Download Morphology of Feedback Neurons in the Mushroom Body of the

THE JOURNAL OF COMPARATIVE NEUROLOGY 404:114–126 (1999)
Morphology of Feedback Neurons in the
Mushroom Body of the Honeybee,
Apis mellifera
BERND GRÜNEWALD*
Institut für Neurobiologie, Freie Universität Berlin, D-14195 Berlin, Germany
ABSTRACT
The anatomy of ␥-aminobutyric acid (GABA)-immunoreactive, recurrent feedback neurons in the mushroom body (MB) of the honeybee, Apis mellifera, was investigated by using
intraneuropilar injections of cobalt ions and light microscopic techniques. Each MB contains
approximately 110 GABA-immunoreactive neurons, and approximately 50% of them are
feedback neurons, i.e., they connect the MB output regions—the ␣-lobe, ␤-lobe, and pedunculus—with its input regions—the calyces. Their somata are located in the lateral protocerebral
lobe, and their primary neurites project medially and bifurcate near the ␣-lobe. In the
␣-lobe feedback neurons form narrow banded, horizontal arborizations in the dorsal and
median ␣-lobe; each cell innervates a certain ␣-lobe layer. The neurons form additional
branches in the pedunculus and the ␤-lobe. All calycal subcompartments—the lip, collar, and
basal ring—are innervated by feedback neurons. However, individual feedback neurons
innervate exclusively a certain subcompartment in both the median and lateral calyx. Due to
the arrangement of intrinsic Kenyon cells, each calycal subcompartment is connected to its
specific, corresponding layer in the ␣-lobe. Feedback neurons interconnect the ␣-lobe and the
calyces in either a corresponding or a noncorresponding fashion. With respect to their
branching pattern in the ␣-lobe, the basal ring and the collar neuropil receive input from
feedback neurons innervating the corresponding dorsal and median ␣-lobe layers. By
contrast, the lip region, which receives olfactory antennal input, is innervated by feedback
neurons with arborizations in a noncorresponding dorsal ␣-lobe layer. J. Comp. Neurol.
404:114–126, 1999. r 1999 Wiley-Liss, Inc.
Indexing terms: insects; neuroanatomy; protocerebrum; cobalt hexamine chloride; inhibitory
feedback neurons
The mushroom bodies (MB) are central neuropil structures in the protocerebrum of the insect brain. They are
involved essentially in learning behavior and memory
formation in insects (Apis: Menzel et al., 1974; Erber et al.,
1980, 1987; Hammer and Menzel, 1995; Drosophila: Heisenberg et al., 1985; De Belle and Heisenberg, 1994;
Connolly et al., 1996). Since their discovery (Dujardin,
1850), their anatomical organization, their connectivity
within the insect protocerebrum, and their behavioral
development have been described (Kenyon, 1896; Vowles,
1955; Goll, 1967; Steiger, 1967; Pearson, 1971; Schürmann, 1974; Strausfeld, 1976; Weiss, 1981; Mobbs, 1982,
1984; Rybak and Menzel, 1993; Withers et al., 1993;
Fahrbach et al., 1995; for review, see Mobbs, 1985; Schürmann, 1987; Menzel et al., 1994). Detailed neuroanatomical studies of individual MB neurons or groups of neurons,
however, are rare. This study describes the anatomy of
␥-aminobutyric acid (GABA)-immunoreactive feedback
neurons within the MB neuropil of the honeybee, Apis
r 1999 WILEY-LISS, INC.
mellifera. This neuron group is particularly interesting,
because it forms a major component of the insect MB and
may be involved in olfactory learning (Erber et al., 1987;
Hammer, 1997) and olfactory information processing (Laurent and Naraghi, 1994; Stopfer et al., 1997).
Each MB of the worker honeybee consists of some
170,000 densely packed and parallel arranged, intrinsic
Kenyon cells (Witthöft, 1967). Their dendritic arborizations build up the median and lateral calyces, which were
identified ultrastructurally as the main input regions of
the MB (Schürmann, 1974) and are subdivided into the
Grant sponsor: Deutsche Forschungsgemeinschaft; Grant numbers: Pf
128/6 and SFB 515/C5.
*Correspondence to: Bernd Grünewald, Institut für Neurobiologie, Freie
Universität Berlin, Königin-Luise-Str. 28–30, D-14195 Berlin, Germany.
E-mail: [email protected]
Received 18 March 1998; Revised 25 August 1998; Accepted 28 August
1998
MB FEEDBACK NEURONS IN THE HONEYBEE
main subcompartments: lip, collar, and basal ring. Kenyon
cell axons form the two stalks of the pedunculus and
branch halfway in the pedunculus to form the ␣-lobe and
the ␤-lobe, which are the main output regions of the MB
(Schürmann, 1974). Each calycal subcompartment receives input from a certain sensory modality. The lip
region receives olfactory input, the collar receives visual
input, and the basal ring receives mixed chemosensory
and visual input (Mobbs, 1982, 1984; Homberg, 1984;
Gronenberg, 1986). The topographic organization of the
MB indicates that the Kenyon cells maintain this modalityspecific input structure. Their radial arrangement in the
concentric calycal subcompartments is transmitted into a
stratified orientation in the ␣-lobe, and each calycal subcompartment can be allocated to a discrete band in the
␣-lobe, its corresponding zone (Mobbs, 1982; Rybak, 1994).
Therefore, it was concluded that olfactory information is
processed mainly in the ventral parts of the ␣-lobe that
correspond to the lip region, that visual information is
processed mainly in the median ␣-lobe corresponding to
the collar region, and that basal ring information is
represented in the dorsal layers of the ␣-lobe. Output
neurons typically arborize in the ␣-lobe, in the pedunculus,
or in the ␤-lobe and connect the MB with other brain areas
(Mobbs, 1984; Schürmann, 1987; Rybak and Menzel, 1993).
They respond to a large variety of stimuli, and individual
neurons often respond to different sensory modalities
(Erber, 1978; Homberg and Erber, 1979; Gronenberg,
1987; Mauelshagen, 1993). Similar response patterns were
observed in MB output neurons of other species, like
crickets and cockroaches (Schildberger, 1981, 1983, 1984;
Li and Strausfeld, 1997). This indicates that, at least at
the MB output level, the separation between sensory
modalities is abolished. Rather, output neurons form large
dendritic arborizations and, thus, would be capable of
integrating information of different sensory modalities
(Mauelshagen, 1993; Rybak and Menzel, 1993). Therefore,
the MB has been regarded as a sensory integration center
(Erber et al., 1987).
