Abstract
The influence of neuronal activity on cortical axon branching was studied by imaging axons of layer 2/3 neurons in organotypic slice cultures of rat visual cortex. Upper layer neurons labeled by electroporation of plasmid encoding yellow fluorescent protein were observed by confocal microscopy. Time-lapse observation of single-labeled axons showed that axons started to branch after 8-10 d in vitro. Over the succeeding 7-10 d, branch complexity gradually increased by both growth and retraction of branches, resulting in axon arbors that morphologically resembled those observed in 2- to 3-week-old animals. Electrophysiological recordings of neuronal activity in the upper layers, made using multielectrode dishes, showed that the frequency of spontaneous firing increased dramatically ∼10 d in vitro and remained elevated at later stages. To examine the involvement of spontaneous firing and synaptic activity in branch formation, various blockers were applied to the culture medium. Cultures were silenced by TTX or by a combination of APV and DNQX but exhibited a homeostatic recovery of spontaneous activity over several days in the presence of blockers of either NMDA-type or non-NMDA-type glutamate receptors alone. Axonal branching was suppressed by TTX and AMPA receptor blockade but not by NMDA receptor blockade. We conclude that cortical axon branching is highly dynamic and that neural activity regulates the early developmental branching of upper layer cortical neurons through the activation of AMPA-type glutamate receptors.
Introduction
During neural circuit formation, an intriguing axonal behavior is the formation of elaborate branches in target regions. This developmental event permits neurons to make extensive synaptic connections selectively with appropriate target cells.
In most vertebrate species, it is believed that there are at least two modes in the morphological development of axons (Acebes and Ferrus, 2000). The first is stereotyped branching, as is found in the topographic- and layer-specific elaboration of retinotectal projections and cortical connections (Sanes and Yamagata, 1999; Yamamoto et al., 2002; McLaughlin et al., 2003a). In these systems, early axonal branching is regulated by branch-promoting and -inhibiting factors expressed in specific locations and stages (Yamagata and Sanes, 1995; Inoue and Sanes, 1997; Castellani et al., 1998; Yamamoto et al., 2000; Yates et al., 2001; Hindges et al., 2002; Mann et al., 2002; Sakurai et al., 2002; McLaughlin et al., 2003b).
The other mode involves modification by neuronal activity. Both sensory-evoked and spontaneous firing have been demonstrated to influence wiring patterns in the brain (Hubel et al., 1977; Reh and Constantine-Paton, 1985; Shatz and Stryker, 1988; Hata and Stryker, 1994; Penn et al., 1998). Several studies have indicated that neural activity affects individual axonal branching in vivo (Udin, 1983; Sretavan et al., 1988; Cline and Constantine-Paton, 1990; Antonini and Stryker, 1993, 1996; O'Rourke et al., 1994; Antonini et al., 1999; Cohen-Cory, 1999; Hata et al., 1999; Ruthazer et al., 2003). However, little is known about the mechanisms by which neural activity influences axonal branching.
To explore the role of neural activity on axonal branching, we have focused on the development of horizontal axons in the mammalian neocortex. Horizontal axons are long-range axon collaterals that horizontally run in layer 2/3 and form branches in the same layer (Gilbert and Wiesel, 1979; Rockland and Lund, 1982; Burkhalter and Charles, 1990; McGuire et al., 1991; Lohmann and Rorig, 1994). It has been reported that these axon arbors develop during postnatal stages by growth and retraction (Callaway and Katz, 1990; Durack and Katz, 1996). Moreover, it is likely that neuronal activity is involved in this process, because the organization of these connections is altered by deprivation of visual experience (Callaway and Katz, 1991; Lowel and Singer, 1992) and by blockade of action potentials (Ruthazer and Stryker, 1996).
