KCC2 expression promotes the termination of cortical interneuron migration in a voltage-sensitive calcium-dependent manner

Summary

The molecular mechanisms controlling the termination of cortical interneuron migration are unknown. Here we demonstrate that prior to synaptogenesis, migrating interneurons change their responsiveness to ambient GABA from a motogenic to a stop signal. We found that during migration into the cortex, ambient GABA and glutamate initially stimulate the motility of interneurons through both GABA_A and AMPA/NMDA receptor activation. Once in the cortex, up-regulation of the potassium-chloride co-transporter KCC2 is both necessary and sufficient to reduce interneuron motility through its ability to reduce membrane potential upon GABA_A receptor activation which decrease the frequency of spontaneous intracellular calcium transients initiated by L-type Voltage-Sensitive Calcium Channels (VSCC) activation. Our results suggest a novel mechanism whereby migrating interneurons determine the relative density of surrounding interneurons and principal cells through their ability to sense the combined extracellular levels of ambient glutamate and GABA once GABA_A receptor activation becomes hyperpolarizing.

INTRODUCTION

Balance between excitation and inhibition in cortical circuits is dictated in part by the relative number of excitatory glutamatergic pyramidal neurons and inhibitory GABAergic interneurons. This balance is of critical importance for the proper function of the adult neocortex (Rubenstein and Merzenich, 2003). Although the mechanisms stimulating the motility and guiding the migration of cortical interneurons are beginning to be unraveled (Flames et al., 2004; Marin et al., 2001; Polleux and Ghosh, 2002; Poluch et al., 2003; Powell et al., 2001), the extracellular cues and signaling pathways instructing when and where cortical interneurons stop migrating are currently unknown.

The mode of migration of pyramidal neurons and interneurons differs greatly, and these differences include the cellular constrains leading to the termination of their migration. Pyramidal neurons are born from asymmetric divisions of radial glial progenitors in the ventricular zone of the dorsal telencephalon (Noctor et al., 2001), migrate radially towards the pial surface by translocating along radial glial processes (Kriegstein and Noctor, 2004; Rakic, 1972) and terminate near the top of the CP by detaching from their glial substrate (Dulabon et al., 2000; Pinto-Lord et al., 1982). On the other hand, interneurons migrate dynamically in a saltatory, start-stop fashion from the medial and caudal ganglionic eminences (the MGE and CGE respectively), to the dorsal telencephalon where they migrate tangentially through the marginal zone (MZ) and intermediate zone (IZ) (Ang et al., 2003; Lavdas et al., 1999; Marin and Rubenstein, 2001; Marin et al., 2001; O'Rourke et al., 1992; O'Rourke et al., 1995; Polleux et al., 2002; Tanaka et al., 2006). Although interneurons can transiently fasciculate with radial glial fibers during their invasion of the cortical plate (CP) (Polleux et al., 2002), they are most frequently seen moving tangential to the direction of radial glial processes even within the CP (O'Rourke et al., 1995; Polleux et al., 2002; Bortone and Polleux unpublished observations). Therefore, unlike pyramidal neurons, for which detachment from the radial glial scaffold at the top of the CP is thought to be a determining factor, the absence of a required substrate for interneuron migration, obfuscates the spatial and temporal mechanisms that might underlie the termination of their migration.

Some phenotypic features of cortical interneurons are genetically specified by the expression of transcription factors including Arx, Dlx1/2, Nkx2.1 and Lhx6 in the medial ganglionic eminence (MGE) (Anderson et al., 1997; Colombo et al., 2007; Kitamura et al., 2002; Lavdas et al., 1999; Sussel et al., 1999; Zhao et al., 2008). Lhx6-expressing interneurons originate from the MGE and primarily differentiate into the parvalbumin-positive subpopulation of cortical interneurons (Cobos et al., 2005; Cobos et al., 2006; Liodis et al., 2007; Zhao et al., 2008), which comprises basket cells and chandelier cells making restricted synaptic contacts on the soma and axon initial segment of pyramidal neurons, respectively.

Gamma-aminobutyric acid (GABA), the primary inhibitory neurotransmitter of the central nervous system, has been proposed to play multiple roles in controlling neuronal migration. GABA receptors are typically categorized into three types: GABA_A, GABA_B, and GABA_C. GABA_A and GABA_C are ionotropic receptors composed of five heteromeric subunits and are predominantly permeable to chloride ions. These two ion channels can be distinguished pharmacologically as GABA_C receptors are not inhibited by bicuculline, a GABA_A antagonist. GABA_B receptors are metabotropic (G-protein coupled receptors) and are therefore not dependent on the electrochemical equilibrium of chloride ions for their function. Tangentially migrating interneurons synthesize GABA and also possess the means to release it through a non-vesicular mechanism (Manent et al., 2005) that might involve reversal of GABA transporters (Conti et al., 2004). Migrating interneurons also have the capacity to respond to GABA in an autocrine/paracrine manner since they express both ionotropic GABA_A and metabotropic GABA_B receptors during their migration (Cuzon et al., 2006; Lopez-Bendito et al., 2003; Lujan et al., 2005), yet the function of this complete signaling package in migrating interneurons is poorly understood (Represa and Ben-Ari, 2005).

In fact, GABA has been suggested to exert many, sometimes conflicting, effects on neuronal proliferation, migration and differentiation in the developing central nervous system (Andang et al., 2008; Behar et al., 1996; Behar et al., 1998; Bolteus and Bordey, 2004; Cuzon et al., 2006; Lopez-Bendito et al., 2003; LoTurco et al., 1995; Marty et al., 1996). Here we report that upon reaching the cortex, depolarization through activation of GABA_A receptors or activation of NMDA/AMPA glutamate receptors induce calcium transients that stimulate interneuron motility. These effects require activation of L-type VSCCs. Interestingly, several days after reaching the cortex, migrating interneurons upregulate the potassium/chloride (K⁺/Cl⁻) exchanger KCC2 (also called Solute Carrier Family 12, Member 5; SLC12A5) which controls the reversal potential of Chloride (Cl⁻) ions and therefore determines the developmental switch from depolarizing to hyperpolarizing following GABA_A receptor activation (Ben-Ari, 2002). We demonstrate that expression of KCC2 is both necessary and sufficient to induce responsiveness to GABA as a stop signal by negatively regulating the frequency of spontaneous intracellular calcium transients in migrating interneurons. Therefore KCC2 expression acts as a switch to induce a voltage-sensitive, calcium-mediated reduction of interneuron motility. Our results suggest for the first time that, once GABA_A receptor activation becomes hyperpolarizing in migrating cortical interneurons, these cells are able to sense and integrate the ambient extracellular levels of GABA and glutamate as a way to determine when to stop migrating.

RESULTS

The majority of cortical interneurons terminate migration during the first postnatal week

To first document when interneurons stop migrating in the developing mouse cortex in vivo, we performed real-time confocal imaging of acute cortical slices isolated from a fluorescently labeled BAC-transgenic mouse line provided by the GENSAT consortium (Gong et al., 2003). In this transgenic mouse, enhanced green fluorescent protein (EGFP) is expressed under the control of the regulatory elements of LIM homeobox 6 (Lhx6), a transcription factor expressed specifically in MGE progenitors (Lavdas et al., 1999). These Lhx6-EGFP mice express EGFP in approximately 65% of cortical interneurons in vivo. Most of these Lhx6-positive interneurons will differentiate into parvalbumin-positive large basket cells and chandelier cells postnatally (Cobos et al., 2006). Time-lapse experiments of these Lhx6-EGFP mice revealed a gradual reduction in interneuron migration from E15 to P7 as shown in Figures 1B–D (Movies S1–S3). Although some interneurons have terminated their migration in the CP (CP) by E15 (15.1±2.8%; Figure 1E), the vast majority continues to translocate. By P1, 57.8±1.9% of interneurons stopped migrating whereas almost all cortical interneurons have terminated migration by P7 (91.6±1.4%).

