In the rodent primary visual cortex, maturation of GABA inhibitory circuitry is regulated by visual input and contributes to the onset and progression of ocular dominance (OD) plasticity. Cortical inhibitory circuitry consists of diverse groups of GABAergic interneurons, which display distinct physiological properties and connectivity patterns. Whether different classes of interneurons mature with similar or distinct trajectories and how their maturation profiles relate to experience dependent development are not well understood. We used green fluorescent protein reporter lines to study the maturation of two broad classes of cortical interneurons: parvalbumin-expressing (PV) cells, which are fast spiking and innervate the soma and proximal dendrites, and somatostatin-expressing (SOM) cells, which are regular spiking and target more distal dendrites. Both cell types demonstrate extensive physiological maturation, but with distinct trajectories, from eye opening to the peak of OD plasticity. Typical fast-spiking characteristics of PV cells became enhanced, and synaptic signaling from PV to pyramidal neurons became faster. SOM cells demonstrated a large increase in input resistance and a depolarization of resting membrane potential, resulting in increased excitability. While the substantial maturation of PV cells is consistent with the importance of this source of inhibition in triggering OD plasticity, the significant increase in SOM cell excitability suggests that dendrite-targeted inhibition may also play a role in OD plasticity. More generally, these results underscore the necessity of cell type-based analysis and demonstrate that distinct classes of cortical interneurons have markedly different developmental profiles, which may contribute to the progressive emergence of distinct functional properties of cortical circuits.
- γ-aminobutyric acid
during a brief postnatal period, the closure of one eye can permanently shift the response property of neurons in the primary visual cortex (V1) to favor inputs from the open eye, i.e., ocular dominance (OD) shift (Hubel and Wiesel 1970). OD plasticity has been a premier model to study how sensory experience shapes the development of cortical circuits during a critical period. To shift eye preference during monocular deprivation, visual cortical neurons must first be able to detect the imbalance of converging visual inputs, relayed to the cortex as altered spiking patterns in thalamic axons, before they can engage a cascade of molecular, cellular, and circuitry mechanisms to weaken the deprived eye-associated inputs and strengthen the open eye-associated inputs (Hensch 2005). GABAergic interneurons are crucial in shaping and detecting the precise spatiotemporal patterns of electrical signaling in cortical circuits and in regulating synaptic plasticity (Markram et al. 2004). Accumulating evidence has indicated that proper functioning of GABAergic inhibitory neurons within V1 are critical to establish the necessary physiological milieu that enables OD plasticity. Mice lacking the synaptic isoform of the GABA-synthetic enzyme GAD65 show no OD plasticity, a deficit that can be rescued by cortical infusion of a GABAA receptor agonist (Hensch et al. 1998). In addition, genetic (Huang et al. 1999) and pharmacological (Fagiolini and Hensch 2000) enhancement of the maturation and function of GABA inhibition in V1 induces a precocious critical period. However, the cellular and circuitry mechanisms by which the maturation of cortical inhibition promotes OD plasticity are not well understood.
Synaptic inhibition in the neocortex is achieved by diverse groups of interneurons, which mediate GABA transmission at discrete spatial and temporal niches during circuit operation and demonstrate distinct physiological properties and connectivity patterns (Markram et al. 2004; Burkhalter 2008). Although our understanding of this diversity is far from complete, previous studies have established a major dichotomy in the inhibitory control of pyramidal neurons. Interneurons that innervate pyramidal cell dendrites are responsible for controlling the efficacy and plasticity of glutamatergic inputs that terminate in the same dendritic domain (Miles et al. 1996; Tamas et al. 1997). On the other hand, interneurons targeting the perisomatic region control action potential (AP) generation, timing, and synchrony in pyramidal cell populations (Cobb et al. 1995; Miles et al. 1996). Whether or how these two major sources of inhibition differentially engage molecular and cellular plasticity mechanisms and contribute to OD plasticity is unclear. Previous studies have focused on parvalbumin-expressing (PV) interneurons, which are fast spiking and innervate the perisomatic region of pyramidal neurons (Miles et al. 1996; Tamas et al. 1997). For example, the morphological maturation of perisomatic innervation from PV cells correlates with the timing of the critical period (Chattopadhyaya et al. 2004). In addition, PV interneurons signal through α1-subunit-containing GABAA receptors, and inhibition through these receptors appears critical for OD plasticity (Fagiolini et al. 2004). Furthermore, the homeoprotein orthodenticle homolog 2, which is able to trigger plasticity when transported from the retina to V1, is prominently taken up by PV cells (Sugiyama et al. 2008). These results indicate an important role of PV interneurons and perisomatic inhibition in the onset of OD plasticity. On the other hand, the role of dendrite-targeted inhibition through somatostatin-expressing (SOM) interneurons has not been well studied. OD plasticity ultimately involves structural rewiring of excitatory synapses onto the dendritic spines of pyramidal neurons (Oray et al. 2004; Mataga et al. 2004). Given the powerful role of dendrite-targeted inhibition in controlling synaptic integration (Miles et al. 1996; Perez-Garci et al. 2006), dendritic Ca2+ spikes (Murayama et al. 2009), plasticity (Ballard et al. 2009), and learning (Collinson et al. 2002; Maubach 2003), it is likely that this source of inhibition also contributes to aspects of OD plasticity.
