Spatial distribution of inhibitory synaptic connections during development of ferret primary visual cortex

Abstract

Intracortical inhibition in the primary visual cortex plays an important role in creating properties like orientation and direction selectivity. However, the development of the spatial pattern of inhibitory connections is largely unexplored. This was investigated in the present study. Tangential slices of layers 2/3 of ferret striate cortex were prepared for whole-cell patch clamp recordings, and presynaptic inhibitory inputs to pyramidal neurons were scanned by local photolysis of Nmoc-caged glutamate. Inhibitory synaptic currents (IPSCs) were first detected around postnatal day (P) 17. They originated locally around the recorded cells. Both the number and the total areas supplying the inhibitory inputs increased thereafter and peaked at the time around and shortly after eye opening (P29–37). A refinement period then followed in which the areas providing the majority of inhibitory inputs shrank from 600 µm around the recorded neurons to 200–300 µm in more mature animals (≥P38). The amplitude of IPSCs increased progressively with increasing age. Long-range inhibitory inputs (>600 µm) were present around eye opening and they often developed into a clustered patchy pattern in more mature animals (≥P38). In summary, our results show a refinement and clustering in the spatial pattern of inhibitory connections during postnatal development of ferret visual cortex.

Introduction

The primary visual cortex is characterized by a columnar architecture in which neurons preferring similar stimulus features cluster to form functional domains (Hubel and Wiesel 1962, 1963; Shmuel and Grinvald 1996; Weliky et al. 1996). The selective response properties of visual cortical neurons have spurred many innovative experimental and theoretical efforts to understand the underlying synaptic circuitry (Sillito 1979; Sillito et al. 1980; Heggelund 1981; Berman et al. 1991; Bonhoeffer and Grinvald 1991; Woergoetter and Koch 1991; Somers et al. 1995; Ferster et al. 1996; Weliky et al. 1996; Crook et al. 1997; Roerig and Kao 1999). There is now overwhelming evidence indicating an important role for intracortical synaptic connections, and in particular intracortical inhibitory connections, in the generation of stimulus feature selectivity of visual cortical neurons (Sillito 1984; Martin 1988; Chapman and Stryker 1992; Eysel 1992; Berman et al. 1992; Vidyasagar et al. 1996; Roerig and Kao 1999; Roerig and Chen 2002). For example, it has been shown that connections between inhibitory neurons and excitatory neurons with similar (Woergoetter and Koch 1991; Berman et al. 1991; Somers et al. 1995; Roerig and Chen 2002) or orthogonal (Sillito 1979; Sillito et al. 1980; Heggelund 1981) orientation preferences sculpt the pattern of orientation selectivity. Similarly, converging inputs to excitatory cells from inhibitory neurons located in regions preferring similar and opposite directions appear to contribute to the emergence and sharpening of direction selectivity (Eysel et al. 1988; Crook et al. 1997; Roerig and Kao 1999).

Inhibitory connections are thus of tremendous functional importance. However, while there are numerous anatomical and electrophysiological studies focusing on the development of excitatory neurons and their patterns of connectivity (Rockland et al. 1982; Callaway and Katz 1990; Komatsu and Iwakiri 1991; Lubke and Albus 1992) and inhibitory (Albus et al. 1991; Komatsu and Iwakiri 1991; Kisvarday 1992; Albus and Wahle 1994; Azouz et al. 1997; Gao et al. 1999; Roerig and Chen 2002), very little is known about the development of inhibitory connections. This is in part due to methodological limitations: interneurons are smaller, i.e., harder to record from, and fewer in number (less than 20% of cortical neurons). In addition, they represent a highly diverse group in terms of connectivity, molecular markers, andfunctional properties (Kawaguchi and Kubota 1997; Beaulieu and Somogyi 1990) and show a wide range of morphological and physiological diversity.

Sensory experience during the postnatal critical period is essential for the normal maturation of visual cortical circuits and function (Hubel and Wiesel 1962). Although most studies have been devoted to the modification of excitatory circuits (Rittenhouse et al. 1999),there are indications that GABAergic circuits also change (Winfield 1983). Indeed, one hypothesis is that the maturation of inhibitorycircuits plays an important role in timing the critical periodfor the modifications of excitatory connections (Kirkwood and Bear 1994; Hensch et al. 1998; Huang et al. 1999). Evidence for the slow cortical maturation of inhibitory mechanisms derives primarily from anatomical studies. In rodents, the numbers of inhibitory synapses (Blue and Parnavelas 1983;Miller 1986) and levels of GABA-synthesizing enzymes increase until postnatal week 5 (Huang et al. 1999). Inhibitory circuits continue to mature after birth in the visual cortex, and there is a dramatic postnatal increase in the number of GABAergic inputs converging onto layer II/IIIpyramidal cells. These findings indicate that the GABAergic circuitry in the visual cortex is plastic and can be shaped by experience.

