Abstract

The neocortex contains diverse populations of excitatory neurons segregated by layer and further definable by their specific cortical and subcortical projection targets. The current study describes a systematic approach to identify molecular correlates of specific projection neuron classes in mouse primary somatosensory cortex (S1), using a combination of in situ hybridization (ISH) data mining, marker gene colocalization, and combined retrograde labeling with ISH for layer-specific marker genes. First, we identified a large set of genes with specificity for each cortical layer, and that display heterogeneous patterns within those layers. Using these genes as markers, we find extensive evidence for the covariation of gene expression and projection target specificity in layer 2/3, 5, and 6, with individual genes labeling neurons projecting to specific subsets of target structures. The combination of gene expression and target specificity imply a great diversity of projection neuron classes that is similar to or greater than that of GABAergic interneurons. The covariance of these 2 phenotypic modalities suggests that these classes are both discrete and genetically specified.

Introduction

Primary somatosensory cortex (S1) forms connections with a wide variety of cortical and subcortical target structures subserving different features of sensorimotor information processing. Projections from S1 target other cortical regions, including neighboring ipsilateral secondary somatosensory cortex (S2), primary motor cortex (M1), and frontal cortex, multiple contralateral cortical areas, and the perirhinal cortex, which provides input to the hippocampal formation (Aronoff et al. 2010). Subcortically, S1 projects to the striatum (ipsilaterally and contralaterally), thalamus, tectum, and brainstem. These cortical and subcortical projections are made by specific projection neuron classes primarily in the supragranular (layer 2/3) and infragranular layers (5 and 6), and a variety of excitatory projection neuron classes have been described based on morphology, functional properties, and projection targets. Layer 2/3 neurons project primarily to other cortical regions, although they also send projections to the striatum (Reiner et al. 2003). Layer 5 projects primarily to striatum and deep brain (corticofugal) targets, and layer 6 primarily to the thalamus, although substantial corticocortical projections also exist in both layers (Mitchell and Macklis 2005; Chakrabarti and Alloway 2006). Similar to the well-described diversity of cortical GABAergic interneuron types (Gupta et al. 2000; Ascoli et al. 2008), many projection neuron classes in S1 can be defined on the basis of their target specificity, and understanding their specific properties will be key to understanding information processing through cortical circuits.

Selective expression of marker genes has been used for many years in classification of cortical inhibitory interneurons (Kubota et al. 2011). Nearly mutually exclusive expression of a small set of canonical markers can discriminate among GABAergic subclasses with different physiological and morphological properties as well as distinct developmental origins (Kawaguchi and Kondo 2002; Rudy et al. 2011). Identification of these markers has been highly valuable in allowing selective genetic targeting and functional manipulation (Doyle et al. 2008; Madisen et al. 2010; Taniguchi et al. 2011). While much less well established, molecular studies have begun to identify large numbers of genes with specificity for particular cortical layers and excitatory neuron subclasses within each layer (Sugino et al. 2006; Molyneaux, Arlotta, Menezes, et al. 2007; Doyle et al. 2008; Belgard et al. 2011; Bernard et al. 2012). For example, a number of genes with laminar specificity were identified through initial mining of the Allen Mouse Brain Atlas, a genome-wide ISH resource {(Lein et al. 2007 #2876; also see, Ko et al. 2013 #3152; Tebbenkamp AT and Borchelt DR 2010 #3153; Davis FP and Eddy SR 2009 #3155)}. Two more recent studies identified large cohorts of layer-enriched genes through selective laminar isolation and transcriptional profiling in macaque monkey using laser microdissection and DNA microarrays (Bernard et al. 2012), and in mouse using RNA-seq methods (Belgard et al. 2011). Projection neurons with distinct anatomical targets display distinct molecular signatures as well. For example, Pcp4 and Cck labeling differentiates corticocortical versus corticothalamic neurons in layer 6 of rat cortex (Watakabe et al. 2012). Transcriptional profiling of genetically labeled layer 5 neurons in cingulate cortex showed differential gene expression compared with retrogradely labeled corticothalamic neurons in layer 6 (Sugino et al. 2006), and between genetically labeled layer 6 corticothalamic, layer 5a corticocortical/corticostriatal, and layer 5b corticofugal neurons (Doyle et al. 2008). Similarly, comparison of FACS-sorted corticocortical versus corticofugal projection neurons during development identified differential gene expression patterns (Arlotta et al. 2005). One of these genes, Fezf2, is critical for proper differentiation of corticofugal neurons, and Fezf2 null mice fail to form subcortical projections (Molyneaux, Arlotta, and Macklis 2007).

Apart from these few examples, relationships between gene expression and projection target specificity is largely uncharacterized. Furthermore, some of the most broadly used markers for specific layers are not expressed selectively in neurons with a specific projection target. For example, Etv1, which is frequently used to label layer 5 pyramidal neurons, is expressed both in corticocortical and corticofugal neurons, but not by all neurons of either population {(Hevner et al. 2003 #2712; Yoneshima et al. 2006 #3150)}. Similarly, SMI-32, FNP-7, and N200 can distinguish between Type I and Type II neurons in layer 5 of rat S1, but cannot distinguish among neurons projecting to different subcortical targets {(Voelker et al. 2004 #3151)}. To better understand the diversity of projection neuron classes and to allow selective genetic manipulation of those classes, identification of additional genes and a direct correlation with projection targets is needed.

The current study aimed to systematically identify genes likely to be expressed in specific excitatory projection neuron classes by first mining publicly accessible genome-wide ISH data in the Allen Mouse Brain Atlas (Lein et al. 2007) for genes with laminar specificity. We then used a double-labeling approach to identify genes with nearly mutually exclusive expression, and therefore likely expressed in distinct classes. Finally, we combined retrograde labeling with ISH to assign projection target specificity to a large cohort of laminar genes in layers 2/3, 5, and 6. The results demonstrate striking correlations between marker genes and projection targets, and illustrate a highly diverse population of excitatory projection neuron classes with selective molecular signatures.

Materials and Methods

Manual Annotation of Laminar Gene Expression Patterns in S1

All available ISH image series in the Allen Mouse Brain Atlas (Lein et al. 2007) were examined for heterogeneous cellular expression in S1. ISH images with poor data quality were excluded from the analysis. Laminar expression patterns were scored for ∼3300 image series (http://help.brain-map.org/display/mousebrain/Documentation), and scores for 1053 genes with most robust heterogeneous patterns predominantly in neurons are supplied as Supplementary Table 1. Cellular expression patterns in S1 were visually scored in layers 1, 2/3, 4, 5, 6, and 6b for density (0–4) and level (0–5), based on quantified heat maps described in (Lein et al. 2007), as well as for nonuniform distribution across the depth of each layer (sublaminarity: superficial, middle, or deep). Density scores represented sparse (1), scattered (2), medium (3), and high (4) cellular densities relative to total cells in that layer. Scores of 1 and 2 represent either GABAergic interneurons (<20% of total neurons), glia, or sparse excitatory neurons. Scores of 3 or 4 must at least contain the predominant excitatory neurons in that layer. Layer 1 was scored only with density of 1 or 2 to reflect the low cellular density and lack of excitatory cells in that layer.

Animals and Tissue Processing

All experimental procedures were approved by the Allen Institute for Brain Science Animal Care and Use Committee. Male C57BL/6J mice (Jackson Labs West) between postnatal day 52 and 58 were used for all experiments. For all ISH studies, serial 25-µm fresh-frozen cryostat sections were systematically collected at regular intervals from the entire brain, or from a region containing S1, and hybridized to gene-specific probes. For retrograde labeling studies, when the injection site fell outside of the sections containing S1, 40-µm-thick sections through the injection site were also collected. These sections were used for target verification and were not processed for ISH.

Fluorescent In Situ Hybridization

In situ hybridization (ISH) data were generated using a semiautomated, nonisotopic digoxygenin (DIG)-based platform as described previously (Lein et al. 2007; Thompson et al. 2008). For single-probe fluorescent ISH (FISH), riboprobes were labeled with DIG-UTP and/or dinitrophenyl-11-UTP (DNP; Perkin Elmer, Waltham, MA, USA). For double-probe FISH (dFISH), a DNP-labeled probe and a DIG-labeled probe were hybridized to the tissue simultaneously. Signal amplification was performed for each individual probe using either anti-DIG-HRP with tyramide-biotin, or anti-DNP-HRP with tyramide-DNP. For single-probe FISH, signal was visualized using streptavidin-Alexa Fluor 488 or streptavidin-Alexa Fluor 647 (Life Technologies/Molecular Probes). For dFISH, signal was visualized using streptavidin-Alexa Fluor 488 (Life Technologies/Molecular Probes) and anti-DNP-Alexa Fluor 555 (Life Technologies/Molecular Probes). Image capture was performed using an automated fluorescent microscopy platform (Thompson et al. 2008). Briefly, mosaic images were created from stitched single images captured using a Plan Apo ×10 objective and appropriate filter cubes (Semrock, Rochester, NY, USA) on a DM6000B Leica microscope (Leica Microsystems, Wetzlar, Germany).

