Topography of Four Classes of Kenyon Cells in the Mushroom

THE JOURNAL OF COMPARATIVE NEUROLOGY 399:162–175 (1998)
Topography of Four Classes
of Kenyon Cells in the Mushroom Bodies
of the Cockroach
MAKOTO MIZUNAMI,1* MASAYUKI IWASAKI,2 RYUICHI OKADA,1
AND MICHIKO NISHIKAWA2
1Laboratory of Neuro-Cybernetics, Research Institute for Electronic Science,
Hokkaido University, Sapporo 060–0812, Japan
2Faculty of Science, Fukuoka University, Fukuoka 814–0180, Japan
ABSTRACT
Mushroom bodies (MBs), which are higher centers in the insect brain, are implicated in
associative memory and in the control of some behaviors. Intrinsic neurons of the MB, called
Kenyon cells, receive synaptic inputs from axon terminals of input neurons in the calyx. Axons
of Kenyon cells project into the pedunculus and to the a and b lobes, where they make synaptic
connections with dendrites of extrinsic (output) neurons. In this study, we examined the
morphology of Kenyon cells in the cockroach by using Golgi stains and found that they can be
classified into four classes (K1, K2, K3, and K4), according to the diameter, location, and
morphology of the cell bodies, dendrites, and axons. The somata of Kenyon cells of different
classes occupy different concentric zones; Kl cells occupy the most central zone, and K4 cells
occupy the most peripheral zone. The main processes of Kenyon cells of different classes also
occupy different concentric zones in the calyx. Dendrites of K2 and K3 cells are distributed
throughout the calycal neuropil, whereas those of K1 and K4 cells cover the outer and inner
halves of the depth of the neuropil, respectively. In the pedunculus and the a and b lobes,
axons of Kenyon cells of different classes occupy different zones, although the separation is not
complete. A class of extrinsic neurons in the a lobe has dendrite-like arbors that cover the
zones where either K1, K2, or K3 are located. These neurons probably transmit signals of each
class of Kenyon cells. We conclude that, in the cockroach, four classes of Kenyon cells
subdivide the cell body region, pedunculus, and lobes of the MBs, whereas subdivision is less
prominent in the calycal neuropil. J. Comp. Neurol. 399:162–175, 1998. r 1998 Wiley-Liss, Inc.
Indexing terms: insect; brain; neuroanatomy; associative memory; higher center
Mushroom bodies (MBs) are paired neuropil structures
that are located in the center of insect brain. Intrinsic
neurons (Kenyon cells) of the MBs extend dendritic arbors
into the calyces and receive synaptic input from terminals
of input neurons (Steiger, 1967; Schürmann, 1974). The
axon of each Kenyon cell runs through the pedunculus and
then bifurcates at its base; one extends dorsally in the a
lobe, and the other extends medially in the b lobe. In the
pedunculus and lobes, Kenyon cells make synaptic connections with dendrites of extrinsic (output) neurons (Frontali
and Mancini, 1970; Schürmann, 1970). Axons of extrinsic
neurons exit the MB and project to various brain areas
(Rybak and Menzel, 1993; Li and Strausfeld, 1997).
MBs have been implicated in associative memory and in
some forms of motor control (Huber, 1960; Erber et al.,
1980; Mizunami et al., 1993; Davis, 1996). In the honey
bee, Erber et al. (1980) noted that local cooling of the MBs
results in an impairment of consolidation of olfactory-
r 1998 WILEY-LISS, INC.
association memory. In the fruit fly Drosophila, studies
using structural and biochemical mutants suggested that
MBs play an essential role in olfactory memory (Heisenberg et al., 1985; Nighorn et al., 1991; de Belle and
Heizenberg, 1994) and also in courtship behavior (Hall,
1994). In crickets and grasshoppers, electric stimulation
evoked various forms of behavior or its elements; thus, it
was concluded that MBs are responsible for the selection
and coordination of complex behavior patterns (Huber,
1960; Wadepuhl, 1983). In the cockroach, bilateral abla-
Grant sponsor: Ministry of Education, Science, Culture, and Sports of
Japan; Grant sponsor: The Japan Securities Scholarship Foundation;
Grant sponsor: Narishige Zoological Science Award.
*Correspondence to: Dr. Makoto Mizunami, Laboratory of NeuroCybernetics, Research Institute for Electronic Science, Hokkaido University, Sapporo 060-0812, Japan. E-mail: makoto@ncp8.es.hokudai.ac.jp
Received 29 December 1997; Revised 6 May 1998; Accepted 10 May 1998
KENYON CELLS IN THE COCKROACH MUSHROOM BODIES
163
Fig. 1. Frontal sections of a cockroach brain that was stained with osmium-ethyl gallate exhibiting the
gross morphology of a mushroom body. Posterior (A) and anterior (B) sections were photographed.
P, pedunculus; a, a lobe; b, b lobe; MC, medial calyx; LC, lateral calyx. Scale bar 5 200 µm.
Fig. 2. Reconstructions of four classes of Kenyon cells (K1, K2, K3,
and K4) from frontal sections of Golgi-stained brains. The somata of
the K1 cells are located deep within the calycal cup. The somata of the
K2 cells are located at the surface of the calycal cup, those of the K3
cells are at the rim of the cup, and those of the K4 cells are at or just
outside the rim of the cup. The main processes of four classes of
Kenyon cells are arranged concentrically at the base of each calyx: K1
is at the innermost part, and K4 is at the outermost part. NL, neuropil
layer; KFL, Kenyon fiber layer; CBL, cell body layer; D, dorsal; V,
ventral; L, lateral; M, medial.