The spatial distribution and potential developmental refinement of inhibitory connections in the developing visual cortex has so far not been investigated. The present study was designed to examine the spatial patterns of inhibitory inputs to layer 2/3 pyramidal neurons in the developing visual cortex of the ferret, an excellent model system for studies of the developing visual system. Ferrets are born relatively immature as compared with other species, such as cat and rat (Luskin and Shatz 1985), which makes it an ideal model for developmental studies. Inhibitory inputs were recorded in tangential slices by whole-cell patch clamp recordings combined with laser scanning photostimulation (Katz and Dalva 1994). We addressed the following questions: (a) Does the spatial distribution of inhibitory inputs change during development and is there evidence for a patchy distribution of inhibitory synaptic connections resembling that of horizontal excitatory connections in layer 2/3? (b) Does a developmental refinement or rearrangement of inhibitory input patterns correlate with the onset of visual experience? This would indicate that the refinement of inhibitory connections is involved in the refinement of cortical maps and response properties of cortical neurons.

Materials and methods

All animal procedures were performed in accordance with the animal care and use guidelines of the NIH, with the approval of the Institutional Animal Care and Use Committee.

A total of 34 animals were used. Animals were divided into four age groups based on developmental stages. These age groups were P7–16 (before thalamic fiber ingrowth, 12 animals), P17–28 (before eye opening, ten animals), P29–37 (around eye opening, seven animals), and P38 and beyond (after eye opening, five animals). A total of 98 neurons in layer 2/3 of the primary visual cortex were recorded successfully from 81 slices. Of these, 54 neurons were identified morphologically as pyramidal cells based on dendritic and axonal morphology. The remaining cells were either nonpyramidal neurons or could not be found in histological sections. Only recordings from pyramidal neurons were analyzed. The number of cells in each age group was: P7–16, 18 cells; P17–28, 16 cells; P29–37, 12 cells; P38 and above, 8 cells.

Slice preparation

Ferrets (P7 to adult, Marshall Farms, New Rose, NY) were deeply anesthetized with Nembutal (100 mg/kg, i.p.). Animals were then decapitated and blocks of tissue containing the primary visual cortex (area 17) from both hemispheres were removed. The block was placed into chilled standard artificial cerebrospinal fluid (composition: 125 mM NaCl, 5 mM KCl, 1.25 mM KH₂PO₄, 1.3 mM MgSO₄, 3.2 mM CaCl₂, 10 mM dextrose, 25 mM NaHCO₃) oxygenated with a mixture of 95% O₂ and 5% CO₂. Tangential slices (400 μm thickness) of area 17 were prepared with a vibratome and were maintained at room temperature in a chamber containing standard ACSF for 1 h before use. Slices containing layer 2/3 were used for patch clamp recording.

Patch clamp and photostimulation

Individual slices were transferred to a recording chamber mounted on the stage of an upright microscope (BX50WI, Olympus Optical, Tokyo, Japan). Slices were continuously superfused with standard ACSF oxygenated with a mixture of 95% O₂ and 5% CO₂ at room temperature. Electrophysiological recordings from single neurons were performed using standard whole-cell patch clamp methods. Patch electrodes (5–10 MΩ) were filled with internal solution (110 mMd-gluconic acid, 110 mM CsOH, 11 mM EGTA, 10 mM CsCl, 1 mM MgCl_2, 1 mM CaCl_2, 10 mM HEPES, 1.8 mM GTP, 3 mM ATP, pH 7.2) and 0.5% Biocytin (Sigma) was added. Voltage clamp recordings were conducted using a Warner PC-501 amplifier (Warner Instrument Corp. USA). Recordings were filtered at 1 kHz and digitized at 8 kHz. Resistances ranged from 11 to 17 MΩ, and a 30–50% compensation was usually achieved using the amplifier adjustments. Synaptic inputs to the recorded neurons were scanned by local photolysis of Nmoc-caged glutamate, which was applied to the bath solution once a stable recording had been established (see details in Roerig and Chen 2002). This chemical was designed to resist spontaneous hydrolysis while yielding reasonable photorelease and kinetics (Rossi et al. 1997). Upon photolytic uncaging, light-induced cleavage of Nmoc-caged glutamate (on a sub-millisecond time scale) occurs followed a subsequent light-independent, pH-dependent decarboxylation (on the millisecond time scale). Because of the possible difference among animals of different ages in terms of ability to drive presynaptic cells to fire action potentials, a high concentration of Nmoc-caged glutamate (1 mM) was used to increase the firing probability. An argon/krypton ion laser (Spectraphysics Stabilite 2017) was used as a UV light source for local uncaging. The laser beam was coupled into a 50-μm-diameter fiber optic mounted to a motorized XY-stage. The fiber was moved within an oil droplet below the recording chamber. The laser beam was manually focused into the middle of the slice preparation (Fig. 1). Opening of the external shutter, scanning of the laser beam, and data acquisition were controlled by a National Instruments AD board (AT-MIO/AI E-10) and custom-written software (Labview, National Instruments).

