Premature expression of KCC2 in embryonic mice perturbs neural development by an ion transport-independent mechanism


Zachi Horn, 1Neonatal Research Unit, as above.


During neuronal maturation, the neuron-specific K–Cl co-transporter KCC2 lowers the intracellular chloride and thereby renders GABAergic transmission hyperpolarizing. Independently of its role as a co-transporter, KCC2 plays a crucial role in the maturation of dendritic spines, most probably via an interaction with the cytoskeleton-associated protein 4.1N. In this study, we show that neural-specific overexpression of KCC2 impairs the development of the neural tube- and neural crest-related structures in mouse embryos. At early stages (E9.5–11.5), the transgenic embryos had a thinner neural tube and abnormal body curvature. They displayed a reduced neuronal differentiation and altered neural crest cell pattern. At later stages (E11.5–15.5), the transgenic embryos had smaller brain structures and a distinctive cleft palate. Similar results were obtained using overexpression of a transport-inactive N-terminal-deleted variant of KCC2, implying that the effects were not dependent on KCC2′s role as a K–Cl co-transporter. Interestingly, the neural tube of transgenic embryos had an aberrant cytoplasmic distribution of 4.1N and actin. This was corroborated in a neural stem cell line with ectopic expression of KCC2. Embryo phenotype and cell morphology were unaffected by a mutated variant of KCC2 which is unable to bind 4.1N. These results point to a role of KCC2 in neuronal differentiation and migration during early development mediated by its direct structural interactions with the neuronal cytoskeleton.


KCC2 is a neuron-specific isoform of the K–Cl co-transporters. Its developmental upregulation is temporally associated with maturation of postsynaptic GABAergic inhibition in central neurons (Rivera et al., 1999; reviewed in Blaesse et al., 2009). Functional expression of KCC2 during neuronal development leads to a decrease in the intraneuronal Cl concentration and, consequently, to a hyperpolarizing shift in the reversal potential of GABAA receptor-mediated currents (EGABA) from depolarizing values that are characteristic for immature neurons.

During embryonic development, GABA has a neurotrophic role through its depolarizing and sometimes excitatory actions (Owens & Kriegstein, 2002; Wang & Kriegstein, 2009). KCC2 expression precedes the functional EGABA shift in several neuronal systems such as the spinal cord (Li et al., 2002; Stein et al., 2004; Delpy et al., 2008), the auditory brainstem (Balakrishnan et al., 2003; Blaesse et al., 2006) and hippocampal cultures (Khirug et al., 2005).

Ectopic expression of KCC2 in immature neurons shifts EGABA to more negative levels (Chudotvorova et al., 2005; Lee et al., 2005). Interestingly, a premature shift in the GABA response by ectopic KCC2 expression has been reported to impair the morphological maturation of cortical neurons in rats (Cancedda et al., 2007). Furthermore, overexpression of KCC2 from the onset of development has been shown to perturb neuronal differentiation and axonal growth in zebrafish (Reynolds et al., 2008). These studies demonstrate the importance of a spatiotemporal regulation of the inception of KCC2-mediated Cl transport activity. In addition, it has been demonstrated that KCC2 plays a pivotal morphogenic role in dendritic spine formation and this structural function does not require the transport activity of KCC2 (Li et al., 2007; for a similar ion transport-independent structural role of the Na–K–2Cl co-transporter 1 see Walters et al., 2009). Whether KCC2 has a structural role during early embryonic development has not been elucidated.

Here, we report that KCC2 alters neuronal differentiation and motility through an ion transport-independent mechanism. We employed a tissue-specific promoter to overexpress three different KCC2 constructs in neuronal progenitors of transgenic mouse embryos and a neural stem cell line. The embryos and the cell cultures were severely affected by two of these constructs, coding for a transport-active and a transport-inactive KCC2 protein, which were both found to interact with the cytoskeleton-associated protein 4.1N. This was in contrast to a point-mutated variant of KCC2 that did not interact with 4.1N. Our findings suggest that KCC2 may regulate early neuronal development through structural interactions with the actin cytoskeleton.

Materials and methods

DNA constructs

The human nestin 2nd intron (hnestin) 1852 vector was generated from the hnestin 1852/LacZ plasmid (Lothian & Lendahl, 1997). A thymidine kinase (tk) promoter sequence was inserted downstream of the hnestin sequence. A 3348-bp fragment spanning the open reading frame of the mouse KCC2 sequence and flanked by XhoI and HindIII sites was generated by PCR from a cDNA clone purchased from RZPD (; I.M.A.G.E. Consortium [LLNL] cDNA CloneID 6838880). The upstream primer was 5′-TAA CTC GAGATG CTC AAC AAC CTG ACG and the downstream primer was 5′-GAC AAG CTT TCA GGA GTA GAT GGT GAT G (the XhoI and HindIII sites are, respectively, the first and second underlined sections and the start codon is indicated in italics). The KCC2-N-terminal deletion (ΔNTD; deletion of amino acids 1-100) sequence in a internal ribosome entry site–enhanced green-fluorescent protein (IRES-EGFP) vector was kindly given to us by Dr Claudio Rivera, and the cysteine-to-alanine substitution in amino acid 568 (KCC2-C568A) sequence in a pGEMHE vector was a kind gift from Dr David B. Mount. The KCC2-full-length (FL), KCC2-ΔNTD and KCC2-C568A fragments were subcloned into pBluescript SK (Stratagene, La Jolla, CA, USA) and sequenced. The fragments were then excised and subcloned into the hnestin 1852/tk promoter vector for pronuclear injection and a pcDNA3 vector for cell culture experiments.