A necessary step toward further understanding the role of these two major classes of GABAergic interneurons in OD plasticity is a characterization of their functional maturation during the critical period. It has been known for decades that maturation of the GABAergic system in the rodent visual cortex follows a postnatal time course (Luhmann and Prince 1991). However, previous studies have primarily used methods such as spontaneous, miniature, or field-evoked inhibitory postsynaptic current (IPSC) recordings (Bosman et al. 2002, 2005; Heinen et al. 2004; Morales et al. 2002), which cannot distinguish the synaptic sources of inhibition. To date, the physiological maturation of PV interneurons during the critical period has not been fully characterized, and the maturation of SOM interneurons has not been explored. Using two green fluorescent protein (GFP) reporter mouse lines, here, we present the first comparison of the maturation of inhibition provided by PV and SOM interneurons before and during the critical period for OD plasticity. We found that both classes of interneurons demonstrate substantial physiological maturation from eye opening to the peak of OD plasticity, but with distinct trajectories. While PV cells demonstrated significant maturation in their characteristic fast signaling properties, which plateaued after the onset of critical period, SOM cells demonstrated a profound and steady increase in their excitability, which continued to the peak of OD plasticity.
To identify PV and SOM cells, we used two transgenic mouse lines, B13 and GIN, respectively, that have been previously used to identify these classes of neurons. The B13 line (Dumitriu et al. 2007; Goldberg et al. 2008; Ango et al. 2008; Daw et al. 2010) expresses enhanced GFP (EGFP) driven by the Pv gene. EGFP in the B13 line is expressed selectively in ∼50% of PV cells in the neocortex (Dumitriu et al. 2007). The GIN line expresses EGFP driven by the Gad1 promoter (Oliva et al. 2000), and EGFP is restricted to a subclass of SOM neurons (Oliva et al. 2000; Ma et al. 2006; Halabisky et al. 2006). EGFP in the GIN line is expressed in SOM neurons in both superficial and deep layers of the neocortex (Oliva et al. 2000; Ma et al. 2006), labeling approximately one-third of SOM cells in layer II/III (Ma et al. 2006). Mice were treated in accordance with Cold Spring Harbor Laboratory guidelines on animal husbandry and care/welfare. Experiments were performed on animals between 15 and 30 days after birth [postnatal day (P)15 and P30], as indicated.
Acute brain slices were prepared at the appropriate ages. Animals were deeply anesthetized with avertin (tribromoethanol in amyl hydrate, intraperitoneal injection, 0.2 ml/g), and decapitated. Brains were rapidly removed and placed into ice-cold, oxygenated cutting solution, containing (in mM) 110 choline chloride, 2.5 KCl, 25 NaHCO3, 1.25 NaH2PO4, 0.5 CaCl2, 7 MgCl2, 25 glucose, 11.6 ascorbic acid, and 3.1 pyruvic acid bubbled with 95% O2-5% CO2. The anterior one-third of the brain and the posterior section containing the cerebellum were removed with coronal cuts. The brains were then glued to the slicing block, anterior face down. Slices were prepared in the choline-based cutting solution on a Microm HM650V (Walldorf, Germany). Coronal slices contained V1 and were 350 μm thick. Slices were transferred to artificial cerebrospinal fluid (aCSF) and incubated at 32–34°C for at least 30 min. aCSF contained (in mM) 126 NaCl, 2.5 KCl, 25 NaHCO3, 14 glucose, 1.25 NaH2PO4, 1 MgSO4, and 2 CaCl2 bubbled with 95% O2-5% CO2 to pH 7.4. For recording, slices were transferred to a recording chamber continuously perfused with oxygenated aCSF and maintained at 28–30°C.
GFP-positive cells were identified in layer II/III of the visual cortex and patched with a recording pipette containing 0.2% biocytin. These slices were then incubated overnight at 4°C in 4% paraformaldehyde in PBS (pH 7.4). After fixation, slices were rinsed in PBS (3 times for 5 min) and then incubated overnight in Alexa fluor 568-conjugated streptavidin (1:1,000, Invitrogen) with 0.3% Triton X-1000 in PBS. Slices were then rinsed in PBS (3 times for 5 min) and mounted in Vectashield mounting medium (Vector Labs). Fluorescently labeled neurons were imaged using a Zeiss LSM 510 confocal microscope and reconstructed using Neurolucida (MicroBrightField).
