J Comp Physiol A (2002) 188: 359–370 DOI 10.1007/s00359-002-0310-1
O R I GI N A L P A P E R
D. Mu¨ller Æ R. Abel Æ R. Brandt Æ M. Zo¨ckler R. Menzel
Differential parallel processing of olfactory information in the honeybee, Apis mellifera L. Accepted: 22 March 2002 / Published online: 30 April 2002 Springer-Verlag 2002
Abstract Two distinct neuronal pathways connect the first olfactory neuropil, the antennal lobe, with higher integration areas, such as the mushroom bodies, via antennal lobe projection neurons. Intracellular recordings were used to address the question whether neuroanatomical features affect odor-coding properties. We found that neurons in the median antennocerebral tract code odors by latency differences or specific inhibitory phases in combination with excitatory phases, have a more specific activity profile for different odors and convey the information with a delay. The neurons of the lateral antennocerebral tract code odors by spike rate differences, have a broader activity profile for different odors, and convey the information quickly. Thus, rather preliminary information about the olfactory stimulus first reaches the mushroom bodies and the lateral horn via neurons of the lateral antennocerebral tract and subsequently odor information becomes more specified by activities of neurons of the median antennocerebral tract. We conclude that this neuroanatomical feature is not related to the distinction between different odors, but rather reflects a dual coding of the same odor stimuli by two different neuronal strategies focusing different properties of the same stimulus. Keywords Olfaction Æ Spatial-temporal coding Æ Projection neurons Æ Intracellular recordings Æ Insect
D. Mu¨ller (&) Æ R. Abel Æ R. Brandt Æ R. Menzel Institut fu¨r Biologie-Neurobiologie, Freie Universita¨t Berlin, Ko¨nigin-Luise-Strasse 28–30, 14195 Berlin, Germany E-mail:
[email protected] Tel.: +49-30-83856284 Fax: +49-30-83855455 M. Zo¨ckler Konrad-Zuse-Institut fu¨r Informationstechnik, Freie Universita¨t Berlin, Takustrasse 7, 14195 Berlin, Germany
Abbreviations AL antennal lobe Æ KC Kenyon cell Æ l-ACT lateral antennocerebral tract Æ m-ACT median antennocerebral tract Æ LH lateral horn Æ MB mushroom body Æ OB olfactory bulb Æ PN projection neuron
Introduction Olfactory stimuli are coded by a spatio-temporal pattern of activity at the primary level of sensory integration, the olfactory bulb (OB) in vertebrates and the antennal lobe in insects (AL). Olfactory receptor axons originating from the same receptor type converge in one glomerulus of the olfactory bulb (Mombaerts 2001) and the antennal lobe as it has been shown via receptor neuron backfills (Hansson et al. 1992; Mustaparta 1996) and confirmed by membrane receptor gene expression (Vosshall et al. 2000). Evidence has accumulated that odors are coded in glomerular activity patterns (Galizia and Menzel 2000; Xu et al. 2000), indicating that multiple glomeruli encode particular chemical compounds as well as odor mixtures. A glomerular activity pattern may be translated into a cross-fiber-activity pattern in relay neurons (Kauer 1991; Hildebrand and Shepherd 1997) enriched by temporal components, e.g., synchrony effects of spike activity (Laurent 1996). In all studied insects, the inner antenno-cerebral tract (Kenyon 1896; synonym in honey bees: the median antennocerebral tract, m-ACT), connects the mushroom body (MB) and the lateral horn (LH) with the antennal lobe. Similarly, axons of the lateral ACT (l-ACT) innervate the lateral horn and mushroom body but in a reverse sequence: first the lateral horn and then the lip region of the mushroom body calyces. The l-ACT (also called the outer antenno-cerebral tract) has been described in various insect species, but the neurons of this tract may show different features in different species. At least in some species a sub-tract connects the antennal lobe with the mushroom body (Mobbs 1982; Homberg et al. 1988; Malun et al. 1993).
