The primary antennal sensory centers (antennal lobes) in the brain of the honeybee are highly compartmentalized into discrete spheres of synaptic neuropil called glomeruli, many of which can be identified according to their predictable size and location. Glomeruli undergo significant changes in volume during the lifetime of the adult worker bee, at least some of which are activity dependent. This study tests the commonly expressed assumption that increases in neuropil volume are accompanied by an underlying increase in the number of synapses present in the tissue. A combination of light and electron microscopy was used to determine total synapse number within two glomeruli, T1-44 and T4-2(1). The Cavalieri direct estimator of volume was applied to 1.5 μm sections of resin-embedded brains. Selected sections were then re-embedded and prepared for transmission electron microscopy. Synapse densities were determined using the physical disector method on electron micrographs. Synapse density and glomerulus volume were combined to give an unbiased estimate of the total number of synapses. In glomerulus T1-44, a significant increase in volume was accompanied by a significant increase in the total number of synapses. In contrast, synapse counts in T4-2(1) remained unchanged, despite a significant increase in the volume of this glomerulus. These results demonstrate that synapse proliferation in antennal lobes of the adult worker bee is highly site specific. Although volumetric changes and changes in synapse number both contribute to the structural plasticity of the antennal lobes, these two components of plasticity appear to be independent processes.
- Apis mellifera
- olfactory bulbs
- transmission electron microscopy
- structural plasticity
It is a commonly accepted assumption that an increase in neuropil volume reflects an increase in the number of synapses present (Devoogd et al., 1985; Turner and Greenough, 1985;Black et al., 1990; Withers et al., 1993; Klintsova et al., 2000), but, in many regions of the brain, this has yet to be clearly established. The antennal lobe of the adult worker honeybee provides an ideal model in which to test the validity of this assumption.
Honeybee antennal lobes receive sensory information from the antennae: multifunctional organs that house olfactory receptors, mechanoreceptors, and contact chemoreceptors, as well as receptors for detecting changes in temperature, carbon dioxide levels, and humidity (Esslen and Kaissling, 1976; Masson and Mustaparta, 1990). A large proportion of these sensory receptor neurons project to the antennal lobes of the brain, forming discrete subunits of synaptic neuropil, called glomeruli (Pareto, 1972) (see Fig.1). Many of the glomeruli can be identified (Arnold et al., 1985;Flanagan and Mercer, 1989a; Galizia et al., 1999a), and some have been shown to undergo age- and activity-related changes in volume (Winnington et al., 1996; Sigg et al., 1997). T1-44, for example, a large glomerulus located in the dorsal region of the antennal lobe, undergoes significant growth during the first week of adult life. Thereafter, the volume of this glomerulus remains relatively constant, despite significant changes in the behavioral activities of the bee. In contrast, glomerulus T4-2(1), which is located in the posterior region of the lobe, not only increases in volume significantly during the first week of adult life but exhibits additional increases in volume in bees of foraging age (Winnington et al., 1996). These two glomeruli are innervated by different tracts of primary sensory neurons and exhibit marked differences in their internal architecture (Suzuki, 1975; Arnold et al., 1985; Flanagan and Mercer, 1989a). Activities performed by bees have differential effects on the structural plasticity of these two glomeruli. Behavioral reversion studies show, for example, that the volume of T4-2(1) is smaller in foragers that revert to nursing duties than in bees that remain as foragers (Winnington et al., 1996), and induction of precocious foraging is accompanied by a premature increase in the volume of this glomerulus (Sigg et al., 1997). These particular forms of behavioral manipulation have no significant impact on the volume of glomerulus T1-44.
The aim of this study was to determine whether age- and activity-related changes in volume are accompanied by changes in the number of synapses within the glomerular neuropil of the antennal lobes of the honeybee. Here we combine stereology and transmission electron microscopy (TEM) to determine total synapse number in T1-44 and T4-2(1). In T1-44, an increase in glomerulus volume is correlated with an increase in synapse number, whereas in glomerulus T4-2(1), despite a significant increase in volume, the synapse number is unchanged.