The expression cassettes were excised from the vector backbone, purified, and used for pronuclear injection of fertilized mouse (B6D2F1) oocytes. Pronuclear injection and implantation of oocytes into pseudopregnant mice was performed by Karolinska Center for Transgene Technologies. Pregnant dams with embryos at embryonic days (E)9.5–18.5 were killed by spinal dislocation, and the embryos were rapidly dissected out. Pups were collected at birth [postnatal day (P)0]. The transgenic embryos and pups were identified by PCR, using yolk sac or tail DNA as a template. A sense primer complementary to hnestin was combined with an antisense primer complementary to the KCC2 sequence. KCC2 expression assayed by immunohistochemistry (see below) verified an overexpressed protein. Animals were treated according to European Communities Council guidelines (directive 86/609/EEC).


Embryos were fixed for 4 h or overnight in 4% paraformaldehyde in phosphate-buffered saline (PBS), pH 7.4, and thereafter cryoprotected overnight in 30% sucrose in PBS. The embryos were then embedded in mounting medium (Tissue-Tek) and rapidly frozen, and 12-μm sections were serially collected in a cryostat (Leica CM3050 S; Leica Microsystems Nussloch GmbH, Germany). Sections were rinsed in PBS and blocked and permeabilized in 5% donkey serum (Jackson Immunoresearch Laboratories, West Grove, PA, USA), 1% bovine serum albumin (Sigma-Aldrich, St Louis, MO, USA) and 0.3% Triton X-100 (Sigma-Aldrich) in PBS for 45 min, followed by overnight incubation with the primary antibody in a moist chamber. See Table 1, for a full list of the primary antibodies used. The 4.1N antibody was a kind gift from Dr Kari Keinänen (Li et al., 2007). The following day, the sections were washed in PBS and then incubated for 1.5 h with secondary Cy3- or FITC-conjugated antibodies (Jackson Immunoresearch Laboratories) at a 1 : 400 dilution. When the distribution of actin microfibers was investigated, 50 μg/mL FITC- or TRITC-conjugated phalloidin (Sigma-Aldrich) was added to the solution. After subsequent PBS washes, the sections were mounted in Vectashield Hard Set mounting medium (Vector Laboratories, Burlingame, CA, USA). Primary antibodies were titrated to determine the optimal dilutions, and control slides were included with the respective primary antibody omitted. The sections were analyzed in a fluorescent (Zeiss AxioExaminer D1; 10 × and 40 × objectives) or confocal (Leica TCS-SP; 40 × objective) microscope.

Table 1.   Antibodies used for immunohistochemistry and immunocytochemistry
AntibodySpeciesDilutionSourceCat no.
Anti-β-tubulin III/TuJ1Mouse1 : 500Covance, Denver, PAMMS-435P
Anti-PSA-NCAMMouse1 : 200Millipore, Billerica, MAMAB5324
Anti-DoublecortinGuineapig1 : 1000Millipore, Billerica, MAAB2253
Anti-Caspase-3Rabbit1 : 500Cell Signaling Technology, Danvers, MA9661
Anti-Phospho-histone-3Rabbit1 : 200Upstate Biotechnology, Lake Placid, NY06-570
Anti-KCC2Rabbit1 : 300Upstate Biotechnology, Lake Placid, NY07-432
Anti-KCC2Mouse1 : 100UC Davis/NINDS/NIMH NeuroMab Facility, Davis, CAN1/12
Anti-SOX-10Rabbit1 : 200Affinity Bioreagents, Golden, COPA1-17132
Anti-AP-2αMouse1 : 50Santa Cruz Biotechnology, Santa Cruz, CAsc-25343
Anti-Connexin-43Mouse1 : 50Santa Cruz Biotechnology, Santa Cruz, CAsc-13558
Anti-4.1NRabbit1 : 500Dr Kari Keinänen

Cell culturing

The mouse neural stem cell line C17.2 was cultured in Dulbecco’s modified Eagle’s medium (Invitrogen, cat no. 41966) supplemented with 10% fetal bovine serum (FBS), 5% horse serum, 2 mm l-glutamine and 1% penicillin–streptomycin–fungizone (all supplements from Invitrogen). Cells at 80-90% confluency were transfected with the EGFP, KCC2-FL, KCC2-ΔNTD and KCC2-C568A expression vectors using Lipofectamine 2000 (Invitrogen) according to the manufacturer’s instructions.


At 24 or 48 h after transfection, cells were fixed with 4% paraformaldehyde and then permeabilized and blocked in 7% non-fat dry milk and 0.1% Triton X-100 in PBS. Incubation with primary antibodies was done at 4°C overnight. See Table 1 for antibody details. The following day, the cells were rinsed and secondary antibodies were incubated for 1.5 h. Endogenous actin was visualized with FITC- or TRITC-phalloidin (Sigma-Aldrich) diluted to 50 μg/mL in the same solution as the secondary antibody. Thereafter the cells were rinsed in PBS and mounted in Vectashield Hard Set mounting medium (Vector Laboratories), before analysis by fluorescent (Zeiss AxioExaminer D1; 40 × objective) or confocal (Leica TCS-SP; 40 × objective) microscopy.


Transfected C17.2 cells were extracted in ice-cold lysis buffer [50 mm Tris, pH 7.4, 150 mm sodium chloride, 1% NP-40, 1 mm EDTA and 1 ×  protease inhibitor cocktail (Roche)] and the extracts were incubated with 3 μg of a rabbit (Upstate) or monoclonal (NeuroMab) KCC2 antibody. Immunoprecipitates were collected on Protein G Sepharose Fast flow beads (GE Healthcare Biosciences, Uppsala, Sweden) by overnight rotation, washed with lysis buffer, resuspended in 2 × Laemmli sample buffer, and subjected to SDS-PAGE followed by Western blot analysis using anti-4.1N and anti-KCC2 antibodies at a 1 : 2000 dilution.

