;Zhang et al. 2015).
Bumblebees are characterised by a large surface due to their large and hairy bodies, which probably increases the probability to touch anthers and stigmas and the amount of pollen transferred. Bumblebees' long tongues (Goulson et al. 2005) make them better pollinators of flowers with deep corollas (Hobbs et al. 1961) and probably induced positioning on the studied flowers favourable for contact with anthers and stigmas. Furthermore, like for other Fabaceae (Córdoba & Cocucci 2011), bumblebees were strong enough to open flowers of O. viciifolia and thus could reach the nectar hidden between the base of the banner and keel petals. By pushing their heads and tongues towards the nectar, keel petals folded down while stigma and anthers were released and touched the bumblebees' ventral side, enabling pollination. In B. officinalis dehiscence of anthers is introrse and pollen grains are held within the cone-like anthers (Corbet et al. 1988). This controls pollen removal by flower visitors, since large amounts of pollen are released only in response to sonication (Buchmann 1983;De Luca & Vallejo-Marín 2013). Bumblebees are well known to perform buzz pollination (Buchmann 1983;King & Buchmann 2003). They were observed to sonicate on B. officinalis in this study (see also Corbet et al. 1988), which probably increased the amount of pollen released and transferred to subsequent flowers in B. officinalis.
Other bees analysed in this study were generally smaller, less hairy and had shorter tongues, probably causing morphological and behavioural mismatches, which restricted pollen transfer compared to bumblebees in most cases. Bees of the family Halictidae seemed of inappropriate size to flowers of all three plant species, not touching stigmas while drinking nectar, but rather occasionally on their way to the nectaries or when foraging on pollen. Honeybees had to crawl underneath stigmas of M. sylvestris to reach the nectar or inserted the tongue laterally into flowers of M. sylvestris and O. viciifolia, reducing the contact with anthers and stigmas (see also Comba et al. 1999). Furthermore, honeybees were observed to intensively groom pollen from their bodies and discard it after visiting flowers of M. sylvestris, making the pollen unavailable for transfer (Inouye et al. 1994). In addition, sonication is not performed by bees of the genus Apis (Buchmann 1983;King & Buchmann 2003) and only very rarely within the Megachilidae (Neff & Simpson 1988), probably reducing their pollination efficiency on B. officinalis. In contrast to the literature (Richards & Edwards 1988;Goplen et al. 1991), A. mellifera was not a successful pollinator of O. viciifolia in this study, probably because closed flowers of O. viciifolia could not be opened easily, as has been shown for other Fabaceae (Córdoba & Cocucci 2011). Bees of the genus Megachile, however, have relatively long tongues and were strong enough to trigger the lever mechanism of O. viciifolia, as shown for other papilionate flowers (Córdoba & Cocucci 2011), explaining a pollen deposition efficiency similar to the ones of the bumblebee groups.
In summary, A. mellifera and Halictidae on M. sylvestris and A. mellifera on B. officinalis were pollinators of the respective plant species, although much less efficient than the bumblebee groups. According to the observed behaviours, the inefficient pollinators can be classified as (pollen or nectar) thieves or base workers instead (sensu Inouye 1980), feeding on floral resources without pollinating. All analysed flower-visitor groups of M. sylvestris were efficient pollinators, whereas two groups of B. officinalis and O. viciifolia were inefficient pollinators. This may reflect an increasing specialisation of the floral morphology from M. sylvestris to B. officinalis to O. viciifolia, which hinders pollination by less suitable flower visitors.
The efficient pollinators deposited all at least as many pollen grains on stigmas as ovules present per flower (see Material and Methods). Hence, all efficient pollinators can potentially provide pollination services for maximum seed set of the plant species studied in our system. Differences in foraging behaviour within a plant or within single flowers may still affect conspecific pollen grain quality with regard to viability or the degree of kinship, both affecting germination success (Snow et al. 1996). Similarly, low flower constancy increases transfer of heterospecific pollen (Waser 1986), which could lead to stigma clogging and prevent conspecific pollen to germinate. Thus, including fertilisation, e.g. seed set and germination rates, is an important next step to determine the required quantity and quality of conspecific pollen, and the effect of heterospecific pollen deposited by efficient pollinators (Garibaldi et al. 2014;Zhang et al. 2015).