Fig. 1

Schematic representation of the photostimulation setup. The laser beam was coupled into a 50-μm-diameter fiber optic mounted to a motorized XY-stage. The fiber was moved within an oil droplet below the recording chamber. Synaptic inputs were scanned at 50-μm spacing. The duration of the shutter opening was 10 ms. Inhibitory and excitatory currents were separated by recording at a holding potential of −20 mV

The acquisition period for each laser stimulus (flash) was 1 s and the interstimulus interval was 3–5 s. In each acquisition period (1,000 ms), the shutter opening occurred at 500 ms and the laser flash duration was 10 ms. The prestimulus acquisition period of 500 ms before shutter opening served to monitor spontaneous synaptic inputs to the recorded neurons. The spacing of stimulation sites was 50 µm. Typical maps consist of 500 to more than 2,000 stimulation sites, with sampling done in a “four-pass” mode, i.e., only every fourth spot was stimulated during one round and the map was filled in during four acquisition cycles.

The holding potential for the patch clamp recording was set at −20 mV to distinguish between excitatory and inhibitory synaptic inputs (Katz and Dalva 1994). At this holding potential, the outward inhibitory currents occurred as upward deflections (inhibitory postsynaptic currents: IPSCs) whereas inward excitatory events occurred as downward deflections (excitatory postsynaptic currents: EPSCs). Since most cells would fire and escape voltage clamp at this very depolarized holding potential, QX314 (100 μΜ) was added to the internal solution to block sodium channel. Depending on the stimulation site, either EPSC alone or IPSC alone or a mixture thereof could occur (Fig. 2A,B). IPSCs were almost abolished after bath application of GABAzine, a GABA_A antagonist (Fig. 2C). In the photostimulation maps shown in the Results, only IPSCs are represented since our focus was on inhibitory maps. The current detection threshold was defined as 2x standard deviation. Direct photostimulation of the postsynaptic cell body resulted in spiking or a large, short-latency (~8 ms) inward current. In most cases, direct currents can be easily distinguished from postsynaptic currents by their short latency and large amplitude. Fig. 3 shows such latency differences. Direct activation had short latencies (mean 8.1 ms, standard deviation 2.45 ms, n=324 traces), whereas postsynaptic currents had much longer latencies (mean 55 ms, standard deviation 15.61 ms, n=7,561 traces). To further distinguish between direct activation and postsynaptic currents, tetrodotoxin (TTX, 1 µM) was added to the bath solution in two experiments to block synaptic transmission. In these cases, postsynaptic currents (EPSCs and IPSCs) were blocked while direct activation remained unaffected. In each slice only one to two cells were recorded to facilitate alignment and unambiguous assigning of input maps to postsynaptic cells.

Fig. 2 A

Different types of postsynaptic currents were evoked depending on the site of photostimulation. Upper and middle traces show mixed EPSCs and IPSCs. The lower trace shows an EPSC. B Consistency of evoked EPSCs (upper traces) and IPSCs (lower traces) when the same presynaptic site was stimulated repetitively, with an interval of 3–5 ms. In both examples there was little variation in current onset latency, amplitude, or duration. C Currents evoked by photostimulating the same sites before and after bath application of GABAzine. The IPSC was almost abolished after GABAzine application (upper traces), whereas EPSC was little affected (lower traces)

Fig. 3

The distribution of onset latency for direct activation (solid circles) and synaptic currents (blank circles)

To test the consistency of the evoked currents, the same spot was photostimulated repeatedly. Figure 2B (upper traces) shows an overlay of five EPSCs evoked by stimulation of a presynaptic site repeatedly. No significant differences were observed among traces in terms of latency, amplitude, or peak time of the evoked currents. A similar result was found for IPSCs, as shown in the lower traces in Fig. 2B. Evoked currents were still reproducible with an interval of 30 min between stimulations of the same site. In addition, very similar maps were obtained from one cell in repetitive mapping experiments during a recording period of up to 2 h.

Histology

Following the recording session, slices were fixed for 1 day in 4% paraformaldehyde in phosphate-buffered saline (pH 7.4) and then transferred into phosphate-buffered saline solution containing 30% sucrose. Slices were resectioned at 70 µm on a vibratome and processed for immunohistochemical visualization of biocytin-filled cells using a standard avidin-biotin (Elite) kit and DAB as chromagen. Labeled cells were inspected for identification of their cell type. Sections containing labeled pyramidal neurons were photographed for later alignment with photostimulation maps.

Data analysis and image alignment

All data shown here are from layer 2/3 pyramidal neurons. Before patch clamp recording, a slice overview image (including the position of the micropipette) was taken using a SONY XC-75 CCD camera and a SNAPPY (play) video frame acquisition module in conjunction with SNAPPY software. This slice image was used to aid the alignment of histological sections with synaptic input maps obtained by photostimulation (Fig. 4). The position of the stained postsynaptic cells in the histological sections and the direct activation area in the photostimulation map served as additional marks. The alignment was made using the layer menu of Adobe Photoshop. Only linear scaling and rotation were applied to images during alignment.

