K-Cl Cotransporter 2–mediated Cl⁻ Extrusion Determines Developmental Stage–dependent Impact of Propofol Anesthesia on Dendritic Spines

ANNUALLY, millions of infants undergo general anesthesia worldwide. An increasing number of clinical and experimental observations suggest that when administered during early life, anesthetic drugs trigger long-term morphologic and functional alterations in the central nervous system.^1–4  Notably, general anesthetics can induce long-term interference with physiologic patterns of synaptic growth. Most studies regarding the impact of these drugs on synaptogenesis have been conducted on the rodent cortex, where the most intense phase of synapse formation takes place during the first postnatal month.⁵  In humans, this spans the period from the third trimester of gestation to the first few years of life.^6,7  In rats, when administered at early stages of the peak synaptogenic period (around postnatal days [PNDs] 5 to 10), general anesthetics rapidly led to a significant and permanent decrease in dendritic spine density.^8–10  In contrast, when applied at later stages of development (around PND 15), these very same drugs have no such effects or can even lead to an increase in the number of synaptic contacts.^11,12  The molecular and cellular mechanisms underlying these developmental stage–dependent differences remain as yet unknown.

Most general anesthetics act as inhibitors of glutamate receptors or potentiators of γ-aminobutyric acid receptor type A (GABA_A)–mediated signaling.¹³  During early postnatal development in rodents, including the initial stages of synaptogenesis, intraneuronal Cl⁻ concentrations ([Cl⁻]_i) are high, and GABA_A receptor activation results in Cl⁻ efflux with consequent postsynaptic depolarization, sometimes triggering action potentials. In contrast, during later stages of synaptogenesis,¹⁴  developmental up-regulation of the neuron-specific potassium-chloride (K-Cl) cotransporter 2 (KCC2) results in a reduction of [Cl⁻]_i, which is necessary for functionally inhibitory GABA_A signaling.^15,16 

Propofol, one of the most widely used GABA_A-enhancing agents for induction of anesthesia, is also one with negligible inhibition of glutamate receptors at clinically relevant concentrations.¹³  Here, our hypothesis is that the qualitative impact of propofol on the density of spines depends on the efficacy of KCC2-mediated Cl⁻ extrusion, which in turn sets the efficacy of GABA_A receptor-mediated inhibition.¹⁵  By employing in utero electroporation of an N-terminally truncated KCC2 variant that lacks K-Cl transport functionality but retains cytoskeletal interactions,^5,17,18  we show that the efficacy of KCC2-mediated Cl⁻ extrusion, but not its transport-independent morphogenic functions,^15,19  determines the effect of propofol anesthesia on dendritic spines of cortical pyramidal neurons. To our best knowledge, the current results provide the first evidence that the increase in KCC2-mediated Cl⁻ extrusion efficacy acts as a cellular mechanism to account for the developmental stage–dependent impact of GABA_A receptor-targeting general anesthetics on neuronal connectivity in the cerebral cortex in vivo.

The experiments were conducted according to the guidelines of the Swiss Federal Veterinary Office (Bern, Switzerland), with the approval of the cantonal Animal Welfare Committee (Geneva and Lausanne, Switzerland), and permission of the National Animal Ethics Committee of Finland (Helsinki, Finland) and the local Animal Ethics Committee of the University of Helsinki (Helsinki, Finland). Wistar rats and ICR mice were housed and bred in the animal facilities under a 12-h light–dark cycle and temperature-controlled (22° ± 2°C) conditions. Food and water were available ad libitum. Animals were randomly assigned to experimental conditions.

Acute 400-μm coronal whole-brain slices containing the somatosensory cortex (SSC) were prepared from PND 5 to 20 rats and mice as described previously.⁵  Before the experiments, the slices were allowed to recover for 45 min at 34°C in a physiologic solution containing 124 mM NaCl, 3.5 mM KCl, 2 mM CaCl₂, 25 mM NaHCO₃, 1.1 mM NaH₂PO₄, 2 mM MgSO₄, and 10 mM d-glucose and equilibrated with carbogen (95% O₂ and 5% CO₂).