Fluorescent Retrograde Labeling

Mice were anesthetized using 1–3.5% inhalant isoflurane or an intraperitoneal injection of 2.5% Avertin and then placed in a stereotaxic device (myNeurolab.com, St. Louis, MO, USA). The skin along the midline of the skull was opened using a scalpel and a surgical drill was used to create 1–3 small holes in the skull. Pulled glass pipettes attached to a picospritzer were used to make 1–4 injections (0.1–0.4 µL) into each brain region targeted. Stereotaxic coordinates were obtained from Paxinos adult mouse brain atlas (Paxinos and Franklin 2004) (Supplementary Table 2). In order to retrogradely label distinct populations of layer 5 projection neurons in S1, we made injections of Cholera toxin B conjugated to Alexa 555 (CTB-555) (Life Technologies/Invitrogen, Carlsbad, CA, USA) into contralateral S1 (callosal), striatum (corticostriatal), thalamus (corticothalamic), superior colliculus (corticotectal), and trigeminal nucleus (corticotrigeminal neurons), respectively (Supplementary Fig. 1). In order to use mice efficiently and to retrogradely label 2 potentially related populations of projection neurons in layers 2/3, 5, and 6 of S1, for a subset of experiments injections of CTB-555 or CTB-488 were made in pairs of structures, including S1, striatum, thalamus, superior colliculus, pons, trigeminal nucleus, frontal cortex, M1, and S2. Mice were kept alive for 5–7 days following surgery and then sacrificed.

Brains with retrograde labeling were processed for single-probe FISH as described above. ISH signal was visualized using streptavidin-Alexa Flour 488 for single target injections or streptavidin-Alexa Fluor 647 for dual target injections to allow fluorescently tagged CTB and riboprobe to be visualized in the same tissue.

Quantification of FISH data

Double FISH

Metamorph v. 7.7 (Sunnyvale, CA, USA) was used to quantify dFISH colabeling. Three-color (Dapi, Alexa-488 and Alexa-555) mosaic images were analyzed. Using the DAPI signal, a region of interest within layer 5 and within the boundaries of S1 was manually traced on both sides of the brain in each image based on the Allen Coronal Reference Atlas (Dong 2008). Green and red cells were identified and counted by viewing each channel separately. Colabeled cells were then counted based on the identification of cells in each individual channel, and the % of the total number of counted cells that were either single- or double-positive for the 2 genes was calculated. The density of labeled cells per 100 µm2 of gene-positive cells in each category was also quantified per layer. For each pair of genes, 3–4 sections in S1 of both hemispheres were analyzed in 2 replicates, with 200 µm separation between sections.

Retrograde Labeling + FISH

Essentially the same analysis procedures were used to quantify colabeled cells for single retrograde tracing studies combined with FISH. The percentage of CTB-labeled cells positive for the gene of interest was calculated. For each projection neuron population and gene combination, we analyzed 2–4 sections with 200 µm spacing between sections for 3–6 brains for the data presented in Figure 6 (Supplementary Table 3) and for 1–6 brains for the data presented in Figures 7, 8, and 9. Statistical significance was determined using ANOVA and a student's t-test. For the qualitative analysis presented in Figures 7, 8, and 9, the level of colabeling was assessed per gene/projection target combination by carefully viewing images of 4–8 tissue sections containing S1. Each RGB image was opened using Photoshop or Metamorph software, and the presence of colabeling was evaluated by first identifying CTB+ cells (in the red or green channel) and then toggling back and forth between channels to determine whether these cells colocalized with labeling for the gene of interest (in the blue channel).

Results

Specificity and Diversity of Layer-Specific Gene Expression Patterns in S1

To identify genes with heterogeneous cellular expression patterns across layers of S1, we first performed a systematic screening of cellular level ISH expression patterns across the ∼20 000 genes assayed in the Allen Mouse Brain Atlas (Lein et al. 2007). This screen identified 1053 genes displaying robust, nonubiquitous, and predominantly neuronal patterns, which were scored for their relative cellular density (0–4; maximum of 2 for layer 1), expression level (0–5), and sublaminar enrichment across cortical layers 1, 2/3, 4, 5, 6, and 6b (see Materials and Methods section and Supplementary Table 1). Remarkably, 950 patterns were observed among the 1053 genes scored with this method. It is likely that some proportion of this variation may be due to probe- and experimental condition-related variability in expression levels, while labeled cell density, though still sensitive to signal thresholding, may be less subject to these factors. Using only the cellular density scores, we still observe 508 distinct cellular patterns (Supplementary Table 1), thereby illustrating the great complexity of gene expression in the cortex. The most frequently observed density patterns are shown in Figure 1a. Interestingly, the majority of the most common expression patterns, while selective for neuronal cells based on their original identification, are broad in their laminar distribution. For example, the most frequently observed pattern was near pan-neuronal (maximum density in all layers; observed for 28 different genes). Selective labeling of either nearly all excitatory neurons (high density in all layers except layer 1, which does not contain excitatory neurons; observed for 19 genes) or putative inhibitory neurons (low density in all layers including layer 1; observed for 18 different genes) were also frequent. Several highly specific laminar patterns were also among the most frequently observed (red arrows in Fig. 1a), including selective expression in layer 5 (19 genes) and layer 6b (7 genes), indicating that the neuron subtypes in these layers are the most genetically distinct among cortical neurons.

Figure 1.

Global molecular patterning across layers of S1. Laminar (a,b) and sublaminar (c) distribution of the 1053 most heterogeneous, predominantly neuronal genes, based on scored densities of labeled neurons in each layer of S1. (a) Forty-three most commonly observed patterns. Red arrows indicate layer-specific expression in layer 5 and 6b. (b) Frequency of laminar enrichment in each layer and combination of layers, as a function of the degree of enrichment (difference in density score between layer(s) of interest compared with all other layers). (c) Frequency of sublaminar enrichment among scored genes for layers 2/3, 4, 5, and 6, broken down into superficial, middle, and deep partitions.

Figure 1.

Global molecular patterning across layers of S1. Laminar (a,b) and sublaminar (c) distribution of the 1053 most heterogeneous, predominantly neuronal genes, based on scored densities of labeled neurons in each layer of S1. (a) Forty-three most commonly observed patterns. Red arrows indicate layer-specific expression in layer 5 and 6b. (b) Frequency of laminar enrichment in each layer and combination of layers, as a function of the degree of enrichment (difference in density score between layer(s) of interest compared with all other layers). (c) Frequency of sublaminar enrichment among scored genes for layers 2/3, 4, 5, and 6, broken down into superficial, middle, and deep partitions.

We next asked which laminar patterns were most common and whether there tended to be coexpression among any layers. Figure 1b shows the frequency of specific laminar enrichment patterns at different levels of relative enrichment based on density scores. The majority of all possible laminar patterns were observed at some level of enrichment, although some patterns were far more frequent. For example, selective enrichment in layers 2/3, 5, 6, and 6b, containing predominantly projection neurons, was far more frequent than enrichment in the local circuit-containing layers 4 and 1. Similarly, one of the most common patterns involving multiple layers was enrichment in all projection neuron layers combined (2/3, 5, 6, 6b), perhaps reflecting expression patterns associated with the formation and maintenance of these projections. As noted above, selective 6b expression was among the most commonly observed patterns, but enrichment in all other excitatory neuron-containing layers except 6b was also among the most frequently observed patterns.

In addition to laminar specificity, a significant proportion of genes were not uniformly expressed throughout the depth of each layer, but rather relatively enriched in the superficial, middle or deep aspects of that layer (Fig. 1c). For example, nearly 20% of scored genes in layer 2/3 showed sublaminar enrichment, with the majority representing the superficial aspect adjacent to layer 1 (putative layer 2). Approximately one-quarter of layer 5 genes were selective for either deep or superficial aspects with a small proportion restricted to the middle of layer 5. Approximately 15% of layer 6 genes were sublaminar as well. A much smaller proportion of genes expressed in layer 4 were nonuniform, with the majority showing enrichment in deep layer 4. Overall, these gene expression data illustrate a remarkable laminar specificity, degree of combinatorial complexity, and sublaminar complexity suggestive of a high degree of cortical neuronal heterogeneity.

Based on these analyses, many genes with enriched expression in each cortical layer were identified for further analysis. These genes displayed a wide range of cellular densities and sublaminar patterning. For example, Figure 2 shows genes enriched in layer 2/3 (top row), with pink highlighting indicating the portion of the layer with the majority of labeled neurons. Some genes appear to be expressed by the majority of neurons in this layer (e.g., Rasgrf2 and Gucy1a3). In contrast, other genes are restricted to the most superficial (Rfx3, Trpc6) or deep (Grid2ip) part of layer 2/3. Presumably these latter genes delineate layer 2 from layer 3, which are difficult to discriminate in rodents based on traditional histological methods.

Figure 2.

Spectrum of laminar gene expression patterns in projection neuron layers of S1. Colorimetric ISH images from the Allen Mouse Brain Atlas demonstrating the diversity of expression patterns observed in layers 2/3, 5, 6, and 6b. Black lines mark layer boundaries, with pink coloring delineating the portion of the layer of interest with densest labeling to highlight sublaminar enrichments. Genes in each layer demonstrate a wide range of cell densities and sublaminar distributions. Scale bar: 500 μm.

Figure 2.

Spectrum of laminar gene expression patterns in projection neuron layers of S1. Colorimetric ISH images from the Allen Mouse Brain Atlas demonstrating the diversity of expression patterns observed in layers 2/3, 5, 6, and 6b. Black lines mark layer boundaries, with pink coloring delineating the portion of the layer of interest with densest labeling to highlight sublaminar enrichments. Genes in each layer demonstrate a wide range of cell densities and sublaminar distributions. Scale bar: 500 μm.

Layer 5 neurons (Fig. 2, rows 2 and 3) were selectively labeled by a particularly large number of genes that varied widely in their densities and sublaminar distributions. Selective enrichment in more superficial layer 5 (5a) likely relates to intracerebral projection neurons (e.g., corticostriatal, callosal), while enrichment in deeper layer 5 (5b) likely reflects the predominant long-range (corticofugal) projection neurons (Molnar and Cheung 2006). We identified genes that had dense expression through the entire layer (Etv1, Hsd11b1), genes with superficial enrichment (Trib2), and a relatively larger number of genes with deep layer 5 enrichment (e.g., Bcl6, Npr3). In addition, several genes had extremely sparse deep expression (Anxa1 and Chrna6).

