We have quantitatively analyzed the three-dimensional dendritic structure of an identified nonspiking interneuron, Local Directionally Selective (LDS) interneuron in the terminal abdominal ganglion of the crayfish, Procambarus clarkii Girard, using a confocal laser scanning microscopic system. Comparison of the cell size before and after fixation followed by dehydration and clearing revealed that the diameter of dendritic branches shrank by about 15% while the branch length showed no significant change. The soma and dendritic diameter in vivo were obtained from the measurement in the cleared preparation multiplied by the shrinkage factor.
The total length of dendrites and their membrane area as well as their electrotonic structure were found to vary from preparation to preparation. The spatial distribution of fine dendritic branches was variable, but that of major branches was constant among preparations. The interneuron further shared bilaterally asymmetrical dendritic structure: those branches on the soma side commonly showed larger membrane area and larger number of branching than those on the opposite side. The results indicate that the functional difference in the synaptic activity so far reported between dendrites on the soma side and those on the opposite side is paralleled by their differences in the dendritic morphology and electrotonic structure.
Many nerve cells in the central nervous system of invertebrates are individually identifiable according to their physiological and morphological characteristics (Hoyle, 1975). Nonspiking interneurons, however, which have been reported to be present in the central nervous system of a variety of invertebrates (e.g. Burrows, 1976; Heitler et al., 1980; Pearson et al., 1975; Spencer, 1988; Takahata et al., 1981), are not identifiable in general; cells with a similar pattern of dendritic branching and soma location form a cluster so that they can not be recognized individually (Goodman, 1978; Hisada et al., 1984). One exception is the LDS (Local Directionally Selective) interneuron which is located in the terminal abdominal ganglion of the crayfish Procambarus clarkii and involved in the tailfan mechanosensory system (Krenz et al., 1985; Reichert et al., 1982; Reichert et al., 1983). It has been reported that this cell can be individually identified by their structure and physiological properties including voltage-dependent inward and outward rectification (Takahashi et al., 1995; Takahata et al., 1995).
The LDS interneuron in every animal shares a common basic structure, having a single transverse segment crossing the midline and 2 major branches directly bifurcating from the transverse segment on both side. The interneuron can be repeatedly recognized from animal to animal according to this common basic structure, but, we reported in the previous papers that it showed a certain degree of morphological variability (Takahashi et al, 1995; also see Reichert et al, 1983). It has been long known in the central nervous system of invertebrates that the identified cells show morphological variability regarding their axonal and dendritic branching patterns (Burrows, 1973; Goodman, 1978; Macagno et al, 1973; O'Shea et al, 1974; Pearson et al, 1979; Pearson et al, 1980; Stretton et al, 1973; Treistman et al, 1976; Tweedle et al, 1973). However, the structural analyses in these cells including the LDS interneuron have so far been based mostly on the comparison of camera lucida drawings of the projected image of these ceils and therefore remained qualitative.
The dendrites of central neurons function in not only collecting synaptic inputs but also integrating them electrotonically based on their membrane properties and three-dimensional structure (Borst et al, 1994; Rall et al, 1992; Shepherd et al, 1990). Since the output of the LDS interneuron is directly controlled by graded synaptic potentials instead of action potentials (Reichert et al, 1983), its function depends on how the shape of synaptic potentials is transformed in the course of spread from the input to output synaptic regions. Quantitative analyses of the morphological variability in the dendritic structure of the LDS interneuron is therefore crucial for understanding its functional roles in mechanosensory information processing. In the present study, we quantitatively investigated the three-dimensional structure of the LDS interneuron by a confocal laser scanning microscopic system, which has been demonstrated to enable three-dimensional cell morphometry (Carlson et al, 1989; Wallén et al, 1988).