Wound assay

This method has been described previously (Lindqvist et al., 2010; see also Liang et al., 2007). Briefly, subconfluent C17.2 cells were transfected and then allowed to reach 100% confluency. The cells were then treated with 10 μm Mitomycin C (Sigma-Aldrich) for 3 h to arrest the cell cycle. A scratch was introduced through the cell layer using a pipette tip. The medium was changed to serum-reduced (1% FBS) to keep the cells from dividing, and a line was drawn underneath the culture dish perpendicular to the scratch. Pictures were taken just above or below the line under a light phase-contrast microscope (Nikon Eclipse TE200; 10 × objective), immediately (T = 0) and after 18 h (T = 18 h).

Data analysis

For quantification of β-tubulin III/TuJ1, phospho-histone-3, doublecortin, PSA-NCAM and Caspase-3 (Fig. 4 and Supporting information, Fig. S3), the length and width of the neural tube was measured based on micrographs using the measuring tool in Adobe Photoshop CS (Adobe Systems Inc., San Jose, CA, USA). Positive cells were counted manually and a mark was made on each cell to avoid double counting. The number of cells was divided by the total area of the neural tube. The area unit for the neural tube measurements is mm2.

Figure 4.

 KCC2 overexpression reduced neuronal differentiation. (A–D) TuJ1 labelling in wild-type and transgenic embryos. The neural tube of wild-type embryos displayed numerous TuJ1-positive cells with tangential projections (arrowheads in A) close to the pial surface. (B) KCC2-FL and (C) KCC2-ΔNTD embryos had a reduced proportion of TuJ1-positive cells compared to wild-type littermates. The cells in these transgenic embryos had radial projections (arrows in B and C) but few tangential projections. (D) KCC2-C568A embryos did not differ from wild types. Quantification is shown in (M). (E–H) Doublecortin (Dcx; green) and KCC2 (red) double staining showing endogenous (arrowheads in E) and exogenous KCC2 protein expressed by doublecortin-positive cells. (F) KCC2-FL and (G) KCC2-ΔNTD but not (H) KCC2-C568A embryos displayed a reduced proportion of doublecortin-expressing cells in the neural tube (quantified in N). (I–L) Similarly to TuJ1 and doublecortin, PSA-NCAM expression was reduced in KCC2-FL and KCC2-ΔNTD, but not in KCC2-C568A, embryos (quantified in O). A higher proportion of cells expressed PSA-NCAM in the ventricular and intermediate zones in KCC2-FL and KCC2-ΔNTD embryos (quantified in P). Migrating neuronal cells with leading processes (arrowheads) are shown as insets in I–L. Asterisks in I–L mark the lumen. *< 0.05 and **< 0.01. Scale bars, 20 μm.

When quantifying AP-2α and SOX-10 (Fig. 5), an arbitrary score from 1 to 3, depending on the extent of the neural crest cell groups (supporting Fig. S1), was given to each transverse section with a detectable neural crest. The scores were then summed for each embryo and divided by the size of the embryo.

Figure 5.

 Defective neural crest migration in KCC2-overexpressing embryos. Wild-type embryos at E9.5 displayed abundant expression of (A) AP-2α and (A’) SOX-10 outside the neural tube (arrows). Shown here is the level of the hindbrain and optic pit (OP). (B) KCC2-FL and (C) KCC2-ΔNTD embryos had smaller groups of cells expressing AP-2α and SOX-10 (arrows), and scattered clusters of neural crest were common (arrowheads). (D) KCC2-C568A embryos did not differ from wild-type embryos. (E) Quantification of the extent of neural crest cell groups adjusted by the size of the embryos. *< 0.05. Scale bar, 100 μm.

Imagej (National Institutes of Health; was used to measure the Western blot band intensities (Fig. 8). For quantification of the wound assay results (Fig. 9), both the number of migrating cells and the percentage of area covered were calculated. Adobe Photoshop CS was used to measure the distance between the edges of the wound at T = 0. The same area in images at T = 18 h was identified. The measured distance between the edges, combined with a fixed length of the scratch, yielded a rectangular field. The cells within the field were marked and counted manually, and then divided by the area. The percentage of the re-colonized area was determined using Imagej. For this, binary (black and white) images were generated from the original photomicrographs and the rectangular selection tool was used to create a rectangular field encompassing the wound area at T = 0. Using the X and Y coordinates from the bounding rectangle, the corresponding area was identified in T = 18 h images and the area fraction was calculated using the measuring tool. At least three experiments with triplicates in each were performed.

Figure 8.

 KCC2-FL and KCC2-ΔNTD, but not KCC2-C568A, interacted with the actin cytoskeleton via 4.1N. (A–D) C17.2 cells transfected with control (pEGFP), KCC2-FL, KCC2-ΔNTD and KCC2-C568A plasmids were stained with TRITC-phalloidin. (A′–D′) Co-expression with EGFP or KCC2 is shown in green. (A) Control-transfected cells displayed stress fibres of actin in the cytoplasm (arrows). (B) KCC2-FL and (C) KCC2-ΔNTD expression reduced the stress fiber formation and resulted in diffuse actin staining and aggregates in the cytoplasm (arrowheads). (D) Cells expressing KCC2-C568A did not differ from control-transfected cells. (E–H) 4.1N labelling of transfected cells. A circumferential expression of 4.1N was detected in control-transfected cells (arrows in E). Some cytoplasmic staining could also be observed (arrowhead in E). A more diffuse 4.1N distribution was observed in the cytoplasm of cells transfected with (F) KCC2-FL and (G) KCC2-ΔNTD (arrowheads). The circumferential 4.1N expression was maintained after transfection with KCC2-C568A (arrows in H), although some cytoplasmic staining was evident (arrowhead in H). (E′–H′) Co-expression with EGFP or KCC2 is shown in green and DAPI-staining in blue. (I) Actin levels were not changed in KCC2-transfected cells. (J and K) Immunoprecipitation of the KCC2 protein showed that 4.1N was co-precipitated with KCC2-FL and KCC2-ΔNTD. The co-precipitation was less effective for KCC2-C568A. Protein levels were normalized to those of KCC2. The sizes of the 4.1N and KCC2 proteins are ∼120 and 140 kDa respectively. The KCC2-ΔNTD protein is smaller due to truncation. ***< 0.001. Scale bars, 10 μm.

