We developed an organotypic slice culture model that maintains the anatomical organization of the SVZ-OB system (Figure 1A) to study the migration and the first steps of integration of neuronal precursors in the OB circuit at early postnatal periods. For this study, the slices were obtained from 7-day-old mice and maintained for up to 12 days in vitro (div). Immunohistochemical staining performed at 6 div with antibodies directed against a neuronal precursor marker (doublecortin (DCX); n = 3 slices; Additional file 1) and more mature neuronal markers (NeuN, tyrosine hydroxylase (TH) and GABA; n = 3 slices for each; Additional files 1 and 2) were in agreement with the expected distribution patterns observed at similar postnatal ages, as visually compared to the Allen Institute for Brain Science atlas for P14 animals (and P7 when available) for TH and DCX. Similarly, our labeling corresponded to previously published data from adult slices for GABA in the RMS 12 and NeuN in the OB 28 29 . Together, these observations suggested that overall OB organization was preserved in our slice model.

A total of 132 cultured slices were transduced at 1 div with lentiviral vectors to induce the expression of GFP and to manipulate their levels of NKCC1 expression in SVZ precursor cells. After 1 to 2 days post-injection (dpi), strong GFP labeling was visible at the injection site (Figure 1A,B). Several GFP⁺ cells were distributed along the RMS shortly after vector injection (>3 dpi; Figure 1C). These labeled cells displayed typical morphology of migrating neuroblasts, with a small and elongated cell body (6 to 8 μm), including leading and trailing processes 30 . The first GFP⁺ cells reached the GCL 5 days after vector injection (Figure 1D). Six days after injection, scattered cells were observed throughout the bulb, including in the glomerular layer (Figure 1E). After reaching the appropriate layer, the morphology of GFP⁺ cells changed, with cells transforming into more mature OB interneurons including mainly GC and less numerous PGCs.

The use of lentiviral vectors precludes precise birth-dating of newborn neurons. Nonetheless, from morphological and physiological points of view, the population of neuroblasts studied was homogeneous during the course of our analysis (around 10 days). We used the measurement of input membrane resistance (IR) as an index of cell maturation 31 and recorded GC in addition to RMS neuroblasts for comparison. In the RMS, we did not find a correlation between the IR and the number of dpi (n = 15 from 11 slices; P = 0.36; Additional file 3A), while a significant decrease in IR over time (P = 0.05) was seen for GFP⁺ cells recorded in the GCL (n = 21 from 19 slices; Additional file 3B), consistent with the expected maturation of newborn GCs.

To test the potential effects of NKCC1 on neuroblast migration and their electrophysiological properties, we used two different strategies. The first consisted of the pharmacological blockade of NKCC1 using the broad-spectrum blocker bumetanide 32 33 . To check for potential unspecific effects of bumetanide, we repeated all our measurements in cells transduced with a well-characterized RNA interference sequence to knock down the expression of NKCC1 21 23 34 .

To validate the use of the short hairpin RNA (shRNA) sequence designed and characterized by Ge et al. 21 , we performed immunohistochemical staining for NKCC1 on RMS neuroblasts. We compared the staining in cells transduced with either a lentiviral vector encoding GFP and a non-encoding control sequence (hereafter called the control) or a vector encoding GFP and the shRNA for NKCC1 (hereafter called shNKCC1). Two sets of experiments were done. First, we compared the number of neuroblasts coexpressing GFP and NKCC1 in slices transduced with either the control or the shNKCC1 encoding sequences after 1 and 6 dpi (Figure 2). Second, we compared the number of GFP⁺/NKCC1⁺ cells in both the RMS and the GCL at 6 dpi transduced with either one of the sequences (Additional file 4). In both sets of experiments, we found a significant difference between groups using the Kruskal-Wallis test: values for each experiment were H = 33.82 (P = 0.1 × 10^-4) and H = 38.13 (P = 0.1 × 10^-4). Subsequent Mann-Whitney comparisons between the groups showed a significantly higher proportion of neuroblasts positive for NKCC1 in slices transduced with the control sequence compared to those transduced with shNKCC1 at 6 dpi: values changed in the first experiment from 0.69 ± 0.05 (n = 85 cells) to 0.13 ± 0.07 (n = 41 cells), P = 1 × 10^-3 and in the second experiment from 0.80 ± 0.09 (n = 58 cells) to 0.25 ± 0.11 (n = 63 cells), P = 1 × 10^-4. In contrast, the proportion of NKCC1⁺ GCs was small in control slices (0.22 ± 0.05; n = 27 cells) and did not significantly change in shNKCC1 slices at 6 dpi (0.05 ± 0.03; n = 17 cells; P = 0.09). In slices injected with the shNKCC1-encoding vector, we observed a significant reduction in the proportion of GFP⁺/NKCC1⁺ neuroblasts when comparing results at 1 and 6 dpi (from 0.81 ± 0.03 to 0.13 ± 0.07; n = 22 and 41 cells; P = 1 × 10^-3). Meanwhile, no difference was observed between 1 and 6 dpi in slices injected with the control vector (0.73 ± 0.03 and 0.69 ± 0.05; n = 67 and 85 cells; P = 0.25). The significance reported in this section takes into account the adjustment for P-values using the Bonferroni method.