Feedback connections are major components in the
insect MB (ants: Goll, 1967; crickets: Schürmann, 1973;
Schildberger, 1983; flies: Strausfeld, 1976; honeybees:
Mobbs, 1982; Bicker et al., 1985; locusts: Leitch and
Laurent, 1996; moths: Homberg et al., 1987; orthopterans:
Weiss, 1978). In the honeybee MB, they connect the ␣-lobe,
the ␤-lobe, and the pedunculus with the ipsilateral median
and lateral calyx. The feedback neurons form spines in the
lobes and in the pedunculus and form bleb-like structures
in the calyces (Mobbs, 1982; Gronenberg, 1987; Rybak and
Menzel, 1993), indicating postsynaptic terminals in the
lobes and presynaptic endings in the calyces for these
neurons. Feedback neurons, thus, would transmit information from the output of the MB back to its input regions.
Because they are immunoreactive to GABA (Bicker et al.,
1985), feedback neurons probably provide an extrinsic
inhibitory feedback loop of the MB. It is still unclear
whether feedback neurons interconnect corresponding or
noncorresponding zones of the MB. This is interesting,
because, if there is information transfer between noncorresponding MB zones through feedback neurons, then this
may explain several aspects of the multimodal response
behavior of MB output neurons. Therefore, this study
investigates the anatomy of feedback neurons in the
honeybee MB with special attention to the innervated
areas in the ␣-lobe and the calyces. A preliminary account
115
Fig. 1. Schematic diagram of the mushroom body (MB) within the
honeybee brain. Only the right brain hemisphere is represented. The
extrinsic feedback loop is indicated, and arrows represent the putative
flow of information. The corresponding MB zones—lip and ventral
␣-lobe, collar and median ␣-lobe, basal ring and dorsal ␣-lobe—are
each indicated by the same gray shading. Arrows indicate calycal
regions that presumably are innervated by feedback neurons. ␣,
␣-Lobe; ␤ ␤-lobe; MC, median calyx; LC, lateral calyx; Pe, pedunculus;
Li, lip; Co, collar; Br, basal ring; AL, antennal lobe; LO, lobula; ME,
medulla; CB, central body; OC, lateral ocellus. Dorsal (d) and lateral
(l) directions are indicated. Scale bar ⫽ 100 µm.
of this work has appeared in abstract form (Grünewald,
1995).
MATERIALS AND METHODS
Animals and preparation
Worker honeybees, Apis mellifera carnica, were caught
between 9:00 a.m. and 10:00 a.m. at the hive entrance with
a perspex pyramid hold at a distance of 50–100 cm from
the hive entrance. Thus, mostly departing foragers were
caught. The age of each bee was estimated by checking
external criteria (e.g., hairs on the thorax) and internal
morphologic features (e.g., color and number of head
glands). Young bees were excluded from further analyses.
After immobilization by cooling, the bees were fixed in
small metal tubes by a stripe of sticky wax between head
and thorax. After recovering, animals were fed with a drop
of sucrose solution (25% concentration). To allow for stable
and reproducible positioning of the injection pipette, strong
movements of the brain were minimized by using the head
preparation developed by Mauelshagen (1993). For this,
bees were decapitated, the isolated heads were mounted
on a perfusion chamber that was supplied continuously
Fig. 2. Arborizations of feedback neurons in the ␣-lobe and the
pedunculus of the mushroom body. A: Group of several cobalt-stained
neurons. Stained profiles are restricted to the dorsal (d) and median
␣-lobe (m). No staining is observed in the ventral portion (v). Arrow
indicates the ␣-exit. B: The ␣-lobe is subdivided into six horizontal
bands. Dorsal bands 1 and 2 comprise axons of Kenyon cells that form
dendrites in the basal ring neuropil of the calyces. Median band 3
corresponds to the collar region, and ventral bands 4–6 correspond to
the lip region of the ipsilateral calyx (according to Mobbs, 1982).
C: Some feedback neurons form fine, narrow, horizontal arborizations
in the dorsal band 2 of the ␣-lobe (arrows). D–F: Serial sections of a
group of three feedback neurons photographed at 60 µm (D), 150 µm
(E), and 250 µm (F) below the anterior surface of the brain. These
neurons innervate median band 3 of the ␣-lobe. Entrance point of
feedback neurites into the ␣-lobe is indicated (arrow in D). Profiles
that branch radially around the ␣-lobe are also stained (star in D).
Arrows in E and F indicate the stained neurite bundle that runs
within the protocerebrocalycal tract toward the calyces. Weakly
stained profiles in the ventral ␣-lobes (stars within the ␣-lobe in C and
D) belong to nonfeedback neurons. Frontal sections. Dorsal (d),
median (m), and lateral (l) directions are indicated by arrows in this
and all following figures. For other abbreviations, see Figure 1. Scale
bars ⫽ 50 µm.
MB FEEDBACK NEURONS IN THE HONEYBEE
117
with aerated saline [135 mM NaCl, 5 mM KCl, 10 mM
MgCl2, 1.6 mM CaCl2, and 80 mM tris(hydroxymethyl)aminomethane (Tris), pH 7.25]. After stabilizing the brain by
carefully dissecting the esophagus with its connecting
muscles, the anterior surface of the brain was exposed
around the injection site. Injection pipettes (borosilicate
glass; 1.0 mm outer diameter, 0.58 mm inner diameter)
were pulled with a horizontal micropipette puller (P87;
Sutter Instruments, Novato, CA). Pipette tips were filled
with an aqueous 4% solution of cobalt-hexaminechloride
(Sigma, St. Louis, MO), and the tissue resistance of
pipettes ranged between 20 M⍀ and 100 M⍀. Capillaries
were positioned in the mediolateral ␣-lobe, which was easy
to recognize visually. They were lowered vertically into the
brain, and the actual depth of the pipette was monitored
by using a potentiometer connected to the fine adjustment
of the micromanipulator (Leitz, Wetzlar, Germany). Due to
the small tip diameter of pipettes, neural activity could be
recorded extracellularly (10 ⫻ preamplification). Thus, the
site at which the bundle of feedback neurons entered the
␣-lobe at a depth of 80–120 µm depth was identified, and
the injection pipette could be positioned precisely.
