Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

J Comp Physiol A (2006) 192: 691700

DOI 10.1007/s00359-005-0091-4

O R I GI N A L P A P E R

Sayaka Hori Hideaki Takeuchi Kentaro Arikawa


Michiyo Kinoshita Naoko Ichikawa
Masami Sasaki Takeo Kubo

Associative visual learning, color discrimination, and chromatic


adaptation in the harnessed honeybee Apis mellifera L.

Received: 14 March 2005 / Revised: 22 December 2005 / Accepted: 22 December 2005 / Published online: 20 January 2006
 Springer-Verlag 2006

Abstract We studied associative visual learning in har-


nessed honeybees trained with monochromatic lights
Introduction
associated with a reward of sucrose solution delivered to
The excellent visual capacity of the honeybee Apis
the antennae and proboscis, to elicit the proboscis
mellifera L. was rst discovered by von Frisch (1967)
extension reex (PER). We demonstrated ve properties
who demonstrated that bees see and discriminate colors
of visual learning under these conditions. First, antennae
in the appetitive context of food search. Honeybees are
deprivation signicantly increased visual acquisition,
able to discriminate various colors, shapes, and patterns
suggesting that sensory input from the antennae inter-
in a visual conditioning paradigm in which free-ying
feres with visual learning. Second, covering the com-
bees are trained to y towards training/test targets and
pound eyes with silver paste signicantly decreased
are rewarded with sucrose solution if they select the
visual acquisition, while covering the ocelli did not.
correct target (von Frisch 1967; Zhang et al. 2004;
Third, there was no signicant dierence in the visual
Giurfa et al. 2001; Giurfa 2003; Horridge 2003; Stach
acquisition between nurse bees, guard bees, and forag-
et al. 2004). Recently, Giurfa et al. (2001) used honey-
ers. Fourth, bees conditioned with a 540-nm light stim-
bees conditioned with a series of colors and patterns to
ulus exhibited light-induced PER with a 618-nm, but not
demonstrate that bees can form concepts of sameness
with a 439-nm light stimulus. Finally, bees conditioned
and dierence, and can build and generalize congural
with a 540-nm light stimulus exhibited PER immediately
pattern representations (Stach et al. 2004). Such high-
after the 439-nm light was turned o, suggesting that the
order cognitive visual performances of the honeybee are
bees reacted to an afterimage induced by prior adapta-
supported by a remarkable small brain (approximately
tion to the 439-nm light that might be similar to the 540-
1-mm3 volume) containing only approximately 9.6105
nm light.
neurons (Witthoft 1967).
How does the honeybee brain perceive and process
visual information? Three types of photoreceptors have
been identied in the honeybee retina; Short (S),
Medium (M), and Long (L), with maximum sensitivity in
the ultraviolet (S or ultraviolet receptor; kmax=350 nm),
Electronic Supplementary Material Supplementary material is
available for this article at http://dx.doi.org/10.1007/s00359-005- blue (M or blue receptor; kmax=440 nm), and green (L
0091-4 and is accessible for authorized users. or green receptor; kmax=540 nm) regions of the spec-
trum, respectively (Menzel and Blakers 1976). The
S. Hori H. Takeuchi (&) T. Kubo photoreceptor neurons project to the optic lobe, the vi-
Department of Biological Sciences, Graduate School of Science,
The University of Tokyo, 113-0033 Bunkyo-ku, Tokyo, Japan sual center of the insect brain. Neurons that respond to a
E-mail: takeuchi@biol.s.u-tokyo.ac.jp specic color or motion have been identied in the optic
Tel.: +81-3-58414448 lobe (Menzel and Backhaus 1991). Some visual inter-
Fax: +81-3-58414448 neurons project to the collar and/or a part of the basal
K. Arikawa M. Kinoshita ring region of the mushroom body (MB), the insect
Graduate School of Integrated Science, brain region is considered to be involved in high-order
Yokohama City University, 236-0027 Yokohama, Japan sensory processing (Mobbs 1982; Ehmer and Gronen-
berg 2002). The molecular mechanisms and neural
N. Ichikawa M. Sasaki
Faculty of Agriculture, Tamagawa University, network underlying visual processing in the honeybee,
194-8610 Machida, Tokyo, Japan however, are poorly understood.
692

