Odor discrimination in classical conditioning of proboscis extension in two stingless bee species in comparison to Africanized honeybees
- First Online:
- Cite this article as:
- Mc Cabe, S.I., Hartfelder, K., Santana, W.C. et al. J Comp Physiol A (2007) 193: 1089. doi:10.1007/s00359-007-0260-8
- 239 Views
Learning in insects has been extensively studied using different experimental approaches. One of them, the proboscis extension response (PER) paradigm, is particularly well suited for quantitative studies of cognitive abilities of honeybees under controlled conditions. The goal of this study was to analyze the capability of three eusocial bee species to be olfactory conditioned in the PER paradigm. We worked with two Brazilian stingless bees species, Melipona quadrifasciata and Scaptotrigona aff. depilis, and with the invasive Africanized honeybee, Apis mellifera. These three species present very different recruitment strategies, which could be related with different odor-learning abilities. We evaluated their gustatory responsiveness and learning capability to discriminate floral odors. Gustatory responsiveness was similar for the three species, although S. aff. depilis workers showed fluctuations along the experimental period. Results for the learning assays revealed that M. quadrifasciata workers can be conditioned to discriminate floral odors in a classical differential conditioning protocol and that this discrimination is maintained 15 min after training. During conditioning, Africanized honeybees presented the highest discrimination, for M. quadrifasciata it was intermediate, and S. aff. depilis bees presented no discrimination. The differences found are discussed considering the putative different learning abilities and procedure effect for each species.
KeywordsAssociative learningOdor discriminationStingless beeHoneybeeClassical conditioning
In order to efficiently obtain food from flowers, nectar-collecting insects learn the association between the food, which serves as a positive reinforcement, and several floral cues, such as the floral odor (for review see Menzel 1999). Learning the food odor during foraging facilitates the return to a profitable flower patch (von Frisch 1919). The contingency odor-reward can also take place under laboratory conditions. Studies on olfactory classical conditioning have been reported for several insect species (honeybees: Takeda 1961; Bitterman et al. 1983; bumblebees: Laloi et al. 1999; cockroaches: Watanabe et al. 2003; moths: Daly and Smith 2000; hymenopterous parasitoids: Kaiser et al. 2003; flies: Tully and Quinn 1985; Chabaud et al. 2006), showing that they are all able to associate an odor with a reward. However, the best-established model is the honeybee. Harnessed honeybees are trained to associate an odor with a reward of sucrose. The obtainment of sucrose solution occurs by touching the bee’s antenna that elicits the proboscis extension response (PER) (Frings 1944). This response is a reflex used as unconditioned response and it is conditioned (Takeda 1961; Bitterman et al. 1983) in a classical (Pavlovian) way.
Olfactory conditioning in controlled laboratory conditions, such as the PER paradigm, allows the study of insects’ cognitive abilities, enabling posterior correlations between behavior and the underlying physiological mechanisms of learning processes (Menzel 1999). In highly organized insect societies it also helps to understand behavioral social issues such as information transfer and communication between nestmates (Farina et al. 2005; Grüter et al. 2006) because of the ability of some species to transfer appetitive associations between different behavioral contexts (Gerber et al. 1996; Sandoz et al. 2000; Chaffiol et al. 2005).
There are only few learning studies under controlled conditions that were performed on honeybee-related species, a particularly interesting one being that of Pessotti and Sénéchal (1981) who studied visual operant conditioning in several species of Brazilian stingless bees. They showed that species from the genera Melipona and Scaptotrigona were able to associate color with food. When studying olfactory learning, Abramson et al. (1999) attempted to accomplish classical (absolute) conditioning with Melipona scutellaris using the PER protocol, but even after 12 trials, these bees did not show any conditioned response toward the rewarded odor.
