Abstract

The long-distance corticocortical connections between visual and nonvisual sensory areas that arise from pyramidal neurons located within layer V can be considered as a subpopulation of feedback connections. The purpose of the present study is to determine if layer V pyramidal neurons from visual and nonvisual sensory cortical areas that project onto the visual cortex (V1) constitute a homogeneous population of cells. Additionally, we ask whether dendritic arborization relates to the target, the sensory modality, the hierarchical level, or laterality of the source cortical area. Complete 3D reconstructions of dendritic arbors of retrogradely labeled layer V pyramidal neurons were performed for neurons of the primary auditory (A1) and somatosensory (S1) cortices and from the lateral (V2L) and medial (V2M) parts of the secondary visual cortices of both hemispheres. The morphological parameters extracted from these reconstructions were subjected to principal component analysis (PCA) and cluster analysis. The PCA showed that neurons are distributed within a continuous range of morphologies and do not form discrete groups. Nevertheless, the cluster analysis defines neuronal groups that share similar features. Each cortical area includes neurons belonging to several clusters. We suggest that layer V feedback connections within a single cortical area comprise several cell types.

Introduction

Layer V pyramidal neurons are recognizable by their basic morphology, which consists of a pyramid shaped soma, an apical dendrite extending towards the pial surface and a basal skirt (De Felipe and Farinas 1992). Layer V pyramidal neurons exhibit a range of dendritic morphologies, even within a single cortical area, as previously shown in the mouse (Tsiola et al. 2003). The dendritic arborization of these neurons appears to be related to their connections and by the cortical area to which they belong. For instance, in monkeys, it was shown that callosal projection neurons possess longer apical and basal dendrites and more dendritic spines than ipsilaterally projecting neurons (Soloway et al. 2002) and that the complexity of the dendritic tree increases with the hierarchical level of cortical areas of a same sensory modality (Elston and Rosa 2000). In the mouse, it was furthermore demonstrated that dendritic arbors are simpler in the visual cortex than in the somatosensory cortex (Benavides-Piccione et al. 2006; Groh et al. 2010).

The length of the apical dendrite is often used to classify layer V pyramidal neurons in 3 morphological groups: tall-tufted, tall-simple, and short (Larsen and Callaway 2006). These 3 subtypes of pyramidal neurons are associated with specific electrophysiological properties and connectivity. In the neocortex, tall-tufted pyramidal neurons, also known as thick, tufted or type I, are intrinsically bursting cells, whereas tall-simple and short neurons, also called slender or type II, are mostly regular spiking (Chagnac-Amitai et al. 1990; Mason and Larkman 1990; Kasper et al. 1994; Hattox and Nelson 2007). While tall-tufted neurons are involved in corticofugal connections, tall-simple and short neurons participate in corticocortical connections (Schofield et al. 1987; Games and Winer 1988; Hallman et al. 1988; Hubener and Bolz 1988; Hubener et al. 1990; Kasper et al. 1994; Larsen et al. 2007).

Feedback corticocortical connections are not a homogeneous population (Rockland 2004). Although, they mostly arise from infragranular layers, some neurons from supragranular layers are also involved (Rockland and Pandya 1979; Felleman and Van Essen 1991). The relative contribution of supragranular and infragranular neurons to corticocortical connections is related to the hierarchical distance between 2 connected cortical areas (Barone et al. 2000; Vezoli et al. 2004; Reid et al. 2009). For 2 equivalent hierarchical levels, retrogradely labeled neurons will be equally distributed in infragranular and supragranular layers. When neurons of origin are almost exclusively restricted to infragranular layers, the cortical area of origin has been assigned to a higher hierarchical level than the other one (Felleman and Van Essen 1991). Furthermore, layer V appears to strongly contribute to long-distance feedback connections in the mouse, although many feedback neurons are also located within layer VI (Bai et al. 2004) (Charbonneau V, Laramée ME, Boucher V, Bronchti G, Boire D, unpublished data). Finally, because several subtypes of layer V pyramidal neurons have been described (Molnar and Cheung 2006), one can predict the existence of several types of feedback connections originating from this layer. So far, the existence of subtypes of layer V pyramidal neurons has not been taken into account to study the organization of feedback corticocortical connections.

As mentioned above, the complexity of the apical dendrite has been the main morphological criterion to classify layer V pyramidal neurons. With this approach, 2 main groups have been described: 1) tall-tufted and 2) tall-simple and short. However, this parameter alone does not account for the structural complexity of the entire dendritic arborization. The developmental pattern of layer V pyramidal neurons strongly suggests that there are likely 4 separate functional dendritic compartments namely the basal dendrite, apical trunk, oblique dendrite, and tuft dendrite (Romand et al. 2011). Therefore, the use of several dendritic morphological parameters, pertaining to the apical, oblique and basal dendrites, may provide a more comprehensive analysis and promote successful classification of layer V pyramidal neurons. Mathematical tools that ordinate objects in multidimensional space, such as the principal component analysis (PCA) and that objectively group similar objects using multiple descriptors such as cluster analysis, are appropriate for this approach. These analyses have previously been used to demonstrate the existence of distinct subgroups of layer II/III (Benavides-Piccione et al. 2006), V (Tsiola et al. 2003), and VI (Chen et al. 2009) neurons in the mouse cerebral cortex.

In this study, we document the organization of the subgroup of corticocortical feedback connections in layer V, using as comparison the morphological properties of commissural and noncommissural layer V pyramidal neurons that project onto the primary visual cortex (V1) from the primary auditory (A1) and somatosensory (S1) cortices and from the lateral (V2L) and medial (V2M) parts of the secondary visual cortex. If sensory modality, hierarchy, and laterality influence the morphology of the dendritic arborization, we would predict several subgroups of pyramidal neurons, with the implication of several types of feedback connections. If, however, determining factors for the dendritic arborization are associated with specific connectivity and functional properties, layer V pyramidal neurons that project onto V1 are more likely to be morphologically similar.

Materials and Methods

All experiments were carried out in level 2 biosafety facilities in accordance with National Institutes of Health (NIH) Guidelines for Research Involving Recombinant DNA Molecules and Guidelines for the Care and Use of Laboratory Animals (NIH Publication No. 80-23).

Injections and Tissue Processing

To study the morphology of the entire dendritic arbors of individual layer V pyramidal neurons in A1, S1, V2M and V2L that project onto V1, we injected an adenovirus that expresses enhanced green fluorescence protein (EGFP) under a synapsin promoter (AdSynEGFP) in V1 in C57BL/6 mice. With this technique, a Golgi-like retrograde labeling of complete dendritic arbors was achieved (Tomioka and Rockland 2006; Fuentealba et al. 2008; Ichinohe et al. 2008; Laramee et al. 2011; Papp et al. 2012).

Seven C57BL/6 mice (90–120 days) were used for this study. They were anesthetized using a solution of chloral hydrate (400mg/kg body weight). When a surgical level of anesthesia was achieved, a small opening in the skull and dura was performed and 0.5 μL of AdSynEGFP (1.5 × 1012 pfu/mL) was pressure injected in V1 (3.8-mm posterior and 2.6-mm lateral to Bregma) at a depth of 600 μm. All injections were performed at a rate of 0.06 μL/min through a 30-μm tip glass pipette glued onto a Hamilton syringe. After the surgery, animals were returned to their nest for recovery. Fourteen days later, they were anesthetized with urethane (1.5 mg/kg body weight) and perfused transcardially with 4% paraformaldehyde (PFA) in 0.1 M phosphate buffer (PB) at pH 7.4. The brains were harvested, postfixed in 4% PFA solution for 2–3 h and placed for 24 h in a 30% sucrose solution for cryoprotection. The next day, they were frozen and cut in 50 μm coronal serial sections with a freezing microtome. Sections were rinsed in phosphate buffered saline (PBS) at pH 7.4 and blocked in a 5% normal goat serum with 0.5% Triton-X-100 and 0.1 M PBS (PBS-TX) for 2 h. They were afterward placed in the primary antibody solution of 0.5 μg/mL of rabbit anti-GFP antibody (Tomioka and Rockland 2006), overnight at room temperature. After a wash in PBS, sections were incubated in the biotinylated goat anti-rabbit secondary antibody solution (Vector Labs; 1:200) for 2 h. After washes in PB, they were placed in an ABC solution (Vector Labs; 1:200) for another 2 h and then reacted with 3-3′-diaminobenzidine and intensified with Nickel-ammonium sulfate. Finally, sections were washed in PB and mounted on gelatinized slides, dehydrated, and counterstained with cresyl violet before being coverslipped.