In contrast, the mechanisms underlying olfactory bees could broadly discriminate long wavelength (from
processing have been extensively characterized, be- 500 to 650 nm) from short wavelength (from 400 to
cause harnessed honeybees can be used in olfactory 450 nm) regions of the spectrum. Interestingly, acqui-
conditioning of the proboscis extension reex (PER) sition in this visual conditioning paradigm seemed to
(Takeda 1961; Bitterman and Menzel 1983). The be robust as bees showed no sign of extinction even
advantage of this conditioning paradigm using har- after 30 consecutive CS presentations without any
nessed bees is that, in contrast to visual learning, bees reward at 5-min intervals (Kuwabara 1957).
are immobilized, thus allowing access to their nervous Since Kuwabaras (1957) pioneering work, the idea
system, which can be exposed by carefully cutting the of conditioning PER in harnessed bees with light
cuticle of the head capsule. In this way, olfactory stimuli was essentially abandoned. Fifteen years later,
conditioning of PER can be coupled with experimental Masuhr and Menzel (1972) also reported that bees
techniques, such as electrophysiology (Mauelshagen deprived of their antennae could learn to associate light
1993; Hammer 1993; Abel and Menzel 2001; Muller stimuli and sucrose reward, but no further research was
et al. 2002), calcium imaging (Faber et al. 1999; Faber continued along this line. In 1998, Gerber and Smith
and Menzel 2001; Sachse and Galizia 2003; Sandoz demonstrated that color illumination had an indirect
et al. 2003; Galizia et al. 1999; Guerrieri et al. 2005), role in honeybee olfactory learning, but did not dem-
pharmacology (Muller 1996, 2000; Grunbaum and onstrate specic acquisition of light-reward associa-
Muller 1998; Hammer and Menzel 1998; Lozono et al. tions. On the other hand, it is a common observation
2001), and double-stranded RNA-directed RNA that PER cannot be conditioned in harnessed bees
interference (Farooqui et al. 2003, 2004). Electro- presented with colored cardboards. In such experi-
physiologic techniques were used to identify some key ments, however, antennal deprivation has generally
neurons involved in the CS and US pathways in been neglected. It is therefore possible that such
olfactory-PER associative learning (Hammer 1993; deprivation constitutes a critical factor for acquisition
Mauelshagen 1993; Muller et al 2002). Pharmacologic of the light-reward association, a fact that Kuwabaras
methods were used to demonstrate that long-term (1957) work did neither study nor consider. Addition-
olfactory memory formation requires key signal ally, Kuwabaras (1957) results did not include the
transduction molecules, such as nitric oxide synthase, appropriate controls for associative learning, such as
cyclic AMP-dependent protein kinase A, and protein the unpaired presentation of CS and US, where no
kinase C (Muller 1996, 2000; Grunbaum and Muller associative learning is expected. Moreover, the extent
1998; Lozono et al. 2001). Such a vast amount of to which bees learned to associate chromatic or ach-
research on olfactory learning contrasts with the romatic cues with sucrose reward also remains unclear.
scarcity of information concerning visual learning in In this context, separating the contributions of the
bees, especially with respect to the cellular and compound eyes and of the ocelli to the performances
molecular substrates. Clearly, progress in this area has observed could be a rst step in distinguishing between
been impeded by the fact that visual learning in bees these alternatives.
requires free-ying bees, making it dicult to simul- In the present study, we aimed at reproducing
taneously access their brains. Nevertheless, the rst Kuwabaras (1957) basic experiment in a controlled way
study on PER conditioning was performed using vi- so that a robust protocol for visual associative learning
sual stimuli instead of olfactory stimuli as the CS can be made available under laboratory conditions. We
(Kuwabara 1957). Kuwabara (1957) immobilized bees modied some basic features of this original paradigm
by clipping their wings and elicited the PER by and trained bees with dierent monochromatic lights
touching the tarsae with sucrose solution contained in paired with sucrose solution delivered to the proboscis.
a small spoon. Because the bees responded by PER to We then compared explicitly CSUS paired and
the stimulation of water vapor from the spoon, unpaired groups to determine whether changes in per-
Kuwabara (1957) amputated the antennae to suppress formance over conditioning trials are truly associative.
this confounding factor. Thus, Kuwabara used bees We studied the eect of antenna deprivation on this
that were deprived of their antennae and trained with performance and determined the dierential contribu-
dierent spectral stimuli (from 410 to 630 nm). From tions of ocelli and compound eyes, and whether or not
6 to 38 learning trials were necessary for the bees to bees with dierent task specializations dier in their rate
learn the light-reward association and such a large of acquisition. Here, we demonstrated that bees can
range seemed to be wavelength-dependent, as learning learn to associate light stimuli with sucrose reward; this
long-wavelength stimuli required fewer trials than learning is truly associative and that antennal depriva-
short-wavelength stimuli. The use of a small number tion is critical for such learning to occur. We also
of bees per trained light condition (N=10) and the showed that visual input through the compound eyes is
lack of appropriate controls in which the CS and US necessary and that visual capabilities, such as color
were presented unpaired in his study, however, makes discrimination and chromatic adaptation, can be ana-
it dicult to draw denitive conclusions from the lyzed in harnessed bees by exploiting the PER, as in
results. Kuwabaras (1957) results suggested that olfactory conditioning.
693