Fig. 4

Slice preparation and photostimulation. A Plane of sectioning for tangential slices. Only sections through layers 2/3 were used in our experiments. B Tangential slice through the upper layers prepared from a P48 ferret. The section contains the biocytin-filled postsynaptic cell. C Higher power view of the recorded (postsynaptic) layer 2/3 pyramidal cell. D Example of a stimulated (“mapped”) area overlaid on a histological section of the recorded slice (here from a P35 animal). A live slice image was used to aid the alignment. E Morphology of another biocytin-labeled pyramidal neuron recorded in a tangential slice. The photostimulation map of this neuron is displayed in F. Each colored square in E corresponds to an area of 50×50 μm. The various colors are used to represent the amplitude of synaptic input to the recorded cell, as indicated by the color bar

Photostimulation data were analyzed using custom software written in LabVIEW (National Instruments). Current events were inspected and manually marked. Only synaptic currents with onset occurring within 100 ms after shutter opening were selected for further analysis. This time window was determined based on our current clamp data showing that all photostimulation-evoked action potentials in presynaptic neurons occurred within this time frame (Roerig and Chen, unpublished). For each inhibitory current event, the delay to onset latency and peak amplitude was manually detected. The number and amplitude of spontaneous events in each trace were also analyzed for later calculation of normalized synaptic input values (Dantzker and Callaway 2000). The analyzed data were then fed into a second program in which the selected events and their corresponding positions from which they were evoked were retained. This program yielded a photostimulation map in which the peak amplitudes (not the average amplitude) of current events were color-coded for each stimulation site. The current amplitude shown in this map was not corrected for spontaneous activity and differences of series resistance between each recording session. However, these two factors were taken into account when the overall synaptic input values were calculated, as described below.

To compare the spatial distribution of inhibitory inputs for each age group, the area surrounding the recorded cell was divided into concentric annuli spaced 200 µm apart. A normalized evoked input (NEI) value was calculated for each annulus of each recorded cell to compensate for differences in spontaneous events and series resistance among different recording sessions (Dantzker and Callaway 2000). For each annulus of 200 µm, the normalized input values NEI were calculated:

$$ {\text{NEI}} = \frac{{\overline{{f_{E} }} \overline{{a_{E} }} - \overline{{f_{C} }} \overline{{a_{C} }} }} {{\overline{{a_{C} }} }} $$

where f _E is the mean number of evoked events, a _E is the mean amplitude of evoked currents, f _c is the mean number of spontaneous events and a _c is the mean amplitude of spontaneous currents.

The calculated values of the corresponding annuli from each cell of the same age group were then pooled and averaged. Plotting of histograms as well as statistical comparison of distributions was done using Sigmastat and Sigmaplot (Jandel).

Results

The main purpose of this study was to examine the spatial pattern of intracortical inhibitory connections in developing ferret visual cortex. To this end, the locations around the recorded neurons were scanned using the photostimulation method. A synaptic input map for each recorded neuron was thus obtained. These maps were color-coded, representing the amplitude of evoked currents. We defined local inputs as those PSPs elicited within 400 μm from the target cell. Those that originated at between 400 and 600 μm and at >600 μm were defined as intermediate and long-distance inputs, respectively. Results are described in the order of ascending age groups.

Early development before thalamic fiber ingrowths (P7–16)

At the early stage of this group, evoked currents mostly originated from a small area (within ~250 µm) around the recorded neurons. The majority of these currents resulted from direct activation of the recorded cell. Figure 5A shows results from a P7 ferret. The upper panel shows a color-coded map with numbered squares. The corresponding current traces for these numbered squares are displayed in the lower panel. Stimulations of the soma of the recorded cell (white square) evoked a short-latency (10 ms), large-amplitude (25 pA) direct activation (trace 1). This inward current had a flat plateau lasting for a few hundred milliseconds. This type of long-lasting direct current was characteristic of pyramidal neurons during the first two postnatal weeks. Similar current traces but with longer latencies were also observed (traces 2–4). They resulted from photostimulation of the dendritic trees of the recorded neuron. EPSCs could also be evoked at this early age from areas close to the cell body (within ~250 µm). However, they were usually of small amplitude (8–10 pA, trace 5) and were usually intermixed with and masked by direct activation. The amplitudes of direct activation and excitatory synaptic currents increased with age. Figure 5B shows a representative example from a later stage (P15) of this group. As compared with P7 (panel A), direct activation (trace 1) and EPSCs (traces 2–5) increased in amplitude and EPSCs were evoked from a larger area.

