The origin of neocortical nitric oxide synthase-expressing inhibitory neurons

Inhibitory neurons are critical for regulating effective transfer of sensory information and network stability. The precision of inhibitory function likely derives from the existence of a variety of interneuron subtypes. Their specification is largely dependent on the locale of origin of interneuron progenitors. Neocortical and hippocampal inhibitory neurons originate the subpallium, namely in the medial and caudal ganglionic eminences (MGE and CGE), and in the preoptic area (POA). In the hippocampus, neuronal nitric oxide synthase (nNOS)-expressing cells constitute a numerically large GABAergic interneuron population. On the contrary, nNOS-expressing inhibitory neurons constitute the smallest of the known neocortical GABAergic neuronal subtypes. The origins of most neocortical GABAergic neuron subtypes have been thoroughly investigated, however, very little is known about the origin of, or the genetic programs underlying the development of nNOS neurons. Here, we show that the vast majority of neocortical nNOS-expressing neurons arise from the MGE rather than the CGE. Regarding their molecular signature, virtually all neocortical nNOS neurons co-express the neuropeptides somatostatin (SST) and neuropeptide Y (NPY), and about half of them express the calcium-binding protein calretinin (CR). nNOS neurons thus constitute a small cohort of the MGE-derived SST-expressing population of cortical inhibitory neurons. Finally, we show that conditional removal of the transcription factor Sox6 in MGE-derived GABAergic cortical neurons results in an absence of SST and CR expression, as well as reduced expression of nNOS in neocortical nNOS neurons. Based on their respective abundance, origin and molecular signature, our results suggest that neocortical and hippocampal nNOS GABAergic neurons likely subserve different functions and have very different physiological relevance in these two cortical structures.