A protease approach was used to assess the surface expression of KCC2.²⁰  SSC microslices, obtained from rats at PNDs 5, 10, 15, and 20, were incubated on ice (less than 4°C) for 60 min in the presence of cod trypsin (2 U/ml; Zymetech, Iceland) in carbogen-bubbled standard physiologic solution. The cleavage reaction was stopped using 100 μM protease inhibitor phenylmethylsulfonyl fluoride, after which the microslices were analyzed using immunoblotting.²¹  SSC microslices were homogenized in radioimmunoprecipitation assay buffer containing protease inhibitors. Subsequently, 30 μg protein in Laemmli sample buffer (Bio-Rad, USA) was incubated at 37°C for 30 min, electrophoretically separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis, and transferred to nitrocellulose membranes (PerkinElmer, USA). The membranes were blocked in phosphate-buffered saline (PBS) or 4% milk for 1 h at room temperature. Incubation with the respective antibodies (in PBS or 4% milk) was performed overnight at 4°C. Primary antibodies used were rabbit anti-KCC2 (C-terminal; 1:1,000; Millipore, USA) and mouse anti–β-tubulin (1:5,000; Sigma, USA). The horseradish peroxidase–conjugated secondary antibodies, donkey anti-rabbit immunoglobulin G (1:5,000; Amersham GE Healthcare, USA) and goat anti-mouse immunoglobulin G (1:1,000; Dako Agilent, USA), were applied for 1 h at room temperature in PBS or 4% milk with agitation. Immunoreactivity was detected and quantified as before.²¹  Surface-expressed plasmalemmal KCC2 was quantified on the basis of a cod trypsin–cleaved KCC2 fragment (~100 kDa). Total KCC2 was quantified as the sum of the immunodensity signals of the noncleaved ([internal] ~135 kDa) and cleaved KCC2. Total and surface KCC2 signals were normalized to β-tubulin.

Changes in membrane potential in response to somatic GABA_A receptor activation were assessed using local ultraviolet-flash photolysis (372 nm; 0.8 to 2.0 mW; ~10 μm spot diameter; 5- to 50-ms flash; 0.033 Hz) of caged GABA (100 μM; DPNI-GABA; Tocris, United Kingdom) in gramicidin (40 μg/ml) perforated-patch current clamp recordings in brain slices obtained from the rat SSC at PNDs 10 and 15. Patch pipettes (resistance 3 to 4 MΩ) were filled with solution containing 150 mM KCl, 10 mM HEPES, and 5.6 mM KOH; pH 7.2. Recordings from cells with a stable access resistance in the range of 80 to 300 MΩ were accepted. Alexa Fluor 488 (40 μM; Thermo Fisher, USA) was included in the pipette solution to monitor patch integrity via confocal fluorescence microscopy. Propofol was applied from a 100 mM stock in dimethyl sulfoxide to the extracellular solution to yield a final concentration of 5 μM.²²  Dimethyl sulfoxide vehicle alone had no effect on GABA uncaging–evoked responses. Propofol was applied 15 min before recording the membrane responses.

To assess the efficacy of Cl⁻ extrusion mediated by KCC2 in pyramidal neurons of SSC layer 2/3, we used our standard assay where a constant somatic Cl⁻ load (19 mM) is imposed on the neuron in whole-cell patch configuration and GABA_A receptor-mediated currents are elicited using local uncaging of GABA (1 mM) at the soma and 50 μm away at the apical dendrite.²³  Brain slices, obtained from rats and mice between PNDs 5 and 20, were superfused with carbogen-bubbled physiologic solution containing the following drugs: 10 μM bumetanide (Tocris; to block Na-K-2Cl cotransport), 0.5 μM TTX, 1 μM CGP 55845, 10 μM CNQX, and 20 μM D-AP5 (all from Abcam, United Kingdom). The pipette solution consisted of 18 mM KCl, 111 mM K-gluconate, 0.5 mM CaCl₂, 2 mM NaOH, 24 mM KOH, 10 mM glucose, 10 mM HEPES, 2 mM Mg-ATP, 5 mM BAPTA, and 0.1 mM Alexa Fluor 488; pH 7.3. Pyramidal neurons were identified using infrared microscopy based on their morphology and transfected neurons via confocal microscopy of enhanced green fluorescent protein (EGFP) fluorescence.¹⁸  Data obtained from cells with resting membrane potential less than −60 mV and stable access resistance throughout the recording were included in the analysis. Membrane potential values were corrected for series resistance effect and for a calculated liquid junction potential of 14 mV.²⁴ 