Most of the Neotropical highly eusocial bees belong to the tribe Meliponini that includes hundreds of species grouped in several genera and subgenera (Camargo and Pedro 1992). These stingless bees have been described to communicate food source location and profitability using diverse communication systems (Lindauer and Kerr 1958; Hrncir et.al. 2000; Aguilar and Briceño 2002; Hrncir et.al. 2004; Nieh 2004). Information transfer about floral odors within stingless bee colonies has also been reported (Lindauer 1956; Nieh et al. 1999; Aguilar et al. 2005). Honeybee queens of African origin were introduced to Brazil 50 years ago, and following a hybridization process they have proliferated, taking over most of the Neotropical and also some temperate regions in the South and North of the Americas (for a recent review of the genetics of the Africanization process see Whitfield et al. 2006). So, not only are the Meliponini a very diverse tribe and closely related to the less diverse Apini (with only 11 species in a single genus; Michener 2000), but they also share with the Africanized honeybees part of their resources and biotopes (Wilms et al. 1996). This presents an excellent opportunity to analyze their learning abilities simultaneously.
Specifically, the three species chosen for this study present very different foraging and recruiting behavior. Honeybees undoubtedly are very efficient foragers due to their complex communication systems, such as the dance behavior to recruit inactive nestmates (von Frisch 1967). They are also exceptionally good learners associating incidental environment cues with the respective resources (for review see Menzel and Müller 1996). Like some other species of the trigonine group, Scaptotrigona species have a recruiting system that greatly relies on pheromone trails (Lindauer and Kerr 1958; Schorkopf et al. 2007). This kind of recruitment suggests that neither visual nor olfactory (floral) learning abilities may be very relevant to find the food source, since a bee must only follow a conspecific odor trail to the goal. Melipona species stand in the middle of these two extremely different strategies. They have been described as using a general alerting strategy that consists of jostling and sound production to transmit information about the food source inside the hive (Nieh 1998a; Hrncir et al. 2000). In addition, pheromone marks are deposited near the food source to help recruited individuals in the final approach (Lindauer and Kerr 1958; Nieh 1998a).
The main goal of this study was to analyze the capability of the native stingless bees Melipona quadrifasciata and Scaptotrigona depilis to be olfactory conditioned in the PER paradigm, evaluating responses to different floral odors after a five-trial differential conditioning (five rewarded-odor presentations and five nonrewarded-odor presentations). This behavioral study was simultaneously performed with the already thoroughly studied honeybee kept within the same habitat.
Materials and methods
The experiments were performed on the campus of the University of São Paulo, Ribeirão Preto, Brazil, between August and December 2005 and between July and November 2006. We used one Africanized Apis mellifera hive; four colonies of Melipona quadrifasciataanthidioides and three colonies of Scaptotrigona aff. depilis (subsequently referred to as M. quadrifasciata and S. aff. depilis). The stingless bee colonies had already been established in this environment before the starting of experiment and were kept in a meliponary adjacent to the apiary of Africanized honeybees. All colonies contained a queen, brood and ample stored food.
First, we examined whether M. quadrifasciata and S. aff. depilis bees presented a clear proboscis extension reflex, which is a basic requirement to initiate later on a conditioning procedure in the PER paradigm (Frings 1944). Once this was established, this response was used to investigate a bee’s sensitivity to increasing concentrations of sucrose. This protocol had been previously used with honeybees (e.g., Page et al. 1998; Pankiw and Page 1999), but not yet with stingless bees. Sucrose sensitivity was recorded for each species at least once a week from October to December 2005, during which time a comparative learning study was carried out.
Bees of each species were always collected at the respective hive entrances using plastic tubes. They were immobilized by chilling and mounted in appropriately cut pipette tips that restrained the body movement but allowed free movement of the antennae and mouthparts (Frings 1944). They were offered water to drink and then kept in an incubator (28°C, 60% relative humidity, and darkness) for at least 1 h before the tests.
Prior to performing the assay, bees were offered water in order to avoid confounding effects of thirst. Bees were assayed using a concentration series of 0.1, 0.3, 1, 3, 10, 30 and 50% weight/weight (w/w) sucrose solution following the protocol described by Page and coworkers (1998). Between each sucrose solution trial, all bees were tested for their response to water as a control on the potential effects of repeated sucrose stimulation that could lead to sensitization or habituation affecting subsequent responses (Page et al. 1998). The lowest concentration at which an individual bee responds by protruding her proboscis is interpreted as her sucrose response threshold (SRT).