Our immunohistochemical studies show that the shRNA interference strategy used in this study is efficient to reduce the expression of NKCC1 in neuroblasts, and coincides with its previously proven efficacy in the hippocampus 21 34 and the cortex 23 . Moreover, our results show that the expression of NKCC1 is reduced when the newborn cells leave the RMS to integrate the GCL.

We explored whether the diminution in the activity of NKCC1 could impair the migration of neuroblasts in the RMS. For this, we compared the tangential migration of cells transduced with the non-target shRNA-GFP (control) against control cells in slices incubated in 10 μM bumetanide 1 hour before and during the time lapse recording. A similar approach was employed with cells transduced with the shNKCC1-GFP virus. We calculated the average migration speed and the total displacement (∆X) along the horizontal axis of the RMS in 202 cells (from 9 slices) for controls; 61 cells (4 slices) incubated in bumetanide, and 83 cells (from 8 slices) transduced with shNKCC1, in 7 independent experiments. Control neuroblasts migrated with an average speed of 42.2 ± 2.0 μm/h. The migration speed became significantly slower in cells incubated in bumetanide (32.4 ± 4.7 μm/h; P = 0.02) or transduced with shNKCC1 (31.9 ± 3.1 μm/h, P = 0.007) (Figure 3A; Additional files 5 and 6).

Figure 3B illustrates RMS neuroblasts migrating back and forth with respect to the OB. To investigate whether NKCC1 was required for determining the net tendency of the cells to migrate towards the OB, we compared the total displacement of cells along the x-axis, and used it as an index for directionality (Figure 3C). The initial cell position was set to zero and the values of displacement were considered negative when cells moved towards the injection site (that is, the SVZ) and positive when cells headed towards the bulb. We found no significant differences in the values of ∆x between control cells and cells incubated in bumetanide or transduced with the shNKCC1 (P = 0.9 and 0.6, respectively).

GABA acting on GABA_A receptors reduces the migration of neuroblasts 12 . In order to test whether the effects observed by reducing the activity of NKCC1 were due to changes in the action of GABA_A receptors, as a consequence of a potential drop in [Cl^-]_i, we analyzed the effects of blocking GABA_A receptors with 10 μM gabazine in slices transduced with the control sequence, in slices transduced with the control sequence and treated with bumetanide, and in slices transduced with the shNKCC1 sequence.

We analyzed 69 cells transduced with the shNKCC1 sequence (from 4 slices), 43 cells treated with bumetanide (from 3 slices) and 112 cells transduced with the control sequence (from 5 slices); all were recorded in the presence of gabazine during the experiments. Our results were compared using a two-way ANOVA, which showed significant differences between the group injected with the control sequence, the group injected with control sequence and treated with bumetanide and the shNKCC1 sequence group (P = 1 × 10^-6), as well as a significant effect of gabazine (P = 1 × 10^-6). The differences between individual groups were then evaluated using a Tukey posthoc test. The comparisons show an increase in the average migration speed due to the GABA_A receptor blockade in the three groups (Figure 4A). The average speed of control cells treated with gabazine was significantly faster than in the non-treated control group (53.9 ± 1.5 compared to 42.2 ± 2 μm/h; P = 1 × 10^-3; Figure 4A). Gabazine also significantly accelerated the speed of migration when comparing bumetanide groups (from 34.4 ± 2.8 to 45.5 ± 2.5 μm/h; P = 0.01) and shNKCC1-treated cells (from 31.9 ± 3.1 to 42.5 ± 1.6 μm/h; P = 0.01). The cumulative distribution curve shifted to the right in all the groups treated with gabazine, compared with their respective controls (Figure 4B). The Kolmogorov-Smirnov test revealed no significant difference in the distribution of the average speed between control groups and bumetanide or shNKCC1 groups treated with gabazine (P = 0.99 and 0.13). Overall, the analysis shows a very significant effect of NKCC1 pharmacological and genetic blockade as well as a significant effect of GABA_A-receptor blockers on the speed of migration of neuroblasts. Since the interaction between these two variables was not statistically significant (P = 0.78), we conclude that the role of NKCC1 on the maintenance of migration speed is independent of the modulation exerted by GABA_A receptors.