Fig. 5

Representative examples (A P7, B P15) from the P7–16 age group showing the results from photostimulation. For each panel, the color-coded photostimulation map is shown at the top and the current traces for the numbered squares on the maps are shown below. The white square marks the location of the recorded neuron. In A, evoked events originated from the area 200–250 μm around the recorded cell. Most of them represented direct activation. The direct activation in young animals (P7–P12) usually lasted for more than 500 ms. Small amplitudes of EPSCs (trace 5) were also observed. No IPSC was encountered. In B, the excitatory synaptic input (EPSC) map is shown at the top. Direct activation (trace 1) was evoked by stimulation of the recorded cell directly (white square). Traces 2–5 show representative examples of EPSCs resulting from stimulation of multiple presynaptic sites. No IPSCs were detected

IPSCs, on the other hand, were not detected at P7. Even in more mature animals, IPSCs were not observed in most cases. In some cases, one to three small IPSCs (<8 pA) were recorded. Their paucity prevented us from generating a meaningful synaptic input pattern. Thus, at this developmental stage, while excitatory connections were already forming, a discernible pattern of inhibitory connections was not observed.

Age group before eye opening (P17–28)

A recognizable pattern of inhibitory inputs was first detected at P17. At this age, IPSCs were evoked from a small region (within ~200–300 µm) around the recorded cell (Fig. 6A). Their amplitudes were also small (maximum 15 pA). As age increased, both the number of presynaptic sites and the amplitude of IPSCs increased (Fig. 6B, P22; 6C, P28). At P28 (Fig. 6C), the region providing the majority of inhibitory inputs expanded to an area of ~500 µm distance around the recorded neuron. However, as the color-coded maps indicate, presynaptic sites closer to the recorded neurons usually provided stronger inputs to their target cells. In addition, IPSCs evoked from a similar distance away from the recorded cells usually had similar amplitudes. Long-distance inhibitory inputs (>600 µm) were rarely seen. These results suggest that inhibitory inputs were undergoing a constructive phase at this age group.

Fig. 6

Representative examples showing inhibitory synaptic input pattern recorded from the age group before eye opening (A P17, B P22, C P28). Inhibitory synaptic inputs (IPSCs) were first detected at the age of P17 (A). The IPSCs were evoked by stimulation of a small region (~200 μm) around the recorded cell and their amplitude was small. With increasing age, the number of effective stimulation sites increased and the average IPSC amplitude increased (B, C)

Age group around eye opening (P29–37)

Figure 7 shows results from representative examples of this age group (panel A, P30; panel B, P35). As seen in panel A (P30), the previous constructing phase of inhibitory connections continued and reached a peak at around P30–32. At this postnatal age, the area supplying the vast majority of inhibitory inputs further extended into a larger area of 600 µm around the recorded cells. The amplitude of IPSCs also increased as compared with those in the previous age group. In addition, long-distance inhibitory inputs (>600 µm) were present. However, they were few in number and patches of presynaptic sites generating IPSCs were rarely seen (panel A).

Fig. 7

Representative examples demonstrating inhibitory synaptic input maps from an age group around the time of eye opening (A P30, B P35). At the early stage of this group (P30, A), the majority of local inhibitory inputs originated from an area ca. 600 μm around the recorded neuron. At the late stage of this group (P35, B), the effective stimulation sites shrank to an area ca. 300 μm around the recorded cell. In addition, long-distance inhibitory inputs (>600 μm) appeared, although in small numbers. As compared with the previous age group, the amplitude of these local IPSCs increased

At the late stage of this age group, as seen in panel B (P35), the constructive phase of inhibitory connections appeared to pass into a rearrangement or modification period. This is indicated by the fact that the number of intermediate inputs (400–600 µm) began to decrease. In addition, IPSCs evoked from a similar distance from the recorded neuron showed a greater variation in amplitude (compare P35, Fig. 7B, with P28, Fig. 6C), indicating rearrangement or modification of synaptic inputs. Concurrent with these changes, the amplitudes of the inhibitory inputs (0–400 μm) increased, suggesting more efficient and/or more stable connections. The number of long-range inhibitory inputs also increased. Thus, around the period of eye opening, the intracortical inhibitory connections undergo an early constructing phase and a later rearrangement or modification phase.

Age group after P38 (after eye opening)

Figure 8 shows results from representative cases of this age group (panel A, P38; panel B, P56). Compared with the previous age groups, the area supplying the majority of inhibitory inputs decreased from ~600 µm to 200–300 µm (panel B) from the recorded neuron. The average amplitude of these local inhibitory inputs, on the other hand, increased progressively as the animals became more mature. Thus, while the total number and areas supplying inhibitory inputs decreased, the potency of local inputs increased, indicating increased input specificity. Long-distance inhibitory inputs (>600 μm), while not commonly seen in the age group around the period of eye opening (P29–37), were encountered frequently in this age group. More distant inhibitory inputs up to 2 mm could also be found (panel B). Distant inputs tended to form a clustered pattern. On average, each cluster consisted of three to eight effective stimulation sites. IPSCs from the same cluster usually exhibited different kinetics, signaling that they originated from close but different presynaptic sites. It should be noted, however, that although the number of long-range inhibitory inputs increased as animals became more mature, their total number and average amplitude were much lower than those of the local inputs.