Variation in the efficiency within flower-visitor groups may be explained by differences between species within groups (King et al. 2013) or between individuals within species (Jauker et al. 2016). This seems especially important for bumblebees, showing a wide intraspecific range of body sizes and tongue lengths (Goulson et al. 2002;Willmer & Finlayson 2014). Since only female bees collect pollen, morphological and behavioural differences between sexes may also lead to intraspecific differences in bees (Ne'eman et al. 2006). Future studies including analyses of pollinator performances at the lowest taxonomic rank possible and even within species will be necessary for further insights into pollination variability.

Visitation frequencies and pollinator effectiveness
Most visitors to flowers of M. sylvestris, B. officinalis and O. viciifolia were bees. The flower-visitor groups analysed for pollen deposition were among the most frequent visitors. Apis mellifera dominated the number of flower visits to M. sylvestris and B. officinalis, whereas bumblebees of the B. lapidarius complex were the most frequent visitors to flowers of O. viciifolia. Differences in dominant visitors between plant species suggest differing flower preferences and high flower constancy of A. mellifera (Hill et al. 1997) and bumblebees of the B. lapidarius complex (Chittka et al. 1997;Raine & Chittka 2005;Zych & Stpiczyńska 2012). It also exemplifies, that the overall most abundant pollinator species, the generalist honeybee, is not necessarily the most frequent visitor to all present plant species. For M. sylvestris and B. officinalis, A. mellifera was the most effective pollinator, its high abundance compensating for the relatively low pollinator efficiency per single visit. For O. viciifolia, bumblebees of the B. lapidarius complex were the most effective pollinators, again based on higher abundances than the similarly efficient B. sylvarum and bees of the genus Megachile at the single-visit level. Hence, in this study the most frequent flower-visitor group of each plant species was most effective in depositing conspecific pollen per time unit, even though not showing the highest pollinator efficiency per single visit. In combination with previous studies, this suggests that the most frequent visitors are often the most important pollinators overall, even when being poor or equally efficient pollinators per single visit J Poll Ecol 21(1) (Olsen 1997;Vázquez et al. 2005;Madjidian et al. 2008;Rader et al. 2009;Zych et al. 2013). Even though pollen limitation was of minor importance in our study and several pollinator species can potentially cause maximum seed set, the results suggest that visitation rate could play a more important role in plant reproduction than pollination success on a per-interaction basis (Vázquez et al. 2005;Sahli & Conner 2006). Thus, pollinator abundance seems a similar important conservation measure for ecosystem functioning as species richness for biodiversity (Kleijn et al. 2015).
Visitation frequencies, however, did not explain pollinator effectiveness entirely. Flower visitors not contributing to pollination at all were similarly frequent as legitimate pollinators in both B. officinalis (Megachile and Halictidae) and O. viciifolia (A. mellifera). Furthermore, differences in visitation frequencies between flower-visitor groups are not proportional to the differences in pollinator effectiveness. For example, honeybees were 15 times more frequent on M. sylvestris than bumblebees of the B. lapidarius complex, but only five times more effective.
Therefore, it seems more reliable and informative to use direct measurements when determining pollinator performances, instead of deducing effectiveness from visitation frequencies alone (Mayfield et al. 2001;Javorek et al. 2002;Ballantyne et al. 2015). Such information reveals differences in the density dependence of pollination success between pollinators and is ultimately necessary for estimating consequences of pollinator decline for plant population persistence. Revealing true pollinator performances will allow to construct more informative plant-pollinator networks (Ballantyne et al. 2015), find appropriate species for crop pollination (Westerkamp & Gottsberger 2000) and thus conserve and apply pollination service best (Garratt et al. 2014).

Conclusion
In conclusion, considerable differences in pollinator performance occurred even in closely related taxa. Not all bees were pollinators of the studied plants and pollinators differed in their pollen deposition efficiency per single visit. Although the most frequent flower visitors were most effective, pollinator effectiveness could not be explained by visitation frequencies alone. These findings emphasise the need to connect visitation frequencies to stigmatic pollen deposition to reveal true pollination performances of flower visitors.