To explore the effects of GABA_A signaling on the directionality of the migration, we analyzed the displacement of the neuroblasts along the x-axis, as explained in the previous section. We did not observe any effect on direction after the blockade of GABA_A receptors in any of the groups treated with gabazine - that is, control, bumetanide and shNKCC1 groups - compared to their respective non-treated groups (P = 0.8, 0.56 and 0.7, respectively; Figure 4C). Altogether, our results show that neither NKCC1 nor the GABA_A-mediated activity are important in determining the directionality of RMS migrating neuroblasts.

Concerning the speed of migration, our experiments show that activation of GABA_A receptors diminishes the migration speed in a way compatible with previous observations 12 . Moreover, the experiments performed in the presence of gabazine showed that the effects of NKCC1 on neuroblast migration do not depend on GABA_A signaling.

Our whole-cell recordings show that migrating neuroblasts do not receive spontaneous synaptic inputs, confirming previous studies 13 . The lack of synaptic inputs in neuroblasts holds even after treatment with alpha-latrotoxin to trigger massive neurotransmitter release (data not shown, n = 3 cells from 3 slices) 35 36 . In contrast, newly formed GCs displayed spontaneous synaptic activity, which started at 6 dpi and significantly increased at 10 dpi.

Due to the lack of synaptic activity in neuroblasts, the action of GABA on the migration speed must depend on a tonic influence of this neurotransmitter. This mechanism has been suggested 12 but the tonic currents have never been recorded. Thus, we recorded the spontaneous activity of GFP⁺ cells (5 to 11 dpi) held at -60 mV before and after applying the GABA_A receptor blocker gabazine (10 μM). GFP⁺ cells were either migrating neuroblasts located in the RMS (Figure 5A) or, for comparison, neurons located in the GCL (Figure 5B). In the GCL we paid particular attention to avoid recording from migrating GFP⁺ cells, most likely representing newborn PGCs en route to the glomerular layer. Our results confirm the tonic action of GABA_A receptors in neuroblasts and extend the observation to GCs. Moreover, they show that 10 μM gabazine, the same concentration used in our migration studies, efficiently blocks the tonic GABA_A-mediated currents, which became evident from a diminution in holding current of 4.0 ± 0.7 pA in migrating neuroblasts (n = 10 cells from 9 slices; P = 9e-4, paired t-test) and of 3.1 ± 1.2 pA in GCs (n = 6 cells from 6 slices; P = 0.02, paired t-test) (Figure 5C) 37 .

In order to understand the effects observed on neuroblast migration, we decided to determine whether and, if so, how the activity of NKCC1 affects the action of GABA_A receptors and the overall excitability of the neuroblasts. For this we measured E_GABA (the reversal potential for GABA_A mediated responses) in RMS cells transduced with the control sequence and incubated in 10 μM bumetanide or in cells transduced with the shNKCC1 sequence. As a reference, we repeated our measurements in newborn cells at a more mature stage, that is, in GFP⁺ GCs.

Gramicidin-perforated patch-clamp recordings were carried out to determine [Cl^-]_i, and the strength of GABA_A-mediated responses. A voltage-ramp from -80 to +20 mV was applied in the presence and absence of a focal application of 10 μM muscimol (pressure ejected near the cell soma). The resulting currents were subtracted to calculate E_GABA (Figure 6Ai,Bi). Our results show that NKCC1 is an important modulator of GABA action in neuroblasts, since the diminution of its activity by either of the means tested produced a significant hyperpolarization of E_GABA. E_GABA in migrating RMS neuroblasts shifted from -29.8 ± 1.7 mV (n = 6 cells from 6 slices) in control to -49.7 ± 2.8 mV in bumetanide (n = 9 from 6 slices; P = 0.01) and to -56.7 ± 2.3 mV in shNKCC1-treated neuroblasts (n = 8 from 7 slices; P = 4 × 10^-4) (Figure 6Bii).

In contrast, new GCs had E_GABA values significantly more hyperpolarized than neuroblasts in the control (-57.5 ± 4.1 mV; n = 13 from 13 slices; P = 6.9 × 10^-5; Figure 6Aii) and this value was not sensitive to either bumetanide -58.6 ± 3.6 mV (n = 12 from 10 slices; P = 0.99) or shNKCC1 treatment (-52.8 ± 3.3 mV; n = 6 from 5 slices; P = 0.94) (Figure 6Bii).

