Cerebral Cortex Advance Access originally published online on August 17, 2005
Cerebral Cortex 2006 16(5):696-711; doi:10.1093/cercor/bhj015
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Dendritic Branch Typing and Spine Expression Patterns in Cortical Nonpyramidal Cells
Division of Cerebral Circuitry, National Institute for Physiological Sciences, Aichi Okazaki 444-8787, Japan
Address correspondence to Yasuo Kawaguchi, Division of Cerebral Circuitry, National Institute for Physiological Sciences, Myodaiji, Okazaki, 444-8787, Japan. Email: yasuo{at}nips.ac.jp.
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
To understand the dendritic differentiation in various types of cortical nonpyramidal cells, we analyzed quantitatively their dendritic branching and spine expression. The dendritic internode and interspine interval obeyed exponential distributions with type-specific decay constants. The initial branching pattern, internode interval and spine density at the light microscopic level divided nonpyramidal cells into three dendritic types, correlated with axonal, neurochemical and firing types. The initial branching pattern determined the overall vertical spread of dendrites. Basket cell subtypes with different firing and chemical expression patterns were distinct in the vertical and horizontal spatial spread, providing diverse input territories. Internode densities of dendritic spines, as well as those of axonal synaptic boutons, did not correlate with the tortuosities and intervals, suggesting a tendency to distribute synapses homogeneously over the arbor. Dendritic spines identified at the electron microscopic level were different in length and shape among subtypes. Although the density was lower than that of pyramidal cells, spines themselves were also composed of several morphological types such as mushroom and multihead ones, which were expressed differentially among subtypes. Correlation of dendritic branching characteristics with differences in spine structure suggests distinct ways to receive specific inputs among the subtypes.
Key Words: basket cell • double bouquet cell • fast-spiking cell • interneuron • Martinotti cell • neurogliaform cell
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The neocortex is composed of horizontal laminae and radial columnar structures (Szentágothai, 1975
; Peters and Sethares, 1996; Mountcastle, 1997) and contains various types of pyramidal and nonpyramidal cells (Ramón y Cajal, 1911; Jones, 1975, 1984; Fairén et al., 1984; White, 1989; Somogyi et al., 1998). The fundamental subtypes of nonpyramidal cells, GABAergic interneurons, have been gradually revealed (Kawaguchi and Kubota, 1997; Somogyi et al., 1998; Cauli et al., 2000; Gupta et al., 2000; Kawaguchi and Kondo, 2002; Markram et al., 2004), but it remains to be investigated how selective and precise the connections of the nonpyramidal cell subtypes are within the layered and vertical structure of the neocortex (Silberberg et al., 2002; Douglas and Martin, 2004). Quantitative representation of local circuit morphology constitutes a crucial step toward revealing the wiring principle in neocortex. To understand how cortical neurons form circuits, how their axons and dendrites elongate, branch and form synapses should be investigated (Abeles, 1991; Braitenberg and Schütz, 1998; Anderson et al., 2002; Thomson and Morris, 2002).
Previously we identified nonpyramidal cell subtypes by quantitative morphological characteristics in addition to the firing patterns (Karube et al., 2004). Among the identified subtypes, local axonal morphologies including synaptic boutons were quantitatively compared. Both the internode and interbouton intervals have been shown to obey an exponential distribution irrespective of subtypes, but the interval means are different among the subtypes. On the other hand, axonal branching angles are similar in both distribution pattern and average among subtypes. The dendritic branching pattern is also diverse among nonpyramidal cell subtypes (Jones, 1975; Feldman and Peters, 1978), but its local phenotypes have been investigated quantitatively only in a few cases (Parnavelas and Uylings, 1980; McMullen et al., 1984; Jin et al., 2001). We assumed that local dendritic morphologies could be governed by common rules and described by the quantitative parameters, and that the nonpyramidal cells subtypes may be grouped into some dendritic types.
Spine protrusions of pyramidal cells may be generated for the selective synapse formation (Stepanyants et al., 2002; Yuste and Bonhoeffer, 2004). Some GABAergic nonpyramidal cells also form spines along their dendrites (Feldman and Peters, 1978; Kawaguchi, 1993), although their density is lower than those of excitatory thick pyramidal cells and spiny stellate cells (Larkman, 1991; Lund and Holbach, 1991; Table 2). The morphological form and distribution of spines remains to be investigated in nonpyramidal cells. The spine density, distribution pattern along the dendrite and their morphological characteristics may be related to their functional roles in nonpyramidal cells. We assumed that spine formation pattern might be different among subtypes.
