Two specific populations of GABAergic neurons originating from the medial and the caudal ganglionic eminences aid in proper navigation of callosal axons

The corpus callosum (CC) plays a crucial role in interhemispheric communication. It has been shown that CC formation relies on the guidepost cells located in the midline region that include glutamatergic and GABAergic neurons as well as glial cells. However, the origin of these guidepost GABAergic neurons and their precise function in callosal axon pathfinding remain to be investigated. Here, we show that two distinct GABAergic neuronal subpopulations converge toward the midline prior to the arrival of callosal axons. Using in vivo and ex vivo fate mapping we show that CC GABAergic neurons originate in the caudal and medial ganglionic eminences (CGE and MGE) but not in the lateral ganglionic eminence (LGE). Time lapse imaging on organotypic slices and in vivo analyses further revealed that CC GABAergic neurons contribute to the normal navigation of callosal axons. The use of Nkx2.1 knockout (KO) mice confirmed a role of these neurons in the maintenance of proper behavior of callosal axons while growing through the CC. Indeed, using in vitro transplantation assays, we demonstrated that both MGE‐ and CGE‐derived GABAergic neurons exert an attractive activity on callosal axons. Furthermore, by combining a sensitive RT‐PCR technique with in situ hybridization, we demonstrate that CC neurons express multiple short and long range guidance cues. This study strongly suggests that MGE‐ and CGE‐derived interneurons may guide CC axons by multiple guidance mechanisms and signaling pathways. © 2013 Wiley Periodicals, Inc. Develop Neurobiol 73: 647–672, 2013

ABSTRACT: The corpus callosum (CC) plays a crucial role in interhemispheric communication. It has been shown that CC formation relies on the guidepost cells located in the midline region that include glutamatergic and GABAergic neurons as well as glial cells. However, the origin of these guidepost GABAergic neurons and their precise function in callosal axon pathfinding remain to be investigated. Here, we show that two distinct GABAergic neuronal subpopulations converge toward the midline prior to the arrival of callosal axons. Using in vivo and ex vivo fate mapping we show that CC GABAergic neurons originate in the caudal and medial ganglionic eminences (CGE and MGE) but not in the lateral ganglionic eminence (LGE). Time lapse imaging on organotypic slices and in vivo analyses further revealed that CC GABAergic neurons contribute to the normal navigation of callosal axons. The use of Nkx2.1 knockout (KO) mice confirmed a role of these neurons in the maintenance of proper behavior of callosal axons while growing through the CC. Indeed, using in vitro transplantation assays, we demonstrated that both MGE-and CGE-derived GABAergic neurons exert an attractive activity on callosal axons. Furthermore, by combining a sensitive RT-PCR technique with in situ hybridization, we demonstrate that CC neurons express multiple short and long range guidance cues. This study strongly suggests that MGE-and CGE-derived interneurons may guide CC axons by multiple guidance mechanisms and signaling pathways. V C 2013 Wiley Periodicals, Inc. Develop

INTRODUCTION
The largest commissural tract in the human brain is the corpus callosum (CC) with over 200 million axons that transfer information between the two cerebral hemispheres. Agenesis of the corpus callosum (AgCC) including complete absence, as well as, hypogenesis (partial absence) of the CC is a complex neurodevelopmental disorder. More than 50 different types of CC agenesis have a genetic component and are associated with recognizable syndromes (Richards et al., 2004;Paul et al., 2007). Agenesis of the CC can result from defects during specific steps of callosal development such as cortical neuronal proliferation, specification or migration, as well as axonal guidance through the CC (Richards et al., 2004;Paul et al., 2007). Numerous patients with AgCC exhibit reduced capacity in abstract reasoning, problem solving, generalization and category fluency. Patients with AgCC are also deficient in the comprehension of idioms, proverbs, vocal prosody, and narrative humour (Doherty et al., 2006;Paul et al., 2007). This may be explained by the fact that the CC is the main path for coordinating syntactic information in the left hemisphere and prosodic information in the right hemisphere. Moreover, structural CC abnormalities have been noticed in some patients with schizophrenia, autism, Tourette's syndrome and attention deficit hyperactivity disorder, even though definitive association with these disorders still remains to be demonstrated (Hardan et al., 2000;Seidman et al., 2005;Plessen et al., 2006;Hallak et al., 2007;Paul et al., 2007).
During development, callosal axons navigate through a complex environment and make several critical decisions before reaching their target (Richards et al., 2004;Chedotal, 2011;Kalil et al., 2011). Growth cones, specialized sensory-motor structures at the leading edge of axons, are able to sense and to interpret different attractive and repellent guidance cues (secreted/diffusible or bound factors) in their environment (Dickson, 2002). The "guidepost" cells are placed at defined and critical positions within the growing pathway of axons and serve as temporary target points providing the axonal projections with the information required for appropriate guidance (Jones et al., 2002;Dickson and Zou, 2010). The current model proposes that the three midline structures, namely, the glial cells of the glial wedge (GW), the glial sling and the indusium griseum (IG) coordinate midline cortical axon guidance by secreting guidance factors which channel the CC axons into the correct path (Silver and Ogawa, 1983;Silver et al., 1993;Shu and Richards, 2001;Shu et al., 2003a,b,c;Smith et al., 2006). These guidance signaling factors include Netrin1/DCC, Slit2/Robo1, Semaphorin/Neuropilin-1 (Npn-1), ephrins/Eph and Wnt/Ryk (Serafini et al., 1996;Fazeli et al., 1997;Gu et al., 2003;Hu et al., 2003;Keeble et al., 2006;Mendes et al., 2006;Andrews et al., 2006Andrews et al., , 2007Lopez-Bendito et al., 2007;Ren et al., 2007;Bush and Soriano, 2009;Islam et al., 2009;Piper et al., 2009;Hutchins et al., 2011). In addition, observations in mice and humans showed that many neurons are present in the "glial" sling (Shu et al., 2003a;Ren et al., 2006;Jovanov-Milosevic et al., 2010), and scattered neurons were identified in the cat CC during early post-natal life (Riederer and Innocenti, 1992;Riederer et al., 2004). We have previously shown that two transient subpopulations of these neurons, one glutamatergic and one GABAergic, occupy a strategic position in the mouse CC and contribute to the guidance of growing callosal axons (Niquille et al., 2009). In particular, we found that both neuronal populations can attract callosal axons and that this activity is dependent in part on Sema3C function in glutamatergic neurons.
Recently, the meninges have also been shown to play a distinct role in the proper formation of the CC by regulating the onset of cortical neurogenesis in embryos (Siegenthaler et al., 2009). It is believed that the interaction of the meninges, via the secretion of BMP7, with the cingulate neurons, that express Wnt3, regulates CC formation (Choe et al., 2012). BMP7 acts as a negative regulator of callosal axon growth, and the callosal axons overcome this inhibition via the expression of Wnt3 (Choe et al., 2012). Interestingly, BMP7 is also required to control CC midline guidepost cells differentiation (Sanchez-Camacho et al., 2011). Also, Wnt3 expression in the callosal axons is regulated by GDF5 produced by the Cajal-Retzius cells, and GDF5 is further regulated by its inhibitor, Dan produced by the meninges (Choe et al., 2012). In addition, we have showed recently that the regulatory factor RFX3 plays a crucial role during early brain development by indirectly regulating GLI3 activity, which leads to FGF8 upregulation and to disturbed distribution of guidepost neurons that are required for CC morphogenesis (Benadiba et al., 2012). Therefore, these findings raise the possibility that the study of genes controlling CC guidepost cells differentiation or positioning could shed further light on callosal agenesis.
Here, we studied in details the guidepost GABAergic neurons of the CC and showed that they form a heterogeneous population that is generated in the ventral telencephalon and migrates to populate transiently the CC region at the time when callosal axons are approaching the midline. One of these neuronal populations is derived from the MGE, since they are generated from Lhx6-expressing precursors. The other population is generated in the CGE as shown by their labeling in 5-HT 3a -GFP-expressing transgenic mice. Time-lapse microscopy reveals that GABAergic neurons in CC migrate actively until E14.5 and reduce their mobility from E16.5 to E18.5 as callosal axons crossing the midline make frequent contacts on their surface. Grafting experiments demonstrate that both MGE-and CGE-derived GABAergic neurons exert a direct attractive action on callosal axons. Interestingly, GABAergic neurons of the CC were found to express several guidance factors known to regulate axonal outgrowth, branching formation or cell migration. Moreover, using Nkx2.1 2/2 mice that lack MGE-derived GABAergic interneurons, we confirmed that the guidepost neurons are required for proper navigation of callosal axons across the CC midline.
Together, these results reveal that two distinct classes of transient guidepost GABAergic neurons are important for callosal axon guidance and lay ground for new strategies to characterize the etiology of CC agenesis in humans.

