Discrimination of small sugar concentration differences helps the nectar-feeding bat Leptonycteris yerbabuenae cover energetic demands

When foraging, animals are often confronted with a choice between simultaneously present food sources of unequal energetic quality. To make a beneficial decision, the available options must be sensed, memorized and then compared (Cnaani et al., 2006; Nachev et al., 2013b; Nachev and Winter, 2012). Optimal foraging theory predicts that foragers should prefer food sources that yield more energy per unit handling time (Schoener, 1971), which particularly applies to foragers that consume immobile food like floral nectar (Sih and Christensen, 2001). By increasing the benefits and decreasing the costs of feeding, foragers are able to maximize their net energy gain (von Helversen and Winter, 2003). Nectar is offered by flowers as an energy reward in exchange for pollination services (Heinrich and Raven, 1972) and is regularly consumed by insects and vertebrates, including nectar-specialist birds and bats (Nicolson, 2007). While searching for food, these nectar feeders may encounter flowers offering nectar with different energetic content (Baker et al., 1998). A nectar feeder's ability to select optimal food sources is of adaptive significance for survival and evolutionary fitness (Ritchie, 1990). To understand how nectar feeders decide to approach a flower, it is crucial to determine how sensitive they are to changes in the nectar's energetic content, which represents food quality.

The reward value of a nectar source is mainly a function of its sugar concentration and volume (Cnaani et al., 2006; Harder and Real, 1987). Across taxa there is a general pattern that nectarivores favour the more concentrated and thus sweeter nectar solution when offered the choice between different sugar concentrations (Nachev et al., 2013a). In bats and birds, this holds true for sugar concentrations up to ∼50% (Roces et al., 1993; Tamm and Gass, 1986). Surprisingly, it is as yet unknown how well most of these nectar specialists can discriminate different sugar concentrations. Nectar-specialist bats and hummingbirds both feed from plants that offer nectar with sugar concentrations up to 35% (Martínez del Rio et al., 2001; Rodríguez-Peña et al., 2016). In an early study by Hainsworth and Wolf (1976), different hummingbird species were offered the choice between nectars that differed by at least 1.74% in sugar concentration. While some birds were able to discriminate this difference, others failed to do so. Rufous hummingbirds (Selasphorus rufus) were able to discriminate a difference of 5% but only at and below a mean sugar concentration of 32.5% (Tamm and Gass, 1986). A follow-up study showed that the hummingbirds' discrimination ability varied with mean sugar concentration and was best at intermediate sugar concentrations of 20% (Blem et al., 2000). Nectar-feeding bats of the genus Glossophaga discriminated a 5% sugar concentration difference if the mean sugar concentration was below 15% (Nachev and Winter, 2012). Lesser long-nosed bats (Leptonycteris yerbabuenae) discriminated a sugar solution of 18% from one of 15% in a study by Rodríguez-Peña et al. (2007). However, most of these studies offered large differences in concentration and did not determine the discrimination threshold, i.e. how small sugar concentration differences can be before an animal fails to discriminate between them. A fine-tuned sensory ability to discriminate very small concentration differences would be helpful in choosing higher quality food sources.

For the first time in any nectarivorous mammal, we systematically tested this discrimination threshold, focusing on the highly specialized nectar-feeding bat L. yerbabuenae. It feeds mainly on nectar and pollen from flowers of the (sub-)families Bombacoideae, Cactaceae and Agavoideae, many of which in turn depend heavily on pollination by bats (Horner et al., 1998; Stoner et al., 2003). The nectar offered by these plant species represents a temporally and spatially variable and highly heterogeneous food source: intra-plant variation in sugar composition and concentration is often very large (Pacini and Nepi, 2007), resulting in substantial differences in the offered amount of energy between individual flowers (Rodríguez-Peña et al., 2016). Like all nectarivores, nectar-feeding bats have very high energy demands. Their metabolic rates are much higher than those of other similar-sized mammals (Nagy et al., 1999; von Helversen and Winter, 2003). This is mainly because of their costly hovering flight mode while feeding from flowers (Voigt and Winter, 1999). Additionally, their metabolism is fuelled directly by immediately ingested sugar, mostly without converting it to fat (Voigt and Speakman, 2007). This forces nectar-feeding bats to consume large amounts of nectar every day (von Helversen and Winter, 2003). To keep the daily energy uptake constant, nectar-feeding bats increase their volumetric intake when nectar sugar concentration is low. This intake response (Castle and Wunder, 1995) is described by a simple mathematical model: the power function V=a×c^b, where V is nectar intake, c is concentration and the intercept a and the exponent b are constants (Ayala-Berdon and Schondube, 2011; Martínez del Rio et al., 2001). If a species always consumes the same amount of energy, it achieves compensatory feeding (i.e. the slope of the power function equals −1). Species unable to do so will take up less energy when feeding on dilute solutions (i.e. the slope of the power function is larger than −1) and thus face physiological constraints at lower nectar sugar concentrations. The presence of such constraints can lead to changes in the bats' foraging behaviour that affect the bat–plant interaction (Ayala-Berdon et al., 2011). Reasons for physiological constraints in bats include coping with the management of large quantities of water, limitations in gut capacity and the rate of sugar assimilation (Ayala-Berdon and Schondube, 2011).