All recordings were performed in layer II/III in coronal cut slices and were restricted to V1. Dual whole cell recordings were performed on a two-channel Multiclamp 700B amplifier (Molecular Devices, Sunnyvale, CA). For paired and single whole cell recordings, interneurons were identified by GFP expression under a narrow-band GFP filter set (Chroma Technology, Brattleboro, VT) in an Axioskop FS2 upright microscope (Zeiss, Thornwood, NY) with an ORCA-ER camera (Hamamatsu, Hamamatsu City, Japan). The GFP-positive cell was subsequently visualized with differential interference contrast (Zeiss). For paired recordings, a nearby pyramidal neuron (<50 μm) was visually identified by a triangular soma with a distinct apical dendrite oriented toward the pial surface. Patch pipettes were pulled from borosilicate glass on a Flaming-Brown micropipette puller (Sutter Instruments, Novato, CA) and had a resistance of 2–4 MΩ. In both paired and single cell recordings, the intracellular solution for the interneuron contained (in mM) 130 K-gluconate, 6 KCl, 2 MgCl2, 2 EGTA, 2 HEPES, 2.5 Na-ATP, 0.5 Na-GTP, and 10 Na-phosphocreatine (pH 7.25, 285–295 mosM). The interneuron recording was performed in current clamp for all experiments. For paired recordings, a high internal Cl− concentration was used to magnify IPSC responses in pyramidal neurons to aid with analysis. The pyramidal neuron internal solution contained (in mM) 65 K-gluconate, 65 KCl, 2 MgCl2, 2 EGTA, 2 HEPES, 2.5 Na-ATP, 0.5 Na-GTP, and 10 Na-phosphocreatine (pH 7.25, 285–295 mosM). The postsynaptic pyramidal neuron was held in voltage clamp at −75 mV. Zolpidem was used to assess GABAA receptor α-subunit composition and was purchased from Tocris (Ellisville, MO).
To assess intrinsic properties, interneurons were stimulated with increasing 1-s-long current steps. PV cells were assessed with 50-pA steps starting from −200 pA increasing up to +700 pA, and SOM cells were assessed with 20-pA steps starting from −100 pA increasing up to +200 pA. Maximum current injections of +700 and +200 pA for PV and SOM cells, respectively, were chosen because these were the highest levels that would reliably not produce spike inactivation. For paired recordings, the presynaptic interneuron was stimulated with a brief suprathreshold current injection (0.8–1.2 nA, 3–5 ms). With the exception of short-term plasticity experiments, 1 spike was initiated per trial, with an intertrial interval (ITI) of 5 s. For analysis, 25–50 trials were averaged. For short-term plasticity experiments, presynaptic interneurons were stimulated at 20 Hz, with an ITI of 10 s.
Analysis and Electrophysiological Parameters
All analysis was performed offline with the ClampFit 9.0 (Molecular Devices) program.
Only stable recordings were included for analysis. Data were discarded if series resistance was >25 MΩ or varied by >25%, if resting membrane potential (VM) was greater than −50 mV several minutes after break in, or if membrane resistance (RM) was <50 MΩ. Pairs of neurons were considered synaptically connected if the averaged IPSC was >2 pA.
The membrane time constant (τM; in ms) was determined by a monoexponential fit of the hyperpolarizing voltage response (∼10 mV) to a suitable current injection, from resting level to minimum point of sag, if present.
RM (in MΩ) was the slope of the linear portion of voltage responses to a series of negative current steps (current-voltage response curve).
Membrane capacitance (CM; in pF) was determined from the following equation: CM = τM/RM.
VM (in mV) was the stable membrane potential, as determined with no current injection a few minutes after the seal was broken.
AP threshold (VT; in mV) was membrane potential when the rate of rise equaled 5 V/s, as measured in response to the smallest current step able to evoke a spike.
The excitability index (EI; in pA) incorporated multiple properties of a cell to estimate intrinsic excitability, as determined from the following equation: EI = (VT − VM)/RM.
Spike half-width (in ms) was the width measured at half-amplitude, between VT and the peak of the AP, measured in response to the smallest current step able to evoke a spike.
Afterhyperpolarization (AHP) amplitude (in mV) was the voltage difference between VT and the most negative point reached after an AP, measured in response to the smallest current step able to evoke a spike.
Spike frequency (in Hz) was the inverse of the first interspike interval (ISI).
Frequency-current slope (in Hz/pA) was the slope of the linear portion of the frequency-current response curve using the initial spike frequency.
Spike frequency adaptation (dimensionless) was the ratio of the last ISI to the fourth ISI. It was assessed at +700 pA for PV cells and +200 pA for SOM cells.
Amplitude (in pA) was measured from baseline (average of 5 ms) to the peak of the averaged IPSC.
The rise time of averaged IPSCs (RTAvg; in ms) was the time from 10–90% of the rising phase of the averaged IPSC.