In Utero Electroporation
To perform in utero electroporation, we adapted our protocol from the one described previously (Cancedda et al., 2007). E14.5 timed-pregnant GAD67-GFP mice were anesthetized with isoflurane (3.5% during induction and 2.5% during surgery) and embryos were exposed within the intact uterine wall after sectioning the abdominal wall. The DNA solution containing the expression vector encoding the red fluorescent protein (pCAG-IRES-Tomato) (2 mg/mL) together with the dye Fast Green (0.3 mg/mL; Sigma, St. Louis, MO) was pressure injected focally (1-2 mL) into the lateral ventricle of embryos through a glass micropipette using a PicoSpritzer III (Parker Hannifin, Cleveland, OH). Each embryo, while still being within the uterus, was placed between tweezer-type electrodes (System CUY-650P5 Nepa Gene Co, Chiba, Japan) and was electroporated with five electrical square unipolar pulses (amplitude: 45 V; duration: 50 ms; intervals: 950 ms) powered by a BTX electroporator apparatus (model BTX ECM 830; Harvard Apparatus, Holliston, MA). By orienting the tweezer-type electrodes, we were able to preferentially electroporate dorsal pallium precursors that give rise in part to pyramidal callosally-projecting neurons (Borrell et al., 2005). The embryos were quickly placed back into the abdominal cavity and the muscular and skin body wall layers were sutured. The embryos were allowed to develop until E16.5.

Slice Cultures
For E14.5 to E16.5 CC slice cultures, we used an in vitro model of CC organotypic slices that has been previously described (Niquille et al., 2009). Embryos were dissected in ice cold dissecting medium (MEM Gibco ref. 11012-044 with 15 mM glucose and 10 mM Tris pH 7-9). Brains were isolated and embedded in 3% low-melting point agarose (Invitrogen). Coronal sections (250 mm thickness) were made using a vibratome and slices at the level of the CC were collected in cold dissecting medium. CC slices were cultured on nucleopore Track-Etch membrane (1 mm pore size; Whatman) or millicell inserts (0.4 mm pore size; Millipore) in slice culture medium (SCM) (BME/HBSS (Invitrogen) supplemented with 1% glutamine, 5% horse serum, and Pen/Strep) (Lopez-Bendito et al., 2006).

Transplantation Experiments
To define the origin of CC GABAergic neurons, we performed transplantation experiments of LGE, MGE and CGE ventral regions known to generate cortical GABAergic neurons. For transplantation experiments, slices from E14.5 and E16.5 GAD67-GFP 1 embryos were selected and small explants comprising the ventricular (VZ) and subventricular (SVZ) proliferating zones of the LGE, MGE, and CGE were manually excised using tungsten needles and transplanted into the corresponding zones of GAD67-GFP 2 host slices.
To determine the possible function of LGE-, MGE-and CGE-derived CC GABAergic neurons in attracting callosal axons, we transplanted small explants of E16.5 GAD67-GFP 1 LGE, MGE or CGE into the septum of E16.5 GAD67-GFP 2 host slices. The study of early callosal axon navigation was performed after insertion of small crystals of 1,1'-dioctadecyl 3,3,3 0 ,3 0 -tetramethylindocarbocyanine perchlorate (DiI) into the frontal cortex to label cortical axons. After 36 to 72 h, the slice cultures were fixed and we analyzed neuronal migration or callosal axon trajectories by using confocal microscopy. In this assay, cases with axons growing along GAD67-GFP 1 neurons originating from the MGE or CGE transplants were counted as positive results for attraction.

In Vitro Focal Electroporation
We applied an in vitro focal electroporation method using a Nepagene apparatus to transfect an expression vector into telencephalic slices (Stuhmer et al., 2002). E14.5-E16.5 organotypic coronal slices from WT or GAD67-EGFP embryos were obtained as described above. The expression vectors encoding the green (pCAG-IRES-EGFP) or the red (pCAG-IRES-Tomato) fluorescent proteins were electroporated at a concentration of 1mg/ml in sterile PBS 1X (Sigma, St. Louis, MO). Expression vectors were focally injected through a glass micropipette into the LGE or the MGE using a PicoSpritzer III (Parker Hannifin, Cleveland, OH) and the slices were electroporated between platinum Petri-dish and cover square platinum electrodes (System CUY-700-1 and CUY-700-2; Nepa Gene Co, Chiba, Japan) with two unipolar square pulses (amplitude: 100 V; duration: 5 ms; intervals: 500 ms) generated by a CUY21EDIT Electroporator (Nepa Gene Co, Chiba, Japan).

Confocal Time-Lapse Microscopy
For imaging neurons, slices were cultured on nucleopore Track-Etch membranes (1 mm pore size; Whatman) in tissue dishes containing 3 mL of SCM medium (BME/HBSS supplemented with glutamine, 5% horse serum and Pen/Strep). CC GABAergic interneuron migration paths and dynamics were studied in coronal CC slices of GAD67-GFP transgenic mice from E14.5 to E18.5. To visualize callosal axons, the plasmid encoding the red fluorescent protein (pCAG-IRES-Tomato) was electroporated in utero into the dorsal pallium progenitors of E14.5 embryos that generate callosally projecting neurons. In some additional experiments, small dye crystals were implanted within the frontal cortex of GAD67-GFP sections made from E16.5 to E18.5. We imaged growth cone dynamics and outgrowth properties of frontal callosal axons that were found to grow continuously from the lateral extension of the CC to the midline after E16.5. After 15 to 20 h in culture, slices were placed in an open perfusion chamber on the stage of an upright Leica SP5 confocal microscope. Temperature was maintained at 37 C with the aid of a microscope incubator system (Life Scientific, Switzerland) and slices were perfused with SCM medium containing a gas mixture of 5% CO2/95% O2. GAD67-GFP positive CC neurons and callosal axons labeled with Tomato were imaged with a 203, a 403, or a 603 immersion lens at 15 or 20 min intervals using the fast scan function of the Leica SP5 confocal microscope resonant scanner. Image captures and all peripherals were controlled with Leica software. Pictures were processed and converted into *.AVI movies using Imaris and Metamorph 6.0. Migration parameters (basal rate of migration, frequency and duration of the intervening pauses) were determined by measuring movements every 15 to 20 min during 5 h while the neurons were migrating through the CC white matter. Using the tracking function of Metamorph software 6.0, a total of 70 GAD67-GFP neuron traces were analyzed at E14.5 and at E16.5; and 35 traces at E17.5.

Immunocytochemistry
Embryos were collected after caesarean section and quickly killed by decapitation. Their brains were dissected out and fixed by immersion overnight in a solution of 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4) at 4 C.
Postnatal mice were profoundly anesthetized and perfused with the same fixative, and their brains post-fixed for 4 h. Brains were cryoprotected in a solution of 30% sucrose in 0.1 M phosphate buffer (pH 7.4), frozen and cut in 50 mm thick coronal sections for immunostaining.
DAB immunostaining was performed as follows: Endogenous peroxidase reaction was quenched with 0.5% hydrogen peroxide in methanol, and unspecific binding was blocked by adding 2% normal horse serum during preincubation and incubations in the Tris-buffered solutions containing 0.3% Triton X-100. The primary antibodies were detected with biotinylated secondary antibodies (Jackson ImmunoResearch, West Grove, PA) and the Vector-Elite ABC kit (Vector Laboratories, Burlingame, CA). The slices were mounted on glass slides, dried, dehydrated, and covered with Eukitt.