MATERIALS AND METHODS
Worker bees were collected from a hive located at the Department of Zoology (University of Otago, Dunedin, New Zealand). Pupal bees at stage 9 of metamorphic adult development (P9) were identified according to accepted criteria (Jay, 1962) and were removed from brood cells immediately before dissection. Newly emerged adults were collected within 2 hr of emergence from their brood cells. To collect 4-d-old bees, a colored tag was glued to the thorax of newly emerged bees, which were then returned to the hive. Bees were considered to be 0 d old for the first 24 hr after emergence, and their age was measured in daily increments thereafter. Forager bees returning from pollen foraging trips were collected at the entrance of the hive. The age of the pollen foragers was not known; however, previous work on colonies maintained under similar conditions has shown that >85% of foragers are 3 weeks of age or older (Sigg et al., 1997). Bees from all age groups were derived from the same queen.
Brain tissue was initially fixed for 1 hr within the head capsule in 4% paraformaldehyde, 2% glutaraldehyde, and 0.04% CaCl2 in 0.225 m cacodylate buffer. Brains were then removed from the head capsule, placed in fresh fixative, and held overnight at 4°C. After primary fixation, samples were washed several times in 0.225 m cacodylate buffer and postfixed in 1% OsO4 solution in 0.225m cacodylate buffer for 1 hr at room temperature. The tissue was then rinsed through several buffer changes, followed by several distilled water washes. Samples were en bloc stained using 2% uranyl acetate solution in distilled water for 1 hr, rinsed, dehydrated through an ethanol series, and embedded in Agar 100 epoxy resin. Blocks were polymerized for 48 hr at 60°C. Semithin sections (1.5 μm) and ultrathin sections (see below) were cut on a Reichert Ultracut E ultramicrotome (Leica, Vienna, Austria). The 1.5 μm sections were collected on glass slides and stained with methylene blue. These sections were then used for estimating volume using the Cavalieri method. Six 1.5 μm sections from within the T4-2(1) glomerulus (Fig.1) were chosen in a random systematic manner from throughout the structure. These sections were re-embedded in Agar 100 resin and polymerized for 48 hr at 60°C. The block face was trimmed to include one entire antennal lobe in the ultrathin sections. Ultrathin sections (80 nm) were cut from these re-embedded blocks, with serial sections collected on Formvar-coated copper slot grids. Sections were post-stained with uranyl acetate and lead citrate in an LKB (Bromma, Sweden) Ultrostainer. The sections were then viewed and photographed in a Philips 410LS transmission electron microscope.
To ensure that synapses were identified consistently and reliably, both within and between animals, the following synapse identification criteria were established: (1) the presence of parallel opposed membranes, (2) evidence of presynaptic and postsynaptic specialization, and (3) the presence of at least three synaptic vesicles at the presynaptic terminal (Fig.2). All three criteria had to be fulfilled before a synapse was included for analysis. An individual synapse was defined as having a continuous presynaptic specialization with vesicles directly aligned at the presynaptic terminal, regardless of the number of associated postsynaptic processes. Synapses were not included for analysis if presynaptic and postsynaptic processes could not be distinguished from one another as a result of plane of section or orientation. The percentage of synapses not included for analysis was ∼5% for all animals. To examine synapse identification consistency over time, data from some animals were reanalyzed at a later date, and resulting synapse counts were compared. Synapse counts were within 10% of each other and therefore were considered consistent.
Establishing an appropriate stereological sampling regimen
Stereological methods were used in this study to ensure accurate and unbiased quantification of synapse numbers in antennal lobes of the brain. To establish an appropriate sampling regimen, however, it was necessary to obtain some indication of the size and distribution of synapses in the antennal lobes of the bee. Synapse size data were used to determine the distance between disector pairs, and distribution data were used to determine the intensity of sampling required. Synapse complexity was examined also to provide a preliminary assessment of potential changes in connectivity.
Synapse size and complexity. Synapse size was calculated from measurements of how many consecutive sections a synapse was observed to span; the number of postsynaptic processes associated with each presynaptic process provided a measure of synapse complexity. Because both of these features may be age- or activity-dependent, size and complexity of synapses was examined for each glomerulus in a newly emerged adult worker, as well as in a forager bee. Tissue was prepared for electron microscopy as described above. For each animal (n = 4 in total), up to 50 serial sections were collected onto Formvar-coated copper slot grids. The average section thickness, measured using Small's minimal folds technique (Geinisman et al., 1996), was 80 nm. Serial electron micrographs were taken on a Phillips 410LS transmission electron microscope operating at 80 kV. Images were collected onto 35 mm film at a magnification of 3200× and enlarged onto 12 × 16 inch photographic paper. For each glomerulus, two different areas were randomly sampled; one area included the outer margins of the glomerulus, and the second area sampled the glomerulus core. Only synapses entirely contained within the stack of electron micrographs were included for analysis. A total of 176–375 synapses was analyzed per glomerulus.
