Two dispersers are better than one: a ‘bird-fig’ attracts bats via nocturnal scent

The plant genus Ficus is a keystone resource in tropical ecoystems. One of the unique features of this group is the modification of fruit traits in concert with various dispersers, the so-called fruit syndromes. The classic example of this is the strong phenotypic differences found between figs with bat and bird dispersers (color, size, and presentation). The ‘bird-fig’ Ficus colubrinae represents an exception to this trend since it attracts the small frugivorous bat species Ectophylla alba at night, but during the day attracts bird visitors. Here we investigate the mechanism by which this ‘bird-fig’ attracts bats despite its morphology which should appeal solely to birds. We performed feeding experiments with Ectophylla alba to assess the role of fruit scent in the detection of ripe fruits. Ectophylla alba was capable of finding ripe figs by scent alone under exclusion of other natural sensory cues. This suggests that scent is the key signal in the communication between Ectophylla alba and Ficus colubrinae. Analyses of odor bouquets from the bat- and bird-dispersal phases (i.e. day and night) differed significantly in their composition of volatiles. This indicates that an olfactory signal allows a phenotypically classic ‘bird-fig’ to attract bat dispersers at night thus to maximizing dispersal.


METHODS
STUDY ANIMAL-Ectophylla alba is a small-bodied leaf-nosed bat species (Phyllostomidae) that 1 1 5 is distributed from northern Honduras to north-eastern Panama (Rodriguez-Herrera et al. 2008 It modifies leaves, predominantly of plants of the genus Heliconia, to construct shelters where it 1 1 7 roosts in social groups of typically four to eight individuals (Brooke 1990 leaves in the area of TRC and selected single males for the feeding experiments in order to 1 2 0 prevent lactating or pregnant females or juveniles from isolation of the social group. All 1 2 1 individuals that were not considered for further experiments were set free immediately in close 1 2 2 proximity to the roost. Following the capture, a single male was released into a flight tent 1 2 3 (Eureka; ground area 4 x 4m, height 2.5m) several hours before sunset. At nightfall we installed a 1 2 4 freshly cut branch of Ficus colubrinae that yielded a range of fruits of different stages of maturity 1 2 5 into the flight tent. In order to adjust to the foraging situation we allowed the bat to feed on ripe 1 2 6 fruits. After the consumption of five fruits we started choice trials in order to test whether E. alba 1 2 7 relies mainly on olfaction or echolocation/vision for the short-range localization of ripe fruits. On 1 2 8 one side of the branch we presented a strong olfactory cue to the bat that lacked visual or echo-1 2 9 acoustic properties of natural figs, i.e. we presented a tissue bag that was filled with ten ripe figs 1 3 0 (similar methods have been used to test for the response of bats to olfactory cues in absence of 1 3 1 natural fruit shape or surface structure: Kalko and Condon (1998) presented cotton saturated with from red clay that were similar to natural F. colubrinae fruits in terms of form, color, and fruit 1 3 5 presentation (in branch forks). We rated E. alba's behavior as a positive response to the presented 1 3 6 object when repeated approximation flights to or a landing next to the object followed by a 1 3 7 directed movement to it occurred. In total, we tested six individual bats. Every bat was tested 1 3 8 7 only once in order to avoid bias caused by learning effects. It was not possible to record data 1 3 9 blind because our study involved focal animals. We documented bat behavior using an infrared 1 4 0 camera (Sony Night-Shot DCR-HC42E, Sony, Japan) that was connected to a video recorder 1 4 1 (GV-D 900E, Sony, Japan). We stored recordings on MiniDV video tapes (DVM60PR3, Sony, 1 4 2 Japan). containing activated charcoal (activated charcoal, Supelco, Orbo 32 large) that was installed 1 5 0 upstream in order to filter-clean the pulled atmospheric air. After passing the glass chamber daily sampling, respectively, and lasted for eight hours with a flow rate of ca. 100mL min -1 .