|
To reveal dendritic branching and spine formation rules and their differences among the previously identified subtypes of GABAergic nonpyramidal cells, we determined the following dendritic characteristics quantitatively: (i) spatial spread differences among subtypes; (ii) probability density functions describing the internode interval, terminal branch length, branching angles and interspine interval; (iii) the spine distribution pattern along the dendrite; (iv) dendritic spine morphologies; and (v) the relation between dendritic spine formation and the trajectory of the elongating branch, in comparison with synaptic bouton formation in axons. As a result, we found basic dendritic types among the nonpyramidal cells and correlation of dendritic branching characteristics with differences in spine structures in addition to its density.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Electrophysiology
The experiments were performed on young Wistar rats (19–23 days postnatal). Sections of frontal cortex were cut to a thickness of 300 µm and immersed in a buffered solution containing (in mM) NaCl, 124.0; KCl, 3.0; CaCl2, 2.4; MgCl2, 1.2; NaHCO3, 26.0; NaH2PO4, 1.0; and glucose, 10.0, aerated with a mixture of 95% O2 and 5% CO2. Cortical cells were recorded in a whole-cell mode at 30–31°C using a x40 water immersion objective. The electrode solution consisted of (in mM) potassium methylsulfate, 120; KCl, 5.0; EGTA, 0.5; MgCl2, 1.7; ATP, 4.0; GTP, 0.3; HEPES, 8.5; and biocytin, 17. Input resistances of cells were determined by linear regression of zero and three voltage shifts up to 20 mV negative to rest induced by hyperpolarizing current pulses (duration, 500–600 ms), and by subtraction of the series resistance. Membrane time constants were determined by passing hyperpolarizing current pulses inducing voltage shifts of 6–15 mV negative to rest. Spike-widths at half amplitude were measured for spikes elicited within 10 ms by depolarizing current pulses. Spike afterhyperpotentials were measured from the spike onset potential. By linear discriminant analysis of defined firing types (R, free software; http://cran.r-project.org/), we obtained the first discriminant function from the above physiological parameters and calculated discriminant scores for individual cells (Wang et al., 2002).
Histology and Immunohistochemistry
Tissue slices were fixed by immersion in 4% paraformaldehyde, 1.25 or 0.05% glutaraldehyde and 0.2% picric acid in 0.1 M phosphate-buffered (PB) solution. Slices were cut at thickness of 50 µm. Each slice (a set of sections) was further treated by one of the following two procedures.
- Some slices were incubated with avidin–biotin–peroxidase complex (Vector, Burlingame, CA) in Tris–HCl buffered saline (TBS) with or without 0.04% Triton X-100 (TX), and reacted with 3,3'-diaminobenzidine tetrahydrochloride (DAB) (0.05%) and H2O2 (0.003%).
- The other slices were processed for fluorescence immunohistochemistry. The slices were incubated with the primary antibodies in TBS containing 2% bovine serum albumin, 10% normal goat or horse serum and 0.5% (or 0.04%) TX. After washing in TBS, they were incubated in fluorescent secondary antibodies, followed by incubation with Alexa 350 streptavidin (S-11249, Molecular Probes, Eugene, OR) in TBS. After examination for fluorescence, the slices were incubated with avidin–biotin–peroxidase complex, and reacted with DAB and H2O2.
They were then postfixed in 1% OsO4 in PB, dehydrated and flat-embedded on silicon-coated glass slides in Epon. After fixation, dehydration and embedding in Epon, slices shrank to about 90% in length (87 ± 4% in slice flat surface and 90 ± 2% in slice thickness; 11 slices; Karube et al., 2004). The shrinkage was not corrected in the analysis.
After reconstruction by Neurolucida, stained cells were photographed using a x100 objective and serially sectioned using an ultramicrotome. The thickness of ultra thin sections was calibrated by a color laser 3D profile microscope (VK-9500; Keyence, Japan). Electron micrographs were taken with a Hitachi H-7000 electron microscope (EM), using tilting up to 60°. EM images of the labeled terminals and associated structures were captured using a CCD camera and reconstructed three-dimensionally (Visilog; Noesis, France).
Quantitative Morphology
Somata, axons, dendrites, boutons and spines of stained cells were reconstructed three-dimensionally, using the Neurolucida system (MicroBrightField, Williston, VT) with a x60 objective and analyzed with NeuroExplorer (MicroBrightField). We measured the following morphological parameters at the light microscopic (LM) level: (i) vertical spread (ratio of dendrite length outside the 100-µm-thick horizontal slab centering on the soma to the total length; Fig. 1); (ii) horizontal spread (ratio of dendrite length between 100- and 400-µm-diameter columns to the length within a 400-µm-diameter column centering on the soma; Fig. 1); (iii) number of primary dendrites (dendrites issued from the soma); (iv) vertical direction bias (VDB; calculated from the frequency distribution of vertical directions, Fig. 2): if all vertical directions were between 80 and 90°, and between 0 and 10°, the VDB would be 1 and 0, respectively; (v) spine density (LM) (total spines divided by the total dendritic length; /10 µm); (vi) internode interval [lengths between two successive nodes (branch points) along the dendrite including those from the soma origin to the first node]: an order was assigned to each internode interval in a centrifugal way; (vii) internode tortuosity (ratio of the internode interval to the direct distance); (viii) terminal branch length (length of dendritic segment between the tip point and the node just prior to it): the tip point is an authentic ending within the slice, not an ending due to the slice preparation; (ix) branching aperture (the angle between daughter branch vectors, Fig. 4) (Karube et al, 2004); and (v) branching tilt (the angle between the parent branch vector and the vector midway between daughter branch vectors).
|
), T is the membrane time constant (ms), W is the spike width at a half amplitude (ms) and A is the spike afterpotential (mV). Non-FS cells were >0.1; LS and FS cells were <–0.3; cells that were <–2.5 were FS cells.
|
|
Dendritic surface protrusions were investigated at both the LM and EM level. At the LM level, protrusions during Neurolucida reconstruction were regarded as spines, from which the spine density and interspine interval (interval between dendritic roots of successive spines) were obtained. To see the relationship of the internode tortuosity with internode spine or axonal bouton density, we used the dendritic or axonal branches with at least one spine or bouton.