Validation of Injection Sites

All injection sites in V1 were validated with reference to cortical areal cytoarchitecture (Caviness 1975). AdSynEGFP injections were required to be limited to V1 and to label all cortical layers without encroaching on to the subcortical white matter (Fig. 1A). Retrograde labeling in the lateral geniculate nucleus was also required to validate the localization of injection sites. Five of the 7 injected mice met these criteria and were used for this study.

Figure 1.

AdSynEGFP injections and labeling. (A) Injection site in V1 (asterisk). AdSynEGFP injections covered all cortical layers without encroaching on the subcortical white matter. Note the column of retrograde labeling in V2L. Nissl counter-staining was used to delineate areal borders. (B) Retrograde labeling of a callosal layer V pyramidal neuron from A1. (C) High magnification of rectangle in (B) to show the high resolution labeling of dendrites and dendritic spines. Scale bars: 1 cm in A, 50 μm in B and 10 μm in C.

Figure 1.

AdSynEGFP injections and labeling. (A) Injection site in V1 (asterisk). AdSynEGFP injections covered all cortical layers without encroaching on the subcortical white matter. Note the column of retrograde labeling in V2L. Nissl counter-staining was used to delineate areal borders. (B) Retrograde labeling of a callosal layer V pyramidal neuron from A1. (C) High magnification of rectangle in (B) to show the high resolution labeling of dendrites and dendritic spines. Scale bars: 1 cm in A, 50 μm in B and 10 μm in C.

Selection of Neurons

A total of 63 neurons were selected from the 5 mice with validated AdSynEGFP injections in V1 (Table 1). Layer V pyramidal neurons were selected for further analysis, if they were sufficiently well isolated to allow complete 3D reconstruction of the dendritic arbor. Localization of pyramidal cells was based on cytoarchitectonic features of the cortical area to which they belong (Caviness 1975). The number of neurons reconstructed in each mouse is depicted in Table 2.

Table 1

A total of 63 layer V pyramidal neurons have been reconstructed from A1, S1, V2M, V2Lant, and V2Lpost, of both hemispheres

Cortex Ipsilateral Contralateral Total 
A1 10 19 
S1 10 
V2M 10 
V2L ant 12 
V2L post 12 
Total 32 31 63 
Cortex Ipsilateral Contralateral Total 
A1 10 19 
S1 10 
V2M 10 
V2L ant 12 
V2L post 12 
Total 32 31 63 
Table 2

Presentation of the number of neurons, for each mouse, that were reconstructed for each cortical area

 Mouse number 
Cortex 
A1 
S1 
V2M 
V2Lant 
V2Lpost 
Total 22 15 10 
 Mouse number 
Cortex 
A1 
S1 
V2M 
V2Lant 
V2Lpost 
Total 22 15 10 

Three-Dimensional Reconstruction

All neurons were reconstructed under a 100X objective (oil immersion objective, 1.4 NA) using an Olympus BX51W1 microscope coupled to CCD digital camera and customized with the Neurolucida software (MicroBrightField Bioscience). All neurons were reconstructed from a complete set of serial sections. To ensure completeness of reconstructions, all dendrites were required to be continued onto adjacent sections until a round ending was found or until it was impossible to find the continuation of the branch onto the next section. Because of the tissue processing, all sections shrinked to an average thickness of 20 μm, which correspond to a reduction of 2.5 times of the original section thickness. Therefore, after the dendritic arborization was completely reconstructed, a shrinkage correction of 2.5 in the z-plane was applied to the tracing using the shrinkage correction tool from the Neurolucida software. No correction was applied for the x and y planes. The analyzed morphological parameters (25) were all available in Neurolucida Explorer (MicroBrightField Bioscience) (Table 3).

Table 3

Twenty-five morphological parameters analyzed for each reconstructed neuron

  PC1 PC2 PC3 PC4 PC5 PC6 
 Eigenvalues  7.56 5.14 3.70 2.18 1.84 1.13 
 % variance 30.25 20.55 14.79 8.73 7.35 4.53 
Soma Area 0.67 −0.12 −0.29 0.10 −0.23 0.12 
Basal dendrites Quantity 0.06 −0.49 −0.21 0.59 −0.33 0.07 
Number of endings 0.24 −0.04 −0.13 0.84 0.03 −0.32 
Number of spines 0.71 −0.29 −0.31 0.12 0.42 −0.21 
Total length 0.62 −0.33 −0.28 0.38 0.41 −0.22 
Volume 0.73 −0.42 −0.45 0.08 0.14 0.01 
Mean terminal distance from the cell body 0.53 −0.26 0.12 −0.22 0.61 0.10 
Diameter 0.61 −0.43 −0.49 −0.22 −0.08 0.18 
Apical trunk Number of endings 0.32 0.86 0.03 0.24 −0.11 −0.05 
Number of spines 0.72 0.40 −0.11 −0.23 0.22 −0.21 
Total length 0.57 0.74 0.06 −0.02 0.06 0.02 
Volume 0.82 −0.09 −0.13 0.11 −0.37 0.20 
Mean terminal distance from the cell body 0.36 0.07 0.57 0.42 0.17 0.46 
Diameter 0.50 −0.49 −0.42 −0.22 −0.37 0.26 
Distance of the bifurcation 0.18 −0.14 0.58 0.47 0.22 0.43 
Apical tuft Number of endings 0.34 0.85 0.04 0.25 −0.07 −0.09 
Number of spines 0.54 0.69 −0.14 −0.23 0.10 0.02 
Total length 0.48 0.78 −0.02 −0.04 0.06 0.14 
Volume 0.67 0.52 −0.22 −0.14 −0.16 0.24 
Oblique dendrites Quantity 0.39 −0.26 0.71 −0.14 −0.22 −0.02 
Number of endings 0.36 −0.04 0.77 0.07 −0.34 −0.25 
Number of spines 0.71 −0.32 0.42 −0.21 −0.03 −0.27 
Total length 0.61 −0.24 0.64 −0.11 −0.18 −0.24 
Volume 0.84 −0.32 0.16 −0.17 −0.25 −0.09 
Distribution along the apical trunk 0.15 −0.34 0.50 −0.22 0.47 0.20 
  PC1 PC2 PC3 PC4 PC5 PC6 
 Eigenvalues  7.56 5.14 3.70 2.18 1.84 1.13 
 % variance 30.25 20.55 14.79 8.73 7.35 4.53 
Soma Area 0.67 −0.12 −0.29 0.10 −0.23 0.12 
Basal dendrites Quantity 0.06 −0.49 −0.21 0.59 −0.33 0.07 
Number of endings 0.24 −0.04 −0.13 0.84 0.03 −0.32 
Number of spines 0.71 −0.29 −0.31 0.12 0.42 −0.21 
Total length 0.62 −0.33 −0.28 0.38 0.41 −0.22 
Volume 0.73 −0.42 −0.45 0.08 0.14 0.01 
Mean terminal distance from the cell body 0.53 −0.26 0.12 −0.22 0.61 0.10 
Diameter 0.61 −0.43 −0.49 −0.22 −0.08 0.18 
Apical trunk Number of endings 0.32 0.86 0.03 0.24 −0.11 −0.05 
Number of spines 0.72 0.40 −0.11 −0.23 0.22 −0.21 
Total length 0.57 0.74 0.06 −0.02 0.06 0.02 
Volume 0.82 −0.09 −0.13 0.11 −0.37 0.20 
Mean terminal distance from the cell body 0.36 0.07 0.57 0.42 0.17 0.46 
Diameter 0.50 −0.49 −0.42 −0.22 −0.37 0.26 
Distance of the bifurcation 0.18 −0.14 0.58 0.47 0.22 0.43 
Apical tuft Number of endings 0.34 0.85 0.04 0.25 −0.07 −0.09 
Number of spines 0.54 0.69 −0.14 −0.23 0.10 0.02 
Total length 0.48 0.78 −0.02 −0.04 0.06 0.14 
Volume 0.67 0.52 −0.22 −0.14 −0.16 0.24 
Oblique dendrites Quantity 0.39 −0.26 0.71 −0.14 −0.22 −0.02 
Number of endings 0.36 −0.04 0.77 0.07 −0.34 −0.25 
Number of spines 0.71 −0.32 0.42 −0.21 −0.03 −0.27 
Total length 0.61 −0.24 0.64 −0.11 −0.18 −0.24 
Volume 0.84 −0.32 0.16 −0.17 −0.25 −0.09 
Distribution along the apical trunk 0.15 −0.34 0.50 −0.22 0.47 0.20 

Note: Only those with a loading equal to or greater than 0.70 in one of the 6 first PCs were kept for subsequent analyses (in bold). Below each PC, the corresponding eigenvalue and percentage of variance are indicated.