Single RT-PCR
After microdissection, GABAergic neurons of the CC and of the bordering cortical area, labeled for GAD67-GFP, are isolated by FACS sorting and mRNAs for guidance cues are screened by sensitive RT-PCR (Rufer et al., 2005). The cDNA amplification and screen is performed in six steps: (1) the 300 GFP 1 and 300 GFP 2 cells are isolated by FACS sorting, (2) the lysis of the sorted cells and the reverse transcription of total mRNA into cDNA are performed simultaneously, (3) the cDNA is precipitated, (4) the homopolymeric 3'-oligo(dA) tailing is done on precipitated cDNA, (5) the cDNAs are globally amplified by PCR with the aid of oligo-dT Iscove primers, and (6) the genes for guidance cues are screened by PCR using specific primers. GAD67-GFP 1 neurons (labeled as R2) corresponding to 11.5% of the total CC cells population are purified by flow cytometry. The expression profile of the flow-cytometer sorted GFP 1 and GFP 2 cell fractions ascertained with specific primers against the various cell makers and guidance cues of interest. Using primers for GAD67, we can verify the selectivity of the FACS sorting method since GFP 1 neurons but not GFP 2 cells express GAD67. Using selective primers for guidance cues, GFP 1 and GFP 2 CC cells were screened for: (1) trophic factors (NGF, BDNF, FGF), (2) secreted guidance factors (Netrins, Slits, Semaphorins, ephrins) and their receptors (Ephs and Npns), as well as (3) cell adhesion molecules from the L1 family of CAMs (L1, NrCAM, NCAM, Ncadh) and from the TAG1/ axonin-1 family of GPI-linked CAMs (TAG1).

In Situ Hybridization
Sema 3A plasmid was linearized with Not1 (New England Biolabs) for antisense RNA synthesis by T3 polymerase (Promega). Sema3C plasmid was linearized with EcoRI (New England Biolabs) for antisense RNA synthesis by T7 polymerase (Promega) and with XhoI (New England Biolabs) for sense RNA synthesis by T3 polymerase (Promega). TAG-1 plasmid was linearized with XbaI (New England Biolabs) for antisense RNA synthesis by T3 polymerase (Promega) and with HindIII (New England Biolabs) for sense RNA synthesis by T7 polymerase (Promega). NrCAM plasmid was linearized with EcoRI (New England Biolabs) for antisense RNA synthesis by T7 polymerase (Promega). EphA4, Npn-1, EphB1 plasmids were linearized with SacI (New England Biolabs) for antisense RNA synthesis by T3 polymerase (Promega). ephrinB2 and eph-rinA3 plasmid were linearized with BamH1 (New England Biolabs) for antisense RNA synthesis by Sp6 polymerase (Promega). ephrinB3 plasmid was linearized with HindIII (New England Biolabs) for anti-sense RNA synthesis by T7 polymerase (Promega). EphrinA1 plasmid, EphA3 and EphA5 were linearized with Kpn1 (New England Biolabs) for antisense RNA synthesis by T7 polymerase (Promega). ephrinA4 plasmid and EphB3 were linearized with SacII (New England Biolabs) for antisense RNA synthesis by T3 polymerase (Promega). EphrinA5 plasmid was linearized with BamH1 (New England Biolabs) for antisense RNA synthesis by T7 polymerase (Promega). Slit2 plasmid was linearized with Xba1 (New England Biolabs) for antisense RNA synthesis by T7 polymerase (Promega). For in situ hybridization, brains were dissected and fixed by immersion overnight at 4 C in a solution containing 4% paraformaldehyde (PFA) in PBS. Hundred micrometers free-floating vibratome sections were hybridized with digoxigenin-labeled cRNA probe as described before (Garel et al., 1997). To combine in situ hybridization and immunofluorescence, Fast Red (Roche) was used as an alkaline phosphatase fluorescent substrate.
Imaging DAB stained sections were imaged with a ZeissV C Axio-plan2 microscope equipped with 103, 203, or 403 Plan-NEOFLUAR objectives and coupled to a CCD camera (Axiocam MRc 1388 3 1040 pixels). Fluorescent-immunostained sections were imaged using confocal microscopes (Zeiss LSM 510 Meta, Leica SP5 or Zeiss Qasar 710) equipped with 103, 203, 403 oil Plan-NEOFLUAR and 633oil, 1003 oil Plan-Apochromat objectives. Fluorophore excitation and scanning were done with an Argon laser 458, 488, 514 nm (blue excitation for GFP and Alexa488), with a HeNe laser 543 nm (green excitation for Alexa 594, CY3, and DiI), with a HeNe laser 633 nm (excitation for Alexa 647 and CY5) and a Diode laser 405 nm (for Hoechst-stained sections). Z-stacks of 10 to 15 planes were acquired for each CC coronal section in a multitrack mode avoiding crosstalk. Z-stacks of 40 to 50 sections were acquired for each CC section for the creation of isosurfaces with Imaris 6.0 software (Bitplane Inc.).
All three-dimensional Z stack reconstructions and image processings were performed with Imaris 6.0 software. To create real three-dimensional data sets we used the mode "Surpass." The generation of iso-surfaces (object defining a surface surrounding voxels located between two threshold values) allowed us to visualize the contours of cells and of growth cones. The colocalization between two fluorochromes was calculated and visualized by creating a yellow channel. Figures were processed in Adobe Photoshop CS2 and schematic illustrations were produced using Adobe Illustrator CS2.

CC GABAergic Neuronal Population Analysis
In 50 mm CC sections of Lhx6-Cre/R26R-YFP, Nkx2.1-Cre/ Dlx1-Venus fl , GAD67-GFP, and 5-HTR 3a -GFP embryos, GABAergic neurons of the CC were counted at E16.5 and E18.5 as the number of cells labeled respectively for the YFP, Venus, or GFP from at least five sections per condition. To study the density of GABAergic neurons through the CC, the values were quantified on four planes of the CC section stacks and were reported per surface unit (number of cells/mm 2 ). The cell densities were determined in two CC areas (midline and lateral) and in the entire CC area (total). Based on our work, we considered that the CC is divided into two regions: medial and lateral. The medial part is bordered dorsally by the IG and the longitudinal fissure and ventrally by the glial wedge and the dorsal limit of the septum. The lateral part comprises the white matter bordered by the cingulate cortex dorsally, and by the ventricular zone between the glial wedge and the mediodorsal angle of the lateral ventricle toward the ventricular side.

Ultrastructure
The E16.5 and E18.5 embryos were killed by decapitation. Brains were dissected and fixed by immersion for 24 h at 4 C in a solution containing 2% glutaraldehyde and 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4) with 2% sucrose. The brains were then rinsed in 0.1 M cacodylate buffer (pH 7.4), post-fixed at room temperature for 2 h in 1% OsO 4 , dehydrated in series of graded ethanol and embedded in Epon. The regions containing the CC of the embedded brains were trimmed and mounted on blocks to cut semi-thin and ultrathin sections. The ultrathin sections were mounted on formvar-coated single slot grids and contrasted with 2% uranyl acetate and 0.2% lead citrate.
For preembedding immunocytochemistry, embryonic brains were fixed by immersion in a solution containing 4% paraformaldehyde and 0.1% glutaraldehyde in 0.1 M phosphate buffer (pH 5 7.4) supplemented with 2% sucrose for 24 h. Fifty micron-thick sections cut with a Vibratome were immunoreacted with the same protocol as for light microscopy DAB staining, except that the detergent was omitted in all the solutions. Following the DAB reaction, sections were dehydrated and embedded in Epon. The plastic embedded specimens containing a DAB reacted section were prepared for ultra-thin sectioning following the same protocol as above.

Statistical Analysis
For all analysis, values from at least three independent experiments were first tested for normality. Values that followed a normal distribution were compared using Student's t-test. Values that did not follow a normal distribution were compared using Mann-Whitney and Kolmogorov-Smirnov non-parametric tests.