Cobalt histology and microscopy
Cobalt ions were injected iontophoretically for 10–30
minutes with 10–20 nA depolarizing current pulses (200msec duration; 1–3 Hz) applied with a stimulator (SD9;
Grass Instrument Co., Quincy, MA) that was connected to
an intracellular preamplifier (Simmonds, Cambridge,
United Kingdom). Thus, small amounts of cobalt ions were
injected into the ␣-lobe to stain only a few cells and to
reduce background staining around the injection site.
After diffusion of cobalt ions (1–3 hours), brains were
excised and placed in phosphate-buffered saline (PBS), pH
7.4, followed by precipitation of cobalt with aqueous 5%
ammonium sulfide solution (4 minutes). Subsequently,
specimens were fixed in Carnoy’s solution (45 minutes)
and block intensified (Bacon and Altman, 1977; Rybak and
Menzel, 1993). After dehydration in graded ethanol, the
specimens were embedded with propylene oxide in Durcupan (Fluka, Deisenhofen, Germany), and frontal and
sagittal sections were cut (20–50 µm, Leitz microtome).
Serial sections were reconstructed by using a camera
lucida attachment and photographed (Agfapan black-andwhite, 25 ASA film; Gevaert NV, Montsel, Belgium) at a
Wild-Polyvar microscope (Leica, Bensheim, Germany). For
this study, 28 specimens were evaluated. Each consisted of
1–20 marked neurons; thus, more than 200 neurons were
stained and analyzed. According to their branching patterns, the feedback neurons could be classified into different classes. However, estimates on the total numbers of
neurons per class were not determined, because such
estimates are unreliable and can be misleading. This is
due to the fact that, depending on the pipette position, one
group of neurons may incorporate more cobalt ions than
another or one group may be stained more homogeneously
than the other. In addition, in those specimens in which many
feedback neurons were stained, the different groups could not
be distinguished unambiguously.
RESULTS
General morphology
The feedback neurons belong to a group of cells termed
A3 neurons by Rybak and Menzel (1993). By counting
Fig. 3. Camera lucida drawing of ␣-lobe projections of feedback
neurons. A: Two neurons with tree-like arborizations in dorsal ␣-lobe
band 1. B: Group of three neurons that innervate band 3 and form
additional branches dorsally around the ␣-lobe. Scale bar ⫽ 100 µm.
stained somata and fibers in some specimens (n ⫽ 3),
it was estimated that they formed a subpopulation of
about 50–60 fibers in the ventral somata cluster of the A3
neurons (A3-v, according to Rybak and Menzel, 1993) out
of a total of approximately 110 GABA-immunoreactive MB
neurons, which is generally consistent with earlier
observations (Bicker et al., 1985; Rybak and Menzel, 1993).
The somata of the feedback neurons are located in a
cluster ventrally in the anterior lateral protocerebral lobe
in a depth of about 80 µm below the anterior brain surface,
close to the border of the lobula (Figs. 1, 7, 9). The primary
neurite projects dorsally and medially and bifurcates
at the dorsolateral margin of the ␣-lobe. One branch
loops ventrally and enters the ␣-lobe laterally at a depth
of 80–120 µm (␣-exit). This branch gives rise to stratified, horizontal arborizations within the ␣-lobe perpendicular to the Kenyon cell axons and sends several
collaterals posteriorly. These run parallel to the Kenyon
cells within the ␣-lobe, bifurcate at the posterior border of
118
B. GRÜNEWALD
Fig. 4. Arborizations of feedback neurons in the ␤-lobe, pedunculus, and basal ring of the calyces. A,B: Some neurons form processes in
the ␤-lobe that are located at the ␤-lobe margins. Two different
specimens. C: Most neurons form loose networks of fine profiles
throughout the whole ␤-lobe extension. In the distal pedunculus,
cobalt staining of feedback neurons is limited to the finger-like
structures that are formed by those Kenyon cell profiles that build up
the ␣-lobe. The stained pedunculus areas correspond to the median
and dorsal ␣-lobe layers. D: Camera lucida drawing of pedunculus and
␤-lobe innervations of a single, cobalt-stained neuron (reconstruction
of the projections of the same neuron in the lip region of the calyces is
shown in Fig. 5D). Arborizations in the proximal median pedunculus
(arrowhead) probably were stained incompletely. E: Group of feedback
neurons that innervates a dorsal portion of the basal ring region
(arrows). Stained profiles also are seen in the proximal pedunculus
(arrowheads). Frontal sections. For abbreviations, see Figure 1. Scale
bars ⫽ 50 µm in A–C and E, 100 µm in D.
the ␣-lobe (about 200 µm deep), and innervate the ␤-lobe
and the pedunculus. The other branch runs outside the
MB within the protocerebral-calycal tract (p.c.t.) in a
dorsal and posterior direction, bifurcates between the
median and lateral calyx, and sends collaterals into the
inner ring tracts (i.r.t.) that run between the peduncular
stalk and the calyx (Figs. 5, 6). Each feedback neuron
innervates both the median and the lateral calyx (see, e.g.,
Fig. 4E). From the i.r.t., several fine branches project
distally into the calycal subcompartment, which is innervated homogeneously (Figs. 5D, 6D). In addition to the
branches within the MBs, the feedback neurons form
small-diameter branches with few and fine, ring-like arborizations around the dorsal ␣-lobe. Some neurons also
possess tree-like arborizations at the bifurcation site between the calyces (Figs. 7, 9). The projection areas of the
feedback neurons are restricted to the ipsilateral brain
hemisphere: The neurons do not project to the contralateral brain site.
Projection areas in the MB compartments
␣ -Lobe. The ␣-lobe is organized in several horizontal
layers from dorsal to ventral (bands 1–6; Mobbs, 1982),
with each layer comprising the axons of Kenyon cells that
MB FEEDBACK NEURONS IN THE HONEYBEE
119
Fig. 5. Cobalt staining of feedback neurons in the calyces. A,B: Two
groups of neurons were stained that innervate either the collar (stars)
or the basal ring (arrowheads). Photomicrographs of two 50-µm frontal
sections of the same specimen (B is 100 µm farther posterior than A).