Each sucrose solution was tested by touching the bee’s antennae with a toothpick soaked in the solution, and it was considered as a positive response only if the bee fully extended her proboscis. We recorded the proportion of bees of each species that responded to at least the 50% sucrose solution. We also obtained a sucrose-response score for each bee, which is the number of consecutive concentrations in the series for which the bee responded by extending her proboscis. This score correlates with the response threshold because animals normally respond to all concentrations above their threshold (Pankiw et al. 2004). A bee that only responded to the contact with the 50% sucrose solution was given a score of one, whereas a bee that responded to the 0.1% solution and onwards to all seven subsequent concentrations received a score of seven. A gap of no response to only one concentration in the responses series was tolerated. If a bee did not respond to any sugar concentration (score of zero), it was not considered in the analysis.
Differential PER conditioning
In order to analyze learning abilities in different eusocial bee species that coexist in a Neotropical environment we performed a classical differential conditioning experiment (Bitterman et al. 1983). In the same way as in the SRT-experiment, bees from each species were collected at the hive entrance. Once captured, they were offered a tiny droplet of 50% w/w sucrose solution and placed in an incubator. After 1.5 h in captivity, bees were harnessed as described earlier and were fed a 50% w/w sucrose solution for about 3 s before being kept in the incubator for at least 2 h. We subjected the harnessed bees to a standard differential PER conditioning protocol in which two pure odors were presented, one rewarded (rewarded conditioned stimulus, CS+) with 50% w/w sucrose solution (unconditioned stimulus, US) and the other non-rewarded (non-rewarded conditioned stimulus, CS−). A device that delivered a continuous airflow was used for odorant application. To apply the odorant stimulus, by means of an electronic valve, the air flux was redirected to pass through a syringe with a filter paper soaked in 4 μl of odor. The bees were exposed to these stimuli five times each in a pseudo-randomized order (CS−, CS+, CS+, CS−, CS−, CS+, CS−, CS+, CS+, CS−). The inter-trial interval lasted 10–15 min between CS presentations. The same interval was maintained between the last trial and the testing phase. Only those bees that showed the unconditioned response (the reflexive extension of the proboscis after applying a 50% w/w sucrose solution to the antennae, UR) and that did not respond to the mechanical airflow stimulus were used. Trials lasted for 46 s and consisted of 20 s of airflow, 6 s of odor (CS) and 20 s of airflow. During rewarded trials, the reward (US) was delivered during the last 3 s of the CS, when the bees had extended the proboscis (PER) as a response to contacting their antenna with the sucrose solution. A conditioned response (CR) was considered only if the bee responded by fully protruding her proboscis during the first 3 s of odor presentation, without need of touching her antenna with the sucrose solution. Bees that responded to the first presentation of the CS (spontaneous response, SR) were eliminated from the PER conditioning experiment. This aimed to avoid working with bees that had any sort of uncontrolled previous experience with the floral odors.
To evaluate if the bees had short-term memory, after they had all gone through the learning assay, we left the harnessed bees in the incubator for 15 min and then subjected them to a non-rewarded presentation of both odors.
To obtain the maximum possible data in each phase (SR, training and test), we evaluated all the bees that were responsive during the phase, even if they ceased to respond in the following phases. In each figure the number of bees taken into account is indicated.
Because of the low response found in stingless bees during the differential conditioning runs, we decided to repeat the learning assays between July and November 2006. We repeated the previously described protocol (i.e., 6-s odorant stimulus) and now, based on the work of Laloi and coworkers (1999) who studied olfactory learning in bumblebees, we also added a series in which the odorant stimulus lasted for 12 s. This served to test whether the stimulus duration had any influence on their learning ability. In these tests, trials lasted for 46 s and consisted of 20 s of airflow, 12 s of odor (CS) and 14 s of airflow. During rewarded trials, the reward (US) was delivered during the last 3 s of CS, as previously described.
Two pure floral odors were used in the learning experiments, linalool (LIO) and phenylacetaldehyde (PHE). Both odors are natural components of flower scents (Knudsen et al. 1993) and were obtained from Sigma-Aldrich, Steinheim, Germany.