attached as a roof providing a 5 cm high ceiling in the


Materials and methods center of the arena (Fig. 1). One side of the arena was
closed with a piece of white paper (76 cm2) to mini-
Preparation of the bees mize interference by uncontrolled visual information.
Light was provided by a 150 W high stable Xenon
European honeybees (Apis mellifera L.) maintained at lamp (Hamamatsu Photonics LC5 and L8253). A light
the University of Tokyo were used for the experiments. guide was used to illuminate the arena. The exit pupil
Colonies were purchased from the Kumagaya Honeybee of the light guide was set 32 cm above the head of the
Farm (Saitama, Japan). For all the experiments except harnessed bee. Monochromatic light was provided by
the one in which workers were compared according to attaching an interference lter (618, 540, or 439 nm;
the division of labor, ying workers were randomly half-bandwidth=ca. 10 nm) at the exit pupil. In most
collected near the hive entrance on sunny days from experiments, bees were conditioned with a 618-nm light
spring to autumn. In the winter, the honeybee colonies (Figs. 3, 4, 5) whose intensity was measured at the
were kept in a greenhouse maintained at 1325C and bees head (Sanso Radiometer 470-D) and adjusted to
the workers were collected similarly. To test for poten- 2.411014 photons/cm2/s using the shutter. This wave-
tial dierences in acquisition due to dierent task spe- length was chosen because it selectively activates L
cialization, we collected foragers, guard bees, and nurse receptors. In free-ying honeybees, L receptors are
bees. Worker bees exhibit age-polytheism such that the important for visual detection and recognition of
labors of workers change according to their age after chromatic stimuli (Giurfa et al. 1996, 1997; Hempel de
eclosion: young workers (nurse bees) are engaged in Ibarra and Giurfa 2003) and for orientation detection
nursing their brood, middle-aged workers (guard bees) (Giger and Srinivasan 1996; Stach et al. 2004). The
in guarding the colony from natural enemies, and older upper wavelength limit of L receptors is 650 nm,
bees (forgers) in foraging for nectar and pollen (Winston whereas that of M receptors is 550 nm (Menzel and
1987). Foragers were captured at the hive entrance and Blakers 1976). Therefore, L receptors are selectively
were collected as workers with pollen loads on their hind stimulated by light in the 550 to 650-nm wavelength
legs. Guard bees were collected as workers that attacked range. For wavelength discrimination and the color
tweezers presented as a decoy at the hive entrance. Nurse adaptation experiment (Figs. 6, 7), the intensity of each
bees were collected as workers that had inserted their monochromatic light was measured at the bees head
heads into brood combs at least twice successively or for and adjusted to 2.411014 photons/cm2/s. The temper-
more than 20 s. To collect nurse bees, frames holding the ature change in the arena due to the monochromatic
sheets of honeycomb were taken out and observed out- illumination was less than 0.2C, even after 10 min of
side of the hive box. continuous illumination.
Captured bees were anesthetized by cooling on ice. Approximately 30 s before each conditioning trial, a
Following Kuwabaras (1957) procedure, both antennae honeybee was placed in the arena (i.e., under the light-
were cut with ne scissors at the base of the scapus, projecting screen). Two groups of bees were condi-
taking care not to pull them. Bees with leaking hemol- tioned. In the explicitly paired group, light (CS) was
ymph were omitted (<1%) from the experiments. The temporally paired with sucrose solution (US). In the
bees were then harnessed in 500-ll plastic centrifuge unpaired groups, the CS was not paired with the US
tubes with small amounts of cotton at the bottom to (Fig. 2a). In the paired groups, bees were exposed to
support the bees weight. The harnessed bees were sta- either a 618- or 540-nm light (CS) for 7 s. Four seconds
bilized using a collar placed between their heads and after the onset of the CS, 1 ll of 1.4 M sucrose
thoraces. The bees were then kept on a 16h:8h light solution (US) was delivered to the honeybees by means
photoperiod (photo-phase 05:0021:00 h) at 25C and of a micropipette directly touched to the proboscis to
6070% humidity, and fed 10 ll of 1.4 M sucrose evoke the PER, and the bees were fed for 3 s (Fig. 2b).
solution every morning and evening. Two days after To avoid a response to the remaining sucrose solution,
being placed in the tubes, harnessed bees that performed the collar was covered with white plastic tape, which
prompt PER when sucrose solution touched the pro- was changed whenever sucrose solution was spilled on
boscis were selected for conditioning experiments. More it. In the paired group, 20 conditioning trials were
than 90% of the total bees passed this screening. presented with a 20-min inter-trial interval (ITI). In the
unpaired groups, the CS and US were presented in
separate, alternating trials separated by 10 min. As for
Associative visual PER conditioning the paired group, the CS lasted 7 s and the US 3 s. To
compare both groups in an equivalent condition with
The selected bees were trained with 10 conditioning respect to an ITI and the number of placements in the
trials per day for 2 days (total 20 trials). Each of the conditioning arena, we added 20 additional placement
harnessed bees was placed in the experimental arena in trials in the paired group interspersed with the condi-
a room illuminated with 40 W uorescent lamps. The tioning trials. In such placement trials, bees were placed
arena consisted of a tube holder (1072.5 cm3) and a in the arena, but did not receive any stimulation.
sheet of translucent light-projecting screen that was We measured the percentage of PER over trials (see
694

Fig. 1 Schematic diagram of


the apparatus used for light-
PER associative learning. Bees
are harnessed in a small tube
and placed in the experimental
arena. The light stimulus is
indirectly irradiated from above
the light-projecting screen using
a Xenon lamp with a
monochromatic lter

Electronic Supplementary Material Table S1) and Corporation, AG-400-100, Japan) to suppress light
analyzed whether performance changed signicantly input. Paint was applied at least twice using a ne brush
using repeated measures ANOVA. The data for bees and checked under a dissection microscope to conrm
that died during the experiments were not included in that the eyes were entirely thickly covered. A non-treated
the analysis. group of bees served as a control. We measured the
percentage of PER over trials and analyzed whether
performance changed signicantly using a one-way
The contribution of ocelli and compound eyes to visual repeated measures ANOVA.
PER conditioning

To determine the visual input mediating visual PER Wavelength discrimination and chromatic adaptation
conditioning, we trained bees in an explicitly paired
manner (see previous), but painted either the compound A group of bees was conditioned with a 540-nm light
eyes and/or the ocelli with silver paste (The Nilaco stimulus explicitly paired with sucrose solution. A bee

Fig. 2 a Stimulus conditions.


