Long‐range projections from sparse populations of GABAergic neurons in murine subplate

Abstract The murine subplate contains some of the earliest generated populations of neurons in the cerebral cortex, which play an important role in the maturation of cortical inhibition. Here we present multiple lines of evidence, that the subplate itself is only very sparsely populated with GABAergic neurons at postnatal day (P)8. We used three different transgenic mouse lines, each of which labels a subset of GABAergic, ganglionic eminence derived neurons. Dlx5/6‐eGFP labels the most neurons in cortex (on average 11% of NEUN+ cells across all layers at P8) whereas CGE‐derived Lhx6‐Cre::Dlx1‐Venusfl cells are the sparsest (2% of NEUN+ cells across all layers at P8). There is significant variability in the layer distribution of labeled interneurons, with Dlx5/6‐eGFP and Lhx6‐Cre::R26R‐YFP being expressed most abundantly in Layer 5, whereas CGE‐derived Lhx6‐Cre::Dlx1‐Venusfl cells are least abundant in that layer. All three lines label at most 3% of NEUN+ neurons in the subplate, in contrast to L5, in which up to 30% of neurons are GFP+ in Dlx5/6‐eGFP. We assessed all three GABAergic populations for expression of the subplate neuron marker connective tissue growth factor (CTGF). CTGF labels up to two‐thirds of NEUN+ cells in the subplate, but was never found to colocalize with labeled GABAergic neurons in any of the three transgenic strains. Despite the GABAergic neuronal population in the subplate being sparse, long‐distance axonal connection tracing with carbocyanine dyes revealed that some Gad65‐GFP+ subplate cells form long‐range axonal projections to the internal capsule or callosum.


| INTRODUCTION
Subplate neurons, located between the white matter and layer 6a, are amongst the earliest born neurons of the mouse cortex (Angevine & Sidman, 1961;Bystron, Blakemore, & Rakic, 2008;Price, Aslam, Tasker, & Gillies, 1997;Smart, Dehay, Giroud, Berland, & Kennedy, 2002). Whereas many subplate neurons are generated in the ventricular zone of the cortical neuroepithelium, the subplate compartment also contains cells derived from the rostral medial telencephalic wall (RMTW; García-Moreno, López-Mascaraque, & De Carlos, 2008;Pedraza, Hoerder-Suabedissen, Albert-Maestro, Molnár, & De Carlos, 2014)-to date the only brain structure that has been shown to give rise to both interneurons and projection neurons in rodents. The RMTW region gives rise to particularly early born neurons (neurogenesis is finished by E12.5 in this region) that migrate tangentially into the upper part of the cortical neuroepithelium, where they differentiate into projection neurons eventually located within the subplate and GABAergic neurons in the infragranular layers (but excluding the subplate).
Together, this can be taken as evidence of mature GABAergic neurons located within the young postnatal rodent subplate.
In addition to their electrophysiological characterization, subplate neurons have been identified by location within the cortical plate, by their early birth date and their enriched or selective gene expression (Allendoerfer & Shatz, 1994;Kanold & Luhmann, 2010). For embryonic mouse brains, there are at least nine genes with subplate restricted cortical expression and expression in the "superplate" in Reeler brains (Oeschger et al., 2012). Of these, RCAN2 protein colocalizes with Gad67-GFP in the embryonic (E17.5) and postnatal (P8) subplate (Oeschger et al., 2012), but it only labels a small proportion of all cells in the subplate. For postnatal mouse brains, there are at least an additional five molecules identified that label cells in the subplate region in normal brains, and cells in the superplate in Reeler brains. These include CTGF, NURR1, CPLX3, Tmem193, and MoxD1  (Akbarian et al., 1996). These NADPH diaphorase positive cells develop by 15 GW in human frontal cortex, in the emergent subplate (Yan, Garey, & Jen, 1996). NADPH diaphorase positive cells in cat white matter and layer 6b can have long-range projections (Higo, Udaka, & Tamamaki, 2007), and long-range projecting, nNOS+ cells were also found in the white matter of monkeys (Swiegers et al., 2018;Tomioka & Rockland, 2007).
Here we show, that the most abundant postnatally expressed subplate marker-connective tissue growth factor (CTGF, also known as CCN2), is not present in GABAergic interneurons. CTGF is therefore presumed to be expressed in glutamatergic projection neurons. We quantified the distribution of labeled caudal ganglionic eminence (CGE)derived Lhx6-Cre::Dlx1-Venus fl GABAergic neurons, medial ganglionic eminence (MGE)-derived Lhx6-Cre::R262R-YFP GABAergic neurons and lateral ganglionic eminence (LGE)/MGE-derived Dlx5/6-IRES-eGFP GABAergic neurons in the mouse primary somatosensory cortex at postnatal day (P)8. Surprisingly, we present data demonstrating that GABAergic neurons are less common in subplate than in other cortical layers in the P8 mouse. None-the-less, some of the sparse GABAergic neurons in the subplate as labeled by Gad65-GFP possess long-range axonal projections.