The energy uptake in L. yerbabuenae is also limited by physiological constraints: Ayala-Berdon and Schondube (2011) showed that wild-caught individuals are unable to take up the necessary amount of energy when feeding on sugar concentrations below 15%. To select flowers that offer a high sugar concentration they need to be able to perceive and compare minute differences in sugar concentration. This is particularly important at low sugar concentrations where discrimination should be better, as the energetic benefits from selecting more concentrated nectar are greater at low absolute concentrations than at high. Therefore, the uptake of low concentrated nectar should be avoided, because it limits the bats' ability to store energy and maintain body mass (Ayala-Berdon and Schondube, 2011). We therefore hypothesized that L. yerbabuenae will discriminate small sugar concentration differences in the lower range of nectar concentrations (<15%) it faces in its natural habitat. To test this, we offered bats the choice between paired feeders containing sugar solutions of different concentrations and determined their individual discrimination thresholds across the range of concentrations found in their natural foraging plants.

Treatment of the experimental animals complied with all institutional and national laws on animal care and experimentation. A specific ethical approval was not required because of the observational nature of the study, which caused no harm, pain or distress to the bats. This assessment was confirmed by the animal welfare officer of the University of Tübingen and the local authority (Regierungspräsidium Tübingen).

We tested 12 (6 male, 6 female) adult lesser long-nosed bats (Leptonycteris yerbabuenae Martínez and Villa-R 1940). All bats were born in captivity and held in our laboratory colony for at least 8 months prior to the start of the study. Animals were weighed (27.0±0.7 g) and visually inspected to confirm they showed no signs of illness or pregnancy before taking part in the experiment. Bats were weighed again after they had participated in all trials and had gained, on average, 1.6±0.5 g during the experimental time period. They were housed under a reversed 12 h:12 h light:dark cycle with constant air temperature (26±1°C) and relative humidity (55±5%). All bats of the colony were fed daily with bee-collected dry pollen as a nitrogen and amino acid supply (Howell, 1974) and a variety of maintenance diet solutions, each offering a total sugar concentration of 18% (w/w). Individual bats were marked with a reflective forearm ring.

For the experiment, 2–3 bats were transferred to a separate room (3.45×3.45×2.95 m). Environmental conditions were identical to those of the housing room. The bats had 3 days to habituate to the room and the feeder set-up, during which they were fed with one of the maintenance diet solutions. The bats had no difficulties finding the feeding place. The walls and the floor of the room were covered with anechoic foam. The feeding station (1 m above the floor) was placed at the side of the room, opposite to the bats' resting place. In order to drink from a feeder, bats had to hover in front of it. Each experimental session started within 30 min of the beginning of the dark phase by placing the two feeders on the feeding station (two-alternative free choice test). The experiment lasted 6 h each day, mirroring the natural foraging time in the wild (Horner et al., 1998), after which the two feeders were removed again. Above the feeding station we attached a camera (Genie HM 1024, Teledyne Dalsa) with a zoom lens and an 850 nm infrared spotlight (Infralumis, AMG) to the ceiling that allowed us to record without any visible light. Videos were recorded on a laptop directly from the incoming camera signal with the software Gecko (version 2.031, Vision Experts).

During each experimental day, we presented the bats with a choice between a pair of feeders placed next to each other on the feeding station, 17 cm apart. Each feeder contained an ad libitum amount of sugar water made of a 1:1 mixture of glucose and fructose (≥99.5%, Roth), differing only in their concentration. The two choices differed equally above and below mean concentrations of 5%, 10%, 15%, 20% and 25% (w/w). For each mean concentration, we determined the smallest difference that each bat could discriminate in steps of 0.5%. We chose 0.5% as the smallest difference to test because this value is in the range of the absolute gustatory threshold of L. yerbabuenae for sugar detection (Ayala-Berdon et al., 2013). We had already determined in a preliminary experiment with four bats (not used here) the approximate discrimination threshold at each concentration mean. The sequence in which these mean concentrations was presented and the initial placement of the two feeders on the feeding station were randomized. In total, each bat was offered 13 sugar solution pairs (Table S1) and the bats remained in the experimental room until we had offered them all test solution pairs, which took about 2–3 weeks. If a bat did not discriminate the largest difference that we initially offered at any mean concentration, this bat was tested again after the end of the experiment, increasing Δc in steps of 0.5% until it was able to discriminate. For example, if a bat did not discriminate the largest difference of 5% that we had offered it at the mean concentration of 20% (see Table S1), we continued testing it with increasingly greater concentration differences (5.5%, 6%, 6.5%, etc.) until it was able to discriminate. The same procedure was applied when a bat was able to discriminate a smaller difference than we had initially presented.