The paired group rst received
a light stimulus (conditioned
stimulus, CS; 7 s) and was then
presented with sucrose
(unconditioned stimulus, US;
3 s). Note that the CS and US
co-terminated. The CSUS
pairings were presented with 20-
min inter-trial intervals (ITIs).
In the unpaired group, the CS
(7 s) and US (3 s) were
administered consecutively
every 10 min. b Response of a
conditioned honeybee under
the paired-stimuli condition. 1
Before the light stimulus (CS); 2
PER was evoked by the 618-nm
light stimulus as indicated by a
yellow arrow within the rst 4 s
of the stimulus; 3 the bee was
then fed with sucrose solution
for 3 s. A conditioned bee can
be viewed in Electronic
Supplementary Material Video
S1
695

Results

Establishment of light-PER associative learning

We analyzed the performance of the explicitly paired


and the unpaired groups of bees deprived of their
antennae. Both groups were exposed to a 618-nm stim-
ulus. In the paired group, light stimulation was followed
by sucrose solution; in the unpaired group, light and
sucrose were not temporally associated. Figure 3 shows
the performance of both groups in terms of percentage
PER over the 20 trials. The reactions of individual bees
Fig. 3 Light-PER associative learning in the antennae deprived are shown in Electronic Supplementary material Table
honeybee using 618-nm monochromatic light. In the conditioning S1. A conditioned bee can be viewed in Electronic
procedure, 20 conditioning trials were presented with an approx-
imately 20-min ITI (N=28). In the unpaired group (10 trials/day; Supplementary Material Video S1. In the paired group,
N=28), the CS and US were presented consecutively at 10-min the responsiveness to the CS increased signicantly over
intervals. Proboscis extension was recorded during the rst 4 s of trials (one-way repeated measures ANOVA: F=6.2124,
each CS period without US presentation. The data for bees that P<0.00001), and reached a level of approximately 40%
died during the experiments were not included in the analysis. *
Signicant dierence, P<0.001

was used for wavelength discrimination and chromatic


adaptation experiments, if the CS induced PER three
successive times prior to the presentation of the US
during trials (see Electronic Supplementary Material
Table S2). The bees were tested with a 439-nm mono-
chromatic light, a 540-nm monochromatic light (the
CS), and a 618-nm monochromatic light. Stimulus pre-
sentations lasted 7 s each and were separated by a
20-min interval. These three tests were performed under
extinction conditions (i.e., no reward delivered after
light stimulation). We recorded the number of bees that
exhibited light-induced PER and the number of bees
that extended their proboscis within 7 s after the light
was turned o. We analyzed whether percentage of PER
in the six conditions (439-nm ON, 540-nm ON, 618-nm
ON, 439-nm OFF, 540-nm OFF, and 618-nm OFF)
changed signicantly using a one-way repeated measures
ANOVA.

Statistics

Repeated measures ANOVA was used for between-


group (two-way ANOVA) as well as within-group (one-
way ANOVA) comparisons. Although parametric
analysis of variance is usually not allowed for dichoto-
mous data, such as those of the PER, the Monte Carlo
studies indicate that it is permissible to use ANOVAs for
dichotomous dependent variables under certain condi- Fig. 4 a Eect of antennae deprivation on light-PER associative
tions, which are met by our data: equal cell frequencies learning. Proboscis extension rate for the antenna + (N=20) and
antenna (N=20) bees were plotted. *Signicant dierence,
(Lunney 1970, Giurfa and Malun 2004). ANOVA, P<0.01. b Eect of ocelli and/or compound eye occlusion on
however, can only indicate whether there is a signicant light-PER associative learning. Proboscis extension rates for the
dierence among the levels of the groups. If a signicant four groups were plotted. Number of bees tested were: N=29 for
main eect or interaction was found, the nding was bees with covered compound eyes (compound eye ocellus +);
followed with post hoc tests, the Newman-Keuls tests N=29 for bees with covered ocelli (compound eye + ocellus );
N=29 for bees with both eyes covered (compound eye ocellus );
for repeated measures. The a-level was set to 0.05 (two- N=29 for intact bees (compound eye + ocellus +). * Signicant
tailed) for all analyses. dierence, P<0.05, **P<0.005
696

trials was not observed in the control group (Electronic


Supplementary Material Fig.S1. Two-way repeated
measures ANOVA indicated signicant dierences in
the PER rates between the control (N=10) and test
group (N=10) (F=5.690, P<0.05). These results
strongly suggested that sensory input from the antennae
(olfactory and/or mechanical stimulation) interferes
with the light-PER associative learning in this condi-
tioning paradigm. This might explain why attempts to
achieve visual PER conditioning without antennae
deprivation have been generally unsuccessful.

Contribution of ocelli and compound eyes

Fig. 5 Comparison of learning ability of nurse bees, guard bees, To examine the possible dierential contribution of
and foragers. Proboscis extension rates for the three groups were ocelli and compound eyes to visual PER conditioning,
plotted. Number of bees tested were: N=20 for nurse bees; N=20 we compared the performance of four groups of bees:
for guard bees; N=20 for foragers. There were no signicant
dierences (P>0.1) one in which the ocelli were painted with silver paste,
another in which the compound eyes were painted, a
third group with both the ocelli and compound eyes
on the 16, 17, and 19th trials, even though decay in painted, and a fourth group of intact bees (Fig. 4b).
performance occurred on the 18 and 20th trials. Trials Responsiveness to the CS over trials in two of the groups
14 and 16 through 20 were signicantly dierent from (the intact group and the ocelli painted group) increased
trial 1 (Newman Keuls, post hoc tests, P<0.01). In signicantly (one-way repeated measures ANOVA:
contrast, there was no signicant increase in the F=7.976 and 4.0526, respectively, P<0.00001), while
unpaired group (one-way repeated measures ANOVA: that in the other two groups (the compound eyes painted
F=0.8201, P=0.683). Furthermore, two-way repeated group, and the compound eyes and ocelli-painted group)
measures ANOVA indicated a signicant dierence be- did not (one-way repeated measures ANOVA:
tween the paired and the unpaired groups (F=17.1941, F=1.3952 and 1.4493, respectively). Two-way repeated
P<0.001). The dierence indicated that the increase in measures ANOVA indicated that there were signicant
responsiveness in the paired group was due to associa- dierences between the compound eye-painted group
tive learning of the lightsucrose reward contingency. compared with intact bees (F=11.7170, P<0.005),
Thus, PER can be conditioned using monochromatic whereas there were no signicant dierences between the
lights paired with a sucrose reward. ocelli-painted group and intact bees (F=0.2225,
P>0.5). These results indicate that the visual informa-
tion received by the compound eyes was crucial in this
Eect of antennae deprivation conditioning paradigm.