INTRODUCTION
Brain activity is regulated by the interaction of two major types of neural cells: excitatory neurons that use the neurotransmitter glutamate and inhibitory neurons that use the neurotransmitter γ-aminobutyric acid (GABA). Inhibitory neurons are critical for maintaining the excitatory-inhibitory balance necessary for the effective transfer of information while preventing runaway excitation, and consequently play important roles in regulating network activity (Pouille and Scanziani, 2001;Markram et al., 2004;Haider and McCormick, 2009). Severe disruption of GABAergic inhibition profoundly alters neocortical activity patterns and leads to seizures, while milder changes in inhibitory neuron number and function are strongly linked to psychiatric disorders, such as schizophrenia and autism (Volk et al., 2002;Yau et al., 2003;Belmonte et al., 2004;Levitt et al., 2004;Cossart et al., 2005;Dani et al., 2005;Caceda et al., 2007;Gonzalez-Burgos and Lewis, 2008;Morris et al., 2008). Inhibition is thus hypothesized to be a critical regulator of normal brain function and a key cause of dysfunction in disease. Inhibitory function relies on the existence of a variety of GABAergic inhibitory neuronal subtypes.
Most inhibitory neurons display locally projecting axons and have been consequently named interneurons. Besides this morphological variety, different classes of inhibitory neurons exhibit distinct intrinsic membrane properties, molecular markers, connectivity, and synaptic specializations, and therefore likely mediate specific roles within cortical circuits . In the murine system, although many characteristics are not acquired until several weeks postnatally, the specification of GABAergic neurons in the mature animal is largely established during embryonic development, through the function of specific transcription factors exclusive to different spatial regions (Flames et al., 2004;Butt et al., 2005Butt et al., , 2008Wonders and Anderson, 2006;Fogarty et al., 2007;Wonders et al., 2008;Gelman et al., 2009).
While the majority of neocortical GABAergic neurons project a highly ramified axon locally, these interneurons are not the sole source of inhibition within the neocortex. A small portion of GABAergic neurons projects axons to distant neocortical areas, including regions in the ipsilateral hemisphere and contralateral hemispheres (McDonald and Burkhalter, 1993;Gonchar et al., 1995;Aroniadou-Anderjaska and Keller, 1996;Chowdhury et al., 1996;Salin and Prince, 1996;Kimura and Baughman, 1997). A study combining labeling of GABAergic projection neurons, retrograde labeling and axon tracing methods revealed that the vast majority of neocortical long-range projecting GABAergic neurons belong to a group of cells expressing the neuronal isoform of nitric oxide synthase (nNOS) (Tomioka et al., 2005). This result suggests that even though low in number (Gonchar et al., 2007), neocortical nNOS GABAergic neurons have the potential to strongly influence neocortical networks. Most of the studies on the origin and function of nNOS cells have focused on the hippocampus where nNOS is expressed in a numerically large population of GABAergic neurons (Fuentealba et al., 2008;Tricoire et al., 2010Tricoire et al., , 2011. However, the function and origin of nNOS-expressing inhibitory neurons in the neocortex remain largely unexplored.
In the neocortex, near all parvalbumin (PV) and somatostatin (SST) expressing inhibitory neuron subtypes are produced in the MGE (Wichterle et al., 1999;Butt et al., 2005;Fogarty et al., 2007;Wonders et al., 2008;Xu et al., 2008). The genetic cascade necessary for proper specification, differentiation, and development of MGE-derived interneurons is starting to be elucidated. Nkx2.1 has been shown to direct the MGE-subtype specification and Lhx6 is necessary for the proper migration and maturation of both PV-and SST-expressing interneurons (Sussel et al., 1999;Liodis et al., 2007). The transcription factor Sox6, which is also expressed in MGE-derived interneuron lineages (Azim et al., 2009;, acts downstream of Lhx6 and is required for the positioning and maturation of PV cells, and to a lesser extent, SST cells . In addition to the MGE, the CGE is the second largest source of neocortical inhibitory neurons, contributing approximately 30% of all GABAergic neurons (Lee et al., 2010). Recently, it has been shown that all CGE-derived neocortical interneurons specifically express the serotonin receptor 3a (5-Ht3a) (Lee et al., 2010;Vucurovic et al., 2010), while the transcription factors CoupTF1/2 are widely but not selectively expressed within the CGE (Sousa et al., 2009). The 5-Ht3a-expressing CGE-derived GABAergic neocortical interneurons, includes the entire vasoactive intestinal peptide (VIP) and cholecystokinin (CCK)-expressing population as well as the entire SST-negative populations of calretinin (CR) and Reelin (RLN)-expressing interneurons (Nery et al., 2002;Lee et al., 2010;Miyoshi et al., 2010).
It has been recently shown that different parts of the POA selectively express Nkx5.1 and Dbx1, respectively, and produce a small and highly diverse fraction of neocortical GABAergic interneurons (Gelman et al., 2009(Gelman et al., , 2011. The interneurons types produced in the Dbx1 population are largely overlapping with the MGE-derived subtypes while it is possible that the Nkx5.1 expressing population is included in the population expressing 5-Ht3a (Gelman et al., 2009(Gelman et al., , 2011Lee et al., 2010).
Similarly to neocortical interneurons, PV-and SST-expressing hippocampal interneurons originate in the MGE, whereas hippocampal interneurons expressing CCK, CR, and VIP are produced in the CGE (Tricoire et al., 2011;Keimpema et al., 2012). Neuropeptide Y (NPY) expressing hippocampal interneurons encompass a mixed repertoire of subtypes originating from the MGE, CGE and the POA (Gelman et al., 2009). Despite largely originating in the MGE, the majority of nNOS-expressing hippocampal GABAergic neuron subpopulations do not overlap with PV-or SST-expressing interneurons (Fuentealba et al., 2008;Tricoire et al., 2010Tricoire et al., , 2011, and few nNOS-positive cells coexpress CR (Jinno and Kosaka, 2002a). In the hippocampus, nNOS is expressed in the majority of neurogliaform cells (NGC) and Ivy cells (IvCs) interneuron subtypes (Fuentealba et al., 2008;Tricoire et al., 2010). Within the hippocampus, nNOS colocalizes with a variety of markers (Fuentealba et al., 2008;Szabadics and Soltesz, 2009;Tricoire et al., 2010) and has been reported to be the numerically largest interneuron population (Fuentealba et al., 2008). It remains unknown if there are any homologs of the nNOS-expressing neurogliaform and Ivy interneurons within the neocortex.
Here we show that contrary to the hippocampus, nNOSexpressing cells constitute a small minority of the total neocortical GABAergic neurons. In order to investigate the developmental origin of nNOS-expressing cells, we did genetic fate mapping using cre-drivers specific for different domains of the MGE (Nkx2.1 Cre and Lhx6 Cre ) as well as the 5-Ht3a EGFP -reporter line that labels the CGE-derived lineage. Our findings suggest that nNOS-expressing neocortical GABAergic neurons originate from the MGE, mainly from the most dorsal domain of the MGE (dMGE). In the hippocampus, similarly to what has been previously shown, we observed that a majority of the nNOS interneurons are derived from the MGE, but a significant portion originate within the CGE and/or POA (Tricoire et al., 2010). Both neocortical and hippocampal nNOS-expressing cells also express NPY, but in contrast to the hippocampus virtually all neocortical nNOS neurons express SST, and about half express CR. Finally, we show that the transcription factor Sox6 has a role in the differentiation of neocortical nNOS cells, since loss of Sox6 leads to a total absence of SST expression in neocortical nNOS cells, and possibly stunts the development of neurites.
Our results show that nNOS-expressing neocortical and hippocampal inhibitory neurons have different origins, suggesting that neocortical and hippocampal nNOS-expressing cells constitute an example of unrelated subtypes, both having acquired/retained the expression of nNOS. Understanding the functional role of this sparse population of neocortical long-range projection inhibitory neurons and the relevance of the nitrinergic signaling in cortical circuits are questions of considerable interest.