| Immunohistochemistry
Immunohistochemistry was done on sections from 4% PFA-fixed brains cut coronally at 50 μm on a vibrating microtome (VT1000S, Leica) and stored in 0.1 M PBS with 0.05% sodium azide until use.
Fluorescent triple immunohistochemistry was done on three to four free-floating sections per brain and antibody combination, all in the region of primary somatosensory cortex.

| Definition of subplate and other cortical layers
For the purposes of this study, the subplate zone or layer was defined as a 50 μm thick band directly above the white matter as evident from DAPI or NEUN staining (Hoerder-Suabedissen & , and is equivalent to layer 6b. This corresponds well to the region labeled by CCN2/CTGF, which only rarely labels cells in the underlying white matter. To assess the distribution of GFP or Venus labeled neurons across the entire cortical gray matter depth in primary somatosensory cortex (S1), 50 μm thick bands were placed within each cytoarchitecturally defined layer (i.e. Layer 2-6), well away from the layer boundaries. DAPI counterstaining or NEUN immunohistochemistry was used to identify layer boundaries.

| Imaging and analysis
For analysis of GFP+ cell distribution or colocalization of CTGF and GFP, image stacks along the white matter-cortex boundary or through the entire depth of cortex were acquired on a confocal laser scanning microscope (LSM710, Zeiss) in primary sensory cortex. Imaging intensity and filter-cutoffs were selected to minimize bleed-through. All images were subsequently imported into Photoshop CS3 and intensity and contrast adjusted.
To determine the proportion of GFP+ interneurons in each layer, NEUN+ cells that were also GFP+ (or Venus+) were identified. Quantification was based on one to three 50 μm thick bands per layer. To determine the proportion of CTGF+ subplate cells that were also interneurons, NEUN+CTGF+ that were also GFP+ (or Venus+) were identified.
For the analysis of carbocyanine-labeled sections, an epifluorescence microscope (DMR; Leica, Germany) was used. DiI-labeled cells were identified and analyzed for GFP fluorescence. There was no visible bleed-through of either GFP onto the cy3 filter used for DiI or of DiI into the cy2 filter used for GFP on this microscope.

Figures for publication were assembled and contrast and intensity
adjusted as a whole using Photoshop CS3 and later versions.