164
Fig. 3. A–F: Morphology of four classes of Kenyon cells observed in
the calyx of Golgi-impregnated brains. Frontal sections. The main
processes of a pair of K1 cells (A) run along the inner surface of the
Kenyon fiber layer (KFL) of the calyx. They extend a few branches into
the neuropil layer (NL), each of which forms a small, spherical
dendritic field at its tip (arrowheads). A pair of K2 cells (B) extends
branches that repeatedly arborize to cover a wide area in the NL. The
M. MIZUNAMI ET AL.
main process of a K3 cell (C) runs in the outermost part of the KFL.
The main process of a K4 cell (D) runs along the NL. In the
preparations shown in E and F, a number of Kenyon cells could be
identified as K1, K2, K3, or K4 (labelled as 1–4) on the basis of the
locations and diameters of the cell bodies and the main processes.
Scale bar 5 30 µm.
KENYON CELLS IN THE COCKROACH MUSHROOM BODIES
Fig. 4. Some morphological features of K1 cells. A: Some K1 cells
extend side branches that are intermingled among each other at the
inner surface of the Kenyon fiber layer (KFL; arrowheads; sagittal
section). B: Axons of K1 cells run in the innermost part of the KFL in
tion of the MBs caused an impairment of navigation based
on visual spatial memory (Mizunami et al., 1993). Wire
electrode recordings of the activities of extrinsic neurons of
the lobes of the MBs of freely behaving cockroaches
showed that activities of some neurons precede the initiation of specific motor actions, suggesting that the MBs play
a role in the control and possibly in the planning of motor
actions (Mizunami et al., 1993). The neural mechanisms
underlying these higher functions remain to be clarified.
In the accompanying paper (Mizunami et al., 1998), we
report that approximately 15 modular subunits are maintained throughout the output neuropil, i.e., the pedunculus and the a and b lobes, of the MBs of the cockroach.
Each subunit consists of a pair of dark and light slabs, each
of which is formed by the axons of a distinct subset of
intrinsic neurons (Kenyon cells). A class of extrinsic neurons of the pedunculus and lobes possesses segmented,
dendrite-like arborizations that interact with every other
slabs, i.e., with only dark or light slabs; therefore, it was
concluded that the slabs are units that transmit output
signals from the MBs.
In this study, we examined the morphologies of Kenyon
cells in the MBs of the cockroach by using Golgi stains, and
we report here that they can be classified into four
morphological classes. The somata of four classes of Kenyon
cells occupy distinct parts in the cell body region, and their
axons also occupy distinct parts in the pedunculus and
lobes, but there is a high degree of overlap in the distribution of their dendrites in the calycal neuropil. We conclude
that modular subunits in the pedunculus and the lobes are
further organized into four larger groups, according to
morphological classes of Kenyon cells.
Previous studies using Golgi stains in bees (Mobbs, 1982),
ants (Goll, 1967), flies (Strausfeld, 1976), moths (Pearson,
1971), and crickets (Schürmann, 1973) showed that Kenyon
cells can be classified into a few morphological types and
that these types often subdivide the whole or a part of the
MBs. The present findings on cockroaches are compared
with those noted for other insects in order to discuss general
and specific features of the organization of insect MBs.
165
the calyx (arrows) and continue to the most-medial part in the
pedunculus (frontal section). The dorsal side is at the top in A and B.
Scale bars 5 30 µm in A, 100 µm in B.
MATERIALS AND METHODS
Adult male and female cockroaches (Periplaneta americana) from a colony raised in our laboratory were used.
The cockroaches were maintained at 25–27°C and in 12
hour light/12 hour dark conditions. Each cockroach was
anesthetized with cooling on ice, and the head was removed and mounted on a small dish. Impregnations were
carried out according to the Golgi/Colonnier and mixed
Golgi rapid/Colonnier methods (Strausfeld, 1980), the
details of which are described in the accompanying paper
(Mizunami et al., 1998). Golgi-impregnated brains were
dehydrated and embedded in soft Araldite and sectioned at
50–110 µm.
The osmium-ethyl gallate procedure was followed according to Wigglesworth (1957). Isolated heads were mounted
in a dish. The head capsules were opened and immersed in
cacodylate-buffered 2% glutaraldehyde and 1% paraformaldehyde for 24 hours. The brains were then dissected out
and osmicated in 1% osmium tetroxide for 1–2 hours at
4°C. The tissues were washed in buffer and then transferred to 0.5% ethyl gallate for 2 hours at 4°C. The
specimens were dehydrated, embedded in soft Araldite,
and sectioned at 10–12 µm. For reduced silver staining, a
variation of Otsuka (1962) and the original methods of
Bodian impregnation (Bodian, 1936) were used for 10–12
µm paraffin sections. Preparations were observed under
real time three dimensions (Edge Scientific Instruments,
Santa Monica, CA), Nomarski interference contrast (Olympus, Tokyo, Japan), and conventional microscopes. Observations at 1,0003 were made by using an oil-immersion
objective lens. Drawings were made with the aid of a
camera lucida.