Synapse distribution. Analysis of total synapse number in structures in which synapse distribution is regional or heterogeneous requires a more intensive sampling regimen than sampling from a structure in which synapses are homogeneously distributed. For this reason, it was also important at the outset to obtain some idea of the distribution of synapses within T1-44 and T4-2(1). Tissue was prepared for electron microscopy as described above. The prepared blocks were initially trimmed to enable sections to be cut from the anterior margin of the glomerulus (referred to as anterior sections) and then retrimmed to allow sections to be cut from the middle of the glomerulus (referred to as core sections). Ultrathin sections were collected onto Formvar-coated copper slot grids. Sections were examined in a Philips CM100 transmission electron microscope. Images were collected onto plate film at a magnification of 15,000× and enlarged onto 8 × 10 inch photographic paper. A montage of electron micrographs was made from an anterior section and a core section, and the positions of all synapses were marked. Montages were scanned into the digital software package Adobe Photoshop (Adobe Systems, San Jose, CA) and converted into images showing synapse position and synapse density. A 25 μm2 grid was placed randomly over the glomerular profiles, and color coding was used to indicate how many synapses fell within each 25 μm2 area. The T1-44 glomerulus was prepared in this way for one forager bee. If there was a difference in synapse distribution between the cortical layer and core region of this glomerulus, it was thought that this difference would be apparent from one individual. For glomerulus T4-2(1), anterior and core sections were prepared from a 4-d-old worker.
Volume estimates. Volume estimates of the two glomeruli T1-44 and T4-2(1) were made using sections of the antennal lobe examined at the light microscope level. A set of random systematic sections through each glomerulus was stained, and the volume was determined using point counting methods with the Cavalieri estimator of volume (Gundersen et al., 1988). Images of frontal sections of the antennal lobe were projected from an Olympus Optical (Tokyo, Japan) BHS system microscope to a Panasonic (Secaucus, NJ) WV-CM140 video monitor using a Panasonic WV-CL500 video camera. An acetate sheet with a grid pattern of 20 mm2 was placed on the image at random angles, and the number of grid intersections falling on the structure of interest was counted. A random systematic set of sections at intervals of 4.5 μm was used for glomerular volumes and at 15 μm spacing for antennal-lobe volumes. A minimum total of 100 points was counted over at least 10 sections. The total volume was determined from the point counts using the following: Vol(object)= ∑P (object) · t ·a (p), where ∑P (object) is the total number of grid points counted for the glomerulus, t represents section separation, and a (p)is the area associated with each point (Gundersen et al., 1988).
Synapse density. A subset of six sections was selected in a random systematic manner from the entire section set for analysis using TEM. Within each selected section, the points used for TEM analysis were determined by placing a 15 mm2transparent grid over an image of the glomerulus. Selection of every sixth grid point, in a random systematic manner, ensured that at least 20 sample points were selected throughout the glomerulus. Synapse density was determined using the physical disector method on pairs of electron micrographs taken from consecutive thin sections at the selected sample points. Serial section analysis showed that some synapses were present on one section only; thus, adjacent section pairs were used to ensure that disector separation did not exclude small synapses from analysis. The thickness of ultrathin sections was based on an average thickness, determined by Small's method of minimal folds, as described by Geinisman et al. (1996). A counting frame 215 × 175 mm was placed over the micrographs to denote the disector area. The efficiency of the physical disector was maximized for each pair of sections by reversing the role of sections as reference section and look-up section. In this way, 20 sample points provided 40 disector pairs, with a total disector volume of 135 μm3. Density (Nv) was calculated using the following equation (Geinisman et al., 1996):Nv = ∑Q −/∑v(dis), where ∑Q − is the total number of synapses counted over all disectors, and ∑v(dis) is the total disector volume over all disectors.
Total synapse number. The total number of synapsesN(syn) in the total reference volume of antennal-lobe glomeruli V(ref) was calculated for each animal using the following formula (Geinisman et al., 1996): N(syn) =V(ref) · Nv.