The rise time of individual IPSCs (RTIndv; in ms) was the average of the 10–90% rise time determined from at least 10 individual IPSC responses that could be resolved from noise. This measure was used to avoid potential error from jitter in presynaptic spiking or IPSC delay but could only be measured for PV→pyramidal (PV→Py) IPSCs due to the very small amplitude of SOM→pyramidal (SOM→Py) IPSCs.
Decay time (in ms) was determined by a monoexponential fit of the decaying phase of averaged IPSCs.
The paired-pulse ratio (PPR; dimensionless) was the ratio of the second IPSC amplitude to the first IPSC amplitude, stimulated at 20 Hz.
All data are reported as means ± SD. Significance was determined by one-way ANOVA with a post hoc Tukey's test for all data with greater than two comparison groups. For comparison of two groups, two-tailed unpaired Student's t-test was used. For comparison of distributions, the Kolmogorov-Smirnov test was used. The significance level was set at P < 0.05.
Targeted whole cell current-clamp recordings were made from PV and SOM neurons to analyze membrane and spiking properties. Dual whole cell recordings were made from PV→Py and SOM→Py pairs to analyze synaptic properties. Three age groups were chosen to study the maturation of inhibition in relation to the onset of OD plasticity: the precritical period after eye opening (P15–P17), early in the critical period (P22–P24), and the peak of the critical period (P28–P30) (Gordon and Stryker 1996). All statistical comparisons were across these three age groups unless otherwise noted.
Morphology of GFP-Labeled Cells
Morphological reconstructions were performed to determine which types of cells were labeled in the B13 and GIN lines and to confirm previous reports (see methods). Three cells each in animals from the young age group (P15–P17) and mature age group (P28–P30) in the B13 and GIN lines (total of 12 cells) were reconstructed (Fig. 1). Reconstructed cells in the B13 line (Fig. 1A) generally had dense local axonal arborization restricted to layer II/III, with occasional horizontal or vertical collaterals. This is consistent with identity as nest or small basket cells but not with large basket cells (Wang et al. 2002; Markram et al. 2004). In the GIN line (Fig. 1B), all reconstructed cells had axonal projections up toward the pial surface and extensive arborization within layer I. This supports previous work identifying GIN line cells as Martinotti (MN) cells (Ma et al. 2006).
Cell Type-Specific Maturation of Membrane Properties
PV and SOM neuron membrane properties demonstrated distinct developmental profiles over the period studied (Fig. 2). PV cells showed a decrease in τM from eye opening (P15–P17, 12.2 ± 5.3 ms) to the early critical period (P22–P24, 8.2 ± 3.2 ms, P < 0.05), when it reached a steady state (Fig. 2A). On the other hand, in SOM cells, τM increased between eye opening (P15–P17, 29.3 ± 13.0 ms) and the peak of the critical period (P28–P30, 43.2 ± 12.3 ms, P < 0.05; Fig. 2B).
The mechanisms of the changes in τM were different in PV and SOM cells (Fig. 3). In PV cells, RM did not change significantly over the period studied (Fig. 3A); therefore, the reduction in τM of PV cells may be due to a decrease in CM between P15–P17 (87.7 ± 22.5 pF) and P22–P24 (67.7 ± 14.2 pF, P < 0.01; Fig. 3A, 2). In contrast to PV cells, SOM cells showed a substantial RM increase across all three ages studied (P15–P17: 299 ± 99 MΩ and P28–P30: 445 ± 79 MΩ, P < 0.01; Fig. 3B), but CM did not change significantly over the developmental time period tested (Fig. 3B, 2). This increase in RM likely underlies the developmental increase in τM observed in SOM cells.
The VM of PV cells was stable over the period studied (Fig. 4A, 1). The VM of SOM cells was stable between P15–P17 and P22–P24 but became significantly more depolarized before the peak of the critical period at P28–P30 (P22–P24: −61.3 ± 7.7 mV and P28–P30: −55.9 ± 5.1 mV, P < 0.05; Fig. 4B, 1). Since VM was measured after the experimental internal solution would have replaced the natural cellular contents, this observed depolarization may reflect a developmental reduction of K+ permeability at rest rather than a change in ion concentrations. A reduction of open K+ channels in SOM cells may explain both the developmental depolarization of VM as well as the developmental increase in RM (Cameron et al. 2000).
Two approaches were used to quantify cell excitability: 1) cumulative distribution of the minimum current injection required for at least one spike to occur (100-pA intervals for PV cells and 20 pA intervals for SOM cells) and 2) EI [expressed in pA and determined by the equation EI = (VT − VM)/RM, as described above]. A lower EI would reflect a more excitable cell, and vice versa.
No changes in cell excitability were observed in PV cells; however, SOM cells demonstrated a substantial increase in excitability with maturation. The current required to evoke spikes in PV cells did not change with development (Kolmogorov-Smirnov test, P = 0.89; Fig. 4A, 2), nor did EI change significantly over the period studied (Fig. 4A, 2). In SOM cells, likely due to increased RM and more depolarized VM, a significant shift to the left in the distribution of current required to evoke spikes was observed with maturation (Kolmogorov-Smirnov test, P < 0.01; Fig. 4B, 2), along with a twofold reduction of EI (P15–P17: 53.6 ± 25.3 pA and P28–P30: 25.9 ± 7.5 pA, P < 0.01; Fig. 4B, 2).