Computer-Assisted Analysis

A total of 25 morphological parameters have been analyzed and used to compare 63 layer V pyramidal neurons from 10 different cortical areas. An objective computer-assisted method was required to compare all reconstructed neurons and to group them based on their most significant morphological features. Therefore, as in previous studies performed in the mouse (Tsiola et al. 2003; Benavides-Piccione et al. 2006; Chen et al. 2009), PCA and cluster analysis were used.

Principal Component Analysis

We were interested here in showing the dispersion of neurons as described by their entire dendroarchitecture. This can best be achieved by gathering a host of morphological parameters or descriptors on each object, in this case, neurons. Examining scatterplots of the dispersion of these objects with respect to all possible pairs of descriptors is neither efficient nor informative of the whole body of available data. Therefore, PCA was used to visualize the distribution of cells in multivariate space. In order to decide how many components should be maintained for further analysis, we chose the Kaiser–Gutman criterion that states that one should interpret only those components with an eigenvalue larger than the mean eigenvalues. When PCA is performed on the correlation matrix, the average of the eigenvalues being 1, only the components whose eigenvalues are larger than 1 should be interpreted (Legendre P and Legendre L 1998). Therefore, from all available principal components (PCs), only those with a loading equal to or greater than 1 were selected for further analysis. A PCA is more influenced by parameters that have high absolute values, such as the apical dendrite length and less so by small absolute values, such as the number of oblique dendrites. Therefore, each morphological parameter was normalized using logarithmic or square root transformations, depending on which of these resulted in the best Gaussian distribution. The choice of the significant PCs was confirmed with the screeplot of all PCs, also available with the PCA. These significant PCs were afterward used to produce a scatterplot matrix, to observe the distribution of each neuron in the multidimensional space. Finally, using the absolute loadings of each variable for every PCs, the morphological parameters that had a loading equal to or greater than 0.70 in at least one PC were selected for further analysis (in bold in Table 3).

Cluster Analysis

This technique groups neurons that are sufficiently similar to each other in a multidimensional space, according to the selected clustering method. Here, Ward’s method with squared Euclidean distance was used to classify neurons so that each group has a minimal variance. Following this analysis, the frequency distribution of cell types was compared between cortical areas in order to determine whether each cortical area comprises similar populations of neurons using a Chi-square analysis (SPSS version 16.0).

Sholl Analysis

The Sholl analysis provides a synthetic representation of the distance of dendrites and spines, from the soma. The Neurolucida Explorer (MicroBrightField Bioscience) software was used to perform this analysis. The distance between each radius was set to 30 μm, and the starting point was located at the cell body. Apical and basal dendrites were analyzed separately. The apical dendrite includes the apical trunk, oblique dendrites, and the apical tuft. The total dendritic length and spine number were directly obtained from the analysis, and spine density was calculated by dividing the number of spines per 30 μm radius over the total dendritic length within that radius. The spine density was expressed as the number of spines per 10 μm.

Statistical Analysis

A one-way analysis of variance (ANOVA) was used to compare groups obtained with the cluster analysis. This was performed in an attempt to understand the classification of the neuronal population on the basis of their dendritic arbors. Although it could be considered inappropriate to perform between-group ANOVA comparisons because the same sum of squares are calculated in the cluster analysis (Chen et al. 2009), this analysis was performed only to identify which morphological parameters significantly differ between the cluster-defined groups. A one-way ANOVA was also used to compare the distribution of dendritic spines between commissural and noncommisural neurons. In order to have more stringent statistical conclusions, all statistical analyses were performed with a significance level of P < 0.01 using SPSS 16.0 software for Windows.

Results

Injection Site and Retrograde Labeling

To be considered for further analysis, AdSynEGFP injections in V1 had to be large enough to label all cortical layers without encroaching on the subcortical white matter and be limited to V1 (Fig. 1A). From the 7 mice that received injections, 5 met these criteria; the injections were located at the V2M/V1 border in the other 2 mice. AdSynEGFP injections resulted in a high resolution Golgi-like labeling of the dendritic arbors (Fig. 1B) and dendritic spines (Fig. 1C), confirming the efficacy of this viral tracer for complete 3D reconstructions (Tomioka and Rockland 2006; Fuentealba et al. 2008; Ichinohe et al. 2008; Laramee et al. 2011; Papp et al. 2012) and quantification of spines (Ichinohe et al. 2008; Laramee et al. 2011).

Retrogradely labeled neurons were mainly located in infra-granular layers, mostly layer V, and only neurons from this layer were reconstructed. In all 5 mice, retrogradely labeled neurons were found in many sensory and nonsensory cortical areas, such as medial and lateral secondary visual areas, primary and secondary auditory cortices, primary and secondary somatosensory cortices, motor areas, frontal areas, cingulate cortex, retrosplenial cortex, temporal, and posterior parietal associative areas. This pattern of retrograde labeling after injection in V1 is consistent with previous studies in rat (Miller and Vogt 1984) and mice (Charbonneau V, Laramée ME, Boucher V, Bronchti G, Boire D, unpublished data). Furthermore, 2 clusters of neurons were observed within V2L: one in the anterior part and one in the posterior part of this cortical area (Fig. 2). These neurons were therefore analyzed separately, as V2Lant and V2Lpost, because they might represent 2 different extrastriate visual areas, namely the anterolateral (AL) and lateromedial (LM) areas (Olavarria and Montero 1989; Coogan and Burkhalter 1993; Wang and Burkhalter 2007) that are involved in distinct sensory processes (Wang et al. 2011).

Figure 2.

Schematic representation of the localization of injection site into V1 and retrograde labeling within V2L. To visualize the AdSynEGFP injection site in V1 (black spot), the borders were traced on subsequent coronal sections and represented in a dorsal projection view of the posterior neocortex in 3 representative cases (A, B, and C). They were all centrally located. In the 5 mice used in this study, retrogradely labeled neurons in V2L were grouped into 2 patches (circles in V2L) identified as V2Lant and V2Lpost. (D) Identification of V1, V2M, and V2L onto the canvas that has been used in A, B, and C. Bottom right, identification of the axes: a, anterior and m, medial.

Figure 2.

Schematic representation of the localization of injection site into V1 and retrograde labeling within V2L. To visualize the AdSynEGFP injection site in V1 (black spot), the borders were traced on subsequent coronal sections and represented in a dorsal projection view of the posterior neocortex in 3 representative cases (A, B, and C). They were all centrally located. In the 5 mice used in this study, retrogradely labeled neurons in V2L were grouped into 2 patches (circles in V2L) identified as V2Lant and V2Lpost. (D) Identification of V1, V2M, and V2L onto the canvas that has been used in A, B, and C. Bottom right, identification of the axes: a, anterior and m, medial.

Various morphologies were qualitatively observed. Some neurons had a highly arborized apical tuft within layer I, whereas other had a slim tuft that only reached layers II/III (see Fig. 4, cells #43 vs. #28). There were neurons with oblique dendrites distributed along the length of the apical trunk (see Fig. 4, cell #29 and #34), whereas others had highly ramified oblique dendrites originating from its base (see Fig. 4, cell #43 and Fig. 5). Basal dendrites were, for the most part, restricted to layer V, but they were highly arborized in some neurons and very sparse in others (see Fig. 4, cells #43 vs. #21). Because of this diverse range of morphologies, it was almost impossible to compare the neurons from V2Lant, V2Lpost, V2M, A1 and S1 that project onto V1 based on qualitative observations only. It was thus necessary to use unbiased quantitative measurements, such as the length of the apical dendrite, PCA, cluster analysis, and Sholl analysis, to perform an objective classification.

Figure 4.

Reconstructed neurons. Of the 63 reconstructed neurons, 20 are shown here. Each neurons is identified with a cell number that correspond to the one found in the cluster analysis (Fig. 6) and to its score in the first principal component (PC1). The values ranged from −1.80 to 2.40, and cells are shown for approximate 0.21 increments. The factors most contributing to the first PC loading are the volume of the oblique dendrites (0.839), of the apical dendrite (0.820), and of the basal dendrites (0.730). Note the range of neuronal morphologies that can be observed. Scale bar: 100 μm.

Figure 4.