In vivo GENETIC FATE MAPPING
To perform genetic fate mapping of neocortical and hippocampal GABAergic neurons transgenic males heterozygous for the driver line Dlx6 Cre (GENSAT project at Rockefeller University) were crossed to female homozygous for the RCE EGFP allele reporter allele (Sousa et al., 2009). To fate map MGE-derived lineages, males heterozygous for the transgenic driver lines Lhx6 Cre (Fogarty et al., 2007) and Nkx2.1 Cre  were crossed to female mice homozygous for the RCE EGFP reporter allele (Sousa et al., 2009). Fate mapping of the CGE-derived lineage was accomplished by using the BAC-transgenic line 5-Ht3a EGFP (GENSAT project at Rockefeller University).

TISSUE PREPARATION FOR IMMUNOCYTOCHEMISTRY
The brains of juvenile mice (P21) were fixed by transcardiac perfusion of 4% paraformaldehyde (PFA)/phosphate buffered saline (PBS) solution followed by a one hour post-fixation on ice with 4% PFA/PBS solution. Brains were rinsed with PBS and cryoprotected by using 30% sucrose/PBS solution overnight at 4 • C. Tissues were embedded in Tissue Tek, frozen on dry ice, and cryosectioned at 20 μm thickness.

NEOCORTICAL nNOS-EXPRESSING CELLS CONSTITUTE A SMALL MINORITY OF NEOCORTICAL GABAergic NEURONS
Previous work has shown that nNOS cells in the hippocampus and neocortex express markers characteristic of inhibitory neurons (Jinno et al., 2002;Gonchar et al., 2007;Fuentealba et al., 2008). In order to determine the percentage of nNOS neurons that are GABAergic, as well as the percentage of GABAergic neurons that express nNOS we performed immunostaining for nNOS in brain slices where the entire GABAergic cortical neuronal population was labeled with EGFP, using genetic fate mapping of cells expressing the pan-GABAergic gene Dlx6 (Batista-Brito et al., 2008). Genetic fate mapping experiments using Cre/loxP technology are based on the combined use of a driver allele (whereby the promoter of a gene of interest drives the expression of a Cre recombinase) and a floxed-reporter allele (whereby the expression of the reporter gene is triggered by a recombination event mediated by the Cre recombinase). Upon recombination, the expression of the reporter gene allows for selective and cumulative labeling of cells sharing the common expression of the gene of interest at any point in their history. Here we used the Dlx6 Cre driver mouse line combined with the RCE EGFP reporter line, whereby Cre activity removes the floxed-STOP cassette at the Rosa locus, resulting in permanent EGFP expression in cells that have once expressed Dlx6. We show that all the neocortical and hippocampal interneurons (100 ± 0.9% and 99.5 ± 1.1%, respectively) expressing nNOS also express EGFP. Therefore, we conclude that nNOS labeling in the neocortex and hippocampus is restricted to GABAergic neurons (Figures 1A-C). The number of neocortical GABAergic neurons expressing nNOS is much lower relative to the number of nNOS-expressing cells in the hippocampus (Figures 1A-C). Within the neocortex only 1.7 ± 0.3% of the total number of Dlx6 fate mapped cells express nNOS, compared to 15.9 ± 3.2% in the hippocampus ( Figure 1D). All our analyses were confined to cells that express high levels of nNOS, also known as nNOS type I neurons (Yan et al., 1996;Judas et al., 1999;Lee and Jeon, 2005). For simplicity, and because the used antibody does not allow the identification of cells expressing low levels of nNOS (type II nNOS neurons), we will refer to the cells that unambiguously express high levels of nNOS as nNOS-expressing cells throughout the manuscript. Neocortical nNOS neocortical neurons are preferentially distributed in deeper layers (V-VI), although some are located in the superficial layers (II/III) ( Figure 1A). Furthermore, a few nNOS-expressing neurons are also located in the white matter. Neocortical nNOS neurons display highly elaborated processes as described elsewhere (Tomioka et al., 2005). By contrast, nNOS staining is present throughout the mouse hippocampus at P21, including all layers of the Ammon's horns CA1-3 as well as in the dentate gyrus' layers (except the granular layer) ( Figure 1B).

