FOCUSED REVIEW

published: 15 September 2009

doi: 10.3389/neuro.01.028.2009

Reconstruction of virtual neural circuits

in an insect brain
Shigehiro Namiki 1,2, S. Shuichi Haupt 1, Tomoki Kazawa1,
Akira Takashima1, Hidetoshi Ikeno 3 and Ryohei Kanzaki 1*
1
Research Center for Advanced Science and Technology, The University of Tokyo, Tokyo, Japan
2
Graduate School of Life and Environmental Sciences, University of Tsukuba, Ibaraki, Japan
3
School of Human Science and Environment, University of Hyogo, Hyogo, Japan

The reconstruction of large-scale nervous systems represents a major scientific and engineering
Edited by: challenge in current neuroscience research that needs to be resolved in order to understand
Alexander Borst, Max Planck Institute
of Neurobiology, Germany
the emergent properties of such systems. We focus on insect nervous systems because they
Reviewed by: represent a good compromise between architectural simplicity and the ability to generate a
Uwe Homberg, Universität Marburg, rich behavioral repertoire. In insects, several sensory maps have been reconstructed so far.
Germany
Hiromu Tanimoto, Max Planck Institute We provide an overview over this work including our reconstruction of population activity in the
of Neurobiology, Germany primary olfactory network, the antennal lobe. Our reconstruction approach, that also provides
* Correspondence: functional connectivity data, will be refined and extended to allow the building of larger scale
neural circuits up to entire insect brains, from sensory input to motor output.
Keywords: digital atlas, identifiable neurons, neuron database, virtual brain

Ryohei Kanzaki is a Professor

INTRODUCTION nervous systems compared to mammals, yet rich
at the Research Center for Advanced The brain is the most important information behavioral repertoires. Here we describe basic
Science and Technology (RCAST), processing and control system in more highly techniques for the reconstruction of sensory
The University of Tokyo, Japan. developed animal organisms. Because of the maps and future directions toward the modeling
He received his Ph.D. degree in 1986, complexity of such systems, individual studies of entire behaviorally relevant neural circuits and
and was appointed Associate Professor
at the Institute of Biological Sciences,
have normally focused on specific brain regions. whole insect brains.
University of Tsukuba, Japan, in 1991, For instance, we have reconstructed the neural
following post-doctoral studies in the population activity in the antennal lobe of the INSECT NERVOUS SYSTEM
laboratory of John G. Hildebrand at the silkmoth in response to odorant stimuli (Namiki Experimental studies using mammalian brains are
University of Arizona, Tucson, USA.
and Kanzaki, 2008). An important point in this a predominant line of research in modern neuro-
He was recruited as a full Professor at the
Department of Mechano-Informatics, work has been that it is possible to integrate the science. However, numerous studies using simpler
Graduate School of Information Science data from different experiments. This makes it invertebrate systems have contributed much to
and Technology, The University of Tokyo possible to gradually build a database represent- the very foundations of neuroscience (Hawkins
in 2004, and then moved to RCAST ing a neural system to some extent. However, can et al., 1983; Hodgkin and Huxley, 1952; Reichardt,
in 2006. His research focuses on
we reasonably hope to understand how nervous 1957). Some of the key reasons for using insects
neuroscience using insect brains
as model systems. To reveal the design systems work without considering them as enti- in brain research are:
principles of insect brains and transfer ties that can only be dissociated with some com- Insect brains are convenient model systems to
this knowledge to engineering promises? Are not the non-neural constituents understand complex nervous systems in general.
applications, his laboratory is set up as of an animal’s body and the environmental fac- Although insects are unlikely to possess cognitive
a multidisciplinary environment,
including biology, informatics,
tors it is exposed to be included in our analyses? functions anywhere near those of mammals, they
engineering, agriculture, and chemistry. To work toward such an integrative approach, display rich behavioral repertoires comprising
[email protected] we have focused on insects, which have simple for example mating behavior, collision avoid-

Frontiers in Neuroscience www.frontiersin.org September 2009 | Volume 3 | Issue 2 | 206