In utero electroporation of timed-pregnant rats and mice with embryonic day 17.5 and embryonic day 14.5 embryos, respectively, was done as before.^5,18  The full-length KCC2 (KCC2-FL), N-terminally truncated KCC2 (KCC2-ΔNTD), anti-KCC2 small hairpin RNA (shKCC2), and the inward-rectifier potassium channel Kir2.1 complementary DNAs have been described and used by us previously for in utero electroporation and transfection.^5,18,25,26  An EGFP-coding plasmid bearing a modified chicken β-actin promoter with a cytomegalovirus immediate-early enhancer (CAG) and an internal ribosome entry site, pCAG-IRES-EGFP, was coinjected to fluorescently tag electroporated cells, yielding a highly efficient cotransfection rate.⁵  As before,^5,18,25  the KCC2 and Kir2.1 complementary DNAs were expressed under the CAG promoter. We have recently shown that in utero electroporation of the vector expressing EGFP permanently labels layer 2/3 pyramidal neurons in the SSC without perturbing dendritic arbor and spine development of these cells.⁵  Experiments on in utero electroporated animals were performed and analyzed in a blinded manner with respect to the genetic construct used.

General anesthesia was induced in PND 10 rats by intraperitoneal injection of propofol (40 mg/kg; Fresenius, Germany) as described previously.¹⁰  This first dose of the drug induced sedation (i.e., loss of the righting reflex) between 45 and 60 min. After the initial dosing regimen, animals received an hourly intraperitoneal injection (four in total) of propofol (20 mg/kg) to maintain sedation for 6 h. Control animals received intraperitoneal injections of a lipid vehicle solution (20% Lipofundin; Fresenius) at equivalent volumes and frequency. All pups underwent the same maternal separation and handling as the anesthetized animals and were kept in individual cages for the duration of the experimental procedure. Body temperature was monitored and maintained at 37° to 38°C using a heating pad (Harvard Apparatus, USA). We have shown previously that this anesthesia protocol maintains blood gas parameters in the physiologic range.¹⁰  Notably, the currently used anesthesia protocol does not result in changes in cortical KCC2 protein expression levels within the present experimental timeframe²⁷  and is expected^28,29  to yield brain concentrations within the low micromolar range that specifically target GABA_A and glycine receptors.¹³ 

For dendritic spine analysis, animals were euthanized immediately at the end of the 6 h propofol anesthesia via transcardiac perfusion of paraformaldehyde and slices were prepared from acutely fixed brain tissue as described previously.⁵  Briefly, 300-μm coronal sections containing the SSC (bregma, −0.6 to −2 mm; lambda, 5.2 to 3.8 mm both at PND 10 and at PND 15) were cut on a vibratome in ice-cold PBS (pH 7.4). Sections were stained for 30 min with methylene blue to allow visualization of neuronal somata. Pyramidal neurons in layer 2/3 of the SSC were loaded iontophoretically with 5% Lucifer Yellow (Sigma) via sharp micropipettes using a negative current of 70 nA. To reveal detailed dendritic arbor architecture of the iontophoretically injected neurons, brain slices were incubated with primary and secondary antibodies against Lucifer Yellow and EGFP as described previously.⁵ 

Only neurons with proper Lucifer Yellow filling of distal dendritic tips were analyzed. An LSM 510 META confocal microscope (Zeiss, Germany) equipped with a ×63 oil-immersion objective was used for dendritic spine analysis. Spine analysis was performed as before on acquired stacks of images using a homemade plug-in written for OsiriX software (Pixmeo, Switzerland).⁵  As reported by us previously,⁵  and in contrast to a previous study,²⁵  we did not observe any effect after overexpression of KCC2-FL or Kir2.1 on the dendritic branch number or total arbor length of layer 2/3 SSC pyramidal neurons, as assessed by iontophoretic Lucifer Yellow loading and immunolabeling (see also Pfeffer et al.³⁰ ).