Figure 9.

 KCC2 affected cell migration in vitro. Migration of transfected C17.2 cells was analyzed in a wound assay. Pictures (A–D) were taken directly after the scratch (T = 0) and (A′–D′) after incubation for 18 h (T = 18 h). Cells transfected with (B) KCC2-FL or (C) KCC2-ΔNTD migrated at a lower extent than (A) control-transfected cells. (D) Transfection with KCC2-C568A did not affect migration. Overlay of binary images from T = 0 (white) and T = 18h (red) are shown in A′′–D′′. (E and F) Quantification of the number of invading cells is shown in E and the re-colonized area fraction is shown in F, *< 0.05; ns, not significant. Scale bar, 50 μm.

Microsoft Excel 2003 was used for the data quantification and statistical analysis. Differences between wild-type and transgenic conditions were determined using Welch’s unpaired t-tests for unequal variances, with significance set at < 0.05 (two-sided). For the embryos, only littermates were compared between groups. Data are presented as means with error bars representing the SDs.


Endogenous KCC2 expression commenced in neuronal progenitors

The developmental KCC2 expression was analyzed in wild-type mouse embryos from E9.5 to E15.5 (n = 4 per age). The KCC2 protein was already detectable in the posterior part of the neural tube at E9.5 (Fig. 1A). Cells expressing KCC2 were observed in the periphery of the neural tube and were also β-tubulin III/TuJ1-positive, implying that KCC2 can be expressed by neurons at early stages of differentiation. The expression was also found in a subset of neural crest cells outside the neural tube (Fig. 1A′). At E11.5, cells expressing KCC2 were observed in the metencephalon and more caudally (Fig. 1B). At E13.5, the KCC2 expression reached the mesencephalon and diencephalon (Fig. 1C). In addition, KCC2 was found in neural crest cells forming the trigeminal and facial ganglia (Fig. 1C′). By E15.5, KCC2 was also observed in the basal telencephalic plate and olfactory bulb (Fig. 1D). This demonstrates that KCC2 is expressed in early neuronal cells during embryonic development and this precedes, by several days, previously shown time points for the hyperpolarizing shift in EGABA (Herlenius, 2001; Stein et al., 2004; Ren & Greer, 2006; Delpy et al., 2008).

Figure 1.

 Expression of KCC2 during development. Endogenous expression of the KCC2 protein is shown from E9.5 to E15.5 in mouse embryos. (A) At E9.5, KCC2 was detected in the marginal zone of the posterior neural tube (enlarged boxes in A′′ and A′′′), and in a subset of neural crest cells (enlarged box in A′), co-expressed with β-tubulin III/TuJ1 (arrows in A′–A′′′). A small number of KCC2-expressing cells were not TuJ1-positive (arrowheads in A′ and A′′). (B) KCC2 was expressed in the metencephalon at E11.5. The expression was still found in the marginal zone (arrows in B′) and neural crest (arrowheads in B′) in β-tubulin III/TuJ1-positive cells (B′′). (C) At E13.5, KCC2-expression reached the mesencephalon and the diencephalon (arrows in C) and was also expressed in neural crest cells forming peripheral ganglia (arrows in C′). (D) The KCC2 protein was detected in the basal telencephalic plate and olfactory bulb at E15.5. V, trigeminal ganglion; VII, facial ganglion; di, diencephalon; mes, mesencephalon; met, metencephalon; tel, telencephalon; BTP, basal telencephalic plate; CP, cortical plate; OB, olfactory bulb. The asterisks indicate the lumen. Scale bars, 20 μm (A–A′′′), 50 μm (B′ and B′′), 100 μm (C′, C′′ and D′), 200 μm (B, C and D).

KCC2 overexpression impaired neural development in vivo

To study the early role of KCC2 in the developing neural tube, we created transgenic mouse embryos using pronuclear injection of a DNA vector containing the KCC2 coding sequence under control of the nestin second intron promoter. This promoter yields a tissue-specific overexpression in neural progenitors from ∼E7 in the mouse (Lothian & Lendahl, 1997; Shariatmadari et al., 2005). We employed three different variants of KCC2: full-length (KCC2-FL), N-terminal-deleted (KCC2-ΔNTD) and point-mutated (cysteine-to-alanine substitution in amino acid 568; KCC2-C568A). The two latter forms have previously been shown to be inactive as K–Cl co-transporters (Li et al., 2007; Reynolds et al., 2008). Notably, both KCC2-FL and KCC2-ΔNTD can interact with the actin cytoskeleton to promote the formation of dendritic spines (Li et al., 2007).

Transgenic embryos were identified with PCR and immunohistochemistry. The KCC2 protein was overexpressed exclusively in the neural tube of these embryos, with a patchy expression pattern throughout the whole tube (Fig. 2A–D), although a higher expression was detected at the level of hindbrain and caudally (not shown).

Figure 2.

 Overexpression of KCC2 in the neural tube. (A) Endogenous KCC2 expression in the neural tube of E9.5 embryos was restricted to the marginal zone. (B–D) Overexpression of KCC2-FL, KCC2-ΔNTD and KCC2-C568A in the neural tube produced a patchy expression pattern. Enlarged sections of KCC2-positive cells are shown as insets in each image. Scale bar, 20 μm.