Namiki et al. Building a vrtual insect brain

Reconstruction ance, goal-oriented behaviors, and sophisticated neurons can be merged across individuals (Bjaalie
In this review, we use the term spatial navigation capabilities. Insects are the et al., 2005). Other cases, in which a close cor-
“reconstruction” in a broad sense
for building a virtual counterpart
most successful animal group in evolution with respondence between individuals can even be
of a real biological system. There are the largest number of species. The adaptations carried further are the decapod stomatogas-
multiple levels for reconstruction, that have evolved to many ecological niches are tric nervous system and the nervous system of
including cellular, local network, and extremely valuable as specialized solutions and Caenorhabditis elegans. In both, the number of
whole brain levels. Different models are also likely to be reflected in neural hardware. neurons and their synaptic connectivity are nearly
require different experimental
parameters. To build a single cell
Nevertheless, the basic properties of neurons con- constant across individuals (Marder and Bucher,
model, for example, membrane stituting insect brains are similar to those of their 2007; White et al., 1986).
properties and the three-dimensional mammalian counterparts. Both show similar From these two examples, the importance of
structure of the neuron are often electrical properties (e.g., action potentials) and identifiable neurons in invertebrate systems is
required. To build a network model,
employ similar rules governing synaptic plasticity evident. Identifiable neurons are cells that can
the properties of synapses among
neurons are required. (Cassenaer and Laurent, 2007). At circuit level, be uniquely recognized in each individual of a
Identifiable neurons
extensive similarities in sensory systems have been species, allowing the accumulation of physi-
Neurons or groups of neurons which revealed (Hildebrand and Shepherd, 1997). Thus, ological, morphological, pharmacological, and
exist reproducibly among different the investigation of these relatively simple insect genetic information from different experiments.
individuals of the same species. They brains contributes to the understanding of nerv- The concept of identifiable neurons has given a
generally have the same physiological, ous systems in general. strong impetus to the research in insect brains in
anatomical, and genetic properties
so that we can compare and integrate
Many experiments in insect nervous sys- the past (Comer and Robertson, 2001) and gives
the experimental data from different tems are carried out in behaving preparations. such preparations an advantage over vertebrate
individuals. This feature is one The relation between experimental treatments systems, although some identifiable neurons are
of the major driving forces to focus and behavioral outputs has traditionally placed also known in fish, notably the pair of Mauthner
on neural information processing
insect neuroscience close to biologically moti- cells that are command neurons involved in
in an insect brain.
vated control engineering, i.e., cybernetics. The escape behavior (Eaton et al., 2001).
Neuron database
Database for neuroscience research.
key advantage is the possibility to infer mean-
Neuroscientific data are registered, ing though behavioral performance on a routine RECONSTRUCTION
shared, and analyzed systematically. basis. A hypothesis based on neurophysiologi- OF INSECT SENSORY NETWORK
The database facilitates data mining, cal results can easily be tested using behavioral Due to the availability of identifiable neurons and
communication among researchers experiments, simulations, and reimplementation identifiable structures and its small size, the analy-
and thus increases research efficiency.
We create and maintain the database
in robotics (Delcomyn, 2004; Lindemann et al., sis based on compilation of the physiological and
for neurons of silkmoth named 2008; Ritzmann and Büschges, 2007; Webb, 2002). morphological data has been common in a tra-
Bombyx Neuron Database (BoND), A factor that should also always be considered is ditional insect brain research. The success prob-
a part of which is already available the economic importance of some insect species ability of in vivo intracellular recording is low and
via the Internet.
in sectors such as agriculture and public health. experimenters are not always able to record from
Generally, only a limited amount of data can the neurons of interests. If all recorded neurons
be obtained from a single experiment for various are registered on the database, a comprehensive
reasons, that are technical in nature. For instance, analysis is possible by analyzing the data obtained
only a limited number of cells may be recorded from different experiments performed by different
simultaneously or the experimental paradigm, experimenters. This procedure can be facilitated
such as most anatomical and molecular investi- by data compilation, administration, and visuali-
gations, may only allow the observation of one zation operated on the same platform, utilizing
temporal snapshot. Under these conditions, neuroinformatics. For example, we have devel-
pooling data from different individuals will be oped a neuron database for male silkmoth brain
unavoidable. Whether this is reasonably possible and approximately 1,200 physiological and/or
depends critically on the amount of individual morphological data have been registered over the
variability in the structure of interest or at least last 10 years (Kazawa et al., 2008). The amount of
on the feasibility to exclude individual variability the contents is the largest among databases which
sufficiently from influencing the pooling process focus on a single animal species. The compiled
and thus the final outcome at the desired level data registered by 18 different experimenters have
of observation. In the mammalian neocortex, a achieved several important results. For example,
very good correspondence between individuals 153 antennal lobe local interneurons have been
can be observed in the barrel field of the pri- registered in the database and classified into five
mary somatosensory cortex of rodents, in which types (Seki and Kanzaki, 2008). This classification
a stereotyped pattern of granular zones (barrels), did correspond to the computational classifica-
each of them a single cortical column, represent tion by fractal dimension calculated using the box
defined whiskers on the snout. Using this map, counting method (Urata et al., 2006). Anatomical
physiological and morphological data of single identification and classification have also been