It should be noted that with the EI measure in mature mice (P28–P30), SOM cells were over 10-fold more excitable than PV cells (Fig. 4A, 2 and B, 2). At all ages, SOM cells, compared with PV cells, demonstrated more hyperpolarized VT (no developmental change; data not shown), larger RM, and more depolarized VM. In other words, all three factors used to determine EI would make SOM cells more excitable than PV cells, suggesting that the SOM source of inhibition may be more frequently or more readily engaged. Additionally, these results are consistent with the description of SOM cells as “low threshold spiking” interneurons (Gibson et al. 1999).
Cell Type-Specific Maturation of Spiking Properties
PV cells showed the fast-spiking properties (Kawaguchi and Kubota 1997) described by McCormick et al. (1985). These include low spike frequency adaptation, large fast AHP, narrow AP half-width, and high spiking frequency. These last three features showed significant maturational changes, leading to more typical fast-spiking characteristics, in PV cells.
AP morphology substantially changed with maturation in PV cells but not in SOM cells (Fig. 5). The AP half-width of PV cells decreased between P15–P17 (0.80 ± 0.14 ms) and P22–P24 (0.60 ± 0.31 ms, P < 0.05; Fig. 5A), whereas AHP became stronger between P22–P24 (17.1 ± 4.4 mV) and P28–P30 (21.0 ± 2.7 mV, P < 0.05; Fig. 5A). The development of larger AHP was the only maturational feature of PV cells that did not show changes between P15–P17 and P22–P24. Both AP half-width and AHP of SOM cells were stable over the time period studied (Fig. 5B).
Plots of initial spike frequency versus current injected (frequency-current curves) demonstrated maturational changes for both PV and SOM cells (Fig. 6). In both cell types, the frequency-current curve shifted upward during development. However, in PV cells, the frequency response to a given current input increased with age only for larger current steps (P < 0.05 for 500 pA and P < 0.01 for 600 and 700 pA; Fig. 6A, 2). In SOM cells, higher spike frequencies in older animals were observed across the range of current inputs (P < 0.01 for 40, 80, 120, 160, and 200 pA; Fig. 6B, 2). To determine if changes in RM may contribute to the shift of frequency-current curves, we adjusted these data for developmental differences in RM by multiplying the current input value by the average RM for the specific age group and cell type. This gave an estimation of the predicted membrane potential response (spike activation prevents the full membrane voltage response from occurring) and was expressed in millivolts. Since RM of PV cells did not change with development, adjusting for RM in this cell type had a minimal impact on the relative position of frequency-current curves (Fig. 6A, 3). SOM cells, however, showed a substantial change in RM with development. Adjusting the frequency-current curves of SOM cells for RM eliminated the differences in the frequency response and led to overlapping curves for the different age groups (Fig. 6B, 3). This result suggests that increased RM in SOM cells can largely account for developmental changes in frequency-current response features. Quantification of the frequency-current slope in PV cells showed the increased frequency response occurred between P15–P17 (0.240 ± 0.039 Hz/pA) and P22–P24 (0.429 ± 0.166 Hz/pA, P < 0.01; Fig. 6A, 4); however, in SOM cells, the frequency-current slope increased gradually between P15–P17 (0.457 ± 0.093 Hz/pA) and P28–P30 (0.558 ± 0.092 Hz/pA, P < 0.05; Fig. 6B, 4). Spike adaptation, which is typically low in fast-spiking cells (McCormick et al. 1985), was stable in PV cells over the period studied (Fig. 6A, 4) but increased gradually in SOM cells (P15–P17: 1.94 ± 0.75 and P28–P30: 3.70 ± 2.19, P < 0.05; Fig. 6B, 4).
The increased excitability of SOM cells, and the resultant impact on spiking characteristics, appears unique to this cell type, as it is not observed in pyramidal cells (Oswald and Reyes 2008; Desai et al. 2002) or PV cells (Doischer et al. 2008; Okaty et al. 2009). This may reflect increased involvement of this source of inhibition in more mature cortical circuits. The developmental profile of PV cells was consistent with previous studies (Doischer et al. 2008; Okaty et al. 2009; Kuhlman et al. 2010) and demonstrated the postnatal acquisition of fast-spiking characteristics (McCormick et al. 1985).
Source-Specific Maturation of Inhibitory Transmission
The maturational changes of intrinsic properties in PV and SOM cells determine when, if, and how AP firing occurs. But, the impact of cell spiking is, of course, dependent on communication with postsynaptic cells. Therefore, we decided to study the maturation of synaptic connections in PV→Py and SOM→Py pairs during the same developmental time period (Figs. 7 and 8).