Fig. 8

Representative examples of inhibitory synaptic input maps recorded from the age group after eye opening (A P38, B P56). The presynaptic sites providing local inhibitory inputs shrank further to an area about 250 μm around the recorded neuron in the adult (B), while the IPSC amplitude increased. Long-distance inhibitory inputs increased in number and some of these presynaptic sites aggregated to form a clustered patch pattern. Note that the amplitude of long-distance inhibitory inputs is much smaller than that of the local inputs

Quantitative comparison of inhibitory inputs among different age groups

Results from different age groups were compared by calculating the average amplitude of IPSCs and the average percentage of effective stimulation sites. This was accomplished by pooling IPSCs elicited from the same distance range (at the interval of 200 μm) and from the same age group. A comparison of the percentage of effective stimulation sites (sites responding/total number of stimulation sites) is shown in Fig. 9, and a comparison of mean amplitudes is shown in Tables 1 and 2 and Fig. 10. As is evident from Fig. 9, most IPSCs originated from within 200 μm from the recorded cell in all age groups. In this distance range, there was no statistically significant difference among the groups. However, at the 200- to 400-μm range, the P29–37 group differed significantly from the other two groups. The P29–37 group showed 66.98% effective stimulation sites, whereas the P17–28 and P38 and above groups showed 47% and 45%, respectively. At the range of 400–600 μm, the P29–37 group also had a significantly higher percentage value (40%) than the P17–28 (11.57%) and P38 and above (14.88%) groups. At a distance of >600 μm, the percentage value of the P38 and beyond group was significantly higher than those of the P17–28 and P29–37 groups. These results suggest that inhibitory connections undergo a constructive phase that peaks at around the period of eye opening. They are subsequently rearranged/modified as the animals become more mature.

Fig. 9

Developmental changes in the origin of inhibitory synaptic inputs with regard to their distance from the target cell body. The percentage of effective stimulation sites (those that generate IPSCs in response to photostimulation) at a distance of 200 μm from each pyramidal neuron within each corresponding age group is shown. For within-group comparison, the percentages of IPSCs found at 0–200, 200–400, and 400–600 μm were significantly different from each other (P<0.02). Within the P29–37 group, the percentage of IPSCs found at 400–600 μm was significantly different from that found at 600 through 1,000 μm. For between-group comparison, the P28–37 group differed significantly from the P17–27 and ≥P38 groups at 200–400 and 400–600 μm (P<0.05). The ≥P38 group differed significantly from the other groups at >600 μm (P<0.02). All significance tests were done using one-way ANOVA and the Bonferroni t test

Table 1 Mean amplitude of evoked IPSCs following photostimulation. Amplitude was pooled from each 200-µm annulus from animals of the same age group

Table 2 Normalized evoked inhibitory input values for each 200-µm annulus of each age group (see Materials and methods for calculation method)

Fig. 10

Normalized inhibitory input (NEI) values calculated for each 200-μm distance range for each age group. Within-group comparison showed significant differences between each 200-μm range (P<0.05). Comparisons between age groups revealed significant differences between each age group for each 200-μm range (P<0.05), except at 400–600 μm, at which no significant difference was found between P29–37 and ≥P38 groups

The normalized amplitude values (NEI, see Materials and methods) of IPSCs for each age group are shown in Table 2 and displayed as a histogram in Fig. 10. For all distances, the NEI values increased with increasing age, except in the 400- to 600-μm distance range, where no significant difference was found between the P29–37 and ≥P38 groups. In addition, for all age groups, the NEI values were highest close to the cell body and decreased with increasing distance from the cell body.

IPSC amplitudes increased with age; however, their time course also shortened, which may result from developmental changes in GABA_A receptor properties (Heinen et al. 2004). We have therefore calculated the total IPSC charge transfer for all age groups to account for the change in IPSC efficacy (Fig. 11). Charge transfer increases dramatically around eye opening (P30–P35). In addition, a second significant increase occurs between P40 and adulthood, indicating further strengthening of GABAergic connections. This increase was observed in all distance ranges but was most prominent in local connections (0–400 μm).

Fig. 11

Increase in mean IPSC charge transfer during postnatal development. A steep increase occurred around eye opening (P30–35) and between P40 and adulthood. Intracortical inhibition increases during the period of experience-driven remodeling and refinement of connections and reaches its highest, mature level when critical periods for refinement of maps and synaptic connections are closed. Bars represent mean and standard error; between 78 and 260 evoked synaptic events have been analyzed per age group

Discussion

The present study investigated the spatial pattern of inhibitory connections in the developing visual cortex. In light of the extreme functional importance of inhibitory connections, this is an important aspect of cortical circuit development, which has not been addressed before. Using the technique of photostimulation in combination with patch clamp recording, we have examined the spatial pattern of inhibitory inputs to pyramidal neurons in layers 2/3 of developing ferret striate cortex. Our results indicate that the spatial pattern of intracortical inhibitory connections undergoes developmental rearrangements or modifications that coincide with developmental stages and events, such as eye opening and the initial presence and maturation of orientation maps.