The arrow in A indicates the inner ring tract. The collar is innervated
densely by stained profiles throughout its whole proximodistal extension (B). The lip regions are free of any cobalt-stained profiles.
C: Another stained neuron that arborizes in the collar region of the
calyx. Innervation of the proximal pedunculus by feedback neurons
(arrowheads) reaches up to the calyces. Arrow indicates the inner ring
tract. D: Camera lucida drawing of the innervation of the collar
neuropil by two cobalt-stained feedback neurons with similar morphology. From the inner ring tract (arrowhead), feedback neurons send
several collaterals distally that innervate the whole collar neuropil.
The arrow indicates the neurite that runs within the protocerebrocalycal tract. For abbreviations, see Figure 1. Scale bars ⫽ 50 µm in A–C,
100 µm in D.
form dendrites in a certain calycal subcompartment (Figs.
1, 2B). The feedback neurons arborize exclusively in the
dorsal and median bands of the ␣-lobe, namely, dorsal
bands 1 and 2 and median band 3 (Fig. 2A). In the
anterior-to-posterior ␣-lobe extension, they innervate an
area that extends from about 30 µm to about 100 µm below
the anterior surface of the ␣-lobe. According to their
arborization pattern in the ␣-lobe, three different types of
neurons can be distinguished: first, loosely stratified neurons that form tree-like and less dense arborization patterns in band 1 or 3 (Figs. 2D, 3); second, narrowly
stratified neurons that form dense bands with a small
anterior-to-posterior extension (Fig. 2C) mostly in the
narrow dorsal band 2; and third, neurons that form only a
few, short, sparsely arborized branches in the median
␣-lobe. These neurons arborize rather extensively in the
pedunculus (Figs. 4D, 10). Any individual feedback neuron
arborizes in only one particular ␣-lobe band and innervates this band in its complete median-to-lateral extension. Feedback neurons do not innervate ventral bands
4–6 of the ␣-lobe.
Pedunculus. Kenyon cell axons in the pedunculus are
arranged in a horseshoe-like pattern with concentric rings
that correspond to the calycal subcompartments (Howse,
1974). Most feedback neurons form finger-like, dense, and
richly arborized projections in the pedunculus in addition
to their ␣-lobe arborizations (Figs. 2E,F, 4C, 5C). The
innervated pedunculus area reaches from anterior, just
120
B. GRÜNEWALD
Fig. 6. Innervation pattern of feedback neurons in the calyces.
A–C: Cobalt-stained group of fibers that arborize either in the lip or in
the basal ring region (arrowheads). The collar does not contain stained
profiles. The arrows indicate the inner ring tract. Serial 50-µm-thick
sections were photographed at three planes (B was cut 150 µm
posterior to A, and C was cut 200 µm posterior to A). D: Camera lucida
drawing of a single neuron that arborizes in the lip and forms bleb-like
terminals. It is the same neuron for which the pedunculus and ␤-lobe
projections are shown in Figure 4D. The arrowhead indicates the
protocerebrocalycal tract, and the arrow indicates the inner ring tract.
For abbreviations, see Figure 1. Scale bars ⫽ 50 µm in A–C, 100 µm
in D.
beneath the ␣-lobe (distal pedunculus), to posterior, at the
border to the calyces (proximal pedunculus). In the distal
pedunculus, the neurons form short, fine arborizations
that are organized in layers (Figs. 2F, 4C) with a mediolateral extension that is limited to the finger-like structures
of those Kenyon cell axon bundles that project frontally to
form the ␣-lobe. In the pedunculus, the feedback neurons
innervate those layers that correspond to the median and
dorsal ␣-lobe bands. Thus, feedback neurons may collect
information along their course in the MB from a certain
Kenyon cell subpopulation. The dorsal pedunculus is divided into a lateral and a median stalk connecting the
lobes with the lateral and median calyx. Individual feedback neurons innervate both stalks up to the very dorsoposterior end (Fig. 10), closely adjacent to the calyces.
␤ -Lobe. The ␤-lobe lies perpendicular to the ␣-lobe, is
inclined toward the median, and is directly adjacent to the
contralateral ␤-lobe. The ␤-lobe is innervated by two
different types of feedback neurons: first, neurons that
form diffuse, loosely, and less ramified arborizations
throughout the whole ␤-lobe, forming a loosely woven
network of profiles within the ␤-lobe (Figs. 4C, 7, 9); and
second, neurons that innervate the ␤-lobe by a few collaterals, whose rather dense ramifications are restricted to
the lateral and median margins of the ␤-lobe (Figs.
4A,B,D, 10).
Calyces. The cup-shaped calyces comprise the dendritic terminals of the Kenyon cells, which form the
radially arranged subcompartments of the lip, collar, and
basal ring. The feedback neurons innervate all subcompartments (Figs. 4E, 5, 6); however, the branching area covered
by an individual cell is restricted to one certain subcompartment. Because the feedback neurons innervate both ipsilateral calyces symmetrically, a given subcompartment in the
median calyx and its corresponding subcompartment in
the lateral calyx are innervated by the same neuron group.
MB FEEDBACK NEURONS IN THE HONEYBEE
Fig. 7. Camera lucida drawing of three FN2 neurons connecting
dorsal ␣-lobe band 1 with the lip region of the calyces. The ␤-lobe is
innervated diffusely. These neurons form additional arborizations
around the ␣-lobe (star) and beneath the calyces (arrowhead). Arborizations in the pedunculus are omitted. Oe, esophagus. For other
abbreviations, see Figure 1. Scale bar ⫽ 100 µm.
The collaterals that run within the i.r.t. send several fine
branches at regularly distances into the distal parts of the
calycal subcompartment (see, e.g., Figs. 5A,C,D, 6D, 8).
Due to its asymmetry in size and shape, the lateral calyx is
innervated by approximately 20–30 branches, and the
median calyx is innervated by 10–15 branches. Each
branch ramifies into a fine dendritic tree of about 50–80
µm in diameter. The areas covered by the dendritic trees of
adjacent branches of an individual neuron overlap partially (Figs. 5D, 10). Thus, the complete calycal subcompartment is innervated homogeneously by feedback neuron
projections. In addition to their arborizations within the
MB, some feedback neurons posses fine processes either
dorsally around the ␣-lobe (Figs. 2D, 3B, 7, 9), at the
bifurcation of the neurons ventrally beneath the calyces
(Fig. 7), or at their neurite close to its entrance into the
lateral ␣-lobe (not shown, but previously described by
Rybak and Menzel, 1993).