Reconstructed neurons. Of the 63 reconstructed neurons, 20 are shown here. Each neurons is identified with a cell number that correspond to the one found in the cluster analysis (Fig. 6) and to its score in the first principal component (PC1). The values ranged from −1.80 to 2.40, and cells are shown for approximate 0.21 increments. The factors most contributing to the first PC loading are the volume of the oblique dendrites (0.839), of the apical dendrite (0.820), and of the basal dendrites (0.730). Note the range of neuronal morphologies that can be observed. Scale bar: 100 μm.

Figure 5.

Localization of oblique dendrites. Oblique dendrites (in red) are not evenly distributed along the apical trunk. Instead, they arise from the first microns of the apical trunk, are mainly arborized within layer V and are intermingled with the basal dendrites. This is cell 43 from Figure 4. Scale bar: 100 μm.

Figure 5.

Localization of oblique dendrites. Oblique dendrites (in red) are not evenly distributed along the apical trunk. Instead, they arise from the first microns of the apical trunk, are mainly arborized within layer V and are intermingled with the basal dendrites. This is cell 43 from Figure 4. Scale bar: 100 μm.

Morphological Types

As in Larsen and Callaway (2006), the total length of the apical dendrite, excluding the oblique dendrites, was used to classify neurons as short (less than 1.5 mm), tall-simple (1.5 mm–3.3 mm), and tall-tufted (more than 3.3 mm). Using these classification criteria, only tall-simple (12.70%) and short (87.30%) neurons were found to participate in these feedback corticocortical connections onto V1.

Principal Component Analysis

A total of 25 PCs were extracted. Among them, the first 6 had an eigenvalue equal to or greater than 1 and accounted for 86.20% of the total variance (Table 3). A screeplot was also used to confirm that, after the sixth PC, a plateau was reached, meaning that the remaining PCs did not significantly add to the variance. The loading of each variable was obtained for these 6 PCs and only the ones with an absolute value equal to or greater than 0.70 were kept for all subsequent analyses (in bold in Table 3). These 13 parameters are the ones that accounted for the highest degree of variance in the population of reconstructed neurons and can be considered the most suitable as a basis for classification.

A scatterplot matrix was extracted to represent all neurons in a 3D space using these parameters and the first 3 components, which account for 65.59% of the total variance (Fig. 3). No isolated clusters of neurons were observed, and all neurons were intermingled regardless of the hemisphere, the sensory modality or the hierarchical rank of the cortical area they belong to. The first PC accounts for 30.25% of the total variance and that the second and third PCs account for 20.55% and 14.79%, respectively.

Figure 3.

PCA. (AC) representation of all reconstructed neurons color coded for each cluster in the 3 firsts PCs, which accounted for 65.59% of the total variance of the population. (D) Three-dimensional representation of the 3 firsts PCs. No isolated clusters can be found. Instead, the distribution appears stretched out with respect to the first PC.

Figure 3.

PCA. (AC) representation of all reconstructed neurons color coded for each cluster in the 3 firsts PCs, which accounted for 65.59% of the total variance of the population. (D) Three-dimensional representation of the 3 firsts PCs. No isolated clusters can be found. Instead, the distribution appears stretched out with respect to the first PC.

In an attempt to evaluate the diversity of morphologies found in the population, neurons were represented in the order of their increasing score in the first PC (Fig. 4). The factors most contributing to the first PC are the volumes of the oblique dendrites (0.84), of the apical dendrite (0.82), and of the basal dendrites (0.73). This is evident when the first (cell #21) and last (cell #43) neurons are compared. The first is devoid of oblique dendrites, has a tiny apical tuft and very few basal dendrites. The last one, however, has many oblique dendrites that originate from the base of the apical trunk and arborize extensively within layer V, in an area mostly occupied by basal dendrites (Fig. 5). This neuron also possesses an apical tuft and basal dendrites that are highly ramified. In agreement with the contributing factors of the first PC, this classification of the neuronal population displays the increase in the complexity of the oblique, basal, and apical dendrites.

Cluster Analysis

A cluster analysis based on the Ward’s method, which forms groups that have the smallest degree of variance, was performed (Fig. 6). Six groups were obtained. The first partition of neurons appears to be determined by significant differences in the number of dendritic spines onto the basal dendrites (P < 0.001), the volume of the basal dendrites (P < 0.001), the number of dendritic spines onto the apical trunk (P < 0.001), the total length of the apical trunk (P = 0.010), the volume of the apical trunk (P < 0.001), the number of dendritic spines onto the oblique dendrites (P < 0.001), and the volume of the oblique dendrites (P < 0.001). All these parameters were significantly higher in the second subdivision that comprises groups 5 and 6.

Figure 6.

Cluster analysis. Using the 13 parameters that remain after the PCA (see Table 1), the cluster analysis suggests that there are 6 groups (1–6). None of these groups are representative of a hemisphere, a sensory modality, or a hierarchical rank. Each neuron is identified by the cortical area to which it belongs and by its cell number.

Figure 6.

Cluster analysis. Using the 13 parameters that remain after the PCA (see Table 1), the cluster analysis suggests that there are 6 groups (1–6). None of these groups are representative of a hemisphere, a sensory modality, or a hierarchical rank. Each neuron is identified by the cortical area to which it belongs and by its cell number.

The next subdivision resulted in 3 groups: 1–2, 3–4, 5–6. When these groups were compared, only morphological parameters involving the basal dendrites revealed significant differences. Indeed, groups 1–2 and 3–4 differed by the number of endings (P = 0.009), the number of spines (P < 0.001), and the volume of these dendrites (P < 0.001).

The third and last subdivision resulted in the 6 individual groups. Groups 1 and 2 were highly heterogeneous and no categorization could be found. Groups 3 and 4, however, were distinguished by the number of endings (P < 0.001) and spines (P = 0.001) onto their apical trunk, the total length of the apical trunk (P < 0.001), the number of endings onto the apical tuft (P = 0.001) and the length of the apical tuft (P < 0.001). Groups 5 and 6 differed by the number of spines (P < 0.001).

The different clusters that were obtained do not partition in any way that would correspond to the cortical area in which the neurons are located. Groups 1 and 2 were highly heterogeneous in this respect. This was not the case for the comparison of groups 3 and 4 and 5 and 6. Indeed, in group 4, neurons belonged mainly to primary sensory areas (5/7, 71.43%), whereas group 3 was equally represented by neurons from primary (5/12, 41.67%) and secondary cortical areas. Within groups 5–6, callosal neurons were frequent (16/22, 72.72%), whereas they only account for 36.58% (15/41) of the population from groups 1–2–3–4. Moreover, groups 5 and 6 were differently represented in term of neurons from primary and secondary cortical areas: only 55.56% (5/9) of the neurons from group 5 were from a secondary visual area, whereas 76.92% (10/13) of the neurons found in group 6 came from V2L or V2M.

When the frequency distribution of neurons of the clusters 1–2, 3–4, and 5–6 between cortical areas are compared (Table 4) it is seen that almost all cortical areas include neurons belonging to each cluster. However, neurons from cluster 5–6 are absent in the ipsilateral V2Lant and V2Lpost, whereas these areas contain cells from clusters 1–2 and 3–4. The contralateral V2Lpost comprises only cells from clusters 5–6 and V2M also comprises a greater proportion of neurons from these clusters. Indeed, there is a significant difference in the frequency distribution of cells of clusters 1–2, 3–4, and 5–6 across cortical areas (χ2 = 31.84, P = 0.02, degrees of freedom [df] = 18). This statistical difference is no longer present when the neurons from ipsilateral V2Lant are removed from the analysis (χ2 = 23.77, P = 0.09, df = 16). Similarly, this difference did not reach levels of significance when neurons from the contralateral (χ2 = 20.31, P = 0.21, df = 16) or ipsilateral (χ2 = 25.90, P = 0.06, df = 16) V2Lpost were removed from the analysis. This demonstrates 1) that cortical areas contain subsets of pyramidal neurons in layer V that project to the primary visual cortex that have significantly different dendritic morphologies and 2) that each cortical area can contain different populations of these neurons.