Figure 1. Characterizing the termination of cortical interneuron migration.

(A) Illustration of our time-lapse set-up. See Suppl. Methods for details. (B–D) Merged images showing the decrease in motility of Lhx6-EGFP+ interneurons in cortical slices at embryonic day (E)14.5 (B), post-natal day (P)1 and P7 after ex vivo culture for 1–2 days. Frame corresponding to the start of imaging (t=0) is pseudo-colored in red merged with a frame captured six hours later is pseudo-colored in green. Yellow cell bodies indicate interneurons that did not move during these 6 hours. Arrowheads indicate migrating interneurons while arrows point to sedentary interneurons. (E) Movements of interneurons were quantified in the CP: by P7 the majority of interneurons show no detectable translocation. (F–G) Box plots indicate developmental decline in interneuron motility over time. Individual box indicates mid-50^th percentile range of individual neuron measurements taken. Whiskers extend from 10^th to 90^th percentiles. Horizontal bold line indicates the mean. The average moving speed showed an initial decrease between E15 and P1, but showed no significant decline between P1 and P7 (F).

See Movies S1–3 for the corresponding time-lapse animations.

Because of the saltatory nature of their translocation, quantitative assessment of interneuron migration was performed by computing two distinct parameters: (1) we defined the moving speed as the rate of translocation (microns/hour) when an interneuron is moving i.e. not including the time when it pauses and (2) the pause time or pausing frequency indicates the proportion of time that the interneuron spent pausing during the time-lapse recording. Most investigations using time-lapse microscopy of neuronal migration report only the average speed as defined by distance traveled over a certain period of time (Cuzon et al., 2006). This measure can be ambiguous since changes in average speed can be caused by a change in either moving speed and/or pause time. Using this analysis we show that the moving speed of cortical interneurons only declines modestly E15, P1 and P7 (Figure 1F). However, we found that the increase in pausing frequency primarily accounts for the transition to a sedentary state. This percentage increases significantly between all ages examined (E15, P1, P7; Figure 1G; see Suppl. movie 3). These movies illustrate that interneurons terminate migration primarily by increasing pausing frequency and duration rather than by decreasing their moving speed.

GABA application induces pausing behavior in a subset of migrating interneurons expressing KCC2

Time-lapsing migrating interneurons in slices for prolonged periods of time (>12 hours) presents several challenges including the 3-dimensional (3D) nature of the slice and the high probability of cells moving out of the focal plane. Another strong limitation of using slices is linked to small molecule diffusion which is inherently limited by the thickness of the slice (250 microns). Since the spatial and temporal aspects of small molecule diffusion in slices is difficult to estimate, it would be difficult to know when interneurons were truly exposed to a given pharmacological treatment and at what concentration.

To circumvent these challenges we developed a novel two-dimensional (2D) migration assay. As interneurons will not migrate on coated substrates such as poly-L-lysine/laminin or fibronectin and their migration on collagen and matrigel is limited compared to slices (data not shown), we used dissociations of wild-type pyramidal neurons as a 2D substrate for explants of MGE isolated from EGFP-expressing isochronic littermates (Figure S1A). EGFP+ interneurons migrate from the MGE explant and across this wild-type dissociation just as they would normally in the cortex. We found that this assay enabled longer more accurate observations at higher temporal frequencies and provided immediate pharmacological access to the migrating interneurons.

We initially used this 2D assay to test if interneuron responsiveness to putative stop signals is due to cell-autonomous or cell non-autonomous changes. First, we performed isochronic and heterochronic co-cultures where the age of the cortical substrate for migration varies from 2div to 7div (see Figure S1). Our results demonstrate the after 2div, most interneurons migrate robustly (equivalent to E16.5; Figure S1A; Movie S4) but when both interneurons and their cortical substrate have matured for 7div (equivalent to P1), a significant proportion (approx. 55%) of MGE-derived interneurons have stopped migrating (Figure S1B; Movie S5). This result suggest that: (1) interneurons do terminate their migration in vitro according to a tempo that is comparable to in vivo and (2) the fine structure of the cortex is not necessary for the termination of interneuron migration as this structure is lost in the dissociation process. Furthermore, heterochronic experiments revealed that when `immature' interneurons (E14.5+2div) were plated on `mature' cortical substrate (E14.5+7div), the interneurons migrated at the same rate as in isochronic conditions (E14.5+2div MGE-derived interneurons migrating on 2div cortical substrate) (Figure S1C; Movie S6, quantified in Figure S1D–E). This strongly suggests that migratory responsiveness to putative extracellular stop signals requires some intrinsic maturation of these interneurons such as the up-regulation of an intracellular `gating' factor.

Interneurons possess both the capacity to release and to respond to GABA and yet, in spite of much evidence that GABA can modulate the migration of cortical and hippocampal neurons (Manent et al., 2005; Represa and Ben-Ari, 2005), the exact nature of GABA's effect on cortical interneuron migration, specifically through ionotropic receptors, remains unclear (Behar et al., 1996; Behar et al., 1998; Cuzon et al., 2006). Interneurons from EGFP-expressing MGE explants were allowed to migrate on a substrate of dissociated isochronic wild-type cortical neurons for 4.5div and were time-lapsed for 6 hours prior and 6 hours after the addition of vehicle only (Figure S2B) or GABA (20μM) (Figure S2C). This method allowed us to quantify the migration properties of individual interneurons before and after drug addition (Figure S2D–G). Interneurons that were not present for the entirety of the 12-hour imaging sessions were excluded from the analysis. Interestingly, the moving speed did not decline significantly after addition of GABA while their pausing frequency increased slightly but significantly (7.0±1.5%; data not shown). The increase in pausing frequency of interneurons receiving GABA was significantly higher (p=0.0183) than control receiving only media. This modest yet significant increase, specifically in pausing frequency, while maintaining stable moving speed upon GABA addition strikingly resembles our in vivo observations (Figure 1). In other words, a significant increase in pausing frequency was not only observed during the maturation of cortical interneurons (in vivo and in vitro), but was also observed upon exogenous addition of GABA.

Although a significant effect is observed at the population level, this preliminary analysis revealed a very significant degree of variability in the response of individual interneurons to GABA addition (compare cell 1 and cell 2 in Figure S2D–G). What could account for this significant degree of variability in the migratory response of interneurons to GABA?

KCC2 up-regulation is highly variable among migrating interneurons in vivo

There are at least two intrinsic factors that could contribute to this differential response of migrating interneurons towards GABA: (1) the heterogeneous expression of GABA receptors subunits and (2) the nature of the chloride (Cl⁻) gradient across the membrane dictating the direction of chloride ion flow upon opening of ionotropic GABA_A receptors. However, GABA_A receptors are the most diverse ligand-gated ion channel with 21 different subunits (Fritschy and Brunig, 2003). Single cell RT-PCR analysis on migrating interneurons revealed that the majority of cortical interneurons express combinations of α(3–5), β(1–3) and γ1 subunits as they are migrating (Cuzon et al., 2006). However, the functional relevance of this relative diversity of GABA_A receptor subunit composition for neuronal migration is unclear at this point.