At the EM level, processes >300 nm from the dendritic shaft were tentatively called protrusions including spines (EM). From reconstructions of serial ultrathin sections, we measured the interspine interval, direction change of successive spine roots and spine length. To measure the width and thickness of spine heads and the spine neck circumference, we selected spines reconstructed well.
Frequency histograms were fitted by the Gaussian, exponential or gamma distribution. The shape parameter of the gamma distribution determines the curve pattern including the skew symmetry.
Statistical Analysis
Data are given as mean ± SD. Statistical analysis was performed using StatView (SAS, North Carolina). Analysis of variance (ANOVA) was used for confirmation of significant differences among subtypes in individual parameters, followed by post-hoc Tukey-HSD test for the group comparison (
<0.01 or 0.05). For comparing two cell classes, the Mann–Whitney U-test was used (P < 0.01 or 0.05). The goodness of curve fitting was evaluated by the square of correlation coefficient (r2) between the measured values and the calculated ones from the fitting curve.
We applied principal component analysis (PCA) to the standard scores of morphological variables (R, free software) and obtained the principal components (PCs) (Sultan and Bower, 1998). We mapped nonpyramidal cells by PC1 and PC2 to reveal cell clusters with similar characteristics.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Nonpyramidal Cell Identification and their Subtypes
In contrast to pyramidal cells, most nonpyramidal cells lacked apical dendrites and had fewer spines (Figs 1A, 2B and Table 2). Previously we identified nine subtypes among 91 nonpyramidal cells by three types of firing patterns and quantitative axon morphometric parameters (Karube et al., 2004). According to this systematic classification scheme, we quantitatively identified the following six subtypes (88 cells) in this paper and analyzed dendritic characteristics in detail: (i) fast-spiking (FS) basket cells; (ii) late-spiking (LS) neurogliaform cells; (iii) non-FS Martinotti cells; (iv) non-FS large basket cells; (v) non-FS small basket cells; (vi) non-FS double bouquet cells (Fig. 1A and Table 1). Non-FS cells used here were also immunohistochemically characterized (Table 1). The above six subtypes were selected because most GABAergic cells are contained in these subtypes (Kawaguchi and Kubota, 1997; Karube et al., 2004). In layer II/III of rat frontal cortex, parvalbumin and somatostatin cells, separate populations, account for 
40% and 20% of GABAergic cells, and their considerable parts are occupied by FS basket and Martinotti cells, respectively (Kubota et al., 1994; Kawaguchi and Kubota, 1997; Y. Hirai and Y. Kawaguchi, unpublished observations). VIP cells, separate from parvalbumin and somatostatin cells, account for 20% of GABAergic cells, and their considerable parts are occupied by double bouquet and small basket cells. CRF cells belong to VIP cells in layer II/III. These suggest that the above subtypes dominate the large part of GABAergic populations.
|
Linear discriminant analysis of physiological parameters mostly differentiated FS, LS and non-FS cells used for this study (error = 4.7%), confirmed by leave-one-out cross validation (error = 1.6%). The three firing types showed different discriminant scores of the first discriminant function (Fig. 1C).
Dendritic Spatial Spread
Morphometric parameters of dendritic branching and spine formation were compared among subtypes in layer II/III (Fig. 2A). The parameters of FS basket and Martinotti cells were also compared between layer II/III and V (Table 3). To reveal dendritic field differences among subtypes, we compared the vertical and horizontal spread which represented how loose or compact dendritic branches were around the somata (Fig. 1B). Each subtype had a characteristic combination of horizontal and vertical spread centered on the soma (Fig. 4I). The dendritic fields of neurogliaform and FS basket cells were vertically more compact than those of the other cells. The dendritic fields of large basket cells were horizontally more diffuse than those of neurogliaform, double bouquet, small basket and FS basket cells.
|
Initial Branching Pattern and Spine Density
The number of primary dendrites ranged from 2 to 11. Neurogliaform and FS basket cells had more primary dendrites than double bouquet, small basket and large basket cells (Figs 2A, 4A). Initial dendritic direction was used to calculate a vertical direction bias (Fig. 2C). Dendrites of double bouquet, small basket and large basket cells were more vertically oriented than FS basket and neurogliaform cells (Figs 2A, 4B). Neurogliaform and FS basket cells did not show any direction preference.
Martinotti cells were much higher in spine density at the LM than the other nonpyramidal cell subtypes (Fig. 2A,B, 4C), but clearly lower than pyramidal cells (Fig. 2B and Table 2). Layer V Martinotti cells were lower in spine density than layer II/III ones (Table 3). These spine densities counted at the LM level are lower than those of excitatory spiny stellate cells (Lund and Holbach, 1991; Datwani et al., 2002), layer II/III (Larkman, 1991; Mataga et al., 2004; Bock et al., 2005) and layer V (Riccio and Matthews, 1985; Larkman, 1991) pyramidal cells (Table 2).