Methodological considerations

Slice preparation

The aim of this study was to analyze the development of inhibitory connections and their spatial arrangement since these parameters are of crucial importance to the development of functional response properties. We therefore made a strong effort to confine recording and stimulation sites to the upper cortical layers in tangential slices. However, owing to the curvature of the cortex we cannot completely rule out that some slices contained parts of other layers and that some intralaminar connections may have been stimulated. Although this potential problem may have contaminated some of our developmental results, the cresyl violet control stains and biocytin fills of recorded cells showed that the majority of recording sites and stimulation sites were in the upper layers.

Spatial resolution of the photostimulation method

The spatial extent of activated neurons around the laser spot is dependent on many factors, including the caged glutamate kinetic profile (in particular, its concentration over time), the laser flashing time, and the extent of dendritic trees in the surrounding neurons. Not all of these factors can be determined and, thus, spatial resolution had to be tested empirically in our system.

We performed voltage clamp in cell-attached mode on three pyramidal neurons (one from P28, two from P48). The Ca²⁺ concentration in bath solution was reduced to block synaptic transmission (“low calcium” in mM: 0.4 CaCl₂, 4.0 MgSO₄, all other ion concentrations were the same as in normal ACSF). One example is shown in Fig. 12. As is evident from the photostimulation map, spiking of the recorded neurons occurred only when its soma or proximal dendrites were directly stimulated. Activation of distal dendrites only elicited direct activation and no response was elicited when uncaging occurred immediately adjacent to the recorded neurons. This was observed for all three neurons tested. Indeed, recording under normal conditions also showed that spiking or direct activation could only be elicited from one or two locations directly under the recording pipette. In view of these observations, it appears that the effective stimulation area in the horizontal plane is at most less than 100 µm × 100 µm. We believe that this level of resolution suffices for the purpose of comparison between different age groups. This is under the assumption that if there was over- or underestimation of the spatial resolution, it occurred in all the cells that were recorded under the same condition.

Fig. 12 A

Morphology of a pyramidal neuron tested in cell-attached mode. Ca²⁺ in bath solution was reduced to block synaptic transmission. B “Map” of evoked currents recorded from this cell. Spiking (black color) of the cell occurred when the laser spot hit the soma or proximal dendrites. Activation of distal dendrites only elicited direction (other colors). No response was observed when areas immediately adjacent to the cell were activated

Polysynaptic responses

Polysynaptic responses could occur if a glutamate-activated excitatory neuron elicited spiking of its target interneuron, which in turn projected to the recorded neuron. This seems unlikely because, as mentioned earlier, we never observed firing of recorded neurons by photostimulation of their presynaptic inputs. Spiking was only seen when the recorded neurons were directly stimulated. This occurred also in cell-attached mode with normal bath Ca²⁺ concentration. In the visual cortex, extracellular recording in slice preparations also showed that action potentials could only be elicited when the recorded neuron was directly stimulated by the laser beam (Katz and Dalva 1994). In addition, we observed very little jitter in the response latencies upon repetitive stimulation of the same input, which is indicative of a monosynaptic connection. The length of the latency can be explained in part by the duration of uncaging and the concentration built up at the stimulation site. It should also be noted that the uncaging time for the more stable Nmoc-caged glutamate is longer than for commercially available compounds (see Rossi et al. 1997). However, the most critical factor is the time of spike generation in the presynaptic neuron. Once a spike has been generated, the remaining latency component is only determined by action potential propagation time and synaptic delay. If the spike occurs late following the laser flash (which we frequently observed by directly stimulating neurons while recording spikes in current clamp mode), the overall latency will be long. Response latencies of 40–100 ms or longer have, furthermore, been observed by other authors using the photostimulation techniques [e.g. Sawatari and Callaway 2000; Dantzker and Callaway 2000; Shepherd et al. 2003)].

Are spatial patterns of inputs at different ages merely the reflection of brain growth?

It is not likely that the disparate spatial patterns for different age groups are the result of cell density change as brain grows. If this were the case, we would expect the density of local inputs to be decreased in more mature animals. However, our results indicated that the percentage of local inputs (200 µm) did not differ significantly among the three groups examined (Fig. 9).

Source of inhibition

Application of GABAzine to the bath solution almost abolished the IPSCs that occurred before 100 ms (Fig. 2C). Therefore, it is assumed that the IPSCs recorded resulted mostly from activation of GABA_A receptors. The slow, G-protein-coupled, metabotropic GABA_B receptor has been reported to be involved in shaping the response properties of individual neurons (Allison 1996). However, this component was difficult to measure in our study because whole-cell patch clamp recording dialyzes second messenger compounds out of neurons.