THE MAJORITY OF NEOCORTICAL nNOS GABAergic NEURONS ORIGINATE WITHIN THE dMGE
In order to determine the origin of nNOS-expressing neocortical inhibitory neurons, we fate mapped neurons produced in the ventral telencephalon and located in the somatosensory cortex of P21 mice. Fate mapping of the MGE was accomplished by using the Nkx2.1 Cre  and Lhx6 Cre (Fogarty et al., 2007) driver lines in combination with the RCE EGFP reporter allele. In Nkx2.1 Cre ; RCE EGFP mice, EGFP labeled cells originate in most of the MGE (except for its most dorsal part; dMGE), and the POA Sousa et al., 2009). In Lhx6 Cre ; RCE EGFP mice cells originating within the entire MGE, and POA are permanently labeled with EGFP (Fogarty et al., 2007). Lhx6 is also expressed in blood vessels, therefore, fate mapping using this line also labels blood vessels with EGFP (Fogarty et al., 2007). We labeled CGE-derived cells with the transgenic line 5-Ht3a EGFP (Lee et al., 2010).

Frontiers in Neural Circuits
www.frontiersin.org July 2012 | Volume 6 | Article 44 | 5 Since 5-Ht3a EGFP labels cells migrating out from the CGE, our results suggest that the CGE do not produce nNOS inhibitory neurons of the neocortex. In contrast, 93 ± 3.3% of nNOS cells in the somatosensory cortex of Lhx6 Cre ; RCE EGFP mice express EGFP (Figures 2B,D), leading us to conclude that the vast majority of the neocortical nNOS cells originate within the MGE. Surprisingly, only 28.5 ± 4.3% of nNOS cells also express EGFP when we used the Nkx2.1 Cre ; RCE EGFP transgenic line. Comparison of the results from the Nkx2.1 Cre ; RCE EGFP and Lxh6 Cre ; RCE EGFP lines indicate that 70-75% of the total of nNOS cells in the somatosensory cortex originates within the dMGE, while the remaining 20-25% originates from the Nkx2.1 Cre domain. Moreover, this fate mapping strategy also confirmed that nNOS inhibitory neurons only account for a small portion of the MGE-and dMGEderived cortical interneurons (respectively, 0.7 ± 0.19% of fate mapped interneurons in Nkx2.1 Cre ; RCE EGFP mice and 2.6 ± 1.1% of fate mapped interneurons in Lhx6 Cre ; RCE EGFP mice) ( Figure 2E).

HIPPOCAMPAL nNOS GABAergic NEURONS DO NOT ORIGINATE WITHIN THE dMGE
We used the fate mapping approaches previously described to test if hippocampal nNOS neurons have the same origin as their neocortical counterparts. In accordance with previous studies (Tricoire et al., 2010), we observed that the majority of hippocampal nNOS neurons are derived from the MGE, as indicated by the observation that 76.5 ± 1.9% and 77.9 ± 5.7% of nNOS inhibitory neurons coexpress EGFP in the transgenic lines Nkx2.1 Cre ; RCE EGFP (Figures 3A-C,J) and Lxh6 Cre ; RCE EGFP ( Figures 3D-F,J), respectively. Comparison of the results from the Nkx2.1 Cre ; RCE EGFP and Lxh6 Cre ; RCE EGFP lines indicate that, contrary to what is observed in the neocortex, only a small minority (less than 2%) of hippocampal nNOS cells originate from the dMGE. Furthermore, we observed in the hippocampus that 43.1 ± 0.4% of nNOS neurons were co-labeled with EGFP in the transgenic 5-Ht3a EGFP line (Figures 3G-I,J). While it has been previously recognized that the CGE produces some hippocampal nNOS neurons by using the transgenic lines Gad65 GFP (Tricoire et al., 2010) and Mash1 CreER (Tricoire et al., 2010), the contribution of the CGE towards nNOS hippocampal neurons perhaps have been underestimated. This could be due to the fact that the Gad65 GFP lineage does not recapitulate the totality of the CGEderived lineage of interneurons, and due to the mosaic inducible nature of the Mash1 CreER line. However, the fact that 77.9 ± 5.7% (from the Lhx6 Cre ) and 43.1 ± 0.4% (from 5-Ht3a EGFP ) adds up to more than a 100% it is likely that 5-Ht3a EGFP is less a specific CGE-marker in the hippocampus compared to the cortex. Another possible explanation, which is not mutally exclusive of these observations, is that cell arising in the POA are both EGFPpositive in the Lhx6 Cre ; RCE EGFP and in the 5-Ht3a EGFP lines in the hippocampus. If the latter is the case, up to 21% of the nNOS population could be POA-derived, a much higher number than in the cortex.
Interestingly, we observed that despite the striatum being typically thought of as a structure less related to the neocortex than the hippocampus, the origins of striatal and neocortical nNOS cells are more similar than the origins of hippocampal and neocortical nNOS cells. The vast majority of striatal nNOS cells originate from the MGE, as indicated by Lhx6 Cre fate mapping (Compare Figure 4A to B, Figure 4C) (Gittis et al., 2010). Similarly to the neocortex, the dMGE is a major source of striatal nNOS neurons, producing about 40% of this population (substract 45.9% of nNOS cells fate-mapped with the line Nkx2.1 Cre ; RCE EGFP to 86.7% of nNOS cells fate mapped with the line Lhx6 Cre ; RCE EGFP ) ( Figure 4C).