MATERIALS AND METHODS
Animals and preparation
Adult crayfish Procambarus clarkii Girard of both sexes, ranging from 8 to 10 cm in body length, were obtained commercially from a local dealer (Sankyo Labo, Sapporo). They were kept in laboratory tanks with running water until used in the experiment. The terminal (6th) abdominal ganglion was isolated from the rest of the body together with its ganglionic nerve roots. The ganglion was pinned down with the dorsal side up in a silicone-lined chamber which was filled with crayfish saline (Van Harreveld, 1936) at room temperature (20—25°C). The ganglionic sheath was left intact.
Electrophysiologyand cell identification
Glass microelectrodes were filled with 3% Lucifer yellow dissolved in 1M LiCI. The electrode resistance was about 35MΩ ranging 30–40 MΩ. The microelectrode was connected to a high input-impedance amplifier (Nihon-Kohden, CEZ-3100) whose output was sent to a storage oscilloscope (Tektronix 5113), a magnetic data recorder (TEAC, XR-50H) and a personal computer (NEC, PC-9801RX). Pin electrodes were placed in contact with the second and third ganglionic roots on both sides for their extracellular stimulation. They were coupled with an electronic stimulator through isolators. The impaled cell was first identified to be the LDS interneuron on the basis of physiological criteria (Takahata et al, 1995) and finally confirmed by its morphology revealed by intracellular injection of a fluorescent dye Lucifer yellow CH (Stewart, 1978). For later analyses of the electrotonic structure of the cell, hyperpolarizing step current pulses were injected intracellularly through the recording microelectrode using the discrete current clamp method (Wilson et al, 1975). The sampling frequency in the present study was 1.0–2.0 kHz. Transient responses of the cell to step current injection were used to extract the membrane time constant (Takahashi et al, 1995). The interneuron was always impaled in the thick transverse segment on or near the midline. Methodological details on the physiological experiments are provided elsewhere (Takahata et al, 1995).
After the physiological experiment was finished, the dye was iontophoretically injected into the cell by applying hyperpolarizing current pulses (5 –10 nA, 500 msec in duration, 1 Hz in frequency) for 15–30 min. The site of electrode impalement into the cell was monitored in situ under the dissecting microscope by the epiillumination method (Takahashi et al, 1995). The preparation was then fixed in 10 % formaldehyde for 6 minutes, dehydrated in an ascending series of ethanol (each 2 minutes in 70, 95, 100%) and cleared in methyl salicylate. In some experiments, the preparation was examined before fixation under a confocal laser scanning microscope for measuring the diameter and length of dendritic branches in vivo. The results were compared with those obtained from the same preparation after fixation to estimate the shrinkage factor due to the clearing process.
In the morphological analysis 13 LDS interneurons were successfully stained. Each cell was optically sectioned using a confocal laser scanning microscope system (Sarastro 2000, Molecular Dynamics) including an Argon-ion laser, a Nikon Optiphoto II epifluorescence microscope and a Silicon Graphics Iris Indigo XS24 workstation. The preparation was always set with the dorsal side up on the microscope stage so that the X- and Y-axis in the three-dimensional coordinate system correspond respectively to the transverse and longitudinal axis of the animal body. The laser line of 457 nm in wavelength was selected from the Argon laser generator (25 mW power) having 3 lines (457, 488, 514 nm). The laser beam was focused to a point by a microscope lens and scanned the specimen with the aid of a two-mirror system driven by a galvanometric (Y-axis) and a stepping motor (X-axis). The fluorescent light from the specimen passed back through the microscope objective lens, dichroic beamsplitter mirror and achromat lens to the confocal pinhole aperture which was set to 50 μm to prevent out-of-focus light from reaching the photomultiplier detector. A long-pass barrier filter (510 nm) was placed behind the pinhole aperture.