Dendritic Branching and Termination
The frequency histogram of dendrite internode intervals followed an exponential distribution (Fig. 3A1,B). In correspondence to the exponential, the slope of the regression line of the interval SD versus mean for individual cells was close to 1 (Fig. 3A2). The decay constant calculated from the fitted exponential correlated with the interval mean for each subgroup (Fig. 3B and Table 4). The internode intervals of neurogliaform and FS basket cells were shorter than those of large basket, double bouquet and Martinotti cells (Figs 2A, 4D). Layer V FS basket cells were longer in internode interval than layer II/III ones (Table 3). These indicate that dendritic branching occurs superficially according to a Poisson process, but the mean interval varies among nonpyramidal cell types (20–60 µm; Table 4).
|
, decay constant of the fitted exponential; r2, square of correlation coefficient between the measured values and the calculated ones from the fitting curve. (C1,2) Frequency histograms of terminal branch lengths of neurogliaform and Martinotti cells, respectively. (D1,2) Internode intervals at each branching order of neurogliaform and Martinotti cells, respectively. Total intervals at each order (number in parenthesis) were pooled for individual subtypes. Intervals were normalized by the internode interval mean of each cell. The interval was independent of the order, but the terminal branch was longer than the interval.
|
The intervals were independent of branching order (Fig. 3D), but the terminal branch length was longer than the mean internode interval (Fig. 3C,D), especially in FS basket and LS neurogliaform cells (Table 4). The terminal branch length distributions were fitted to an exponential (Fig. 3C; Table 4). Both internode interval and terminal branch length followed an exponential distribution, but were independently determined. As shown in the dendrogram (Fig. 2A), terminal branches distant from the soma were not necessarily short, suggesting that dendritic termination might not be determined by the elongation history. These data indicate that branching frequency and elongation termination are differentially regulated in nonpyramidal cell dendrite patterning.
In a similar way to the axon, the dendritic bifurcations obeyed a Gaussian distribution in their aperture angle (mean, 80°; Fig. 4E and Table 4) and a gamma distribution (shape parameter = 2) in their tilt angle (mean, 30°; Fig. 4F and Table 4). They did not show significant differences among subtypes, indicating that the manner of generating daughter branches at the node is common to all nonpyramidal cell types.
Three Dendritic Types
Since the initial branching pattern (primary dendrite number and initial direction, Fig. 4G, upper) and the formation of nodes and spines (internode interval and spine density, Fig. 4G, lower) were significantly different among subtypes, we compared the nonpyramidal cell subtypes using principal component analysis of these four parameters. Because the first two principal components (PC1 and PC2) could account for a large amount of the total variance (77%), we mapped layer II/III nonpyramidal cells by them (Fig. 4H). This procedure revealed three clusters on the PC1–PC2 plane: (i) FS basket and LS neurogliaform cells; (ii) non-FS Martinotti cells; and (iii) non-FS large basket, small basket and double bouquet cells. This map was made from the parameters of dendritic morphology, but the clusters on the PC plane were correlated with grouping based on the firing pattern, neurochemical content and axonal characteristics. Thus, nonpyramidal cells subtypes can be grouped into some dendritic types.
FS basket and LS neurogliaform cells were similar in vertical spreads, as non-FS large basket, small basket and double bouquet cells (Fig. 4I). These suggest that vertical elongation patterns are similar within each dendritic type in the above. In each type, horizontal extension patterns were heterogeneous. Three types of basket cells showed different combinations of the vertical and horizontal elongation patterns, suggesting their distinct input characteristics.
Spine Distribution of Martinotti Cells
Spine densities at the LM level were variable among subtypes. Martinotti cells had higher dendritic spine density than the other types (Figs 2A, 4C). In Martinotti cells, the spine density was low within 50 µm from the soma, and progressively increased to its peak 100 µm along the dendrite from the soma (Fig. 5A,B). Following this peak, the density remained relatively constant for several hundred micrometers along the dendrite, and decreased gradually again towards the tip. This indicates a uniform spine distribution on Martinotti cell dendrites, except along the initial and tip portions of the dendritic tree. The spine density of Martinotti cells was between one-fifth and one-third of pyramidal cells (Table 2 and Fig. 2B), but this distribution pattern was similar to spine distributions in pyramidal cells in the rat (Larkman, 1991; Trommald et al., 1995) and the primate (Elston and Rosa, 1997; Elston et al., 1999).
|
The interval between successive spines (interspine interval) was measured at the LM level in dendrites of Martinotti cells. The distribution was well fitted to an exponential, except for intervals shorter than 1 µm (Fig. 5A3; r2 = 0.98 ± 0.01 in layer II/III, 0.94 ± 0.06 in layer V). The interval distribution between 100 and 300 µm from the soma was similarly fitted to an exponential (Fig. 5A4). We speculated that the interspine interval might observe an exponential distribution, but that spines with intervals of <1 µm could not be counted accurately at the LM level, due to their masking by the adjacent spine.
To measure the interspine intervals more correctly, we reconstructed 15 dendritic segments (length, 11.2–28.2 µm; total reconstructed length, 231.4 µm: 6.8% of whole dendritic length, 3420.7 µm) from serial ultra thin sections of a Martinotti cell (Fig. 6A). These dendritic segments were selected because they were relatively higher in spine density. We realized that at LM level some spines could not be counted, and spines with two or three heads could not be differentiated (Fig. 6B). We found 251 dendritic protrusions longer than 300 nm by EM. In the same dendritic regions, 112 spines were counted during Neurolucida reconstruction using a x60 objective (45% of protrusions by EM).
|
Among the 251 protrusions, 211 (84.2%) were considered to be spines. The remaining were treated as filopodia (13.9%) and fan-like processes (2%) (see below). Synaptic specializations were found in
70% of spines by EM observation (Fig. 6B2). Comparing the EM reconstructions with the LM photograph taken using a x100 objective, we could relate 177 spine heads by EM with spines by LM (71% of spine heads by EM). Presumably, spines with the short length or perpendicular to the section could not be seen by LM. In the eight segments (total length, 135.4 µm; surface area, 370.7 µm2) of a Martinotti cell used for the EM analysis, at >100 µm from the soma the protrusion density on the dendritic shaft was 1.15/µm (per length) and 0.42/µm2 (per surface area). The spine head density was 1.26/µm (per length).