Due to the mechanistic and exploratory nature of this work, no statistical power analysis was used to guide sample-size estimation. Based on our previous experience, 3 to 4 animals (8 to 12 cells per animal, all of which were from different litters) were used per experimental group when dendritic spine densities were considered as the outcome measure. This approach allowed us to analyze between 907 and 3,371 dendritic spines in each experimental group that were viewed as independent (i.e., uncorrelated) measurements. In biochemical and electrophysiologic experiments, 3 to 11 animals were used in each experimental group. The sample size of each experimental group is stated in the figure legends. Normality was tested using the D’Agostino and Pearson test for each distribution, and α was set to 5% for all tests. Normally distributed data were analyzed using two-tailed Student’s t test; for non-Gaussian distributions, the Mann–Whitney rank sum test was used. For multiple comparisons, statistical significance was determined using one-way or two-way ANOVA with repeated measures, both followed by Bonferroni post hoc tests. Wilcoxon signed-rank test was used for paired comparisons of KCC2 immunofluorescence signal. Prism Software 5.0a (GraphPad, USA) and Sigmaplot 12 (Systat, USA) were used to analyze the data. Data are expressed as mean ± SD. No data were subject to loss.

Flash photolysis of caged GABA was used to investigate the effect of GABA_A receptor activation on membrane potential in PND 10 and 15 rat SSC layer 2/3 pyramidal neurons in gramicidin-perforated current clamp recordings with zero current (fig. 1A). In line with previous data obtained using single-channel recordings,³¹  somatic GABA uncaging at PND 10 in the majority of the recorded neurons (11 of 14) depolarized the resting membrane potential, often triggering spiking. Bath application of propofol at a clinically relevant concentration (5 μM)²²  potentiated the depolarizing response and facilitated action potential triggering in neurons, which did not fire under control conditions (n = 3; fig. 1A). In a minority of PND 10 neurons with a hyperpolarizing response to GABA, the response was likewise potentiated by propofol (not shown). By PND 15, in line with the gradual decrease in GABA_A receptor-mediated depolarization in this neuronal population,³¹  membrane potential responses to GABA were either hyperpolarizing (n = 3) or shunting (n = 9) and action potential firing could not be triggered in any of the recorded neurons (n = 12), even at maximal light intensity and a 10-fold flash duration. As in PND 10 neurons, propofol potentiated the GABA-evoked change in membrane potential (n = 3; fig. 1A).

It is evident from data shown in figure 1, B and C, that endogenous KCC2 protein expression is low between PNDs 5 and 10 in the rat SSC (PND 5, 5.2 ± 5.6% of PND 20; PND 10, 15.7 ± 13.1% of PND 20). A robust increase in the expression levels of total KCC2 protein is observed by PND 15 (70.4 ± 47.2% of PND 20; P < 0.001 vs. PND 10). The increase in the mean value from PND 15 to 20 was not statistically significant (P = 0.075; fig. 1, B and C). During neuronal development, KCC2 is also heavily influenced by posttranslational modifications, increasing its surface expression and kinetic activation.¹⁵  To quantify the plasmalemmal KCC2 protein pool, we employed molecular shaving of cell surface proteins by cold-adapted cod trypsin, which enables assessment of changes in the plasmalemmal KCC2 pool under conditions of arrested membrane trafficking.^20,23  As was the case for total KCC2, a robust up-regulation in plasmalemmal KCC2 was observed between PNDs 10 (12 ± 9% of PND 20) and 15 (49.7 ± 18.6% of PND 20; P < 0.001). A further significant increase in plasmalemmal KCC2 took place between PNDs 15 and 20 (P < 0.001; fig. 1, B and C).

To compare the developmental expression patterns of KCC2 with its transporter functionality, we electrophysiologically assessed the Cl⁻ extrusion capacity of KCC2 in layer 2/3 pyramidal neurons at PNDs 5, 10, and 15 (fig. 1, D and E). Under the present experimental conditions, a fixed 19 mM somatic Cl⁻ load imposed via the whole-cell patch pipette was calculated (Goldman–Hodgkin–Katz [GHK] voltage equation) to generate a reversal potential of GABA_A receptor-mediated currents (E_GABA) of −47.7 mV. In the absence of a Cl⁻-extruding mechanism to counteract the diffusion of Cl⁻ from the pipette, the local neuronal E_GABA is expected to be close to the above value obtained from the GHK equation. The mean somatic and dendritic E_GABA obtained using local GABA uncaging from PND 5 neurons was very similar (−48.1 ± 1.8 and −49.1 ± 1.8 mV, respectively; P = 0.451; fig. 1, D and E), indicating inefficient Cl⁻ extrusion capacity at this age, in line with the above KCC2 immunoblot data. In agreement with the gradual increase in KCC2 expression during the second postnatal week in the SSC (fig. 1, B and C), by PND 10, a modest but significant difference (P < 0.001) was observed between the mean somatic (−51.1 ± 2.1 mV) and dendritic (−56.7 ± 2.7 mV) E_GABA values (fig. 1E). By PND 15, when GABA uncaging no longer triggered action potential firing, a robust negative shift was observed in mean E_GABA values both at the soma (−62.6 ± 3.1 mV; P < 0.001 to value at PND 10) and the apical dendrite (−72.7 ± 4.6 mV; P < 0.001 to value at PND 10; fig. 1D).