To analyze whether antenna deprivation is a critical


factor for successful visual PER conditioning, we tested Comparison of the visual learning capacity according
two groups of bees using a 618-nm light as the CS: one to the division of labor of the workers
group was completely deprived of their antennae and the
other was not (Fig. 4a). Although the responsiveness to In the above experiments, we used ying workers col-
the CS over trials in the intact group increased slightly lected randomly near the hive entrance. As the PER rate
(one-way repeated measure ANOVA: F=1.7504, was not greater than 4050% in our conditioning par-
P<0.05) and reached a level of 520%, none of the adigm, we hypothesized that visual learning ability
trials diered signicantly from trial 1 (Newman Keuls, might dier depending on the division of labor of the
post hoc tests). Two-way repeated measures ANOVA workers. To examine whether the learning ability of the
indicated signicant dierences in the PER rates workers diered depending on their labor, we collected
between these two groups (F=7.9502, P<0.01). To nurse bees, guard bees, and foragers separately based on
exclude the possibility that physical damage due to their behavior, and subjected them to the conditioning
antenna deprivation by scissors might enhance the paradigm under the paired condition using a 618-nm
acquisition of a visual learning, we also tested another light stimulus (Fig. 5). Responsiveness to the CS over
control group of bees that were deprived of the distal trials in all three groups increased signicantly (one-way
regions (one segment) of both antennae. The bees were repeated measures ANOVA, F=4.1271 (nurse bees),
assumed to have experienced physical damage similar to F=4.0427 (guard bees), and F=2.2561 (foragers),
that in the test group, but to still receive sensory input P<0.00001). Two-way factor ANOVA demonstrated
from the antennae. As a result, the increase in CRs over that there were no signicant dierences among the
697

Fig. 6 Wavelength discrimination and color adaptation. Bees irradiation at 540 nm (panel 2) and 618 nm (panel 8), whereas it
conditioned to a 540-nm light stimulus were irradiated with 540- was evoked just after the 439-nm light was turned o (panel 6), as
nm (panels 13), 439-nm (panels 46), or 618-nm light stimuli indicated by the yellow arrows. Wavelength discrimination and
(panels 79). Panels 1, 4, and 7 bees before irradiation; panels 2, 5, chromatic adaptation in a harnessed bee can be viewed in
and 8 bees irradiated with light (CS); panels 3, 6, and 9 bees just Electronic Supplementary Material Video S2
after the light was turned o. Note that PER was evoked during

three groups (Fig. 5), indicating that these workers have 439-nm monochromatic light adjusted to the same
almost equivalent ability for light-PER associative photon numbers (2.411014 photons/cm2/s). We used
learning, irrespective of their roles in the colony bees that had three successive CS-induced PERs prior to
(F=0.7101, P>0.1). presentation of the US during trials. Approximately
50% of the bees in the paired group with a 540-nm CS
(N=37) had three successive PERs during 20 trials,
Wavelength discrimination and chromatic adaptation indicating that honeybees can also be conditioned to
light stimuli with a 540-nm stimulus (see Electronic
Free-ying bees have visual capabilities, such as color Supplementary Material Table S2). There was no sig-
discrimination (Giger and Srinivasan 1996; Hempel de nicant dierence between the ratio of bees that exhi-
Ibarra and Giurfa 2003) and chromatic adaptation bited PER upon irradiation with a 618-nm (60%) or
(Neumeyer 1981). We examined whether these visual a 540-nm light stimulus (78%) (N=18) (Figs. 6, 7,
capabilities can be demonstrated using harnessed bees in Newman Keuls, post hoc tests, P>0.05). In contrast,
our paradigm. First, we irradiated bees that had been none of the bees conditioned to the 540-nm light stim-
conditioned to a 540-nm light stimulus with 618- and ulus exhibited PER upon irradiation with a 439-nm light
698

responding to an afterimage induced by a prior adap-


tation to 439-nm light, which is similar to 540-nm light.