A G test was used to compare PER frequencies of sucrose responses between species, and multiple comparisons were performed using the Dunn-Sidak correction (Sokal and Rohlf 1995). When analyzing the sucrose-response scores, Kruskal–Wallis tests were used because the assumptions of normality were not met (Zar 1999). No statistical analysis was performed on SR-values obtained during the differential conditioning runs because of the very low number of responses found for the PER paradigm, which meant that one of the prerequisites for a contingency analysis was not met (expected frequencies should be higher than 5, Sokal and Rohlf 1995). Performance during conditioning was analyzed using a Discrimination Index (DI) that was calculated for each bee as the difference between her response to the CS+ minus the response to the CS− during the last pair of trials| (the fifth pair) and during the testing period. In this way the index could take values of −1, if the bee responded only to the CS−, 0 if the bee responded equally to both odors, or 1, if it only responded to the rewarded odor. The DIs of the different species were compared using Kruskal–Wallis test. Multiple comparisons were performed by means of Dunn’s comparison.
Regarding the testing phase, where short-term memory was evaluated, we compared for each group the proportion of responses to the rewarded odor and to the non-rewarded one by means of Fisher’s exact test. To see if the discrimination levels attained during the training were sustained in the test phase we compared the discrimination indices during the testing phase with the ones obtained during the fifth (last) trial for each bee by means of Wilcoxon’s matched pair test.
Differential PER conditioning
To evaluate as to what extent these bee species had learned to discriminate between the rewarded and the unrewarded odor and to ascertain that only one of them was associated with the reward, we calculated a DI as the difference between the positive responses to the CS+ and the CS−. This index was evaluated in the last (fifth) training presentation of the odors. Despite the low number of responses found in M. quadrifasciata bees, their mean discrimination indices differed from zero, a fact that means that the index achieved at the last training period was higher than that for the first pair of trials (Hypothesis testing, Student t test for M. quadrifasciata Phe as CS+: t = 5.36, N = 75, P < 0.01, for M. quadrifasciata Lio as CS+: t = 4.52, N = 70, P < 0.01). This clearly demonstrates that M. quadrifasciata workers learned in the PER paradigm to discriminate floral odors.
We compared the indices between A. mellifera and M. quadrifasciata each with both odors (PHE and LIO) that were used as CS+. There were significant differences between the four groups [Kruskal–Wallis test: H(3, N = 295) = 50.42, P = 0.0001; Fig. 3] and evaluation by means of multiple comparisons made it clear that in each species there were no differences regarding the odor used as CS+ (Dunn’s multiple comparisons: A. mellifera CS+ Phe vs. A. mellifera CS+ Lio: Q = 0.15, n.s.; M. quadrifasciata CS+ Phe vs. M. quadrifasciata CS+ Lio: Q = 0.68, n.s.). However, the discrimination indices for A. mellifera bees were significantly higher than those for the M. quadrifasciata bees (Dunn’s multiple comparisons A. mellifera CS+ Phe vs. M. quadrifasciata CS+ Phe: Q = 5.74, P < 0.05; A. mellifera CS+ Phe vs. M. quadrifasciata CS+ Lio: Q = 6.37, P < 0.05; A. mellifera CS+ Lio vs. M. quadrifasciata CS+Phe: Q = 5.60, P < 0.05; A. mellifera CS+Lio vs. M. quadrifasciata CS+ Lio: Q = 6.23, P < 0.05).
The test phase trials were performed 15 min after the differential conditioning. We compared for each species and with each of the odors the proportion of response to the CS+ and to the CS−. In all the groups, the response to the CS+ was higher than to the CS− (Fisher exact test: A. mellifera CS+ Phe: p < 0.01; A. mellifera CS+ Lio: p < 0.01; M. quadrifasciata CS+ Phe: p < 0.01; M. quadrifasciata CS+ Lio: P < 0.01), showing that there is still discrimination between the odors. To ascertain if the discrimination level attained during the training phase was sustained during the test phase, we calculated the DIs during testing and compared them with that of the respective last pair of training trials. There were no differences for any of the four groups (Wilcoxon’s matched pair test for A. mellifera CS+ Phe: T = 99, N = 71, P = 0.56; for A. mellifera CS+ Lio: T = 32, N = 74, P = 0.11; for M. quadrifasciata CS+ Phe: T = 48, N = 71, P = 0.49; for M. quadrifasciata CS+ Lio: T = 39, N = 59, P = 1).