In SSC layer 2/3 pyramidal neurons from PND 15 to PND 20 rats, coelectroporated in utero with plasmids encoding EGFP and shKCC2,²⁶  somatic (−49.4 ± 2.9 mV) and dendritic (−50.5 ± 2.8 mV) E_GABA values were close to that predicted by the GHK equation using the pipette [Cl⁻], indicating no effective Cl⁻ extrusion in the absence of KCC2 (fig. 1E). Somatic (−60.1 ± 5.8 mV) and dendritic (−71.1 ± 7.3 mV) E_GABA values in neighboring nontransfected control layer 2/3 pyramidal cells (n = 9 neurons; 3 slices; 3 animals) of in utero electroporated rats were not different from respective values in naive PND 15 animals (P = 0.251 and 0.822, respectively).

Similar to the results obtained from rats, recordings with 19 mM Cl⁻ in the pipette performed in the mouse SSC demonstrated a striking developmental enhancement of Cl⁻ extrusion capacity of layer 2/3 pyramidal neurons (not illustrated). Also here, both the somatic and dendritic E_GABA values underwent a negative developmental shift from −52.7 ± 1.8 to −64.5 ± 4.6 mV (P < 0.005) and from −55.3 ± 2.9 to −78.6 ± 6 mV (P < 0.005), respectively, as seen in neurons recorded at PNDs 6 to 7 (n = 8 neurons; 8 slices; 6 animals) and PNDs 15 to 20 (n = 13 neurons; 12 slices; 11 animals). Taken together, our biochemical and electrophysiologic data show that robust up-regulation of transport-functional KCC2 protein takes place by the end of the second postnatal week in rodent SSC layer 2/3 pyramidal neurons.

In line with our previous observations on layer 5 pyramidal neurons in the rat medial prefrontal cortex,¹⁰  we found that a 6 h anesthesia at PND 10 with propofol results in 34 ± 4% (P = 0.03) and 35 ± 6% (P = 0.004) reduction in spine densities of apical and basal dendritic shafts of SSC layer 2/3 pyramidal neurons, respectively (fig. 1, F and G). Analysis of the maximal cross-sectional diameter of the spine head, as an approximation of spine head volume and thus of synaptic strength,³²  revealed that the propofol-induced decrease in spine density was primarily caused by a decrease in the number of protrusions with small head diameters (fig. 1H). In contrast to PND 10, 6 h propofol exposure at PND 15 induced a significant increase in spine densities in both apical (53 ± 9%; P = 0.001) and basal (35 ± 4; P = 0.004) dendrites (fig. 1, I and J), which was mainly due to the increase in the number of protrusions with small heads (fig. 1K).

Taken together, the above results lead to the prediction that premature expression of KCC2 is expected to alleviate or prevent the propofol-induced reduction in dendritic spines at PND 10. To test this, we coelectroporated in utero at embryonic day 17.5 the EGFP plasmid with a vector expressing KCC2-FL into rat neural progenitors, which later give rise to layer 2/3 pyramidal neurons. As shown previously, this approach results in reliable coexpression of EGFP and KCC2, leading to a persistent premature enhancement of neuronal Cl⁻ extrusion capacity and hence in negative shifts in E_GABA values of SSC pyramidal neurons of rodents as young as PND 0.^18,25  Indeed, in PND 10 rat layer 2/3 pyramidal neurons cotransfected with KCC2-FL and EGFP (KCC2-FL/EGFP+), Cl⁻ extrusion capacity was significantly enhanced compared to neighboring EGFP− nontransfected controls (somatic E_GABA KCC2-FL/EGFP+: −66.5 ± 4.3 mV vs. EGFP−: −53.9 ± 1 mV; dendritic E_GABA KCC2-FL/EGFP+: −74.2 ± 4.9 mV vs. EGFP−: −59 ± 1.4 mV; P < 0.001 for both; n = 10 neurons [KCC2-FL/EGFP+]; n = 7 neurons [EGFP−]) and was similar to the level observed in naive PND 15 pyramidal neurons (fig. 1E). In line with the above, quantification of the somatic plasmalemmal KCC2 immunofluorescence of KCC2-FL/EGFP+ coelectroporated pyramidal neurons from anesthesia-naive PND 10 animals revealed a 3.6 ± 2.3-fold increase in KCC2 (P < 0.001; n = 33 neurons per group; 5 slices; 3 animals) signal compared to neighboring EGFP− controls (fig. 2A).