We collected embryos at E9.5, E11.5, E13.5, E15.5 and E18.5, and newborn pups (P0). The number of transgenic embryos decreased drastically with age and only wild-type and KCC2-C568A mice survived to birth and postnatally. KCC2-FL and KCC2-ΔNTD embryos died between E13.5 and E15.5 (= 2 and = 1, respectively) and had a number of abnormalities including underdeveloped brain structures and cleft palate (Fig. 3B and C; see Table 2 for details). KCC2-C568A mice at E13.5 (= 2) were not different from wild-type littermates (Fig. 3D). Due to the necrotic tissue in KCC2-FL and KCC2-ΔNTD embryos at these stages, we went on to study embryos at E9.5 and E11.5.

Figure 3.

 Phenotypes of wild-type and transgenic embryos. (A–D) KCC2-FL and KCC2-ΔNTD embryos died at E13.5 and had underdeveloped brain structures such as the hindbrain (arrow in B) and midbrain (arrow in C). They also displayed cleft palates (arrowheads in B and C). (D) KCC2-C568A embryos at E13.5 did not differ from (A) wild-type embryos. (E and F) At E11.5, ectopic KCC2 expression resulted in a reduced size and defective body flexures. Higher magnification of the embryo in F is shown in F’. Embryos displayed enlarged olfactory pits (arrow in F’) and a smaller telencephalon and hindbrain (arrowheads in F’). A wild-type E11.5 embryo is shown in E. (G–J) Compared to (G) wild-type embryos at E9.5, (H) KCC2-FL and (I) KCC2-ΔNTD embryos were smaller in size and had aberrant body curvatures (arrowheads). Several embryos also displayed facial abnormalities such as a smaller mandibulum (arrow in H). Note the small size of the forebrain relative to the optic pit (arrow in I). Some KCC2-C568A embryos at E9.5 displayed milder defects such as a slightly smaller brain (J) but most had normal phenotypes. Scale bars, 1 mm.

Table 2.   Comparison of developmental defects in the transgenic embryos
Phenotypes of embryosNumber of embryos displaying phenotype
E9.5–E11.5(= 6)(= 8)(= 6)
 Small brain and spinal cord562
 Defective body flexure551
 Enlarged olfactory pits220
 Small mandible441
 Open neural tube posteriorly010
E13.5–E15.5(= 2)(= 1)(= 2)
 Reduced size210
 Underdeveloped hindbrain210
 Underdeveloped midbrain110
 Small telencephalon210
 Lack of blood circulation210
 Cleft palate110

KCC2-FL (= 6) and KCC2-ΔNTD (= 8) embryos at E9.5 and E11.5 had smaller brains and spinal cords than did wild-type littermates (Fig. 3E–J). They often displayed a loose appearance with the body improperly flexed (Fig. 3F, H and I) or even completely outstretched (supporting Fig. S2). Some transgenic embryos also had aberrant facial structures seen as a small mandibulum or enlarged olfactory pits (Fig. 3F and H). Only two out of the six KCC2-C568A embryos at E9.5 displayed abnormalities similar to, although less than, KCC2-FL and KCC2-ΔNTD embryos. However, the phenotypes of KCC2-C568A embryos were, overall, milder (Fig. 3J) and two-thirds of the embryos had a normal phenotype. Moreover, KCC2-C568A mice survived until birth (> E18.5) and even postnatally (supporting Fig. S2). The phenotypes are summarized in Table 2.

These results show that ectopic expression of KCC2-FL and KCC2-ΔNTD has severe effects on neural development, whereas KCC2-C568A only affects development to a milder extent. The finding that KCC2-FL and ion transport-inactive KCC2-ΔNTD overexpression yielded similar effects indicates that functional K–Cl co-transport is not the cause of the abnormalities seen in the transgenic embryos.

Neuronal differentiation was reduced in KCC2-FL and KCC2-ΔNTD embryos

To examine the abnormalities in the brain of the transgenic embryos, we analyzed cryosections at E9.5. The neural tube was thinner in KCC2-FL (78% of wild-type; = 0.0003, = 6) and in most KCC2-ΔNTD embryos (80% of wild-type; = 0.240, not significant, = 4) compared to wild-type littermates (= 4 and = 3, respectively). However, neurulation was completed in all embryos except for one KCC2-ΔNTD embryo, which displayed an open neural tube posteriorly (supporting Fig. S2). Immunostaining for the early neuronal marker TuJ1 revealed the morphology of differentiating neuronal cells (Fig. 4A–D). In both wild-type and transgenic embryos, TuJ1-positive cells in the neural tube had radial processes and were found mostly in proximity to the pial surface. However, the neuronal cells in wild-type embryos displayed more protrusions in the tangential direction than did the cells in KCC2-FL and KCC2-ΔNTD embryos. In addition, there was a reduced number of TuJ1-positive cells in the neural tube of KCC2-FL (77% of wild-type; = 0.005, = 4) and KCC2-ΔNTD (66% of wild-type; = 0.016, = 4) embryos, but no significant difference in KCC2-C568A embryos (92% of wild-type; = 0.465, = 3) compared to wild-type littermates (= 3 per group; Fig. 4M).

As a reduced differentiation could be due to a decrease in proliferation, we stained for the mitotic marker phosphohistone-3. However, the number of cells positive for phosphohistone-3 did not differ between the neural tubes of transgenic embryos and wild-type littermates (supporting Fig. S3). To further analyze whether the reduced differentiation could be due to increased apoptosis, we examined the expression of caspase-3. We found a small number of apoptotic cells scattered in the neural tube of both the wild-type and transgenic embryos (supporting Fig. S3). There was no detectable increase in apoptosis in the transgenic embryos. These findings indicate that overexpression of KCC2-FL and KCC2-ΔNTD reduced the number of TuJ1-positive cells without affecting proliferation or apoptosis.