NEOCORTICAL nNOS NEURONS BELONG TO THE MGE-DERIVED SOMATOSTATIN-POSITIVE POPULATION OF GABAergic INHIBITORY NEURONS
In order to eludicate the neurochemical profile of the dMGEderived nNOS-expressing inhibitory neurons, we performed immunostaining for nNOS and SST, PV, NPY, CR, and RLN, which are characteristic markers of inhibitory neuron subtypes. Double immunostaining performed with the antibodies nNOS and SST or PV in the juvenile somatosensory mouse cortex (P21) shows that virtually all nNOS cells co-express SST but not PV (data not shown). A further genetic fate mapping analysis carried out using the SST Cre ; RCE EGFP line confirmed that all neocortical nNOS cells belong to the SST-expressing population, and none of them express PV (Figures 5A,B). In addition, we did triple labeling, to discern between the SST/CR and SST/RLN coexpressing populations versus the STT-negative/CR and /RLN expressing cells that are CGE-derived . It is worth mentioning that NPY expression also seems to be shared by subpopulations of several different non-overlapping subtypes that have been shown to originate from the MGE, CGE, and POA (Fogarty et al., 2007;Xu et al., 2008;Gelman et al., 2009;Lee et al., 2010;Miyoshi et al., 2010). All the nNOS/SST double positive cells also co-express NPY (Figures 5B,C), 60 ± 15% express CR (Figures 5B,D), and none of the nNOS/SST double positive cells expressed RLN (Figures 5B,E). This is in contrast to the SST population in general, among which only around 25% co-express CR. This further supports a dorsal MGE origin for nNOS neocortical neurons since this population has been shown to also be enriched in SST/CR double positive interneurons (Sousa et al., 2009). In contrary, the majority of hippocampal nNOS cells do not express SST (Figure 5F), except for a small proportion of nNOS interneurons, mainly located in the dentate gyrus hilus ( Figure 5F1). Interestingly, virtually all the striatal nNOS neurons also express SST (Figure 5G), once again suggesting that striatal, but not nNOS hippocampal neurons, are related to nNOS neocortical neurons in terms of origin and phenotype.

THE TRANSCRIPTION FACTOR Sox6 IS REQUIRED FOR THE SPECIFICATION OF NEOCORTICAL nNOS GABAergic NEURONS
We have previously shown that the transcription factor  on the activity of the transcription factor Lhx6, which is specifically expressed in all MGE-derived interneurons (Du et al., 2008). We reported that conditionally removing Sox6 in MGE interneurons leads to a 30% reduction of the number of neocortical neurons expressing SST, and a complete loss of SST/CR-expression in mutant cells. In the present study, to address if in Sox6 mutants the nNOS subtype was among the affected SST population, we removed Sox6 from neocortical MGE-derived interneurons while simultaneously fate mapping the cells. To this end we used a conditional allele of Sox6 (Sox6 F/F ) (Dumitriu et al., 2006) crossed with the Lhx6 Cre driver line and the RCE EGFP reporter line. The conditional mutant and control generated are Sox6 F/F ; Lhx6 Cre ; RCE EGFP and Sox6 F/+ ; Lhx6 Cre ; RCE EGFP , respectively. In Sox6 F/F ; Lhx6 Cre ; RCE EGFP mutants we could still identify nNOS neurons, however, nNOS somatic expression appeared to be diminished relatively to control animals (Figures 6A,B). While the total number of neocortical nNOS neurons tends to be slightly decreased in control vs mutant (4.8 ± 1.8 nNOS+ cells/mm 2 vs 3.4 ± 2.1 nNOS+ cells/mm 2 , respectively), nNOS cells were rarely found in the superficial layers of Sox6 conditional mutants. We next asked if the neurochemical markers NPY, SST, and CR were affected in neocortical mutant nNOS neurons. The somatic nNOS staining, in combination with the EGFP expression of fate mapped cells allowed for the identification of Sox6 mutant MGE-derived inhibitory neurons. Sox6 mutant nNOS neurons no longer express SST ( Figure 6C) or CR (Figure 6D), but do retain the expression of NPY ( Figure 6E). While cortical nNOS expressing neurons are affected by the conditional loss of Sox6, their hippocampal counterparts remain largely unaffected in the conditional Sox6 mutants (data not shown).
In agreement to what we reported previously ), we observed a non-cell autonomous increase in the number of cells expressing NPY, as well as a diffuse and non-cell autonomous increase in the background of nNOS staining (data not shown). Furthermore, the cell morphology of mutant nNOS cells seemed to be severely affected, with these cells possessing considerably less dendritic and axonic arborization (compare Figures 6A1 and  6B1). However, it is not clear if this represents an alteration in the morphology of the cells, or if it is the result of weaker immunostaining. A more careful characterization, such as filling the mutant cells with a dye will be necessary to resolve this issue. However, since this cell population is so sparse, we did not succeed in identifying nNOS cells for detailed morphological analysis.
As previously described, Sox6 F/F ; Lhx6 Cre ; RCE EGFP mutants develop generalized seizures by P16 followed by death of the animals between P17 and P19 from prolonged seizures, most likely due to a severe loss of inhibition from the affected PVexpressing basket cells . To test if the phenotype observed in nNOS cells was a non-cell autonomous consequence of seizure activity, we generated the mutant mice, Sox6 F/F ; SST Cre ; RCE EGFP , in which Sox6 is only removed in SST-expressing MGE-derived inhibitory neurons, leaving the PV interneurons unaffected. Sox6 F/F ; SST Cre ; RCE EGFP mice appear to be phenotypically normal and do not show premature death nor develop obvious seizures. We confirmed that similarly to what we observed in Sox6 F/F ; Lhx6 Cre ; RCE EGFP mutants, nNOS neocortical neurons in Sox6 F/F ; SST Cre ; RCE EGFP animals still fail to express SST and CR (data not shown) and have reduced nNOS expression at the soma level (Figures 6F,G), as well as seemingly less elaborated arborization of their processes (Figures 6F1,G1). Our results suggest that the transcription factor Sox6 is necessary for the proper specification of nNOS neocortical neurons, and possibly necessary for the development of complex arborization of their processes (Figure 7).