Another critical modulator of ionotropic GABA function, the reversal of Chloride (Cl⁻–electrochemical potential in neurons, is predominantly controlled by the expression of one protein, the K⁺/Cl⁻ co-transporter, KCC2 (Rivera et al., 1999). Early in development, the expression of this Cl⁻ extruder is low relative to other Cl⁻ transporters such as NKCC1, resulting in a high intracellular concentration of Cl⁻. The driving force of this gradient causes a depolarizing efflux of chloride upon the binding of GABA to ionotropic GABA_A receptors. As a neuron matures, the up-regulation of KCC2 extrudes Cl⁻, which leads to a reversal of its Cl⁻ electrochemical driving force (Ben-Ari, 2002; Payne et al., 1996; Rivera et al., 1999) and therefore to a hyperpolarizing influx of Cl⁻ which underlies the inhibitory function of GABA in the mature CNS.

We tested if KCC2 expression is heterogeneous among migrating cortical interneurons by performing immunofluorescent staining in vitro and in vivo. Our in vitro analysis reveals drastically different levels of KCC2 expression among MGE-derived interneurons (Figure S3A–D). Also of note, pyramidal neurons, which composed the dissociated substrate, show very little KCC2 expression at the same age (Figure S3D). To confirm the physiological relevance of this variability of KCC2 expression observed in vitro, we performed immunofluorescent staining for KCC2 on Lhx6-EGFP cortical interneurons in vivo (Figure S3E–H). These results were consistent with the variability observed in vitro. Surprisingly, at P0 variability in KCC2 expression was observed in every layer of the neocortex including the CP (Figure S3E–H). In order, to compare the time course of KCC2 expression in interneurons and pyramidal neurons, we labeled the latter population using dorsal telencephalic electroporation at E14.5. This electroporation technique specifically labels radially migrating pyramidal neurons with no cross labeling of tangentially migrating interneurons (Figure S4A–C). After 4 days in culture, allowing both sub-populations of neurons to migrate, these slices were fixed and immunostained for KCC2, which revealed little to no KCC2 expression in pyramidal neurons in situ (Figure S3L–N) at an age when a large proportion of interneurons already express KCC2 (Figure S3I–K).

The 2D in vitro assay was adapted to more directly isolate and quantify KCC2 expression in these two distinct neuronal sub-populations. Pyramidal neurons progenitors were electroporated with monomeric red fluorescent protein (mRFP) at E14.5, dissociated and cultured in vitro with MGE-EGFP explants (Figure S4F–I). Upon fixation, immunostaining at 4div (Figure S3O–T) and 8div (Figure S3U–Z) confirmed that KCC2 is up-regulated in interneurons significantly earlier than in pyramidal cells (Figure S3AA). By 8div, KCC2 protein expression in interneurons was still twice that of pyramidal cells (p<0.0001; Figure S3AA). In addition, there is significantly less variability in the level of KCC2 expression among pyramidal neurons than among migrating interneurons (Figure S3AA).

We performed a time-course of KCC2 expression in the developing neocortex of Lhx6-EGFP in vivo to determine if interneurons up-regulated KCC2 before or after entering the cortex (Figure S5). Upon the initial entry of interneurons into the cortex at E14.5, little to no KCC2 expression was observed in any cortical layer. By E16.5 some staining is detected almost exclusively in Lhx6-EGFP+ interneurons in the CP. By P0, KCC2 expression is observed in cortical interneurons present in all layers, primarily associated with Lhx6-EGFP+ interneurons, with the most intense staining occurring in the CP. This demonstrates that migrating interneurons up-regulate KCC2 expression (1) non-synchronously approximately one week after reaching the cortex and (2) do so at least one week before pyramidal neurons.

KCC2 expression is correlated with responsiveness to GABA as a stop signal in migrating interneurons

To test the hypothesis that heterogeneous KCC2 expression underlies the differential responses to extracellular GABA application in migrating cortical interneurons, we combined the quantification of individual interneuron's migratory responses to GABA addition (Figure S2) with post-hoc immunofluorescent staining of KCC2 in the same interneurons (Figure 2A–J; Movie S7). Co-cultures were time-lapsed for 12 hours (6h before and 6h after addition of GABA or other drugs). Immediately after the acquisition of the last frame, these time-lapsed cultures were fixed and stained for KCC2, MAP2 and EGFP (Figure 2G–J). Each interneuron tracked during time-lapse and pharmacological treatment was re-located and imaged, enabling KCC2 protein expression to be correlated to the migratory response of individual interneurons following pharmacological treatment (Figure 2B–J).

Figure 2. GABA decreases motility of interneurons expressing high levels of KCC2 via GABA_A receptor activation.

(A) E14.5 wild-type neocortex was dissociated providing a 2-D substrate for the migration of MGE-derived EGFP-expressing interneurons.

(B–F) Time-series of migrating interneurons pseudo-colored at equally spaced time frames before extracellular application of 20μM GABA (B, white; 180 minutes after start of imaging i.e. 180 minutes before application of GABA; C, blue 360 minutes after start of imaging corresponding to time of GABA application) and after application of GABA (D, red, 540 minutes after start of imaging i.e. 180 minutes after GABA application; E, green 720 minutes after start of imaging i.e. 360 minutes after GABA application). See Movie S7 for the animated version of this panel.

(G–J) Interneurons cultures that were time-lapsed where fixed immediately after the last frame, and immunostained for MAP2, EGFP and KCC2. The same cells that were time-lapsed were then reimaged to assess their level of KCC2 expression.

(K) This technique allows the matching of individual interneuron responses to pharmacological treatments with KCC2 expression in individual interneurons. Interneuron responses to GABA application were binned into low and high KCC2 populations (green and red, respectively) which demonstrate a significant shift in pausing frequency for the high KCC2-expressing subpopulation of interneurons but not for the low KCC2-expressing interneurons.

(L–M) Box-whisker plots representation of interneuron responses to drug application binned according to KCC2 expression levels. Throughout the paper box plots in red represents putative responses to `hyperpolarizing' GABA<sub>A</sub> receptor activation (high KCC2 expressing interneurons) whereas green box plots indicate response to putative `depolarizing' GABA_A receptor activation. No significant changes in moving speed were detected in either sub-population following 20μM GABA (L) or GABA with concurrent application of GABA_A receptor antagonist bicuculline methiodide (BMI; 10μM; data not shown). (M) Low KCC2 interneurons showed no significant increase in pause time upon GABA application, while high KCC2 interneurons showed a significant (p=0.0004) increase in pausing after GABA application. Co-application of BMI with GABA abolished the effect of GABA alone (p=0.0308), leading to no significant difference between pre and post-drug pausing frequency. Conversely BMI co-application did cause any significant increase in pausing frequency for interneurons expressing low level of KCC2.

This analysis revealed two different sub-populations of interneurons with markedly different response to GABA application. We binned our migration data according to the KCC2 fluorescence of the individual interneurons. Interneurons with KCC2 fluorescence lower than the average value for that imaging session were binned into the low KCC2 group while those higher than the average were binned into the high KCC2 group. Interneurons expressing low levels of KCC2 showed no significant changes in moving speed (Figure 2L) or percentage of time moving following GABA application (green dots in Figure 2K and see also left box plot in Figure 2M). Strikingly, migrating interneurons expressing high levels of KCC2 responded to GABA addition by significantly increasing the time their pausing frequency (upward shift in box plot in Figure 2L, 2K and 2M). As observed when characterizing the termination of migration in situ, this GABA-induced increase in percentage of pausing time occurs without a significant decrease in moving speed (Figure 2L).