Generation and Migration of CC GABAergic Neurons Before Callosal Axon Ingrowth
Our previous work has implicated GABAergic neurons in the formation of the CC (Niquille et al., 2009). In order to further characterize the precise function of the CC GABAergic neurons in the formation of callosal pathways, we analyzed their distribution during development in relation to callosal axons. GABAergic guidepost neurons of the CC were identified using the GAD67-GFP mouse line in which the green fluorescent protein (GFP) is reliably expressed by GABAergic neurons (Tamamaki et al., 2003;Lopez-Bendito et al., 2004). At E14.5, migrating GAD67-GFP 1 neurons were properly positioned to guide callosal axons after they leave the cortex overlying the lateral region of the CC [ Fig. 1(A)]. At E15.5, the callosal axons reach the midline and by E16, the first axons that pioneer the formation of the CC begin to cross the midline and others that are originating from the frontal cortex just start to approach the CC lateral border (Ozaki and Wahlsten, 1998;Rash and Richards, 2001;Niquille et al., 2009). From E15.5 to E18.5, the callosal axons were surrounded by numerous GAD67-GFP 1 neurons that have invaded the entire CC white matter [ Fig. 1(B,C)]. From P7 to P14, CC GABAergic neurons disappeared progressively in a spatiotemporal gradient from the medial to the lateral CC regions [ Fig. 1(D,E)] (Niquille et al., 2009). These results show that during embryonic development transient GABAergic guidepost neurons, positioned at specific locations within the CC, delineate callosal axon paths.
The temporal profile of the embryonic CC GABAergic neuronal production was studied in the embryonic CC with birth-dating experiments. Several studies have mentioned that cortical GABAergic interneurons are mainly generated from the ventral pallium between E12.5 and E16.5 (Anderson et al., 2001;Marin and Rubenstein, 2001;Gelman et al., 2009;Miyoshi et al., 2010;Rubin et al., 2010;Vucurovic et al., 2010). Therefore, 5-bromo2 0 -deoxy-uridine (BrdU) injections were performed at E12.5, E14.5, and E16.5 in GAD67-GFP mouse and brains were processed for BrdU immunostaining when the CC is being formed i.e. at E16.5 and E18.5 [ Fig. 1 (F-J)]. We found that the GAD67-GFP 1 neurons that first reach the CC midline at E16.5 were mainly produced at E12.5 and a very few at E14.5 [ Fig.  1(F,G,K)]. Similarly also at E18.5, the CC was populated by GAD67-GFP 1 GABAergic neurons generated mainly between E12.5 and E14.5 [ Fig. 1(H-K)]. GAD67-GFP 1 neurons generated at E12.5 were found in both the medial and the lateral parts of the E18.5 CC [ Fig. 1(H-K)]. GAD67-GFP 1 neurons generated at E14.5 were distributed throughout the entire E18.5 CC, but mainly in the medial parts [ Fig.  1(I-K)]. GABAergic interneurons born at E16.5 were rarely observed throughout the CC medial and extreme lateral parts (CC wings, star in Fig.  1(J1,J2)]. This study indicates that CC GABAergic guidepost neurons are generated principally from E12.5 to E14.5 and migrate into the CC concomitantly to callosal axon ingrowth. Thus, the position and the timing of appearance of these GABAergic interneurons raise the possibility that they might be actively involved in the guidance of callosal axons.

CC Guidepost GABAergic Neurons Originate from the MGE and the CGE
We next characterized the source and molecular identity of these GABAergic neurons that converge to the CC commissure during embryonic development. The embryonic ventral telencephalon, or subpallium, contains distinct progenitor domains, the lateral, medial and caudal eminences (LGE, MGE, and CGE) and the preoptic area (Corbin et al., 2001;Marin and Rubenstein, 2001;Flames et al., 2004;Ghanem et al., 2007). In vivo fate mapping studies have revealed that the MGE and the CGE are the primary sites of genesis of a majority of cortical interneurons a majority of cortical interneurons Nery et al., 2002;Butt et al., 2005). The MGE was believed to be the major contributor in generating cortical GABAergic neurons but the CGE domain has recently been shown to account for 30-40% of them (Miyoshi et al., 2010;Rubin et al., 2010;Vitalis and Rossier, 2010;Vucurovic et al., 2010;Rudy et al., 2011). To determine the precise subpallial origin of CC GABAergic neurons, we have performed fate mapping studies using transplantation experiments, in vitro electroporation and genetic lineage tracing (Figs. 2 and 3).
First, transplantation experiments in organotypic slices were used to follow the fate of CC GABAergic neurons derived from different subpallial domains.
LGE, MGE, or CGE explants from E14.5 GAD67-GFP 1 mice containing GABAergic precursors, were transplanted into corresponding sites of unlabeled host telencephalic slices including the main body of the CC  (Sussel et al., 1999), which is specifically expressed in this region [ Fig. 2(E)]. At E14.5, electroporation in the MGE consistently gave rise to GAD67-GFP 1 neurons co-labeled in red by Tomato that had migrated into the CC white matter [ Fig. 2(D1,D2)]. By contrast, electroporation in the LGE or septum (not shown) did not generate GAD67-GFP 1 /Tomato 1 neurons appearing in the CC at any age [ Fig. 2(C)].

Figure 1
To further establish the relative contribution of the MGE and CGE domains toward the generation of CC GABAergic guidepost neurons, we performed genetic fate mapping analysis (Fig. 3). To selectively fate map MGE-derived postmitotic GABAergic neurons, we crossed Lhx6-Cre mice with the reporter line Rosa26-lox-STOP-lox-YFP (Srinivas et al., 2001;Fogarty et al., 2007). To visualize LGE-/CGEderived GABAergic neurons, we used a subtractive genetic labeling strategy described previously (Rubin et al., 2010), in which Dlx1-lox-Venus-lox mice referred as Dlx1-Venus fl transgenic mice were crossed with Nkx2.1-Cre mice. This approach allowed us to visualize GABAergic neurons generated by Dlx1positive subpallial domains except Nkx2.1-derived cells. The Venus 1 GABAergic neurons have been shown to originate from LGE/CGE domains (Rubin et al., 2010).
We compared the distribution of MGE-derived YFP 1 GABAergic neurons in the CC of Lhx6-Cre/ R26R-YFP embryos with the distribution of LGE-/ CGE-derived Venus 1 neurons in the embryonic CC at E16.5 and at E18.5. While LGE/CGE-derived GABAergic neurons have been shown to be generated and reach the cerebral cortex later than MGEderived neurons (Lavdas et al., 1999;Anderson et al., 2001;Miyoshi et al., 2010;Rubin et al., 2010;Vucurovic et al., 2010), both populations were present in the CC white matter as early as E16.5 [ Fig. 3(E,I)]. Consistent with previous studies showing that the two neuronal populations initially follow distinct migratory paths (Rubin et al., 2010), MGE-derived YFP 1 GABAergic neurons migrated via the MZ, SVZ/IZ and SP routes [ Fig. 3(E)] whereas LGE/ CGE-derived Venus 1 GABAergic neurons travelled exclusively within the SVZ/IZ [ Fig. 3(I)]. After quantification, MGE and LGE/CGE were found to generate an equal number of CC GABAergic neurons (YFP 1 neuronal density compared with Venus 1 neuronal density at E16.5: p > 0.05 for medial, lateral and entire CC) (Supporting Information Fig. S1A). Comparison of the medial and lateral parts of the CC showed that, at E16.5, MGE-and LGE-/CGE-derived GABAergic neurons localized preferentially within the lateral developing CC (neuronal density in the medial CC compared with neuronal density in the lateral CC at E16.5: p < 0.001 for both neuronal populations) [ Fig. 3(F,J), green lines]. Two days later at E18.5, both neuronal populations accumulated in abundance within the entire white matter and were distributed homogenously through out the medial and lateral CC (neuronal density in the medial CC  Fig. 3(K), red lines]. The density of the LGE-/ CGE-derived Venus 1 neuronal population significantly increased throughout the entire CC at E18.5 indicating that CGE-derived neurons most probably continued to migrate to the CC after E16.5 (Venus 1 neuronal density at E16.5, Fig. 3(J) compared with Venus 1 neuronal density at E18.5, Fig. 3(K); p < 0.05 in the entire CC; red lines]. The MGE and the LGE/CGE domains provide equal numbers of CC GABAergic neurons in the E18.5 CC (YFP 1 neuronal density compared with Venus 1 neuronal density at E18.5: p > 0.05 for medial, lateral and entire CC) (Supporting Information Fig. S1B). We furthermore confirmed that all CC GABAergic neurons originated either from the MGE or the LGE/CGE, by comparing the total number of GAD67-GFP 1 CC neurons (Supporting Information Fig. S1C and S1D) to the sum of YFP 1 neurons and Venus 1 neurons in the CC at E16.5 and E18.5. These results show that all the CC GABAergic guidepost neurons originate in the MGE or the LGE/CGE.
To further discriminate between an LGE or a CGE origin, we took advantage of the 5-HTR 3a -GFP mice that express GFP in cortical GABAergic neurons derived specifically from the CGE at embryonic ages (Inta et al., 2008;Lee et al., 2010;Vucurovic et al., 2010). Firstly, we studied the relative distributions of CGE-derived 5-HTR 3a -GFP 1 neurons in the CC of E16.5 and E18.5 [ Fig. 3(M,N,O,P)]. Secondly, we compared the densities of 5-HTR 3a -GFP 1 and of Venus 1 neurons throughout the CC at the two ages [Supporting Information Fig. S1(A,B)]. Apart from an additional band of 5-HTR 3a -GFP 1 neurons present throughout the cortical SP, the CC neuronal distribution was very similar for both neuronal populations indicating that the CGE, in contrast to the LGE, is a major contributor of CC GABAergic interneurons [compare  Fig. S1(A,B)]. Our results therefore show that the guidepost GABAergic neuronal population of the CC is more heterogeneous than previously thought. The CC white matter contains transient GABAergic interneurons that originate equally from two different subpallial domains: the MGE and the CGE.