| Enrichment of interneuron markers in the subplate-microarray data
The rodent subplate layer contains a diverse population of cells, containing both pallium derived projection neurons (TBR1+, presumed glutamatergic) and ganglionic eminence derived interneurons (GABAergic; Hevner & Zecevic, 2006). Moreover, it is one of several compartments in which ganglionic eminence derived interneurons migrate tangentially during cortical development (Parnavelas, 2000) and recently subplate has been implicated in regulating radial migration (Ohtaka-Maruyama et al., 2018). We have previously used microarray profiling of the mouse subplate layer at E15, E18 and P8 with the aim of identifying different cell types within the subplate . Of the commonly used interneuron molecular markers (see Supporting Information Table S1 for full list), Gad1 (Glutamate decarboxylase 1 [brain, 67 kDa]; encodes Gad67), Cck (Cholecystokinin) and Sst (Somatostatin) are expressed at higher levels in the subplate compared to adjacent cortical plate or Layer 6 in embryonic development, but not postnatally, based on the previously published microarray studies in mouse (Hoerder-Suabedissen et al., 2009Oeschger et al., 2012). Cck is expressed in a subplate-enriched pattern from E14 to E18, but more broadly in subplate and L6 by P4. Similarly, Sst expression is primarily in the preplate derived structures of the subplate and marginal zone at E15. Sst expressing cells are more abundant in the subplate than in other cortical layers at E18, but by P4, Sst expressing cells are distributed across all infragranular layers, and SST+ cells are essentially absent from subplate by P8 Marques-Smith et al., 2016). Gad1 expressing cells are sparsely present in the anterior subplate at E15 and E18, but present throughout the cortex by P4.
The calcium-binding protein Calretinin, often found in GABAergic cells, has been historically used as a subplate marker during embryonic cortical development, because it labels cells in the preplate derivative structures subplate and marginal zone in the E16 rat cortex (Fonseca, DelRio, Martinez, Gomez, & Soriano, 1995).
Calretinin is expressed in the rat subplate from E16 to P3, but expression levels decline sharply after that time point (Fonseca et al., 1995). However, the above gene expression profiling did not identify Calb2 as differentially expressed between subplate and Layer 6 (or cortical plate) at any of the developmental ages analyzed. In support of this, Figure 1 shows that Calbindin (CB), but not Calretinin cells are relatively more abundant in the E19 mouse subplate than in the cortical plate.
On the other hand, most of the molecular markers of subplate neurons identified by the above gene expression profiling approach label non-GABAergic cells in mouse. Both UNC5C and CDH10 expression is restricted to the embryonic subplate, but neither protein is detectable in GABAergic cells (Oeschger et al., 2012). Similarly, NURR1 and CPLX3 label partially overlapping populations of postnatal subplate neurons in mouse, but neither is detectable in GABAergic cells (Hoerder-Suabedissen et al., 2009). Furthermore, the CB+ cells located in the E19 subplate are neither CTGF nor Lpar1-eGFP positive (inset Figure 1; not quantified). Colocalization with CPLX3 was not  (Butt et al., 2007;Liodis et al., 2007). Lhx6-Cre::R26R-YFP labels all MGE-derived GABAergic neurons with yellow fluorescent protein (YFP; Fogarty et al., 2007).
Distal-less homeobox 1 (Dlx1) is expressed in GABAergic neuron progenitors as well as migrating GABAergic neurons (Butt et al., 2007). In the Dlx1-Venus fl transgenic mouse line, the green fluorescent protein Venus is expressed in all cells derived from subpallial progenitor zones including medial, lateral and caudal ganglionic eminence.
However, in the presence of CRE-recombinase, driven by the MGEexpressed Lhx6 promoter, Venus is excised, giving a subtractive labeling of cells from the LGE/CGE only (Rubin et al., 2010).
Dlx5/6-eGFP, contains the transgene Cre-IRES2-eGFP under the mouse distal-less homeobox5/6 enhancer id6/id5. eGFP expression from this enhancer is restricted to the SVZ and mantle zone of the MGE and LGE (Stenman et al., 2003), and labels postmitotic neurons in their final location in cortex, too.     Of these, none were CTGF positive (see Figure 3 for an example).
Thus, CTGF is a subplate marker that labels non-GABAergic, presumed glutamatergic projection neurons in the postnatal subplate.
As CTGF protein levels only become detectable in the subplate from E18, it is unlikely to be labeling GABAergic neurons migrating within the subplate at younger ages, but this was not confirmed.