Layer 6 (Fig. 2, row 4) also exhibited a range of laminar expression patterns, including genes expressed across the entire layer (e.g., Foxp2) and genes that had expression restricted to superficial (e.g., Tnnc1) and deep layer 6 (Bmp3, Sulf1). Gnb4 had particularly strong expression in lateral cortex, and became sparser as it extended into the border region between S1 and S2.

Finally, a number of genes showed strong specificity for layer 6b (bottom row), which is derived from the developmental subplate zone (Fig. 2). This layer varies extensively across species (Montiel et al. 2011), and is known to contain a diverse cell population with selective gene expression (Hoerder-Suabedissen et al. 2009; Oeschger et al. 2012). This diversity is mirrored by expression patterns ranging from dense (e.g., Ctgf) to quite sparse (e.g., Trh).

Systematic Double Labeling Identifies Layer 5 Neuron Markers with Mutually Exclusive Expression

Layer 5 has the most diverse set of efferent target structures among cortical layers, as well as particularly selective and diverse gene expression patterns. To identify genes that label these anatomically and functionally distinct populations of layer 5 neurons, we first sought to identify genes that are never or rarely coexpressed in the same neurons. To that end, we selected a set of 10 layer 5-enriched genes (all shown in Fig. 2) that represent the spectrum of observed patterns (dense, sparse, 5a, 5b) and developed a robust method for studying coexpression relationships by double fluorescent ISH (dFISH). For dFISH, 2 genes were detected simultaneously with probes conjugated either to Alexa Fluor555 (Red) or Alexa Fluor488 (green). All possible combinations of these 10 markers (45 combinations, each with a fluor-reversal for a total of 90 pairwise analyses) were performed. Figure 3 shows representative dFISH data for a subset of these comparisons, and quantitative coexpression results for the entire 10 × 10 layer 5 coexpression matrix are shown in Figure 4aj. These genes showed a highly complex series of coexpression relationships. Many genes showed mutually exclusive expression, which we focus on below. On the other hand, the most common outcome of these experiments was partial co-expression. Overall the degree of overlap was surprisingly low, and we did not observe any case where greater than 50% of total labeled cells in any pairwise comparison expressed both genes (Fig. 4).

Figure 3.

Colabeling identifies layer 5 enriched genes with overlapping and mutually exclusive expression. (ai) Fluorescent ISH data for the gene pair indicated in a coronal section through S1, with raw dFISH data in the left panels and quantification of single and double labeling on the right. The left-most dFISH panel is a low-magnification image with layers indicated, and the right panels show high-magnification single and double labeling within layer 5 corresponding to the boxed region. Histograms shows quantitative data for single (red, green) and double (yellow) labeled cells as mean percentages of total labeled cells (±SEM). Panels (af) show minimal to no overlap between Deptor, Bcl6, Trib2, and Slc17a8. Panels (gi) show partially overlapping expression between Bcl6 and Vat1l, Hsd11b1, and Npr3, respectively. Scale bars: 125 μm (low magnification), 125 μm (high magnification).

Figure 3.

Colabeling identifies layer 5 enriched genes with overlapping and mutually exclusive expression. (ai) Fluorescent ISH data for the gene pair indicated in a coronal section through S1, with raw dFISH data in the left panels and quantification of single and double labeling on the right. The left-most dFISH panel is a low-magnification image with layers indicated, and the right panels show high-magnification single and double labeling within layer 5 corresponding to the boxed region. Histograms shows quantitative data for single (red, green) and double (yellow) labeled cells as mean percentages of total labeled cells (±SEM). Panels (af) show minimal to no overlap between Deptor, Bcl6, Trib2, and Slc17a8. Panels (gi) show partially overlapping expression between Bcl6 and Vat1l, Hsd11b1, and Npr3, respectively. Scale bars: 125 μm (low magnification), 125 μm (high magnification).

Figure 4.

Quantification of systematic dFISH analysis among 10 layer 5-selective genes. Each graph shows the relationship between a “parent” gene (red) and the remaining 9 layer 5-selective genes (green) as the percentage of the total number of cells (±SEM) that were single- or double-labeled (yellow) by dFISH. Gene comparisons are shown from top to bottom roughly ordered from the least to the most selective. Etv1 was least selective (a), and exhibited some coexpression with all other genes assayed, while Hsd11b1 9 (b) and Vat1l (c) partially overlapped with all genes except Slc17a8. Bcl6, Deptor, Trib2, and Slc17a8 were much more selective in their expression, and displayed mutually exclusive or largely nonoverlapping (Trib2) expression as illustrated in Venn diagram form (l), based on the labeled cell densities in layer 5 (mean number of neurons per 100 µm2 ± SEM). The mean density of labeled cells for each gene in layer 5 is shown in (k).

Figure 4.

Quantification of systematic dFISH analysis among 10 layer 5-selective genes. Each graph shows the relationship between a “parent” gene (red) and the remaining 9 layer 5-selective genes (green) as the percentage of the total number of cells (±SEM) that were single- or double-labeled (yellow) by dFISH. Gene comparisons are shown from top to bottom roughly ordered from the least to the most selective. Etv1 was least selective (a), and exhibited some coexpression with all other genes assayed, while Hsd11b1 9 (b) and Vat1l (c) partially overlapped with all genes except Slc17a8. Bcl6, Deptor, Trib2, and Slc17a8 were much more selective in their expression, and displayed mutually exclusive or largely nonoverlapping (Trib2) expression as illustrated in Venn diagram form (l), based on the labeled cell densities in layer 5 (mean number of neurons per 100 µm2 ± SEM). The mean density of labeled cells for each gene in layer 5 is shown in (k).

We identified 4 genes with largely nonoverlapping expression (Fig. 3af). Bcl6 was rarely coexpressed with Deptor (0.25 ± 0.25%), Slc17a8 (0.38 ± 0.38%), and Trib2 (4.55 ± 0.75%) (Fig. 3ac). Most Trib2-positive cells are not labeled by these genes, although there is some overlap with Bcl6 and Deptor (Fig. 3c,f). Slc17a8 was the most selective, only showing appreciable overlap with Etv1 among the entire marker panel (16.4 ± 2.4%; Fig. 4j). Relationships among the cell populations labeled by these 4 genes are summarized in the Venn diagram in Figure 4l, showing near mutual exclusivity for Bcl6, Deptor, and Slc17a8, and minimal overlap of Trib2 with all 3 of these genes. Several other genes, including Vat1l (Figs 3g and 4c), Npr3 (Figs 3i and 4e), and Layn (Fig. 4f), showed similar patterns to Bcl6 with little to no overlap with Deptor, Trib2, and Slc17a7 suggesting that they label a similar cell population. However, labeling with these genes only partially overlapped with Bcl6, indicating that there are additional gene-defined layer 5 neurons not captured in the Venn diagram in Figure 4l. In summary, this approach identified a set of genes labeling largely nonoverlapping populations of layer 5 neurons. We hypothesized that these genes could molecularly define functionally distinct excitatory projection neuron classes.

Combined Retrograde Labeling of Layer 5 Projection Neurons with ISH Demonstrates Correlations Between Projection Targets and Molecular Phenotypes

To directly test whether Bcl6, Deptor, Trib2, and Slc17a8 label different classes of neurons in S1, we combined FISH using Alexa Fluor 488-conjugated probes with retrograde labeling using Cholera toxin B (CTB) conjugated to Alexa Fluor 555. We retrogradely labeled 5 different targets of S1 layer 5 projection neurons using CTB injections, including ipsilateral striatum (corticostriatal), thalamus (corticothalamic), superior colliculus (corticotectal), and contralateral S1 (callosal), and trigeminal nucleus (corticotrigeminal). CTB-labeled tissue was then processed for ISH for each of the 4 genes, allowing simultaneous visualization of retrogradely labeled neurons and cells labeled by each gene. Figure 5 shows a representative image from each of these experiments for Deptor, Trib2, Bcl6, and Slc17a8 and each projection target (20 gene/target combinations). Each panel contains an image of retrogradely labeled cells, images of ISH-labeled cells, the overlay image of the 2, and a schematic of single retrogradely labeled (red) and double-labeled (yellow) cells in these particular tissue sections. These schematics clearly illustrate the relationship between genes and projection targets. Bcl6, which is expressed in deep layer (5b), labeled the majority of deep brain-projecting neurons (corticothalamic, corticotectal, and corticotrigeminal), but did not label neurons projecting to the striatum and contralateral cortex (Fig. 5c). Deptor had essentially a converse pattern to that of Bcl6, labeling a significant portion of corticostriatal and callosal neurons in the middle and superficial portions of layer 5, but did not label deep brain-projecting neurons (Fig. 5a). Similarly, Trib2 primarily labeled corticostriatal and callosal neurons located in the most superficial portion of layer 5 (Fig. 5b). Interestingly, Slc17a8 rarely labeled neurons from any of these targets (Fig. 5d).

Figure 5.

Combined layer 5 marker ISH with fluorescent retrograde labeling demonstrates correlations between gene expression and target specificity. S1 FISH labeling for Deptor (a), Trib2 (b), Bcl6 (c), and Slc17a8 (d), combined with CTB retrograde labeling from the striatum (corticostriatal; Row (1), contralateral S1 (callosal; Row (2), thalamus (corticothalamic; Row (3), superior colliculus (corticotectal; Row (4), and trigeminal nucleus (corticotrigeminal; Row (5). From the left within each panel, images show retrogradely labeled cells, cells labeled by ISH, the fluorescent overlay with high-magnification inset in layer 5, and a schematic of single retrogradely labeled cells (red) and colabeled cells (yellow). In the schematic view, layer 5 is delineated by dotted lines, and cells labeled by ISH but not CTB are not displayed. Deptor was expressed by a substantial portion of corticostriatal and callosal neurons, and rarely or not at all by deep brain-projecting neurons (corticothalamics, corticotectals, and corticotrigeminals). Trib2 also labeled a high proportion of corticostriatal and callosal neurons, but few of the deep brain-projecting neurons. In contrast, Bcl6 only rarely labeled these intracerebral populations, but labeled the majority of deep brain-projecting neurons. Slc17a8 did not label any of these populations except a small proportion of corticostriatal neurons. Scale bar: 100 μm.