Discussion

In the present study, we established light-PER associa-


tive learning using harnessed bees. The bees were con-
ditioned to light stimuli of two dierent wavelengths,
540 and 618 nm. The 540-nm wavelength is the sensi-
tivity maxima of L receptors and 618 nm is exclusively
received by L receptors. Our results are consistent with
previous reports of L receptor involvement in visual
learning and recognition in free-ying bees (Giger and
Fig. 7 Number of bees that exhibited color discrimination and Srinivasan 1996; Giurfa et al. 1996; Hempel de Ibarra
chromatic adaptation. Light ON shows the percentage of bees and Giurfa 2003; Stach et al. 2004). We used this con-
that exhibited PER during light irradiation, while Light OFF
shows the percentage of bees that exhibited PER just after the light
ditioning paradigm to demonstrate some properties of
was turned o. *Signicant dierence, P<0.05 visual associative learning of the harnessed bees. First,
antennae deprivation signicantly enhanced visual
acquisition in this paradigm. The importance of antenna
stimulus (Figs. 6, 7). These results indicate that the 540- deprivation in this conditioning paradigm has been
nm light-conditioned bees generalized between 618- and neglected for almost half a century since Kuwabaras
540-nm light stimuli, whereas they did not generalize report (1957). Although removing the antennae with
between 540 and 439 nm (Newman Keuls, post hoc tests, scissors risks serious injury, the motivation of the bees
P<0.01). To exclude the possibility that bees used total might not be severely aected, as shown by the fact that
brightness or contrast dierences instead of wavelength, they learned and showed normal responsiveness to the
we performed a similar experiment, changing the light US and survival rate. A general decrease in motivation
intensity. We irradiated bees that had been conditioned should result in no learning and/or a signicant decrease
to the 540-nm light stimulus (2.411014 photons/cm2/s) of the response to the US. In our experiment, all bees
with a 439-nm light stimulus with a higher intensity (1.4 deprived of their antennae responded to the US during
times; 3.421014 photons/cm2/s) or decreased intensity training, demonstrating that it was not the case.
of 540-nm monochromatic light (0.41 times; Next, we demonstrated that blocking input to the
9.761013 photons/cm2/s, 0.15 times; 3.691013 pho- compound eyes signicantly impaired light-PER asso-
tons/cm2/s). Only 4.5% of the bees exhibited PER upon ciative learning, suggesting that the visual input received
irradiation with an increased 439-nm light stimulus by the compound eyes is important for learning in this
(N=22). In contrast, 90.0 and 63.6% of bees condi- paradigm. In the honeybee, ocelli perceive the general
tioned to the 540-nm light stimulus exhibited PER upon luminosity of the environment, whereas compound
irradiation with a decreased 540-nm light stimulus eyes allow color, pattern, and shape perception among
(0.41 times, and 0.15 times, respectively; see Electronic others. In our conditioning paradigm, from the various
Supplementary Material Fig. S2. N=22; Newman cues, chromatic and achromatic, provided by the
Keuls, post hoc tests, *P<0.05, **P<0.01). Further- monochromatic lights used as conditioned stimuli, the
more, we also exposed bees to these test lights in a bees probably learned to associate chromatic cues with
dierent order and the result was similar to previous sucrose reward. Our results argue against the possibility
results (data not shown). The results clearly indicated that the honeybees are conditioned with sensory stimuli
that the bees responded to dierences in wavelength, but other than light, such as temperature change and/or
not brightness or contrast. mechanical stimuli.
During this experiment, the bees conditioned to a Next, we demonstrated that the workers have essen-
540-nm light stimulus exhibited PER immediately after tially equivalent learning capacities, irrespective of their
the 439-nm light was turned o, although they did not role in the colony. In contrast, phototaxis of the workers
exhibit PER during the 439-nm light irradiation. The is more developed in forager bees than in nurse bees
percentage of bees that exhibited PER immediately after (Ben-Shahar et al. 2003). In addition, nurse bees are
the cessation of the 439-nm light was approximately inferior to foragers in olfactory learning (Ray and
65% (N=18; Figs. 6, 7). After cessation of the 540- and Ferneyhough 1997, 1999). Our results suggest that the
618-nm light, approximately 15 and 10% of bees neural and/or molecular basis underlying visual learning
exhibited PER, respectively (Figs. 6, 7). These results is not identical to that of phototaxis and/or olfactory
can be accounted for by the eect of chromatic adap- learning. There are two possible reasons why we cannot
tation demonstrated in free-ying bees (Hempel de detect any dierence between foragers and nurse bees in
Ibarra and Giurfa 2003): the bees might exhibit PER this experiment. One is that the nurse bees might develop
699