DISCUSSION
In the present study, we characterized the contribution of nNOSexpressing cells in cortical GABAergic populations, as well as their place of origin and genes involved in their development. We show that contrary to the hippocampus, where nNOS interneurons are a numerous GABAergic class, nNOS (type I) neurons in the neocortex only constitute a very small portion of the GABAergic population. In accordance with previous studies, we observed that the majority of hippocampal nNOS GABAergic neurons originate within the MGE (Tricoire et al., 2010(Tricoire et al., , 2011, however, a considerable percentage of this neuronal type also arises from the more ventral structures CGE and/ or POA. In contrast, the vast majority of neocortical nNOS GABAergic neurons are produced in the MGE, mainly in the dMGE. Hippocampal nNOS GABAergic neurons are also more diverse than their neocortical counterparts with regards to their molecular heterogenity. Neocortical nNOS cells appear to be a rather homogeneous population consisting perhaps of only two sub-populations. All nNOS-expressing neocortical inhibitory neurons express SST and NPY while 60% of them express CR. However, hippocampal nNOS GABAergic neurons express a variety of molecular markers such as NPY, SST, VIP, CR (Tricoire et al., 2010), and can further be subdivided on neurogliaform cells (NGC) and Ivy cells (IvC) accordingly to their morphology and the laminar location on their soma (Fuentealba et al., 2008). Finally we show that while neocortical nNOS cells require the transcription factor Sox6 in order to develop their mature morphology and express characteristic molecular markers, the maturation of the hippocampal nNOS cells appears to be unaffected by Sox6 loss of function. This again strengthens the evidence supporting that nNOS GABAergic neurons in these two distinct, but highly related cortical structures, are generated through independent developmental programs. Interestingly neocortical nNOS neurons appear to be more closely related to their striatal counterparts.

NEOCORTICAL nNOS GABAergic NEURONS MAINLY ORIGINATE IN THE DORSAL MGE
Neocortical GABAergic neurons arise from three telencephalic transient embryonic structures, the MGE, CGE, and POA ). In Figure 8, we show a schematic of genes expressed in those regions, which are important for the production and specification of neocortical and hippocampal GABAergic neurons. The MGE expresses Nkx2.1 (Sussel et al., 1999) (red domain in Figure 8). In order to fate map the MGE we used the BAC line Nkx2.1 Cre , in which Cre expression is driven by the Nkx2.1 promoter . While Nkx2.1 protein is expressed throughout the MGE and POA, Cre expression in the Nkx2.1 Cre driver spares the most dorsal part of the MGE (dMGE) (purple domain in Figure 8) . The absence of Cre expression in the dMGE of this BAC transgenic line might be due to positional effects or the absence of a critical part of the Nkx2.1 promoter. However, because the expression of Nkx2.1 mRNA is weaker in the dMGE than the rest of the MGE (Sussel et al., 1999), it has been suggested that the absence of Cre expression in the dMGE reflects a weaker Nkx2.1 gene activation in this domain . The fact that the expression of another independently produced Nkx2.1 Cre BAC line is also absent from the dMGE (Kessaris et al., 2006;Fogarty et al., 2007) further supports this idea. Fate mapping of the totality of the MGE can be accomplished by using the BAC transgenic line Lhx6 Cre (Fogarty et al., 2007). Lhx6 is expressed in postmitotic neurons derived from the MGE, as well as the POA (blue domain in Figure 8) (Fogarty et al., 2007;Du et al., 2008). The most significant insight into the developmental origin of neocortical nNOS GABAergic was provided by the comparison between the fate mapping of Nkx2.1-and Lhx6-derived lineages. A minority of the neocortical nNOS cells (about 30%) could be accounted by Nkx2.1 fate mapping, but virtually all of the neocortical nNOS cells could be accounted by Lhx6 fate mapping (about 95%). Since the only progenitor region labeled by Lhx6 but not Nkx2.1-fate mapping is the dMGE, our data strongly suggests that the dMGE is the source of the majority of neocortical nNOS neurons.
In our experience, no nNOS-positive neocortical neurons were labeled by the 5-Ht3a EGFP line, which labels the CGE and possibly part of the POA (Lee et al., 2010). Given that previous studies have shown that a small fraction of the cortical nNOS-expressing cortical neurons originate in the Dbx1 domain of the POA (Gelman et al., 2009(Gelman et al., , 2011, we conclude that the 5-Ht3a EGFP BAC transgenic allele is unlikely expressed in the Dbx1-positive domain within the POA. This is further supported by the presence of markers typical for MGE (and thus not 5-Ht3a EGFP ) such as PV and SST in the Dbx1-fate mapped population.