Notably, after the KCC2-FL/EGFP coelectroporated rats had been exposed to a 6 h propofol anesthesia at PND 10, analysis of KCC2-FL/EGFP+ neurons filled iontophoretically with Lucifer Yellow no longer revealed significant differences in mean dendritic spine densities on both apical (+17 ± 3%; P = 0.06) and basal dendritic shafts (+6 ± 8%; P = 0.36 vs. KCC2-FL/EGFP+ of vehicle-treated PND 10 animals; fig. 2, B–D). Dendritic spine head diameters were comparable between propofol- and vehicle-treated KCC2-FL/EGFP coexpressing neurons (fig. 2E). In contrast, in these same PND 10 animals, propofol still induced a significant decrease in the number of both apical (−30 ± 5%; P = 0.003) and basal (−27 ± 9%; P = 0.012) dendritic spines in adjacent EGFP− pyramidal neurons (fig. 2B–D), which express low native levels of KCC2 (fig. 2A). These results thus indicate that premature expression of KCC2 in immature pyramidal neurons prevents the decrease in dendritic spine density after propofol anesthesia at PND 10. Control experiments showed that electroporation of an EGFP-expressing plasmid into rat pyramidal neurons per se does not affect dendritic spine responses to propofol (fig. 3). Furthermore, we have recently shown that the current propofol anesthesia protocol does not result in changes in cortical KCC2 protein expression levels within the current experimental timeframe.²⁷ 

Previous work, using the EGFP signal alone for neuronal reconstruction, reported that premature expression of KCC2 perturbed arborization of rat SSC layer 2/3 pyramidal neurons, as seen at PNDs 6 and 14.²⁵  However, in a more recent work⁵  and also in the current study, using a higher fidelity method for complete visualization of dendritic arbors, Lucifer Yellow filling of neurons, no such effects were observed on branch point number or total dendritic length of these neurons as seen at PNDs 10, 15, or 90. In line, we found no difference between KCC2-overexpressing and neighboring control PND 10 rat layer 2/3 pyramidal neurons in their cell capacitance (72.5 ± 13.3 vs. 66.4 ± 16.1 pF; P = 0.40; n = 10 and 8 neurons, respectively), an electrophysiologic parameter that is directly proportional to the neuronal surface area.

To investigate whether the KCl cotransporter function of KCC2, but not interaction of its C-terminal domain with the cytoskeleton,^5,17,18,33  is required to prevent propofol-induced dendritic spine loss, we took advantage of the ion transport-deficient KCC2-ΔNTD construct.^5,17,18  As seen in figures 2 and 4, and in line with our previous work,⁵ in utero electroporation of either KCC2-FL or KCC2-ΔNTD into pyramidal neurons results in a significant increase in dendritic spine density compared to adjacent nontransfected neurons. However, in contrast to neurons expressing the KCC2-FL construct (fig. 2), propofol still induced a significant decrease in dendritic spine densities in both apical (−53 ± 3%; P = 0.002) and basal (−50 ± 3%; P = 0.004) dendrites of cells transfected with KCC2-ΔNTD (fig. 4A–C). The magnitude of this dendritic spine loss was comparable to that observed in adjacent nontransfected cells (apical, −45 ± 7%, P = 0.0001; basal, −49 ± 3%, P = 0.0004), and as in controls, it was primarily due to a decrease in the number of spines with small head diameters (fig. 4D).