Table 4

Number and percentage (in parenthesis) of neurons from each Cortical area with respect to their classification as obtained with the cluster analysis

Hemisphere Cortical area Group 1–2 Group 3–4 Group 5–6 Total 
Ipsilateral A1 3 (30.0) 4 (40.0) 3 (30.0) 10 
S1 2 (40.0) 2 (40.0) 1 (20.0) 
V2M 1 (20.0) 2 (40.0) 2 (40.0) 
V2L ant 5 (83.3) 1 (16.7) 0 (0.0) 
V2L post 2 (33.3) 4 (67.7)  0 (0.0) 
Contra lateral A1 5 (55.5) 2 (22.2) 2 (22.2) 
S1 2 (40.0) 2 (40.0) 1 (20.0) 
V2M 1 (20.0)  0 (0.0) 4 (80.0) 
V2L ant 1 (16.7) 2 (33.4) 3 (50.0) 
V2L post  0 (0.0)  0 (0.0) 6 (100.0) 
Total  22 (34.9) 19 (30.1) 22 (34.9) 63 
Hemisphere Cortical area Group 1–2 Group 3–4 Group 5–6 Total 
Ipsilateral A1 3 (30.0) 4 (40.0) 3 (30.0) 10 
S1 2 (40.0) 2 (40.0) 1 (20.0) 
V2M 1 (20.0) 2 (40.0) 2 (40.0) 
V2L ant 5 (83.3) 1 (16.7) 0 (0.0) 
V2L post 2 (33.3) 4 (67.7)  0 (0.0) 
Contra lateral A1 5 (55.5) 2 (22.2) 2 (22.2) 
S1 2 (40.0) 2 (40.0) 1 (20.0) 
V2M 1 (20.0)  0 (0.0) 4 (80.0) 
V2L ant 1 (16.7) 2 (33.4) 3 (50.0) 
V2L post  0 (0.0)  0 (0.0) 6 (100.0) 
Total  22 (34.9) 19 (30.1) 22 (34.9) 63 

Note: Neurons from a given area were not restricted to one group of the cluster.

Sholl Analysis

The spatial organization of the dendritic arbors was also compared between commissural and noncommissural projecting neurons for all cortical areas. The total dendritic length, spine number, and spine density were obtained for basal (Fig. 7) and apical (Supplementary Fig. 1) dendrites, using the Sholl analysis. For almost all neurons, the basal dendrites are contained in 210-μm spheres. In V2Lpost, however, there are 2 neurons that possess long basal dendrites that extend as far as 450 μm from the cell body. In all cortical areas, neurons have the highest total dendritic length, ranging from 588.7 to 1905.8 μm (both in contralateral S1), around the 90-μm bin, and there are neurons with short and long dendrites, which suggest heterogeneity of basal dendritic arbors. The same is also true for the spine number. Some neurons possess up to 617 spines (contralateral S1), whereas others have as few as 60 spines (ipsilateral A1). A high heterogeneity was also found when the spine density was compared. In all cortical areas, the spine density ranges between 1 and 4 spines/10 μm. Finally, when the same analysis is performed over the apical dendrite, several apical morphologies are also found (see Supplementary Fig. 1).

Figure 7.

Sholl analysis of basal dendrites. The dendritic length (left row), number of spines (middle row), and spine density (right row), as obtained with the Sholl analysis, were compared between commissural (gray line) and noncommisural (black line) neurons. Radius size increment is 30 μm for all parameters and spine density is measured as number of spines per 10 μm of dendritic length within each 30 μm radius. All y-axis are set to equivalent values, for each row, to compare areas. Neurons projecting onto V1 are characterized by short or long dendrites with few or many spines. This morphological heterogeneity is observed in all cortical areas.

Figure 7.

Sholl analysis of basal dendrites. The dendritic length (left row), number of spines (middle row), and spine density (right row), as obtained with the Sholl analysis, were compared between commissural (gray line) and noncommisural (black line) neurons. Radius size increment is 30 μm for all parameters and spine density is measured as number of spines per 10 μm of dendritic length within each 30 μm radius. All y-axis are set to equivalent values, for each row, to compare areas. Neurons projecting onto V1 are characterized by short or long dendrites with few or many spines. This morphological heterogeneity is observed in all cortical areas.

Spine Distribution

The laminar distribution of spines was also compared between ipsilaterally and contralaterally projecting neurons because it will affect the type of inputs a neuron receives. Their distribution onto the apical trunk, apical tuft, oblique dendrites, and basal dendrites was analyzed with respect to cortical layers (Fig. 8). Only neurons in the contralateral V2Lant had a greater number of dendritic spines, distributed on basal dendrites within layer V (P = 0.006).

Figure 8.

Spine distribution. For all dendritic compartments, the number of spines per cortical layer is compared between noncommissural (straight line) and commissural neurons (dashed line). Error bars: SEM. **P < 0.01, *P < 0.05.

Figure 8.

Spine distribution. For all dendritic compartments, the number of spines per cortical layer is compared between noncommissural (straight line) and commissural neurons (dashed line). Error bars: SEM. **P < 0.01, *P < 0.05.

Discussion

In 3 mice, all retrogradely labeled neurons were charted in cortical areas V2L, V2M, A1 and S1, and other nonsensory areas. Direct heteromodal corticocortical connections onto V1 have already been reported in macaque (Falchier et al. 2002; Rockland and Ojima 2003; Clavagnier et al. 2004; Borra and Rockland 2011), cat (Innocenti et al. 1988; Sanchez-Vives et al. 2006; Hall and Lomber 2008), mice (Larsen et al. 2009), and prairie vole (Campi et al. 2010). As in previous studies, the laminar distribution of retrogradely labeled neurons found here, namely in the supragranular and, more frequently, infra-granular layers, suggests that these connections are of the feedback type (Charbonneau V, Laramée ME, Boucher V, Bronchti G, Boire D, unpublished data). In primates, feedback corticocortical connections originate from supragranular and infragranular neurons with a clear predominance of infragranular neurons located in layer VI (Rockland and Pandya 1979; Felleman and Van Essen 1991; Falchier et al. 2002; Clavagnier et al. 2004; Rockland 2004). In rodents, a similar pattern is shown but infragranular neurons are located in both layer V and VI (Bai et al. 2004; Budinger et al. 2006). These projections neurons are numerous in both layers.

Layer V pyramidal Cell Subtypes

Larsen and Callaway (2006) used the morphology and length of the apical dendrite to classify mouse layer V pyramidal neurons in 3 neuronal groups: short neurons have apical dendrites shorter than 1.5 mm, tall-simple neurons have apical dendrites length ranging from 1.5 mm to 3.3 mm and tall-tufted neurons have apical dendrites longer than 3.3 mm. Using these criteria, only tall-simple and short neurons were shown to contribute to corticocortical connections onto V1. This is consistent with previous studies that used the length of the apical dendrite (Games and Winer 1988; Hallman et al. 1988; Hubener and Bolz 1988; Hubener et al. 1990; Kasper et al. 1994; Larsen et al. 2007) or the ratio of the total length of the apical dendrite over the total length of the basal dendrites (Groh et al. 2010) to classify pyramidal neurons. Tall-simple and short neurons are recognized as regular spiking (Chagnac-Amitai et al. 1990; Mason and Larkman 1990; Kasper et al. 1994; Hattox and Nelson 2007) and their projections are usually directed toward other cortical areas. In contrast, tall-tufted neurons send axon collaterals onto several subcortical structures, such as the thalamus and superior colliculus (Schofield et al. 1987; Games and Winer 1988; Hallman et al. 1988; Hubener and Bolz 1988; Hubener et al. 1990; Kasper et al. 1994; Larsen et al. 2007). All reconstructed neurons that project to V1 were tall-simple or short. Our results therefore suggest that feedback corticocortical connections onto V1 from several cortical areas are functionally similar, on the basis of the electrophysiological properties and connection specificity.

Most of the reconstructed neurons exhibited a slender apical tuft that extended throughout superficial layers in addition to several basal dendrites. These morphological properties indicate that the reconstructed neurons belong to the subgroup 1B of layer V pyramidal neurons (Tsiola et al. 2003). This is consistent with what has been found in V1 and S1, where neurons involved in corticocortical connections also belong to subgroup 1B (Groh et al. 2010). However, we cannot rule out the possibility that the neurons in layer V found to project to area V1 may further differentiate into several subtypes defined by different protein expression patterns (Molnar and Cheung 2006). The qualitative observations presented here further suggest that the structure and distribution of basal and oblique dendrites might also be relevant in assessing the diversity of layer V pyramidal neurons. This is consistent with the observation that layer V pyramidal cells that express the Kv3.1 potassium channel have more numerous oblique dendrites than those that do not express this receptor (Akemann et al. 2004).

Principal Component Analysis

The PCA was used because it allows for a multidimensional representation of the neuronal population based on the most relevant features of their dendritic arborization. To our knowledge, the representation of the distribution of layer V pyramidal neurons in a multidimensional space, using the PCA, has never been performed before. This technique has, however, been used to represent the distribution of layer II/III neurons from V2 and the secondary somatosensory (S2) and motor (M2) cortical areas in the mouse (Benavides-Piccione et al. 2006). The authors showed that neurons from each of these cortical areas form distinct groups. Also, the PCA was used to successfully sort granule cells, interneurons, CA1 pyramidal neurons, and CA3 pyramidal neurons in individual groups in the hippocampus of the rat (Cannon et al. 1999). Altogether, these results suggest that the PCA is a powerful technique to classify neurons using their most relevant morphological features.