Frontiers in Neuroscience www.frontiersin.org September 2009 | Volume 3 | Issue 2 | 207

Namiki et al. Building a vrtual insect brain

Digital atlas done in neuronal populations in the mushroom belonging to the structure under investigation can
An anatomical map in a vector format. body, the center of learning and memory, and be employed to map data into a common geom-
The similarity of the nervous systems
among different individuals turns
the lateral accessory lobe, the premotor network etry for pooling and reconstruction of a sensory
such an atlas into an invaluable tool in the insect brain, based on 109 and 120 reg- representation.
facilitating data comparison and istered neurons, respectively (Fukushima and The antennal lobe, the first order olfactory
integration. Kanzaki, 2009). These studies made use of the processing center in insects, contains clearly ana-
data obtained by different experimenters. tomically segregated structures, called glomeruli.
In previous work, we have collected single-cell Each primary sensory neuron on the antennae
recording data obtained from different experi- expresses one of approximately 60 types of olfac-
ments under the same experimental conditions tory receptor proteins in the silkmoth. Notably,
(Namiki and Kanzaki, 2008). We have obtained sensory neurons which express the same types
recordings from 38 identified projection neu- of receptor proteins send axonal projections
rons and covered 32 of the total number of to the same glomeruli and make synapses with
60 ± 2 glomeruli. We have subsequently merged central neurons. As such, glomeruli are the func-
these data with reference to an anatomical map tional units of olfactory processing (Vosshall and
(Kazawa et al., 2009) and reconstructed stimulus- Stocker, 2007). Individual glomeruli have charac-
induced population activity. The reconstruction teristic size, shape, and relative positions and are
approach has been utilized to analyze several sen- identifiable across different individuals (Rospars,
sory representations in insects and is particularly 1983). Because of the constancy of glomerular
suitable when the structure of interest has clear organization, a digital atlas of the antennal lobe
anatomical landmarks. For example, the antennal can be made (Galizia et al., 1999a; Kazawa et al.,
lobe and the central body are composed of ana- 2009). Utilizing such anatomical maps, for exam-
tomically defined subunits facilitating reconstruc- ple, response profiles of sensory neurons have
tion (Heinze and Homberg, 2007; Namiki and been reconstructed on the digital atlas in the fly
Kanzaki, 2008). In contrast, the reconstruction (Hallem and Carlson, 2006).
of seemingly relatively unstructured neuropils We have performed a reconstruction of odor-
such as the lateral protocerebrum requires more evoked dynamics among projection neurons
involved image processing techniques (Jefferis (Namiki and Kanzaki, 2008), which are the sec-
et al., 2007; Tanaka et al., 2004; Troyer et al., ond order neurons in the insect olfactory system
1994). For example, in the study of a mechano- and functional equivalents of mitral/tufted cells in
sensory representation in the terminal abdomi- the mammalian brain. At first, we have examined
nal ganglion of crickets, the neuropil structures the morphological characteristics of this neuro-
of interest are not defined by a clear anatomical nal population and revealed that most of them
delineation and maps have to be generated based innervated to only a single glomerulus. This indi-
on landmarks not related to the structure itself. cates that by sampling this neuronal population,
This type of integration process for experimental we are able to measure the output information
data from different individuals to reconstruct a of the antennal lobe with single glomerular reso-
sensory map has first been applied to the cricket lution. Projection neurons innervating the same
cercal system. Crickets have a pair of append- glomeruli show similar olfactory response profiles
ages on the abdomen, the cerci, which bear sets as shown in other insect species (Reisenman et al.,
of filiform hairs with different preferences for 2005; Wilson et al., 2004). Due to this observation,
wind direction and frequency. Primary sensory the reconstruction approach, i.e., integrating data
neurons of these hairs send axonal projections from different individuals to assemble a virtual
into the terminal abdominal ganglion. Notably, neural circuit, is feasible in the moth antennal lobe.
sensory neurons with similar response profiles Utilizing the digital atlas of this system (Kazawa
have similar axonal projection patterns in the et al., 2009), we identified the glomeruli inner-
ganglion (Bacon and Murphey, 1984). The group vated by stained projection neurons. By mapping
digitized the neuronal morphology and pooled the dynamics of firing rate onto the glomeruli, we
the data following a linear transformation with reconstructed the olfactory representation based
reference to the outer surface of each ganglion. on a population of projection neurons (Figure 1).
Using these methods, it was possible to reveal the In the reconstructed map, odor-evoked firing rate
detailed topographic architecture of the system changes are represented as color in each glomer-
in which different types of information are proc- ulus. Temporal aspects of network dynamics in
essed in distinct subregions of the projection area the virtual projection neuron population can
of the filiform hairs’ mechanoreceptors (Jacobs be analyzed. For example, this population may
and Theunissen, 1996, 2000; Troyer et al., 1994). discriminate given olfactory stimulations at each
This case study has proven that landmarks not time (Figure 2). From the reconstructed map, it