We have shown previously that unlike the case with KCC2-FL, in utero electroporation of KCC2-ΔNTD does not result in enhancement of Cl⁻ extrusion capacity of rodent SSC layer 2/3 pyramidal neurons, as seen in mice at the end of the first postnatal week.¹⁸  To rule out the unlikely scenario that this is an age-dependent effect confined to the first postnatal week, we assessed the Cl⁻ extrusion capacity of SSC layer 2/3 pyramidal neurons overexpressing FL-KCC2, KCC2-ΔNTD, or EGFP alone in slices from mice at PNDs 15 to 20 (fig. 4, E and F). As expected, mean somatic and dendritic E_GABA values (−60.9 ± 4.5 and −73.4 ± 6.6 mV, respectively) in neurons expressing KCC2-ΔNTD were not significantly different from respective values in neurons expressing EGFP alone (−64.1 ± 4.6 and −79 ± 5.7 mV; P = 0.789 and 0.159, respectively) or from those measured in neighboring EGFP−nontransfected control neurons (P = 0.597 and 0.252, respectively; fig. 4, E and F). In accordance with our previous results,^17,18  these data clearly show that the KCC2-ΔNTD construct encodes for a persistently ion transport–inactive variant of KCC2 in rodent cortical pyramidal neurons in vivo. As shown for rats (cf. fig. 1E), mouse neurons transfected with the KCC2-FL construct had significantly more negative somatic (−72.2 ± 5.9 mV; P < 0.001 vs. EGFP alone) and dendritic (−85.4 ± 4.8 mV; P = 0.037 vs. EGFP alone) E_GABA values (fig. 4F). Taken together, these results indicate that the K-Cl cotransport function of KCC2 is required for preventing propofol-induced dendritic spine loss.

To further explore the possibility that the ion transport function of KCC2 is required for protecting pyramidal neurons against propofol-induced spine loss, we carried out in utero electroporation experiments using a construct encoding the inward-rectifier potassium channel Kir2.1. Upon overexpression, Kir2.1 decreases intrinsic neuronal excitability by persistently hyperpolarizing the membrane potential,^25,34  thereby functionally mimicking the effect of KCC2-dependent properties of GABA_A receptor actions on neuronal excitability. We first investigated whether the overexpression of Kir2.1 from early developmental stages in rat SSC layer 2/3 pyramidal neurons affects the development of spines in these cells. As seen in figure 5, when assessed at PNDs 10 and 15, Kir2.1-transfected neurons displayed dendritic spine densities and spine head diameters comparable to control neurons both at apical and basal dendrites, indicating that overexpression of Kir2.1 per se does not interfere with physiologic patterns of dendritic spinogenesis during the brain growth spurt. We next exposed animals in utero electroporated with Kir2.1 to propofol anesthesia on PND 10. We found no effect of propofol on the mean density of dendritic spines in neurons expressing the Kir2.1 construct (apical, +32 ± 25%, P = 0.27; basal, +30 ± 30%, P = 0.18 compared to control, Kir2.1-transfected nonexposed neurons; fig. 6). In contrast, propofol still induced a significant decrease in the density of spines in neighboring nontransfected EGFP− neurons of the same animals (apical, −55 ± 16%, P = 0.004; basal, −46 ± 5%, P = 0.009 compared to nontransfected nonexposed control cells).These data strongly support the idea that decreasing neuronal excitability is sufficient to prevent the propofol-induced decrease in dendritic spine density in the early postnatal period.

We and others have shown that early exposure to general anesthetics can rapidly alter the number of dendritic spines and synaptic contacts in various cortical regions in a manner that is persistent and qualitatively dependent on the stage of brain development at the time of exposure,^8–12,35  as reviewed by Vutskits.¹  Most general anesthetics, notably propofol, act as allosteric potentiators of GABA_A receptors.¹³  The current results suggest a fundamental developmental mechanism whereby the up-regulation of the multifunctional neuron-specific KCC isoform KCC2,¹⁵  acting as a Cl⁻-extruding K-Cl cotransporter, controls the age-dependent effects of propofol anesthesia on dendritic spines in developing cortical principal neurons by setting the driving force for GABA_A receptor-mediated currents. We demonstrate that the propofol anesthesia–induced decrease in dendritic spine density at PND 10 is prevented in neurons precociously expressing KCC2. In contrast, such an effect is not seen in neighboring control neurons or neurons expressing a transport-inactive KCC2 variant.