The interspine intervals at the EM level, both those obtained from all segments and those from segments >100 µm from the soma, including intervals <1 µm, fitted well to an exponential distribution (Fig. 6C). The spine protrusion direction was independent of that of the previous one (Fig. 6D). Taken together, these data indicate that spines of Martinotti cell dendrites are produced at superficially random intervals and directions along the dendrite, except along the initial and tip portions.
In the segments >100 µm from the soma, the interspine interval was 0.96 ± 0.07 µm (n = 125) at EM and 1.88 ± 1.58 µm (n = 69) at LM level using a x60 objective. By assuming that interspine intervals may be 50% of those measured during Neurolucida reconstruction, we estimated the mean interspine intervals from light microscopically obtained intervals in Martinotti cells. In the constant region (100–300 µm away) of Martinotti cell dendrites the LM mean interval was 2.92 ± 0.39 µm in layer II/III cells and 3.86 ± 0.93 µm in layer V cells. The corrected mean was 1.5 ± 0.2 µm (range, 1.2–1.8) in layer II/III cells and 2.3 ± 0.6 µm (1.3–2.7) in layer V cells.
Trajectories and Spine Formation in Comparison with Axonal Bouton Formation
Dendrites do not elongate in straight lines, but meander between nodes (Fig. 7A1). Dendritic internode intervals (length along the dendrite) linearly increased with the direct internode distance irrespective of cell types (correlation coefficient, 0.98–0.99; regression coefficient, 1.25–1.45). The distributions of internode tortuosities (ratios of the internode interval to the direct distance) were skewed (Fig. 7B). The internode tortuosity was only weakly correlated with the internode interval (Fig. 7A2; correlation coefficient, 0.24–0.5). These data indicate that dendritic branch trajectories are independent of the branch lengths.
|
If a dendrite or an axon actively searched for its specific inputs or targets, the tortuosity would be expected to vary among subtypes that have different synaptic preferences. Indeed the dendritic tortuosity of Martinotti cells was larger than that of other types of cell (Fig. 7B). Layer II/III and V Martinotti cells were similar in their dendritic tortuosity. This indicates that Martinotti cells extend dendrites in a more meandering manner. The distributions of axon internode tortuosities were also skewed (Fig. 8B). FS basket cells had smaller axonal tortuosity than the other types (
< 0.05), indicating that FS basket cells extend axons in a straighter manner.
|
The tortuosity varied not only among subtypes, but also among the internode intervals of the same cell. By meandering, a dendrite may be able to find synaptic targets that it may not find if elongating in a straight trajectory; tortuosity may thus reflect a searching activity for specific targets (Stepanyants et al., 2004
). Since most boutons and spines participated in synapses (Karube et al., 2004; see also above), they can be used as indicators for synapses. Some internodes (13.6%) did not have spines, but their total length made up of a small proportion (2.35%) of the whole length. Therefore we excluded the internode interval without spines for the analysis. To see the correlation between the trajectory and synapse formation, we examined the relation between the spine or bouton internode density and the internode tortuosity and interval. In Martinotti cell dendrites, internode spine density did not correlate with internode tortuosity (Fig. 7C). The correlation of the internode spine density with the interval was also weak (Fig. 7C, box). The distribution of internode spine densities could be fitted to a Gaussian distribution (Fig. 7C, insets).
We compared axonal bouton formation between layer II/III Martinotti and FS basket cells, because these two types of axons were significantly different in the internode bouton density, tortuosity and interval (Fig. 8A–C). Some internodes did not have boutons (17.74% in FS basket cells, 10.75% in Martinotti cells), but their total length was very short (4.95% in FS basket cells, 2.44% in Martinotti cells). Therefore we excluded the internode interval without boutons for the analysis. Martinotti cells had larger mean bouton density, axonal tortuosity and internode interval than FS basket cells (Fig. 8C). In Martinotti and FS basket cell axons, the internode bouton density did not correlate with the tortuosity (Fig. 8D1,2). The correlation of the internode bouton density with the interval was also weak (Fig. 8D, box). The distribution of internode bouton densities of both groups could be fitted to a Gaussian, and had similar coefficients of variation to that of the internode spine densities of Martinotti cells (0.4; Figs 7C, 8D, insets). These data indicate that meandering and formation of spines or boutons were independently determined among individual branches of nonpyramidal cells, and that axonal or dendritic branches create boutons or spines with a probability characteristic of individual nonpyramidal cell subtypes, irrespective of their tortuosity and length. In other words, synapses are generated along the nonpyramidal cell protrusion at the same probability, specific for each cell type, independently of the manner of elongation.