RESULTS
We obtained approximately 250 brains in which Kenyon
cells had been impregnated successfully by using Golgi
procedures. In about 50% of these brains, profiles of
166
M. MIZUNAMI ET AL.
Fig. 5. A–E: Serial horizontal sections in which axons of K1 cells
originating from various parts of the medial calyx are traced to the a
lobe. Axons of K1 cells converge at the base of the calyx (A). Axon
bundle of K1 cells from the medial calyx (MC) and that from the lateral
calyx are arranged side by side at the head of the pedunculus (P; B,C)
and occupy the most medial part in the pedunculus (C,D). In B, M and
L indicate axons of Kenyon cells derived from the medial and lateral
calyces, respectively. They continue to the most posterior part in the a
lobe (D,E). Anterior is at the top, and lateral is on the right. Scale
bar 5 50 µm.
individual Kenyon cells could be traced for their whole
extent in serial sections. In the remaining brains, such
large numbers of Kenyon cells were stained that it was not
feasible to trace individual cells. By observing the former
preparations, we classified Kenyon cells into four types
(K1, K2, K3, and K4) on the basis of 1) the diameter and
position of the cell bodies, 2) the diameter and the position
of the main processes in the calyx, and 3) the morphology
of dendritic arborizations. Axons of the Kenyon cells of
different types occupy different zones in the pedunculus
and lobes, although the segregation is not complete. The
presence of four types of Kenyon cells was also evident in
the latter mass-impregnated brains. Here, we first describe the morphology of four classes of Kenyon cells in the
KENYON CELLS IN THE COCKROACH MUSHROOM BODIES
167
calyx; then, we proceed to a description of the morphology
of their axons in the pedunculus and lobes.
Morphology of four classes of Kenyon cells
in the calyx
Each MB of the cockroach contains a lateral and a
medial calyx (Fig. 1A). Each calyx can be divided into inner
and outer layers (Sanchez, 1933; Weiss, 1974). The inner
layer, which we refer to as the Kenyon fiber layer (KFL),
consists of a large number of main processes of Kenyon
cells running in parallel (Figs. 1, 2). Ventrally, the KFL
continues to the pedunculus (Fig. 1A). The outer layer,
which we refer to as the neuropil layer (NL) or calycal
neuropil, is filled with numerous microglomeruli, where
axon terminals of input neurons make synaptic contacts
with dendrites of Kenyon cells (Schürmann, 1974; Weiss,
1974).
K1 cells. The somata of K1 cells are 6–8 µm in
diameter and fill the bottom half of the calycal cups (Figs.
2, 3A,E,F). The main processes originating from the cell
bodies enter the inner surface of the KFL. Immediately
after entering the KFL, some K1 cells give off side branches,
which are connected to one another at the inner surface
of the KFL (Fig. 4A). The main processes of K1 cells
run in the innermost part of the KFL (Figs. 2, 4B). Each
K1 cell extends three to five processes into the NL, each
giving rise to several short arborizations at its tip to form
a spherical dendritic field, the typical diameter of which
is 15–30 µm (Fig. 2). Each dendritic field appears to
cover a single microglomerulus. The dendrites are rich in
spines.
Observations of a large number of K1 cells have shown
that dendrites of K1 cells are distributed in all radial and
concentric zones of the calycal neuropil (see Fig. 5A).
Dendritic arbors of K1 cells are distributed more densely
in the outer half of the depth of the neuropil (Figs. 2, 3A).
K2 cells. The somata of K2 cells are located dorsal to
those of K1 cells and fill a superficial area within the
calycal cups (Figs. 2, 3E,F). The diameters of the somata
are 6–8 µm, similar to those of K1 somata. Their main
processes run outside those of K1, in the central zone of the
KFL (Figs. 2, 3A). Each process extends several branches
in the NL, which send off arborizations to sparsely cover a
large number of microglomeruli (Figs. 2, 3B). The arborizations exhibit numerous spines.
Observations of K2 cells in a number of Golgi-impregnated brains have shown that dendrites of K2 cells are
distributed in all radial and concentric zones of the calycal
neuropil (see Fig. 7A). They also are distributed almost
evenly throughout the depth of the neuropil.
K3 cells. The cell bodies of K3 cells are slightly larger
(7–9 µm in diameter) than those of K1 or K2 cells (Figs. 2,
3C,E,F), and they are located peripheral to those of K2, at
the rim of the calycal cups. The main processes of K3 cells,
which appeared thicker than those of K1 and K2 cells,
enter the KFL and run in the outermost part of the KFL
(Fig. 2). Each process gives off several branches in the NL,
which arborize to sparsely cover a large number of microglomeruli. The arborizations of K3 cells are rich in spines.
It was often difficult to distinguish K2 and K3 cells from
their dendritic morphology. K2 and K3 cells, however, can
be distinguished by the diameters and the positions of
their cell bodies, main processes, and axons.
Fig. 6. The positions of four classes of Kenyon cells (K1–K4) in the
a lobe are mapped into the positions of each modular subunit
(M1–M15). The position of each modular subunit from M1 to M15 in
the a lobe is based on observations of reduced silver preparation
(Mizunami et al., 1998). M1 (shown as 1 on the abscissa) is located at
the posterior end, and M15 (15) is at the anterior end. The ordinate is
the number of preparations in which at least one Kenyon cell of each
class is seen at the position corresponding to that of each modular
subunit from M1 to M15. The number of preparations examined are 23
for K1; 17 and 20 for K2 at M1–M3 and M13–M15, respectively; 19 for
K3; and 10 for K4. Bars above the graph indicate typical distribution
areas of K1, K2, K3, and K4 cells, defined as modular subunits in
which at least one cell for each type has been observed in more than
one in three of the preparations examined.
Observations of a number of preparations have shown
that dendrites of K3 cells cover the whole radial and
concentric zones of the calycal neuropil (see Fig. 8A). They
also are distributed almost evenly throughout the thickness of the neuropil.
K4 cells. The somata of K4 cells are 8–10 µm in
diameter, the largest among the four classes, and they are
located peripheral to those of K3 cells, at or just outside the
rim of calycal cups (Figs. 2, 3D–F). The number of K4 cells
impregnated in each MB is less than ten, which is at least
ten times fewer than those of other classes. The main
processes of K4 cells are the thickest among the four
classes. Most K4 cells run in the NL and give off short
collaterals of 10–30 µm in various directions in the KFL.