Is this response to GABA as a stop signal in migrating interneurons expressing KCC2 mediated by activation of GABA_A receptors? Co-application of 20μM GABA with 10μM bicuculline methiodide (BMI), a competitive GABA_A receptor antagonist, abolished the effect of GABA application on the pausing frequency in interneurons expressing high levels of KCC2 (Figure 2M). Interestingly, co-application of GABA+BMI significantly increased the pausing frequency in interneurons expressing low levels of KCC2 (Figure 2M). These results are also evident when viewing the averaged, frame-by-frame data for interneurons subdivided into high and low KCC2 expressing interneurons (Figure S5E). Our results show that (1) expression of KCC2 is correlated with the ability of interneurons to respond to GABA as a stop signal and (2) that in migrating cortical interneurons expressing low levels of KCC2, ambient GABA exerts a tonic, motogenic function stimulating inerneurons migration as shown by bicuculline-induced reduction of motility in low-KCC2 expressing cells. Therefore, KCC2 might act as a `gate control' transforming GABA from a motogenic to a stop signal in migrating interneurons.

KCC2 expression is required to reduce interneuron motility in response to GABA_A receptor activation

Our results raise three important questions: (1) is premature, forced KCC2 expression in migrating interneurons sufficient to induce responsiveness to GABA as a stop signal? (2) does the early, motogenic function mediated by GABA_A receptor activation require low KCC2 expression? and (3) is the transport activity of KCC2 required to confer the responsiveness to GABA as a stop signal?

To test the sufficiency of KCC2 expression in GABA-induced increase in pausing frequency, E14.5 wild-type MGEs were electroporated with human KCC2-IRES-EGFP expression plasmids (KCC2*; Figure 3A). This construct successfully up-regulates KCC2 expression compared to endogenous expression (Figure 3 B–D; 6–8 fold increase over endogenous expression, data not shown). Expression of exogenous KCC2 has been shown in the past to successfully extrude chloride and convert the effect of GABA from depolarizing to hyperpolarizing (Cancedda et al., 2007; Fiumelli et al., 2005). To knockdown endogenous KCC2 expression we designed a short hairpin (sh)RNA targeting mouse KCC2 mRNA (shKCC2) but not human KCC2* which allowed us to perform rescue experiments. We confirmed the knockdown of endogenous KCC2 in migrating interneurons by electroporating E14.5 MGE explants with either a control construct expressing a non-specific shRNA or shKCC2 (Figure 3A and 3E–K) and immunostaining for KCC2 after 7div i.e. when the vast majority of interneurons have normally up-regulated KCC2 (Figure S3). None of these treatments decreased interneuron viability has measured by expression of activated caspase-3 (Figure S7).

Figure 3. Manipulating KCC2 expression is sufficient to reproduce effects observed in binning interneurons high and low KCC2 expressing populations.

(A–D) KCC2 was over-expressed in E14.5 wild-type MGE explants by electroporation of EGFP-IRES human KCC2 construct (KCC2*). (B–D) Expression was verified by high immunofluorescence after 4.5div.

(E–J) Short hairpin RNAi targeted against mouse KCC2 (shKCC2) is very effective to knockdown endogenous KCC2 expression in interneurons. (E–G) Interneuron electroporated with a construct encoding EGFP at E14.5 was immunostained for KCC2 (red) at 7div. (H–J) Interneuron electroporated with a control EGFP and a construct encoding shKCC2 at E14.5 shows no KCC2 expression at 7div.

(K) Quantification shows a significant reduction in KCC2 with use of shKCC2. Measurements of background KCC2 immuno-reactivity show KCC2 is approaching undetectable background levels with introduction of shKCC2. Fluorescence was measured in 12 bits (value range of 0–4095).

(L) Similar to interneurons expressing low levels of KCC2, interneurons expressing shKCC2 (KCC2 knockdown) respond to BMI application (10 μM) by pausing significantly more frequently (p<0.0001). KCC2* rescues the knockdown by pausing them more with GABA_A receptor activation (10 μM muscimol; p<0.0001) and less with blocking the receptors (p<0.0001). Box plots are color-coded to represent putative depolarized (green), hyperpolarized (red) or mixed (yellow) interneuron population depending on the treatment. Light-grey shading shows the 25^th–75^th percentile range of appropriate control.

Using these tools we then determined the responsiveness of interneurons expressing experimentally-controlled levels of KCC2 to GABA_A receptor activation. Knocking down endogenous expression of KCC2 (shKCC2) or premature expression of KCC2 (shKCC2 + KCC2*) reproduced the responses found with endogenously low and high KCC2 expression, respectively (Figures 2 and 3). GABA_A receptor activation using muscimol (10μM), caused no significant change in the pausing frequency in interneurons expressing low levels of KCC2 (shKCC2) while interneurons expressing KCC2* paused significantly more than control isochronic interneurons expressing only EGFP (Figure 3L). Conversely, antagonizing GABA_A receptors with BMI (10μM) caused low KCC2-expressing shKCC2 interneurons to pause more frequently and decreased the pausing frequency of KCC2* expressing interneurons (Figure 3L).

KCC2 function on interneuron motility requires its co-transporter properties

KCC2 has recently been shown to mediate some of its function by interacting with the actin cytoskeleton though the ability of its intracellular C-terminus to bind the actin-binding protein 4.1B (Li et al., 2007). To determine if the effects of KCC2 expression are due to these cytoskeletal interactions or its K⁺/Cl⁻ co-transporter activity, we attempted to rescue the effects of knocking down KCC2 (shKCC2) using full length KCC2 or an N-terminal deletion (KCC2*-NTD). This deletion was previously shown to ablate the transport activity of KCC2 without affecting its cytoskeletal interactions (Li et al., 2007). A C-terminal deletion could not be used to eliminate cleanly the cytoskeletal interactions of KCC2 (Li et al., 2007) since it has also been shown to affect the transport activity of KCC2 under iso-osmotic conditions (Mercado et al., 2006). KCC2*-NTD expresses at levels identical to full length KCC2 (Figure 3B–D; Figure S8A–C). Unlike what is observed with KCC2* expression, the motility of interneurons expressing NTD-KCC2* were unaffected by muscimol application and showed a significant increase in pausing frequency upon BMI application (Figure S8D). The inability of NTD-KCC2* to phenocopy full-length KCC2* demonstrates that the transport properties of KCC2 are required for its effect on inducing responsiveness to GABA as a stop signal in migrating interneurons.

Forced hyperpolarization reduces interneuron motility

Our results so far (and some other presented below) strongly suggest that depolarizing signals such as GABA acting through GABA_A receptors prior to KCC2 up-regulation stimulate interneuron migration. Conversely, upon KCC2 up-regulation, GABA_A receptor activation, which should hyperpolarize interneurons caused a significant increase in pausing frequency. One possibility is that interneurons integrate ambient hyperpolarizing and depolarizing signals to determine when to stop migrating. To test this, we electroporated an inward rectifying potassium channel (K_ir 2.1) along with shKCC2 (in order to restrict our analysis to interneurons expressing low levels of KCC2). K_ir 2.1 has been previously shown to constitutively reduce membrane potential by inducing potassium efflux (Cancedda et al., 2007). Our results show that expression of K_ir 2.1 significantly increases pausing frequency in migrating interneurons compared to expression of shKCC2 alone or control EGFP expression (Figure 3L). These results indicate that forced hyperpolarization is sufficient to increase pausing frequency in migrating interneurons.

KCC2 expression is sufficient to terminate interneuron migration

To further reinforce the relevance of KCC2 expression with the termination of interneuron migration rather than simply decrease of their motility, KCC2 expression was measured at later time point in culture and correlated with migration properties. E14.5 EGFP+ MGE-derived interneurons were cultured for 7div on isochronic wild-type dissociated cortical substrate, time-lapsed for 6 hours, fixed, and immunostained for KCC2 (Figure 4; Movie S8). At this time point, we observed a perfect correlation between KCC2 expression and the absence of cell body translocation in cortical interneurons strongly suggesting that KCC2 up-regulation is tightly coupled with the termination of interneuron migration and not only with decreased motility.