Morphological Differences of Dendritic Spines among Subtypes
To see the spine morphological differences among subtypes, we reconstructed dendrites of three nonpyramidal cells: Martinotti (see above), double bouquet (28 dendritic segments; length, 4.5–32.6 µm; total reconstructed length, 486.4 µm: 12.3% of whole dendritic length, 3970.3 µm) and FS basket cells (14 segments; length, 11.2–30.1 µm; total length, 240 µm: 8.2% of whole dendritic length, 2927.7 µm). Among protrusions >300 nm, we found spines and other protrusions such as filopodia and fan-like processes (Fig. 9B,C). The filopodia were slender protrusions without head differentiation and observed in all three cells. Fan-like processes were found in double bouquet and FS basket cells.
|
The spine length was different among the three cell types (Martinotti, 1.55 ± 1.14 µm; double bouquet, 1.03 ± 0.78 µm; FS basket, 0.62 ± 0.29 µm; P < 0.01; Fig. 9D). Martinotti cells had many protrusions >1 µm (61%, versus 37% in double bouquet cells and 10% in FS basket cells). There are various forms of spines in nonpyramidal cells like pyramidal cells, including multi-head, mushroom, thin and stubby types (Peters and Kaiserman-Abramof, 1970
; Harris et al., 1992; Fiala and Harris, 1999; Figs 6B, 9A). Some spines were hardly classified into one of these categories. Martinotti cells had branched spines with two heads (Fig. 6B3; 9%) and three heads (Figs 6B3,4, 9A; 0.9%). In addition to multi-branch spines, the Martinotti cell had unbranched spines with two swellings (double-swelling spine, 2.4%; Fig. 9A). Thus the Martinotti cell had multi-head spines (12.3%), including double-branch, triple-branch and double-swelling spines (Fig. 9F). One of two branches of a Martinotti double-branch spine had three swellings. This kind of multiple swelling was not found in the FS basket and double bouquet cells.
Mushroom-type spines (Peters and Kaiserman-Abramof, 1970; Harris et al., 1992; Fiala and Harris, 1999; Figs 9A,B, 10A), with larger head width than neck diameter, were common in the Martinotti cell (Fig. 9E) (length, 1.81 ± 0.65 µm, n = 43; Fig. 10B), but there were fewer in the double bouquet cell (length, 1.59 ± 0.58 µm, n = 17; Fig. 10B) and very few in the FS basket cell (Fig. 9E). We measured dimensions of mushroom spine heads and necks. The heads of mushroom spines were not spherical in most cases, but flattened (Fig. 10A,C). The neck diameter was calculated from the circumference, assuming a circle. The neck diameter was similar to the spine head thickness (Fig. 10C).
|
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
To understand the formation rule of cortical local circuits, it is necessary to reveal how the dendrite and axon make daughter branches and distribute synaptic junctions along them. Following quantitative determination of dendritic morphological parameters for each nonpyramidal cell subtype (those of Martinotti cells shown in Fig. 6E), we identified basic dendritic types (Fig. 4), correlated with axonal, neurochemical and firing types, and also with vertical dendritic spreads. Concerning the synapse distribution along neurites, nonpyramidal cells might tend to distribute synapses homogeneously over the arbor except the initial and tip portions of dendrites because internode densities of dendritic spines, as well as those of axonal synaptic boutons, did not correlate with the tortuosities and intervals. From the differences in spine density, length and head morphology, each subtype may develop specialized forms of synaptic structures according to its strategy in connection formation within the intracortical circuit.
Functional Implications of Subtype Differences in Dendritic Morphology
On the basis of their initial dendritic branching pattern, and their node and spine formation, nonpyramidal cells were divided into three groups. The primary dendrite number, initial direction bias, mean internode interval and spine density can be used for identifying these major three dendritic types. Group I (including FS basket and LS neurogliaform cells) had more primary dendrites issuing at all directions and shorter internode intervals. Group III (double bouquet, small basket and CCK large basket cells) had fewer primary dendrites with preferred direction to the pia and white matter and longer internode interval. Group II (Martinotti cells) showed intermediate characteristics except for a higher spine density. These three groups are correlated with the expression of peptides and calcium binding proteins (Kubota et al., 1994; Kubota and Kawaguchi, 1997; Kawaguchi and Kondo, 2002; Karube et al., 2004). Group I cells did not express any of the peptides tested here, but FS cells are known to express parvalbumin. Group II expressed somatostatin, while Group III expressed VIP, CCK, CRF and calretinin in layer II/III. In addition to the functional differences, these dendritic variations may reflect developmental differences. Indeed, parvalbumin-containing cells (belonging to Group I) and somatostatin-containing cells (Group II) originate within the medial ganglionic eminence, whereas calretinin cells (Group III) appear to derive from the caudal ganglionic eminence (Xu et al., 2004). Nonpyramidal cells of the same type are electrically connected through gap junctions (Gibson et al., 1999). FS cells and LS neurogliaform cells, respectively, are frequently electrically coupled (Galarreta and Hestrin, 1999; Amitai et al., 2002; Chu et al., 2003; Fukuda and Kosaka, 2003; Simon et al., 2005). The dendritic morphology of these subtypes, namely more primary dendrites and frequent branching without direction preference could contribute to the frequent formation of gap junctions.
Subtype-specific Stochastic Branching and Ending of Dendrites
The aperture and tilt angles were very similar among subtypes and also similar to those of axons, and followed the Gaussian and gamma distributions, respectively (Karube et al., 2004). The average angles were similar between the two types of neurites (aperture, 80 ± 30°; tilt peak 30–40°). Internode intervals of both dendrites and axons followed an exponential distribution, and their means were different among the subtypes. These findings suggest that the dendritic and axonal branchings obey the same rules, and that only the branching probabilities are differentially regulated between subtypes or between dendrites and axons (Karube et al., 2004; Binzegger et al., 2005). The internode interval means of FS basket cells were different between layers II/III and V, suggesting that the same type of cell may change its dendritic branching frequency, according to its laminar position.
