Figure 4. (a) Mean number of seeds produced per fruit in the non-robbed control (blue, N=85) and robbed (red, N=59) plants. (b) Number of pollen grains exported from anthers in non-robbed control (blue, N=43) and robbed (red, N=50) plants.
Discussion
With our experimental approach we show how the addition of high levels of nectar robbing can have a cost to the reproductive output of a plant colonising a new area. This cost is in terms of seed production, with robbed plants producing 25% fewer seeds compared with non-robbed controls. Below we discuss how this result can be related to changes in pollinator behaviour and the potential implications of these costs for the naturalised populations exposed to novel levels of robbing.
Visiting bumblebees altered their behaviour when interacting with robbed plants by having a significantly reduced visit rate and visiting a smaller proportion of flowers in the inflorescence compared to non-robbed plants. This is consistent with the fact that bumblebees tend to spend less time on an unrewarding inflorescence, and departure can be triggered by encountering one or more unrewarding flowers (Best & Bierzychudek 1982; Heinrich 1979). We also found that bumblebees reduced the length of floral visits on robbed flowers compared to non-robbed flowers. This could potentially be due to a lower volume of nectar present in robbed flowers taking less time to drink and bumblebees leaving sooner (Hodges & Wolf 1981). Richardson (2004) also found that bumblebee visit duration was reduced in robbed flowers compared with unrobbed flowers. Other studies have found that robbing is associated with reduced visitation rates, suggesting that robbing holes and damage to the flower is off-putting to visitors (Goulson et al. 2007; Bronsteinet al ., 2017) and mutilation of the flower can even alter the visiting pollinator assemblage (Varma et al ., 2020). In our experiment (as in non-native field populations of D. purpurea ) holes were made at the base of the corolla, which in field conditions may be out of visual range for legitimate visitors to detect and be repelled by, especially for the fast paced foraging of hummingbird pollinators. However, since there is evidence that hummingbirds can use visual cues to discriminate against robbed flowers (Lara & Ornelas 2001), it would be interesting to assess to what degree hummingbird pollinators in the introduced range are dissuaded from pollinating by the robbing holes as well as from a lack of nectar resulting from robbing (Irwin 2000).
The two visitors to our experimental plants were Bombus hortorumand B. pascourum , with the former visiting 3 times more often, consistent with what is observed in nature in the UK (Broadbent and Bourke 2012; Mackin et al ., 2021). Interestingly, B. pascuorum visited individual flowers for a longer duration, regardless of robbing treatment. This could be explained in part by the shorter tongue length of B. pascuorum , causing them difficulty feeding inD. purpurea flowers and so they take a longer time to complete a visit. In any case, both bumblebee species showed the same patterns of reduced visitation to robbed flowers compared to non-robbed ones.
The change in bumblebee behaviour could be contributing to the lower reproductive output we find for the female component of reproduction in robbed plants, with intense levels of nectar robbing causing less frequent and shorter visits which ultimately reduces pollen deposition and therefore seed production. This idea is supported by several studies that find the duration of visits by bumblebees positively correlates with pollen deposition (Cresswell 2000; Kudo 2003; Thøstesen & Olesen 1996). This is not always the case; in a study by Richardson (2004) bumblebees spent less time in robbed flowers but visit duration did not correlate with amount of pollen deposited. Other authors finding similar results to ours (Irwin & Brody 1998; Lara & Ornelas 2001), suggest that reduced attractiveness of flowers can lead to a reduction in the pollinator visitation rate and a lower seed production. However, we cannot rule out the possibility that resource depletion resulting from manual robbing also contributed to the reduced seed production. In contrast to findings we present here, many studies suggest robbing can have limited or no negative effects on the female component of fitness (Andalo et al., 2019; Carrió & Güemes 2019; dos Santos et al ., 2020; Maloof 2001; Richardson 2004; Varma et al ., 2020; Varma & Sinu 2019; Zimmerman & Cook 1985). This lack of an effect on reproductive output could be due to the legitimate pollinators still visiting the plant and saturating the stigmas with enough pollen so the plant can achieve full seed set (Heiling et al ., 2018; Stoutet al ., 2000). One potential caveat in this study is that our ability to detect differences in mean seed production could be biased by the standardised three-hour visitation periods per day in our experiment, as opposed to longer periods of visitation which could lead to full seed set. However, even with this restriction on the amount of visits plants could receive, these experimental results are consistent with the comparatively low seed set we have observed in the non-native populations (as we discuss below).