Next, we examined a possible effect on neuronal migration. Doublecortin labeling showed migrating neurons in the neural tube and neural crest. The pattern resembled that of TuJ1 with positive cells distributed mainly in the marginal zone (Fig. 4E–H). Similar to TuJ1, doublecortin-expressing cells were significantly reduced in the neural tube of KCC2-FL (42% of wild-type; = 0.025, = 3) and KCC2-ΔNTD (31% of wild-type; = 0.048, = 3) embryos compared to wild-type (= 3 per group) and KCC2-C568A (= 3) embryos. Moreover, we stained for polysialylated neural cell adhesion molecule (PSA-NCAM). PSA-NCAM-positive cells displayed radial projections similar to TuJ1- and doublecortin-expressing cells and were found both in the ventricular and marginal zones of the neural tube, with a higher abundance in the posterior part (Fig. 4I–L). KCC2-FL (= 4) and KCC2-ΔNTD (= 3) embryos had a significantly lower number of PSA-NCAM-positive cells (66 and 62% of wild-type; = 0.044 and = 0.023, respectively), while KCC2-C568A embryos (= 3) did not differ from their wild-type littermates (= 3 per group; Fig. 4 O). In addition, KCC2-FL and KCC2-ΔNTD embryos displayed a larger proportion of PSA-NCAM-positive cells in the ventricular and intermediate zones relative to the marginal zone than did wild-type littermates (30 and 26% more than wild-type; = 0.012 and = 0.0496, respectively; Fig. 4P). These findings suggest that radial migration of neuronal cells may be delayed in KCC2-FL and KCC2-ΔNTD embryos.

Neural crest migration was perturbed in KCC2-FL and KCC2-ΔNTD embryos

The phenotypes of the KCC2-FL and KCC2-ΔNTD embryos indicate disturbances in neural crest cell migration. Neural crest contributes to both the facial bone structures and the bone marrow that produces blood cells (Inoue et al., 2004; Nagoshi et al., 2008). To investigate the distribution of migrating neural crest cells, E9.5 embryos were labelled with the neural crest cell markers AP-2α and SOX-10 (Inoue et al., 2004). In wild-type embryos (= 3 per group), several transverse sections in the hindbrain area showed a large amount of labelled neural crest cells outside the neural tube (Fig. 5A). SOX-10-positive cells were found both inside the neural tube, in a migrating stream projecting from the tube, and in areas further away from the tube. AP-2α-positive cells were mainly located in the areas with longer distances from the neural tube, and co-localized with SOX-10-positive cells, indicating that AP-2α expression turns on at later migratory stages. KCC2-FL (= 4) and KCC2-ΔNTD (= 3) embryos had a lower proportion of transverse sections with detectable neural crest (63 and 70% of wild-type; = 0.019 and = 0.011, respectively) and often displayed a diffuse pattern of these cells (Fig. 5B and C). In contrast, KCC2-C568A embryos (= 4) did not differ from wild-type embryos in the proportion of sections with neural crest (95% of wild-type; = 0.846) nor the neural crest cell pattern (Fig. 5D).

Connexins mediate early, direct and rapid communication between cells (Jaderstad et al., 2010) and play a key role in radial neuronal migration (Elias et al., 2007). Wild-type staining of connexin-43 showed a focused expression in cell processes of neural tube and neural crest cells (Fig. 6A). However, KCC2-FL and KCC2-ΔNTD embryos displayed numerous cells with a loss of this polarized expression pattern and with a more circumferential distribution of connexin-43 (Fig. 6B and C). This indicates that cell polarization, an essential feature of developing and migrating cells, might be disturbed in KCC2-FL and KCC2-ΔNTD embryos.

Figure 6.

 Neural crest cells displayed a loss of polarization in KCC2-FL and KCC2-ΔNTD embryos. (A) Connexin-43 was expressed in the leading process of migrating cells in and outside the neural tube of wild-type embryos. (B–D) A large fraction of the neural crest cells in (B) KCC2-FL and (C) KCC2-ΔNTD, but not (D) KCC2-C568A, embryos displayed a loss of this polarized expression of Connexin-43. Enlarged sections from the neural crest regions in A–D are shown in A′–D′. Scale bars, 50 μm (A–D), 10 μm (A′–D′).

KCC2 overexpression altered the distribution of the actin cytoskeleton

KCC2 has been shown to interact with the actin cytoskeleton in an ion transport-independent manner (Li et al., 2007). We therefore labelled the actin cytoskeleton in the E9.5 embryos using phalloidin. Wild-type embryos displayed an enriched actin labelling at the adherens junctions lining the neural tube (Fig. 7A and E). This pattern was partly lost in KCC2-FL (Fig. 7B and F) and KCC2-ΔNTD (Fig. 7C and G) embryos and, instead, intense cytoplasmic actin staining was observed in several areas of the neural tube. The aberrant distribution of actin was particularly evident in the most affected embryos. No difference in the actin pattern could be detected in KCC2-C568A embryos (Fig. 7D and H).

Figure 7.

 KCC2 overexpression correlated with an aberrant pattern of cytoskeletal proteins in the neural tube. (A–H) Phalloidin labelling displayed the actin distribution in wild-type and transgenic embryos. (E–H) Higher magnifications. (A) Wild-type embryos displayed enhanced actin labelling in the adherens junctions near the lumen (arrow). (B and F) KCC2-FL and (C and G) KCC2-ΔNTD embryos had abnormal cytoplasmic distributions of actin in the cells of the neural tube (arrowheads). (D and H) KCC2-C568A embryos had an actin distribution similar to wild-type embryos. (I–L) 4.1N staining of wild-type and transgenic embryos. (I) The cytoskeleton-associated protein 4.1N was observed in the adherens junctions and as a thin circumferential line around the cells in the neural tube of wild-type embryos. (J and K) KCC2-FL and KCC2-ΔNTD embryos displayed a 4.1N-labelling pattern that resembled the actin-staining. (L) KCC2-C568A embryos did not differ from wild types in the 4.1N expression pattern. Higher magnifications of the 4.1N-staining are shown as insets in I–L. Scale bars, 20 μm (in A for A–D and I–L; in E for E–H and insets in I–L).