Similar to PV cell intrinsic properties, PV→Py synaptic transmission showed significant maturation in numerous characteristics after eye opening (Fig. 7A). These developmental changes largely stabilized by the early critical period. In contrast, SOM→Py synaptic transmission did not demonstrate any significant changes over the time period studied (Fig. 7B). Synaptic features at both PV→Py and SOM→Py connections are quantified and compared in Fig. 8.
The kinetic properties of IPSCs became faster in PV→Py pairs. IPSC rise times decreased, mostly between P15–P17 (RTAvg: 0.77 ± 0.14 ms and RTIndv: 0.71 ± 0.14 ms) and P22–P24 (RTAvg: 0.65 ± 0.11 ms, P < 0.05; RTIndv: 0.57 ± 0.14 ms, P < 0.05). IPSC decay times also decreased and followed a similar time course (P15–P17: 12.68 ± 3.85 ms and P22–P24: 8.94 ± 1.41 ms, P < 0.01). IPSC amplitude showed a nonsignificant increase (amplitude varied substantially). All of these changes stabilized by P22–P24, with little change apparent between P22–P24 and P28–P30.
To determine if a shift in GABAA receptor subunit composition may underlie the maturational changes in IPSC kinetics, the α1-subunit-specific agonist zolpidem (Munakata et al. 1998) was applied (0.4 μM) (Ali and Thomson 2008) to some PV→Py paired recordings in young (P15–P17) and mature (>P22) slices. Numerous studies have demonstrated a shift in α-subunits from α2/3- to α1-subunits during postnatal maturation (Laurie et al. 1992; Fritschy et al. 1994; Heinen et al. 2004; Hashimoto et al. 2009). α1-Subunit-containing GABAA receptors have faster deactivation properties (Lavoie et al. 1997) and contribute to faster IPSC decay kinetics (Vicini et al. 2001; Bosman et al. 2005). In paired recordings, the application of zolpidem led to a significantly larger increase in IPSC half-width in mature (>P22: 29.3% ± 20.1) compared with young animals (P15–P17: 2.5% ± 6.1, P < 0.05 by Student's t-test; Fig. 7C). This suggests that a relative increase in α1-subunit-containing GABAA receptors contributes to the faster IPSC kinetics observed in older animals. The application of zolpidem did not consistently potentiate IPSC amplitude in any age group (data not shown), likely due to the saturation of GABAA receptors on layer II/III pyramidal neurons (Hajos et al. 2000).
In contrast to PV→Py pairs, SOM→Py connections did not show any maturational changes in IPSC properties. The lack of developmental changes in IPSC kinetics is likely due to the use of GABAA receptor subtypes different from those of perisomatic targeting interneurons; dendritic targeting interneurons signal primarily though α5-subunit-containing GABAA receptors (Ali and Thomson 2008). Although expression of α5-subunits decreases with age (Heinen et al. 2004; Yu et al. 2006), these do not appear to be replaced with a different α-subunit, and therefore kinetic properties would not be expected to change.
In addition to the features of individual IPSCs, short-term synaptic plasticity can also be a key determinant in the function of different types of synaptic connections (Reyes et al. 1998; Markram et al. 1998). For example, excitatory input to PV cells shows short-term depression, whereas excitatory input to SOM cells shows short-term facilitation, possibly resulting in differential engagement of these two sources of inhibition depending on network activity (Reyes et al. 1998). To assess the maturation of short-term plasticity at the output synapses of PV and SOM cells, we measured the PPR at 20 Hz in both PV→Py and SOM→Py pairs (Fig. 7D). Nearly every single paired recording, at both types of synapses and at all ages, demonstrated synaptic depression (one PV→Py pair in the P28–P30 age group was slightly facilitating, PPR = 1.02, but this was the only exception). A gradual increase in PPR occurred in PV→Py pairs from P15–P17 (0.65 ± 0.09) to P28–P30 (0.81 ± 0.16, P < 0.05). This is the only PV synaptic feature that did not appear to stabilize by the early critical period and may reflect an increased ability to sustain activity in the PV inhibitory network. The mechanism of developmental reduction in synaptic depression at these synapses is unclear and would not be explained by increased PV expression (Caillard et al. 2000; Muller et al. 2007). SOM→Py pairs did not demonstrate any changes in PPR.
The distinct developmental profiles of PV and SOM cells contribute to the physiological differences in these two sources of inhibition in the mature V1 (Fig. 8). The mature PV network, by inhibiting proximal cell regions via fast α1-subunit-containing GABAA receptors, is generally expected to provide fast synaptic inhibition. However, in young animals (P15–P17), the decay times of PV→Py and SOM→Py IPSCs were indistinguishable (Fig. 8C). Additionally, mature (P28–P30) PV→Py connections demonstrated less synaptic depression than SOM→Py connections, but this also was a feature that only appeared later in development as a result of maturation of PV synapses (Fig. 8C). Throughout the time period studied, PV→Py IPSCs, compared with SOM→Py IPSCs, had larger amplitudes and faster rise times.