We controlled for possible bias in our measurements due to the concentration of Cl^- in the patching solution. The comparison of E_GABA values between independent groups of cells using pipette solutions containing either a low or a high [Cl^-] (solutions 1 and 2 in Materials and methods) did not show any difference (P = 0.47) so the results were pooled. Furthermore, an additional control using immunohistochemical labeling for KCC2 showed that the hyperpolarization of E_GABA in neuroblasts transduced with shNKCC1 is not due to a compensatory expression of KCC2, since no positive KCC2 labeling was observed in either GFP⁺ neuroblasts or GFP⁺ GCs (Additional file 7).

We used E_GABA values to estimate [Cl^-]_i in both neuroblasts and GCs in the control and after treatments (Figure 6C). According to the effects on E_GABA, the calculated values of [Cl^-] decreased from 43.5 ± 3.1 mM in control neuroblasts to 21.4 ± 2.3 mM (P = 6.14 × 10^-5) in bumetanide- and 15.9 ± 1.3 mM (P = 1.09 × 10^-5) in shNKCC1-treated cells. In newly formed GCs the values of [Cl^-] in bumetanide- (16.4 ± 2.6 mM; P = 0.99) and shNKCC1-treated cells (17.2 ± 2.8 mM; P = 1) were not significantly different from those in the control (17.5 ± 2.7 mM). The lack of effect of the treatments diminishing NKCC1 activity on GCs agrees with the low expression of NKCC1 observed in the immunohistochemical analysis (Additional file 4B,C).

The values of E_GABA and [Cl^-] were constant in each cell type for the duration of our study as no significant correlation between E_GABA and dpi was observed in Spearman correlation tests (r = -0.11 for neuroblasts and r = -0.22 for GCs; Additional file 8).

To understand the net effect of the activation of conductance on a cell's membrane potential, it is useful to determine the value of the potential at rest. The resting membrane potential (RMP) was measured in the cell-attached mode as the equivalent of the reversal potential of K⁺ conductances using isotonic intra- and extracellular concentrations of this ion 38 . We did verify that, in our system, this method was sensitive to predictable changes in RMP by measuring the reversal of K⁺ currents before and after raising the extracellular [K⁺] to 10 mM. This manipulation produced an average 3.4 ± 1.6 mV depolarization in the four cells tested.

We then measured the RMP in migrating neuroblasts and GCs in control slices. Remarkably, we did not find significant differences in RMP values between these cell groups: -71.5 ± 3.6 mV (n = 6 from 5 slices) and -71.0 ± 4.4 mV (n = 6 from 6 slices; P = 0.93, t-test), respectively. Nonetheless, the comparison of RMP in neuroblasts treated with bumetanide or shNKCC1 showed a depolarization due to the diminution of NKCC1 activity. The measured values were -59.8 ± 3.2 mV (n = 6 from 5 slices; P = 0.07) for bumetanide and -58.7 ± 2.9 mV (n = 9 from 5 slices; P = 0.02) in cells expressing shNKCC1 (Additional file 9). In contrast, in GCs the diminution of NKCC1 activity did not produce significant changes in RMP compared to control; values of -67.0 ± 1.0 mV (n = 2 from 2 slices; P = 0.90) in bumetanide- and -69.4 ± 6.1 mV (n = 5 from 5 slices; P = 0.97) in shNKCC1-treated cells.

We used the E_GABA values from muscimol-mediated currents and the RMP values to calculate the driving force for GABA-mediated responses. In neuroblasts, both the hyperpolarization of E_GABA as well as the depolarization of about 11 mV in their RMP, observed in bumetanide and shNKCC1 groups, resulted in an important decrease in the driving force for GABA responses, with the strongest effects on the shNKCC1-treated group. The calculated driving force value in control neuroblasts (41.7 ± 2.8 mV) was significantly diminished to 10.2 ± 2.9 mV (P = 2.84 × 10^-5) in bumetanide- and to 2.0 ± 2.6 mV (P = 1.03 × 10^-5) in shNKCC1-treated cells. Moreover, as expected from the lack of effect of the diminution of NKCC1 activity on both E_GABA and RMP in GCs, we did not observe any effect on the driving force between control GCs (13.5 ± 4.2 mV) and the bumetanide- (10.81 ± 3.6 mV; P = 0.98) or shNKCC1-treated groups (16.6 ± 4.8 mV; P = 0.99) (Figure 6Aiii).