Frontiers in Neuroscience www.frontiersin.org September 2009 | Volume 3 | Issue 2 | 208

Namiki et al. Building a vrtual insect brain

Figure 1 | Reconstruction of the population activity of the antennal lobe projection neurons. Data were obtained
by in vivo intracellular recording and staining with glass microelectrodes. Odor-evoked firing activity is monitored (A),
averaged across trials, and mean firing rate for each time window is calculated (B). Row and column mean time and
glomeruli innervated by the stained projection neurons, respectively. Colors in the figure represent mean firing rate change
for each projection neuron in response to cis-3-hexen-1-ol. Scale on the vertical axis represents the time period of odor
presentation (500 ms). Dye-filled neurons are imaged by confocal laser scanning microscopy (C), and the glomeruli
innervated by the stained neurons are identified (D). Pooled data are integrated with geometrical information using a digital
atlas (E). Reconstructed odor-evoked activity by a virtual projection neuron population in response to cis-3-hexan-1-ol
is shown. Upper and lower images show anterior and posterior views of the reconstructed dynamics. The number at the
bottom is the time elapsed after the onset of odor presentation. Five time points are shown. D, dorsal; L, lateral; M, medial.

Frontiers in Neuroscience www.frontiersin.org September 2009 | Volume 3 | Issue 2 | 209

Namiki et al. Building a vrtual insect brain

Figure 2 | Temporal evolution of olfactory representations. By applying dimensionality reduction techniques,

odor-evoked activities among projection neuron (PN) population are mapped into a 2D space (A). Illustration
of the trajectories representing PN population activity over time in response to cis-3-hexen-1-ol (green), linalool
(blue), and citral (red). For each odor, 47 PN spike sequences were divided into consecutive 50-ms time bins
and the number of spikes in each bin was counted. We viewed the 47-dimensional vector as the odor representation
by the PN population. To visualize the 47-dimensional vector, we applied principal component analysis to the vector
and used the first two principal components for the graph. Each representation evolved in an odor-specific manner.
(B) Each odor is tested three times. Euclidean distances are calculated for each odor pair. Gray lines represent
individual trials and the red line represents the average. Gray box shows odor presentation period (500 ms).
PC, principal component.