The Student's t test was used to compare the reference volumes and to compare the total synapse numbers between age groups. One-way ANOVA was used when more than two age groups were being compared. When a significant difference between age groups was identified, post hoc Tukey's tests were used. Two-way ANOVA models were used to compare the reference volumes and synapse numbers, with age group (4-d-old or pollen forager) and glomerulus type as fixed factors. Confidence intervals were calculated separately for each glomerulus type for mean differences in volume and synapse number for the two age classes. One-way ANOVA was used in subsequent comparisons of total antennal-lobe volume and glomerular neuropil volume among four age groups for T4-2(1). When significant differences between age groups were identified, post hoc comparisons were made using Tukey's test. All analyses were performed using the Minitab software package, and a p value of <0.05 was considered to be statistically significant. Coefficients of error for each level of sampling were used to estimate the sampling precision, as per Geinisman et al. (1996). The contribution of the variation associated with the stereological sampling to the overall group variation was determined from the total intra-animal coefficient of error divided by the coefficient of variation of N(syn), as per Geinisman et al. (1996).
Synapse size and complexity
A range of both synapse size (in terms of how many consecutive sections a synapse was observed to span) and complexity (in terms of how many postsynaptic processes were associated with a single presynaptic process) was observed. Because some synapses were observed to be present only in a single 80 nm section, consecutive adjacent sections were used as disector pairs to ensure that all synapses had an equal chance of being included in the analysis. Not all synapses were so small, and some synapses were observed to span seven or more consecutive serial sections; however, our results suggest that, in both newly emerged bees and foragers, synapses of this size are rare (Fig.3 A,B). There was also considerable diversity observed in the postsynaptic complexity of individual synapses (Fig. 2 A–C), ranging from a single postsynaptic process (Fig. 2 A) to a maximum (in one observed instance) of eight postsynaptic processes observed aligned with a single presynaptic process (Fig.2 C). A majority of synapses, however, had either two or three postsynaptic processes (Fig. 3 C,D). In both T1-44 and in T4-2(1), the number of postsynaptic processes in newly emerged and forager bees was similar (Fig. 3 C,D). We could find no preliminary evidence of changes in postsynaptic process number.
Cobalt backfills of primary sensory afferent neurons reveal a distinct cortical layer of sensory input in glomerulus T1-44 (Fig.1 C), and a higher density of synapses within the cortical layer than in the central region of T1 glomeruli has been reported previously (Gascuel and Masson, 1991). To ensure that an appropriate sampling protocol was adopted in the current study, it was considered essential to first determine whether synapses were concentrated in specific areas within individual glomeruli.
The density of synapses in an anterior section of the T1-44 glomerulus, containing predominantly the cortical layer, was compared with the density of synapses in a section from the glomerular core, likely to contain both cortical and inner core regions. The number of synapses per unit area (25 μm2) was based on synapse distribution data obtained from electron micrograph montages (Fig. 4). Areas of relatively high synapse density (6–20 synapses per 25 μm2) represented 70% of the total area in the anterior section in the T1-44 glomerulus (Fig.4 A) and 75% of the total area in the section from the glomerular core (Fig. 4 B). Areas of high synapse density were distributed across the entire glomerular profile, in both the so-called cortical layer of the glomerulus and the core region of the glomerulus. No concentration of synapses could be identified within the cortical layer of glomerulus T1-44. Synapse distribution was examined also in glomerulus T4-2(1), which does not exhibit an obvious cortical layer of primary sensory afferent terminals (Fig.1 D). Comparison of an anterior section (Fig.4 C) with a section from the glomerular core (Fig.4 D) revealed that, at both depths of section, pockets of high synaptic density (11–15 and 16–20 synapses per 25 μm2) were dispersed among areas of low synaptic density.
An intensive sampling regimen, as used in this study, ensures that areas of both high and low synapse concentration will have an equal chance of being sampled, allowing an accurate estimation of total synapse number to be determined.
Neuropil volume and total synapse number
To compare two groups of worker bees of different age and behavioral status, 4-d-old workers and pollen foragers were used. Coefficients of error were calculated as an indication of the precision of stereological estimates, at each level of sampling. In all groups analyzed in this study, the majority of the observed error was attributable to biological rather than sampling variation (Table1).
Volumetric analysis showed no significant difference in the volume of the glomerular neuropil or central neuropil of the antennal lobe between 4-d-old bees and foragers (Fig.5 A,B), nor was any significant difference in the mean total antennal-lobe volume found between these two groups. However, the percentage of the antennal lobe occupied by glomerulus T1-44 and by glomerulus T4-2(1) was significantly higher in foragers than in 4-d-old workers (Fig.5 C,D).