Different classes of feedback neurons
The feedback neurons were classified with respect to
their branching areas in the ␣-lobe layers and in the
calycal subcompartments. According to the initial question
of whether the feedback neurons interconnect corresponding or noncorresponding output and input regions, four
121
different classes of feedback neurons (FN1–FN4) can be
distinguished (Table 1).
FN1 neurons connect dorsal band 1 of the ␣-lobe with
the basal ring of the calyces. In the ␣-lobe, their arborizations are less dense, and additional branches are send into
the pedunculus. This group of feedback neurons, which
has been described previously (Rybak and Menzel, 1993),
interconnects corresponding MB zones. FN2 neurons connect dorsal band 1 of the ␣-lobe with the lip region of the
calyces (Fig. 7), and their tree-like arborizations in the
␣-lobe loosely innervate the whole extension of band 1 (Fig.
3A). Some of these neurons form fine processes dorsally
around the ␣-lobe and small dendritic trees at the bifurcation point of the main branch between the calyces. FN2
neurons interconnect noncorresponding MB zones. FN3
neurons project into dorsomedian band 2 of the ␣-lobe and
into the basal ring of the calyces (Figs. 2C, 4E, 8). They
form condensed, narrow, stratified arborizations in the
␣-lobe and also send collaterals into the pedunculus. FN4
neurons connect median band 3 of the ␣-lobe with the
collar of the calyces (Figs. 2D,E, 5C, 9). They densely
innervate this ␣-lobe layer, project toward the pedunculus,
and innervate the ␤-lobe by diffusely arborized processes.
FN3 and FN4 neurons connect corresponding MB zones.
The results indicate that the basal ring receives inhibitory input from its corresponding ␣-lobe bands 1 and 2 and
that the collar receives input from corresponding band 3.
By contrast, the lip is innervated by feedback neurons
(FN2 neurons) of the noncorresponding ␣-lobe band 1,
whose corresponding calycal subcompartment would be
the basal ring.
In addition to these four feedback neuron classes, another type of feedback neuron exists that connects the
pedunculus with the calyces but does not form banded
arborizations in the ␣-lobe. These neurons densely innervate the pedunculus from its distal part at the origin of the
␣-lobe to its proximal part at the calyces. In the median
␣-lobe, these neurons form only few, short, and sparsely
arborized branches (Figs. 4D, 10). This group, which
consists of several different types of neurons connecting
the pedunculus with the lip and collar region of the
calyces, was not characterized further.
DISCUSSION
Technical considerations
Extracellular, intraneuropilar injections of cobalt ions
were employed successfully to reveal the neuronal connectivity of the MB (Strausfeld and Hausen, 1977; Mobbs,
1984; Rybak and Menzel, 1993). The original technique
employed (e.g., see Rybak and Menzel, 1993) to study the
neural connectivity of the MB of the honeybee was modified in this study to stain fewer neurons, to achieve
maximum resolution of fine arborizations, and to inject
cobalt solution directly into the bundle of feedback neurons. Cobalt ions were injected iontophoretically, instead
of pressure injection, by using pipettes with very small tip
diameters. This reduced the pool of injected cobalt ions and
probably injured fewer cells, which resulted in staining
fewer neurons that were distinguishable even in the area
around the injection site. Therefore, reconstructions could
be performed at the level of small cell groups with arborizations that could be identified unambiguously. Spontaneous
neuronal discharge activity could be recorded extracellularly because of the small tip diameters of the pipettes and
the use of an attached intracellular amplifier. Thus, pipette
122
B. GRÜNEWALD
Fig. 8. Anatomical reconstruction of two FN3 neurons that connect the narrow band 2 of the dorsal
␣-lobe with the basal ring of the calyces. Pedunculus arborizations are omitted. Innervation of the ␤-lobe
was stained very weakly in this specimen and was not reconstructed. For abbreviations, see Figure 1.
Scale bar ⫽ 100 µm.
positioning was very precise, and the feedback neuron
bundle was identified at its entrance into the ␣-lobe,
which increased the probability of specifically and exclusively staining feedback neurons. The use of the head
preparation strongly reduced movements of the whole
brain by eliminating rhythmic hemolymph pumping and
muscle contractions. This allowed for stable positioning of
the injection pipette for an extended period of time.
Organization of the MB
Sensory input into the honeybee MB calyces is organized
in a modality-specific way. The lip region is innervated by
projection neurons from the ipsilateral antennal lobe that
transmit chemosensory information into the MB (Mobbs,
1982; Homberg, 1984). The collar is innervated by fiber
tracts from the ipsilateral lobula and medulla that transmit visual information into the MB (Mobbs, 1982; Gronenberg, 1986). The basal ring receives mixed sensory input
from various parts of the brain (Mobbs, 1982). This inputspecific organization is maintained throughout the MB
(Fig. 11) and within the layers of the pedunculus and the
␣-lobe, where information is processed onto MB extrinsic
output (Mobbs, 1982; Rybak and Menzel, 1993). Individual
output neurons often respond to a large variety of stimuli
MB FEEDBACK NEURONS IN THE HONEYBEE
123
Fig. 9. Three FN4 neurons arborize in median ␣-lobe band 3 and
connect this layer with its corresponding collar neuropil of the calyces.
This type of feedback neuron was stained frequently during this study.