DIVERSE ORIGINS OF nNOS NEOCORTICAL AND HIPPOCAMPAL GABAergic NEURONS AND THE USE OF NEUROCHEMICAL MARKERS
Fate mapping studies have shown that the neocortical interneuron subtypes arising from the MGE and CGE are, for the most part, mutually exclusive . For instance the hippocampal and neocortical PV-expressing fast spiking basket cells are solely produced in the MGE (Butt et al., 2005;Tricoire et al., 2011), while the CGE is the source of the VIP-expressing bipolar interneuron populations (Butt et al., 2005;Lee et al., 2010;Tricoire et al., 2011). This suggests that interneuron precursors are subjected to distinct genetic programs initiated within the eminences that at least in part, determine their subtype identity. However, some molecular markers, such as the neurotransmitter NPY, are not specific for a particular interneuron subtype. Both neocortical and hippocampal NPY-expressing inhibitory neurons comprise a wide variety of subtypes (Karagiannis et al., 2009;Tricoire et al., 2011) and are derived from the MGE, CGE, and POA (Fogarty et al., 2007;Gelman et al., 2009;Lee et al., 2010;Tricoire et al., 2011). Furthermore, the level of expression of NPY is highly dependent on neuronal activity (Sperk et al., 1992;Schwarzer et al., 1995;Baraban et al., 1997). Taken together, these observations suggest that NPY should be used with caution as a marker of subtype identity.
Here we show that similar to PV neurons, neocortical nNOS cells are solely derived from the MGE and constitute a rather homogeneous subset of long range projections SST-expressing neurons (Tomioka et al., 2005), therefore, suggesting that in the neocortex nNOS might be a good indicator of neuronal subtype. By contrast, hippocampal nNOS cells are produced from a number of subpallial regions, have different morphologies and express a variety of neurochemical markers (Tricoire et al., 2010(Tricoire et al., , 2011. This suggests that using nNOS as a subtype marker must be done carefully, especially when comparing structures. Moreover, the comparison between subtype diversity of hippocampal and neocortical cells expressing nNOS suggests that the location of birth of nNOS neurons is not the sole factor determining their marker expression, and that environmental cues present in the neocortex and hippocampus likely influence the expression of nNOS. It is likely that, except for nNOS expression, neocortical and hippocampal nNOS neurons are unrelated, and involved in different functions. Interestingly, neocortical nNOS neurons appear to have more in common with their counterparts in the striatum. Although being the largest striatal subtype, nNOS neurons are largely derived from the Nkx2.1 domain and similar to the cortex express both SST and NPY.