With an objective lens of × 10 (NA 0.45), each section was resolved into either 512×512 pixels with the size of 2 × 2 μm or 1024 ×1024 pixels of 1×1 μm in 256 gray scales. Scanning was repeated for 3-4 times at the same plane for averaging to improve the signal-to-noise ratio. The scanning plane was then changed by 1 or 2μm in the Z-axis direction corresponding to the dorso-ventral axis of the animal body, using a stepping motor attached to the fine control of the microscope. Sectioning of the whole cell including the cell body and all dendritic branches resulted in a series of 147 to 267 sections with the stepsize of 2μm. They were stored in a magneto-optical disk for later analyses.
The spatial distribution of individual dendritic branches was studied by examining the three-dimensional structure of the whole cell. We used a software (ImageSpace, version 3.0; Molecular Dynamics) implemented on the Sarastro 2000 system, to measure the diameter and length of each dendritic branch on the serial sections by re-displaying them on the workstation video screen. Each dendritic branch was represented by one cylinder or a connected sequence of cylinders each with specific diameter and length values. The whole LDS interneuron could be represented as an assembly of 418–645 cylinders. The length of each cylinder was calculated trigonometrically from the three-dimensional coordinates of the centers of its both ends. The diameter and length of dendrites in vivo were obtained by multiplying the measurements in the cleared specimen with the shrinkage factor (see above). The coordinates were stored together with diameters in a database running on a personal computer for quantitative presentation of the cell by a dendrogram (Fig.3). The dendrograms in this paper represent the real branch length, diameter and branching pattern of the interneuron in the three-dimensional space. The cell body was represented by an ellipsoid of revolution characterized by a revolution axis passing the attachment site of the dendrite with the cell body.
Cell shrinkage by clearing
Morphology of an LDS interneuron is compared as the horizontal projection image before and after the clearing process in Figure 1A, B. After fixation, dehydration and clearing, the interneuron apparently shrank to become smaller than before the clearing process. In order to examine the degree of the cell shrinkage, we have measured the length and diameter of major dendritic branches in 8 LDS interneurons from different preparations before and after the clearing process. Since the interneuron showed common basic branching pattern (see introduction), we selected 14 sites for the measurement of dendritic diameter and 4 sites for the measurement of length (Fig. 1C). The longest and shortest diameters of the soma were also examined. The diameter of dendritic branches and the soma were found to show significant shrinkage by the rate of 14.6 ± 1.3 % (mean ± SE) after clearing (P< 0.01 with Student's paired t-test). The length of dendritic branches decreased by 3.0±2.7% (Fig. 1D), but the difference was not statistically significant (P> 0.05). We found no significant difference in the shrinkage rate among different measurement sites and among different preparations (P > 0.05). It was thus suggested that the diameter of dendritic branches of the LDS interneuron shrank uniformly over the entire cell while the dendritic length remained unchanged. In the following part of this paper, the diameter of dendritic branches in vivo was determined by multiplying the measurements in the cleared preparation with the shrinkage factor of 1.17 (1/0.854).
Three-dimensional morphometry of the LDS interneuron
Projection images of an LDS interneuron are shown in Figure 2. The projection on the X-Y plane, i.e., the horizontal plane, was obtained by summing up all the optical sections with the stepsize of 2 μm (Fig. 2A1). The projection on the XZ plane, i.e., the frontal plane (Fig. 2A2), and that on the Y-Z plane, i.e., the sagittal plane (Fig. 2A3), were obtained computationally by summing up the each profiles of optical sections with the interpolation function turned off. The cell was divided into numerous cylinders on the basis of the diameter and turning point of the dendrite. The diameter and length of each cylinder were measured in the three-dimensional coordinate system using the stack of optical sections. The cell structure was then reconstructed as an assembly of cylinders each of which representing the corresponding part of dendrites. The projection images of the reconstructed cell on the X-Y, X-Z and Y-Z planes are shown in Figure 2B1-3 respectively. The real diameter and length in vivo of each dendritic branch, together with the soma, are represented as the Sholl's dendrogram (Sholl, 1953) in Figure 2C. In this study, beaded swellings in the dendrites on the side opposite to the soma as observed by the silver-intensification method (Kondoh et al, 1986) were not observed, although their size was reported to be large enough, ranging from 1.0 to 5.8 μm, for the spatial resolution of the confocal microscope. The dendrogram in the present study therefore does not include information on them.