Figure 4. KCC2 is strongly correlated with the termination of interneuron migration.

(A) E14.5 EGFP-MGE explants were placed on E14.5 wild-type dissociations and cultured for 7div. These interneurons were then time-lapsed for 6 hours (B–D), fixed and immunostained for KCC2 (F–H). (E) Box plots show 10^th, 25^th, 50^th, 75^th and 90^th percentiles KCC2 fluorescence in moving versus sedentary interneurons. Light-grey shading shows the 25^th–75^th percentile range of KCC2 expression in the motile population of interneurons. Binning interneurons into moving and non-moving populations reveals significantly higher KCC2 expression in the sedentary population (p<0.0001). Note yellow `co-labeling' in time-lapse representation (D) corresponds with yellow co-labeling of KCC2 with interneurons (H). Suppl. Movie S8 provides an example of the raw time-lapse data.

If our model is correct, precocious KCC2 expression should be sufficient to induce premature termination of interneuron migration. E14.5 wild-type slices were electroporated specifically in the MGE with control (EGFP only) or KCC2-IRES-EGFP constructs at E14.5 and cultured for 4 days ex vivo (Figure 5A). Control electroporations show robust migration from the MGE into the cortex (Figure 5B), while KCC2 over-expressing interneurons display limited migration mostly restricted to the ventral telencephalon (Figure 5C). The quantification demonstrates a near two-fold reduction in the percentage of MGE-derived interneurons reaching the dorsal telencephalon in KCC2-expressing compared to control slices (29.0±5.3% vs. 54.4±5.0%, respectively; Figure 5D). These results demonstrate that premature expression of KCC2 is sufficient to reduce interneuron motility to the cortex in the presence of endogenous levels of GABA.

Figure 5. KCC2 expression is necessary and sufficient for the inhibition of cortical interneuron migration.

(A) Wild-type embryonic cortical slices (300 microns thick) were injected in the MGE with either control EGFP constructs or EGFP-IRES-KCC2* expressing constructs and subsequently electroporated. Inset shows early EGFP expression restricted to the MGE after 1div.

(B–C) While EGFP controls (B) show robust migration of interneurons from the striatum into the dorsal telencephalon, precocious expression of KCC2* in slices reduces migration to the cortex by approximately 2-fold.

(D) Significant decrease (p<0.0001) in the percentage of MGE-derived interneurons migrating into the cortex in KCC2-expressing interneurons compared to control.

(E–G) MGEs electroporated with either (E) control plasmid, (F) shKCC2 or (G) shKCC2+KCC2* were explanted on wild-type cortical dissociated cultures for 7div and time-lapsed for 6 hours. Pictures show initial frame (t=0) in red and frame captured 3 hours later in green. Yellow labeled cells indicates sedentary interneurons.

(H) Box plots show change in pausing frequency of cortical interneuron populations expressing the indicated constructs, before and after the indicated drug treatment Light-grey shading shows middle 50^th percent range of appropriate control population. Quantification shows a significant decrease in the pausing frequency of interneurons expressing shKCC2 (p<0.0001) and a significant rescue with KCC2* (p<0.0001). See also Movies S9–11 for representative examples.

To test if KCC2 expression is required to stop interneuron migration, E14.5 MGE explants were electroporated with control shRNA, shKCC2 or shKCC2 rescued with human KCC2*. These interneurons were time-lapsed at 7div when most interneurons express KCC2 and have stopped migrating in vitro and in vivo (see Figure 1 and Figures S1 and S3). By 7div, only few interneurons are still migrating in control electroporations (30.5±3%; Figure 5E and H; Movie S9) corresponding approximately to the percentage of interneurons migrating in vivo at P1 (E14.5+7days; see Figure 1). Importantly, knocking-down KCC2 almost doubles the proportion of migrating cortical neurons to 55.5±3.5% (Figure 5F and H; Movie S10). KCC2* expression successfully rescued the termination of interneuron migration, leaving 93.2±2.1% in a sedentary state (Figure 5G and H; Supplementary movie S11). Therefore expression of KCC2 is necessary and sufficient for proper termination of interneuron migration.

KCC2 expression in migrating interneurons negatively regulates the frequency of intracellular calcium transients

Calcium has often been tied to the process of cellular motility (Zheng and Poo, 2007). The frequency of intracellular calcium dynamics was shown to control positively the rate of migration of cerebellar granule cells (Komuro and Kumada, 2005), and to negatively regulate the rate of extension of axonal growth cones (Gomez et al., 1995; Gomez and Spitzer, 1999; Kater and Mills, 1991)). We hypothesized that regulating the outcome of GABA_A receptor activation from depolarizing to hyperpolarizing could alter the intracellular calcium dynamics of migrating interneurons through differential activation of VSCCs, coupled or not to release of calcium from internal stores. Calcium responses have been observed in migrating cortical interneurons, but only upon pharmacological manipulations such as the application of muscimol, NMDA and kainate (Soria and Valdeolmillos, 2002). Therefore, at this point, the spatial and temporal features of spontaneous intracellular calcium dynamics in migrating cortical interneurons have not been reported.

In order to visualize intracellular calcium dynamics in migrating interneurons, we electroporated mRFP in E14.5 MGE progenitors and cultured these explants on isochronic dissociated cortical neurons as a 2D substrate for 4div. These co-cultures were then acutely loaded with the cell-permeant calcium-sensitive dye, Oregon Green BAPTA-AM and spontaneous calcium transients could clearly be monitored in these migrating interneurons (Figure 6A–E; Movie S12). The occurrence of these calcium transients was not observed in all migrating interneurons, we nevertheless confirmed the existence of naturally-occuring calcium transients during the course of interneuron migration. In order, to determine if KCC2 expression was directly responsible for the diverse range of calcium dynamics observed in migrating interneurons, we compared the calcium dynamics of migrating interneurons where we knocked down KCC2 expression (shKCC2) to interneurons where expressing KCC2*. Movie S13 shows calcium dynamics of KCC2 knockdown interneurons in four independent fields. Most interneurons (9 of 16 cells) with knocked down KCC2 expression display several calcium transients per hour (see Figure 6F). This is in stark contrast to the absence (0 of 14 cells) of similar transients in isochronic migrating interneurons expressing KCC2* (E14.5+4div; see Movie S14; see also Figure 6G).

Figure 6. Calcium signals in tangentially migrating interneurons are reduced with KCC2 up-regulation.

(A–D) Migrating cortical interneurons show spontaneous calcium transients. (A) An mRFP-electroporated E14.5 MGE interneuron was loaded with Oregon green BAPTA-AM and time-lapsed. Pseudo-colored images show calcium signal at low (B) and high (C) periods of activity in the outlined migrating interneuron. (D) Pseudo-colored strips show a kymograph obtained by an orthogonal re-section through the cell body during the course of the time-lapse. (E) Calcium signals observed in control conditions shows wide range of dynamics.

(F–J) KCC2 expression reduces the calcium transients in the 0.003-0.03Hz range. Calcium signals from Oregon Green BAPTA-loaded interneurons electroporated with either mRFP and a plasmid encoding shKCC2 (F), mRFP-IRES-KCC2 (G) or shKCC2+10μM BMI (H) are shown. Neither KCC2 over-expressing nor shKCC2+BMI interneurons show these types of calcium transients. See Movies S12–15 for corresponding time-lapse. (I) A spectral analysis done on individual interneurons and averaged for each group reveals a significant decrease in calcium signaling in the 0.003–0.03Hz frequency range upon either KCC2 over-expression or GABA_A receptor blockage. (J) The relative power spectral densities were binned into 0.003–0.03Hz and >0.03Hz categories.