THE TRANSCRIPTION FACTOR Sox6 IS REQUIRED FOR THE SPECIFICATION OF NEOCORTICAL nNOS GABAergic NEURONS
Previous work has shown that loss of function of Nkx2.1 at the progenitor level results in a decreased number of nNOSexpressing neurons migrating tangentially into the cortex during development (Anderson et al., 2001), and a vast reduction of nNOS interneurons in the hippocampus at adult ages (Tricoire et al., 2010). We have previously shown that loss of the transcription factor Sox6, a gene downstream of Nkx2.1 and Lhx6 affects the radial migration and maturation of PV-and to a lesser extent SST-expressing neocortical inhibitory neurons. PV expression in fast-spiking basket cells is by and large lost in Sox6 mutants while the effect on SST expression is much more modest. One SST-positive population that is affected by the loss of Sox6 is the SST/CR double positive population where CR expression is lost completely . Given that in the neocortex nNOS neurons constitute a subpopulation of SST-neurons, we investigated whether the neurochemical characteristics, number, positioning or morphology of nNOS cells would be affected when Sox6 is conditionally removed in Lhx6 expressing cells. Removal of Sox6 had a marked effect on nNOS expressing neocortical inhibitory neurons. While nNOS persists (at low but detectable levels), Sox6 mutant nNOS-expressing cells no longer express SST or CR, and are possibly less numerous, as indicated by the fact that in mutant animals nNOS neocortical neurons tend to be absent from the superficial layers. Furthermore, it appears based on nNOS staining, that mutant nNOS neurons display underdeveloped arborization of their processes. This contrasts with our previous results showing that even though Sox6 mutant PV cells are immature, they still develop normal morphology . Taken together, our data show that removal of Sox6 in nNOS neocortical neurons compromises their specification and also perhaps their morphology. Since the phenotype of nNOS mutant cells remained the same when Sox6 was selectively removed in SST-expressing neurons, we argue that the effect is cell autonomous, or at least not caused by disrupted cortical network activity associated with PV-expressing fast spiking interneuron dysfunction. Similar to the neocortex, nNOS cells deprived of Sox6 gene function in the striatum have reduced SST staining (data not shown). However, we did not detect any obvious defects in hippocampal nNOS interneurons in Sox6 mutants (data not shown).

FUNCTIONAL ROLE OF nNOS GABAergic NEURONS
In the hippocampus (Jinno and Kosaka, 2002b), and likely in the neocortex, nNOS-expressing GABAergic neurons are the major source of nitric oxide (NO). NO regulates synaptic plasticity of glutamatergic and GABAergic synapses (Garthwaite and Boulton, 1995;Arancio et al., 1996;Burette et al., 2002;Szabadits et al., 2007;Garthwaite, 2008), and in the hippocampus potentiates GABA release (Li et al., 2002;Yu and Eldred, 2005). Even though neocortical nNOS neurons are distinct, and possibly differently regulated than their hippocampal counterparts, nNOS cells in the neocortex also function to control the level of GABA release and synaptic plasticity (Hardingham and Fox, 2006;Dachtler et al., 2011). While the number of nNOS neurons in the neocortex is small, their impact on the neocortical network through NO release might be significant as they are highly ramified, and project long distances rostro-caudally and medio-laterally, connecting neocortical areas up to 6-8 mm apart (Tomioka et al., 2005). In fact, it has been recently shown that specifically damaging nNOS neurons in the neocortex results in dysinhibition of the entire neocortical network and strong alterations of the spatio-temporal dynamics of the neocortex (Shlosberg et al., 2012). Cortical nNOS expressing neurons are also implicated in regulating local blood flow by transforming neuronal signals into vascular responses (Cauli et al., 2004), suggesting that neurogliaform-like neurons that express nNOS are likely to mediate cortical vasodilatations and/or vasoconstrictions that happen in vivo (Karagiannis et al., 2009). Moreover, it has also been shown that nNOS neurons are active during sleep (Gerashchenko et al., 2008). It has been reported that the functional activation of nNOS neurons during sleep is restricted to the cortex (Pasumarthi et al., 2010), suggesting that neocortical nNOS are major players in homeostatic sleep regulation (Gerashchenko et al., 2008;Pasumarthi et al., 2010). Importantly, while the regulation of the levels of NO is critical, an excessive production of NO can lead to neurotoxicity Frontiers in Neural Circuits www.frontiersin.org July 2012 | Volume 6 | Article 44 | 13 (Dawson and Dawson, 1998), suggesting that the nNOS population in addition to normal brain function can result in disease states when perturbed.
In the future, in order to understand the contribution of nNOS expressing inhibitory neurons in the neocortical circuit, it would be interesting to investigate the intrinsic properties of nNOS neocortical neurons, as well as their synaptic inputs and outputs. However, in order to address these questions in such a sparse neuronal type, it will be necessary to specifically target nNOS cells, a goal that is achievable using recently developed genetic tools, such as the inducible Cre-diver line, nNOS creER (Taniguchi et al., 2011). It would be particularly interesting to compare the electrophysiological properties of nNOS neocortical neurons to those of other SST expressing cells. Our hypothesis is that similar to nNOS cells in the striatum (Kawaguchi et al., 1995;Centonze et al., 2002) and SST neocortical interneurons (Kawaguchi and Kubota, 1996), nNOS cells in the neocortex will exhibit low threshold-spiking characteristics. However, due to the very different range of their output domains, we suspect that their functional roles in the network may be very different from those of SST + /nNOS − neocortical interneurons. To address the functional role of neocortical nNOS neurons at both synaptic and circuit levels, it would be very interesting to selectively activate and suppress their activity using genetic methods such as optoor pharmaco-genetics.