Figure 5.

Combined layer 5 marker ISH with fluorescent retrograde labeling demonstrates correlations between gene expression and target specificity. S1 FISH labeling for Deptor (a), Trib2 (b), Bcl6 (c), and Slc17a8 (d), combined with CTB retrograde labeling from the striatum (corticostriatal; Row (1), contralateral S1 (callosal; Row (2), thalamus (corticothalamic; Row (3), superior colliculus (corticotectal; Row (4), and trigeminal nucleus (corticotrigeminal; Row (5). From the left within each panel, images show retrogradely labeled cells, cells labeled by ISH, the fluorescent overlay with high-magnification inset in layer 5, and a schematic of single retrogradely labeled cells (red) and colabeled cells (yellow). In the schematic view, layer 5 is delineated by dotted lines, and cells labeled by ISH but not CTB are not displayed. Deptor was expressed by a substantial portion of corticostriatal and callosal neurons, and rarely or not at all by deep brain-projecting neurons (corticothalamics, corticotectals, and corticotrigeminals). Trib2 also labeled a high proportion of corticostriatal and callosal neurons, but few of the deep brain-projecting neurons. In contrast, Bcl6 only rarely labeled these intracerebral populations, but labeled the majority of deep brain-projecting neurons. Slc17a8 did not label any of these populations except a small proportion of corticostriatal neurons. Scale bar: 100 μm.

For each gene/target combination we counted the number of CTB-labeled neurons within layer 5 of S1, and quantified the mean percentage of CTB-labeled cells that expressed that gene (Fig. 6). Bcl6 labeled the great majority of corticothalamic (87.0 ± 7.14%), corticotectal (92.0 ± 4.99%), and corticotrigeminal (87.4 ± 6.54%) neurons. The majority of intracerebral corticostriatal and callosal neurons rarely expressed Bcl6 (7.45 ± 3.93% and 1.12 ± 1.40%, respectively). Conversely, Deptor, which is expressed in the border region between layer 5a and 5b (Fig. 5a), primarily labeled corticostriatal (66.63 ± 5.35%, Fig. 6e) and callosal neurons (51.24 ± 5.07%, Fig. 6d). Similarly, Trib2, which is expressed primarily in layer 5a, labeled a high proportion of corticostriatal (57.19 ± 8.25%, Fig. 6e) and callosal (70.66 ± 14.13%, Fig. 6d) neurons, but few of the deep brain-projecting neurons. Interestingly, the proportion of neurons projecting to these 3 corticofugal structures that expressed Trib2 and Deptor decreased with distance from the cortex. A small proportion of corticothalamic neurons expressed Trib2 (23.9 ± 14.80%) and Deptor (5.8 ± 12.33%), while an even smaller proportion of corticotectal neurons expressed Trib2 (10.06 ± 8.54%). The greatest molecular specificity was seen for the most distant target: 87.38 ± 6.54% of corticotrigeminal neurons expressed Bcl6, while none expressed Trib2 and only 0.90 ± 1.4% expressed Deptor (Fig. 6c). Remarkably, in 6 of 6 specimens examined Slc17a8 only labeled a very small proportion of corticostriatal neurons (2.88 ± 0.95%, Fig. 6e), suggesting that most Slc17a8-positive neurons project to structures not examined in this study. In summary, Bcl6 labels the majority of deep brain-projecting neurons, Slc17a8 labels only a small percentage of corticostriatal neurons, and Deptor and Trib2 mainly label intracerebral projecting neurons. As these 2 latter genes only minimally overlap based on the colabeling data above, this result suggests there are at least 3 different molecularly distinct classes of intracerebrally projecting neurons (i.e., Deptor+, Trib2+, and Deptor/Trib2+).

Figure 6.

Mutually exclusive layer 5 genes distinguish broad classes of projection neurons. Graphs display the mean % of CTB-labeled cells in S1 projecting to the thalamus (a), superior colliculus (b), spinal trigeminal nucleus (c), contralateral S1 (d), or ipsilateral striatum (e) that also express Slc17a8, Bcl6, Deptor, and Trib2. (a) The great majority (∼85%) of corticothalamic neurons expressed Bcl6, with a smaller percentage that expressed Trib2 (∼20%) and Deptor (∼5%) and no cells that expressed Slc17a8. The majority of corticotectal neurons (b) and corticotrigeminal neurons (c) also expressed Bcl6 (∼90%) and rarely any of the other genes. Callosal (d) and corticostriatal (e) neurons exhibited the opposite expression pattern. Few of these cells expressed Bcl6, while ∼50% and ∼60% of callosal and corticostriatal neurons, respectively, expressed Deptor and Trib2. Approximately 5% of corticostriatal neurons expressed Slc17a8. *P < 0.01.

Figure 6.

Mutually exclusive layer 5 genes distinguish broad classes of projection neurons. Graphs display the mean % of CTB-labeled cells in S1 projecting to the thalamus (a), superior colliculus (b), spinal trigeminal nucleus (c), contralateral S1 (d), or ipsilateral striatum (e) that also express Slc17a8, Bcl6, Deptor, and Trib2. (a) The great majority (∼85%) of corticothalamic neurons expressed Bcl6, with a smaller percentage that expressed Trib2 (∼20%) and Deptor (∼5%) and no cells that expressed Slc17a8. The majority of corticotectal neurons (b) and corticotrigeminal neurons (c) also expressed Bcl6 (∼90%) and rarely any of the other genes. Callosal (d) and corticostriatal (e) neurons exhibited the opposite expression pattern. Few of these cells expressed Bcl6, while ∼50% and ∼60% of callosal and corticostriatal neurons, respectively, expressed Deptor and Trib2. Approximately 5% of corticostriatal neurons expressed Slc17a8. *P < 0.01.

Given the relationships observed between projection target and marker gene expression, we expanded this approach to systematically examine the target specificity of neurons expressing a larger panel of layer 5-enriched genes. This expanded panel included 8 out of 10 of the gene/projection target set described above plus 14 additional genes, in combination with a larger set of layer 5 target structures. In addition to the callosal and corticofugal targets described above, we examined the molecular specificity of layer 5 projections to the pons and to different cortical regions including ipsilateral frontal cortex, M1 and S2 (Fig. 7a). The results of systematic retrograde labeling from 9 different layer 5 target structures combined with FISH labeling for 22 genes are summarized in Figure 7b. In this figure (and the subsequent 2 figures), genes are grouped roughly by the sets of target structures they mark. In each case, the estimated proportion of retrogradely labeled cells that expressed each gene is shown in a color-coded grid which indicates coexpression proportions ranging from sparse to a large proportion of retrogradely labeled cells. Images were scored in a very conservative fashion to emphasize even minimal colabeling. The level of colabeling was binned into 4 categories: none (white), sparse (light blue) when rare colabeled cells could be identified, moderate (royal blue) when the incidence of colabeling was frequent but less than the majority, and high (dark blue) when the majority of cells were colabeled.

Figure 7.

Target specificity of molecularly defined projection neuron classes in layer 5 of S1. (a) Schematic of different cortical and subcortical targets of layer 5 projection neurons tested. (b) Summary of projection targets for layer 5 neurons expressing specific marker genes, based on CTB retrograde labeling with FISH for that gene. The approximate proportion of neurons projecting to each target labeled by each gene is indicated by color, including no labeling (white), sparse (light blue), low to medium (blue), and high (dark blue). Generally, individual genes labeled neurons projecting to specific combinations of possible projection targets. A subset of genes selectively labeled corticofugal neurons, while the majority of genes labeled some striatally projecting neurons. Several genes selectively labeled neurons projecting to frontal cortex among the possible cortical targets. (c,d) Target specificity for neurons expressing Chrna6 (c) and Anxa1 (d). For each gene, double labeling with FISH (green) and CTB retrograde labeling (red) from superior colliculus (corticotectal, left), thalamus (corticothalamic, middle), and spinal trigeminal nucleus (corticotrigeminal, right) is shown. Within each panel, left image is a low-magnification image spanning the thickness of S1, while the right panels show high-magnification single and double labeling corresponding to the boxed regions in layer 5. (e) Quantification of the percentage of retrogradely labeled neurons from each target expressing Chrna6 and Anxa1, demonstrating that a small subset of projection neurons express each gene, specificity for specific subcortical targets, and selectivity for superior colliculus projecting neurons for Chrna6. Animal numbers for Chrna6 analysis, S2/Frontal: N = 3, Callosal/Striatum/Pons: N = 4, Thalamus/Trigeminal: N = 5, Superior Colliculus: N = 9; animal numbers for Anxa1 analysis, S2/Frontal/Callosal: N = 1, Striatum/Pons/Trigeminal: N = 3, Thalamus: N = 4. Scale bars in (c,d): low magnification: 250 μm, high magnification: 75 μm.

Figure 7.