A comparison between cell types - neuroblasts versus GCs - in the control condition shows a significant decrease in the driving force for GABA_A-mediated responses in newborn interneurons when reaching the bulb (P = 5.28 × 10^-5; Figure 6Aiii,Biii). Nonetheless, GABA maintained a depolarizing action on both cell types. Our results indicate that NKCC1 expression is diminished in GCs from the time of their arrival at the OB. Therefore, in sharp contrast to neuroblasts, neither the action of GABA nor the values of RMP depend on NKCC1 activity in developing GCs.

Finally, we verified whether gabazine was able to change the RMP in neuroblasts, as would be expected from the measured driving forces for muscimol responses. In control cells, gabazine induced a 2.3 ± 0.8 mV (n = 3 cells) hyperpolarization of the RMP while only a 0.45 ± 0.04 mV hyperpolarizing shift was observed in shNKCC1 neuroblasts (n = 2). The lesser effect of gabazine on RMP in shNKCC1 cells supports the diminished driving force for muscimol observed in these cells. Moreover, the lesser effect on RMP observed in shNKCC1 cells might be related to the decreased effect of gabazine on the acceleration of neuroblast migration in the bumetanide and shNKCC1 cells (Figure 4).

In the present study we used organotypic slice cultures to compare the role of NKCC1 on the excitability of neuroblasts and in maturing newborn neurons. The use of this model seems particularly appropriate for the present study since both the neuroblast migration and GC integration occur in the organotypic control, in a similar way to those reported from acute models 12 31 . First, the speed of migration in our system is remarkably similar to previous values measured in acute slices 12 . Moreover, the labeled precursors evolve in the organotypic culture, first as a contingent of neuroblasts migrating in the RMS to increasingly populate the OB. Once in the GCL, the cells change their morphology and show functional maturation (Figure 1; Additional file 3), including the development of synaptic activity. A similar system has been previously used to study the control of differentiation in dopaminergic cells 54 .

Our results show that the early maturation of OB interneurons, at least when considering GCs, is paralleled by changes in both GABAergic signaling and NKCC1 function. After reaching the bulb, the neural precursors acquire a mature morphology and establish synaptic interactions with pre-existing neurons, some of which are GABAergic 55 . Our results show that GABA action evolves from a non-synaptic, temporally and spatially diffuse tonic influence in the RMS to include synaptic contacts on GCs during the first weeks of development. The tonic influence of GABA was preserved in GCs during the entire duration of our study, regardless of the development of synaptic contacts. These results are in agreement with a previous report showing that, in early postnatal tissue, GABA produced a depolarizing influence, capable of shunting the firing activity on GCs 56 . Correspondingly, we detected a small tonic current and a slightly positive driving force for Cl^- in GCs (Figures 5 and 6A iii). Slightly positive driving forces for GABA are frequently observed in immature interneurons 57 58 59 and are believed to favor neuronal development 14 .

The manipulations affecting the activity of NKCC1 showed that the strong depolarizing action of GABA in RMS neuroblasts largely depends on this Cl^- transport system. Meanwhile, in GCs, GABA exerted a much weaker depolarizing driving force that was not affected by the manipulations blocking the activity of NKCC1. These results are confirmed by the immunohistochemical data showing a much higher proportion of neuroblasts expressing NKCC1 compared to GCs in slices injected with the control sequence, and a significant drop of NKCC1⁺ neuroblasts but not GCs in cells transduced with shNKCC1 (Additional file 4). In accordance with the functional and anatomical evidence of low expression of NKCC1 on GCs, we failed to find any effect of the manipulation of this Cl^- transporter on the development of GC dendrite projections (data not shown). Thus, it seems unlikely that GC synaptic connectivity in the OB is shaped by NKCC1, in contrast to the dentate gyrus 21 .

Wang and colleagues 56 showed that, after the second postnatal week, GABA hyperpolarizes the GC, held at -60 mV, at the time that KCC2 mRNA appears. We observed a strong hyperpolarization of E_GABA when GFP⁺ cells reached the GCL. However, GABA action did not become hyperpolarizing during our study, possibly owing to delays in this process due to in vitro conditions, or because our recorded cells were young and did not yet express KCC2 (Additional file 7B). Several reports have shown that KCC2 mRNA expression is abundant in mature neurons, whilst negligible expression was detected in neuronal progenitors 17 60 . It is conceivable that, at later stages of maturation, KCC2 is involved in the synaptic connectivity of GCs, as it is in other systems 40 61 . Nonetheless, our results indicate that the first significant decrease in [Cl^-]_i in GCs, and therefore the simultaneous hyperpolarization of E_GABA, results from a functional down-regulation of NKCC1.