The ␤-lobe is innervated diffusely by these neurons. Peduncular
arborization are not shown. This neuron type forms additional fine
and less arborized, ring-like branches dorsally around the ␣-lobe
(arrow). Arrowhead indicates the ␣-exit point. For abbreviations, see
Figure 1. Scale bar ⫽ 100 µm.
of different modalities in the honeybee MB (Erber, 1978;
Homberg and Erber, 1979; Mauelshagen, 1993) and in
crickets (Schildberger, 1981, 1983, 1984) and cockroaches
(Li and Strausfeld, 1997). This indicates that information
from different sensory sources is integrated either at the
level of the MB extrinsic neurons or farther upstream
within the MB, at the level of the Kenyon cells. The finding
that feedback neurons of the honeybee MB connect noncorresponding MB zones, like the dorsal ␣-lobe with the lip
region of the calyx, indicates that there is information
transfer between noncorresponding zones of the MB. Therefore, the feedback neurons may mediate one suggested
function of the MB as a sensory integration center (Erber
et al., 1987; Menzel et al., 1994).
of different sensory stimuli, including olfactory and gustatory stimuli (Erber, 1978; Homberg and Erber, 1979;
Gronenberg, 1987; Grünewald, 1995). Individual neurons
often respond to stimuli of different sensory modalities.
In the calyces, feedback neurons probably form output
synapses onto Kenyon cells. Feedback neurons possess
bleb-like structures at their calycal terminals, indicative of
presynaptic endings (Gronenberg, 1987; Rybak and Menzel, 1993). Cultured Kenyon cells of honeybees express
GABA-mediated chloride currents (Rosenboom et al., 1994).
In the MB of Calliphora, an antibody against a GABAA
receptor subunit stains Kenyon cells (Brotz et al., 1997).
Thus, GABA release from feedback neurons may induce
hyperpolarizing currents in postsynaptic Kenyon cells.
Furthermore, intracellular recordings from Kenyon cells
of Schistocerca MB reveal inhibitory postsynaptic potentials during odor responses that probably are evoked by
activity in MB feedback neurons (Laurent and Naraghi,
1994), and GABA-immunoreactive processes were found
presynaptic to Kenyon cells in the calyces (Leitch and
Laurent, 1996). However, two observations suggest a more
complex pattern of innervation. First, Leitch and Laurent
Feedback neurons of the MB
Feedback neurons form extensive branching areas in the
lobes and in the pedunculus, where they collect information from large numbers of Kenyon cells. Hence, their
morphology suggests that feedback neurons show rather
broad response spectra. Indeed, intracellular recordings
have revealed that the neurons respond to a large variety
124
B. GRÜNEWALD
Fig. 10. Group of two cobalt-stained neurons that form extensive
projections within the pedunculus but that innervate the ␣-lobe very
sparsely (arrowheads). These neurons innervate the pedunculus
densely and form band-like arborizations at the margins of the ␤-lobe.
They innervate the collar region of the calyces. For abbreviations, see
Figure 1. Scale bar ⫽ 100 µm.
TABLE 1. Classification of Feedback Neurons1
subcompartments. However, the subcompartments are
innervated differentially: The collar and basal ring are
innervated by feedback neurons from the corresponding
median (FN4) and dorsal (FN1 and FN3) ␣-lobe layers, and
the lip region is innervated from a noncorresponding
dorsal ␣-lobe layer (FN2) that corresponds to the basal
ring. This finding suggests that processing of olfactory
information within the MB may be different than for
example, visual information processing. Feedback neurons
do not arborize within ventral ␣-lobe bands 4–6, which
would be the corresponding layers of the lip region and the
region where antennal information is processed (Mobbs,
1982; Rybak, 1994; Joerges and Faber, personal communication). Therefore, the question arose: Where do the feedback neurons receive olfactory input if not from Kenyon
cell axons within the ventral ␣-lobe? Several explanations
may be considered. First, although the lip is the main
input area for olfactory input, Mobbs (1982) described
input from the antennal lobe into the basal ring. Antennal
olfactory information, therefore, also may be present in the
dorsal ␣-lobe bands. There, feedback neurons may receive
olfactory input from intrinsic Kenyon cells of the basal
ring. Second, Kenyon cells from various calycal subcompartments may converge onto single ␣-lobe bands (Rybak and
Menzel, 1993; Rybak, 1994), indicating that, within an
individual layer in the pedunculus or the ␣-lobe, different sensory modalities are processed together (Fig. 11).
Band
1
2
3
␣-Lobe
position
Dorsal
Mediodorsal
Median
Calyx
Basal ring
Collar
Lip
FN12
—
—
FN42
FN23
—
—
FN32
—
1According to their ␣-lobe-to-calyx innervation, feedback neurons form corresponding
and noncorresponding connections.
2Corresponding connections.
3Noncorresponding connections.
(1996) described presynaptic GABA terminals in the lobes
and the pedunculi of locust MB. Second, Brotz et al. (1997)
found antibody staining against a GABAA receptor also in
the axons of Calliphora Kenyon cells. Therefore, Kenyon
cells may receive inhibitory input not only in their main
input areas, the calyces, but also at their axonal projections. In the calyces, the feedback neurons synapse probably onto large numbers of Kenyon cells, because the
dendritic field of a Kenyon cell covers an area about 30 µm
in diameter (Mobbs, 1982), whereas any individual feedback neuron homogeneously innervates a complete calycal
subcompartment. Thus, great numbers of Kenyon cells
may be under the inhibitory control of a single feedback
neuron.
The whole MB input region receives inhibitory feedback, because feedback neurons innervate all calycal
MB FEEDBACK NEURONS IN THE HONEYBEE
Fig. 11. Simplified schematic representation of the internal wiring
of the honeybee MB. Arrows indicate the flow of information. Inputs
into the calycal subcompartments are separated into sensory modalities. This modality-specific organization generally is maintained by
the intrinsic Kenyon cells (K1–K5; Mobbs, 1982). In the ␣-lobe,
Kenyon cells synapse onto output neurons and feedback neurons (the
pedunculus is not shown). Feedback neurons (FN1–FN4) recurrently
convey this information back to the calyces. In addition to the
noncorresponding information transfer by feedback neurons of the
FN2 group (see text), a subgroup of Kenyon cells of the collar region
terminates in the ventral ␣-lobe (Rybak, 1994), indicating noncorresponding connections. Some individual Kenyon cells also have dendritic fields in neighboring calycal subcompartments (hatched area
between collar and lip). These Kenyon cells may transmit olfactory
information into the median ␣-lobe (Rybak and Menzel, 1993). The
synaptic organization of the MB, however, may be more complex than
indicated in the diagram. dors, Dorsal; vent, ventral; med, medial. For
other abbreviations, see Figure 1.