Target specificity of molecularly defined projection neuron classes in layer 5 of S1. (a) Schematic of different cortical and subcortical targets of layer 5 projection neurons tested. (b) Summary of projection targets for layer 5 neurons expressing specific marker genes, based on CTB retrograde labeling with FISH for that gene. The approximate proportion of neurons projecting to each target labeled by each gene is indicated by color, including no labeling (white), sparse (light blue), low to medium (blue), and high (dark blue). Generally, individual genes labeled neurons projecting to specific combinations of possible projection targets. A subset of genes selectively labeled corticofugal neurons, while the majority of genes labeled some striatally projecting neurons. Several genes selectively labeled neurons projecting to frontal cortex among the possible cortical targets. (c,d) Target specificity for neurons expressing Chrna6 (c) and Anxa1 (d). For each gene, double labeling with FISH (green) and CTB retrograde labeling (red) from superior colliculus (corticotectal, left), thalamus (corticothalamic, middle), and spinal trigeminal nucleus (corticotrigeminal, right) is shown. Within each panel, left image is a low-magnification image spanning the thickness of S1, while the right panels show high-magnification single and double labeling corresponding to the boxed regions in layer 5. (e) Quantification of the percentage of retrogradely labeled neurons from each target expressing Chrna6 and Anxa1, demonstrating that a small subset of projection neurons express each gene, specificity for specific subcortical targets, and selectivity for superior colliculus projecting neurons for Chrna6. Animal numbers for Chrna6 analysis, S2/Frontal: N = 3, Callosal/Striatum/Pons: N = 4, Thalamus/Trigeminal: N = 5, Superior Colliculus: N = 9; animal numbers for Anxa1 analysis, S2/Frontal/Callosal: N = 1, Striatum/Pons/Trigeminal: N = 3, Thalamus: N = 4. Scale bars in (c,d): low magnification: 250 μm, high magnification: 75 μm.

These results confirm and expand on the original results (Figs 5 and 6), and 2 features are immediately apparent (Fig. 7b). First, individual genes labeled neurons with a high degree of target specificity, and almost all genes examined showed some degree of target specificity. Only 2 genes, Arhgap26 and S100b, were expressed in neurons projecting to all target regions, and even in that case not all CTB+ neurons were labeled. In other words, few genes labeled neurons projecting to all targets, and pan-layer 5 labeling was not observed among these layer-specific gene markers. Second, the target specificity of layer 5 neurons is highly diverse and combinatorial. The majority of genes tested labeled neurons projecting to one or more corticofugal targets, and within this category nearly all possible combinations of corticofugal targets were observed. For example, among the genes compared by dFISH (Figs 3 and 4), Bcl6, Fam20a, and Layn labeled neurons projecting to all corticofugal targets, while Hsd11b1 was slightly more selective, labeling all corticofugal populations except corticotrigeminal neurons. Trib2 and Deptor labeled only a small proportion of neurons projecting to thalamus and superior colliculus, as described above, but did not label neurons projecting to the newly assayed pons. The newly tested layer 5 genes displayed a variety of additional patterns. Postn selectively labeled corticotrigeminal neurons, and very rarely corticopontine and corticotectal neurons. In contrast, Chrna6 did not label corticotrigeminal neurons, but rather predominantly labeled corticotectal neurons. Among corticofugal targets, Slc17a6 selectively labeled corticopontine neurons (in addition to all intracerebral populations). The majority of genes labeled at least a few neurons projecting to the striatum, and approximately half of the genes labeled neurons projecting to one or more cortical targets.

Several of the genes examined labeled exceedingly sparse neuronal populations in layer 5, and showed corticofugal target specificity. For example, the nicotinic acetylcholine receptor subunit Chrna6 labeled ∼8% (8.51 ± 3.17%) of corticotectal neurons projecting to the superior colliculus and ∼1% of neurons projecting to the striatum (0.55 ± 0.29%), thalamus (1.41 ± 1.41%), and pons (0.72 ± 0.72%), but did not label neurons projecting to the trigeminal nucleus or any of the cortical targets (Fig. 7c,e). Similarly, the Ca2+-dependent phospholipid-binding protein Annexin 1 (Anxa1) labeled ∼1% of neurons projecting to these same structures (0.86 ± 0.86%, 0.95 ± 0.95%, and 1.33 ± 0.83%, respectively), although with no increased proportion of corticotectal neurons (Fig. 7d,e). Therefore, these genes showed molecular specificity and heterogeneity within layer 5 neurons projecting to deep brain targets.

The intracortical projections displayed similar heterogeneity to the corticofugal projections (Fig. 7b). All genes labeling corticocortical projection neurons were expressed in at least some neurons projecting to frontal cortex, and several genes (Gprin3, Tmem200a, Kcnn2, Mylip, Sgcd) labeled only frontal-projecting neurons. With the remaining genes, various other combinations of targets were seen, including all cortical targets (e.g., Slc17a6, Trib2) and all cortical targets with the selective exclusion of M1 and S2 (Sv2c) or S2 only (Deptor, Fam3c). Taking a broad view, fifteen different gene/target patterns were seen just at the level of whether a gene labeled any neurons projecting to these targets. Four different combinatorial patterns were observed within the neocortex, 7 different subcortical patterns were seen among deep brain targets.

Target Specificity of Layer 2/3 Neurons Correlates with Molecular Specificity

We next examined whether selective gene expression in subpopulations of layer 2/3 neurons correlated with intracortical projection specificity. Using the strategy described above, colabeling experiments were performed for 19 genes with enriched layer 2/3 expression in S1 with retrograde labeling from ipsilateral S2, motor, and frontal cortex, as well as from contralateral (callosal) cortex (Fig. 8a). Ten of 19 of these genes labeled neurons projecting to all cortical targets tested (right side of Fig. 8b). The remaining genes labeled neurons projecting to specific targets, or, in one case (Glra3), none of the target areas tested. For example, D8Ertd82e selectively labeled S1 neurons projecting to the proximal area S2. A set of genes including Dusp18 and Grid2ip projected to all targets except motor cortex. Finally, Trpc6-positive neurons projected selectively to ipsilateral frontal cortex, with a small number of M1-projecting neurons labeled as well (Fig. 8c,d). 16.20 ± 2.61% of frontal cortex projecting neurons labeled were positive for Trpc6, compared with <1% (0.66 ± 0.66%) of motor cortex and no S2 or callosal projecting neurons (Fig. 8c). Therefore, this cation channel selectively labels a specific subset of frontal cortex projecting neurons. Taken together, 6 different gene/target patterns were seen for layer 2/3 marker genes, plus one gene that labeled neurons that did not project to targets analyzed in the current study.

Figure 8.

Target specificity of molecularly defined projection neuron classes in layer 2/3 of S1. (a) Schematic of different cortical targets of layer 2/3 projection neurons tested. (b) Summary of projection targets for layer 2/3 neurons expressing specific marker genes, based on CTB retrograde labeling with FISH for that gene. The approximate proportion of neurons projecting to each target labeled by each gene is indicated by color, including no labeling (white), sparse (light blue), low to medium (blue), and high (dark blue). Many genes labeled at least some neurons projecting to all cortical targets, while others selectively labeled neurons projecting to specific cortical regions or subsets of regions. For example, D8Ertd82e selectively labeled neurons projecting to ipsilateral S2, while Trpc6 predominantly labeled frontal cortex-projecting neurons and a small percentage of motor cortex-projecting neurons. Glra3 did not label neurons projecting to any of these target regions. (c,d) Target specificity for neurons expressing Trpc6. Double labeling with FISH (green) and CTB retrograde labeling (red) from frontal cortex. Left image is a low-magnification image spanning the thickness of S1, while the right panels show high-magnification single and double labeling corresponding to the boxed regions in layer 2/3. Arrows denote double-labeled cells. (e) Quantification of the percentage of retrogradely labeled neurons from each target expressing Trpc6, demonstrating specificity for frontal cortex (N = 3 for all cortical populations analyzed). Scale bars: low magnification: 250 μm, high magnification: 50 μm.

Figure 8.

Target specificity of molecularly defined projection neuron classes in layer 2/3 of S1. (a) Schematic of different cortical targets of layer 2/3 projection neurons tested. (b) Summary of projection targets for layer 2/3 neurons expressing specific marker genes, based on CTB retrograde labeling with FISH for that gene. The approximate proportion of neurons projecting to each target labeled by each gene is indicated by color, including no labeling (white), sparse (light blue), low to medium (blue), and high (dark blue). Many genes labeled at least some neurons projecting to all cortical targets, while others selectively labeled neurons projecting to specific cortical regions or subsets of regions. For example, D8Ertd82e selectively labeled neurons projecting to ipsilateral S2, while Trpc6 predominantly labeled frontal cortex-projecting neurons and a small percentage of motor cortex-projecting neurons. Glra3 did not label neurons projecting to any of these target regions. (c,d) Target specificity for neurons expressing Trpc6. Double labeling with FISH (green) and CTB retrograde labeling (red) from frontal cortex. Left image is a low-magnification image spanning the thickness of S1, while the right panels show high-magnification single and double labeling corresponding to the boxed regions in layer 2/3. Arrows denote double-labeled cells. (e) Quantification of the percentage of retrogradely labeled neurons from each target expressing Trpc6, demonstrating specificity for frontal cortex (N = 3 for all cortical populations analyzed). Scale bars: low magnification: 250 μm, high magnification: 50 μm.