Cortical inhibitory circuits consist of diverse classes of interneurons with distinct physiological properties and connectivity patterns (Miles et al. 1996; Kawaguchi and Kubota 1997; Markram et al. 2004; Burkhalter 2008). The maturation profiles of different cell classes and their differential regulation by experience may contribute to the progressive sharpening of functional properties of pyramidal neurons, yet most previous studies on the maturation of cortical inhibition (Bosman et al. 2002, 2005; Heinen et al. 2004; Morales et al. 2002) have not distinguished among different sources of synaptic inhibition. This is the first study, to our knowledge, that directly compares the maturation of two major classes of interneurons during the critical period in V1.
Maturation of Perisomatic Inhibition From PV Interneurons
During the onset phase of the critical period, both the intrinsic and synaptic properties of PV cells become significantly faster and more robust at multiple stages of signal propagation and transmission (Fig. 9). Indeed, the PV τM and AP waveform become faster, and the kinetics of unitary IPSCs become more rapid. The faster membrane properties (lower τM) result from a reduction in CM, which would allow for more rapid integration with enhanced precision in response to synaptic inputs, effectively creating a highly stringent coincidence detector. The rapid AP, along with fast and strong AHP, in PV cells might result from the regulation of ion channel expression and allows PV cells to fire at high frequencies with minimal spike frequency adaptation (Erisir et al. 1999). Additionally, PV cells have the unique property of electrical coupling with one another (Galarreta and Hestrin 1999; Gibson et al. 1999; Beierlein et al. 2000), which could also contribute to the biophysical features of these cells (Veruki et al. 2010).
The more rapid IPSC kinetics (i.e., faster decay) at PV→Py synapses seem to result from increased α1-subunit content of postsynaptic GABAA receptors, as evidenced by the enhanced sensitivity to a low concentration of zolpidem. This is consistent with results from multiple brain regions and species showing high expression of α2- or α3-subunits at early stages, with upregulation of α1-subunits at later times (Fritschy et al. 1994; Bosman et al. 2002; Heinen et al. 2004; Vicini et al. 2001; Hashimoto et al. 2009). Consequently, miniature IPSC or spontaneous IPSC recordings, which do not differentiate between presynaptic sources, demonstrate the development of faster inhibitory signaling (Bosman et al. 2002, 2005; Heinen et al. 2004; Vicini et al. 2001; Kotak et al. 2008; Hashimoto et al. 2009). This developmental switch in subunit composition and IPSC properties appears to be a highly conserved, cell type-specific mechanism to sharpen inhibitory transmission during GABA circuit maturation in different cortical areas and species. The concerted changes in synaptic and intrinsic properties, similar to the maturation of basket cells in the dentate gyrus (Doischer et al. 2008), essentially convert V1 PV cells into fast signaling devices during the onset of OD plasticity.
Previous developmental studies on PV cells in the dentate gyrus (Doischer et al. 2008) and somatosensory cortex (Okaty et al. 2009) included much younger neonatal ages and showed, not surprisingly, more profound maturation of some characteristics, including the strength and reliability of synaptic transmission. Our study focused around the critical period and revealed that faster signaling appears to be the main, if not the only, property of PV cells that shows significant maturation, possibly contributing to the onset of OD plasticity.
During OD plasticity, the rearrangement of synaptic connections in V1 is guided by altered visual inputs from the open and closed eyes, which are encoded in the temporal spiking patterns of thalamic axons. The firing of PV cells can precisely reflect the spiking patterns of excitatory inputs, and groups of synaptic and electrically connected PV cells (Galarreta and Hestrin 1999; Gibson et al. 1999; Beierlein et al. 2000) can discriminate synchronous versus less synchronous inputs (e.g., those from the open or closed eye) with millisecond precision (Galareta and Hestrin 2001). Furthermore, mature PV cells are characterized by their “fast in and fast out” signal transmission (Jonas et al. 2004) and thus are highly effective in converting the pattern of excitatory input into well-timed inhibitory output to the perisomatic region, thereby controlling the spiking pattern of innervated pyramidal neurons. PV fast-spiking interneurons therefore appear ideally suited to detect and transmit precise input spiking patterns in V1, a necessary step in the engagement of plasticity mechanisms (Huang and Di Cristo 2008; Hensch 2005), e.g., spike timing-dependent plasticity, which may already be in place at that time (Kuhlman et al. 2010). Here, we have shown that the intrinsic and synaptic properties of V1 PV cells remain immature after eye opening but are significantly sharpened during the onset of OD plasticity. These results suggest that PV cells may develop improved ability to detect and/or transmit changes in input spiking patterns, e.g., from an open versus closed eye. Therefore, the maturation of the PV interneuron network could potentially enable V1 neurons to engage the plasticity mechanisms involved in OD plasticity.