(K–M) Time-lapsed Lhx6-EGFP interneurons at E15 are shown with initial frame (t=0) in blue, 3 hours frame in green, and 6 hours frame in red. Non-moving cells appear white. Note that chelation of intracellular calcium with 25μM BAPTA-AM significantly increases the number of stationary interneurons (L) relative to control (K). (M) Quantification shows a decrease (Chi-Square Analysis p<0.0001) in the proportion of migrating interneurons following intracellular calcium chelation. See Movies S1 and S16 for corresponding time-lapse.

To test if naturally-occurring calcium transients were due to the depolarizing effects of GABA_A receptors activation in migrating interneurons expressing low levels of KCC2, we knocked down KCC2 and performed calcium imaging in the presence of BMI (10μM). As observed for KCC2 over-expressing interneurons, blocking the depolarizing effect of GABA eliminated the spontaneous calcium transients in migrating interneurons (Figure 6H and Movie S15).

The quantification of the frequency of calcium transient can be rather arbitrary since it depends heavily on threshold definition and signal/noise ratio. To quantify the frequency of these calcium transients in an unbiased way, we performed Relative Power Spectral Density (RPSD) analysis on the recorded calcium traces (see Suppl. Methods for details). This analysis separates time from the frequency component of the raw calcium traces, providing a quantitative representation of what specific frequencies are represented in the recorded calcium signal. This analysis revealed a significantly higher occurrence of `slow' (0.003–0.03Hz) intracellular calcium signal variations in shKCC2 interneurons compared to KCC2*-expressing interneurons (Figure 6I). When these values were binned into 0.003–0.03 Hz versus >0.03mHz, RPSD of shKCC2 interneurons was significantly higher than in KCC2 over-expressing cells (Figure 6J). Interneurons electroporated with shRNA against KCC2 also showed a significantly higher RPSD at 0.003–0.03Hz frequencies compared to >0.03 Hz frequencies, while KCC2 over-expressing interneurons showed no statistical difference in RPSD binned for 0.003–0.03mHz or >0.03mHz categories. Application of bicuculine eliminated the calcium transients observed in KCC2 knockdown interneurons (Figure 6H) demonstrating that in migrating interneurons expressing low or no KCC2, these spontaneous calcium transients are mostly mediated by GABA_A receptor-mediated depolarization.

If there is a causal relationship between the frequency of intracellular calcium transients and the termination of interneuron migration upon KCC2 up-regulation, then prevention of these calcium transients could be sufficient to prematurely stop interneurons migration as previously shown for cerebellar granule cells (Komuro and Kumada, 2005). To chelate intracellular calcium in situ, we incubated acute telencephalic slices isolated from E14.5 Lhx6-EGFP transgenic embryos with 25μM BAPTA-AM for 2 hours before imaging. Intracellular calcium chelation is sufficient to inhibit the migration in cortical interneurons (visualized by white pseudo-color in Figure 6H and 6L, see Movie S1 and S16). The quantification of these movies shows that most (84.9±2.8%) interneurons are migrating in control conditions at E14.5+1div while only 51.4±0.3% of interneurons migrate when intracellular calcium is sequestered (Figure 6M). This result suggests that intracellular calcium transients observed in interneurons are required for their migration and that KCC2 up-regulation is sufficient to ablate these intracellular calcium dynamics leading to termination of interneuron migration.

The effect of GABA-mediated calcium transients on interneuron motility are mediated by activation of VSCCs

We hypothesized that depolarization mediated by GABA_A receptor activation in interneurons expressing no or low KCC2 might lead to opening of VSCCs and therefore to calcium entry. After confirming the expression of L-type VSCC subunits Ca_V1.2 and Ca_V1.3 in migrating cortical interneurons at this developmental stage (Figure S9), we tested this hypothesis by utilizing our 2D interneuron migration assay at E14.5+4.5div in order to test the acute effect of blocking different types of VSCC's in interneuron migration. Antagonists for N-type and L-type VSCC (5μM GVIA ω-conotoxin and 10μM nifedipine, respectively) were tested for effects on pausing frequency in migrating interneurons. While both blockers show a significant effect, the effect of blocking L-type calcium channels using nifedipine was very strong and significantly more potent than blocking N-type VSCCs (Figure 7A–C) leading to interneurons spending on average 85–90% of their time not moving. The acute blockade of L-type VSCCs caused a complete and immediate termination of migration of approximately 40% of interneurons observed (22 of 60) and a significant increase in pausing frequency in the remaining interneurons (Figure 7A–C). Interestingly, applying bayK8644 (an agonist of VSCCs, which increases the duration of `open' state of L-type Ca²⁺ channels) was sufficient to decrease the number of sedentary interneurons at 7div from 70% to 40% (Figure S10) and extend the window of cortical interneuron migration.

Figure 7. Reduced activation of VSCCs by direct pharmacological antagonism or indirectly by blocking glutamate signaling further increases pause times in interneurons.

(A–C) E14.5 interneurons electroporated with shKCC2 and plated on wild-type dissociated cortex increased in pausing with addition of (A) the N-type VSCC antagonist ω-conotoxin GVIA (5μM) at 4.5div (p<0.0001) and an even greater effect with the addition of (B) L-type VSCC antagonist, nifedipine (10μM; p<0.0001). (C) Box plots showing percentage of time pausing for migrating cortical interneurons in conditions shown in A–B. Light-grey shading indicates the 25^th–75^th range of percentiles of control values. See S17 and S18 for corresponding movies. (D–F) KCC2 knockdown interneurons show little effect on pausing when eliminating depolarization due to glutamate signaling (100μM APV and 10μM NBQX) as long as depolarization due to muscimol (10μM) are still present (D,F). Box plots are color coded to indicate whether drug treatment and KCC2 modification would be depolarizing (green) or hyperpolarizing (red) with respect to appropriate control. Note that when KCC2 is up-regulated and muscimol is hyperpolarizing, the effect of eliminating AMPA/NMDA-mediated depolarization is highly significant (E,F). See Suppl. Movies 19 and 20 for corresponding time-lapse.

GABA- or Glutamate-mediated depolarization stimulate the migration of cortical interneurons

Although Figure 5 showed that KCC2 over-expression significantly decreases the motility of migrating interneurons at 7div, Figure 3 indicated that KCC2 expression coupled to GABA application does not cause a complete termination of migration. If GABA-mediated hyperpolarization is responsible for terminating migration, why is the effect of KCC2 up-regulation only partial? These results, combined with the more dramatic effect of antagonizing L-type VSCCs, led us to hypothesize that other ambient depolarizing cues might contribute to the stimulation of interneuron migration even in the absence of GABA-mediated depolarization. An obvious candidate was glutamate-mediated activation of ionotropic glutamate AMPA/NMDA receptors, which are expressed by migrating interneurons (Metin et al., 2000; Soria and Valdeolmillos, 2002).