Target Specificity of Layer 6 Neurons Correlates with Molecular Specificity

Excitatory neurons in layer 6 provide feedback to sensory thalamus, target different regions of the neocortex, and have been described as the most heterogeneous population of neurons in sensory cortex (Briggs 2010). To examine the relationship between gene expression and target specificity for these neurons, we combined expression of 20 layer 6 marker genes with retrograde labeling from ventral thalamus, contralateral cortex, and motor, and frontal cortex (Fig. 9a). As for other layers, we observed selective relationships between marker genes and projection targets. One set of genes specifically labeled corticothalamic neurons in deep layer 6, including Foxp2, and Tle4 (Fig. 9b). Labeling of corticothalamic neurons with Foxp2 is shown in Figure 9d, where ∼30% of these neurons were labeled (Fig. 9f). Other genes labeled both thalamic and corticocortical projection neurons (i.e., Sulf1, Cdh24). As corticocortical and corticothalamic neurons are known to be distinct (Zhang and Deschenes 1997), those genes that label neurons projecting to both targets must label multiple projection types rather than neurons with collaterals in cortical and thalamic targets. While Cyr61 and Nnat did not label neurons projecting to any targets analyzed, likely labeling neurons projecting to targets not tested in this study.

Figure 9.

Target specificity of molecularly defined projection neuron classes in layer 6 of S1. (a) Schematic of different cortical and subcortical targets of layer 6 projection neurons tested. (b) Summary of projection targets for layer 6 neurons expressing specific marker genes, based on CTB retrograde labeling with FISH for that gene. The approximate proportion of neurons projecting to each target labeled by each gene is indicated by color, including no labeling (white), sparse (light blue), low to medium (blue), and high (dark blue). Generally, individual genes labeled neurons projecting to specific combinations of possible projection targets. A subset of genes selectively labeled thalamocortical neurons, while others selectively labeled corticocortical neurons. Several genes selectively labeled neurons projecting to frontal cortex among the possible cortical targets. (cf) Target specificity for neurons expressing Gnb4 and Foxp2. Double labeling with FISH (green) and CTB retrograde labeling (red) from frontal cortex is shown for Gnb4 (c) and from thalamus for Foxp2 (d). Within each panel, the left image is a low-magnification image spanning the thickness of S1, while the right panels show high-magnification single and double labeling corresponding to the boxed regions in layer 6. (e,f) Quantification of the percentage of retrogradely labeled neurons from each target expressing Gnb4 and Foxp2 (mean ± SEM). Gnb4 selectively labels neurons projecting to ipsilateral cortical areas, while Foxp2 selectively labels corticothalamic neurons. For Gnb4 analysis, all cortical targets: N = 3, Thalamus: N = 1. For Foxp2 analysis, Frontal: N = 3, M1/callosal: N = 2, S2/Thalamus: N = 1. Scale bars in (c,d): low magnification: 250 μm, high magnification: 125 μm.

Figure 9.

Target specificity of molecularly defined projection neuron classes in layer 6 of S1. (a) Schematic of different cortical and subcortical targets of layer 6 projection neurons tested. (b) Summary of projection targets for layer 6 neurons expressing specific marker genes, based on CTB retrograde labeling with FISH for that gene. The approximate proportion of neurons projecting to each target labeled by each gene is indicated by color, including no labeling (white), sparse (light blue), low to medium (blue), and high (dark blue). Generally, individual genes labeled neurons projecting to specific combinations of possible projection targets. A subset of genes selectively labeled thalamocortical neurons, while others selectively labeled corticocortical neurons. Several genes selectively labeled neurons projecting to frontal cortex among the possible cortical targets. (cf) Target specificity for neurons expressing Gnb4 and Foxp2. Double labeling with FISH (green) and CTB retrograde labeling (red) from frontal cortex is shown for Gnb4 (c) and from thalamus for Foxp2 (d). Within each panel, the left image is a low-magnification image spanning the thickness of S1, while the right panels show high-magnification single and double labeling corresponding to the boxed regions in layer 6. (e,f) Quantification of the percentage of retrogradely labeled neurons from each target expressing Gnb4 and Foxp2 (mean ± SEM). Gnb4 selectively labels neurons projecting to ipsilateral cortical areas, while Foxp2 selectively labels corticothalamic neurons. For Gnb4 analysis, all cortical targets: N = 3, Thalamus: N = 1. For Foxp2 analysis, Frontal: N = 3, M1/callosal: N = 2, S2/Thalamus: N = 1. Scale bars in (c,d): low magnification: 250 μm, high magnification: 125 μm.

The majority of the remaining genes examined labeled cortically projecting neurons. Strikingly, these genes showed a great deal of specificity for particular cortical targets (Fig. 9b). These patterns included selective labeling of neurons projecting to individual targets or combinations of ipsilateral targets, ipsilateral and contralateral, or, in the case of Gm1441, only contralateral cortex. For example, Gnb4 labeled a subset of neurons projecting to ipsilateral frontal cortex (20.59 ± 13.07%), motor cortex (4.52 ± 3.17%), and S2 (7.69 ± 6.66%), while no labeling was observed for callosal or thalamic projections (Fig. 8c,e). Surprisingly, labeling of a many different combinations of individual cortical regions was observed for different genes. These patterns include frontal cortex only, M1/S2 only, frontal/M1/S2, frontal/M2/callosal, and frontal/callosal. Overall, 7 different cortical patterns were observed.

Discussion

While the diversity of cortical GABAergic neuronal classes has been intensively studied (Markram et al. 2004; Ascoli et al. 2008; Kubota et al. 2011), the complexity of cortical excitatory neuronal classes is not well characterized. The most well-described excitatory neuron classes to date are the pyramidal neurons in layer 5 (reviewed in (Molnar and Cheung 2006)), which have been divided into type I and type II cells. These types differ in their target specificity (intracerebral vs. corticofugal), dendritic morphology (tufted vs. nontufted), and electrophysiological properties (bursting vs. nonbursting), as well as their gene expression and developmental processes (Hevner et al. 2003). These coordinated differences among many phenotypic properties suggest that they represent functionally distinct cell classes that have unique roles in local processing within cortical circuits and in the information they transmit to distant targets. The current study aimed to examine the diversity of cortical excitatory projection neurons in mouse S1 on the basis of coordinated gene expression and projection target specificity. This approach demonstrates the existence of a wide variety of excitatory neuron classes with a strong correlation between these phenotypic features, suggesting a similar degree of complexity to GABAergic interneurons.

Combinatorial Complexity of Cellular Gene Expression Patterns

The phenotypic properties of different cortical neurons are a function of their underlying gene expression, which therefore provides in principle a powerful tool to discriminate functionally distinct cell classes as well as the means to genetically manipulate them (Doyle et al. 2008; Madisen et al. 2010; Taniguchi et al. 2011). The availability of genome-wide ISH data in adult mouse (Lein et al. 2007) made it possible to systematically identify all genes with heterogeneous expression patterns in S1. Analysis of these data allows an unusually broad perspective regarding potential cellular diversity on the basis of the diversity of molecular patterns across the transcriptome. Remarkably, among the 1053 genes selected for their robust, heterogeneous, predominantly neuronal patterns, we identified between 500 (density only) and 950 (density and level) different patterns. Undoubtedly, this result reflects both a diversity of cell types, and a combinatorial complexity whereby any given gene is expressed only in a subset of possible types. These patterns included expression in single cortical layers only, in nearly every possible combination of layers, in scattered (GABAergic, glial, or sparse excitatory) and/or in dense (excitatory) cell populations. Adding to this complexity, roughly 5–25% of genes in layers 2–6 displayed sublaminar expression patterns. This result suggests segregation of neuronal classes within a layer as previously described for projection classes in layer 5 and 6 (Zhang and Deschenes 1997; Molnar and Cheung 2006). While there may be caveats associated with subjectivity in manual scoring of expression patterns and technical issues related to probe sensitivity, there is undoubtedly a high degree of complexity in gene expression patterns that suggests a corresponding complexity of cortical cell classes.

To provide a direct path to genetic targeting of specific cell classes, in the current study we focused the analysis on genes with particularly enriched expression in a single cortical layer. Even within a given layer, a wide variety of cellular patterns were seen. This was particularly true in layer 5 where patterns ranged from high density throughout the depth of layer 5 to sparse and highly sublaminar within layer 5. Taking the approach used successfully to discriminate major GABAergic cell classes using marker genes (Kubota et al. 2011), we did systematic pairwise labeling among genes representing the spectrum of laminar patterns in layer 5. As for GABAergic markers, these laminar genes segregated into several largely nonoverlapping groups. Bcl6, Deptor, Trib2, and Slc17a8 labeled largely distinct subsets of layer 5 pyramidal neurons. The remaining genes tested mostly overlapped with Bcl6 to the exclusion of the other genes. This paradigm, involving broad, largely nonoverlapping molecular classes is very similar to that observed for GABAergic interneurons using the canonical markers Pvalb, Sst, and Vip, or more recently the serotonin receptor Htr3a (Kubota et al. 2011; Rudy et al. 2011). Somewhat surprisingly, we did not find evidence for pan-excitatory neuronal markers among genes with robust laminar specificity, but rather found that individual genes labeled subsets of excitatory neurons.

Gene Expression Correlates with Projection Neuron Target Specificity

Excitatory neurons in somatosensory cortex are well known to project to a wide variety of cortical and subcortical targets (Aronoff et al. 2010). However, it is not clear in many cases whether there are distinct projections to individual target structures or a few classes that send collaterals to multiple structures. For example, most corticofugal neurons in layer 5 send collaterals to multiple structures on the way to the spinal cord (Veinante et al. 2000), but it is unclear whether neurons projecting to different subsets of possible targets should be considered to be distinct functional classes. In some cases, there is evidence for multiple discrete corticofugal types. For example, corticothalamic and corticotrigeminal layer 5 neurons are mostly distinct and show different physiological properties (Hattox and Nelson 2007). Furthermore, individual corticocortical layer 5 neurons may have highly selective axonal projections with distinct, target-selective functional properties (Glickfeld et al. 2013). Finally, projection neurons with different broad target specificity have also been shown to have differential transcriptional profiles (Arlotta et al. 2005; Sugino et al. 2006; Chen et al. 2008; Han et al. 2011).