In the middle of each session (after 3 h), the position of the two feeders was swapped by the experimenter to account for any positional bias. Both feeders were weighed before and after each experimental session. Because visiting a nectar source is rather stereotypic in L. yerbabuenae, the amount taken per visit is likely to be constant for any given concentration (Ayala-Berdon et al., 2011; Howell and Hartl, 1980). We calculated that with each visit a bat drank 0.20±0.003 g of sugar solution, as expected for feeders with almost no flower depth where the sugar solution is filled up to the edge of the feeder (Gonzalez-Terrazas et al., 2012). Therefore, we used the individual number of visits and the mass of the consumed solution to estimate the food intake by each bat under the assumption that the amount taken per visit is constant. We calculated each bats' energy intake based on the amount of consumed sugar, with 1 g of the glucose–fructose mixture yielding 15.61 kJ of energy (Domalski, 1972).

To determine whether a bat was able to discriminate between the two feeders, we counted the number of feeding visits of each bat to each feeder using the freely available software BORIS (version 7.4; Friard and Gamba, 2016). We then determined the discrimination strength by dividing the number of visits to the feeder with the higher concentration by the total number of visits. We did this separately for each half of an experimental session and then averaged the two values. We define bats that showed a discrimination strength value ≥0.62 as being able to discriminate between the two concentrations. We chose this threshold because even for the lowest number of total feeding visits by a single bat (n=93 at 25% sugar concentration), the 95% confidence interval (calculated after Hedderich and Sachs, 2018) lies significantly above chance level if the number of correct choices is ≥57, which equals a discrimination strength of 0.613. Four bats chose one particular feeder and remained with that decision during the course of each experimental session, independent of the presented choices. As established before the start of the experiment, this resulted in discrimination strength values of 0.5 and we excluded the data from these four bats in the discrimination threshold analysis. To test whether the mean sugar concentration and the order in which the pairwise choices were presented had an effect on the bats' food intake, a three-way ANOVA (and subsequent Tukey post hoc test) was performed. Bat identity was included in this model as a random factor to account for individual variation in food intake and for the fact that bats were tested multiple times. All statistical tests were done using JMP (version 14.2.0, SAS Institute). All values are given as means±s.e.m., unless stated otherwise.

We offered 12 L. yerbabuenae the choice between two different concentrations of sugar solution across mean concentrations of 5–25%. In total, we scored 30,836 feeding visits, of which ∼53% occurred in the first half of the 86 experimental sessions. Using the number of visits and the ingested amount of sugar solution, we calculated the daily energy intake of each bat (Fig. 1A). A three-way ANOVA revealed that the mean sugar concentration had a significant effect on energy intake (F_4,139=29.40, P<0.0001), whereas the presentation order had no effect (F_1,139=0.65, P=0.42). On average, each bat consumed sugar corresponding to an energetic equivalent of 74.7±1.4 kJ during the experiment if the mean sugar concentration was ≥10%, corresponding to an intake of 12.4 kJ h⁻¹. This figure is comparable to the average intake of ∼11.6 kJ h⁻¹ from wild conspecifics [data extracted from fig. 2 in Ayala-Berdon and Schondube (2011), averaged over all sugar concentrations ≥15%]. In a previous experiment, our bats had shown a similar total energy intake of 80.4±2.5 kJ when they were allowed to feed for 24 h (unpublished data). We conclude that the bats had enough time to feed and their energy intake was not constrained by the experimental design. The observation that energy intake did not even increase when the bats had the choice between two options with a higher mean sugar concentration indicates that the bats consumed enough carbohydrates to cover their daily energy expenditure. Only when facing the lowest sugar concentration of 5% did bats consume about one-third less energy (50.4±3.3 kJ), a significant difference compared with the energy intake at higher mean sugar concentrations (Tukey post hoc test: all P<0.0001). Based on the average amount of solution taken per feeder visit and the ∼75 kJ of energy a bat needs to take up every day, we calculated the number of feeding visits that are required to cover a bat's daily energy demand (Fig. 1B). Except for the 5% mean sugar concentration, the bats were able to make the necessary number of visits to keep their energetic intake constant. In our experiment, L. yerbabuenae were able to maintain a constant energy intake at sugar concentrations ≥10%, similar to the limit of 15% reported by Ayala-Berdon and Schondube (2011), although their finding was not statistically verified.