To test this, we performed experiments where GABA_A receptors are activated by muscimol in interneurons expressing KCC2 at E14.5+4.5div, only this time in the presence of AMPA and NMDA antagonists (NBQX 10μM and APV 100μM, respectively; Figure 7 D–F). Blocking AMPA/NMDA receptors induced a significant increase (+30%) in the percentage of interneurons terminating migration. This striking result demonstrates that migrating interneurons are under tonic AMPA/NMDA receptor-mediated depolarization, which significantly stimulates their motility. Finally, blocking AMPA/NMDA receptors in shKCC2-expressing interneurons decreases motility (increased pausing) compared to control treatment but, most interestingly, activation of GABA_A receptors in the same shKCC2-expressing interneurons in the presence of NBQX/APV strongly stimulates their motility (Figure 7D–F and Figure S11C). Taken together this demonstrates that in migrating interneurons expressing low or no KCC2, depolarization through both GABA_A or AMPA/NMDA receptors play a cumulative motogenic effect.

DISCUSSION

Our results demonstrates that (1) depolarization through activation of GABA_A receptors (when KCC2 is low) and AMPA/NMDA receptors exerts a complementary motogenic activity on migrating interneurons and (2) that up-regulation of KCC2 is necessary and sufficient to reduce interneuron migration by rendering GABA_A receptor activation hyperpolarizing (Figure 8). Interestingly, these effects are largely mediated through the ability of GABA_A receptor to trigger activation of L-type VSCCs leading to intracellular calcium transients. Our results suggest for the first time that, once GABA_A receptor activation becomes hyperpolarizing in migrating cortical interneurons, these cells are able to sense and integrate the ambient extracellular levels of GABA and glutamate as a way to determine when to stop migrating.

Figure 8. Model – KCC2 expression facilitates the termination of cortical interneuron migration.

(A) Initial migration state. Green `halo' represents motogenic effect of glutamate and GABA. Blue shade of interneurons and pyramidal neurons represents degree of KCC2 expression. Membrane magnification illustrates transporter and channel composition of interneurons throughout the figure. Initially, ambient GABA and glutamate stimulate the migration of interneurons by inducing membrane depolarization and calcium influx through activation of VSCCs. (B) KCC2 up-regulation act as a switch rendering ambient GABA hyperpolarizing, reducing Ca²⁺ influx through VSCCs which significantly reduces interneuron motility. (C) Developmental decrease of ambient glutamate, mostly due to astrocytic re-uptake and confinement to synaptic release, further reduces Ca²⁺ influx and contributes to the termination of interneuron migration by reducing global AMPA/NMDA receptor activation on interneurons.

GABA and the migration of cortical interneurons

GABA has been previously proposed to modulate neuronal migration, although the direction of this modulation often varies depending on experimental approaches (reviewed by (Owens and Kriegstein, 2002)). The opening of ionotropic GABA_A receptors enhances both chemotaxis and chemokinesis in cortical dissociated neurons, although GABA_A receptor activation largely hinders migration of GAD+ cells in the CP (Behar et al., 1996; Behar et al., 1998). In spite of GABA's possible hampering effect on migration, other studies have revealed a positive effect of GABA_A receptor activation on radial (Manent et al., 2005) and tangential migration (Cuzon et al., 2006). By monitoring the response of individual interneurons to GABA and correlating it to KCC2 expression, we uncovered the fact that GABA acts as a motogen for immature, low-KCC2 expressing interneurons but acts as a stop signal for interneurons expressing high level of KCC2 once they reach the cortex. This mechanism might ensure that the paracrine level of ambient GABA (measured as approximately 0.5 μM in vivo) (Cuzon, 2006) does not prematurely stop interneuron migration before they reach the cortex.

Up-regulation of KCC2 in development

Interestingly, premature up-regulation of KCC2 in radially migrating pyramidal neurons does not affect their migration but significantly dampens their level of dendritic growth and branching (Cancedda et al., 2007). These results suggest that in pyramidal neurons, ambient level of GABA is not sufficient to act as a stop signal even if KCC2 is prematurely up-regulated. In fact, this is compatible with our results suggesting that pyramidal neurons naturally upregulate KCC2 4–8 days after interneurons at a time when they have already stopped migrating in vivo.

Our results are the first to document a strong difference in the timing of KCC2 expression between interneurons and pyramidal neurons and this is compatible with a recent study showing an enrichment of KCC2 mRNA using microarray analysis of embryonic interneurons (Batista-Brito et al., 2008). Interestingly, outside the developing cortex, large differences have been found between brain regions during development (Belenky et al., 2008; Belmonte et al., 2004; Gilbert et al., 2007). Furthermore, the chloride gradient has even been found to vary greatly between adjacent neurons within the same structure (Gilbert et al., 2007).

Determining what factors regulate KCC2 expression in cortical interneurons would greatly improve our understanding of the mechanisms regulating the termination of interneuron migration. Several experimental manipulations such as the induction of long-term potentiation, magnesium removal, neuronal stress, seizure kindling and subplate ablation, have been shown to regulate KCC2 expression (Galanopoulou, 2007; Kanold and Shatz, 2006; Rivera et al., 2004; Wake et al., 2007; Wang et al., 2006). Several conflicting lines of evidence have suggested that BDNF (Rivera et al., 2002; Rivera et al., 2004) or GABA itself (Ganguly et al., 2001; Kriegstein and Owens, 2001; Leitch et al., 2005; Ludwig et al., 2003; Rivera et al., 2004; Titz et al., 2003; Toyoda et al., 2003) can regulate KCC2 expression at the transcriptional or translational levels. This may indicate multiple regulatory pathways and/or cell-type and region-specific mechanisms underlying KCC2 transcription and translation. In addition, KCC2 activity can also be modulated by localization, oligomerization and phosphorylation (Adragna et al., 2004; Blaesse et al., 2006; Lee et al., 2007; Wake et al., 2007). Future experiments will determine what factors control KCC2 expression/function in migrating interneurons once they reach the cortex.

Integration of GABA and glutamate signaling by migrating interneurons

Our results indicate that depolarization through ionotropic glutamate receptor activation stimulate interneuron migration as potently as GABA_A receptor activation in interneurons expressing low or no KCC2. Therefore our model (see Figure 8) suggests that interneurons integrate ambient level of GABA and glutamate to determine when to stop migrating. Our results support a model whereby in order to completely stop migrating, GABA_A-receptor mediated hyperpolarization following KCC2 up-regulation must be accompanied by a reduction of tonic AMPA/NMDA receptor activation. This decline might be at least partially due to the up-regulation of glutamate transporters by astrocytes or neurons during this period during early postnatal development (Cavelier et al., 2005; Danbolt, 2001).

Glutamate has been previously shown to evoke responses in migrating interneurons (Metin et al., 2000; Poluch et al., 2001; Poluch et al., 2003). Several studies have also shown that migrating cortical interneurons exhibit transient increase in intracellular calcium upon application of AMPA, kainate and, in some cases, NMDA receptor agonists (Metin et al., 2000; Soria and Valdeolmillos, 2002), although the function of these calcium transients remained unclear. Furthermore, NMDA receptor activation has been shown to stimulate neuronal migration both by controlling calcium dynamics (Komuro and Rakic, 1993). Our results show that upon KCC2 upregulation but before synaptogenesis during the first week of postnatal development, cortical interneurons sense and integrate ambient levels of GABA and glutamate, which might reflect the local density of pyramidal neurons versus interneurons (Figure 8). Therefore, we propose that in migrating interneurons expressing KCC2, the balance between hyperpolarizing ambient GABA and depolarizing ambient glutamate is integrated by the frequency of calcium transients evoked by activation of VSCCs (Figure 8).