The internode interval was independent of the dendritic order, but the terminal branch length was longer than the internode interval, especially in FS basket and LS neurogliaform cells. The terminal branch length may therefore be regulated differently from the internode interval, probably because the terminal branch continues to increase in length well after branching has stopped (McMullen et al., 1988). The dendritic length of GABAergic cells is regulated by trophic factors (Jin et al., 2003; Kohara et al., 2003). Parvalbumin cells change the distribution of synaptic inputs along the dendrite during juvenile period (Yamashita et al., 2003), and are thought to be involved in forming specific connection during the critical period of postnatal development in juvenile animals (Fagiolini et al., 2004). Thus, FS basket and LS neurogliaform cells may change their dendritic field dynamically later in development than the other types (Berghuis et al., 2004), and the longer length of terminal branches possibly reflects this delayed maturation of dendritic morphology in Group I cells. Basket cells of Groups I and III are distinct in their dendritic structures, in addition to physiological, chemical and axonal characteristics (Kawaguchi, 2001; Wang et al., 2002; Karube et al., 2004).
Branch and Synapse Formation Independent of Elongation Trajectories
The interspine interval followed an exponential distribution like the interbouton interval of axons (Braitenberg and Schütz, 1998; Anderson et al., 2002; Karube et al., 2004). Neurites did not show a high tendency to meander (e.g. mean dendritic tortuosity, 1.2–1.4), but most branches made synaptic junctions at the similar density inherent in the cell, irrespective of the elongation trajectories. Considering the connection selectivity of the cortical circuit (Somogyi et al., 1998; Douglas and Martin, 2004; Stepanyants et al., 2004), synaptic sites specific for each cell type seem to be distributed in a random and non-clustered way in the local area such as within a vertical column. This arrangement may be a result of dendritic and axonal elongation to the preferred directions with stochastic branching in addition to random placement of individual cell types (Winfield et al., 1981; Morrison et al., 1984; Anderson et al., 2002; Fukuda et al., 2004). Since neurite extending patterns are restricted by mechanical constraints such as bifurcation of a parent branch with similar branching angles, this kind of cell arrangement and neurite extension pattern may be suitable for the axon to find a specific postsynaptic structure and for the dendrite to meet a specific input (Kalisman et al., 2005). However, synaptic boutons are also generated on the same continuous target apparently according to a Poisson process such as chandelier cell terminals on axon initial segments (Karube et al., 2004). To understand developmental principles of cortical circuit formation, it will be important to reveal how nonpyramidal cells can make synapses on specific postsynaptic targets by distributing synapses at type-specific densities homogeneously along neurites with superficially random intervals, independently of elongation trajectories.
Dendritic Spine Differentiation in Nonpyramidal Cell Dendrites
Martinotti cells had more spines than the other subtypes, especially FS basket cells. At the EM level, spines of Martinotti cell dendrites were longer in length and more in the proportion of mushroom and multi-head types than those of FS basket and double bouquet dendrites. Martinotti and FS basket cells are different in output target selectivity (Kawaguchi and Kondo, 2002). The difference in spine density and length may reflect that of input selectivity (Gibson et al., 1999). Martinotti cell dendrites may protrude spines to connect with specific presynaptic axons distant from them (Stepanyants et al., 2002; Yuste and Bonhoeffer, 2004; Richards et al., 2005). As a result, they made longer spines, and sometimes multi-branch ones, depending on the spatial distribution of specific targets and their activities (Dhanrajan et al., 2004). If specific presynaptic axons might approach FS cell dendrites and make synapses, they would not make spines. The short-term modification of excitatory inputs from pyramidal cell is different among the subtypes: Martinotti cells show facilitation, but FS and double bouquet cells show depression (Markram et al., 1998; Reyes et al., 1998; Rozov et al., 2001; Koester and Johnston, 2005). Each subtype may develop specialized forms of dendritic spines according to the differences in connection formation and synaptic transmission (Passafaro et al., 2003).
In addition to axonal branching and synaptic bouton formation (Karube et al., 2004), dendritic branching and protrusion formation were quantitatively different among the nonpyramidal cell subtypes. The quantitative differences of their input and output structures (branching direction and frequency; target selectivity; synapse density; and local spine morphology) among the subtypes suggest that they develop synaptic connections in the cortical local circuit using a strategy distinct in each subtype. To understand the cortical synapse formation, it is necessary to reveal the connection selectivity of each subtype in further detail.
|
Acknowledgments |
|---|
We thank Ms M. Saito and Mr Y. Ito for technical assistances, and Dr A. Agmon for comments. We thank Dr C.J. Wilson for helpful discussions, Drs. J. Bock and N. Mataga for providing data of spine densities in pyramidal cells, and MicroBrightField for technical support. We are grateful to Dr W. Vale for an antiserum against corticotropin-releasing factor. This work was supported by grants-in-aid for scientific research (15300110, 17022045) from the Ministry of Education, Culture, Sports, Science and Technology of Japan and by Toyota Physical and Chemical Research Institute.
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Abeles M (1991) Corticonics: neural circuits of the cerebral cortex. New York: Cambridge University Press.
Amitai Y, Gibson JR, Beierlein M, Patrick SL, Ho AM, Connors BW, Golomb D (2002) The spatial dimensions of electrically coupled networks of interneurons in the neocortex. J Neurosci 22:4142–4152.