5 7
After sampling, all sorbent tubes were eluted with 0.050 ml of 10:1 pentane/acetone. Eluted gas (2 ml min−1 constant flow). One microliter of each sample was injected splitless at 40°C.  (PCs) with an eigenvalue above one to run a discriminant function analysis (DFA) in order to test 1 8 2 for differences in the scent composition between (1) unripe fruits during night, (2) ripe fruits compounds. Factor loading above 0.5 were considered high. Finally, we compared relative 1 8 6 9 amounts of single compounds of ripe fruits during day and night (groups 2 and 3) using Mann-1 8 7 Whitney U-tests in R 2.15.3 (R Developing Core Team 2015). search flights that lasted between less than one minute and almost two hours (mean ± standard 1 9 6 deviation: 32 ± 43 minutes, n = 6) until the bats approached the branch for the first time. Then the 1 9 7 bats conducted two to nine approximation flight towards the branch over a period of one to 91 1 9 8 minutes (mean ± standard deviation: 19 ± 36 minutes, n = 6) before they landed and consumed a  period of six to 48 minutes (mean ± standard deviation: 16 ± 21 minutes, n = 5, see Table 1) and 2 0 8 one to five approaches the bats either landed on or right next to the bag or landed more than 5 cm were attributed to 17 individual substances, again 13 of which were unambiguously identified by 2 1 3 mass spectrometry (Table 2). Nonanal and 1-tetradecanol contributed the largest share to the 2 1 4 overall bouquet (Fig. 1, Table S1). Three further substances could be assigned to substance sesquiterpenenes, and aromatic compounds. In three cases, two substances contributed to a single 2 1 9 peak in the GC-analysis. In those cases the overlapping substances were represented by a single 2 2 0 value for the following analyses. Two of the identified substances, indene and anthracene, have a 2 2 1 main relevance in industrial applications and were therefore excluded from all further analyses.

2
They were considered environmental pollutants that accumulated on the outside of the fruits over industry. There were no significant differences in relative amounts of indene and anthracene 2 2 5 among day and night in ripe fruits. Medians were lowest in unripe fruits and rising over time while ripening (Fig. S1 & Fig. S2).

7
We performed a PCA that included 12 individual values for the relative amounts of the (2) ripe fruits during day, and (3) ripe fruits during night). Four PCs with an eigenvalue above  Fig. 2). The highest coefficient for DF 1 was attributed to PC2 which in turn had high factor 2 3 4 scores on the sesquiterpenes α -copaene and δ -cadinene + calamenene (Table S2 & Table S3). For 2 3 5 DF 2, PC1 and PC3 had the highest coefficients. PC1 had high factor loading on sesquiterpene A, 2 3 6 β -copaene + naphthalene derivative, α -cubebene + 1,1'-biphenyl and the FAPCs nonanal and 2 3 7 decanal. 1-dodecanol and 1-tetradecanol loaded high on PC3. Seventy-five percent of the original 2 3 8 grouped cases were correctly classified (72.5 % of cross-validated grouped cases). biphenyl) had significantly higher proportions during night (Table 2). Our study shows that scent is an important signal in the communication between Ectophylla alba 2 5 2 and Ficus colubrinae. Ectophylla alba was capable during experimental trials to find ripe figs by figs'. We suggest that this strategy of Ficus colubrinae is an adaptation towards dispersal by 2 5 8 small bats such as Ectophylla alba rather than towards bat dispersal in general, since odor may be 2 5 9 an ideal signal to attract a specific group of bat species.

6 0
Semi-natural feeding trials showed that phyllostomid bats locate fruits by echolocation tissue bag that gave a strong olfactory cue but lacked natural texture, shape, size, or presentation 2 6 5 of figs that might be of importance for detection by echolocation. Therefore, we assume that (Van der Pijl 1957). Korine and Kalko (2005) argue that detection of fruits by downwards 2 7 3 frequency modulated signals which are typical for phyllostomid bats is possible but largely 2 7 4 depends on the fruit presentation and the complexity of the surrounding clutter. However, F.  Olfactory cues enable plants to signal the readiness of fruits for dispersal. Accordingly, 2 8 0 temporal changes in the volatile profile of fruits are common during the process of ripening (e.g. overall composition of the scent bouquet during the process of ripening. Additionally we 2 8 4 observed significant changes among day and night, caused by day-time specific scent production.  colubrinae match the bird-syndrome, scent alone is sufficient for Ectophylla alba to detect the 2 9 2 ripe fruits as shown by our behavioral experiments. Hence, a nightly shift in volatile production 2 9 3 may enable 'bird-figs' to additionally attract certain bat species as dispersers and hence allow for 2 9 4 dispersal during the daytime and at nighttime. To achieve seed dispersal by distinct animal taxa 2 9 5 may result in multiple benefits to a reproducing plant. The contribution to overall seed rain by 2 9 6 birds or bats, respectively, may vary quantitatively across seasons (Galindo-González et al. fruits and seeds may arrive in a more heterogeneous range of microhabitats for germination. were completely missing in our samples. This result was surprising since feeding trials showed