Their dendritic arbors are distributed as discrete patches
within a cylinder, each of which appears to cover a
microglomerulus (Figs. 2, 3D). The dendrites exhibit characteristic clawed endings (Fig. 3D). Some K4 cells run in
the outermost part of the KFL and extend several branches
with clawed endings into the NL.
Observations of a number of K4 cells have shown that
dendritic arborizations of K4 cells are confined mostly to
the inner half of the depth of the NL and that they are
distributed more densely in the most peripheral concentric
zone than in the middle or central concentric zones of the
NL (Figs. 2, 3D). Their dendrites are distributed in all
radial areas of the NL.
Projections of axons of four classes
of Kenyon cells from the calyx to the lobes
The axon of each Kenyon cell descends anteroventrally
through the pedunculus and then bifurcates to form the
168
M. MIZUNAMI ET AL.
Fig. 7. A–C: Serial horizontal sections in which K2 cells with
dendrites that cover various parts of medial calyx are traced to the a
lobe. A population of axons of K2 cells of the medial calyx (MC)
converges at its base (A) and runs at the central zone (area surrounded
by dotted line in B) in the pedunculus (P), where axons of Kenyon cells
of the medial calyx (M) and those of the lateral calyx (L) are arranged
side by side. Axons of K2 cells of the medial calyx then fuse with those
of K2 cells of the lateral calyx as they project to the lobe and occupy the
central part in the a lobe (area surrounded by dotted line in C).
Anterior is at the top, and lateral is to the right. Scale bar 5 50 µm.
a lobe and the b lobe (Fig. 1B). Observations of reduced
silver preparations have shown that the pedunculus and a
lobes consist of approximately 15 modular subunits (M1–
M15): The location of each modular subunit in the a lobe is
reported in the accompanying paper (Mizunami et al.,
1998). Here, the positions of axons of four classes of
Kenyon cells in the a lobe have been mapped into the
positions of modular subunits in order to estimate the
relationship between the modular subunits (M1–M15) and
the four morphological classes of Kenyon cells (K1–K4).
K1 cells. In the pedunculus and lobes, K1 cells are
often seen as a group of 40–100 axons, either as a tightly
packed axonal mass (Figs. 4, 5) or as a thin sheet. The
latter is described in the accompanying paper (Mizunami
et al., 1998). In the preparation shown in Figure 5, a
number of K1 cells with dendrites that are located in
various parts of the medial calyx are traced to the a lobe
through the pedunculus in serial horizontal sections.
Axons of K1 cells converge at the base of the calyx (Fig. 5A)
and form a very tightly packed bundle (Figs. 4B, 5B). K1
axons of the medial calyx run side by side with those of the
lateral calyx (Fig. 5B,C) and occupy the most medial part
in the pedunculus (Figs. 4B, 5D). The axons occupy the
most posterior part in the a lobe (Fig. 5D,E) and the b lobe
(not shown).
The distribution of K1 cells in relation to modular
subunits (M1–M15) was examined in the a lobe. Studies
were made on 23 preparations in which more than 20 cells
were identified as K1 cells from their dendritic morphology
in the calyx and traced to the a lobe. At least one K1 cell
was found in the area corresponding to M1 in all 23
preparations, at least one K1 cell was found in M2 in 16
(70%) preparations, one was found in M3 in six (26%)
preparations; and one was found in M4 in one (4%)
preparation (Fig. 6). There was no preparation in which K1
cells were positioned in M5–M15. To facilitate description,
we define the typical distribution area of a class of Kenyon
cells as the area in which at least one cell was located in
more than one in three of the preparations examined.
Following this definition, the typical distribution area of
K1 cells is M1 and M2.
K2 cells. Typically, K2 cells are seen as a group of
20–80 cells, either as a population of axons that are
dispersed sparsely over a wide area (Fig. 7B,C) or as a thin
sheet in the pedunculus and lobes. In Figure 7, axons of a
population of K2 cells originating from various areas of the
medial calyx are traced to the pedunculus and the a lobe in
serial horizontal sections. These axons converge at the
base of the calyx (Fig. 7A) and occupy a central zone in the
bundle of Kenyon cell axons from the medial calyx
KENYON CELLS IN THE COCKROACH MUSHROOM BODIES
169
Fig. 8. A–D: K3 cells of the medial calyx are traced to the a lobe in
serial horizontal sections. Axons of K3 cells of the medial calyx (MC)
converge at its base (A) and occupy the lateral area (areas surrounded
by dotted lines in B and C) in the pedunculus (P). They run side by side
with K3 cells of the lateral calyx (LC) and then are mixed together and
occupy the anterior area in the a lobe (areas surrounded by dotted
lines in C and D). In B, M and L indicate axons of Kenyon cells of the
medial and lateral calyces, respectively. Anterior is at the top, and
lateral is to the right. Scale bar 5 50 µm.
(Fig. 7B). Axons of K2 cells from the medial calyx are
arranged side by side with those from the lateral calyx and
then are mixed together as they project to the lobes. Axons
of K2 cells occupy the central zone in the a lobe (Fig. 7C)
and the b lobe (not shown).
Even without a quantitative study, it was obvious that
the distribution area of K2 cells includes M4–M12. Thus,
our examination of the distribution of K2 cells was confined to areas corresponding to M1–M3 and M13–M15. For
the former, we selected 17 preparations in which more
than 15 cells were identified as K2 cells, based on their
morphology in the calyx, and traced them to the area
corresponding to M1–M3 in the a lobe. There was at least
one K2 cell in the area corresponding to M3 in all 17
preparations, at least one in M2 in 14 (83%) preparations,
and at least one in M1 in four (18%) preparations (Fig. 6).