360
A third bundle of axons, called the mediolateral antennocerebral tract (ml-ACT) also innervate the lateral horn and larger parts of the lateral protocerebrum but not the calyces, as shown for several insect species (in most insects: middle antenno-cerebral tract, Homberg et al. 1988; Malun et al. 1993; Stocker 1994). In the bee the ml-ACT contains three subtracts, with the third subtract running parallel to the l-ACT (Abel et al. 2001). In the bee the anatomy of the antennocerebral tracts is rather simple. The projection neurons (PNs), which run in two major tracts, the m- and l-ACT, receive input only from single glomeruli in the antennal lobe (uniglomerular PNs, Abel et al. 2001), whereas the smaller number of projection neurons of the ml-ACT receive input from several glomeruli (Fonta et al. 1993). Thus, the mushroom body receives olfactory information only via uniglomerular projection neurons, whereas the lateral horn receives input from both uniglomerular and multiglomerular projection neurons. Considering the similar anatomical features of the m- and the l-ACT neurons one may ask: Why are there two olfactory tracts instead of one? Since the projection neurons of the m- and l-ACT originate in different glomeruli, one might expect different olfactory sensitivity profiles. In fact, the glomeruli served by the two groups of projection neurons are located in two subdivisions of the antennal lobe as defined by the innervation of receptor neuron axons running in four branches of the antennal nerve. Primary axons of branches 2, 3 and 4 (T2, T3, and T4) go to glomeruli served by dendrites of m-ACT neurons, and axons in branch 1 (T1) to glomeruli served by dendrites of l-ACT neurons (Abel et al. 2001). The anatomical separation in the two groups of projection neurons may, therefore, reflect a functional separation between groups of receptor inputs. This separation appears not to be related to the distinction between the general odor system and the sex pheromonal system. Since worker honeybees do not have a macroglomerular complex, comparison with the pheromonal system as described for other insects is difficult. However, the clear separation between the general odor system and the sex pheromonal system as seen at the level of the glomeruli is not seen in antennocerebral tracts (e.g., in the moth, Kanzaki et al. 1989), indicating that the separation of the antennocerebral tracts may not be related to the distinction between general odor and pheromone codes. Here we ask in what respect the response properties of the projection neurons differ in the m- and the l-ACTs. We find an overlap of olfactory sensitivity profiles, but different time courses of their responses to different odors. M- and l-ACT neurons appear to convey odor information to the second odor neuropils in a sequential manner, first general information about the quality of an odor related to e.g. timing or intensity of the stimulus, and then refined information about the particular kind of odor.
Materials and methods Preparation Worker bees (Apis mellifera carnica) were caught at the hive entrance or in an indoor flight room (Praagh 1972), immobilized by cooling, and mounted in plastic tubes. The bees were fed with sucrose solution and kept in the dark at 20C and high humidity. The following day the head was fixed with wax and opened between the median ocellus and the base of the antennae. Glands and tracheal sheaths were removed. A second hole was cut to expose the esophagus. Small droplets of bee physiological saline solution (in mmol l–1: 130 NaCl, 6 KCl, 2 MgCl2, 10 HEPES, 17 glucose, 6 fructose, 160 sucrose, pH 6.7) were applied (for more details see Mauelshagen 1993). Electrophysiology Glass electrodes were pulled with a horizontal puller (P97, Sutter instruments, Novarto, Calif.) and filled with 0.5 mol l–1 KCl or 0.5 mol l–1 K+-acetate depending on the stain used. The electrodes (resistance in the tissue ranged from 120 MW to 300 MW) were positioned at the top of the AL, and lowered posteriorly into the neuropil until a neuron could be penetrated using a micromanipulator (Leitz). A reference electrode, a chlorized silver wire, was inserted into the eye. The recordings were done in bridge mode using an intracellular BRAMP 1 amplifier (NPI Electronics, Tamm, Germany). Data were digitized on a 1401 interface and stored on a PC using spike2 2.3 software (Cambridge Electronic Design, UK). Stimulation The tip of an olfactometer (Galizia et al. 1997) was placed 3 cm from the antenna. To guarantee reproducible stimulation conditions, a computer programmed in Turbo Pascal 4.0 controlled the olfactometer. Odorant stimulation was given for 2 s; 8 s of pure air stimulation were provided between odor stimulations. The first sample of animals received six different odorants. Three different groups of olfactory stimuli were used to characterize neuronal response properties. Citral and geraniol represented pheromone components which are also known as components of different blossoms (Free 1987; Balderrama et al. 1996). 1-hexanol and 1-hexanal are blossom components which are not described as parts of pheromone blends. Natural plant and floral blends were simulated using orange oil and clove oil. The second sample of animals received these six odorants, and in addition, five others. The additional substances were pheromone components 1-octanol, 2-heptanone and isoamyl acetate (Balderrama et al. 1996) and the natural floral blends peppermint and lime tree. A portion of this sample received 2-hexanone and 2-octanone (blossom components which are similar to the pheromone component 2-heptanone) instead of peppermint and lime tree. Every odorant was given at least twice. The odor stimulation was performed in varying sequences to exclude sequence effects. Using these odors allowed us to also search for preferences to chemical properties such as chain length or functional groups, even with a relatively small number of substances. We used pure substances to analyze odor concentration on a level relevant to odorant behavior. It has been shown that in a positive classical appetitive conditioning the highest odorant concentration leads to the best learning and discrimination in honeybees (Pelz et al. 1997). Dye application The tips of the electrodes were filled with 2% Alexa 568 (hydrazide salt; Molecular Probes, Leiden, Netherlands) in 0.5 mol l–1 KCl or 4% tetramethylrhodamine dextrane (Molecular Probes) in
361 0.2 mol l–1 K+-acetate. Pulses of 2 Hz and 0.25 ms duration were applied. Hyperpolarization was used for Alexa and depolarization was used for rhodamine dextrane. Complete labeling of projection neurons required dye-injection for at least 15 min with a current between 2 nA and 5 nA. After intracellular filling, rhodamine dextrane was allowed to diffuse from 3 h up to overnight.