In the current study, we provide evidence that projection neuron target specificity correlates strongly with expression of specific genes. The largely nonoverlapping layer 5 genes derived from our DFISH study-labeled neurons with either intracerebral or subcortical projection targets. This phenomenon seemed to generalize across other genes and cortical layers, in that individual genes selectively labeled neurons projecting to specific targets or combinations of targets. By systematically labeling neurons from the majority of target regions for each layer (e.g., 9 targets for layer 5 neurons) and assaying a large cohort of genes (∼20 for each layer), a picture of the covariance and unanticipated combinatorial complexity of genes and targets emerged. From the gene perspective, most genes assayed only labeled neurons projecting to a subset of all possible target structures, and very rarely only to neurons projecting to a single target. From a target perspective, multiple (in some cases nonoverlapping) genes labeled neurons projecting to each target. In other words, multiple gene-defined classes exist for each target. In one particularly clear example, cells labeled by Deptor and Trib2, each projected selectively to the striatum and contralateral cortex, yet were largely nonoverlapping in their expression. Supporting this idea, double retrograde labeling studies indicate that multiple corticostriatal and callosal projection neuron classes exist (Molyneaux et al. 2009). Interestingly, while individual genes selectively labeled subsets of neurons projecting to a particular target, we did not observe cases where all target-defined cells were labeled by individual genes. The covariance of target specificity and gene expression suggest that the cell classes defined by the intersection of these 2 modalities are deterministic rather than stochastic. Although gene expression in projection neurons could be induced through interactions with their postsynaptic targets, finding molecular specificity for these projection neuron classes implies that they are genetically specified.

Several limitations should be acknowledged when using this approach. First, the retrograde tracers likely do not label all neurons projecting to a particular target. It is therefore not possible to determine if a gene labels the entire population, or if all gene-positive neurons project to certain targets. Rather, we could only determine what proportion of retrogradely labeled cells express a given gene. Furthermore, from these data, it is not possible to establish how many neuron classes, as defined by their target specificity, are labeled for any given gene. For example, a gene that labels 2 target-defined projection neuron populations may represent a single class with axon collaterals in both structures, or 2 classes that each project to a single target. Despite these caveats, gene expression strongly correlated with target specificity, and the number of gene/target patterns observed seems to set a lower bound on the number of projection neuron types in this region of the cortex. Similar complexity has been observed from single neuron fills (Zhang and Deschenes 1997; Veinante et al. 2000) and from double retrograde labeling (Hattox and Nelson 2007), with individual neurons projecting to subsets of all possible targets.

Gene/Target Projection Neuron Relationships

How many classes of gene/target-defined cortical neurons are there? Among 22 genes assayed in layer 5, we observed 15 different target patterns. Given the well-described division between intracerebral and corticofugal layer 5 neurons, those genes labeling both populations are likely expressed in multiple neuronal types. Treating these projection targets separately, we identified 7 different patterns among corticofugal targets and 4 patterns among neocortical targets tested. As many intracerebrally projecting neurons have collaterals in neocortex and striatum, the covariance between these structures likely reflects collateralization to a large degree rather than labeling of multiple projection neuron classes. As mentioned above, there may be another level of heterogeneity as well. Multiple genes with nonoverlapping expression label neurons projecting to the same targets (e.g., Deptor and Trib2), indicating that they label nonoverlapping functional populations. Perhaps not surprisingly, as the parcellation of projection neurons gets increasingly fine-grained, the populations become smaller. We identified several genes that label extremely sparse proportions of neurons projecting to specific targets. For example, the Ca2+-dependent phospholipid-binding protein Anxa1 labeled ∼1% of corticostriatal, corticothalamic, and corticotectal layer 5 neurons. Neurochemical specificity suggests some distinct function, but the role of such sparse cell classes remains to be elucidated and will be an interesting area for exploration using these genes to allow selective targeting and functional manipulation in the future.

Individual genes in layer 2/3 labeled neurons projecting to specific cortical areas and combinations of areas as well. Seven different corticocortical gene/target patterns were seen among the 19 genes assayed. The most specific pattern observed was for the uncharacterized transcript D8Ertd82e, which only labeled neurons projecting to neighboring S2. Trpc6 displayed selectivity for frontal cortex-projecting neurons, although only a subset (∼20%) of these neurons was labeled. The other major pattern involved labeling of neurons projecting to S2, frontal, and callosal targets but not to M1. One gene clearly expressed in layer 2/3 excitatory neurons, the glycine receptor Glra3, did not label neurons projecting to any target tested, indicating specificity for additional S1 targets such as the perirhinal cortex (Aronoff et al. 2010). These data imply a great degree of specificity in information transfer along different functional cortical streams.

Within layer 6, corticothalamic and corticocortical neurons comprise distinct cell classes, and multiple subtypes of corticothalamic and corticocortical cells have been defined from single neuron fills on the basis of target specificity and correlated dendritic morphology (Zhang and Deschenes 1997). A recent study also identified several genes differentially expressed between corticocortical and corticothalamic layer 6 neurons (Watakabe et al. 2012). We found that genes selectively expressed in layer 6 showed particularly striking target specificity, either for corticothalamic neurons or subsets of cortical areas. A number of genes, including Foxp2 and the cholinergic receptor Chrna5, have robust, selective expression in corticothalamic neurons. Despite its specificity for corticothalamic neurons, Foxp2 only labeled ∼30% of retrogradely labeled neurons, indicating another level of heterogeneity among corticothalamic neurons. This gene could perhaps discriminate between corticothalamic neurons that project to the posteromedial nucleus alone or to the posterior group also (Zhang and Deschenes 1997). Many genes were expressed selectively in layer 6 neurons projecting to specific cortical areas as well. As with the other layers, several layer 6 genes were selectively expressed in frontal cortex-projecting neurons. Two genes selectively expressed in the lateral portion of S1 layer 6 were not labeled from any target tested, again indicating specificity for additional projection targets. In total, 8 gene/target patterns were seen in layer 6, assuming that combinations of corticocortical and corticothalamic patterns represent labeling of 2 different cell classes.

Taken as a whole, gene expression combined with target specificity imply a greater diversity of projection neuron classes than generally appreciated, and the covariance of these 2 phenotypic modalities suggests that these classes are discrete and genetically specified. These observations and the identification of marker genes for specific projection types should provide powerful tools for genetic targeting and manipulation of specific functional pathways (Doyle et al. 2008; Madisen et al. 2010; Taniguchi et al. 2011). On the other hand, few genes appeared to label neurons projecting to individual target structures. Although there are technical limitations associated with the single retrograde labeling approach used here, this result suggests that most discrete projection neuron classes send axon collaterals to multiple structures. If this is true, it has implications for cell-type-specific functional manipulation with Cre mouse lines and genetic tools, as it will not permit genetic targeting of neurons projecting to specific individual target structures. In this light, it is critical to understand the diversity of projection neuron types, including defining the number of discrete classes based on their specific axonal projections (to single or multiple targets), their full transcriptional profiles, and their unique morphological, physiological, and connectional properties at the single cell level.

Supplementary Material

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

Notes

The authors thank the Allen Institute founders, Paul G. Allen and Jody Allen, for their vision, encouragement, and support. The authors acknowledge Theresa Zwingman and Maureen Boyle for mining efforts to identify laminar expression patterns in S1; Jolene Kidney, Maureen Howell, Andrew Boe, Tracy Lemon, Naveed Masten, Cecilli Smith, and Thanh Quanh for assistance with tract tracing and sectioning; Nick Dee and Zack Riley for methods development related to combined ISH and retrograde labeling; Amanda Ebbert, Lon Luong, Kimberly Smith, Melissa Reding, and Shiella Caldejon, Sheana Parry, and Emi Byrnes for supporting production and scanning of fluorescent ISH images; Nathanael Motz and Chinh Dang for supporting database needs; and Paul Wohnoutka for supporting data production. Conflict of Interest: None declared.

References

Arlotta
P
Molyneaux
BJ
Chen
J
Inoue
J
Kominami
R
Macklis
JD
Neuronal subtype-specific genes that control corticospinal motor neuron development in vivo
Neuron
 , 
2005
, vol. 
45
 (pg. 
207
-
221
)
Aronoff
R
Matyas
F
Mateo
C
Ciron
C
Schneider
B
Petersen
CC
Long-range connectivity of mouse primary somatosensory barrel cortex
Eur J Neurosci
 , 
2010
, vol. 
31
 (pg. 
2221
-
2233
)
Ascoli
GA
Alonso-Nanclares
L
Anderson
SA
Barrionuevo
G
Benavides-Piccione
R
Burkhalter
A
Buzsaki
G
Cauli
B
Defelipe
J
Fairen
A
, et al.  . 
Petilla terminology: nomenclature of features of GABAergic interneurons of the cerebral cortex
Nat Rev Neurosci
 , 
2008
, vol. 
9
 (pg. 
557
-
568
)
Belgard
TG
Marques
AC
Oliver
PL
Abaan
HO
Sirey
TM
Hoerder-Suabedissen
A
Garcia-Moreno
F
Molnar
Z
Margulies
EH
Ponting
CP
A transcriptomic atlas of mouse neocortical layers
Neuron
 , 
2011
, vol. 
71
 (pg. 
605
-
616
)
Bernard
A
Lubbers
LS
Tanis
KQ
Luo
R
Podtelezhnikov
AA
Finney
EM
McWhorter
MM
Serikawa
K
Lemon
T
Morgan
R
, et al.  . 
Transcriptional architecture of the primate neocortex
Neuron
 , 
2012
, vol. 
73
 (pg. 
1083
-
1099
)
Briggs
F
Organizing principles of cortical layer 6
Front Neural Circuits
 , 
2010
, vol. 
4
 pg. 
3
 