For the latter, we selected 20 preparations in which more
than ten K2 cells were located in the area corresponding to
M13–M15 in the a lobe. At least one K2 cell was located in
the area corresponding to M13 in all 20 examined preparations, and at least one was located in M14 in ten (50%)
preparations, but no K2 cells were found in the area
170
M. MIZUNAMI ET AL.
Fig. 9. Some morphological features of K3 cells compared with K1
and K2 cells. A,B: Two serial frontal sections showing that axons of
K1, K2, and K3 cells occupy separate areas in the pedunculus. The
main processes of K1, K2, and K3 cells occupy the innermost, central,
and outermost parts of the Kenyon fiber layer (A), and their axons
occupy the medial, central, and lateral parts of the pedunculus,
respectively (B). C,D: A photomicrograph of a horizontal section (C)
and its reconstruction (D) showing that axons of K2 and K3 cells from
the medial part of the medial calyx (MC) run along different routes at
the head of the pedunculus (Ped). Axons of K2 cells project centrally,
and those of K3 cells run along the periphery to reach the lateral part.
Some K3 axons take an anterior route (A), whereas others take a
posterior route (P) to reach the lateral part. Dorsal is at the top in A
and B, and anterior is at the top in C and D. Lateral is to the right in
A–D. Scale bars 5 100 µm in A,B, 30 µm in C.
corresponding to M15 in any preparation (Fig. 6). Thus, we
conclude that the typical distribution area of K2 cells is
M2–M14.
K3 cells. In a number of Golgi preparations, 30–100
axons of K3 cells either were distributed densely in a
narrow area, or they formed a thin sheet in the pedunculus
and lobes (Figs. 8B–D, 9A,B). In Figure 8, axons of a
number of K3 cells derived from various parts of the
medial calyx are traced to the a lobe. These converge at the
base of the calyx (Fig. 8A) and occupy the lateral part in
the pedunculus (Fig. 8B–D). They are arranged side by
side with axons of K3 cells from the lateral calyx and are
KENYON CELLS IN THE COCKROACH MUSHROOM BODIES
mixed together as they project to the lobes. They run in the
anterior part of the a lobe (Fig. 8C) and the b lobe (not
shown).
Separate distributions of axons of K1, K2, and K3 cells
in the pedunculus are seen in serial frontal sections of a
Golgi preparation in Figure 9A,B, in which densely packed
axons of K1 cells occupy the medial edge, K2 cells are
distributed sparsely at the center, and dense axons of K3
cells occupy the lateral edge of the pedunculus (Fig. 9B).
At the head of the pedunculus, axons of K3 cells originating from the medial area of the medial calyx (Fig. 9C,D)
and the lateral calyx (not shown) take a distinct route from
those of K2 cells that originate from the same area.
Whereas axons of K2 cells run straight to the center of the
pedunculus, axons of K3 cells take a detour route along the
periphery of the pedunculus until they reach the lateral
parts. This separation provides convincing evidence that
K2 and K3 cells are of separate morphological classes.
Note that the axons of some K3 cells take an anterior
route, whereas the others take a posterior route to reach
the lateral part (Fig. 9C,D). The separation of K3 axons
indicates that the transformation of concentric arrangements of main processes of four classes of Kenyon cells at
the base of the calyx into a linear arrangement in the
pedunculus requires a break point, which is located medially (see Discussion).
Among 19 preparations in which more than 20 cells were
identified as K3 cells based on their morphology in the
calyx and were traced to the a lobe, at least one K3 cell was
seen in the area corresponding to M15 in seven (37%)
preparations, one was seen in M14 in all 19 preparations,
one was seen in M13 in eight (42%) preparations, and one
was seen in M12 in two (10%) preparations (Fig. 6). K3
cells were never observed in the area corresponding to
M1–M11 in any preparation. We conclude that the typical
distribution area of K3 is M13–M15.
K4 cells. Fewer than ten K4 cells were seen in each
MB. Axonal sheets formed by K4 cells were never observed. Figure 10 shows a K4 cell that was reconstructed
from serial horizontal sections from the medial calyx to the
base of the pedunculus. The cell body is not impregnated.
The axon projects to the most lateral part of the pedunculus and continues to the most anterior parts of the a and b
lobes (not shown). Observations of ten preparations in
which one to eight K4 cells were traced to the a lobe show
that all K4 axons are located in the area corresponding to
M15 (Fig. 6) in either the M15 light slab or the M15 dark
slab.
Morphology of four classes of Kenyon cells
in the pedunculus and lobes
Axons of K1 and K2 cells appear to be the thinnest, and
those of K4 cells appear to be the thickest among the four
classes throughout the pedunculus and the lobes. Axons of
all classes of Kenyon cells exhibit varicosities in the a and
b lobes. Axons of K3 and K4 cells exhibit numerous spines
in addition to varicosities throughout the a lobe (Fig. 11)
and the b lobe (not shown). This is in contrast to K1 and K2
cells, which exhibit spines only rarely. In the pedunculus,
axons of all classes of Kenyon cells are smooth and only
rarely have varicosities or spines.
171
Fig. 10. Reconstruction of a K4 cell from serial horizontal sections
of a Golgi-impregnated brain. Its axon reaches the most lateral part of
the pedunculus (Ped). A, anterior; P, posterior; L, lateral; M, medial.
Dendritic morphology of a class of extrinsic
neurons of the a lobe
An extrinsic neuron of the a lobe is shown in Figure
12A,B. It has three sets of dense, dendrite-like, spiny
arborizations that appear to cover three neighboring slabs,
namely, M15 light slab (M15L), M14 dark slab (M14D),
and M14L: One densely covers the entire region of M15L,
the other densely covers the whole M14L, and the remaining one covers a part of the thickness of M14L (Fig. 12B,
arrowheads). Because these areas match the distribution
area of K3 cells, it is likely that this neuron receives
synaptic inputs mainly from K3 cells. The extrinsic neuron
depicted in Figure 12C has segmented, dendrite-like arbors located in an area roughly corresponding to M4–M12.