Guidepost GABAergic Neurons Help Callosal Axons Outgrowth
To better characterize the dynamic properties and migratory profile of CC GABAergic neurons during embryonic development, we performed confocal timelapse video microscopy on GAD67-GFP 1 brain slices. From E14.5 to E15.5, we observed that numerous GAD67-GFP 1 neurons exhibited rapid multidirectional movements within the white matter of the CC (basal rate of migration at E14.5 5 25.0 6 1.0 mm/h, frequency of pauses at E14.5 5 3.5 6 0.3 pauses/10 h) [ Fig. 4
To gain insights into the interactions between CC GABAergic neurons and callosal projections, we analyzed callosal axons navigation in GAD67-GFP 1 brain slices. Callosal pioneering axons were visualized with the red fluorescent protein Tomato after in utero electroporation at E14.5 or by using DiI staining (Fig. 5). Using organotypic slices, we were able to examine callosal axons originating from the cingulate and frontal neocortex and growing continuously from the lateral extension of the CC to the midline [ In order to study the possible involvement of Nkx2.1-derived GABAergic neurons in guiding axons, we crossed Nkx2.1 1/2 mice (Sussel et al., 1999) with heterozygous GAD67-GFP knock-in mice. The sequential crosses allowed us to generate Nkx2.1 1/1 :GAD67-GFP and Nkx2.1 2/2 :GAD67-GFP mice which were used to visualize the GAD67-GFP 1 GABAergic interneurons in wild-type and mutant brains labeled by the GFP antibody. In Nkx2.1 2/2 :GAD67-GFP embryos, we observed as mentioned previously a loss of Nkx2.1-derived glia Kessaris et al., 2006; and personal communication) and a loss of around 60% of GAD67-GFP 1 neurons within the CC (36.8 6 3.36% of GAD67-GFP 1 neurons left in the CC; n 5 6 sections of Nkx2.1 1/1 :GAD67-GFP and Nkx2.1 2/2 :GAD67-GFP mice brains) (not shown). We performed in utero electroporation of a pCAG-IRES-Tomato plasmid into Nkx2.1 1/1 :GAD67-GFP and Nkx2.1 2/2 :GAD67-GFP embryos in order to visualize the callosal axons and GAD67-GFP 1 interneurons (Fig. 6). This experiment allowed us to detect that the callosal axons were displaying subtle differences between the mutant and the control. While Nkx2.1 2/2 :GAD67-GFP mice exhibited an enlarged CC, the general callosal axon navigation paths seemed to be preserved in the mutant compared with the WT brains. The callosal axons crossed the midline normally and progressed to the contralateral hemisphere. However, high magnification views revealed that the tomato-labeled callosal axons were not properly fasciculated at the midline and displayed aberrant branching patterns [ Fig. 6(B) compared with Fig. 6(A), open arrowheads]. The defasciculation of these axons can probably explain the thickening of the CC at the midline. Some of the callosal axons at the midline and post-crossing over of the midline were observed to stray away from the normal path and displayed mistargeting growth trajectories [ Fig.  6(B)]. These individual misguided axons turned within the tract and astray from the normal trajectory. Confocal time-lapse video microscopy unravelled that growth cones of callosal axons in the Nkx2.1 2/2 :GAD67-GFP CC still responded to the remaining GAD67-GFP 1 GABAergic neurons of the CC and progressed along their normal tracks [ Fig.  6(D) compared with Fig. 6(C)] innervating the contralateral hemisphere. However, they formed  The misdirected axons appeared to overlap more often and formed intermingled fascicles, when compared with WT. Some mutant axons displayed repeated retractions exemplified by the quantitative results. Although the WT and Nkx2.1 2/2 axons grew at similar speed, many mutant axons were restricted to smaller area when compared with WT [ Fig. 6(D), arrowheads]. Thus, the reduced GABAergic neuronal localization in the Nkx2.1 2/2 mice lead to callosal axon branching and outgrowth defects in the CC tract.