This study shows that workers of the eusocial Africanized honeybees and the stingless bee Melipona quadrifasciata can be conditioned in a classical (Pavlovian) conditioning paradigm. The use of a differential instead of an absolute classical conditioning procedure excludes the possibility that a sensory priming process might be involved during the experimental procedure to which M. quadrifasciata bees were exposed (Schacter and Buckner 1998; Giurfa 2003; Bouton and Moody 2004). Bees had to decide explicitly against a non-rewarded memorized odor in the PER setup to solve this odor discrimination task, indicating that the underlying learning mechanism is of associative nature (Bitterman et al. 1983; Menzel and Giurfa 2001; Bouton and Moody 2004). Differential conditioning is a typical within-subject control in studies of associative learning. So, even though their learning performance was low compared to that of Africanized honeybee workers, the discrimination indices for M. quadrifasciata bees were different from zero, showing that this bee species responded differentially to the rewarded and the non-rewarded odors. This discrimination is maintained 15 min after training indicating that at least short-term memory is formed in M. quadrifasciata. The fact that M. quadrifasciata showed conditioned responses minutes after training suggests a link between active (short-term) memories (i.e., working memory) and longer-lasting forms of memory (for review see Menzel 1999), a fact not analyzed in the present study.
Differential conditioning in Africanized honeybees
Even though honeybees have been thoroughly studied regarding their learning abilities, most studies worked with European strains (Bitterman et al. 1983; Takeda 1961; Menzel and Müller 1996). Research with Africanized honeybees has been developed using classical absolute conditioning (Abramson et al. 1997; Abramson and Aquino 2002), and, as mentioned before, a differential conditioning is a more demanding task (Giurfa 2003). In the studies performed by Abramson and coworkers, the performance of Africanized and European honeybees were compared, showing the latter to be more rapid learners and reaching higher levels of response. We cannot directly compare those results with ours, but it is plausible that Africanized honeybees could also show lower discrimination than the European strains in a comparative and simultaneous study of classical differential conditioning.
PER protocol for stingless bees
When we compare the honeybees’ performance with that of stingless bees, their apparently higher learning abilities could be due to the fact that the PER paradigm has been especially designed and optimized for honeybees, regarding stimulus intensity, the inter-trial intervals, etc. (Takeda 1961; Bitterman et al. 1983). For example, when working with the bumblebee Bombus terrestris, Laloi and coworkers (1999) needed 12 s of odor stimulation and a high concentration sugar solution for them to learn a floral odor, instead of 6 s. We repeated for both stingless bee species the entire protocol with two odor presentation times, 6 and 12 s. Even though the longer stimulus duration did not increase the response in Melipona quadrifasciata, that apparently had reached a maximum performance in this protocol already, it generated a low discrimination index for one of the odors used as CS+ in S. aff. depilis at the middle point of the training phase. This discrimination level, however, was not significant by the end of the procedure. This could be due to the fact that, maybe since they are smaller bees, they get satiated with just a little sucrose solution, or they get too weak when being harnessed for an extended period. This tendency suggests that after adjusting the experimental procedure it might be possible to achieve better performances of S. aff. depilis in the PER paradigm and thus to re-analyze quantitatively learning abilities in this and other trigonine species in the future.
Different learning abilities between species
The differences in their performances could be related to intrinsic differences in learning abilities between species that might be associated with differences in foraging and recruiting strategies. For a scout bee from all the three species, visual stimuli are especially important during the searching for floral patches at long distances. Scent has been suggested to function during the final approach to the food source, inducing the visiting bee to land on the flower (von Frisch 1914). When returning to the hive, however, differences between the three bee species become apparent, since Scaptotrigona workers lay pheromone trails all along the way. In contrast, Melipona workers leave scent spots only near the food source and A. mellifera workers only occasionally leave scent marks in the feeding place surroundings. Honeybee workers returning to a distant food source or newly recruited workers use direction and distance information acquired either by previous experience or inside the hive while following the dance, but when approaching the food source the floral odor becomes a relevant stimulus (von Frisch 1967). Melipona workers fly in the direction that they are familiar with from previous flights or the one indicated during recruitment by nestmates (Jarau et al. 2000), but in the final approach to the food source there are pheromone marks to indicate the precise location of the feeding place (Nieh 1998b). Scaptotrigona bees, when returning to a particular flower patch or when being recruited, may follow the pheromone trail directly from the hive entrance and they also encounter special scent marks indicating the endpoint at the food source (Schmidt et al. 2003). From these observations it seems plausible that during recruitment, even though in the final approach to the flower the food scent should be important for all species, some of them have extra information: so for honeybees it could be especially relevant to learn the association of the odor and the food, while for M. quadrifasciata this would be of intermediate importance, and for S. aff. depilis, learning a food odor might not be very relevant. Further experiments, such as odor-food choices within an operant context, could help to elucidate whether this is really what is happening, or whether stress conditions during the experimental procedure may be masking response patterns.