Because this area is formed almost exclusively by K2 cells,
the neuron probably receives input from K2 cells. The
segmentation of arborizations indicates that the neuron
interacts with every other slabs, i.e., only either dark or
light slabs (Mizunami et al., 1997). The extrinsic neuron
depicted in Figure 12D extends presumed dendrites in an
area roughly corresponding to M1–M8, the density of
which is very high in M1–M3, where K1 cells are distributed, and much lower in M4–M8, which is formed by K2
cells. This neuron appears to receive input mainly from K1
cells, with some additional input from K2 cells. These
observations suggest that there is a class of extrinsic
neurons that specifically or mainly transmits signals of
each of the K1–K3 cells. We have not yet observed extrinsic
neurons, the dendrites of which are confined to M15,
where K4 cells are distributed.
DISCUSSION
We find that Kenyon cells of the cockroach MBs can be
classified into four classes (K1, K2, K3, and K4), according
to the position, diameter, and morphology of the cell
bodies, dendrites, and axons. The positions of the four
172
M. MIZUNAMI ET AL.
Fig. 11. Morphologies of axons of K1 (A), K2 (B), K3 (C), and K4 (D) cells in frontal sections of the a
lobe. Axons of K1 and K2 cells exhibit varicosities. Axons of K3 and K4 cells are rich in spines
(arrowheads) as well as varicosities. Axons of K4 cells appear to be the thickest, and axons of K1 and K2
cells appear to be the thinnest among the four cell types. Scale bar 5 20 µm.
classes of Kenyon cells are summarized in Figure 13.
These four classes differ in several independent factors;
thus, they represent separate morphological classes rather
than mere variations within a single class. Observations of
Golgi-stained neurons usually do not allow for estimation
of the relative number of each class of neurons, because
Golgi staining often exhibits a prominent selectivity
(Strausfeld, 1980). However, because the area occupied by
each class in the a lobe is K2 . K3 . K1 . K4, it is most
probable that the number of the four classes follows the
same sequence. Here, we regard K2 and K3 cells as the
majority classes and K1 and K4 cells as minority classes.
According to dendritic morphology, the four classes are
grouped into three: K1 cells exhibit small-field, spiny
dendrites; K2 and K3 cells exhibit large-field, spiny dendrites; and K4 cells exhibit small-field, clawed dendrites.
Kenyon cells with spiny and clawed dendrites have been
observed in flies (Strausfeld, 1976), bees (Mobbs, 1982),
moths (Pearson, 1971), and crickets (Schürmann, 1973). It
would be interesting to investigate whether dendrites with
spiny and clawed endings form different types of microglomeruli and make synaptic contact with terminals of different types of input neurons.
Some K1 cells give off side branches, which are tightly
intermingled at the inner surface of the KFL. Note that
these branches are located far from the terminals of input
neurons that are distributed in the outer neuropil layer.
We speculate that these K1 cells form reciprocal synaptic
connections. Connections among dendrites of Kenyon cells
in the calyx have not been noted in any other insects. In
the locust MB, Kenyon cells appear to make reciprocal,
excitatory connections in the pedunculus to synchronize
activities, and the possible functional significance of the
synchronized electrical activity has been discussed (Laurent and Naraghi, 1994).
The somata of Kenyon cells of different classes occupy
different concentric zones in the cell body region; K1 cells
form the most central zone, and K4 cells form the most
peripheral zone (Fig. 13). Somata at the central zone (K1
and K2 cells) are smallest, and those at the most peripheral zone (K4 cells) are the largest. This observation agrees
with previous findings in the cockroaches Blatta orientalis
(Sanchez, 1933) and Periplaneta americana (Weiss, 1974),
in which somata at the peripheral zone appear larger than
those at the central zone. Weiss also noted in his reduced
silver study that a class of Kenyon cells with large somata
near the calycal rim tangentially enter the outer neuropil
layer of the calyx and that their axons occupy the peripheral rim at the head of the pedunculus. These cells are
apparently the same as K4 cells.
The main processes of four classes of Kenyon cells are
arranged concentrically at the bases of calyces and are
converted into a quasilinear arrangement of flattened
layers at the pedunculus and lobes (Fig. 13). This is similar
to the findings by Mobbs (1982) in the honey bee, in which
the calyces are separated into three concentric zones. The
concentric arrangement of Kenyon cells in the calyces is
transformed to a linear arrangement in the pedunculus,
which continues to the lobes. This transformation requires
a break point, which is located medially in the cockroach,
as discussed above for K3 cells (Fig. 9C,D); but is located
posteriorly in the honey bee (Mobbs, 1982). The ant, like
the bee, has its calyces organized in a series of concentric
neuropil compartments; however, in the a and b lobes, the
polar organization is maintained rather than being transformed into a Cartesian organization (Goll, 1967). In the
cricket, Schürmann (1973) noted a maintenance of the
calycal Kenyon cell groupings in the pedunculus and the a
lobe. In the fruit fly Drosophila, there is a subdivision
according to biochemical types of Kenyon cells (Yang et al.,
KENYON CELLS IN THE COCKROACH MUSHROOM BODIES
173
Fig. 12. A-D: Morphologies of dendrite-like arborizations of three
extrinsic neurons seen in horizontal sections of the a lobe. The neuron
depicted in A and B has three sets of dendritic arbors, two of which
cover each of two neighboring slabs, M15 light slab (M15L; A,B) and
M14 dark slab (M14D; B), and the remaining one sparsely covers part
of the thickness of 14L (B, arrowheads). These areas match the
distribution area of K3 cells. The neuron depicted in C has dendritic
arbors that cover an area roughly corresponding to M4–M12, where
K2 cells are located. Note that dendrites of this neuron have a
prominent segmentation and appear to interact with only either dark
or light slabs (Mizunami et al., 1997). The neuron depicted in D has
dendrites that cover M1–M8, the density of which is very high at
M1–M3, where K1 cells are located, and it is much lower at M4–M8,
where K2 cells are located. Anterior is at the top. Scale bar 5 30 µm.