Anderson J, Binzegger T, Douglas R, Martin K (2002) Chance or design? Some specific considerations concerning specific boutons in cat visual cortex. J Neurocytol 31:211–229.[CrossRef][ISI][Medline]
Berghuis P, Dobszay MB, Sousa KM, Schulte G, Mager PP, Härtig W, Görcs TJ, Zilberter Y, Ernfors P, Harkany T (2004) Brain-derived neurotrophic factor controls functional differentiation and microcircuit formation of selectively isolated fast-spiking GABAergic interneurons. Eur J Neurosci 20:1290–1306.[CrossRef][ISI][Medline]
Binzegger T, Douglas RJ, Martin KA (2005) Axons in cat visual cortex are topologically self-similar. Cereb Cortex 15:152–165.
Bock J, Gruss M, Becker S, Braun K (2005) Experience-induced changes of dendritic spine densities in the prefrontal and sensory cortex: correlation with developmental time windows. Cereb Cortex 15:802–808.
Braitenberg V, Schütz A (1998) Cortex: statistics and geometry of neural connectivity. Berlin: Springer.
Cauli B, Porter JT, Tsuzuki K, Lambolez B, Rossier J, Quenet B, Audinat E (2000) Classification of fusiform neocortical interneurons based on unsupervised clustering. Proc Natl Acad Sci USA 97:6144–6149.
Chu Z, Galarreta M, Hestrin S (2003) Synaptic interactions of late-spiking neocortical neurons in layer 1. J Neurosci 23:96–102.
Datwani A, Iwasato T, Itohara S, Erzurumlu RS (2002) NMDA receptor-dependent pattern transfer from afferents to postsynaptic cells and dendritic differentiation in the barrel cortex. Mol Cell Neurosci 21:477–492.[CrossRef][ISI][Medline]
Dhanrajan TM, Lynch MA, Kelly A, Popov VI, Rusakov DA, Stewart MG (2004) Expression of long-term potentiation in aged rats involves perforated synapses but dendritic spine branching results from high-frequency stimulation alone. Hippocampus 14:255–264.[Medline]
Douglas RJ, Martin KA (2004) Neuronal circuits of the neocortex. Annu Rev Neurosci 27:419–451.[CrossRef][ISI][Medline]
Elston GN, Rosa MG (1997) The occipitoparietal pathway of the macaque monkey: comparison of pyramidal cell morphology in layer III of functionally related cortical visual areas. Cereb Cortex 7:432–452.
Elston GN, Tweedale R, Rosa MG (1999) Cellular heterogeneity in cerebral cortex: a study of the morphology of pyramidal neurones in visual areas of the marmoset monkey. J Comp Neurol 415:33–51.[CrossRef][ISI][Medline]
Fagiolini M, Fritschy JM, Löw K, Möhler H, Rudolph U, Hensch TK (2004) Specific GABAA circuits for visual cortical plasticity. Science 303:1681–1683.
Fairén A, DeFelipe J, Regidor J (1984) Nonpyramidal neurons: general account. In: Cerebral cortex. Vol. 1. Cellular components of the cerebral cortex (Peters A, Jones EG, eds), pp. 201–253. New York: Plenum.
Feldman M, Peters, A (1978) The forms of non-pyramidal neurons in the visual cortex of the rat. J Comp Neurol 179:761–793.[CrossRef][ISI][Medline]
Fiala JC, Harris KM (1999) Dendrite structure. In: Dendrites (Stuart G, Spruston N, Häusser M, eds), pp. 1–34. Oxford: Oxford University Press.
Fukuda T, Kosaka T (2003) Ultrastructural study of gap junctions between dendrites of parvalbumin-containing GABAergic neurons in various neocortical areas of the adult rat. Neuroscience 120:5–20.[CrossRef][ISI][Medline]
Fukuda T, Galuske RA, Singer W, Kosaka T (2004) Topological analysis of GABAergic neurons in optically-recorded primary visual cortex. 27th Japan Neuroscience Society Annual Meeting Abstract P1-269.
Galarreta M, Hestrin S (1999) A network of fast-spiking cells in the neocortex connected by electrical synapses. Nature 402:72–75.[CrossRef][Medline]
Gibson JR, Beierlein M, Connors BW (1999) Two networks of electrically coupled inhibitory neurons in neocortex. Nature 402:75–79.[CrossRef][Medline]
Gupta A, Wang Y, Markram H (2000) Organizing principles for a diversity of GABAergic interneurons and synapses in the neocortex. Science 287:273–278.
Harris KM, Jensen FE, Tsao B (1992) Three-dimensional structure of dendritic spines and synapses in rat hippocampus (CA1) at postnatal day 15 and adult ages: implications for the maturation of synaptic physiology and long-term potentiation. J Neurosci 12:2685–2705.[Abstract]
Jin X, Mathers PH, Szabo G, Katarova Z, Agmon A (2001) Vertical bias in dendritic trees of non-pyramidal neocortical neurons expressing GAD67-GFP in vitro. Cereb Cortex 11:666–678.
Jin X, Hu H, Mathers PH, Agmon A (2003) Brain-derived neurotrophic factor mediates activity-dependent dendritic growth in nonpyramidal neocortical interneurons in developing organotypic cultures. J Neurosci 23:5662–5673.
Jones EG (1975) Varieties and distribution of non-pyramidal cells in the somatic sensory cortex of the squirrel monkey. J Comp Neurol 160:205–267.[CrossRef][ISI][Medline]
Jones EG (1984) Laminar distribution of output cells. In: Cerebral cortex. Vol. 1. Cellular