Frontiers in Neuroscience www.frontiersin.org September 2009 | Volume 3 | Issue 2 | 210

Namiki et al. Building a vrtual insect brain

Figure 3 | Spatio-temporal organization of the glomerular network. (A) Functional connectivity as a function
of anatomical distance of glomeruli innervated by a pair of projection neurons (PNs). We used Pearson correlation
coefficients (r) of the responses of pairs of PNs to quantify functional connectivity. Firing rates from stimulus onset
to 1 s after stimulus onset (20 time bins) were used. Pairs of PNs innervating different glomeruli show uncorrelated
activity (r = −0.0556, n = 496). (B) A graph of functional connectivity mapped onto x–z plane of the antennal lobe.
Each node represents a glomerulus. Only strong connectivity is shown, blue and red lines represent positive (>0.5)
and negative (less than −0.5) correlations, respectively. There are many functional connections among glomeruli
in the superficial part of the antennal lobe. Although sampling was somewhat sparse, there is a tendency that strong
functional connections are rare among deep glomeruli in response to the tested set of plant odors.

could be deduced that different odors elicited dis- tion neurons (Laurent, 2002) can apparently only
tinct dynamics and that the functional connectiv- be studied on a per-trial basis using multi-unit
ity among glomeruli was independent of physical recordings.
distance (Figure 3), suggesting distance independ- The variability may partially originate from
ent lateral inhibition among glomeruli as shown in the internal state of the network (Shu et al., 2003;
other species (Olsen et al., 2007; Reisenman et al., Stopfer and Laurent, 1999). In such case, variabil-
2009). As such, we are able to analyze not only ity may be decreased by monitoring physiologi-
a temporal aspect of sensory representation but cal parameters of the animal. For example, local
also the spatial organization with high resolution. field potentials may help for the experimenter to
Similar reconstruction approaches incorporating quantify the state of the network. A classification
temporal information have been applied in other of the responses based on the state of the animal
species as well (Krofczik et al., 2009; Staudacher may enable us to perform more precise recon-
et al., 2009). structions of neural circuits.
The reconstruction method has several dis-
advantages. First, the experimental cost is high TOWARD THE RECONSTRUCTION
in comparison with optical imaging methods OF WHOLE INSECT BRAIN
(Galizia et al., 1999b). Although signal-to-noise Our goal lies in between two approaches: From
ratio of individual recordings is very high, a the reconstruction of the sensory map in the
large number of experiments is required. This antennal lobe and the analysis of its connectivity,
could be improved by automation techniques in we intend to proceed to model circuitry in the
future (Dunlop et al., 2008). Second, even if we brain of the silkmoth that spans from the sensory
use the same experimental conditions, response input to the output of the brain. The ultimate
variability is present even among the same iden- goal will be the emulation of the complete silk-
tifiable neurons in different samples. Individual moth brain in software like blue brain project
variability leads to limitations in the temporal (Markram, 2006). The size of this nervous sys-
resolution at which meaningful information can tem is close to the size of a single neocortical
be extracted. From our results, the firing rate is column. The model can be evaluated by com-
comparable between different neurons of the parison of behavioral output with real insects by
same glomerulus at a bin size of approximately use of insect–robot hybrid systems (Emoto et al.,
50–100 ms. Faster phenomena such as an odor- 2007). The key advantages of the silkmoth for
evoked synchronization among different projec- this endeavor are access to transgenic and physi-

Frontiers in Neuroscience www.frontiersin.org September 2009 | Volume 3 | Issue 2 | 211