Since the spread of synaptic current over the dendrite is dependent on the electrotonic structure of the cell (Rall, 1981; Rall et al, 1992), we further calculated the electrotonic length of each branch by dividing the anatomical length by the space constant (λ) which was obtained from the relationship,Rall, 1977). Rm was obtained from the relationship, tm =Rm .Cm where tm is the membrane time constant as experimentally obtained in this study from the transient response of the cell to step current injection, and Cm, membrane capacitance per unit area, was assumed to be 1 μF/cm2 (Rall, 1977). The electrotonic length of each branch is shown in Figure 2D. Rm was 12.4 ± 1.3 kΩ.cm2, ranging from 5.8 to 20.5.
Dendrograms of ten representative cells are illustrated in Figure 3. Comparison of dendritic structure based on these dendrograms showing anatomical and electrotonic properties of dendrites (Figs. 2C, D and 3) has revealed that individual interneuron differs from each other in several morphometrical parameters by a factor of 3.1 at the largest, including the total membrane area (ranging from 234,045 to 530,427 μm2), the total dendritic length (5,645 to 12,978 μm, the length from the midline to the terminal of the longest dendritic branch (605 to 933 μm), the mean length from the midline to the terminal of all dendritic branches (378.8 to 536.7 μm), the number of branching (48 to 91), maximal electrotonic length from the midline to the dendritic terminal (0.17 to 0.42), and the mean electrotonic length from the midline to the terminal of all branches (0.10 to 0.22; Table 1).
Morphometrical characteristics of the LDS interneuron*
Spatial distribution of dendrites in the neuropile
Superimposition of the three-dimensional structure of 13 LDS interneurons has revealed that the major branches (≥10 μm in diameter) of the interneuron show almost constant spatial distribution in the ganglionic neuropile in every preparation whereas the branching pattern of peripheral dendrites (<10 μm in diameter) show considerable variability from animal to animal (Fig. 4A-C). The position of each cell in the neuropile was normalized by assigning the transverse segment on the midline as the origin of the three-dimensional coordinate system. The neuropilar space was divided into 10 μm cubes and the number of intersection of each cube by the dendrite was counted for all 13 interneurons. The brighter part indicates that it is shared by larger number of interneurons. it is noted that the bright cubes make up a branching pattern common to individual cells especially on the X-Y plane (cf. Figs. 2A and 4A). On the Y-Z (Fig. 4B) and X-Z plane (Fig. 4C), although the bright regions are less demarcated than on the X-Y plane, the branching of major dendrites can be discriminated from the other parts of the cell. The dark part shown with an arrowhead in Figure 4B apparently corresponds to the sensory commissure i in the sixth abdominal neuromere, an axon bundle of sensory neurons crossing the midline (A6SCI; Kondoh et al, 1986). The position of the soma was also variable depending on the preparation.
Spatial distribution of the terminals of each dendritic branch was also found to differ from animal to animal. Dendritic terminals of three representative cells projected on the X-Z plane are illustrated in Figure 5. It is noted that each cell has clusters of terminals in different locations. Thus the three-dimensional morphometry has revealed that the same LDS interneuron shows variability in their dendritic structure in the individuals although it commonly shares basic branching pattern of major dendrites.