Maturation of Dendrite-Targeted Inhibition From Somatostatin-Expressing Interneurons
SOM interneurons likely include several subgroups (Kawaguchi and Shindou 1998; Wang et al. 2004; Ma et al. 2006). The GIN line used in this study (Oliva et al. 2000) labels subsets of SOM cells, especially MN cells (Fig. 1B) (Ma et al. 2006), which are characterized by their ascending axonal projections, with extensive branching in layer I, and slowly accommodating firing pattern with spikes triggered at low threshold. Importantly, MN cells mediate frequency-dependent disynaptic inhibition (FDDI) among nearby pyramidal cells (Silberberg and Markram 2007), and this is a generic circuit motif prevalent across cortical areas and layers (Berger et al. 2009). FDDI has been postulated to gate synaptic plasticity in distal dendrites (Buchanan and Sjostrom 2009). Furthermore, the dendritic encoding of sensory stimuli in pyramidal neurons is highly sensitive to inhibitory control from MN cells (Murayama et al. 2009). However, the maturation of MN cells and other SOM cells has not been examined, especially during the critical period in V1.
We found that the most apparent developmental changes in SOM interneurons are in cellular biophysical properties: a substantial increase in RM underlies a slower τM in mature cells as well as a steepened frequency-current response curve (Fig. 9). Since CM remained constant, the increase of RM likely results from decreased density of leak channels (Cameron et al. 2000). Increased RM, along with depolarized VM, leads to substantially higher excitability of these cells. Additionally, slower membrane properties would increase the time period over which SOM cells can integrate and respond to synaptic inputs.
These maturation profiles of SOM cells are in sharp contrast to those of pyramidal neurons and PV cells. Postnatal development of layer II/III pyramidal cells leads to lower input resistance (Luhmann and Prince 1991; Desai et al. 2002; Kuhlman et al. 2010) and hyperpolarized VM (Luhmann and Prince 1991; Desai et al. 2002; M. S. Lazarus and Z. J. Huang, unpublished observations), which should reduce cell excitability. The substantial increase of excitability of SOM cells may reflect a stronger engagement of this type of inhibition in V1 circuits during the critical period. Compared with PV cells, we noted that while the maturation of fast signaling in PV cells plateaus early in critical period, the increase of SOM excitability continues to the peak of OD plasticity.
Previous studies on the function of inhibition in OD plasticity have mainly focused on PV cells and their role in triggering the onset of the critical period. Whether and how GABAergic inhibition influences the execution of plasticity mechanisms and promotes the progression of OD shift after the critical period onset is poorly understood. In particular, the functional role of dendrite-targeted inhibition and SOM cells is unexplored. OD plasticity ultimately involves physical rewiring of excitatory synapses onto the dendritic spines of pyramidal neurons (Hofer et al. 2009). A major question remains as to how inputs representing the closed and open eyes compete along the apical dendrites for synaptic connections; in particular, it is unclear whether and how dendrite-targeted inhibition contributes to such competition. The intrinsic and membrane properties of mature SOM cells seem ideally suited to represent the strength of excitatory input and to convert it proportionally into inhibitory outputs, which could act to suppress competing inputs along pyramidal cell dendrites. Our results indicate that the maturational increase in the excitability of SOM cells correlates with the progression of OD plasticity, implying a stronger engagement of dendritic inhibition after the onset of the critical period. Genetic manipulation of SOM cells (e.g., using a SOM cell-specific Cre mouse line) offers the opportunity to directly test their function in OD plasticity.
By studying two distinct classes of visual cortical interneurons during the critical period of OD plasticity, we discovered their distinct maturation profiles. The most prominent feature in fast-spiking, perisomatic targeting PV interneurons is the maturation of fast signaling at the onset of the critical period. The most prominent feature in regular-spiking, dendritic targeting SOM interneurons is the maturation of their excitability and therefore the strength of inhibition, which continues to the peak of OD plasticity. In addition, the maturation of PV and SOM cells seem to exhibit a developmentally enhanced dichotomy: PV cells appear increasingly tuned to detect precisely timed inputs, whereas SOM cells develop stronger ability to detect and thus represent an overall level of input activity. Although our present results are descriptive and correlative by nature, they demonstrate that distinct classes of cortical interneurons have different developmental trajectories, which may progressively sharpen functional and plasticity properties in cortical circuits. This finding underscores the necessity of cell type-specific analysis when studying the development of cortical microcircuits and has general implications in other cortical areas and plasticity paradigms.
This work was supported by National Institute of Mental Health Predoctoral Fellowship F30-MH-087036-02 (to M. S. Lazarus) and the Marie and Charles Robertson Fund at Cold Spring Harbor Laboratory (to Z. J. Huang).
No conflicts of interest, financial or otherwise, are declared by the author(s).
The authors thank Dr. Jiangteng Lu and Dr. Arianna Maffei for helpful comments on the manuscript.
- Copyright © 2011 the American Physiological Society