Namiki et al. Building a vrtual insect brain

Standard brain ological approaches and a well-defined behavior, been a long-standing difficulty in any modeling of
A digital atlas of all structures in the pheromone orientation, whose neural substrates neural circuits. We employ the insect nervous sys-
brain. There are several methods to
create a standard brain, they generally
have been studied in detail (Kanzaki et al., 2008). tem in which individual neurons are identifiable
minimize the individual difference This organism is suitable for building a large-scale and comparable from one individual to another. It
among samples. This is important model due to its good accessibility to the electro- therefore becomes possible to integrate data from
for estimating synaptic connectivity physiology of individual neurons and its nervous different experiments in a cumulative way with less
from neuronal morphology data. system is of the adequate scale to allow simulation conceptual problems. Capitalizing on this prop-
of the entire system in real time using currently erty, we will apply multiple techniques to probe
available supercomputer technology. In addition, synaptic connections including dual electrophysi-
neural circuits from sensory to motor neurons ological recording, optical synaptic probing and
via higher order centers in the brain have been transsynaptic tracers. Complementary to this,
investigated in detail by single-cell electrophysiol- connectivity for single-neuron data can be esti-
ogy. The reconstruction approach is well suited to mated through our mapped data base approach
primary sensory neuropils because the stimulus merging data in standard brain coordinates. We
onset provides a clear time reference, but it gradu- map the neurons into a standard brain (Brandt
ally looses temporal resolution in higher order et al., 2005; Kurylas et al., 2008) to use this unified
areas because response variability will increase. structure as a basis to infer potential synaptic con-
However, it is likely that reconstruction tech- tacts (Jacobs and Theunissen, 2000; Stepanyants
niques as outlined above can be applied to motor and Chklovskii, 2005; Tanaka et al., 2004). This
systems in which motor output can serve as time will allow the use of realistic multi-compartment
reference. Motor areas of the silkmoth have been biophysical modeling of a large number of neu-
intensively investigated (Mishima and Kanzaki, rons in the order of the actual number of neurons
1999; Wada and Kanzaki, 2005) and now we are in the silkmoth brain.
collecting single-cell recording data from the neu-
rons connecting sensory and motor areas. ACKNOWLEDGMENTS
For the precise reconstruction of the brain, This research was supported by a grant from a
determining synaptic connectivity is a key problem. Grant-in-Aid for Scientific Research (B) from the
Due to the tangled structure of neurons and the Ministry of Education, Culture, Sports, Science
huge number of contacts they make, this point has and Technology.

REFERENCES Dunlop, J., Bowlby, M., Peri, R.,Vasilyev, D., Hallem, E. A., and Carlson, J. R. (2006). Jefferis, G. S., Potter, C. J., Chan, A. M.,
Bacon, J. P., and Murphey, R. K. (1984). and Arias, R. (2008). High-throughput Coding of odors by a receptor reper- Mar in, E. C., Rohlfing , T.,
Receptive fields of cricket giant electrophysiology: an emerging para- toire. Cell 125, 143–160. Maurer, C. R. Jr., and Luo, L. (2007).
interneurons are related to their digm for ion-channel screening and Hawkins, R. D., Abrams, T. W., Carew, T. J., Comprehensive maps of Drosophila