1995), and ablation of a specific type of Kenyon cell results
in a defect in sexual behavior, whereas that of other types
does not (O’Dell et al., 1995).
In the a and b lobes of the MBs of the cockroach, axons of
K1 and K2 cells exhibited numerous varicosities, whereas
those of K3 and K4 cells exhibited numerous spines as well
as varicosities. Because varicosities are interpreted as
presynaptic, and spines are considered to be postsynaptic
(Strausfeld, 1976), it is likely that K3 and K4 cells are both
pre- and postsynaptic. Frontali and Mancini (1970) noted
in their electron microscopic studies of the cockroach a
lobe that some Kenyon cells make synaptic connections not
only with extrinsic neurons but also with other Kenyon
cells. Li and Strausfeld (1997) noted that some neurons of
the lobes had varicose arborizations and considered these
neurons as input neurons to the lobes. It would be interesting to investigate whether presynaptic elements of K3 and
K4 cells are Kenyon cells, or input neurons, or both.
Although there is a clear subdivision according to morphological classes of Kenyon cells in the cell body region,
174
M. MIZUNAMI ET AL.
Fig. 13. Schematic representations of the position of four classes of
Kenyon cells in the mushroom body of the cockroach. K1, K2, K3, and
K4 cells (labelled as 1–4) are arranged concentrically in the cell body
region. The main processes of four classes of Kenyon cells are arranged
concentrically at the base of the calyx (not shown). In the calycal
neuropil, dendrites of K2 and K3 cells are distributed almost evenly
throughout the neuropil, whereas those of K1 or K4 cells are more
dense in the outer (O) or inner (I) halves, respectively, of the depth of
the neuropil. Axons of four classes of Kenyon cells are arranged
quasilinearly, with some degree of overlap, in the pedunculus and in
the a and b lobes (not shown). A, anterior; P, posterior; L, lateral;
M, medial.
pedunculus, and lobes, subdivision of calycal neuropil
according to dendrites of four classes of Kenyon cells is less
prominent. It is evident that K1 and K4 cells, the minority
types, subdivide the neuropil: Dendrites of K1 and K4 cells
are distributed more densely in the outer and inner halves
of the depth of the neuropil, respectively. Dendrites of K2
and K3 cells, the majority types, however, are distributed
more or less evenly throughout the neuropil, which obscures the subdivision formed by minority types. This
differs from reports for MBs of bees (Mobbs, 1982) and ants
(Goll, 1967), in which there is a prominent subdivision of
the calycal neuropil according to morphological types of
Kenyon cells. In bees, for example, there are three concentric divisions in the calycal neuropil that are related to
sensory modalities of input neurons: The olfactory input
goes to the outermost part (lip), the visual input goes to the
median part (collar), and both olfactory and visual input
and, possibly, mechanosensory inputs go to the central
part (basal ring; Mobbs, 1982).
Recently, Nishikawa et al. (1998) found a subdivision in
the calycal neuropil of the MB of the cockroach according
to the distribution patterns of terminals of different types
of input neurons: Some deutocerebral and protocerebral
input neurons terminate in specific concentric zones in the
calycal neuropil. Thus, it is likely that different subsets of
Kenyon cells of a particular morphological class extend
dendrites in different concentric zones and receive synaptic input from terminals of different types of input neurons. Nishikawa et al. also found that visual input neurons
extend terminal arborizations in the inner layer of the
calycal neuropil. It is interesting to note that dendrites of
K4 cells are almost confined within this layer. In addition,
there is a class of GABA-immunoreactive, multimodal
input neurons that has been found to extend terminal
arborizations in the entire calycal neuropil (Nishino and
Mizunami, 1998; Yamazaki et al., 1998). The patterns of
distribution of dendrites of individual Kenyon cells of each
class need to be studied to clarify further the functional
subdivisions of the calycal neuropil in the MB of the
cockroach.
We conclude that there are at least two distinct types of
structural modules in the MB of the cockroach. The first
consists of approximately 15 repetitive modular subunits,
which are described in the accompanying paper (Mizunami et al., 1998). The second consists of four subdivisions
according to morphological classes of Kenyon cells. It
would be interesting to clarify whether these structural
modules act as functional modules, like the functional
columns in the mammalian cerebral cortex (Hubel and
Wiesel, 1974; Fujita et al., 1992).
ACKNOWLEDGMENTS
We thank Dr. N.J. Strausfeld for helpful discussions and
M. Ohara for comments.
LITERATURE CITED
Bodian, D. (1936) A new method for staining nerve fibres and nerve endings
in mounted paraffin sections. Anat. Rec. 65:89–97.
Davis, R.L. (1996) Physiology and biochemistry of Drosophila learning
mutants. Physiol. Rev. 76:299–317.
KENYON CELLS IN THE COCKROACH MUSHROOM BODIES
de Belle, J.S. and M. Heisenberg (1994) Associative odor learning in
Drosophila abolished by chemical ablation of mushroom bodies. Science
263:692–695.
Erber, J., T. Masuhr, and R. Menzel (1980) Localization of short-term
memory in the brain of the bee, Apis mellifera. Physiol. Entomol.
5:343–358.
Frontali, N. and G. Mancini (1970) Studies on the neuronal organization of
cockroach corpora pedunculata. J. Insect Physiol. 16:2293–2301.