Late development of inhibitory connections

IPSCs were only occasionally observed in the youngest age group (P7–16). Their amplitude was also small (<8 pA). Their paucity prevented us from generalizing a spatial input pattern. Anatomically, however, it has been shown that bands of GABA-reactive neurons are present at cortical and subcortical plates of ferrets as early as the first postnatal day (Gao et al. 1999). Functional GABAergic synaptic connection in neonatal mouse barrel cortex has also been reported (Agmon et al. 1996). The paucity of IPSCs in our study could have been due to the following factors:

• 1. Our current detection threshold is not sensitive enough to detect the IPSCs. This seems unlikely, however, since we did detect IPSCs in some slice preparation in this age group. In addition, EPSCs were observed in all the recorded neurons.

• 2. Inhibitory synapses in young animals have a more depolarized potential, owing to a higher intracellular chloride concentration. At a holding potential of –20 mV, the driving force could have been too small to elicit any IPSCs. We did not directly test this possibility. However, again, we did observe IPSCs in some cases within this age group.

• 3. It has been shown that inhibitory inputs in newborn animals originate only locally (Agmon et al. 1996). It is possible that these local IPSCs were masked by direct activation currents that resulted from stimulation of dendritic arborization of the recorded neurons. However, local IPSCs in more mature age groups were easily detected. Therefore, this assumption holds under the condition that IPSCs in young animals are of very small amplitude.

• 4. Inhibitory neurons in young animals could be less sensitive to glutamate, and thus less excitable. The immaturity of inhibitory interneurons could explain the paucity of IPSCs observed in our study. These neurons are less excitable by other neurons and thus exert less inhibitory effect on their target neurons. The fact that IPSCs were detected in newborn mice but not in ferrets until P17 could have been due to species differences. Unlike monkeys, cats, rats, and mice, ferrets are born before thalamocortical inputs reach the subplate and well before development of the cortical plate is complete (Luskin and Shatz 1985).

The first discernible pattern of inhibitory inputs was present at around P17 (Fig. 6A), which is at least more than 1 week later than the initial appearance of excitatory connections (Dalva and Katz 1994). The late presence of functional inhibition in this study is consistent with other reports (Komatsu and Iwakiri 1991; Luhmann and Prince 1991; Gao et al. 1999) that the development of intracortical inhibitory circuits lags behind that of the excitatory ones. This could partly account for the observation that visual cortical neurons respond to visual stimulation but lack selective responsiveness in newborn animals (Blakemore and Van Sluyters 1975; Buisseret and Imbert 1976). It could also be responsible for the critical time window during which the organization of intracortical circuitry can be strongly influenced by sensory experience (Kirkwood et al. 1997; Micheva and Beaulieu 1997).

Development of inhibitory connections is related to the refinement of response properties in the visual cortex

In adult animals, the intracortical inhibitory circuits are believed to play a major role in shaping the orientation selectivity of cortical neurons (Sillito 1979; Woergoetter and Koch 1991; Berman et al. 1991; Somers et al. 1995; Roerig and Chen 2002). In ferret visual cortex, orientation tuning could be recorded as early as P23 by single-unit recording (Chapman and Stryker 1993) and at P31–36 by optical imaging (Chapman et al. 1996). The latter technique is less sensitive and thus requires stronger underlying intrinsic signal. This time frame corresponds to our observations that inhibitory connection was in the early immature stage at P23 and in the peak of its construction phase around eye opening. The low-contrast initial orientation map matures into an adult-like, high-contrast map at around P39 (Chapman et al. 1996). Our data show that the spatial pattern of inhibitory connections at P38 resembled that of adult animals. At this developmental age, the vast majority of inhibitory inputs originated from immediate areas close to the recorded neurons (200–300 µm). This is consistent with the anatomical organization of dense local axonal collaterals of inhibitory neurons found within layer 2/3 (Beaulieu and Somogyi 1990; Kisvarday et al. 1993). The number of relatively distant inputs (300–600 µm), which were abundant around the period of eye opening, was sharply reduced. Thus, it appears that a smaller, but focused and stronger (i.e., bigger IPSC amplitude) inhibitory input source is essential for cortical neuron feature selectivity.

Chronic cortical optical imaging in developing ferrets indicates that the structure of orientation maps, including the locations of orientation domains, remains remarkably constant during development (Chapman et al. 1996). This implies that the orientation feature selectivity is predefined early in the development and is experience independent (Chapman et al. 1996, Goedecke and Bonhoeffer 1996). The selectivity could well be the product of interplay between excitatory and inhibitory connections during development. Similar to inhibitory connections, excitatory inputs undergo a constructive phase that peaks around or shortly after eye opening (Dalva and Katz 1994). At this period, both the local excitatory and the local inhibitory inputs reach their maximum numbers. As described above, it is at this time that the primitive orientation map is first detected, and it remains relatively unchanged later in normal development (Chapman et al. 1996). Therefore, it is assumed that a primitive excitatory and inhibitory framework has been formed at this period. Its components are subject to modification by substitution, deletion, and insertion commensurate with visual experience. Following subsequent normal development, the relative number of local excitatory inputs declines whereas that of inhibitory inputs increases. This is accompanied by the sharpening of orientation tuning. Thus, it appears that a balance between the relative numbers of local excitatory versus inhibitory inputs affects orientation tuning.