The relationships between glomerulus volume and total synapse number in T1-44 and T4-2(1) are shown in Figures 6 and 7, respectively. The mean volumes of both glomeruli were significantly larger in forager bees than in 4-d-old bees (Figs. 6, 7, white bars). Total synapse number in T1-44 (estimated from the same animals as the volume estimates) was seen to increase significantly in foragers [mean ± SEM N(syn) of 394,349 ± 20,965] compared with 4-d-old workers [mean N(syn) of 204,161 ± 12,890], but a similar effect was not observed for the T4-2(1) glomerulus (Figs.6, 7, black squares). Mean total synapse number is estimated to increase by ∼115,000–265,000 (95% confidence interval) in the T1-44 glomerulus. In contrast, a comparison of foragers (meanN(syn) of 441,101 ± 26,091] and 4-d-old workers [mean N(syn) of 410,243 ± 39,235] shows little difference in synapse number for the T4-2(1) glomerulus. This difference is significantly smaller than that observed in the T1-44 glomerulus (F (1,27) = 5.54;p = 0.026). A 95% confidence interval for the mean difference in synapse number in the T4-2(1) glomerulus for foragers compared with 4-d-old bees (−41,600,103,136) suggests that mean increase in synapse number is likely to be less than ∼100,000.
Because the estimated total synapse number in the T4-2(1) glomerulus of 4-d-old bees (∼400,000) did not differ significantly from that of foragers (Fig. 7), we decided to examine the possibility that the number of synapses in T4-2(1) is established before emergence of the adult bee. For this reason, antennal-lobe volume measurements and synapse counts in the T4-2(1) glomerulus were examined also in newly emerged adults and in bees at pupal stage 9 (P9) of the nine stages of metamorphic adult development (Jay, 1962). The volumes of the central neuropil and glomerular neuropil of the antennal lobe were similar in these two age groups (Fig.8 A,B), and no significant differences could be identified between the two groups in total antennal-lobe volume (Fig. 8 C) or in the proportion of the antennal lobe occupied by glomerulus T4-2(1) (Fig.8 D). Volume estimates of the T4-2(1) glomerulus were also similar in P9 and newly emerged bees, and there was no significant difference between the total synapse numbers recorded in these two groups (Fig. 9).
ANOVA indicated that there were significant differences among the four age groups in which the T4-2(1) glomerulus was examined in the total antennal-lobe volume (F = 12.76; p < 0.001), the glomerular neuropil volume (F = 17.49;p < 0.001), and the proportion of the antennal lobe occupied by the T4-2(1) glomerulus (F = 45.29;p < 0.001). Whereas post hoc tests revealed that antennal-lobe and glomerulus volumes were significantly larger in P9 bees than in 4-d-old workers and pollen foragers, the proportion of the antennal lobe occupied by the T4-2(1) glomerulus was found to be significantly greater in 4-d-old bees and in foragers than in P9 bees. ANOVA provided weak evidence of differences in synapse number among the four age classes (F = 2.71; p = 0.065). However, post hoc tests showed differences between P9 and pollen foragers only. This study provides no evidence of increases in synapse number in the three adult age classes analyzed for this glomerulus. The mean total synapse number in newly emerged bees was 396,103 ± 29,615.
The rigorous stereological methods used in this study ensure accurate and unbiased quantification of structural components of the brain, which allows meaningful comparisons to be made not only between groups within a study but also between the results of the present investigation and those of similar studies on other species.
We showed that increases in T1-44 volume are accompanied by a dramatic increase in the number of synapses in this glomerulus. It is significant, however, that this is not the case for all glomeruli in the antennal lobe of the bee; T4-2(1) increases in volume also during the lifetime of the adult worker bee, but the total number of synapses in this glomerulus shows no significant change. Clearly, it cannot be assumed that increases in neuropil volume will be accompanied by increasing numbers of synapses. In the antennal lobes of the adult worker, these two components of structural plasticity appear to involve independent processes, and synapse proliferation, like volumetric changes (Winnington et al., 1996; Sigg et al., 1997), can be temporally restricted to certain subunits of antennal-lobe neuropil.