Third, the ␤-lobe is a region where feedback neurons
might receive olfactory input. At least some output neurons of
the ␤-lobe respond in an excitatory manner to olfactory stimuli
(Homberg, 1984). Fourth, feedback neurons may get synaptic
input outside the MB, because they form branches in the
surrounding protocerebrum (Figs. 2D, 7, 9; Rybak and Menzel,
1993). MB output neurons that connect the ventral ␣-lobe with
the protocerebrum have overlapping areas with feedback
neurons and may provide additional synaptic input
(Rybak and Menzel, 1993). Recently, Li and Strausfeld (1997) suggested that certain extrinsic neurons in the
MB of cockroaches receive significant input outside the MB.
Possible functions of recurrent inhibition
in the MB
Inhibitory feedback connections are essential neuronal
components in a variety of systems. In the insect MB, they
125
may be involved in olfactory information processing. In
both the locust (Laurent and Naraghi, 1994) and the
honeybee (Stopfer et al., 1997), membrane potentials of
Kenyon cells oscillate in response to odor stimulation. This
odor-evoked oscillatory activity is generated within the
antennal lobe and may be used for olfactory coding (Laurent, 1996). In the MB of the locust, oscillatory activity of
Kenyon cells contains an inhibitory component that probably is mediated by inhibitory feedback neurons (Laurent
and Naraghi, 1994).
Inhibitory feedback loops are a basic feature of brain
structures involved in learning-dependent plasticity, like
the cerebellum (Thompson and Krupa, 1994) and the
hippocampus (Bliss and Collingridge, 1993). During induction of hippocampal long-term potentiation, for example,
GABAergic inhibitory input is reduced, resulting in enhanced postsynaptic depolarization (Bliss and Collingridge, 1993). The MB plays a key role in olfactory memory
formation in insects (for reviews, see Hammer and Menzel,
1995; Davis, 1996; Menzel and Müller, 1996). Inhibitory
feedback neurons may be necessary for the learningdependent response modulations of MB neurons, like the
PE1 neuron, which is an individual, identified MB output
neuron of the honeybee (Mauelshagen, 1993). The PE1
neuron undergoes a decrement in odor response frequency
after associative olfactory learning. Inhibitory feedback
neurons may regulate Kenyon cell activity and, thus, may
modulate synaptic input to the PE1 neuron. Alternatively,
feedback neurons may also control input activity into the
MB by providing presynaptic inhibition of, for example,
olfactory projection neurons. An analysis of the response
characteristics of MB feedback neurons during olfactory
conditioning is currently underway.
ACKNOWLEDGMENTS
The author thanks Dr. Randolf Menzel for his support
and Drs. Randolf Menzel and Jürgen Rybak for critically
reading the paper. The author is grateful to Astrid Klawitter and Sybille Schaare for their expert photographic
assistance.
LITERATURE CITED
Bacon JP, Altman JS. 1977. A silver intensification method for cobalt-filled
neurons in whole-mount preparations. Brain Res 138:359–363.
Bicker G, Schäfer S, Kingan TG. 1985. Mushroom body feedback interneurons in the honeybee show GABA-like immunoreactivity. Brain Res
360:394–397.
Bliss TV, Collingridge GL. 1993. A synaptic model of memory: long-term
potentiation in the hippocampus. Nature 361:31–39.
Brotz TM, Bochenek B, Aronstein K, Ffrench-Constant RH, Borst A. 1997.
␥-Aminobutyric acid receptor distribution in the mushroom bodies of a
fly (Calliphora erythrocephala): a functional subdivision of Kenyon
cells? J Comp Neurol 383:42–48.
Connolly JB, Roberts IJ, Armstrong JD, Kaiser K, Forte M, Tully T, O’Kane
CJ. 1996. Associative learning disrupted by impaired Gs signaling in
Drosophila mushroom bodies. Science 274:2104–2107.
Davis RL. 1996. Physiology and biochemistry of Drosophila learning
mutants. Physiol Rev 76:299–317.
De Belle JS, Heisenberg M. 1994. Associative learning in Drosophila
abolished by chemical ablation of mushroom bodies. Science 263:692–
695.
Dujardin, F. 1850. Memoire sur le systeme nerveux des insects. Ann Sci
Natl Zool 14:195–206.
Erber J. 1978. Response characteristics and after effects of multimodal
neurons in the mushroom body area of the honey bee. Physiol Entomol
3:77–89.
126
Erber J, Masuhr T, Menzel R. 1980. Localization of short-term memory in
the brain of the bee, Apis mellifera. Physiol Entomol 5:343–358.
Erber J, Homberg U, Gronenberg W. 1987. Functional roles of the mushroom bodies in insects. In: Gupta AP, editor. Arthropod brain: its
evolution, development, structure, and functions. New York: John Wiley
&amp Sons, p 485–511.
Fahrbach SE, Giray T, Robinson GE. 1995. Volume changes in the
mushroom bodies of adult honey bee queens. Neurobiol Learn Mem
63:181–191.
Goll W. 1967. Strukturuntersuchungen am Gehirn von Formica. Z Morphol
Ökol Tiere 59:145–210.
Gronenberg W. 1986. Physiological and anatomical properties of optical
input-fibres to the mushroom body in the bee brain. J Insect Physiol
32:695–704.
Gronenberg W. 1987. Anatomical and physiological properties of feedback
neurons of the mushroom bodies in the bee brain. Exp Biol 46:115–125.
Grünewald B. 1995. Sensitization and conditioning—honeybee behaviour
and electrophysiology of mushroom body feedback neurons [abstract].
In: Elsner N, Menzel R, editors. Learning and memory. proceedings of
the 23rd Göttingen Neurobiology Conference. Vol. I, p 29. Thieme,
Stuttgart.
Hammer M. 1997. The neural basis of associative reward learning in
honeybees. Trends Neurosci 20:245–252.
Hammer M, Menzel R. 1995. Learning and memory in the honeybee. J
Neurosci 15:1617–1630.
Heisenberg M, Borst A, Wagner S, Byers D. 1985. Drosophila mushroom
body mutants are deficient in olfactory learning. J Neurogenet 2:1–30.
Homberg U. 1984. Processing of antennal information in extrinsic mushroom body neurons of the bee brain. J Comp Phyiol A 154:825–836.
Homberg U, Erber J. 1979. Response characteristics and identification of
extrinsic mushroom body neurons of the brain. Z Naturf 34c:612–615.
