Bumblebees can detect floral humidity

ABSTRACT Floral humidity, a region of elevated humidity in the headspace of the flower, occurs in many plant species and may add to their multimodal floral displays. So far, the ability to detect and respond to floral humidity cues has been only established for hawkmoths when they locate and extract nectar while hovering in front of some moth-pollinated flowers. To test whether floral humidity can be used by other more widespread generalist pollinators, we designed artificial flowers that presented biologically relevant levels of humidity similar to those shown by flowering plants. Bumblebees showed a spontaneous preference for flowers that produced higher floral humidity. Furthermore, learning experiments showed that bumblebees are able to use differences in floral humidity to distinguish between rewarding and non-rewarding flowers. Our results indicate that bumblebees are sensitive to different levels of floral humidity. In this way floral humidity can add to the information provided by flowers and could impact pollinator behaviour more significantly than previously thought.


7
Floral humidity, an area of elevated humidity within the headspace of the flower, has been 3 8 demonstrated to occur in a number of flower species (Corbet et al., 1979;Nordström et al., 3 9 2017; von Arx et al., 2012). Floral humidity is created by a combination of nectar evaporation 4 0 and floral transpiration (Azad et al., 2007;Corbet et al., 1979;Harrap et al., 2020a;von Arx 4 1 et al., 2012) although the contribution of these two influences may vary between species.  The two types of artificial flowers allowed us to observe whether responses of bees were the better suit bumblebee foraging behaviour. In this way active flowers allowed us to test bee 1 4 3 responses to a stimulus produced in a manner comparable to that study.   experiments (e.g. see (Dyer et al., 2006;Lehrer et al., 1995;von Arx et al., 2012)) were 1 8 9 carried out using both artificial flower types. Secondly, differential conditioning techniques 1 9 0 (e.g. see (Clarke et al., 2013;Dyer and Chittka, 2004;Harrap et al., 2017;Lawson et al., 1 9 1 2018)) were carried out with passive artificial flowers only. This was due to the limits on how 1 9 2 much and how quickly active artificial flowers could be moved about the arena due to their 1 9 3 piping (see construction and experimental details in Supporting Information Appendix 1).

9 4
Both preference and learning (differential conditioning) experiments test the capacity of bees 1 9 5 to detect and respond to floral humidity differences. Additionally, preference experiments 1 9 6 investigate how differences in floral humidity between flowers, in the absence of any other 1 9 7 differences, may influence flower choice of naïve bees with no previous experience of floral 1 9 8 humidity. Learning experiments (differential conditioning) investigate whether bees can 1 9 9 associate differences in floral humidity with corresponding differences in rewards and use 2 0 0 this to inform foraging choices. Therefore, two experiments together assess different ways 2 0 1 1 0 that floral humidity might influence the foraging behaviours bees. Individual bees were not 2 0 2 reused between experiments: an individual bee would only take part in one experiment 2 0 3 (preference or conditioning) as part of a single test group (see below) or on a single type of 2 0 4 artificial flower (active or passive). In preference experiments, individual bees had their preference for floral humidity tested four were the dry flower variant. All artificial flowers were rewarding, containing a 25ߤl 2 1 4 droplet of 30% sucrose solution within their feeding wells. Individual bees were released into the arena alone, and bees were allowed to forage 2 1 6 freely on these artificial flowers, and were free to return to the nest at all times. We 2 1 7 monitored whether bees made contact with the top of artificial flowers (which was recorded 2 1 8 as a landing behaviour), and whether the bee extended its proboscis into the feeding well 2 1 9 ('fed') or left without doing so at each landing. After a bee had fed on a flower, the flower was points. Consequently, active artificial flowers were not taken out of the arena but were 2 2 9 instead moved to a different point. In the rare instances where a bee fed from a flower and 2 3 0 revisited it before it could be moved, then these revisits were not counted. When the bee 2 3 1 returned to the nest all the flowers were removed from the arena, cleaned and returned to 2 3 2 the arena in a new position as described in Supporting Information Appendix 1. For each test bee, this cycle of moving flowers continued until the bee had made at bouts with an average of 7.40 ± 0.67 visits per bout for those presented with passive flowers.  The rate at which bees made a positive response to floral humidity (the 'humidity feeding well, or landing on a dry flower and leaving without extending the proboscis into the 2 4 4 feeding well. The humidity response rate data was bounded between 0 and 1, and so was 2 4 5 arcsine square root transformed to fit test assumptions. We used a two-tailed Wilcoxon 2 4 6 signed-rank test to test whether the median value of the transformed humidity response rate 2 4 7 differed from that expected from random choice (a 0.5 humidity response rate, 0.79 once 2 4 8 arcsine square-root transformed), using R 3.6.3 (R Development Core Team, 2017).