Guidepost GABAergic Neurons Exert Attractive Influences on Callosal Axons
To further understand how CC GABAergic neurons contribute to callosal axon outgrowth, we directly tested whether they could exert an attractive influence on growing axons using grafting experiments in organotypic slices. At E16.5, callosal axons grow through the midline, avoiding the flanking septum and IG which are repulsive for the callosal axons [ Fig. 7(A)]. In contrast, when explants from the CC of E16.5 GAD67-GFP positive slices (containing GABAergic neurons) were grafted into the non-permissive septal region of the wild-type host slice, numerous DiI-labeled callosal axons were deviated from their normal path and attracted toward the CC explant [ Fig. 7(B)]. Moreover, we examined whether the CC region comprising GAD67-GFP 1 GABAergic neurons exerts an attractive influence on cortical axons by placing E16.5 CC explants, comprising the two neuronal populations of interest, adjacent to explants of the medial cortex [ Fig. 7(F)]. Indeed, after 2 days in vitro, axonal outgrowth in the quadrant closest to the CC aggregate was increased compared to the quadrant furthest away from the aggregate, indicating chemoattraction [ Fig. 7(F2,F4)].
We have previously shown that, unlike LGEderived neurons, MGE-derived GABAergic neurons Figure 3 The MGE and the CGE generate CC GABAergic neurons. Immunohistochemistry for GFP on E16.5 (A1-A2, E1-E2, I1-I2, and M1-M2) and on E18.5 (D1-D2, H1-H2, L1-L2, and P1-P2) coronal sections from GAD67-GFP (A1-A2 and D1-D2), Lhx6-Cre/Rosa26R-YFP (E1-E2 and H1-H2), Nkx2.1-Cre/Dlx1-lox-Venus-lox (I1-I2 and L1-L2) and 5-HTR 3a -GFP (M1-M2 and P1-P2) mice. A2, D2, E2, H2, I2, L2, M2 and P2 are magnified views of the lateral CC regions visualized in A1, D1, E1, H1, I1, L1, M1 and P1, respectively. All sections were counterstained with Hoechst that labels cell nuclei in blue. Genetic fate mapping analyses indicate that, at E16.5 and E18.5, the CC comprises numerous GAD67-GFP 1 neurons, MGE-derived YFP 1 GABAergic neurons visualized in Lhx6-Cre/Rosa26R-YFP sections, as well as, LGE/CGE-derived Venus 1 GABAergic neurons visualized in Nkx2.1-Cre/Dlx1-Venus fl sections. In addition, the use of 5-HTR 3a -GFP mice that selectively labels CGE-derived GABAergic neurons confirms the presence of multiple CC GABAergic neurons originating from the CGE and strongly suggests that the LGE/ CGE Venus 1 neuronal population principally originates from the CGE. (B and C) Bars (means 6 SEM) represent the density of all CC GABAergic neurons labeled with the GAD67-GFP and present within the medial (mid), lateral (lat) and entire (tot) CC of E16.5 (B) and of E18.5 (C) GAD67-GFP mice. (F and G) Bars (means 6 SEM) represent the density of MGE-derived YFP 1 neurons present within the medial, lateral and entire CC of E16.5 (F) and of E18.5 (G) Lhx6-Cre/ Rosa26R-YFP mice. (J and K) Bars (means 6 SEM) represent the density of LGE/CGE-derived Venus 1 neurons that populate the medial, lateral and entire CC of E16.5 (J) and of E18.5 (K) Nkx2.1-Cre/Dlx1-Venus fl embryos. (N and O) Bars represent the density of CGE-derived 5-HTR 3a -GFP 1 neurons that occupy the medial, lateral and entire CC of E16.5 (N) and of E18.5 (O) 5-HTR 3a -GFP embryos. (B-C, F-G, J-K and N-O) After quantification, the density of all neuronal populations is significantly higher in the lateral CC compared with the medial CC at E16.5 (***green lines). At E18.5, both part of the CC contain the same density of neurons for each population. At E18.5 compared with E16.5, there is a significant increase of overall neuronal population density in the medial CC (***red lines). This occurs at the expense of the lateral CC where the neuronal density for all populations is decreased after the neurons migrate to the medial CC. are able to attract callosal axons. To determine whether CGE-derived GABAergic neurons of the CC have similar attractive properties, we performed heterotopic graft experiments in organotypic slices. To directly test the roles of these two populations of CC GABAergic neurons, we grafted in the septum of E16.5 host slices, explants coming from E16.5 GAD67-GFP 1 MGE or CGE which both generate the GABAergic neurons of the CC [ Fig. 7(D,E), respectively]. Interestingly, numerous axons left the callosal track, penetrated the repulsive septum and grew in between migrating GAD67-GFP 1 neurons originating from the MGE transplant [ Fig. 7(D), arrowheads, n 5 14/19] or from the CGE transplant [ Fig. 7(E), arrowheads, n 5 11/13]. MGE-derived and CGE-derived neurons attracted callosal axons equally. This attraction was specific for MGE-derived and CGE-derived neurons, since LGE control explants did not attract callosal axons [ Fig. 7(C); n 5 10/12]. These observations indicate that CGE- Figure 5 Dynamic interactions between CC GABAergic neurons and callosal axons between E16.5 to E18.5. (A) Experimental paradigm used to study cell interactions between callosal pioneering axons labeled with the red fluorescent protein Tomato and GAD67-GFP 1 CC neurons. For this purpose, a plasmid encoding the Tomato (pCAG-Ires-Tomato) was electroporated into the dorsal pallium of E14.5 GAD67-GFP 1 living embryos to label neuronal precursors of pyramidal neurons that migrate to the cortex and differentiate for a part into callosal projecting neurons. DNA solution containing the pCAG-Ires-Tomato plasmid was pressure-injected focally into the lateral ventricle of E14.5 embryos. Each embryo, while still within the uterine horn, was placed between tweezer-type electrodes. After in utero electroporation, the embryos were quickly placed back into the abdominal cavity and allowed to develop until E16.5. (B1 and B2) In vitro time-lapse sequences over a period of 100 min (at 20 min intervals) of Tomato-labeled callosal axons and GAD67-GFP 1 neurons on coronal CC slices of E16.5 GAD67-GFP 1 embryos. (B1) Low power view of the CC showing the region of interest where the video imaging was done. (B2) High power views of Tomato 1 callosal axon growing through the CC and making branch extensions and retractions at different time points (t:0 min-t:100 min). Callosal axon branches are seen to make multiple contacts with CC GAD67-GFP 1 guidepost neurons (t:20 min, t:60 min and t:100 min, arrowheads). Contacts either promote axonal growth (t:20 min-t:60 min) or lead to growth cone retraction (t.80 min-t:100 min) and formation of a new branch (t:80 min-t:100 min, open arrowhead). (B3) Illustration of the iso-surfaces obtained from the GAD67-GFP and live Tomato staining in B2 (t:20 min), respectively. (C) Graphical representation displaying the frequency of axonal extension, branching, retraction and pauses per hour for the Tomato-labeled callosal axons that contact CC GAD67-GFP 1 neurons (contact) and for those that do not contact CC GAD67-GFP 1 neurons (non-contact) for time-lapse sequence over a period of 180 min made at 20 min intervals *p < 0.05; comparison between values obtained for axons making contact or not and done in similar portions of E16.5 CC (samples of 10 axons for each case). Axons making contacts make relatively more branching and pauses than axons that do not contact neurons. Interestingly, callosal axons that contact CC GABAergic neurons extend significantly more than those that do make contact. (D1 and D2) Confocal time-lapse imaging of DiI-labeled callosal axons and GAD67-GFP 1 neurons on coronal CC slices of E16.5 GAD67-GFP mice. (D1) Low power view of the CC showing the region of interest where the video recording was done. (D2) High power views of DiI-labeled callosal axons growing along CC GAD67-GFP 1 neurons taken at various time points (t:0 min, t:30 min, t:60 min). (E1 and E2) Illustration of the iso-surfaces obtained from the GAD67-GFP and live DiI staining in D1 and D2, respectively. White color highlights contact between axons and neurons, and stars indicate neurons that remain immobile. IG, induseum griseum. Bar 5 240 mm in B1; 110 mm in D1, E1; 15 mm in B2 and 30 mm in B3, D2, E2.

Figure 6
derived GABAergic neurons directly contribute, in addition to MGE-derived neurons, to an attractive activity on callosal axons.

CC GABAergic Neurons Express Several Guidance Factors and Cell Adhesion Molecules
To characterize the molecular mechanisms underlying the guidance effects of CC GABAergic guidepost neurons on callosal axons, we investigated whether GABAergic neurons express short-or long-range guidance factors (Fig. 8). First, we studied the gene expression profile at E16.5 of GAD67-GFP 1 neurons in CC and the adjacent cortex using a highly sensitive RT-PCR method (Rufer et al., 2005;Speiser et al., 2006) [ Fig. 8(A1)]. After micro-dissection and cell dissociation, 300 GFP 1 and 300 GFP 2 cells of the CC and adjacent cortex were directly isolated by FACS sorting and screened for various guidance factors using RT-PCR together with subsequent amplification with global and specific primers [ Fig.  8(A1,A2)]. We examined the expression profile of the CC neurons for guidance cues that are known to regulate outgrowth, branching or pathfinding in the CC and in other axonal tracts during development [ Fig. 8(A3)]. These include: (1) trophic factors (NGF, BDNF, FGF), (2) secreted guidance factors (Netrins, Slits, Semaphorins, ephrins) and their receptors (Ephs and Npns), as well as (3) cell adhesion molecules from the L1 family of CAMs (L1, NrCAM, NCAM, Ncadh) and from the TAG1/axonin-1 family of GPI-linked CAMs (TAG1) (Stoeckli and Landmesser, 1998;Suter and Forscher, 2000;Wilkinson, 2001;Yu and Bargmann, 2001;Dickson, 2002;Yazdani and Terman, 2006;Egea and Klein, 2007;Mann et al., 2007;Chedotal, 2011). We found that GAD67-GFP 1 neurons did not express Netrins, Slits, or the trophic factors [ Fig. 8(A3), 1 bands). In contrast, they did express Sema3A and Neuropilin receptors (Npn1 and Npn2), several ephrins (ephrinB1, eph-rinB2, ephrinA1, ephrinA4) and Eph receptors (EphB2, EphB3, EphA4), as well as cell adhesion molecules from the L1 family (NrCAM, L1, NCAM1, Ncadh) [ Fig. 8(A3), 1 bands). To confirm that CC GABAergic neurons specifically expressed these factors, we performed in situ hybridization or immunostaining. In accordance with the RT-PCR results, we never detected expression of Netrins, Slits, ephrinA3, ephrinB3, EphB1, nor TAG1 in the CC GAD67-GFP 1 neurons [not shown and Fig. 8(A3)]. NrCAM, eph-rinA1, ephrinA4, and ephrinA5 were only observed at low levels within the cerebral cortex at E16.5 (not shown). Npn-1 while strongly expressed by CC white matter cells was only detected sporadically in a few Figure 6 Branching and outgrowth defects in the callosal axons of Nkx2.1 2/2 :GAD67-GFP mice brains. (A and B) A pCAG-Ires-Tomato plasmid was injected into the lateral ventricle and electroporated into the dorsal pallium, to label the callosal projecting neurons, of E14.5 GAD67-GFP 1 living embryos that were allowed to develop until E18.5. Double immunohistochemistry for the GFP and the Tomato on CC coronal sections from Nkx2.1 1/1 :GAD67-GFP 1 (A) and Nkx2.1 2/2 :GAD67-GFP 1 (B) mice brains at E18.5. Cell nuclei were counterstained in blue with Hoechst. The general callosal axon navigation seems to be preserved in the mutants compared with WT brains (compare B with A). Postcross (A2, B2) and midline panels (A3, B3) are showing higher magnifications of the callosal axons (in red) and the GAD67-GFP 1 interneurons (in green). In the Nkx2.1 2/2 :GAD67-GFP 1 CC, there was a loss of GAD67-GFP 1 neurons and the axons were not properly fasciculated and made aberrant branches that extended in wrong directions (open arrowheads). (C and D) High power views of in vitro time-lapse sequences over a period of 120 min (at 20 min intervals) of Tomato-labeled callosal axons and GAD67-GFP 1 neurons on coronal CC slices of E16.5 Nkx2.1 1/1 :GAD67-GFP 1 (C) and Nkx2.1 2/2 :GAD67-GFP 1 (D) embryos. In the Nkx2.1 2/2 brains, though the callosal axons progressed along normal path, they displayed disoriented branch extensions. The different trajectories of WT and Nkx2.1 2/2 axons within the CC are illustrated by cell tracking paths (E and F, highlighted in white) and three sequential pictures (C and D). (G and H) Graphical representation displaying the axonal outgrowth and the distance from the origin per hour for the WT and Nkx2.1 2/2 Tomato-labeled callosal axons for time-lapse sequence over a period of 5 h made at 20 min intervals ***p < 0.001; comparison between values obtained for WT and Nkx2.1 2/2 axons (samples of 7 axons for each case). Quantification reveals that while axons in WT and Nkx2.1 2/2 axons grow at the same speed, Nkx2.1 2/2 axons slowly move away from the origin due to many back and forward movements. CC, corpus callosum; CCi, cingulate cortex; CFr, frontal cortex; IG, indusium griseum; LV, lateral ventricle; SEP, septum. Bar 5 500 mm in A and B; 40 mm in postcross (A2, B2) and midline panels (A3, B3) and 25 mm in C and D.