In the present study, we examined the role of neuronal activity in branching of horizontal axons using organotypic slice cultures, in which cortical cytoarchitecture and fine structures such as axonal projections and dendritic morphology are preserved (Yamamoto et al., 1992; McAllister et al., 1995; Dantzker and Callaway, 1998). Development of axonal morphology over many days was studied by time-lapse confocal imaging of individual yellow fluorescent protein (YFP)-transfected axons. The influence of neuronal activity on axonal branching was further studied by pharmacological manipulation of glutamatergic transmission and action potential firing, monitored by electrophysiological recordings from multielectrode arrays on which the cultures were grown. The results suggest that neuronal activity contributes to the development of horizontal axon branching by activating synaptic transmission through non-NMDA receptors.
Materials and Methods
Organotypic slice culture. Sprague Dawley rats were used for this study. The first 24 hr after birth was referred to as postnatal day 0 (P0). Coronal slices were dissected from P0 or P1 rat occipital cortex and were cultured using organotypic culture techniques, as reported previously (Yamamoto et al., 1989, 1992). In brief, these slices were cultured in serum-free, hormone-supplemented medium on culture membranes (Millicell-CM PICMORG50; Millipore, Bedford, MA) that were coated with a collagen solution prepared from rat tails. To monitor neuronal activity, cortical slices were grown on multielectrode culture dishes (MEDs) (Alpha MED Sciences, Tokyo, Japan) that were coated with collagen. In both cases, the cultures were maintained at 37°C in an environment of humidified 95% air and 5% CO2.
Axon labeling by electroporation. To label cortical axons, we developed local electroporation and single-cell electroporation methods. In local electroporation, a relatively large number of cells were transfected but localized in a small region (∼0.2 mm in diameter). In the single-cell electroporation method, individual or, at most, a small number of cells were transfected and clearly distinguishable.
A plasmid containing the coding region of enhanced YFP (EYFP) (pCAGGS:EYFP; a generous gift from Dr. Y. Hatanaka, University of Tsukuba, Tsukuba, Japan) was used for the electroporation (Niwa et al., 1991). The plasmid was isolated (Plasmid Maxi kit; Qiagen, Tokyo, Japan) and resuspended at concentrations of 1.0-1.5 mg/ml in Hanks' solution. In both methods, electroporation was performed after 5-6 d in vitro (DIV). Cellular morphology was observed from a few days to >1 week after transfection.
For local electroporation, a glass micropipette (tip diameter, 50-100 μm) was filled with the plasmid solution. The tip of the micropipette was placed between 100 and 300 μm from the pial surface of cortical slices with a micromanipulator (MO-10; Narishige, Tokyo, Japan). Then, a pair of tungsten electrodes (∼300 μm apart) was positioned on both sides of the tip of the micropipette. After pressure injection of the plasmid-containing solution, voltage pulses (100 V; pulse duration, 5 msec; interval, 1 sec; 10 times) were delivered with a square-pulse electroporator (CUY-20; BEX, Tokyo, Japan).
The single-cell electroporation method was modified from Haas et al. (2001). Single-cell electroporation was accomplished using a glass micropipette (tip diameter, 30-50 μm) filled with plasmid-containing solution. A silver wire (0.2 mm diameter) was placed inside the micropipette in contact with the DNA solution. The tip was then placed on the culture surface. Current pulses, generated by a stimulator (Master 8; A.M.P.I.,, Jerusalem, Israel) and a biphasic isolator (BSI-2; BAK Electronics, Germantown, MD), were delivered between the micropipette and a silver wire ground electrode. For delivery of negatively charged DNA, the silver wire inside the micropipette was connected to the negative terminal (anode) of the isolator, and the ground electrode was connected to the positive terminal (cathode). The current passing through the micropipette was monitored across a resistor (100 Ω) in series. The electroporation pulse parameter was three to five trains of 200 square pulses of 1 msec duration at 200 Hz and of 100-150 μA amplitude.