As KCC2 has been shown to bind to the cytoskeleton-associated protein 4.1N (Li et al., 2007), we examined the distribution of this protein in our embryos. This revealed a pattern similar to the actin labelling. Compared to wild-type and KCC2-C568A embryos, which displayed 4.1N labelling in the adherens junctions and as a thin circumferential line around the neural tube cells (Fig. 7I and L), the staining of 4.1N in the neural tube of transgenic KCC2-FL and KCC2-ΔNTD embryos was to a large extent located in the cytoplasm (Fig. 7J and K).

To further analyse the effect of KCC2 on the actin cytoskeleton in neural progenitors in vitro, the neural stem cell line C17.2 (Snyder et al., 1992) was transfected with the KCC2-FL, KCC2-ΔNTD and KCC2-C568A constructs and stained with TRITC-phalloidin (Fig. 8A–D). An EGFP plasmid was used as a control. Actin was displayed as stress fibres protruding inside control-transfected cells. We observed an effect of KCC2-FL and KCC2-ΔNTD, but not KCC2-C568A, on the actin cytoskeleton. This was denoted by a reduction in stress fibres and more aggregates of actin, which were diffusely spread in the cytoplasm of the cells (arrowheads in Fig. 8B and C), suggesting a defective assembly of the G-actin subunits. No difference in the relative levels of actin could be detected by Western blot (Fig. 8I).

Furthermore, transfected C17.2 cells were labelled with 4.1N. In control-transfected cells, 4.1N had a circumferential distribution and was highly expressed in cell-to-cell junctions (Fig. 8E). However, in cells transfected with KCC2-FL and KCC2-ΔNTD, the circumferential 4.1N expression was partly lost and a diffuse cytoplasmic staining was observed (Fig. 8F and G). The distribution pattern of 4.1N was not altered in KCC2-C568A transfected cells (Fig. 8H).

The induced changes in the distribution of 4.1N led us to analyze the binding of the three different KCC2 variants to 4.1N. C17.2 cells were transfected with the KCC2 constructs and the KCC2 protein was precipitated using an anti-KCC2 antibody. Protein loads were normalized to KCC2 and thereafter blotted against 4.1N. The observed bands were in the range of the expected molecular weight: 140 kDa (KCC2-FL and -C568A), 130 kDa (KCC2-ΔNTD) and 120 kDa (4.1N). While a strong 4.1N immunoreactivity was present in the immunoprecipitates deriving from cells transfected either with KCC2-FL or KCC2-ΔNTD, only a weak signal was detected in the KCC2-C568A sample (Fig. 8J). We observed a significantly lower binding to 4.1N for KCC2-C568A than for KCC2-FL or KCC2-ΔNTD (< 0.0001; Fig. 8K).

In summary, the above data show that ectopic expression of KCC2-FL and KCC2-ΔNTD, which can both precipitate 4.1N, altered the distribution of actin and 4.1N. In contrast, the KCC2-C568A mutant, which shows a reduced binding affinity to 4.1N, did not affect the cytoskeleton. Thus, we suggest that the interaction between KCC2 and 4.1N plays a key role in the induction of the developmental defects observed in the transgenic embryos.

KCC2-FL and KCC2-ΔNTD reduced migration in vitro

As KCC2-FL and KCC2-ΔNTD had an effect on migration of neural crest cells, we assessed whether ectopic expression could also affect neuronal migration in vitro. C17.2 cells were transfected with control, KCC2-FL, KCC2-ΔNTD and KCC2-C568A plasmids. After 48 h, a scratch was made through the cell layer and the cells were incubated in serum-reduced medium for 18 h to allow migration in the wound area. In control cultures, the wound area was invaded by a moderate number of cells (Fig. 9A). KCC2-FL (Fig. 9B) and KCC2-ΔNTD (Fig. 9C) transfections significantly reduced the number of migrating cells (73 and 72% of control; = 0.016 and = 0.011, respectively). Transfection with KCC2-C568A (Fig. 9D) did not affect the number of cells in the wound area (96% of control; = 0.627). Thus, KCC2-FL and KCC2-ΔNTD perturbed migration of neuronal cells in vitro, similar to the effect on neural crest migration in vivo.


Our work shows that ectopic expression of KCC2 in mouse embryos leads to disturbances in the actin cytoskeleton, which in turn interferes with neuronal differentiation and migration. The results are consistent with a structural role for KCC2 during early neuronal development that is not dependent on the ion transport function of KCC2.

In several parts of the central nervous system, such as the spinal cord (Delpy et al., 2008) and brainstem (Balakrishnan et al., 2003; Blaesse et al., 2006), KCC2 is expressed before the onset of functional Cl extrusion. Moreover, the levels of KCC2 expression in the auditory brainstem do not change at the periods of the hyperpolarizing EGABA shift (Balakrishnan et al., 2003; Vale et al., 2005). It has been suggested that the early expressed protein is inactive and requires regulation of its localization, state of phosphorylation, or oligomerization for functional activation (Vale et al., 2005; Blaesse et al., 2006; Lee et al., 2007; Hartmann et al., 2009). KCC2 shows a high level of expression in the proximity of excitatory synapses and within dendritic spines (Gulyas et al., 2001) and, more recently, is has been shown that KCC2 promotes the development of spines through interaction with the cytoskeleton-associated protein 4.1N (Li et al., 2007). Thus, KCC2 has a morphogenic role that is independent of its ion transport function. This morphogenic role may explain the early presence of KCC2 prior to the hyperpolarizing EGABA shift.