Figure 7
GABAergic neurons (not shown). Interestingly, we found that Sema3A, Npn-2, and numerous ephrins and Eph receptors (ephrinB1, ephrinB2, EphA4, and EphB2) were expressed at high levels by GAD67-GFP 1 neurons of the CC at E16.5 [ Fig. 8(B-G), arrowheads]. The guidance factors were not uniformly expressed by all CC GABAergic neurons (open arrowheads) indicating that these neurons have distinct guidance properties. The factors were also detected in the GFP 2 pool of CC cells comprising other neurons, glia or precursors [ Fig. 8(A3), -bands; Fig. 8(B2-G2), *]. In accordance with our previous study (Niquille et al., 2009) and our RT-PCR results the Sema3C was only found within the glutamatergic neurons and not the GABAergic neurons of the CC [ Fig. 8(A3) and not shown].
Taken together, our time-lapse and transplantation experiments indicate that CC neurons exert shortand long-range attractive influences on callosal axons, which are mediated by both MGE-and CGEderived GABAergic neurons. In search for candidate molecular signals, we found that CC GABAergic neurons express different combinations of secreted guidance factors and cell adhesion molecules that possibly participate to their guidepost function for callosal axons.

DISCUSSION
Our study reveals that GABAergic neurons are important for regulating axonal guidance in the embryonic CC. We identified two distinct GABAergic neuronal subpopulations, one originating from the MGE and the other from the CGE, which transiently occupy the developing CC. These two neuronal populations exhibit a reduced motility after they have reached the midline and occupy a strategic position for guiding ingrowing callosal axons. Callosal axons while navigating through the CC are seen to make close contacts with the stationary GABAergic neurons. Interestingly, both MGE and CGE GABAergic neuronal populations exert an attractive activity on callosal axons. In search for candidate molecular signals mediating the attractive activity on callosal axons, we found that CC GABAergic neurons express several types of guidance factors and cell adhesion molecules. By revealing a role for GABAergic neurons in the pathfinding of callosal axons, our To estimate these effects, we placed small explants of E16.5 MGE (D1-D3) or E16.5 CGE (E1-E3) that both generate the CC GABAergic neurons into the nonpermissive septal region of E16.5 host slices. (D2-D3 and E2-E3) DiI staining showing that numerous callosal axons penetrate through the septum and are attracted by GAD67-GFP 1 migrating GABAergic interneurons derived from the MGE (D3, arrowheads) or the CGE (E3, arrowheads) transplants. Both neuronal populations are found to exert equivalent attractive activities. All sections were counterstained with Hoechst that labels cell nuclei in blue. (F1-F4) Experimental paradigms used to determine whether the CC (F1) exerts attractive effects on cortical axons. (F2) Immunohistochemical staining for betaIII-tubulin on E16.5 frontal cortex (CFr) co-cultured with E16.5 GAD67-GFP 1 CC. (F3) Immunohistochemical staining for betaIII-tubulin on E16.5 GAD67-GFP 1 explant. Interestingly, callosal axons were seen to leave the cortical explant that comprises GAD67-GFP 1 GABAergic neurons. (F4) Quantification of the axonal guidance responses to E16.5 CC explants comprising GAD67-GFP 1 GABAergic neurons. Data are expressed as a P/D ratio, where P and D are the mean fluorescence 6 SEM intensity in the quadrants proximal and distal to the CC or septum explants. ***Significantly different with p < 0.001. CC GABAergic neurons cause strong chemoattraction on frontal axons. CFr, frontal cortex; IG, induseum griseum; LV, lateral ventricle; SEP, septum. Bar 5 485 mm in A2, B2, C2, D2, E2, F2, F3; 240 mm in A3, B3, C3 and 120 mm in D3, E3. study brings new understanding about neuronal development and new perspectives on pathophysiological mechanisms altering CC formation in humans.

GABAergic Neurons act as Guidepost Cells in the CC
During development, callosal axons make multiple guidance choices to cross the midline and reach their contralateral cortical targets (Richards et al., 2004;Chedotal, 2011). Our previous study revealed that, in addition to the characterized role of glial cells, transient glutamatergic and GABAergic neurons are essential for the control of axon guidance in the developing CC (Niquille et al., 2009). Here, we have specified the origins and functions of the GABAergic neurons of the CC. We have discovered that CC GABAergic neurons arise from two distinct sources, follow different migratory routes and converge to the CC midline before the arrival of callosal axons. Our