Ecological differences between species
From an ecological point of view, S. aff. depilis is described as having a small niche breadth, but this is not a result of narrow floral specialization, since workers of the genus Scaptotrigona use a wide array of food plants (more than 100 genera are exploited, Biesmeijer and Slaa 2006). The approximately ten Scaptotrigona species are all medium-sized bees (6–7 mm) with large colony populations (more than 10,000) (Lindauer and Kerr 1958). Even though aggressive at the hive entrance they lack aggressive behavior during foraging, being easily displaced from their floral patches by bigger or more aggressive pollinators (Roubik 1989; Biesmeijer and Slaa 2006). This produces fluctuations in the resources that they may exploit, and therefore variations in the quality of the food that they bring into and store in their nests. Behavioral threshold responses are adjusted to the concentration of sugar solutions available (for honeybees see Lindauer 1948; Pankiw et al. 2004), so the fluctuations found in the SRT-scores for S. aff. depilis could be due to such variations in food availability. In contrast, A. mellifera and Melipona species are bees of similar size. Both have a broad diet spectrum and present a large overlap in their place and time of foraging (Roubik 1978; Wilms et al. 1996; Wilms and Wiechers 1997). This ability to exploit lots of different plant resources simultaneously or sequentially may make them less prone to show variations in time in their general responsiveness to sugar solutions.
Possible new insect model
The PER protocol has been extensively used in honeybees in a great number of studies, allowing the correlation of behavioral responses obtained under controlled conditions with the underlying physiological mechanisms and even the consideration of ecological aspects (Menzel and Müller 1996; Menzel 1999). The establishment of the PER paradigm in M. quadrifasciata bees is the first step for the eventual instauration of a new insect model to quantitatively study associative learning in Neotropical bees and their biotopes. The importance of including further highly eusocial insect species in such studies stems from the fact that information acquisition has direct implications in the social context, and different levels of sociality may be reflected in the acquisition patterns (Heyes and Galef 1996). The standardization of this learning procedure in a new eusocial insect such as the stingless bee Melipona quadrifasciata, opens the possibility to investigate their ability to use olfactory information gained in a given behavioral context in other contexts, as it was found to occur in honeybees (Gerber et al. 1996; Sandoz et al. 2000; Chaffiol et al. 2005). The transfer of associations between different behavioral contexts is relevant for a social bee, and by using the PER paradigm it is possible to investigate aspects regarding prior experience in the social context, as well as possible mechanisms involved in information transfer (Farina et al. 2005; Grüter et al. 2006). For bees of the Meliponini tribe it has already been reported that odor information transfer within their colonies affects searching behavior of recruits (e.g., Trigona iridipennis: Lindauer 1956; Plebeia tica: Aguilar et al. 2005). This will allow future studies to combine both olfactory experiences at the social level with learning assays under laboratory conditions in common-garden experiments with stingless bees (especially Melipona quadrifasciata) and honey bees. Stingless bees are especially interesting because of their close phylogenetic relationship with the genus Apis (Michener 2000) and their similarity in many ecological aspects. Their incorporation within this behavioral framework, thus, opens a promising field for comparative research on olfactory learning at behavioral, cellular and molecular levels.
We specially want to thank Adelino Penatti for his daily help and technical assistance. We are also grateful to two anonymous reviewers for their insightful comments. This study was supported by funds from ANPCYT (01-12319), University of Buenos Aires (X 036) and CONICET (02049), to W. M. Farina and by a CAPES-SECyT international exchange grant (071/04). We declare that our experiments comply with the current laws of the country in which they were performed.