their visual capacity during the 2-day incubation period web site for details: http://www.hgsc.bcm.tmc.edu/
after xation, because they were kept under laboratory projects/honeybee/). In this visual learning paradigm
conditions in a 16h:8h light photoperiod. Another is that using harnessed bees, in vivo analysis techniques, which
the appetitive motivation of pollen foragers is quite low are used conventionally in olfactory learning (i.e.,
as their appetite is satised prior to returning to the hive. intracellular recording, calcium imaging, and pharma-
It is possible that departing foragers have higher appe- cologic methods), can be applied directly. In addition,
titive motivation and better acquisition performance. gene manipulation techniques, such as RNA interference
Moreover, we were successful in demonstrating color (Farooqui et al. 2003, 2004) and in vivo electroporation
discrimination and chromatic adaptation for the rst (Kunieda and Kubo 2004), can also be applied to the
time in harnessed bees. The harnessed bees could dis- harnessed bees. A combination of behavioral and neu-
tinguish between 540- and 439-nm light stimuli, but not robiologic methods will contribute to clarify the neural
between 540- and 618-nm light stimuli in our condi- circuit and molecular mechanisms underlying visual
tioning paradigm. In the honeybee, both 540- and 618- processing in honeybees.
nm monochromatic light stimulates L receptors almost
exclusively, whereas 439-nm light mainly stimulates M Acknowledgments We would like to express our sincere gratitude to
receptors. Therefore, it is likely that the bees generalized Dr. Martin Giurfa (Universite Paul Sabatier) for his valuable
comments in writing this manuscript and instruction for the
between the 540- and 618-nm light, based on their per- statistical analysis. This work was supported by Grants-in Aid
ceptual similarity. Even more surprisingly, bees condi- from Bio-oriented Technology Research Advancement Institution
tioned to the 540-nm light stimulus exhibited PER as (BRAIN).
soon as the 439-nm light was turned o. This suggests
that adaptation to the 439-nm light induced an after-
image that was indistinguishable from the 540-nm light. References
This is probably equivalent to the successive color
contrast demonstrated in free-ying bees, which has Abel JR, Menzel R (2001) Structure and response patterns of
been explained by adaptation of the spectral receptors olfactory interneurons in the honeybee, Apis mellifera. J Comp
Neurol 437:363383
(Neumeyer 1981). Our results further support that the Ben-Shahar Y, Leung HT, Pak WL, Sokolowski MB, Robinson
bees are not conditioned to an increase in light intensity GE (2003) cGMP-dependent changes in phototaxis: a possible
(i.e., number of photons) and/or an increase in temper- role for the foraging gene in honey bee division of labor. J Exp
ature in this conditioning paradigm. Biol 206:25072515
Bitterman ME, Menzel R (1983) Classical conditioning of pro-
What are the candidate neurons and brain regions boscis extension in honeybees (Apis mellifera). J Comp Psysiol
involved in visual PER associative learning? In olfactory 97:107119
PER associative learning, the octopaminergic VUMmx1 Ehmer B, Gronenberg W (2002) Segregation of visual input to the
neuron (ventral unpaired median cells of maxillary mushroom bodies in the honeybee (Apis mellifera). J Comp
Neurol 451:362373
neuromere 1) is the key neuron that mediates perception Faber T, Joerges J, Menzel R (1999) Associative learning modies
of the US (Hammer and Menzel 1998). The VUMmx1 neural representations of odors in the insect brain. Nat
neuron might also mediate the US in light-PER asso- Neurosci 2:7478
ciative learning, as sucrose was also used as the US in Faber T, Menzel R (2001) Visualizing mushroom body response to
our conditioning paradigm. The VUMmx1 neuron a conditioned odor in honeybees. Naturewissenschaften 88:472
476
innervates the glomeruli of the antennal lobes, the lateral Farooqui T, Robinson K, Vaessin H, Smith BH (2003) Modulation
procerebrum, and lips and basal rings of the MB calyces. of early olfactory processing by an octopaminergic reinforce-
MB calyces can be subdivided into three zones: lips, ment pathway in the honeybee. J Neurosci 23:53705380
collars, and basal rings. The MB basal ring zone receives Farooqui T, Vaessin H, Smith BH (2004) Octopamine receptors in
the honeybee (Apis mellifera) brain and their disruption by
input from both the antennal and optic lobes, whereas RNA-mediated interference. J Insect Physiol 50:701713
the lip and collar zone receive main inputs from the Galizia CG, Sachse S, Rappert A, Menzel R (1999) The glomerular
antennal and optic lobes, respectively (Mobbs 1982, code for odor representation is species specic in the honeybee
1984; Ehmer and Gronenberg 2002). As visual memory Apis mellifera. Nat Neurosci 2:473478
Gerber B, Smith B (1998) Visual modulation of olfactory learning
formation was modulated by antenna deprivation in the in honeybee. J Exp Biol 201:22132217
present conditioning paradigm, the MB basal ring is a Giger AD, Srinivasan MV (1996) Pattern recognition in honeybees:
candidate brain region involved in light-PER associative chromatic properties of orientation analysis. J Comp Physiol A
learning. 178:763769
Using DNA microarray and dierential display Giurfa M, Vorobyev MV, Kevan PG, Menzel R (1996) Detec-
tion of colored stimuli by honeybees: minimum visual angles
methods, we identied several genes selectively expressed and receptor specic contrasts. J Comp Physiol A 178:699
in honeybee MBs that are candidate genes involved in 709
honeybee behavior (Kamikouchi et al. 1998, 2000; Giurfa M, Vorobyev MV, Brandt R, Posner B, Menzel R (1997)
Takeuchi et al. 2001, 2002, 2004). Recently, analysis of Discrimination of colored stimuli by honeybees: alternative use
of achromatic and chromatic signals. J Comp Physiol A
honeybee molecular biology has greatly progressed and 180:235244
the Human Genome Sequencing Center at Baylor Giurfa M, Zhang S, Jenett A, Menzel R, Srinlvasan MV (2001) The
College of Medicine is currently sequencing the honey- concepts of sameness and dierence in an insect. Nature
bee genome (see the Human Genome Sequencing Center 410:930933
700