Homberg U, Kingan TH, Hildebrand JG. 1987. Immunocytochemistry of
GABA in the brain and suboesophageal ganglion of Manduca sexta. Cell
Tissue Res 248:1–24.
Howse PE. 1974. Design and function in the insect brain. In: Browne LB,
editor. Experimental analysis of insect behaviour. Berlin: SpringerVerlag, p 180–195.
Kenyon CF. 1896. The brain of the bee. a preliminary contribution to the
morphology of the nervous system of the arthropoda. J Comp Neurol
6:133–210.
Laurent G. 1996. Dynamical representation of odors by oscillating and
evolving neural assemblies. Trends Neurosci 19:489–496.
Laurent G, Naraghi M. 1994. Odorant-induced oscillations in the mushroom bodies of the locust. J Neurosci 14:2993–3004.
Leitch B, Laurent G. 1996. GABAergic synapses in the antennal lobe and
mushroom body of the locust olfactory system. J Comp Neurol 372:487–514.
Li Y-S, Strausfeld NJ. 1997. Morphology and sensory modality of mushroom
body extrinsic neurons in the brain of the cockroach, Periplaneta
americana. J Comp Neurol 387:631–650.
Mauelshagen J. 1993. Neural correlates of olfactory learning paradigms in
an identified neuron in the honeybee brain. J Neurophysiol 69:609–625.
Menzel R, Müller U. 1996. Learning and memory in honeybees: from
behavior to neural substrates. Annu Rev Neurosci 19:379–404.
Menzel R, Erber J, Masuhr T. 1974. Learning and memory in the honeybee.
In: Barton Browne LB, editor. Experimental analysis of insect behaviour. Berlin: Springer Verlag, p 195–217.
Menzel R, Durst C, Erber J, Eichmüller S, Hammer M, Hildebrandt H,
Mauelshagen J, Müller U, Rosenboom H, Rybak J, Schäfer S, Scheidler
A. 1994. The mushroom bodies in the honeybee: from molecules to
behaviour. In: Schildberger K, Elsner N, editors. Fortschritte der
zoologie, vol. 39: neural basis of behavioural adaptations. Stuttgart:
Gustav Fischer, p 81–102.
B. GRÜNEWALD
Mobbs PG. 1982. The brain of the honeybee Apis mellifera. I. the connections and spatial organization of the mushroom bodies. Phil Trans R Soc
London B 298:309–354.
Mobbs PG. 1984. Neural networks in the mushroom bodies of the honeybee.
J Insect Physiol 30:43–58.
Mobbs PG. 1985. Brain structure. In: Kerkut GA, Gilbert LI, editors.
Comprehensive insect physiology, biochemistry, and pharmacology.
Oxford: Pergamon Press, p 299–370.
Pearson L. 1971. The corpora pedunculata of Sphinx ligustri L. and other
lepidoptera: an anatomical study. Phil Trans R Soc London B 259:477–
516.
Rosenboom H, Goldberg F, Schäfer S, Menzel R. 1994. Ionic currents of
Kenyon cells possibly involved in olfactory learning in insects. Soc
Neurosci Abstr 20:333.19.
Rybak J. 1994. Die strukturelle Organisation des Pilzkörpers und die
synaptische Konnektivität protocerebraler Interneuronen im Gehirn
der Honigbiene, Apis mellifera [dissertation]. Berlin: Freie Universität
Berlin.
Rybak J, Menzel R. 1993. Anatomy of the mushroom bodies in the honey bee
brain: the neuronal connections of the alpha-lobe. J Comp Neurol
334:444–465.
Schildberger K. 1981. Some physiological features of mushroom-body
linked fibers in the house cricket brain. Naturwissen 67:623–624.
Schildberger K. 1983. Local interneurons associated with the mushroom
bodies and the central body in the brain of Acheta domesticus. Cell
Tissue Res 230:573–587.
Schildberger K. 1984. Multimodal interneurons in the cricket brain:
properties of identified extrinsic mushroom body cells. J Comp Physiol
A 154:71–79.
Schürmann F-W. 1973. Über die Struktur der Pilzkörper des Insektengehirns III. Die Anatomie der Nervenfasern in den Corpora pedunculata
bei Acheta domesticus L (Orthoptera): Eine Golgi-Studie. Z Zellforsch
145:247–285.
Schürmann F-W. 1974. Bemerkungen zur Funktion der Corpora pedunkulata im Gehirn der Insekten aus morphologischer Sicht. Exp Brain Res
19:406–432.
Schürmann F-W. 1987. The architecture of the mushroom bodies and
related neuropils in the insect brain. In: Gupta AP, editor. Arthropod
brain. its evolution, development, structure, and functions. New York:
John Wiley & Sons, p 231–256.
Steiger U. 1967. Über den Feinbau des Neuropils im Corpus pedunkulatum
der Waldameise. Z Zellforsch 81:511–536.
Stopfer M, Bhagavan S, Smith BH, Laurent G. 1997. Impaired odour
discrimination on desynchronization of odour-encoding neural assemblies. Nature 390:70–74.
Strausfeld NJ. 1976. Atlas of an insect brain. Berlin: Springer Verlag.
Strausfeld NJ, Hausen K. 1977. The resolution of neuronal assemblies after
cobalt injection into neuropil. Proc R Soc London [Biol] 199:463–476.
Thompson RF, Krupa DJ. 1994. Organization of memory traces in the
mammalian brain. Annu Rev Neurosci 17:51–-549.
Vowles DM. 1955. The structure and connections of the corpora pedunculata in bees and ants. Q J Microsc Sci 96:239–255.
Weiss MJ. 1978. Giant fiber associated with mushroom body neuropiles in
an insect brain: a possible one-cell feedback loop. Soc Neurosci Abstr
4:210.
Weiss MJ. 1981. Structural patterns in the corpora pedunculata of orthoptera: a reduced silver analysis. J Comp Neurol 203:515–553.
Withers GS, Fahrbach SE, Robinson GE. 1993. Selective neuroanatomical
plasticity and division of labour in the honeybee. Nature 364:238–240.
Witthöft W. 1967. Absolute Anzahl und Verteilung der Zellen im Hirn der
Honigbiene. Z Morphol Tiere 61:160–184.