Analysis Traces were analyzed with Igor (Wave Matrix) and Statistica 4.4 (StatSoft). First the averaged background activity was determined by counting the spikes in all 2-s intervals before odorant stimulations for the respective neuron divided by the respective time. Responses were scored as excitatory if the spike frequency during the odor stimulation (number of spikes divided by 2 s) was higher than the averaged background activity plus the standard deviation, and as inhibitory if the spike frequency during odor stimulation was lower than the averaged background activities minus the standard deviation. The responses properties of a neuron to stimulations with different odors has been evaluated, taking into account its variability in response to repeated presentations of the same odor for the same time period and concentration. The variability in response to the same odor was calculated for each neuron by subtracting the average number of spikes or the averaged latency for the second stimulation with stimulus A (all odors to which a particular neuron was sensitive) from the averaged number of spikes or the averaged latency elicited by the same stimulus A for the first stimulation (A1–A2). Accordingly, differences in responses to other odors were obtained by the same procedure using spike frequencies or latencies obtained with stimuli A and B (A–B). Odor ‘‘B’’ indicates all odors to which a particular neuron was sensitive and which differed from A. The statistical procedure was based on the same data acquisition procedure. For each neuron only one value was obtained for the respective column. Thus a neuron had to respond to at least two odors in order to be used for statistical analysis. If a neuron did respond, e.g., to four odors, the two values were obtained as follows: the difference A1–A2 was calculated for each response, then four differences were added up and divided by 4. This result is the value of this neuron for the first column. The value for the other column A–B was obtained by subtracting the response to the first odor from the response to the next subsequent different odor until all possible various pairs (six) were obtained and then the sum obtained was divided by 6. To decide whether the neurons are able to discriminate between the different odorants according to their response strengths or latencies, the average variability in response to the same odorant (A1–A2) was compared to the average difference in response to two odorants (A–B). The comparison was carried out separately for mACT and l-ACT neurons. The time-course of responses were analyzed by counting the number of spikes for single bins (in our case 100-ms bins). The sequence of spike rates for the respective bins was correlated with a second sequence of spike rates evoked by the same odorant (KA1KA2) or with the sequence of spike rates elicited by a different odor (KA-KB). The comparison of the two correlation coefficients was used as a measure of odor specificity coded by the time-course of responses. The comparison is applied separately to m-ACT and l-ACT neurons. Response levels during stimulation were compared between two groups using the Mann-Whitney U-test (U values). The Wilcoxontest was used for within-group comparisons (Z values).
Histology The brains were dissected in bee physiological saline solution and fixed in 4% paraformaldehyde (for 1 h at room temperature) and dehydrated in an ascending alcohol series (50% and 70% for 5 min each and 90% and 99% for 10 min each, then twice in 100% for 10 min).
Confocal microscopy and identification of the recorded neurons: Whole-mount preparations of the brains were viewed with a laser-scanning confocal microscope (Leica, TCS4D). Serial optical sections were imaged at intervals of 1–2 lm through the depth of the tissue and saved as three-dimensional stacks. Where needed, the digitized images were modified only by contrast enhancement. The shapes of the innervated neuropils were viewed by autoflurescence or by staining via primary antibodies (nc46, Synorf1) against synapsin (Reichmuth et al. 1995; kindly provided by Dr. E. Buchner, Wu¨rzburg) and a secondary cy5-conjugated antibody (Molecular Probes, Leiden, the Netherlands). The image segmentation component of the Amira 2.2 software (Indeed, Houston, Tex.) was used to identify the innervated neuropil structures and create three-dimensional polygonal surface models. Because of variability in size and shape, which is enhanced by deformations due to preparation and histology, reconstructions had to be aligned so that homologous structures correlate. To calculate a correspondence between individual data sets, we applied a surface-based deformation algorithm to the neuropil structures (Zo¨ckler et al. 2000). Using thin-plate splines (Bookstein 1991), the deformation field of the surface model was applied to the reconstructed neurons. The PNs were identified as m-ACT neurons if the stained axons leave the AL at its ventromedian border to the protocerbrum at a depth between 170 lm and 190 lm, continue dorsally to the ipsilateral protocerebral neuropil and follows a path along the midline of the brain, bypasses the beta-lobe of the MB posteriorly, and bifurcates in the dorsal protocerebrum. l-ACT neurons leave the AL nearly in the same depth but, unlike m-ACT neurons, continue laterally into the dorsal and lateral part of the protocerebrum (for more details see: Abel et al. 2001). The innervated glomeruli were identified or classified as belonging to one of the four regions of the AL (Flanagan and Mercer 1989) by fitting the coordinates of the innervated region to the atlas (Galizia et al. 1999a).