Twenty-five morphological parameters were analyzed after complete 3D reconstructions of dendritic arbors. Among them, 13 significantly accounted for the variance in our neuronal population. The PCA-based distribution of neurons in multidimensional space resulted in an aggregated scatterplot slightly more elongated in the first PC that accounts for 30% of the total variance with only a few outlying cases. It is noteworthy that no isolated groups were found. When neurons are ordinated along their increasing score in the first PC, mainly influenced by the volume of the oblique, apical, and basal dendrites, it is apparent that the lowest scoring neurons were simple and the highest scoring ones more complex. Such a continuum could emerge from the developmental mechanisms that shape dendritic arbors of cortical neurons. For example, in layer IV, spiny stellate cells start as pyramidal cells; their apical dendrite shrinks and basal dendrites expand in their final morphology (Callaway and Borrell 2011). Early visual deprivation increases the number of pyramidal cells at the expense of stellate cells, indicating that the remodeling of dendrites is experience dependent (Callaway and Borrell 2011; and see Koester and O'Leary 1992 for callosal and corticotectal neurons, and Romand et al. 2011 for changes in somatosensory tall-tufted neurons).

The volume and the number of spines of the apical, basal, and oblique dendrites are the main factors involved in the differentiation of the neurons sampled here and appear in the first PC. In addition, contrary to layer IV spiny stellate neurons in which growth of apical and basal dendrites appears to be inversely correlated, in our sample, they appear to be strongly correlated as show by their positive loadings in the first PC (see Table 3). This pattern is similar to the development of tall-tufted layer V pyramidal neurons of the somatosensory cortex in that, very early in development, there is a rapid growth of all dendritic compartments. Later development is characterized by an increase in the number of basal dendritic segments (Romand et al. 2011). The factors with the highest loadings in the 3 first PCs are thus quite similar to those that are involved in the developmental shaping of dendritic arbors and reflect the strong influence of general growth of the dendritic arbor, multiplication, and pruning of endings and rapid increase in the number of spines.

Cluster Analysis

The cluster analysis is the most common technique to classify neuronal populations. Layer V pyramidal neurons (Tsiola et al. 2003) and neurons from layer VI (Chen et al. 2009) have been sorted in several groups and subgroups. Computer-generated neurons from each cortical layer were also separated in distinct groups (Heumann and Wittum 2009). In addition to the length and complexity of the apical dendrite, the present study includes oblique and basal dendrites in the analysis. This resulted in clusters that do not follow the tall-simple and short neuron dichotomy and may be evidence for finer layer V pyramidal classifications. Protein and gene expression were not considered here but undoubtedly can be useful in refining the classification of neurons (Molnar and Cheung 2006).

Antibodies against latexin have identified subpopulations of pyramidal neurons. In that a small subpopulation of the many retrogradely labeled layer V neurons were latexin+ (Bai et al. 2004). Similar results are shown in A2 and S2 following retrograde tracer injections in A1 and S1, respectively (Bai et al. 2004). This clearly demonstrates that feedback projecting pyramidal cells comprise at least 2 subpopulations, from the perspective of latexin expression. Whether this dichotomy corresponds to the tall-simple and short classification of neurons is not known. More importantly, the proportion of latexin + feedback projecting neurons is different between cortical areas. For layer VI, which has greater number of latexin + retrogradely labeled cells, the percentage of double-labeled neurons varies between 56% in V2L to 88% in S2. It is noteworthy that latexin does not appear to label a significant number of layer V feedforward projection neurons (Bai et al. 2004).

In addition to the PCA, we performed a cluster analysis to determine whether layer V pyramidal neurons that project onto V1 form groups that would correspond to cortical area, laterality, or sensory modality. Six groups were obtained; none of them was formed by neurons observing only one of these criteria. Even if the first division of the cluster resulted in 2 main groups differently represented by commissural and noncommissural neurons, these groups were not homogeneous. We thus cannot conclude that laterality, sensory modality, or cortical area is sufficient to classify layer V pyramidal neurons that project onto V1. What does appear is that in each cortical area, the population of neurons projecting to V1 comprises morphologically diverse neurons.

The basal dendrites of the reconstructed neurons exhibit a range of morphologies within each cortical area. To our knowledge, there are no studies demonstrating an increase in the complexity of the dendritic arbors from low to high order cortical areas in the mouse and our results also do not seem to support this idea. These results are at odds with observations in agouti (Elston et al. 2006), monkeys (Elston and Rosa 1997; Elston et al. 1999b) and humans (Jacobs et al. 1997, 2001), showing that the population of layer II/III pyramidal neurons is homogeneous with respect to the basal dendrite morphology within a given cortical area. This suggests that more complex neurons are still intermingled with simple neurons in mice, as the parcellation of their neocortex might not be as advanced as in primates (Krubitzer 1995).

It is possible that layer V pyramidal neurons that project onto V1 in the mouse are a more homogeneous subgroup of the overall neuronal population in this layer. In monkeys, the complexity of dendritic arbors of retrogradely labeled neurons within the visual system is correlated with the hierarchy of the cortical area (Elston and Rosa 2006). In cats, intermediate segments of neurons projecting onto visual areas from the V1/V2 border are significantly longer that those from neurons in V1 (Vercelli and Innocenti 1993). Here, reconstructed neurons are from visual, auditory, and somatosensory areas and are thus involved in heteromodal corticocortical connections. The correlation between complexity of dendritic arbors and hierarchy might hold true only within a single sensory modality and dramatically so in nonrodent species (Charbonneau V, Laramée ME, Boucher V, Bronchti G, Boire D, unpublished data). Thus, it might not be surprising that the reconstructed neurons cannot be grouped on the basis of the cortical area or hierarchy.

Our results are in agreement with other studies demonstrating evidence of heterogeneous populations of neurons projecting to a single target. For example, there are significant differences in gene expression between supra and infragranular layer callosal projection neurons, and there are combinations of genes that demonstrate subpopulations of callosal projection neurons within individual cortical layers. Furthermore, there is a greater molecular heterogeneity of supragranular than infragranular callosal projection neurons. The molecular dissection of these neuronal subpopulations reveals a greater diversity than was recognized using anatomical criteria (Molyneaux et al. 2009; Fame et al. 2011).

Spine Distribution

The contralateral V2Lpost comprises a greater proportion of neurons belonging to clusters 5–6. This cluster emerges at the first partition and differs from clusters 1–4 in several parameters. In particular, this subdivision of the cluster analysis appeared to be explained by the number of dendritic spines found on the basal dendrites, apical trunk, and oblique dendrites. Because the first group (clusters 1–4) was mainly represented by noncommissural neurons and the second group (cluster 5–6) by commissural neurons, the distribution of spines was compared for all cortical areas with respect to the hemisphere. Only neurons from V2Lant had significantly more dendritic spines onto the basal dendrites.

In V2L, neurons appeared to form 2 distinct patches, one anterior and one posterior, identified as V2Lant and V2Lpost, respectively. The V2Lant and V2Lpost patches seem to correspond to AL and LM extrastriate visual areas, respectively. These 2 areas are involved in distinct pathways: LM is part of the ventral stream and AL of the dorsal stream (Wang et al. 2011). In the marmoset, dendritic arbors of layer II/III pyramidal neurons from the ventral stream have significantly larger dendritic fields, higher spine number, and densities and larger cell bodies compared with those from the dorsal stream (Elston et al. 1999a). Here, neither the PCA nor the cluster analyses split V2Lant and V2Lpost neurons in distinct groups.

Conclusions

The aim of this study was to classify layer V pyramidal neurons involved in long-distance feedback connections onto V1 using the morphology of their dendritic arbors. Reconstructed neurons came from V2Lant, V2Lpost, V2M, A1, and S1 from both hemispheres. We found that they were all tall-simple or short neurons and that it was not possible to sort them in separate groups related to the hemisphere, sensory modality, or hierarchical level of the cortical area to which they belong. The target, in this case V1, could therefore be a major influence on their dendritic arborization. Including basal dendrites and oblique apical dendrites in the morphological analysis results in a classification of neurons beyond the dichotomous tall-simple and short categories proposed for corticocortical projecting pyramidal neurons. This is a more elaborate differentiation of dendrites of layer V pyramidal neurons that contribute to the feedback projection onto the primary visual cortex, a differentiation that likely emerges through experience-dependent shaping of the arborization during development.