dendritic structure. J. Physiol. 352, physiology. Nat. Rev. Drug Discov. 7, and Kandel, E. R. (1983). A cellular higher olfactory centers: spatially seg-
601–623. 358–368. mechanism of classical conditioning regated fruit and pheromone repre-
Bjaalie, J. G., Leergaard, T. B., Lillehaug, S., Eaton, R. C., Lee, R. K. K., and Foreman, M. B. in Aplysia: activity-dependent ampli- sentation. Cell 128, 1187–1203.
Odeh, F., Moene, I. A., Kjode, J. O., and (2001). The Mauthner cell and other fication of presynaptic facilitation. Kanzaki, R., Ando, N., Sakurai, T., and
Darin, D. (2005). Database and tools identified neurons of the brainstem Science 219, 400–405. Kazawa, T. (2008). Understanding
for analysis of topographic organi- escape network of fish. Prog. Neurobiol. Heinze, S., and Homberg, U. (2007). and reconstruction of the mobiligence
zation and map transformations in 63, 467–485. Maplike representation of celestial of insects employing multiscale bio-
major projection systems of the brain. Emoto, S., Ando, N., Takahashi, H., and E-vector orientations in the brain of logical approaches and robotics. Adv.
Neuroscience 136, 681–695. Kanzaki, R. (2007). Insect-controlled an insect. Science 315, 995–997. Robot. 22, 1605–1628.
Brandt, R., Rohlfing, T., Rybak, J., robot – evaluation of adaptation Hildebrand, J. G., and Shepherd, G. M. Kazawa, T., Ikeno, H., and Kanzaki, R.
Krofczik, S., Maye, A., Westerhoff, M., ability. J. Robot. Mechatron. 19, (1997). Mechanisms of olfactory dis- (2008). Development and applica-
Hege, H. C., and Menzel, R. (2005). 436–443. crimination: converging evidence for tion of a neuroinformatics envi-
Three-dimensional average-shape Fukushima, R., and Kanzaki, R. (2009). common principle across phyla. Annu. ronment for neuroscience and
atlas of the honeybee brain and its Modular subdivision of mushroom Rev. Neurosci. 20, 595–631. neuroethology. Neural Netw. 21,
applications. J. Comp. Neurol. 492, bodies by Kenyon cells in the silkmoth. Hodgkin, A. L., and Huxley, A. F. (1952). A 1047–1055.
1–19. J. Comp. Neurol. 513, 315–330. quantitative description of membrane Kazawa, T., Namiki, S., Fukushima, R.,
Cassenaer, S., and Laurent, G. (2007). Galizia, C. G., Mcllwrath, S. L., and current and its application to conduc- Terada, M., Soo, K., and Kanzaki, R.
Hebbian STDP in mushroom bod- Menzel, R. (1999a). A digital-three tion and excitation in nerve. J. Physiol. (2009). Constancy and variability of
ies facilitates the synchronous flow dimensional atlas of the honeybee 117, 500–544. glomerular organization in the anten-
of olfactory information in locusts. antennal lobe based on optical sections Jacobs, G. A., and Theunissen, F. E. (1996). nal lobe of the silkmoth. Cell Tissue
Nature 448, 709–713. acquired by confocal microscopy. Cell Functional organization of a neural Res. 336, 119–136.
Comer, C. M., and Robertson, R. M. (2001). Tissue Res. 295, 383–394. map in the cricket cercal sensory sys- Krofczik, S., Menzel, R., and Nawrot, M. P.
Identified nerve cells and insect behav- Galizia, C. G., Sachse, S., Rappert, A., and tem. J. Neurosci. 16, 769–784. (2009). Rapid odor processing in the
ior. Prog. Neurobiol. 63, 409–439. Menzel, R. (1999b). The glomerular Jacobs, G. A., and Theunissen, F. E. (2000). honeybee antennal lobe network.
Delcomyn, F. (2004). Insect walking and code for odor representation is species Extraction of sensory parameters from a Front. Comput. Neurosci. 2, 9.
robotics. Annu. Rev. Entomol. 49, specific in the honeybee Apis mellifera. neural map by primary sensory interneu- Kurylas, A. E., Rohlfing, T., Krofczik, S.,
51–70. Nat. Neurosci. 2, 473–478. rons. J. Neurosci. 20, 2934–2943. Jenett, A., and Homberg, U. (2008).