The present results show that KCC2 is already endogenously expressed at E9.5 in neuronal cells of mouse embryos. This is earlier than previously shown time points for KCC2 expression (Li et al., 2002; Stein et al., 2004; Delpy et al., 2008) and indicates that neurons require KCC2 at an early stage of maturation. KCC2 is co-expressed with β-tubulin III in the neural tube and neural crest cells, possibly reflecting an involvement in GABA-mediated regulation of neuronal migration (Bolteus & Bordey, 2004). Notably, it has recently been shown that functionally active KCC2 induces migratory arrest in cortical interneurons (Bortone & Polleux, 2009). However, the ion transport-independent structural role of KCC2 and the expression of functionally inactive KCC2 described in our study suggest a dual role for the transporter in neuronal migration. Indeed, we found a reduced migration of a neural cell line transfected with both transport-active KCC2-FL and transport-inactive KCC2-ΔNTD, indicating an ion transport-independent effect on migration.

In line with the results in vitro, KCC2-FL and KCC2-ΔNTD embryos displayed a perturbed neural crest migration and sometimes also smaller mandibles and enlarged olfactory pits at E9.5. This is consistent with the phenotypes of transgenic embryos at later stages showing aberrant facial structures. Neural crest cells migrate from the neural tube to different regions in the body and develop into various structures such as the craniofacial bones, peripheral nervous system, cardiac outflow septum and endocrine glands (Bronner-Fraser, 1993; Inoue et al., 2004). The cause of death of KCC2-FL and KCC2-ΔNTD embryos between E13.5 and E15.5 has not been determined, but the whitish appearance indicates a lack of blood cells. Indeed, neural crest cells have been shown to contribute to the bone marrow where red blood cells are generated (Nagoshi et al., 2008).

We observed reduced expression of β-tubulin III, doublecortin and PSA-NCAM in E9.5 transgenic embryos. The reduction in neuronal cells did not seem to be due to a change in proliferation rate or apoptosis. A reduced differentiation could, however, be caused by a delay in radial migration. The pattern of PSA-NCAM expression displayed a higher proportion of positive cells in the ventricular and intermediate zones, indicating a perturbed radial migration in KCC2-FL and KCC2-ΔNTD embryos. As another study has reported that ectopic KCC2 expression by in utero electroporation at E17–18 does not affect radial migration in perinatal rats (Cancedda et al., 2007), the effect of KCC2 on migration might be age-specific. The reduction in neuronal cells corroborates a previous report showing that KCC2 overexpression reduces neuronal differentiation in zebrafish (Reynolds et al., 2008). However, the authors concluded that this reduction is caused by a negative shift in the GABA response as the use of KCC2-C568A did not produce similar effects. We did not observe any significant effects on neuronal differentiation with KCC2-C568A either, but there was a similar reduction in neuronal cells in KCC2-FL and KCC2-ΔNTD embryos. Thus, we suggest that KCC2 overexpression reduces neuronal differentiation by an ion transport-independent mechanism.

Both KCC2-FL and KCC2-ΔNTD can interact with the actin cytoskeleton by direct structural interaction of the intracellular C-terminus with the actin-binding protein 4.1N (Li et al., 2007). We found aberrant actin and 4.1N patterns in the neural tube of transgenic embryos. The cells of the neural tube had diffuse cytoplasmic levels of actin and 4.1N. Similar results were obtained in the neural cell line C17.2. The cytoplasmic staining in KCC2-overexpressing cells points to a redistribution of the 4.1N protein within the cell, perhaps leading to a defective formation of F-actin. KCC2-C568A did not produce similar effects on the actin cytoskeleton, indicating that the point mutation rendered KCC2 less effective in binding to 4.1N. Indeed, immunoprecipitation of the three variants of the KCC2 protein demonstrated a significantly lower binding of KCC2-C568A to 4.1N (Fig. 8). Previous studies have employed KCC2-C568A as a control for KCC2-FL overexpression (Cancedda et al., 2007; Reynolds et al., 2008). The lack of effects of KCC2-C568A was suggested to be due to inactivation of the ion transport function. However, this interpretation does not exclude a structural effect of KCC2, as our data suggest.

It is not clear whether the C568A mutation interferes with the folding or intracellular trafficking of the protein or resides in an important 4.1N-binding structure. However, the mutation lies within a central domain of the KCC2 protein, and the 4.1N-binding domain has been localized to the C-terminus (Li et al., 2007). As we have detected expression of KCC2-C568A at the protein level, we propose that the mutation has a major influence on the tertiary structure of KCC2, yielding a protein inactive both as an ion transporter and as an interacting partner of 4.1N.

Taken together, our results indicate that KCC2 regulates early neuronal differentiation and migration by effects mediated through direct structural interaction with 4.1N and the actin cytoskeleton. This interaction may be essential for neural tube development.


We wish to thank Ruth Detlofsson, Panagiotis Papachristou, Maria Lindqvist and the Karolinska Center for Transgene Technologies for technical support, and Evan Y. Snyder for the C17.2 cells. This study was supported by grants from the Swedish Research Council, Stockholm County Council, M&M Wallenberg, Sällskapet Barnavård, Swedish Heart and Lung Foundations (E.H.), the Academy of Finland and the Sigrid Jusélius Foundation (K.K.). Z.H. is supported by the League of European Research Universities (LERU). K.K. is a member of the Finnish Center of Excellence in Molecular and Integrative Neuroscience Research.


cysteine-to-alanine substitution in amino acid 568


embryonic day


reversal potential of GABAA receptor-mediated currents


enhanced green-fluorescent protein




human nestin 2nd intron


postnatal day


phosphate-buffered saline


internal ribosome entry site


polysialylated neural cell adhesion molecule


N-terminal deletion