Results Morphology Our results are based on 209 stained neurons of the AL, of which 52 uniglomerular projection neurons connect the AL to the MB. 37 of these neurons leave the AL via the m-ACT, 15 neurons via the l-ACT. The axons of the m-ACT neurons leave the AL and pass the mushroom body posteriorly, sending five collaterals via the inner ring tract to the lip region of the calyces and one collateral to the lateral horn of the protocerebrum. Axon terminals with a large number of bouton-like swellings indicate presynaptic sites in the calyces. A typical example of an m-ACT neuron is shown in Fig. 1a, which receives input from the single glomerulus T3–46. The 15 l-ACT neurons send their axons posteriolaterally through the protocerebrum and first send one collateral to the lateral horn and then five collaterals via the inner ring tract to the calyces. As in m-ACT neurons, the l-ACT neurons are also uniglomerular. The input glomeruli of 8 l-ACT neurons were traced to the T1 region of the AL. In 3 other cases the innervated glomeruli could not be unambiguously identified because they lie on the border between the T1 and T3 regions. The neuron shown in Fig. 1b receives the input from glomerulus T1–27. Axon terminals of this l-ACT
362
neuron contain a larger number of bouton-like swellings (approx. 500) in the calyces than the m-ACT neuron does (approx. 400 in Fig. 1a). However, compared to other stainings the total number of boutons is in the same range with several other m- and l-ACT neurons (8). Figure 1c shows a widely overlapping innervation pattern in the calyces of the MB for the m-ACT and the l-ACT neurons. Again we did not see any difference in the tangential projections of both types of neuron in the lip region of the calyces. But special cases in the innervation pattern can be assumed for dendritic fields of Kenyon cells, which are never more than 50 lm in diameter. In the first 50-lm field (on the right) no blebby structure can be observed; the second field (on the left) is dominated by the m-ACT structures, and the third field (on the bottom) is dominated by the l-ACT structures.
In particularly well-stained cases (9 out of 48 PN), we found projection neurons arborizing just outside the innervated glomerulus with fine branches originating from the axon (indicated by arrows in Fig. 1a and shown in detail in Fig. 1d). These ramifications with 2– 25 boutons occur within the antennal lobe in a region where all axons of projection neurons are located close together and where they form their respective output tracts just outside the glomerular layers. Odor responses Figure 2 shows that m-ACT and l-ACT neurons differed in their background activity (first column) and their odor responses (columns 2–5). The mean background activity of m-ACT neurons is 10.9 Hz [Hz=impulses s–1;
363 b Fig. 1a–d. 3-D reconstruction (Amira) of confocal micrographs showing intracellularly-labeled rhodamine dextran-stained projection neurons (PN). a The drawing of a median antennocerebral tract (m-ACT) neuron (red), a dorso-posterior view, shows the soma, the uniglomerular input from the glomerulus T3–46 in the posterior medial part of antennal lobe (AL; blue) and the entire blebby ramification in the lip region of the calyces (blue) of the mushroom body (MB) (light gray for the pedunculus) and the lateral horn of the protocerebrum (purple). Arrowheads indicate secondary branches with blebb-like structures in the posterior medial antennal lobe. b The drawing of a lateral antennocerebral tract (l-ACT) neuron (yellow), shown from the same perspective. The neuron, which is also uniglomerular, arises from a different origin, the smaller glomerulus T1–27 in the dorso-medial region. The targets in the lip region of the calyces and the lateral horn are reached from the other direction. c The left calyces of the half-brain in b were fitted to the same calyces. The deformation field of the surface model was applied to the reconstructed l-ACT neuron (yellow). A substantial overlap is indicated in the innervation pattern within the calyces with the m-ACT neuron (red). Dotted circles illustrate hypothetical dendritic influences of intrinsic Kenyon-cells for special cases. d A detailed drawing, reduced to the AL, of a second m-ACT neuron showing secondary branches and the innervation of a single glomerulus (T2–1) of the same neuron. The secondary branches cover a region in the posterior medial part outside the glomerular layer. In this region the median as well as the lateral projection neurons begin to form fiber tracts. (p posterior; a anterior; m median; l lateral)
standard deviation (SD)=6.9; n=13;) and is significantly higher than in l-ACT neurons 3.5 Hz (SD=4.3, n=11; U=18; P