Altogether our data suggest that variations in NKCC1 activity can efficiently control various steps of OB postnatal neurogenesis. Accordingly, we show that an important change in the functional expression of this cotransporter accompanies the developmental transition from neuroblasts to developing GCs. NKCC1 is highly active in the RMS precursor cells and its functional expression drops drastically in the newborn GCs, greatly affecting the action of GABAergic signaling on this early period of functional integration. Concerning the neuroblasts in the RMS, our results show, for the first time, an important role of NKCC1 in the control of neuroblast excitability. Our electrical determinations indicate that NKCC1 controls at least two different aspects of neuroblast activity. The first concerns the control of [Cl^-]_i and therefore the integration of the most prominent neurotransmitter signaling system in the RMS, GABA_A-receptor-mediated activity 12 13 31 43 62 63 . The second aspect concerns the regulation of intrinsic excitability and depends on the control of the RMP. Thus, our study suggests that NKCC1 integrates relevant external signaling and the intrinsic neuronal excitability to finely tune the migration speed of RMS neuronal precursors en route to the OB.

Postnatal day 7 C57BL/6J mouse pups (Janvier, Le Genest Saint Isle, France)) were decapitated and the brains removed and placed into ice-cold artificial cerebrospinal fluid (ACSF) containing (in mM) 124 NaCl, 3 KCl, 1.3 MgSO₄, 26 NaHCO₃, 1.25 NaH₂PO₄, 10 glucose and 2 CaCl₂, saturated with 95% O₂/5% CO₂. Sagittal forebrain slices, 300 μm thick, containing the migratory pathway from the SVZ-RMS and OB were prepared with a vibratome (Leica VT1200). All experimental procedures were performed in accordance with the Charter of Fundamental Rights of the European Union (2000/C 364/01), the European Communities Council Directive of 24 November 1986 (86/609/EEC), and European Union guidelines. They were reviewed and approved by our Institutional Animal Welfare Committee.

Each brain slice was transferred to a Millicell-CM Culture Plate Insert (Millipore PICM ORG 50; Billerica, MA, USA ), the excess ACSF was removed and the insert was placed in a 35-mm Petri dish containing 1 ml of medium, as described by Stoppini and colleagues 64 . Culture medium consisted of 46% MEM (31095-029, Invitrogen, Paisley, UK), 25% HBSS (24020-091, Invitrogen), 25% horse serum, 20 nM HEPES, 6 mg/ml D-glucose and gentamycin (Invitrogen, Paisley, UK). The slices were incubated at 37°C in a humid atmosphere of 5% CO_2. The medium was changed three times per week.

We used lentiviral vectors created using the PTRIPDU3 method to express GFP in neural precursors and neuroblasts 65 . The first vector was constructed such that GFP expression was under the control of the PGK promoter. For the second vector, a previously published shRNA sequence for NKCC1, shNKCC1-1 (5'ACACACTTGTCCTGGGATT3'), was introduced under the control of the U6 promoter 21 . This vector was similar to the first, also containing a sequence to induce GFP expression, but under the control of the CMV promoter. As a control, we constructed a third vector in which the shRNA sequence was replaced with a non-target scrambled sequence (5'GCCAGATTTCTCAGGTGATAA3'). Twenty-four hours after culturing the slices, 270 pg of p24 from one of the constructions was injected in the SVZ, in a volume of 18 nl.

GFP⁺ RMS neuroblasts and olfactory bulb GCs, at 3 to 11 dpi, were recorded using the patch-clamp technique in different configurations as detailed below. For data acquisition and electrical control an EPC-10 amplifier (Heka Elektronik, Lambrecht/Pfalz, Germany ) was used. Before the recordings, a piece of Millicell membrane containing the slice was cut out and transferred to a chamber. During the recording session the slices were continuously perfused at a rate of 1.5 ml/minute with extracellular solution containing (in mM) 124 NaCl, 3 KCl, 2.4 CaCl₂, 1 MgCl₂, 25 NaHCO₃, 1.25 NaH₂PO₄, and 10 glucose. The extracellular solution was heated to 35°C at the entrance of the chamber (inline heater TC-324B, Warner Instruments, Inc; Hamden, CT, USA). Glass pipettes with a resistance of 6 to 7 MΩ were used for the recordings.