Common morphological characteristics of the LDS interneuron
Except the basic branching pattern of major dendrites, the LDS interneuron showed morphological variability. But quantitative comparison of the dendritic structure on the side ipsilateral to the soma with that on the opposite side revealed that the interneuron shared bilaterally asymmetrical dendritic structure in every preparation (Table 1; Fig. 6A). Dendrites on the soma side were significantly longer in the total anatomical length, and larger in the total area and in the number of branching than those on the opposite side (P<0.01). In particular, the length of dendrites whose diameter is less than 6 μm is significantly greater on the soma side than on the opposite side. There was no significant difference in the length of dendrites whose diameter was larger than 6 μm between both sides. Thus the difference in the total length of dendrites between the two sides is due to the difference in the length of fine dendrites. The mean electrotonic length between the midline and the most distal dendritic termination of 2 representative major branches on the soma side was also significantly longer (P<0.01) than that of 2 major branches on the opposite side (Fig. 6B, C).
The confocal laser scanning fluorescence microscopy has been demonstrated to provide high-quality three-dimensional images of nerve cells (Brodin et al, 1988; Carlson et al, 1989; Mossberg et al, 1990; Wallén et al, 1988). With the conventional fluorescence microscopy, it is often difficult to quantitatively analyze the cell structure because of a ‘halo’ around strongly fluorescent elements and light scattering during traveling through the specimen (Cullheim et al, 1987; Fleshman et al, 1988; Ulfake et al, 1981). In the confocal microscopy, the specimen is illuminated point-wise by a focused beam of light, and only the light emitted from the focal point is electronically detected (Wilson et al, 1984). Consequently, the out-of-focus information is significantly suppressed to greatly improve image quality. Taking advantage of this confocal laser scanning microscopy, we have quantitatively investigated the dendritic structure of an identified nonspiking interneuron of crayfish and found that it varies from animal to animal while retaining the basic morphological pattern. We will discuss below the functional significance of variability and invariability in the dendritic structure of the interneuron.
Variability in the LDS interneuron morphology
Although morphological variability of the LDS interneuron has been qualitatively known (Reichert et al, 1983; Takahashi et al, 1995), we have quantitatively shown for the first time how the dendritic structure of the interneuron varies in individuals. It was revealed that the total dendritic length including branches on the soma side and those on the opposite side ranged from 234,045 to 530,427 μm2 (N=13) (Table 1). Accordingly, the total membrane area varied among preparations. In addition to other anatomical characteristics, the mean electrotonic length from the midline to the terminal of all branches ranged from 0.10 to 0.22 (Table 1).
It should be noted here that the possibility that some branches in certain cases are not stained incidentally is unlikely from the following reasons: 1) the entire cell structure was stained brightly and the termination of individual branches was demarcated unambiguously from the background in each case, 2) the dye was always injected from the same part of the cell, i.e., the thick transverse segment on the midline, and 3) the cell body which was connected to a major dendritic branch with a long thin process and was most difficult to be stained was always visible. We therefore conclude that the observed difference in the dendritic structure reflects its intrinsic variability rather than any artifactual capriciousness of intracellular staining.
Since the dendrites of the LDS interneuron, as those of other local nonspiking interneurons, function not only in receiving synaptic inputs but also in releasing transmitters for the postsynaptic cells, the variability in dendritic structure is likely to result in functional variability of the interneuron. In particular, the finding that the spatial distribution of fine dendrites (Fig. 4) and their termination sites (Fig. 5) are variable, although that of major branches is rather constant (Fig.4), is functionally relevant since an electronmicroscopic study has revealed that most synapses are distributed over fine dendrites and not at all over major ones (Kondoh et al, 1986). If the spatial distribution of pre- and post-synaptic cell processes remains unchanged, the different branching pattern of fine dendrites of the LDS interneuron would lead to different synaptic connection of the interneuron with those cells. The fact that the mean electrotonic length from the midline to the terminal of all branches varies among preparations (Table 1) further indicates that the spatio-temporal distribution of synaptic potentials over dendrites is also variable although it is not determined solely by the electrotonic length of each branch but affected by the electrotonic structure of the whole cell (Jack, 1979; Rall, 1977, 1981; Rall et al, 1992).