| Some of the GABAergic cells in the subplate have long-range projections
Despite the sparsity of interneurons within the mouse subplate, others have demonstrated that some long-range, ipsilaterally projecting layer 6b cells are Gad67-GFP positive in transgenic mice (Tomioka et al., 2005) and long-range projecting GABAergic neurons have also been described in the cat (Higo et al., 2007) and monkey layer 6b (Tomioka & Rockland, 2007).
Here we assessed whether Gad65-GFP expressing GABAergic neurons in the subplate and white matter also form long-range projec- were GABAergic neurons, compared to 2.4% AE 1.0 (n = 16/732) at P7. Conversely, the percentage of GABAergic, subplate cells with an axon in the midline corpus callosum (presumed to project to the contralateral hemisphere) went up from 4.4% AE 2.1 (n = 5/117) at P2 to 9.4% AE 5.0 (n = 18/251) at P7.
When placing DiI in the internal capsule or the cortex, the smallest possible crystal was used, while relatively large crystals were used for DiI placements into the white matter/ corpus callosum, irrespective of the age or tissue type used.
In conclusion, some long-range projecting cells of the mouse subplate are Gad65-GFP+, and therefore likely to be GABAergic neurons, Our study revealed relatively low density of interneurons within the P8 murine subplate. Overall, it is believed that the mouse neocortex consists of approximately 20% interneurons (Gonchar & Burkhalter, 1997), compared to the 25% to 30% found in primate visual cortex (Jones, 1993). Overall, the density of interneurons is higher in Layers 2-5 than in Layer 1 (Gonchar & Burkhalter, 1997;Xu, Roby, & Callaway, 2010). In agreement with this, Tamamaki et al. (2003) reported that 19% of NEUN+ neocortical neurons are GFP+ in the Gad67-GFP knock-in mouse line (Tamamaki et al., 2003). However, all of these studies restricted their quantification to layer 6a and the thin, compact layer of 6b has not been specifically investigated. The only evidence for sparse GABAergic neuron presence in the subplate prior to our current study comes from Hevner and Zecevic (2006). They report that subplate "projection" neurons (TBR1+) are born between E10.5 and E13.5 whereas GABAergic cells are born both early and late (E10.5 to E16.5). Assuming that BrdU labeling is distributed evenly across both cell types, summing up all the labeled cells in their Figure suggests that TBR1+ cells outnumber GABA+ cells by at least 10 to 1 at birth (Hevner & Zecevic, 2006), which would be more in agreement with our findings. Here we report that less than 5% of neurons in the mouse subplate are GABAergic in each of three transgenic lines at P8. This is likely to still be an overestimate, as our confocal images were taken of specifically selected areas that included at least one GFP+ cell in the subplate. Similarly, when labeling against GABA, the neurotransmitter released by GABAergic neurons, only 6% of NEUN+ neurons in the subplate are labeled. Lastly, the subplate layer appears unusual in the sense that most labeled GABAergic neurons in the subplate were identified in the Lhx6-Cre::Dlx1-Venus fl mouse line, in which overall the fewest GABAergic neurons were labeled.
We quantified only NEUN+ neurons for this analysis, as we noted a relatively large number of GABAergic-marker positive cells with the distinct morphology of migrating cells at the lower border of the subplate. This suggests that cell migration from the SVZ toward the rostral migratory stream is not entirely restricted to the white matter, and that cells can occasionally enter the overlying subplate. We therefore aimed to exclude migratory GABAergic neurons by colabeling with NEUN. This allowed us to determine the overall distribution of labeled GABAergic neurons out of total neurons as well (see Figure 2).
CTGF only labels a small region of the cytoplasm in most neurons, usually between nucleus and base of the primary dendrite. Therefore, we analyzed colocalization on image stacks to exclude cellular arrangements in which the CTGF is derived from one cell but closely apposed to another cell. However, we only found one example of closely adjacent, but nonoverlapping cells and no genuine colocalization between CTGF and any of the GABAergic markers or transgenic labels. This strongly suggests that CTGF is not coexpressed in GABAergic cells of the mouse subplate.
Even in early brain development, nonmigratory GABAergic neurons may not be abundant in the subplate. Of the many subplateselectively expressed genes in embryonic development, only RCAN2 could be demonstrated to colocalize with Gad65-GFP occasionally (Oeschger et al., 2012). Calretinin, often considered a GABAergic marker, labels GABAergic cells in the E16 rat cortical plate, but non-GABAergic cells transiently in the subplate (Fonseca et al., 1995).
In light of the unusual mouse subplate composition, it would be interesting to investigate subplate in large-brained species in more detail (see Swiegers et al., 2018). Previously, it has been noted that NADPH diaphorase labeled cells are largely GABAergic, but specifically within the primate white matter, the majority of NADPHd cells are not GABAergic. Moreover, NADPHd cells within the remaining cortex are often identified as Martinotti cells (Estrada & DeFelipe, 1998;Yan et al., 1996). This in turn, links the interstitial white matter NADPHd cells with Lpar1-eGFP subplate cells in the mouse, as the latter transgene is also expressed in cortical Martinotti cells