Supplementary Material

Supplementary material can be found at: http://www.cercor.oxfordjournals.org/

Funding

National Science and Engineering Research Council (NSERC) of Canada grants to D.B. and G.B. and by a Canadian Foundation for Innovation grant to D.B. and funding from RIKEN Brain Science Organization to K.S.R. M.E.L. is supported by an Alexander Graham Bell (NSERC) fellowship and travel to Japan was supported by the Réseau de la Recherche en Santé de la Vision FRSQ.

We are grateful to Kazumi Ohta for technical expertise concerning viral tracer injections and to Marco Rodriguez for cluster and principal component analyses. Conflict of Interest : None declared.

References

Akemann
W
Zhong
YM
Ichinohe
N
Rockland
KS
Knopfel
T
Transgenic mice expressing a fluorescent in vivo label in a distinct subpopulation of neocortical layer 5 pyramidal cells
J Comp Neurol
 , 
2004
, vol. 
480
 (pg. 
72
-
88
)
Bai
WZ
Ishida
M
Arimatsu
Y
Chemically defined feedback connections from infragranular layers of sensory association cortices in the rat
Neuroscience
 , 
2004
, vol. 
123
 (pg. 
257
-
267
)
Barone
P
Batardiere
A
Knoblauch
K
Kennedy
H
Laminar distribution of neurons in extrastriate areas projecting to visual areas V1 and V4 correlates with the hierarchical rank and indicates the operation of a distance rule
J Neurosci
 , 
2000
, vol. 
20
 (pg. 
3263
-
3281
)
Benavides-Piccione
R
Hamzei-Sichani
F
Ballesteros-Yanez
I
Defelipe
J
Yuste
R
Dendritic size of pyramidal neurons differs among mouse cortical regions
Cereb Cortex
 , 
2006
, vol. 
16
 (pg. 
990
-
1001
)
Borra
E
Rockland
KS
Projections to early visual areas v1 and v2 in the calcarine fissure from parietal association areas in the macaque
Front Neuroanat
 , 
2011
, vol. 
5
 pg. 
35
 
Budinger
E
Heil
P
Hess
A
Scheich
H
Multisensory processing via early cortical stages: connections of the primary auditory cortical field with other sensory systems
Neuroscience
 , 
2006
, vol. 
143
 (pg. 
1065
-
1083
)
Callaway
EM
Borrell
V
Developmental sculpting of dendritic morphology of layer 4 neurons in visual cortex: influence of retinal input
J Neurosci
 , 
2011
, vol. 
31
 (pg. 
7456
-
7470
)
Campi
KL
Bales
KL
Grunewald
R
Krubitzer
L
Connections of auditory and visual cortex in the prairie vole (Microtus ochrogaster): evidence for multisensory processing in primary sensory areas
Cereb Cortex
 , 
2010
, vol. 
20
 (pg. 
89
-
108
)
Cannon
RC
Wheal
HV
Turner
DA
Dendrites of classes of hippocampal neurons differ in structural complexity and branching patterns
J Comp Neurol
 , 
1999
, vol. 
413
 (pg. 
619
-
633
)
Caviness
VS
Jr
Architectonic map of neocortex of the normal mouse
J Comp Neurol
 , 
1975
, vol. 
164
 (pg. 
247
-
263
)
Chagnac-Amitai
Y
Luhmann
HJ
Prince
DA
Burst generating and regular spiking layer 5 pyramidal neurons of rat neocortex have different morphological features
J Comp Neurol
 , 
1990
, vol. 
296
 (pg. 
598
-
613
)
Chen
CC
Abrams
S
Pinhas
A
Brumberg
JC
Morphological heterogeneity of layer VI neurons in mouse barrel cortex
J Comp Neurol
 , 
2009
, vol. 
512
 (pg. 
726
-
746
)
Clavagnier
S
Falchier
A
Kennedy
H
Long-distance feedback projections to area V1: implications for multisensory integration, spatial awareness, and visual consciousness
Cogn Affect Behav Neurosci
 , 
2004
, vol. 
4
 (pg. 
117
-
126
)
Coogan
TA
Burkhalter
A
Hierarchical organization of areas in rat visual cortex
J Neurosci
 , 
1993
, vol. 
13
 (pg. 
3749
-
3772
)
De Felipe
J
Farinas
I
The pyramidal neuron of the cerebral cortex—morphological and chemical characteristics of the synaptic inputs
Prog Neurobiol
 , 
1992
, vol. 
39
 (pg. 
563
-
607
)
Elston
GN
Elston
A
Aurelio-Freire
M
Gomes
LW
Dias
IA
Pereira
A
Jr
Silveira
LC
Picanco Diniz
CW
Specialization of pyramidal cell structure in the visual areas V1, V2 and V3 of the South American rodent, Dasyprocta primnolopha
Brain Res
 , 
2006
, vol. 
1106
 (pg. 
99
-
110
)
Elston
GN
Rosa
MG
The occipitoparietal pathway of the macaque monkey: comparison of pyramidal cell morphology in layer III of functionally related cortical visual areas
Cereb Cortex
 , 
1997
, vol. 
7
 (pg. 
432
-
452
)
Elston
GN
Rosa
MG
Pyramidal cells, patches, and cortical columns: a comparative study of infragranular neurons in TEO, TE, and the superior temporal polysensory area of the macaque monkey
J Neurosci
 , 
2000
, vol. 
20
 pg. 
RC117
 