Figure 8
in vitro and genetic fate-mapping experiments indicate that the GABAergic neurons of the CC originate equally from the MGE and the CGE. Similarly, the MGE and the CGE have been shown to provide cortical GABAergic interneurons Anderson et al., 1997Anderson et al., , 2002Nery et al., 2002;Rubenstein, 2001, 2003;Xu et al., 2004;Butt et al., 2005;Metin et al., 2006;Fogarty et al., 2007;Lee et al., 2010;Miyoshi et al., 2010;Rubin et al., 2010;Vitalis and Rossier, 2010;Vucurovic et al., 2010;Rudy et al., 2011). In addition, grafting experiments indicate that while originating from different sources, both MGE and CGE GABAergic migrating neurons possess similar capacity to chemoattract callosal axons. Interestingly, the different migration paths and distributions of MGE-and CGE-derived GABAergic neurons through the CC at E16.5 may allow these guidepost neurons to play complementary attractive effects on growing callosal axons.
Time lapse analysis allowed us to specify the nature of the cellular interactions between callosal axons and CC GABAergic neurons. In accordance with a previous study made in living CC slices from neonatal hamster (Halloran and Kalil, 1994), we observed that callosal axons exhibit a complex behavior characterized by the extension of new branches that are either stabilized or retracted. Interestingly, our work indicates that this complex growth cone behavior could be related to growth cone interactions with CC guidepost GABAergic neurons. Cell contacts with CC GABAergic neurons were observed more often to be accompanied by new branch formation and increased axonal outgrowth and in fewer cases to correlate with growth cone collapse and branch retraction. The presence of these GABAergic neurons is required for the proper navigation of callosal axons, since the reduction of MGE-derived GABAergic interneurons in the CC of Nkx2.1 2/2 caused branching and outgrowth abnormalities of callosal axons. However, the observed phenotype in Nkx2.1 2/2 is less drastic than the patterning defects seen in Mash1 mutants described by our group previously (Niquille et al., 2009). The differences can be attributed to the fact that Mash1 mutant embryos are nearly devoid of the GABAergic neurons that originate from the ventral telencephalon (Niquille et al., 2009) whereas the Nkx2.1 2/2 embryos loose about 60% of the MGE-derived GABAergic neurons only. These results, thus, unravel that these neurons are important for The cDNA amplification and screen is performed in six steps: (1) the 300 GFP 1 and 300 GFP 2 are isolated by FACS sorting, (2) the lysis of the sorted cells and the reverse transcription of total mRNA into cDNA are performed simultaneously, (3) the cDNA is precipitated, (4) the 3'-oligo(dA) tailing is done on precipitated cDNA, (5) the cDNAs are globally amplified by PCR, and (6) the genes for guidance cues are screened by PCR using specific primers. (A2) GAD67-GFP 1 neurons (labeled as R2) corresponding to 11.5% of the total CC cells population are purified by flow cytometry. (A3) Expression profile of flow-cytometer sorted GFP 1 and GFP 2 cell fractions ascertained with specific primers against the various cell makers and guidance cues of interest. Using primers for GAD67, we can verify the selectivity of the FACS sorting method since GFP 1 neurons but not GFP 2 cells express GAD67. In accordance with the literature, the GFP 2 cells of the CC correspond to VGLUT 1 and Tbr 1 glutamatergic neurons, Emx2 1 precursors and GFAP 1 astrocytes or radial glia. Using selective primers for guidance cues, GAD67-GFP 1 neurons are found to express several long range guidance cues or receptors (ephrinB1/B2/A1/A4 and EphA4/ B2/B3, sema3A, and Npn1/2) and adhesion molecules (NrCAM, NCAM1, NCadh). (B1-G2) In situ hybridization for ephrinB1 (B1-B2), ephrinB2 (C1-C2), EphB2 (D1-D2), EphA4 (E1-E2), and Sema3A (F1-F2) mRNAs combined with immunohistochemical staining for GFP in coronal telencephalon sections of E16.5 GAD67-GFP mice. (G1-G2) Double immunohistochemical staining for Npn-2 and GFP in coronal telencephalon sections of E15.5-E16.5 GAD67-GFP mice. B2, C2, D2, E2, F2 and G2 are high power views of the CC seen in B1, C1, D1, E1, F1 and G1 (open arrows). Consistent with the PCR results, in all the panels GAD67-GFP 1 GABAergic neurons of the CC (arrowheads) and other cells (*) of the CC that are negative for the GFP express strong levels of respective mRNAs. GABAergic neurons are heterogeneous with respect to their expression of various guidance cues since certain GAD67-GFP 1 neurons (arrowheads) but not all (open arrowheads) express ephrinB1/B2, EphB2/A4, Sema3A mRNAs and Npn-2 (shown here). Bar 5 435 mm in B1, C1, D1, E1, F1, G1; 40 mm in B2, C2, D2, F2 and 20 mm in E2, G2. the correct behavior of callosal axons. Altogether, timelapse and transplantation experiments indicate that CC GABAergic neurons exert attraction on callosal axons. This function of CC CGE-and MGE-derived GABAergic neurons as guidepost cells fits in nicely with the finding that thalamo-cortical axon growth depends instead on the ventral migration of LGEderived GABAergic neurons within the ventral telencephalon (Lopez-Bendito et al., 2006). These studies altogether point out the high degree of specificity of each type of guidepost GABAergic neuronal population that are involved in the development of different axonal paths (cortico-cortical vs. thalamo-cortical). Therefore, transient CC guidepost GABAergic neurons constitute a remarkable mechanism that allows the timely arrangement of guidance cues required for callosal axons pathway formation.

Cellular and Molecular Mechanisms of Callosal Axon Guidance
To further understand the molecular mechanisms underlying the guidance effects of CC GABAergic guidepost neurons on callosal axons, we investigated in detail their expression profile for guidance factors.
Our results indicate that CC GABAergic neurons expressed the Sema3A and Npn-2 receptors, several ephrins (B1/B2) and Eph (B2/B3/A4) receptors, as well as cell adhesion molecules (Ncadh, Ncam1) that help to direct axonal outgrowth. In our study, the guidance factors were not uniformly expressed by all CC GABAergic neurons indicating that these neurons form heterogeneous populations concerning their guidance properties. It will be interesting in the future to define whether this heterogeneity is linked to differences of their origin sites (MGE vs. CGE) and whether they have different but complementary roles to play during guidance. This screen nicely complements a previous study showing that the expression of cell surface genes such as Ncam1, Sema3a but not Npn-1 is enriched in E15.5 cortical migrating interneurons (Batista-Brito et al., 2008).
These results are in perfect accordance with previous reports about the expression of several of these guidance factors (ephrins/Eph and semaphorins/Neuropilins) in the CC and surrounding regions (glial wedge, indusium griseum, septum) during development (Gu et al., 2003;Mendes et al., 2006;Bush and Soriano, 2009;Hatanaka et al., 2009;Niquille et al., 2009;Piper et al., 2009;Zhao et al., 2011). Their related receptors (ephrins/Eph and neuropilin-1) are found to be expressed by callosal axons through E15 to E17, the period at which pioneer callosal axons cross and extend to the contralateral hemisphere.
Finally, these numerous guidance factors and their cognate receptors have been implicated in orienting the CC axons at critical choice points.
Previous studies on CC formation highlighted the role of glial cells in guiding callosal axons and proposed that most of these guidance cues are expressed by guidepost glial cells of the CC (Shu and Richards, 2001;Gu et al., 2003;Shu et al., 2003c;Keeble et al., 2006;Mendes et al., 2006;Bush and Soriano, 2009;Islam et al., 2009;Piper et al., 2009). Indeed, studies have nicely demonstrated that the guidepost glia of the glial wedge and indusium griseum direct callosal pathfinding at the midline by secreting guidance factors Wnt5a and Slit2 (Shu and Richards, 2001;Shu et al., 2003c;Keeble et al., 2006). Additional works have shown that for EphB1/B2/B3/A4 located on callosal axons, the potential binding partners, ephrins B1/B2/B3 are expressed by strategic midline guidepost glia and may use forward signal to guide axons (Mendes et al., 2006). Interestingly, ephrin B1 and B2 present on callosal axons will act by reverse signaling after binding EphB2 expressed by the glial cells of the glial wedge and indusium griseum (Mendes et al., 2006;Bush and Soriano, 2009).
However, we previously demonstrated that guidepost glutamatergic neurons of the CC have an attractive activity on callosal axon navigation, partly via Sema3C-Npn1 signaling (Niquille et al., 2009). In addition, we show here that CC axonal guidance requires the additional presence of two other GABAergic neuronal subpopulations. Our screen on guidance molecules clearly indicates that the CC GABAergic neurons express different combinations of guidance molecules or receptors such as Sema3A, ephrinB1 and EphB2 that have been already shown to be involved in CC development (Mendes et al., 2006;Bush and Soriano, 2009;Zhao et al., 2011). Based on these expression profiles and our results, one can hypothesize that during development multiple interactions may take place between the ephrins and their respective Eph receptors either between callosal axons and guidepost glia, and/or between callosal axons and guidepost neurons. In addition, CC guidepost glutamatergic neurons by expressing the Sema3C, and GABAergic neurons by expressing the Sema3A may collaborate in guiding Npn-1 callosal axons. While the Sema3A may act globally as a repulsive signal on cortical axons (Bagnard et al., 1998;Zhao et al., 2011), this repulsive activity may be converted to attraction when the Npn-1 receptor forms transmembrane interaction with L1 or depending on the cyclic nucleotide cAMP/cGMP ratios (Castellani et al., 2002;Bashaw and Klein, 2010). Neuropilin receptors require a signaling co-receptor to mediate semaphorin function (Yazdani and Terman, 2006;Mann et al., 2007). The transducers and signaling pathway that mediate Npn-1 responses on callosal axon guidance remain so far to be defined. Furthermore, the guidance cues expressed by CC GABAergic neurons (e.g. Sema3A) may play a role not only in axonal guidance but also in other functions such as cell migration, dendrite and branch formation, as well as synaptogenesis (Polleux et al., 2000;Mann et al., 2007;Chen et al., 2008;Bashaw and Klein, 2010;Marin et al., 2010).

CONCLUSION
As a model, we propose that the CC formation requires the presence of guidepost glial cells as well as specific guidepost neuronal populations. Glia and neurons by secreting short and long range distance guidance cues cooperate to guide callosal projections. Therefore, the relative contribution of individual neuronal or glial populations in the guidance of callosal axons remains largely to be investigated to understand CC guidance mechanisms. Specific inactivation of genes for guidance factors in guidepost neurons or glia and specific cell ablation strategies will probably allow us in the future to get a better picture of the involved underlying mechanisms.