The ability of immature neurons during migration or during axon pathfinding to use changes of membrane potential as a meaningful signaling event might be a general property. An interesting study has recently shown that axon guidance cues elicit changes in membrane potential at the level of individual axonal growth cones and that conversely, modifying membrane potential changes growth cone response properties to these axon guidance cues (Nishiyama et al., 2008). Finally, changes in the frequency of calcium spikes in immature Xenopus spinal cord neurons alters their neurotransmitter expression suggesting that specific patterns of calcium dynamics plays a role in neurotransmitter homeostasis (Borodinsky et al., 2004). Our results add a new dimension to these findings by showing that migrating interneurons can sense and integrate ambient levels of local hyperpolarization and depolarization as a way to determine when to stop migrating through the ability of changes in membrane potential to activate VSCCs.

Consequence of altered interneuron migration on development

The long-term consequences of altering the window of interneuron migration are unknown at the present time but could be quite drastic for cortical circuit assembly. GABA signaling plays a critical role in determining the critical period for neuronal plasticity in the visual cortex (Hensch, 2005). Future experiments should test if interference with the termination of interneuron migration leads to the precocious development of GABAergic synapses and may therefore prevent the optimal timing of cortical plasticity.

Improper distribution of interneurons may result in cell death in areas where local densities of interneurons are too high (Fuerst et al., 2008) and epileptic forms of activity where interneurons density is too low (Cobos et al., 2005; Li et al., 2008). Not surprisingly, several neurodevelopmental pathologies such as ASD and schizophrenia have been associated with alterations in interneuron number, placement or maturation. Timothy Syndrome, a disorder encompassing multiple dysfunctions including autistic phenotypes, can be caused by a mutation in Ca_V1.2 (CACNA1C), which increases the open times of VSCCs (Splawski et al., 2004). One would expect this mutation to have the same effect on extending interneuron migration as our bayK8644 treatment. A similar investigation uncovered a correlation between autism spectrum disorders and mutations in a T-type calcium channel, Ca_V3.2 (CACNA1H) (Splawski et al., 2006). Future investigations will determine if deficiency in voltage sensitive calcium signaling and/or interference with GABA or glutamate signaling during migration leads to improper placement of cortical interneurons.

Interestingly, during early postnatal development, accute stress can alter the release of GABAergic neurosteroid allopregnanolone, which act as a strong endogenous agonist of GABA_A receptors, and this has been shown to alter the final positioning of GABAergic interneurons in the rat prefrontal cortex (Grobin et al., 2003). We hypothesize that genetic or environmental perturbation of the signaling mechanisms described in the present study might delay the window of cortical interneuron migration and might impact on the onset of inhibitory synaptogenesis in the developing cortex. Future experiments will address how interference with the termination of interneuron migration might affect the maturation and function of the cortical microcircuits.

Experimental Procedures

Animals

All procedures involving animals were approved by the IACUC at UNC-Chapel Hill and were in accordance to NIH guidelines. Transgenic mice expressing EGFP ubiquitously under the control of CMV-enhancer and chicken β-actin promoter (Okabe et al., 1997) were obtained from Jackson Laboratories. Heterozygous Lhx6-EGFP BAC transgenic mice were kindly provided by Drs M-.B. Hatten and N. Heintz (Rockefeller Univ.- GENSAT Consortium; (Heintz, 2004)), were bred on a Balb/C background for at least ten generations and maintained in a 12/12 hours light:dark cycle. Time pregnant mice were obtained by breeding overnight and the following morning is considered as E0.5.

Immunostaining for slices and dissociations

After fixation with PFA 4%, brains, cells or slices were rinsed in PBS 3×15 min, incubated overnight in blocking solution (1 g BSA (A7906, Sigma-Aldrich, St. Louis MO)/10 mL PBS; 0.3% Triton X-100 (X-100, Sigma-Aldrich, St. Louis MO) at 4°C on orbital shaker. The following primary antibodies were applied overnight in blocking solution at 4°C on shaker: chicken polyclonal anti-GFP (A10262; Invitrogen - Molecular Probes, Eugene OR); rabbit polyclonal anti-KCC2 (07–432; Upstate, Temecula CA) against residues 932–1043 of rat KCC2; mouse monoclonal anti-MAP2 (Clone HM-2; Sigma-Aldrich, St. Louis MO); rabbit polyclonal anti-alpha 1C, Ca_V1.2 (C1603; Sigma-Aldrich, St. Louis MO); rabbit polyclonal anti-alpha 1D, Ca_V1.3 (C1728, Sigma-Aldrich, St. Louis MO). Following extensive rinsing in PBS, the following secondary antibodies were applied overnight in blocking solution plus 5% goat serum at 1:1000 at 4°C on shaker: 488 goat polyclonal anti-chicken, A11039; 546 goat polyclonal anti-rabbit, A11035; 647 goat polyclonal anti-rabbit, A21245; 647 goat polyclonal anti-mouse, A21236; Invitrogen - Molecular Probes, Eugene OR). After final wash (5×15min in PBS), slices, sections or cells on coverslips were mounted with GelMount (Biomeda Corp, Foster City CA).

Pharmacology

20μM GABA, made from 20mM stock in ddH₂O (A-5835; Sigma-Aldrich, St. Louis MO) was added to cultures during time-lapse sessions. Other drugs used were: Bicuculline Methoiodide (BMI; 2503; Tocris, Ellisville, MO; 10μM) prepared from 10mM stock in ddH2O. BAPTA-AM (B1205; Invitrogen - Molecular Probes, Eugene OR; 25μM) prepared from 10mM stock in desiccated DMSO (D2650; Sigma-Aldrich, St. Louis MO) added to media underneath slice insert and allowed to load for 2 hours before imaging. Muscimol (M1523; Sigma-Aldrich, St. Louis MO; 10μM) was prepared from 10mM ddH₂O stock solution. BayK8644 (B112, Sigma-Aldrich, St. Louis MO; 10μM) was prepared from 50mM DMSO solution. Nifedipine (N7634; Sigma-Aldrich, St. Louis MO; 10μM) was prepared from 5mM DMSO solution. GVIA ω-conotoxin (C9915; Sigma-Aldrich, St. Louis MO; 5μM) was prepared from 1mM solution in double distilled (dd)H₂O. DL-2-amino-5-phosphopentanoic acid (AVP; A5282; Sigma-Aldrich, St. Louis MO; 100μM) was prepared from 10mM solution in media. NBQX (N171; Sigma-Aldrich, St. Louis MO; 10μM) was prepared from 10mM DMSO solution.

Calcium loading

A stock solution of calcium indicator was made by adding 10 μL DMSO (D2650; Sigma-Aldrich, St. Louis MO) to 50μg of Oregon Green 488 BAPTA-1, AM (OGB-1; O6807; Invitrogen - Molecular Probes, Eugene OR). After vortexing for 1 min 5×2μL aliquots were frozen at −80°C. Warm 9.12μL Pluronic F-127 in 20% DMSO (P3000MP; Invitrogen - Molecular Probes, Eugene OR) was then added to one aliquot of stock solution. After vortexing for 1 min, a 5μM working solution was made by adding 5μL of pluronic - stock solution to 715μL HBSS and vortexing for another minute. See Suppl. Material for details on calcium imaging and quantification.

Ex Vivo Electroporation and Organotypic slice culture

As previously described in Hand et al. (2005). See Supplementary Material for detailed procedure.

Confocal Microscopy

Confocal microscopy on fixed tissue was done as described previously (Hand et al., 2005) using a Leica TCS SL inverted confocal miscroscope equipped with a Marzhauser X-Y motorized stage and a PeCon CO₂- and Temperature-controlled stage incubation chamber. Time-lapse microscopy was done as described previously (Hand et al., 2005) with an imaging frequency of a stack captured every 10 minutes for migration studies. These movies are played back at a rate of 7 frames per second (4200× real time). Calcium imaging movies were obtained by capturing a frame every 4 seconds (see Supplementary Material for details).