Site-specific structural plasticity
The striking differences in the structural plasticity of T1-44 and T4-2(1) suggest that these two glomeruli perform different roles in the antennal lobes of the brain. This is consistent with reports showing that antennal-lobe neurons in the bee respond not only to chemosensory signals (Flanagan and Mercer, 1989b; Sun et al., 1993; Abel et al., 2001) and to changes in humidity (Itoh et al., 1991) but also to mechanical stimulation of the antennae (Flanagan and Mercer, 1989b). The multimodality of the antennal lobes is reflected also in the response characteristics of antennal-lobe projection (output) neurons (Homberg, 1984). Calcium imaging studies provide strong evidence that glomeruli in the dorsal (T1) region of the antennal lobe process olfactory signals (Joerges et al., 1997; Galizia et al., 1999b), but glomeruli in the posterior (T4) region of the lobe, such as glomerulus T4-2(1), are not readily accessible to this form of analysis. Electrophysiological recordings from neurons with processes in T4 glomeruli are also difficult to obtain, but, in a recent study in which three T4 neurons were examined, all three failed to respond to any tested odor (D. Müller, personal communication). Two projection neurons with dendritic arbors in glomeruli in the T4 region of the antennal lobe have been described recently by Abel et al. (2001). These neurons showed mixed excitatory and inhibitory responses to stimulation with orange odor; however, bilateral neurons with arbors in one or two glomeruli in the T4 region of each lobe failed to respond to stimulation with puffs of air, mechanical stimulation, or to any of six odors tested, but did respond to touching the antenna with beeswax (Abel et al., 2001). Although preliminary, this information, combined with the close proximity of the T4 glomeruli to the antennal motor and mechanosensory center of the brain (Pareto, 1972; Suzuki, 1975; Mobbs, 1982; Kloppenburg, 1995) and interconnections with it (Flanagan and Mercer, 1989b), suggest that T4 glomeruli may process sensory signals other than, and perhaps in addition to, olfactory cues. That T1-44 and T4-2(1) are innervated by different tracts of antennal sensory afferent neurons and exhibit marked differences in internal architecture is also consistent with these two glomeruli playing significantly different roles in the antennal lobes.
If T1-44 and T4-2(1) process different types of sensory information, patterns of activity within the two glomeruli are likely to differ significantly over the lifetime of the bee. Worker bees change their behavior with age and use their antennae in different ways. Comb builders, for example, use tactile receptors at the tips of their antennae to measure the thickness of cell walls, foragers rely on antennal olfactory receptors to detect floral odors used in flower recognition, and brood pheromones detected by nurse bees induce them to feed the developing larvae (Winston, 1987; Robinson, 1992). Changes in antennal usage will alter patterns of activity within the antennal lobes, and these events appear to be reflected in the structural plasticity of the lobes.
Consistent with the idea that changes in glomerulus structure are activity rather than age dependent, forager bees used in the present investigation, which were selected on the basis of their behavior (pollen foraging) rather than age, showed remarkably little variation in glomerulus size and synapse counts. In the case of T4-2(1), for example, the level of variation in total synapse counts was smaller among pollen foragers than in same-age (4-d-old) bees.
Volume changes do not predict changes in synapse number
The lack of correlation between changes in volume and total synapse number for glomerulus T4-2(1) does not appear to be an artifact of sampling design; the error estimated to be attributable to the sampling method for T4-2(1) was similar to that associated with the sampling of T1-44, and the major contributor to the total sampling error was identified as coming from biological variation rather than from the sampling protocol (Table 1). In contrast to the situation in T1-44, most synaptic connections in T4-2(1) appear to be established before adult emergence. One possible explanation is that T4-2(1) processes information essential to the survival of newly emerged adults. The detection of temperature, CO2 levels, or humidity are processes that would fall into this category. Alternatively, T4-2(1) may be essential to the performance of behaviors required from the moment of adult emergence, such as social interactions within the colony that require processing of tactile and chemosensory cues. Whereas T4-2(1) appears to be “hard-wired” in terms of synapse numbers, functional plasticity in this glomerulus may be achieved in ways other than by increasing synapse counts. Changing the size or shape of existing synapses (Bailey and Chen, 1983), or remodeling of silent synapses into active ones (Geinisman et al., 2000), may contribute to the functional plasticity of this glomerulus. Changes in the numbers of postsynaptic processes associated with each presynaptic element could also provide plasticity within the circuitry of T4-2(1). Preliminary results suggest, however, that changes in the size and complexity of synapses in T4-2(1) may not explain the activity-dependent changes in volume apparent in this glomerulus (Fig.3 A–D).