Frontiers in Neuroscience www.frontiersin.org September 2009 | Volume 3 | Issue 2 | 212

Namiki et al. Building a vrtual insect brain

Standardized atlas of the brain of the Reichardt, W. (1957). Autokorrelations- Staudacher, E. M., Huetteroth, W., triggered programmed behavior in male
desert locust, Schistocerca gregaria. Cell Auswertung als Funktionsprinzip des Schachtner, J., and Daly, K. C. (2009). A silkmoths revealed by double-labeling of
Tissue Res. 333, 125–145. Zentralnervensystems (bei der optischen 4-dimensional representation of anten- descending interneurons and a motor
Laurent, G. (2002). Olfactory network Wahrnehmung eines Insektes). Z. nal lobe output based on an ensemble neuron. J. Comp. Neurol. 484, 168–182.
dynamics and the coding of multidi- Naturforsch. B 12, 448–457. of characterized projection neurons. J. Webb, B. (2002). Robots in invertebrate
mensional signals. Nat. Rev. Neurosci. Reisenman, C. E., Chirstensen, T. A., Neurosci. Methods 180, 208–223. neuroscience. Nature 417, 359–363.
3, 884–895. and Hildebrand, J. G. (2005). Stepanyants, A., and Chklovskii, D. B. White, J. G., Southgate, E., Thomson, J. N.,
Lindemann, J. P., Weiss, H., Möller, R., and Chemosensory selectivity of output (2005). Neurogeometry and potential and Brenner, S. (1986). The structure
Egelhaaf, M. (2008). Saccadic flight neurons innervating an identified, synaptic connectivity. Trends Neurosci. of the nervous system of the nematode
strategy facilitates collision avoidance: sexually isomorphic olfactory glomer- 28, 387–394. Caenorhabditis elegans. Philos. Trans.
closed-loop performance of a cyberfly. ulus. J. Neurosci. 25, 8017–8026. Stopfer, M., and Laurent, G. (1999). Short- R. Soc. Lond., B, Biol. Sci. 314, 1–340.
Biol. Cybern. 98, 213–227. Reisenman, C. E., Heinbockel, T., and term memory in olfactory network Wilson, R. I., Turner, G. C., and Laurent, G.
Marder, E., and Bucher, D. (2007). Hildebrand J. G. (2009). Inhibitory dynamics. Nature 402, 664–668. (2004). Transformation of olfactory
Understanding circuit dynamics using interactions among olfactory glomeruli Tanaka, N. K., Awasaki, T., Shimada, T., and representations in the Drosophila
the stomatogastric nervous system of do not necessarily reflect spatial prox- Ito, K. (2004). Integration of chemo- antennal lobe. Science 303, 366–370.
lobsters and crabs. Annu. Rev. Physiol. imity. J. Neurophysiol. 100, 554–564. sensory pathways in the Drosophila
69, 291–316. Ritzmann, R. E., and Büschges, A. (2007). second-order olfactory centers. Curr. Conflict of Interest Statement: The authors
Markram, H. (2006). The blue brain project. Adaptive motor behavior in insects. Biol. 14, 449–457. declare that the research was conducted in
Nat. Rev. Neurosci. 7, 153–160. Curr. Opin. Neurobiol. 17, 629–636. Troyer, T. W., Levin, J. E., and Jacobs, G. A. the absence of any commercial or financial
Mishima, T., and Kanzaki, R. (1999). Rospars, J. P. (1983). Invariance and sex- (1994). Construction and analysis of relationships that could be construed as a
Physiological and morphological specific variations of the glomerular a database representing a neural map. potential conflict of interest.
characterization of olfactory descend- organization in the antennal lobes of Microsc. Res. Tech. 29, 329–343.
ing interneurons of the male silkworm a moth, Mamestra brassicae, and a but- Urata, H., Otsuka, A., Isokawa, T., Seki, Y., Received: 02 May 2009; paper pending pub-
moth, Bombyx mori. J. Comp. Physiol. terfly, Pieris brassicae. J. Comp. Neurol. Kamiura, N., Matsui, N., Ikeno, H., and lished: 28 June 2009; accepted: 13 July 2009;
A 184, 143–160. 220, 80–96. Kanzaki, R. (2006). On self-organizing published: 15 September 2009.
Namiki, S., and Kanzaki, R. (2008). Seki, Y., and Kanzaki, R. (2008). map based classification of insect neu- Citation: Front. Neurosci. (2009) 3, 2: 206-
Reconstructing the population activ- Comprehensive morphological iden- rons. In Proceedings of the IEEE Region 213. doi: 10.3389/neuro.01.028.2009
ity of olfactory output neurons that tification and GABA immunoreactiv- 10 Conference, TENCON 2006, IP 5.2, Copyright © 2009 Namiki, Haupt, Kazawa,
innervate identifiable processing units. ity of antennal lobe local interneurons TEN-561. Hong Kong, IEEE Takashima, Ikeno and Kanzaki. This is an
Front. Neural Circuits 2, 1. in Bombyx mori. J. Comp. Neurol. 506, Vosshall, L. B., and Stocker, R. F. (2007). open-access article subject to an exclusive
Olsen, S. R., Bhandawat, V., and 93–107. Molecular architecture of smell license agreement between the authors and
Wilson, R. I. (2007). Excitatory inter- S hu , Y. , Ha s e n s t a u b, A . , a n d and taste in Drosophila. Annu. Rev. the Frontiers Research Foundation, which
actions between olfactory processing McCormick, D. A. (2003). Turning Neurosci. 30, 505–533. permits unrestricted use, distribution, and
channels in the Drosophila antennal on and off recurrent balanced Wada, S., and Kanzaki, R. (2005). Neural reproduction in any medium, provided the
lobe. Neuron 54, 89–103. cortical activity. Nature 423:288–293. control mechanisms of the pheromone- original authors and source are credited.

Frontiers in Neuroscience www.frontiersin.org September 2009 | Volume 3 | Issue 2 | 213