Elston
GN
Rosa
MG
Ipsilateral corticocortical projections to the primary and middle temporal visual areas of the primate cerebral cortex: area-specific variations in the morphology of connectionally identified pyramidal cells
Eur J Neurosci
 , 
2006
, vol. 
23
 (pg. 
3337
-
3345
)
Elston
GN
Tweedale
R
Rosa
MG
Cellular heterogeneity in cerebral cortex: a study of the morphology of pyramidal neurones in visual areas of the marmoset monkey
J Comp Neurol
 , 
1999
, vol. 
415
 (pg. 
33
-
51
)
Elston
GN
Tweedale
R
Rosa
MG
Cortical integration in the visual system of the macaque monkey: large-scale morphological differences in the pyramidal neurons in the occipital, parietal and temporal lobes
Proc R Soc B Biol Sci
 , 
1999
, vol. 
266
 (pg. 
1367
-
1374
)
Falchier
A
Clavagnier
S
Barone
P
Kennedy
H
Anatomical evidence of multimodal integration in primate striate cortex
J Neurosci
 , 
2002
, vol. 
22
 (pg. 
5749
-
5759
)
Fame
RM
MacDonald
JL
Macklis
JD
Development, specification, and diversity of callosal projection neurons
Trends Neurosci
 , 
2011
, vol. 
34
 (pg. 
41
-
50
)
Felleman
DJ
Van Essen
DC
Distributed hierarchical processing in the primate cerebral cortex
Cereb Cortex
 , 
1991
, vol. 
1
 (pg. 
1
-
47
)
Fuentealba
P
Tomioka
R
Dalezios
Y
Marton
LF
Studer
M
Rockland
K
Klausberger
T
Somogyi
P
Rhythmically active enkephalin-expressing GABAergic cells in the CA1 area of the hippocampus project to the subiculum and preferentially innervate interneurons
J Neurosci
 , 
2008
, vol. 
28
 (pg. 
10017
-
10022
)
Games
KD
Winer
JA
Layer V in rat auditory cortex: projections to the inferior colliculus and contralateral cortex
Hear Res
 , 
1988
, vol. 
34
 (pg. 
1
-
25
)
Groh
A
Meyer
HS
Schmidt
EF
Heintz
N
Sakmann
B
Krieger
P
Cell-type specific properties of pyramidal neurons in neocortex underlying a layout that is modifiable depending on the cortical area
Cereb Cortex
 , 
2010
, vol. 
20
 (pg. 
826
-
836
)
Hall
AJ
Lomber
SG
Auditory cortex projections target the peripheral field representation of primary visual cortex
Exp Brain Res
 , 
2008
, vol. 
190
 (pg. 
413
-
430
)
Hallman
LE
Schofield
BR
Lin
CS
Dendritic morphology and axon collaterals of corticotectal, corticopontine, and callosal neurons in layer V of primary visual cortex of the hooded rat
J Comp Neurol
 , 
1988
, vol. 
272
 (pg. 
149
-
160
)
Hattox
AM
Nelson
SB
Layer V neurons in mouse cortex projecting to different targets have distinct physiological properties
J Neurophysiol
 , 
2007
, vol. 
98
 (pg. 
3330
-
3340
)
Heumann
H
Wittum
G
The tree-edit-distance, a measure for quantifying neuronal morphology
Neuroinformatics
 , 
2009
, vol. 
7
 (pg. 
179
-
190
)
Hubener
M
Bolz
J
Morphology of identified projection neurons in layer 5 of rat visual cortex
Neurosci Lett
 , 
1988
, vol. 
94
 (pg. 
76
-
81
)
Hubener
M
Schwarz
C
Bolz
J
Morphological types of projection neurons in layer 5 of cat visual cortex
J Comp Neurol
 , 
1990
, vol. 
301
 (pg. 
655
-
674
)
Ichinohe
N
Hyde
J
Matsushita
A
Ohta
K
Rockland
KS
Confocal mapping of cortical inputs onto identified pyramidal neurons
Front Biosci
 , 
2008
, vol. 
13
 (pg. 
6354
-
6373
)
Innocenti
GM
Berbel
P
Clarke
S
Development of projections from auditory to visual areas in the cat
J Comp Neurol
 , 
1988
, vol. 
272
 (pg. 
242
-
259
)
Jacobs
B
Driscoll
L
Schall
M
Life-span dendritic and spine changes in areas 10 and 18 of human cortex: a quantitative Golgi study
J Comp Neurol
 , 
1997
, vol. 
386
 (pg. 
661
-
680
)
Jacobs
B
Schall
M
Prather
M
Kapler
E
Driscoll
L
Baca
S
Jacobs
J
Ford
K
Wainwright
M
Treml
M
Regional dendritic and spine variation in human cerebral cortex: a quantitative golgi study
Cereb Cortex
 , 
2001
, vol. 
11
 (pg. 
558
-
571
)
Kasper
EM
Larkman
AU
Lubke
J
Blakemore
C
Pyramidal neurons in layer 5 of the rat visual cortex. I. Correlation among cell morphology, intrinsic electrophysiological properties, and axon targets
J Comp Neurol
 , 
1994
, vol. 
339
 (pg. 
459
-
474
)
Koester
SE
O'Leary
DD
Functional classes of cortical projection neurons develop dendritic distinctions by class-specific sculpting of an early common pattern
J Neurosci
 , 
1992
, vol. 
12
 (pg. 
1382
-
1393
)
Krubitzer
L
The organization of neocortex in mammals: are species differences really so different?
Trends Neurosci
 , 
1995
, vol. 
18
 (pg. 
408
-
417
)
Laramee
ME
Kurotani
T
Rockland
KS
Bronchti
G
Boire
D
Indirect pathway between the primary auditory and visual cortices through layer V pyramidal neurons in V2L in mouse and the effects of bilateral enucleation
Eur J Neurosci
 , 
2011
, vol. 
34
 (pg. 
65
-
78
)
Larsen
DD
Callaway
EM
Development of layer-specific axonal arborizations in mouse primary somatosensory cortex
J Comp Neurol
 , 
2006
, vol. 
494
 (pg. 
398
-
414
)
Larsen
DD
Luu
JD
Burns
ME
Krubitzer
L
What are the effects of Severe visual Impairment on the cortical organization and connectivity of primary visual cortex?
Front Neuroanat
 , 
2009
, vol. 
3
 pg. 
30
 
Larsen
DD
Wickersham
IR
Callaway
EM
Retrograde tracing with recombinant rabies virus reveals correlations between projection targets and dendritic architecture in layer 5 of mouse barrel cortex
Front Neural Circuits
 , 
2007
, vol. 
1
 pg. 
5
 
Legendre
P
Legendre
L
Numerical ecology
 , 
1998
2nd ed
Amsterdam (The Netherlands)
Elsevier
Mason
A
Larkman
A
Correlations between morphology and electrophysiology of pyramidal neurons in slices of rat visual cortex. II. Electrophysiology
J Neurosci
 , 
1990
, vol. 
10
 (pg. 
1415
-
1428
)
Miller
MW
Vogt
BA
Direct connections of rat visual cortex with sensory, motor, and association cortices
J Comp Neurol
 , 
1984
, vol. 
226
 (pg. 
184
-
202
)
Molnar
Z
Cheung
AF
Towards the classification of subpopulations of layer V pyramidal projection neurons
Neurosci Res
 , 
2006
, vol. 
55
 (pg. 
105
-
115
)
Molyneaux
BJ
Arlotta
P
Fame
RM
MacDonald
JL
MacQuarrie
KL
Macklis
JD
Novel subtype-specific genes identify distinct subpopulations of callosal projection neurons
J Neurosci
 , 
2009
, vol. 
29
 (pg. 
12343
-
12354
)
Olavarria
J
Montero
VM
Organization of visual cortex in the mouse revealed by correlating callosal and striate-extrastriate connections
Visual Neurosci
 , 
1989
, vol. 
3
 (pg. 
59
-
69
)
Papp
E
Borhegyi
Z
Tomioka
R
Rockland
KS
Mody
I
Freund
TF
Glutamatergic input from specific sources influences the nucleus accumbens-ventral pallidum information flow
Brain Struct Funct
 , 
2012
, vol. 
217
 (pg. 
37
-
48
)
Reid
AT
Krumnack
A
Wanke
E
Kotter
R
Optimization of cortical hierarchies with continuous scales and ranges
Neuroimage
 , 
2009
, vol. 
47
 (pg. 
611
-
617
)
Rockland
KS
Kaas
JH
Collins
CE
Feedback connections: splitting the arrow
The primate visual system
 , 
2004
1st ed
Boca Raton (FL)
CRC Press
(pg. 
387
-
405
)
Rockland
KS
Ojima
H
Multisensory convergence in calcarine visual areas in macaque monkey
Int J Psychophysiol
 , 
2003
, vol. 
50
 (pg. 
19
-
26
)
Rockland
KS
Pandya
DN
Laminar origins and terminations of cortical connections of the occipital lobe in the rhesus monkey
Brain Res
 , 
1979
, vol. 
179
 (pg. 
3
-
20
)
Romand
S
Wang
Y
Toledo-Rodriguez
M
Markram
H
Morphological development of thick-tufted layer v pyramidal cells in the rat somatosensory cortex
Front Neuroanat
 , 
2011
, vol. 
5
 pg. 
5
 
Sanchez-Vives
MV
Nowak
LG
Descalzo
VF
Garcia-Velasco
JV
Gallego
R
Berbel
P
Crossmodal audio-visual interactions in the primary visual cortex of the visually deprived cat: a physiological and anatomical study
Prog Brain Res
 , 
2006
, vol. 
155
 (pg. 
287
-
311
)
Schofield
BR
Hallman
LE
Lin
CS
Morphology of corticotectal cells in the primary visual cortex of hooded rats
J Comp Neurol
 , 
1987
, vol. 
261
 (pg. 
85
-
97
)
Soloway
AS
Pucak
ML
Melchitzky
DS
Lewis
DA
Dendritic morphology of callosal and ipsilateral projection neurons in monkey prefrontal cortex
Neurosci
 , 
2002
, vol. 
109
 (pg. 
461
-
471
)
Tomioka
R
Rockland
KS
Improved Golgi-like visualization in retrogradely projecting neurons after EGFP-adenovirus infection in adult rat and monkey
J Histochem Cytochem
 , 
2006
, vol. 
54
 (pg. 
539
-
548
)
Tsiola
A
Hamzei-Sichani
F
Peterlin
Z
Yuste
R
Quantitative morphologic classification of layer 5 neurons from mouse primary visual cortex
J Comp Neurol
 , 
2003
, vol. 
461
 (pg. 
415
-
428
)
Vercelli
A
Innocenti
GM
Morphology of visual callosal neurons with different locations, contralateral targets or patterns of development
Exp Brain Res
 , 
1993
, vol. 
94
 (pg. 
393
-
404
)
Vezoli
J
Falchier
A
Jouve
B
Knoblauch
K
Young
M
Kennedy
H
Quantitative analysis of connectivity in the visual cortex: extracting function from structure
Neuroscientist
 , 
2004
, vol. 
10
 (pg. 
476
-
482
)
Wang
Q
Burkhalter
A
Area map of mouse visual cortex
J Comp Neurol
 , 
2007
, vol. 
502
 (pg. 
339
-
357
)
Wang
Q
Gao
E
Burkhalter
A
Gateways of ventral and dorsal streams in mouse visual cortex
J Neurosci
 , 
2011
, vol. 
31
 (pg. 
1905
-
1918
)