Chakrabarti
S
Alloway
KD
Differential origin of projections from SI barrel cortex to the whisker representations in SII and MI
J Comp Neurol
 , 
2006
, vol. 
498
 (pg. 
624
-
636
)
Chen
B
Wang
SS
Hattox
AM
Rayburn
H
Nelson
SB
McConnell
SK
The Fezf2-Ctip2 genetic pathway regulates the fate choice of subcortical projection neurons in the developing cerebral cortex
Proce Natl Acad Sci U S A
 , 
2008
, vol. 
105
 (pg. 
11382
-
11387
)
Davis
FP
Eddy
SR
A tool for identification of genes expressed in patterns of interest using the Allen Brain
Atlas. Bioinformatics
 , 
2009
, vol. 
25
 (pg. 
1647
-
1654
)
Dong
HW
Allen reference atlas: a digital color brain atlas of the C57Black/6J male mouse
 , 
2008
Hoboken, NJ
Wiley
Doyle
JP
Dougherty
JD
Heiman
M
Schmidt
EF
Stevens
TR
Ma
G
Bupp
S
Shrestha
P
Shah
RD
Doughty
ML
, et al.  . 
Application of a translational profiling approach for the comparative analysis of CNS cell types
Cell
 , 
2008
, vol. 
135
 (pg. 
749
-
762
)
Glickfeld
LL
Andermann
ML
Bonin
V
Reid
RC
Cortico-cortical projections in mouse visual cortex are functionally target specific
Nat Neurosci
 , 
2013
, vol. 
16
 (pg. 
219
-
226
)
Gupta
A
Wang
Y
Markram
H
Organizing principles for a diversity of GABAergic interneurons and synapses in the neocortex
Science
 , 
2000
, vol. 
287
 (pg. 
273
-
278
)
Han
W
Kwan
KY
Shim
S
Lam
MM
Shin
Y
Xu
X
Zhu
Y
Li
M
Sestan
N
TBR1 Directly represses Fezf2 to control the laminar origin and development of the corticospinal tract
Proc Natl Acad Sci U S A
 , 
2011
, vol. 
108
 (pg. 
3041
-
3046
)
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
)
Hevner
RF
Daza
RA
Rubenstein
JL
Stunnenberg
H
Olavarria
JF
Englund
C
Beyond laminar fate: toward a molecular classification of cortical projection/pyramidal neurons
Dev Neurosci
 , 
2003
, vol. 
25
 (pg. 
139
-
151
)
Hoerder-Suabedissen
A
Wang
WZ
Lee
S
Davies
K
Goffinet
A
Rakić
S
Parnavelas
J
Reim
K
Nicolić
M
Paulsen
O
, et al.  . 
Novel markers reveal subpopulations of subplate neurons in the murine cerebral cortex
Cereb Cortex
 , 
2009
, vol. 
19
 (pg. 
1738
-
1750
)
Kawaguchi
Y
Kondo
S
Parvalbumin, somatostatin and cholecystokinin as chemical markers for specific GABAergic interneuron types in the rat frontal cortex
J Neurocytol
 , 
2002
, vol. 
31
 (pg. 
277
-
287
)
Ko
Y
Ament
SA
Eddy
JA
Caballero
J
Earls
JC
Hood
L
Price
ND
Cell type-specific genes show striking and distinct patterns of spatial expression in the mouse brain
Proc Natl Acad Sci
 , 
2013
, vol. 
110
 (pg. 
3095
-
3100
)
Kubota
Y
Shigematsu
N
Karube
F
Sekigawa
A
Kato
S
Yamaguchi
N
Hirai
Y
Morishima
M
Kawaguchi
Y
Selective coexpression of multiple chemical markers defines discrete populations of neocortical GABAergic neurons
Cereb Cortex
 , 
2011
, vol. 
21
 (pg. 
1803
-
1817
)
Lein
ES
Hawrylycz
MJ
Ao
N
Ayres
M
Bensinger
A
Bernard
A
Boe
AF
Boguski
MS
Brockway
KS
Byrnes
EJ
, et al.  . 
Genome-wide atlas of gene expression in the adult mouse brain
Nature
 , 
2007
, vol. 
445
 (pg. 
168
-
176
)
Madisen
L
Zwingman
TA
Sunkin
SM
Oh
SW
Zariwala
HA
Gu
H
Ng
LL
Palmiter
RD
Hawrylycz
MJ
Jones
AR
, et al.  . 
A robust and high-throughput Cre reporting and characterization system for the whole mouse brain
Nat Neurosci
 , 
2010
, vol. 
13
 (pg. 
133
-
140
)
Markram
H
Toledo-Rodriguez
M
Wang
Y
Gupta
A
Silberberg
G
Wu
C
Interneurons of the neocortical inhibitory system
Nat Rev Neurosci
 , 
2004
, vol. 
5
 (pg. 
793
-
807
)
Mitchell
BD
Macklis
JD
Large-scale maintenance of dual projections by callosal and frontal cortical projection neurons in adult mice
J Comp Neurol
 , 
2005
, vol. 
482
 (pg. 
17
-
32
)
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
)
Molyneaux
BJ
Arlotta
P
Macklis
JD
Molecular development of corticospinal motor neuron circuitry
Novartis Found Symp
 , 
2007
, vol. 
288
 (pg. 
3
-
15
discussion 15–20, 96–18
Molyneaux
BJ
Arlotta
P
Menezes
JR
Macklis
JD
Neuronal subtype specification in the cerebral cortex
Nat Rev Neurosci
 , 
2007
, vol. 
8
 (pg. 
427
-
437
)
Montiel
J
Wang
WZ
Oeschger
F
Hoerder-Suabedissen
A
Tung
WL
García-Moreno
F
Holm
IE
Villalón
A
Molnár
Z
Hypothesis on the dual origin of the mammalian subplate
Front Neuroanat
 , 
2011
, vol. 
5
 
Oeschger
F
Wang
W-Z
Lee
S
García-Moreno
F
Goffinet
A
Arbonés
M
Rakic
S
Molnár
Z
Gene expression analysis of the embryonic subplate
Cereb Cortex
 , 
2012
, vol. 
22
 (pg. 
1343
-
1359
)
Paxinos
G
Franklin
KBJ
The mouse brain in stereotaxic coordinates
 , 
2004
Amsterdam, Boston
Elsevier Academic Press
Reiner
A
Jiao
Y
Del Mar
N
Laverghetta
AV
Lei
WL
Differential morphology of pyramidal tract-type and intratelencephalically projecting-type corticostriatal neurons and their intrastriatal terminals in rats
J Comp Neurol
 , 
2003
, vol. 
457
 (pg. 
420
-
440
)
Rudy
B
Fishell
G
Lee
S
Hjerling-Leffler
J
Three groups of interneurons account for nearly 100% of neocortical GABAergic neurons
Dev Neurobiol
 , 
2011
, vol. 
71
 (pg. 
45
-
61
)
Sugino
K
Hempel
CM
Miller
MN
Hattox
AM
Shapiro
P
Wu
C
Huang
ZJ
Nelson
SB
Molecular taxonomy of major neuronal classes in the adult mouse forebrain
Nat Neurosci
 , 
2006
, vol. 
9
 (pg. 
99
-
107
)
Taniguchi
H
He
M
Wu
P
Kim
S
Paik
R
Sugino
K
Kvitsiani
D
Fu
Y
Lu
J
Lin
Y
, et al.  . 
A resource of Cre driver lines for genetic targeting of GABAergic neurons in cerebral cortex
Neuron
 , 
2011
, vol. 
71
 (pg. 
995
-
1013
)
Tebbenkamp
AT
Borchelt
DR
Analysis of chaperone mRNA expression in the adult mouse brain by meta analysis of the Allen Brain Atlas
PLoS One
 , 
2010
, vol. 
5
 pg. 
e13675
 
Thompson
CL
Pathak
SD
Jeromin
A
Ng
LL
MacPherson
CR
Mortrud
MT
Cusick
A
Riley
ZL
Sunkin
SM
Bernard
A
, et al.  . 
Genomic anatomy of the hippocampus
Neuron
 , 
2008
, vol. 
60
 (pg. 
1010
-
1021
)
Veinante
P
Lavallee
P
Deschenes
M
Corticothalamic projections from layer 5 of the vibrissal barrel cortex in the rat
J Comp Neurol
 , 
2000
, vol. 
424
 (pg. 
197
-
204
)
Voelker
CC
Garin
N
Taylor
JS
Gähwiler
BH
Hornung
JP
Molnár
Z
Selective neurofilament (SMI-32, FNP-7 and N200) expression in subpopulations of layer V pyramidal neurons in vivo and in vitro
Cereb Cortex
 , 
2004
, vol. 
14
 (pg. 
1276
-
1286
)
Watakabe
A
Hirokawa
J
Ichinohe
N
Ohsawa
S
Kaneko
T
Rockland
KS
Yamamori
T
Area-specific substratification of deep layer neurons in the rat cortex
J Comp Neurol
 , 
2012
, vol. 
520
 (pg. 
3553
-
3573
)
Yoneshima
H
Yamasaki
S
Voelker
CC
Molnár
Z
Christophe
E
Audinat
E
Takemoto
M
Nishiwaki
M
Tsuji
S
Fujita
I
, et al.  . 
Er81 is expressed in a subpopulation of layer 5 neurons in rodent and primate neocortices
Neuroscience
 , 
2006
, vol. 
137
 (pg. 
401
-
412
)
Zhang
ZW
Deschenes
M
Intracortical axonal projections of lamina VI cells of the primary somatosensory cortex in the rat: a single-cell labeling study
J Neurosci
 , 
1997
, vol. 
17
 (pg. 
6365
-
6379
)