Giurfa M (2003) Cognitive neuroethology: dissecting non-elemen- Mobbs PG (1984) Neural networks in the mushroom bodies of the
tal learning in a honeybee brain. Curr Opin Neurobiol 13:726 honeybee. J Insect Physiol 30:4358
735 Muller U (1996) Inhibition of nitric oxide synthase impairs a dis-
Giurfa M, Malun D (2004) Associative mechanosensory condi- tinct form of long-term memory in the honeybee, Apis mellifera.
tioning of the proboscis extension reex in honeybees. Learn Neuron 16:541549
Mem 11:294302 Muller U (2000) Prolonged activation of cAMP-dependent protein
Grunbaum L, Muller U (1998) Induction of a specic olfactory kinase during conditioning induces long-term memory in
memory leads to a long-lasting activation of protein kinase C in honeybees. Neuron 27:159168
the antennal lobe of the honeybee. J Neurosci 18:43844392 Muller D, Abel R, Brandt M, Zockler M, Menzel R (2002) Dif-
Guerrieri F, Schubert M, Sandoz JC, Giurfa M (2005) Perceptual ferential parallel processing of olfactory information in the
and neural olfactory similarity in honeybees. PLoS Biol 3:e60 honeybee, Apis mellifera L. J Comp Physiol A 188:359370
Hammer M (1993) An identied neuron mediates the uncondi- Neumeyer C (1981) Chromatic adaptation in the honeybee:
tioned stimulus in associative olfactory learning in honeybees. successive color contrast and color constancy. J Comp Physiol
Nature 366:5963 A 144:543553
Hammer M, Menzel R (1998) Multiple sites of associative odor Ray S, Ferneyhough B (1997) The eects of age on olfactory
learning as revealed by local brain microinjections of octo- learning and memory in the honey bee Apis mellifera. Neuro-
pamine in honeybees. Learn Mem 5:146156 report 8:789793
Hempel de Ibarra NH, Giurfa M (2003) Discrimination of closed Ray S, Ferneyhough B (1999) Behavioral development and olfac-
coloured shapes by honeybees requires only contrast to the long tory learning in the honeybee (Apis mellifera). Dev Psychobiol
wavelength receptor type. Anim Behav 66:903910 34:2127
Horridge A (2003) Visual resolution of the orientation cue by the Sachse S, Galizia CG (2003) The coding of odour-intensity in the
honeybee (Apis mellifera). J Insect Physiol 49:11451152 honeybee antennal lobe: local computation optimizes odour
Kamikouchi A, Takeuchi H, Sawata M, Ohashi K, Natori S, Kubo representation. Eur J Neurosci 18:21192132
T (1998) Preferential expression of the gene for a putative Sandoz JC, Galizia CG, Menzel R (2003) Side-specic olfactory
inositol 1,4,5-trisphosphate receptor homologue in the mush- conditioning leads to more specic odor representation between
room bodies of the brain of the worker honeybee Apis mellifera sides but not within sides in the honeybee antennal lobes.
L. Biochem Biophys Res Commun 242:181186 Neuroscience 120:11371148
Kamikouchi A, Takeuchi H, Sawata M, Natori S, Kubo T (2000) Stach S, Benard L, Giurfa M (2004) Local-feature assembling in
Concentrated expression of Ca2+/calmodulin-dependent pro- visual pattern recognition and generalization in honeybees.
tein kinase II and protein kinase C in the mushroom bodies of Nature 429:758761
the brain of the honeybee Apis mellifera L. J Comp Neurol Takeda K (1961) Classical conditioned response in the honey bee.
417:501510 J Ins Physiol 6:168179
Kunieda T, Kubo T (2004) In vivo gene transfer into the adult Takeuchi H, Kage E, Sawata M, Kamikouchi A, Ohashi K, Ohara
honeybee brain by using electroporation. Biochem Biophys Res M, Fujiyuki T, Kunieda T, Sekimizu K, Natori S, Kubo T
Commun 318:2531 (2001) Identication of a novel gene, Mblk-1, that encodes a
Kuwabara M (1957) Bildung des bedingten reexes von Pavlovs putative transcription factor expressed preferentially in the
Typus bei der Honigbiene, Apis mellica. J Fac Sci Hokkaido large-type Kenyon cells of the honeybee brain. Insect Mol Biol
Univ Zool 13:458464 10:487494
Lozono VC, Armengaug C, Gauthier M (2001) Memory impair- Takeuchi H, Fujiyuki T, Shirai K, Matsuo Y, Kamikouchi A,
ment induced by cholinergic antagonists injected into the Fujinawa Y, Kato A, Tsujimoto A, Kubo T (2002) Identica-
mushroom bodies of the honeybee. J Comp Physiol A 187:249 tion of genes expressed preferentially in the honeybee mush-
254 room bodies by combination of dierential display and cDNA
Lunney GH (1970) Using analysis of dependent variable: an microarray. FEBS Lett 513:230234
empirical study. J Educ Meas 7:263269 Takeuchi H, Yasuda A, Yasuda-Kamatani Y, Sawata M, Matsuo
Masuhr M, Menzel R (1972) In: Wehner R (ed) Information pro- Y, Kato A, Tsujimoto A, Nakajima T, Kubo T (2004) Prepro-
cessing in the visual systems of arthropods. Springer, Berlin tachykinin gene expression in the brain of the honeybee Apis
Heidelberg New York, pp 315322 mellifera. Cell Tiss Res 316:281293
Mauelshagen J (1993) Neural correlates of olfactory learning von Frisch K (1967) The dance language and orientation of bees.
paradigms in an identied neuron in the honeybee brain. Harvard University Press, Cambridge
J Neurophysiol 69:609625 Winston ML (1987) The biology of the honeybee. Harvard
Menzel R, Blakers M (1976) Colour receptors in the bee eye University Press, Cambridge
morphology and spectral sensitivity. J Comp Physiol 108:1133 Witthoft W (1967) Absolute Anzahl und Verteilung der Zellen im
Menzel R, Backhaus W (1991) The perception of colour. In: Hirn der Honigbiene. Z Morphol Oekol Tiere 61:160184
Gouras P (ed) Vision and visual dysfunction, vol 6. Macmillan, Zhang S, Srinivasan MV, Zhu H, Wong J (2004) Grouping of
London, pp 262293 visual objects by honeybees. J Exp Biol 207:32893298
Mobbs PG (1982) The brain of the honeybee Apis mellifera I. The
connections and spatial organization of the mushroom bodies.
Phil Trans R Soc Lond B 298:309354

You might also like