Given the ion transport–independent roles of KCC2 during neural circuitry development,^15,19  it was important to determine whether the effects of KCC2 expression on propofol-induced dendritic spine responses in pyramidal neurons are due to the ion transport function of KCC2 and, thereby, to direct changes in membrane excitability. Two independent experimental approaches indicate that this is the case. First, we utilized in utero electroporation of KCC2-ΔNTD, a KCC2 construct that lacks the KCl cotransport function but retains its interaction with the actin cytoskeleton to promote spinogenesis.^5,17  While this construct increased dendritic spine densities to values comparable to those found with KCC2-FL, it did not prevent the propofol-induced dendritic spine loss. As a second approach, we used in utero electroporation to overexpress the inward-rectifier K⁺ channel Kir2.1. This manipulation has been shown to decrease excitability of SSC layer 2/3 pyramidal neurons by hyperpolarizing their resting membrane potential.^25,34  In line with our data on precocious expression of KCC2, we found that overexpression of Kir2.1 also prevented the propofol-induced decrease in the density of dendritic spines at PND 10. Notably, Kir2.1 overexpression per se had no effect on spine density or on spine head diameter distribution. Interestingly, this manipulation, however, enabled propofol anesthesia at PND 10 to facilitate induction of spines with large diameter, while the number of filopodia-like spines was concomitantly reduced. This suggests that the increase in the depolarizing driving force for GABA_A receptor-mediated currents, which would be expected to result from constitutive hyperpolarizing shift in the resting membrane potential by Kir2.1, in a context-dependent manner, can also drive spine maturation.

There is a growing concern regarding a possible association between early-life anesthesia exposure and subsequently impaired long-term neurocognitive outcome.^1,3,4  The fact that anesthetics can induce age-dependent morphofunctional changes in the developing neural circuitry provides a plausible underlying biologic mechanism.¹  Indeed, mounting evidence indicates that perturbations in neural network development can give rise to permanently impaired neurocognitive function and even to neuropsychiatric disorders.³⁶  These observations strongly suggest a generalized vulnerability of immature neural circuitry to anesthesia exposure in the developing brain. Here, it is relevant to point out that enhancers of GABA_A receptor function are also regularly administered in attempts to suppress seizures in neonates and infants.³⁷ 

In the human supraspinal central nervous system, robust up-regulation of KCC2 starts at the beginning of the third trimester of pregnancy and continues during the first year of life, which is roughly comparable to the period from birth to the first few postnatal weeks in rats and mice.^38–40  Notably, a qualitative change in neocortical GABA_A receptor–mediated signaling is most likely to take place at near-term birth, as evidenced by detailed comparison of rat and human developmental electroencephalogram characteristics and KCC2 expression patterns.^38,40  Thus, propofol anesthesia is likely to affect different targets downstream of GABA_A receptors in a preterm baby when compared to a baby at full term or a more mature infant. Future experiments are needed to determine whether and how these developmental stage–specific effects will influence long-term neurocognitive outcome.

In conclusion, our results provide novel mechanistic insight into how neuronal excitability determines synaptic plasticity in response to GABA_A-enhancing general anesthetics. This mechanism may play a role during neural circuitry assembly not only in the developing brain but also in older children and adults after neuronal trauma, when GABA_A receptor–mediated signaling can redisplay fetal-like excitatory actions.^15,41 

The authors thank Michèle Brunet, University Hospitals of Geneva, Geneva, Switzerland, Merle Kampura, M.Sc., University of Helsinki, Helsinki, Finland, and Mairi Kuris, B.Sc., University of Helsinki, for excellent technical assistance, and Eva Ruusuvuori, Ph.D., University of Helsinki, Patricia Seja, Ph.D., University of Helsinki, and Mari Virtanen, Ph.D., University of Geneva Medical School, Geneva, Switzerland, and University of Helsinki, for comments on the manuscript.

Supported by grant No. 31003A-130625 from the Swiss National Science Foundation, Berne, Switzerland (to Dr. Vutskits), ERA-NET NEURON II CIPRESS from the Academy of Finland, Helsinki, Finland (to Dr. Kaila), the Jane and Aatos Erkko Foundation, Helsinki, Finland (to Dr. Kaila), and the Emil Aaltonen Foundation, Tampere, Finland (to Dr. Puskarjov).

The authors declare no competing interests.