In both T1-44 and T4-2(1), areas of high and low synapse density were distributed throughout the glomerulus neuropil. We could find no evidence in support of a previous report (Gascuel and Masson, 1991) suggesting that the density of synapses is higher in the outer (cortical) layer of T1 glomeruli than in the glomerulus core. However, Gascuel and Masson did not focus specifically on glomerulus T1-44, and it is possible that different T1 glomeruli display different patterns of synapse distribution. Details of the sampling protocol used are unclear, but it appears that Gascuel and Masson did not use random systematic sampling within the entire profile of a glomerulus to ensure that all areas of the glomerulus had an equal chance of being included in the sample. The patchy distribution of synapses evident in glomerulus T1-44 and T4-2(1) (Fig. 4 A–D) has significance for sampling design because an entire cluster (or alternatively very few synapses) could be contained within a single sampling frame, and, by chance, synapse number could be either overestimated or underestimated. Therefore, if only a small number of random samples are taken from any one section, the sampling frame may have by chance only fallen on areas of high or low density, thus giving a false impression of synapse distribution. The use of systematic sampling at a relatively high frequency throughout the entire profile as in the present study ensures that all synapses within a glomerulus have a greater chance of being included for analysis, regardless of a patchy density distribution within the tissue.
Causes and consequences of structural plasticity
Evidence suggests that, in the adult worker bee, increases in brain volume are attributable to the growth and elaboration of existing cells (Devaud and Masson, 1999; Farris et al., 2001) rather than to neurogenesis (Fahrbach et al., 1995; Malun, 1998). In an attempt to isolate the principal driving force(s) of structural plasticity from combined influences of age, behavior, and hormonal control, Sigg et al. (1997) found that structural changes in the antennal lobes of the bee coincide not only with shifts in the tasks being performed but also with behavioral changes associated with olfactory learning and memory. Changes in synapse number at the level of the antennal lobes may reflect storage of olfactory memories. This would be consistent with observations in other species. In the marine mollusc Aplysia californica, for example, long-term memory has been shown to be accompanied by changes in both the number and size of synapses (Bailey and Kandel, 1993). Increases in synapse number have also been associated with learning and memory in the vertebrate brain (Kolb and Whishaw,1998; Klintsova and Greenough, 1999). Mushroom bodies of the insect brain, which have been strongly implicated in olfactory learning and memory (Erber et al., 1987; Davis, 1993; Strausfeld et al., 1995;Menzel and Müller, 1996; Heisenberg, 1998), also exhibit age- and experience-related structural changes (Withers et al., 1993; Durst et al., 1994; Gronenberg et al., 1996; Fahrbach et al., 1998; Gronenberg and Liebig, 1999; Farris et al., 2001), but the relationship between total number of synapses and memory formation in these structures has yet to be determined.
In the somatosensory cortex of the vertebrate brain, use-dependent structural changes (Buonomano and Merzenich, 1998) are accompanied by changes in sensory perception (Sterr et al., 1998a,b). Recent studies suggest that the same may be true in insects. Using theDrosophila mutant gigas, which establishes more synapses than normal (Ferrus and Garcia-Bellido, 1976), increased synapse numbers in antennal-lobe glomeruli have been correlated with enhanced sensitivity to odorants (Acebes and Ferrus, 2001). Sensory experience has been shown also to have a significant impact on volumes and synapse counts in antennal-lobe glomeruli of the fly (Devaud et al., 2001). However, the extent to which changes in synapse number contribute to volume changes in the fly has yet to be determined.
The results of the present investigation show that increases in neuropil volume do not provide a reliable indicator of changes in the total numbers of synapses. Whereas this study has shown an increase in total synapse number in glomerulus T1-44, it is not known whether these changes occur between sensory neurons and postsynaptic antennal-lobe neurons or between local interneurons, projection neurons, and/or centrifugal neurons. It is now for future studies to determine the identity of specific neuronal populations involved in changes in synaptic connectivity.
This work was supported by University of Otago Research Grants B43 and B15. S.M.B. was supported by a Fanny Evans Scholarship. We thank Allan Mitchell and coworkers from the Center for Electron Microscopy at Otago University for technical advice and assistance during the course of this study. Thanks also to Danny Flanagan for providing the cobalt backfills (Fig 1 C,D) and to Robbie McPhee, Richard Easingwood, and Ken Miller for assistance with the illustrations.
Correspondence should be addressed to Alison Mercer, Department of Zoology, University of Otago, P.O. Box 56, Dunedin, 9001 New Zealand. E-mail:.