4 9
Temperature differences between humid or dry flower variants might occur as a 2 5 0 result of evaporative water loss or action of mechanical components within artificial flowers. The 'Control' group was required for checking to what extent bees could use any 2 8 2 miscellaneous cues other than humidity or variables present in the experimental setup to independently of humidity differences, within the specific settings of our setup.

9 2
Bees were observed for 70 flower visits which is well beyond the number of visits 2 9 3 needed for bees to learn a salient cue, and sufficient to demonstrate such learning by a in each test group (45 bees in total).

9 8
After a bee fed on a flower and flew off into another part of the arena, that flower was 2 9 9 carefully removed from the arena through side openings and refilled with sucrose or water, 3 0 0 as appropriate, before being placed back at a different location. This reduced the chance of 3 0 1 bees' associating particular spatial locations with the reward. If a bee suddenly revisited the 3 0 2 flower before it could be moved, then these revisits were not counted. When the bee  Flower visits were determined as 'correct' or 'incorrect' using the same criteria proboscis into the feeding well, or if she did not extend her proboscis into the feeding well 3 0 9 after landing on a non-rewarding flower. Correspondingly, a bee was recorded as making an 3 1 0 incorrect decision if she landed on a non-rewarding flower and extended her proboscis into 3 1 1 the feeding well, or did not extend her proboscis into the feeding well when she landed on a 3 1 2 rewarding flower. The success rate, defined as the proportion of correct visits over the 3 1 3 previous ten visits, was calculated at ten visit intervals (10 visits, 20, 30... etc.) for each bee. As it was bounded between 0 and 1, the success rate data were arcsine square root 3 1 5 transformed to fit test assumptions. Generalised linear models were fitted to this data using procedure is given in supporting information Appendix 3. Artificial flower temperature differences, as measured during the preference experiments, 3 2 7 were negligibly small. Dry passive flowers had a temperature that was 0.31 ± 0.03ºC (mean 3 2 8 ± SEM) higher than in humid passive flowers throughout the experiment. In active artificial  These differences in temperature were below the threshold of temperature detection by bumblebees (Heran, 1952) and are unlikely to elicit a response by bumblebees. In preference experiments, bumblebees showed a higher spontaneous preference for humid 3 3 8 flowers when they were allowed to freely choose between four humid and four dry flowers In the learning experiment, bumblebees were presented with passive flowers differing in corresponding with rewards varied dependent on the three test groups bees were assigned 3 4 8 to (see above). The relationship between foraging success (probing the feeding wells of based on humidity differences. The presence of floral humidity differences between rewarding and nonrewarding flowers (models which had random slopes and intercepts), were not a better fit than those that only influences on success rate between test groups, interacting effects, had a lower AIC (AIC: better fit (Δdeviance = 17.13, df = 2, p < 0.001) than models that forced experience to have suggested preferences for flowers with higher floral humidity. showed an initial preference to more humid flowers but with experience learned to favour 4 0 5 visits to the rewarding flower type, whether rewarding flowers were producing higher or lower 4 0 6 levels of floral humidity. In the latter case, bees were trained against their initial preference,  Our results indicate that floral humidity represents a floral signal or cue that can be used far pollinators like bees are likely to be able to respond to both the amount of humidity produced 4 1 4 by the flower itself (Yokohari, 1983;Yokohari et al., 1982) and the rate of change in humidity 4 1 5 experienced as the bee approaches or passes the flowers (Tichy andKallina, 2014, 2010). flowers, aid the learning and recognition of flowers from those that produce less humidity.