We found that nectar robbing did not negatively affect the male component of reproduction through pollen removal by bumblebees. Other studies found that nectar robbing can include a cost to the male component of fitness in some species (Castro et al., 2008; Irwin & Brody 1999; Irwin & Maloof 2002; Richardson 2004) but not in others (Maloof 2001; Morris 1996; Richman et al ., 2018). With the method we used here, it is unclear how much of the pollen released from anthers actually adheres to bumblebee bodies and reaches other plants. As with using any proxy as a measure of reproductive success, in this case it is difficult to deduct the entire picture as to whether nectar robbing affects male success.
The aim of our experiment was to simulate the conditions of nectar robbing on potted plants of D. purpurea , with the idea that similar effects could be found for plants in the non-native robbed field populations. The reduced seed production following addition of nectar robbing we observe here is consistent with our previous observations showing that non-native plants in populations with nectar robbers have a significantly lower lifetime seed production (average = 40,788 ± 20,644 SD seeds, across three populations in Colombia and Costa Rica; N = 211 plants) compared with native populations with no robbing (; average = 113,812 ± 84,868 SD seeds across two populations in the UK; P < 0.001; see also Mackin et al ., 2021). Although many other factors could be involved, the high levels of nectar robbing could be contributing to the lower average reproductive output in the introduced range. In pollinator surveys in the same naturalised populations in Colombia we found that individual Bombus hortulanus and B. rubicundus bumblebees used a mixed strategy of visiting flowers both legitimately and robbing. This can be common in plant-pollinator mutualisms (Morris 1996) although often bumblebees adhere to a consistent strategy to reduce handling time during a foraging bout (Bronstein et al ., 2017). Depending on the amount of legitimate visitation and pollination the plant still receives, fitness costs will only be incurred once a certain threshold of robbing is reached that is enough to deter legitimate pollinators (or impose metabolic costs in increased production of nectar). Thus, any level of robbing below this threshold could have a negligible effect on plants. A next step could be to test whether the addition of nectar robbing to the plant’s environment causes a detectable reduction in reproductive output in field populations.
With the intensity of nectar robbing varying across populations so radically, there could be considerable differences among populations in robber-mediated selection on floral traits (Castro et al ., 2008; Navarro & Medel 2009). Plant populations experiencing a high level of robbing could evolve local resistance or tolerance to nectar robbing (such as phenological, mechanical or chemical barriers) even at the cost of decreasing the attraction to pollinators and reducing reproductive output compared with other populations (Adler et al ., 2016). It is intriguing that native D. purpurea populations experience low levels of nectar robbing, even in the presence of bumblebee species that are capable of making holes and often rob other plant species (Bombus terrestris, B. lucorum and B. wurflenii ). D. purpurea plants produce high levels of toxic cardenolide compounds (Evans and Cowley 1972) that are also present in the nectar (Palmer-Young et al. 2019). It is possible that toxic compounds in foxglove nectar are differentially toxic to particular visiting species, for example generalist robbers, influencing whether they can feed on the plant as has been seen in other species (Barlow et al ., 2017; Villalona et al ., 2020). Further work into the potential role of nectar toxicity and other floral traits and how their relationship with fitness changes under different intensities of nectar robbing inD. purpurea could give insight into how nectar robbing can affect the trajectory of a plant’s evolution.
Our findings contribute to the growing body of evidence that a changed pollination environment, including nectar robbing, can have strong effects on visitation to a plant and the subsequent reproductive output. The addition of novel floral visitors to a plant’s assemblage is likely to become more frequent as plants and nectivorous animal ranges shift due to human influence (Cheptou et al., 2016; Valiente-Banuetet al ., 2013). Therefore, it is important to understand how plants are likely to respond or change as a result of a new antagonism. Further studies on this system could examine how different nectar robbing communities in different parts of the range of D. purpurea are affecting the plant’s evolution.