Fujita, I., K. Tanaka, M. Ito, and K. Cheng (1992) Columns for visual
features of objects in the monkey inferotemporal cortex. Nature 360:343–
346.
Goll, W. (1967) Strukturuntersuchungen am Gehirn von Formica. Z.
Morphol. Öekol. Tiere 59:143–210.
Hall, J.C. (1994) The mating of a fly. Science 264:1702–1714.
Heisenberg, M., A. Borst, S. Wagner, and D. Byers (1985) Drosophila
mushroom body mutants are deficient in olfactory learning. J. Neurogen. 2:1–30.
Hubel, D.H. and T.N. Wiesel (1974) Sequence regularity and geometry of
orientation columns in the monkey striate cortex. J. Comp. Neurol.
158:267–294.
Huber, F. (1960) Untersuchungen über die Funktion des Zentralnervensystems und insbesondere des Gehirnes bei der Fortbewegung und der
Lauterzeugung der Grillen. Z. Vergl. Physiol. 44:60–132.
Laurent, G. and M. Naraghi (1994) Odorant-induced oscillations in the
mushroom bodies of the locust. J. Neurosci. 14:2993–3004.
Li, Y. and N.J. Strausfeld (1997) Morphology and sensory modality of
mushroom body extrinsic neurons in the brain of the cockroach,
Periplaneta americana. J. Comp. Neurol. 387:631–650.
Mizunami, M., J.M. Weibrecht, and N.J. Strausfeld (1993) A new role for
the insect mushroom bodies: Place memory and motor control. In R.D.
Beer, R.E. Ritzmann, and T. McKenna (eds): Biological Neural Networks in Invertebrate Neuroethology and Robotics. Cambridge: Academic Press, pp. 199–225.
Mizunami, M., M. Iwasaki, M. Nishikawa, and R. Okada (1997) Modular
structures in the mushroom body of the cockroach. Neurosci. Lett.
229:153–156.
Mizunami, M., M. Iwasaki, R. Okada, and M. Nishikawa (1998) Topography
of modular subunits in the mushroom bodies of the cockroach. J. Comp.
Neurol. 399:153-161.
Mobbs, P.G. (1982) The brain of the honeybee Apis mellifera. I. The
connections and spatial organization of the mushroom bodies. Phil.
Trans. Soc. London B. 298:309–354.
Nighorn, A., M.J. Healy, and R.L. Davis (1991) The cyclic AMP phosphodiesterase encoded by the Drosophila dunce gene is concentrated in the
mushroom body neuropil. Neuron 6:455–467.
Nishikawa, M., H. Nishino, M. Mizunami, and F. Yokohari (1998) Functionspecific distribution patterns of axon terminals of input neurons in the
calyces of the mushroom body of the cockroach, Periplaneta americana.
Neurosci. Lett. 245:33–36.
175
Nishino, H. and M. Mizunami (1998) Giant input neurons of the mushroom
body: intracellular recording and staining in the cockroach. Neurosci.
Lett. 246:57–60.
O’Dell, K.M.C., J.D. Armstrong, M.Y. Yang, and K. Kaiser (1995) Functional
dissection of the Drosophila mushroom bodies by selective feminization
by genetically defined subcompartments. Neuron 15:55–61.
Otsuka, N. (1962) Histologisch-entowicklungsgeschichtliche Untersuchungen an Mauthnerschen Zellen von Fischen. Z. Zellforsh. Mikrosk. Anat.
58:33–50.
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.
Rybak, J., and R. Menzel (1993) Anatomy of the mushroom bodies in the
honey bee brain: The neuronal connections of the alpha-lobe. J. Comp.
Neurol. 334:444–465.
Sanchez, D. (1933) Contribution a la connaissance de la structure des corps
fongiformes (calices) et de leurs pedicules chez la blatte commune
(Stylopyga [Blatta] orientalis, L.). Trab. Lab. Invest. Biol. Univ. Madrid
28:149–185.
Schürmann, F.W. (1970) Über die Struktur der Pilzkörper des Insektenhirns. I. Synapsen im Pedunculus. Z. Zellforsch. 103:365–381.
Schürmann, F.W. (1973) Über die Struktur der Pilzkörper des Insektenhirns. 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 pedunculata im Gehirn der Insekten aus morphologischer Sicht. Exp. Brain Res.
19:406–432.
Steiger U. (1967) Über den Feinbau des Neuropils im Corpus pedunculatum der Waldameise. Z. Zellforsch. 81:511–536.
Strausfeld, N.J. (1976) Atlas of an Insect Brain. Berlin: Springer.
Strausfeld, N.J. (1980) The Golgi method. In N.J. Strausfeld and T.A. Miller
(eds): Neuroanatomical Techniques. Berlin: Springer, pp. 132–205.
Wadepuhl, M. (1983) Control of grasshopper singing behaviour by the
brain: Responses to electrical stimulation. Z. Tierpsychol. 63:173–200.
Weiss, M.J. (1974) Neuronal connections and the function of the corpora
pedunculata in the brain of the American cockroach Periplaneta
americana (L.). J. Morphol. 142:21–69.
Wigglesworth, V.B. (1957) The use of osmium in the fixation and staining of
tissue. Proc. R. Soc. B. 147:185–199.
Yamazaki, Y., M. Nishikawa, and M. Mizunami (1998) Three classes of
GABA-like immunoreactive neurons in the mushroom body of the
cockroach. Brain Res. 788:80–86.
Yang, M.Y., J.D. Armstrong, I. Vilinsky, N.J. Strausfeld, and K. Kaiser
(1995) Subdivision of the Drosophila mushroom bodies by enhancertrap expression patterns. Neuron 15:45–54.