The reversal potential of GABA_A-receptor-mediated currents was determined using the perforated patch-clamp configuration with gramicidin - a cation-specific channel - in order to obtain voltage control of the cell while leaving [Cl^-]_i intact 66 . For perforated-patch recordings one of the following patching solutions was used in each experiment: (1) 150 mM KCl, 10 mM Hepes; or (2) 150 mM KGluconate, 10 mM Hepes, both at pH 7.5. A stock of 1 mg/ml of gramicidin was freshly prepared on the day of the experiment and a final concentration of 2 μg/ml was added to the patching solution. The pipette tip was dipped in gramicidin-free patching solution before backfilling it with the gramicidin-containing solution. A giga-seal was established with the cell membrane and the recording session began when a stable access resistance between 50 and 70 MΩ was obtained.

To determine the reversal potential of the currents elicited by the GABA_A receptor agonist muscimol, a 1 s voltage-ramp from -80 to +20 mV and back to -80 mV was applied to cells recorded in the perforated configuration. Muscimol (10 μM) was pressure ejected (10 to 20 psi) near the cell membrane during the ramp. At least three ramps were applied and averaged before and during muscimol application. The averaged currents were subtracted and the resulting trace was used to measure the reversal potential of the muscimol-triggered current. Reversal potentials were determined by linear fitting of the current trace using Igor (Wavemetrics, Inc; Lake Oswego, OR, USA), whereas [Cl^-]_i was estimated from the Nernst equation. Repeated agonist application can influence the apparent Cl^- reversal potential (E_GABA), particularly in immature neurons in which Cl^- regulation mechanisms are less effective 67 . Therefore, we ensured that the interval between agonist applications was sufficiently long (20 s) to avoid shifts in E_GABA.

To determine the cells' RMP we used the method described by Verheugen et al. 38 . Briefly, we kept the cells in the cell-attached mode at -60 mV and applied a voltage-ramp from -100 to +200 mV to activate a voltage-dependent K⁺ conductance. Patching solutions 1 or 2, described above, were used in the absence of gramicidin for these determinations. Both solutions have a concentration of 150 mM K⁺; therefore, the reversal potential of the K⁺ conductance was assumed to be equal to the RMP. The difference between average RMP values and the reversal potential of the muscimol-induced current values was used to determine the driving force for GABA in each condition and cell type.

Tonic currents were recorded in the whole cell mode at -60 mV holding potential. For these recordings solution 1 was supplemented with 2 mM ATPMg, 0.2 mM GTPNa and 0.2 mM EGTA. We recorded the spontaneous activity of the cells held at -60 mV for at least 2 minutes in normal ACSF before applying 10 μM gabazine to the bath, and then recorded spontaneous activity for 5 minutes after the application of the drug. To measure the tonic GABA_A current, we assessed the average holding current necessary to maintain the voltage at -60 mV, during 10 s in control ACSF, starting 40 s before application of gabazine, and 30 s after its application in the bath. Bicuculline blocks SK-type K⁺ channels 68 and may affect the measurement of tonic current. Thus, gabazine is the drug of choice for studying tonic GABA conductance. For one group of cells, GABA_A-dependent tonic currents were determined in the cells recorded first in the perforated-patch configuration and subsequently transformed into whole-cell mode by gentle suction within the recording pipette (n = 6 neuroblasts and 3 GCs). A second group was recorded in whole cell mode in the absence of gramicidin on the intracellular pipette (n = 4 neuroblasts and 3 GCs). No significant differences were observed between the values obtained with either method, neither for the RMS neuroblasts (P = 0.57) nor for the GCs (P = 0.51); thus, the results were pooled.

Only the cells maintaining a stable access resistance (less than 20% variation during the experiment) were analyzed. The series resistance values were between 10 and 25 MΩ and no compensation was applied. The spontaneous currents were filtered at 2.9 KHz with a low-pass Bessel filter and were sampled at 8.33 KHz. To analyze the passive properties of cells, the currents generated by a 20 ms, 20 mV hyperpolarizing pulse were filtered at 5 kHz and sampled at 10 kHz. All values were corrected for the measured liquid junction potential (-2 and -8 mV for solutions 1 and 2, respectively).

All drugs were purchased from Sigma-Aldrich and used at the following concentrations: 10 μM bumetanide was used to block NKCC1 activity. In all the experiments using bumetanide the slices were pre-incubated for 1 h in 10 μM bumetanide and the same concentration was present during our electrophysiological and imaging recordings (Sigma-Aldrich, Saint-Quentin Fallavier, France). 10 μM gabazine and 10 μM muscimol were used as antagonist and agonist of GABA_A receptors, respectively, and 10 nM alpha-latrotoxin was used to trigger neurotransmitter release.