Although the synaptic response of the LDS interneuron has been reported to be variable among preparations (Krenz et al, 1985; Reichert et al, 1983; Takahata et al, 1995), we can not at present relate this functional variability to any morphological one since the physiological experiment on the synaptic activity of the interneuron has so far been done in the condition that the sensory neurons that are presynaptic to the interneuron are stimulated not singularly but in concert (Krenz et al, 1985; Nagayama et al, 1988; Reichert et al, 1982, 1983; Takahashi et al, 1995; Takahata et al, 1995). For understanding the function of dendritic structure in receiving synaptic inputs, synaptic responses of the interneuron to activation of a specific mechanosensory hair as identified by Letourneau (1976) on the tailfan should be compared among LDS interneurons having different patterns of dendritic branching. Further study is needed to determine if the morphological variability of the LDS interneuron leads to any functional variability or simply reflects the morphological differences in the pre- and post-synaptic cells.
Invariability in the LDS interneuron morphology
Physiological experiments have suggested that the dendrites on the side ipsilateral to the cell body function in receiving mechanosensory inputs while those on the contralateral side distribute these inputs to ascending interneurons (Krenz et al, 1985; Reichert et al, 1983). An electronmicroscopic study (Kondoh et al, 1986) has demonstrated that ipsilateral dendrites are exclusively postsynaptic while contralateral ones have both input and output synapses intermingled on the same processes. In the present study, we have shown quantitatively for the first time that the dendritic structure is different between the ipsilateral and contralateral side at the light microscopic level. The ipsilateral dendrites are longer and more complex in branching pattern than the contralateral ones (Table 1). The difference in the whole dendritic length between both sides is mainly due to the difference in the length of fine dendrites (diameter < 6 μm). No significant difference is found in major dendrites (diameter > 6 μm). The spatial distribution of major dendrites was more stable among preparations than that of fine ones (Fig. 4).
The bilateral asymmetry in the dendritic structure of the LDS interneuron could be accounted for by different spatial distribution patterns of input and output cell processes. The interneuron has been reported to receive excitatory monosynaptic mechanosensory inputs from the uropod and telson on the ipsilateral side (Reichert et al, 1983). These inputs originate from mechanosensory cuticular hairs and are carried to the terminal ganglion by sensory nerves entering from the first to 5th roots each having specific receptive field on the tailfan (Calabrese, 1976). The sensory nerve terminals are distributed so extensively in the ganglion (Kondoh et al, 1987) that it is necessary for a cell to stretch out far-reaching dendrites in order to receive all of their inputs. The output target cells of the LDS interneuron have not been identified totally (Reichert et al, 1983), but it is predictable from its bilaterally asymmetrical structure that the dendritic arborization of the target cells is more restricted than that of the LDS interneuron.
Another possibility is that the asymmetric structure might bestow some specific electrotonic characteristics on the interneuron which functions in conveying the ipsilateral synaptic input to the contralateral side. Apparently, the thinner and longer dendrites on the ipsilateral side are disadvantageous for effective spread of synaptic potentials than thicker and shorter dendrites on the contralateral side: potentials are attenuated more seriously and become slower in time course during their passive spread over the dendrite when the distance is longer and the space constant is shorter (Jack, 1979; Rall, 1977). However, the spatio-temporal distribution of synaptic potentials on dendrites is based on not only the anatomical properties of individual dendritic branches but also the three-dimensional branching pattern of the entire cell as well as the membrane properties of dendrites (Jack, 1979; Rall, 1977; Rall et al, 1992). Therefore, comparison of anatomical parameters of individual branches alone is insufficient to understand the synaptic activity of a nerve cell. Functional significance of bilaterally asymmetrical dendritic structure of the LDS interneuron is still open to further analyses using its compartmental model based on three-dimensional dendritic structure and electrophysiological properties of the dendritic membrane (Perkel et al, 1981; Segev et al, 1989).
This work was supported in part by Grants-in-Aid (Nos.07640894, 07554071) from Ministry of Education, Science, Sports and Culture of Japan.