The bats' feeding behaviour followed the intake response pattern: they increased their total daily volumetric food intake at lower sugar concentrations (Fig. S1). The intake response was well described by the power function V=a×c^b. Inclusion of data from trials with the mean sugar concentration of 5% resulted in an exponent of −0.74; without these data, the exponent was −0.99. This result further supports the presence of a physiological constraint at the lowest sugar concentration. Power function fits of individual bats are given in Table S2. Our findings corroborate the previously conducted experiment by Ayala-Berdon et al. (2008) with L. yerbabuenae captured in the wild. Both results, daily energy intake and intake response, demonstrate the bats' inability to take up enough energy at low nectar sugar concentrations.

Our main goal was to determine how well a nectar specialist is able to discriminate between different sugar concentrations at unlimited nectar volumes. By using the bats' inherent preference for sweeter nectars, we determined the smallest detectable concentration difference at sugar concentrations characteristic for the species' main feeding plants (Rodríguez-Peña et al., 2016). Specifically, we wanted to know whether the discrimination performance depends on the mean absolute sugar concentration of the two solutions. As predicted, bats were able to discriminate very small sugar concentration differences and generally showed a better performance at lower absolute sugar concentrations (Fig. 2).

At 5% mean sugar concentration, seven of eight bats were able to discriminate a concentration difference of at least 0.5% (i.e. 4.75% versus 5.25%) and six of them were also able to do this at 10% mean sugar concentration. When the offered sugar concentration increased beyond 10%, the precision of discrimination decreased and between-subject differences also became more evident. For example, at 20% mean sugar concentration, the best-performing bat was still able to detect a difference of 1.5% while another bat only discriminated a 5% difference. It is important to note that we measured discrimination performance and not perceptual discrimination ability. Bats might be able to perceive the difference between solutions but ignore the difference in energetic content, perhaps because the cost of evaluating other nearby options is very low. We therefore think that the measured discrimination thresholds reported here are a conservative approximation of the bats' discrimination ability, as decision-making processes can be erroneous as a result of noise, e.g. caused by exploratory behaviour, and motivational problems (Nachev et al., 2013a). Lack of motivation is also a proximate explanation for why the discrimination becomes worse at sugar concentrations above 20%. Here, we observed some bats picking the higher concentration feeder at the beginning of the experiment, only to fail to change sides once the feeders were swapped. This resulted in discrimination strength values of 0.5, therefore classified as not discriminated. In the vast majority of these trials, bats chose the feeder with the higher sugar concentration when the experiment began but might have been unmotivated to search for the more profitable alternative after the switch. This might be because they had already consumed a substantial amount of sugar and even the solution with the lower sugar concentration provided a sufficient amount of energy. The bats' perceptual discrimination ability at high sugar concentrations can therefore be expected to be higher than evident in our results.

In being able to discriminate a 0.5% difference in sugar concentration, L. yerbabuenae outperforms facultative nectar-feeding bats such as Glossophaga commissarisi (Nachev and Winter, 2012) and even other nectar specialists like hummingbirds (Blem et al., 2000; Hainsworth and Wolf, 1976) and bumblebees (Cnaani et al., 2006; Nachev et al., 2013b). Our results support the pattern that nectarivorous bats and birds have better discrimination power at lower absolute sugar concentrations (Nachev et al., 2013a). In Blem et al.’s (2000) study, hummingbirds exhibited the best discrimination at concentrations found in hummingbird-pollinated flowers (∼20%) and their performance was worse below and above that optimal concentration. We did not find such an optimum curve of discrimination performance. Instead, discrimination performance increased towards lower mean sugar concentrations. The precision of discrimination was particularly fine tuned at sugar concentrations below 10%, where this species also faces physiological constraints in energy uptake. When only diluted nectars are available, it is a critical task for each bat to maintain a positive net energy gain. Failing to do so will have negative effects on fitness and survival probability. In contrast, at higher concentrations it is comparatively easy for bats to cover their energy expenditures and therefore it is less rewarding to discriminate small differences in the nectar's energy content. In our laboratory setting, a sugar concentration of 10% was enough to take up the necessary amount of energy. When deciding for or against a flower, nectar-feeding bats assess concentration and volume of the nectar alongside the energetic costs that arise while foraging, which in turn depend mainly on the distance between individual flowers. By using their excellent discriminatory power for choosing higher concentrated nectar sources, L. yerbabuenae may act as a potential selective pressure on chiropterophilic plants with co-evolutionary consequences for both parties (reviewed by Fleming and Holland, 2018).

We thank Peter Pilz for advice on statistical analysis, Verena Lohmüller for language support and two anonymous reviewers for valuable comments. Some of the experimental animals were kindly provided by York Winter, Humboldt University Berlin.