Organotypic brain slices were fixed in 4% paraformaldehyde before being cryoprotected in 12.5% sucrose, 5% glycerol for 2 h at 4°C and incubated overnight in 25% sucrose, 10% glycerol. Slices were rapidly frozen and thawed three times over liquid N₂ vapor, washed in PBS before a second fixation in paraformaldehyde for 15 minutes, and incubated for 2 h in 10%, NGS 0.1% Triton blocking solution at room temperature. For immunolabeling of chloride cotransporters, slices previously injected with one of the lentiviral vectors were then incubated at 1 or 6 dpi in the primary antibody for NKCC1 (1:200; T4, Developmental Studies Hybridoma Bank) or at 6 dpi in the antibody for KCC2 (1:200; 07-432, Upstate, Billerica, MA, USA) mixed with the anti-GFP (1:1,000; 06-896, Upstate). A different set of organotypic slices at 6 div were incubated in the following primary antibodies: neuronal nuclei marker NeuN (1:200; MAB377, Chemicon, Billerica, MA, USA; n = 3), DCX (1:1,000; Ab18723, Abcam, Cambridge, UK; n = 3), GABA (1:1,000; A2052, Sigma; n = 3) or TH (1:4,000; 22941, ImmunoStar, Hudson, Wisconsin, USA; n = 3). All primary antibodies were diluted in blocking solution and incubated for 48 h at 4°C, followed by washes in PBS. Then, slices were incubated for 2 h in an alexa-conjugated secondary antibody (Alexa Fluor 568 (1:750) for the chloride transporters and Alexa Fluor 488 (1:800) for the rest of the proteins; both from Molecular Probes, Paisley, UK) at room temperature and washed in PBS. NeuN, DCX, GABA and TH stained cultures were then incubated in DAPI for 15 minutes at room temperature and washed in PBS. All the slices were then mounted in ProLong Gold antifade (P36930, Molecular Probes). For analysis, images of 0.5 to 1 μm thick were acquired either in an upright microscope (Zeiss), equipped with a confocal spinning disk unit (Andor, Belfast, UK) or in an inverted microscope (Axioscope FS, Zeiss), equipped with a structured illumination device (Apotome, Zeiss).

To determine the number of GFP⁺ cells in the field that were also positive for NKCC1, we used the Intensity Correlation Analysis package of ImageJ 69 . The region analyzed was first defined by the limits of the GFP signal of a cell in the z-axis. Signal was considered a cell when the morphology corresponded to a round or elliptical cell body of 6 to12 μm with at least one projection. We considered as double labeled those cells showing a positive intensity correlation quotient for both GFP and alexa 568 signals. The number of labeled cells per field (313 × 313 μm) was determined in three independent images for each region of interest (RMS and GCL), and the results were normalized to the total number of GFP⁺ cells analyzed in each image and then averaged (Figure 2; Additional file 4).

The Petri dish containing the cultured slice was transferred to a closed chamber with the temperature control adjusted to maintain the medium at 36°C (DH40i, Warner Instruments, Inc.). During the entire time-lapse session, the tissue was oxygenated using a mixture of 95% O₂/5% CO₂. To prevent evaporation of the culture medium, the internal environment of the chamber was continuously humidified (Humidifying module warmer HMW-1, Warner Instruments, Inc.). Mosaic images of the entire RMS were re-constructed from individual frames (x-, y- and z-axis), acquired every 20 minutes during a period of 3 h. We used an inverted fluorescence microscope (Axioscope FS, Zeiss) equipped with a 20× long-distance objective (LD plan Neofluar, Zeiss). For analysis, a maximal projection of the different planes taken along the z-axis was obtained for each time-lapse. Individual RMS neuroblasts could be identified and their trajectories were tracked using Metamorph (Molecular Devices). Two migration variables were analyzed, the speed and displacement along the x-axis, corresponding to the tangential displacement in the RMS, provided that the rostro-caudal axis of the RMS was positioned parallel to the x-axis (Figure 3B,C). Delta x (∆x), the total displacement of a cell along the rostro-caudal axis of the RMS during an experimental session, was used as an index of the direction of migration in the RMS-OB pathway.

All results are expressed as averages ± standard error of the mean. Experiments involving the comparison of a control and an experimental group were analyzed using the Student's t-test. For multiple comparisons either one- or two-way ANOVA were used with the Tukey method as a post-hoc test; the use of each is detailed in the Results section. For multiple comparisons of the immunohistochemical data Kruskal-Wallis and Mann-Whitney methods were used; in these cases significance was evaluated based on the Bonferroni adjustment of P-values. Differences were considered significant for values of P < 0.05. Distributions were compared using the Kolmogorov-Smirnov test in the R program 70 .