The somata, dendrites, and axons of individual or a small number of cells were clearly labeled by single-cell electroporation. The local electroporation method also labeled cells with axons that traveled horizontally from the upper layers but did not always permit identification of the soma of each axon because of the dense cluster of labeled cells resulting from this method. In both methods, single-labeled horizontal axons that were distinguishable from the others were selected for analysis.
To observe axonal projection pattern in vivo, upper layer cells were transfected similarly with the YFP vector in cortical slices immediately after dissection. The slices were then cultured for 24 hr, which allowed the protein to be synthesized and diffused into processes. This method enabled us to observe cellular morphology at P7 acute slices but not for older animals.
Confocal imaging and quantitative analysis of labeled axon. Cells labeled with YFP were observed using a laser-scanning confocal microscope (MRC-600; Bio-Rad, Tokyo, Japan). YFP fluorescence remained high for >1 week, even with repeated imaging. An objective lens with long working distance (10×, 20×, or 40×; working distance, 5-6 mm) was used. The confocal microscope was equipped with an argon laser (excitation wavelength, 488 nm) and a filter set for fluorescein isothiocyanate. A neutral density filter (1-10% transmittance) was placed in the path of the laser to reduce the intensity of excitation and minimize photodynamic damage produced by the excitation beam. Transmitted light images were also collected to locate the pial surface of the cortical slice. A total of 2-20 optical sections were sampled at different depths (2-10 μm interval). Each image was averaged three or four times to increase the signal/noise ratio. At the end of the experiment, images were taken at lower magnification (4× objective) to determine the locations of axons in the cortical explants.
Labeled axons were reconstructed using NIH Image. They were drawn only if they met the following criteria: (1) they elongated between 100 and 400 μm from the pial surface and had a length of ≥500 μm; (2) they originated from transfected cell clusters or single-labeled pyramidal cells within 100-300 μm from the pial surface; (3) they were brightly labeled and readily distinguishable from others; and (4) they appeared healthy (as determined by the absence of notable membrane blebbing in the cell body, axons, or dendrites). All measurements of axonal morphology, such as branch length and total arbor length, were performed using custom NIH Image software macros. Small processes (<5 μm) from the axon shaft were excluded from the analysis, because lamellipodia- or varicosity-like structures within this size range were often observed. The following parameters were measured: the number of branch points; the axon length defined as the distance from the cell body to the most distal branch tip within upper layers; the branch point density of the arbor (number of branch points per total arbor length); the lengths of individual branch tip segments measured from the branch tips to their nearest branch points; and the order of branch tip segments. For the order of branch tip segments, the unbranched axon has a branch order of zero. Branches originating from the first branch point have a branch order of 1; those branches from the next branch point to the first branch point have branch order of 2; and so on. The maximum order was measured.
Time-lapse imaging. To observe dynamics of axon branching, the culture dishes (Millicell-CM PICMORG50; Millipore) were mounted on a stage of the confocal microscope (Yamamoto et al., 1997). Labeled axons were observed quickly, and then the cultures were put back into the incubator. These observations were performed every 24 hr for 3-7 d. Afterward, similar quantitative analysis was performed (see above). Branch dynamics (growth or retraction) were measured by comparing branches at a given time point with those after 24 hr. The branches that grew or retracted <5 μm for 24 hr were classified as stable.
Recording and quantitative analysis of spontaneous activity using MEDs. To record spontaneous activity of cortical cells, we used techniques similar to those used for organotypic hippocampal slice cultures (Shimono et al., 2002). Spontaneous activity in cultures was recorded through MEDs twice per day at 10 min intervals for 3 min or once per day for 15 min. MEDs have an array of 64 planar microelectrodes arranged in an 8 × 8 pattern. The size of the recording electrodes is 50 × 50 μm with ∼22 kΩ input resistance. The extracellular voltages at the electrodes were amplified by a factor of 10 using the 8-Channel Head Amplifier (SH-MED8; Alpha MED Sciences), further amplified by the 8-Channel main amplifier (bandwidth filter, 0.1-20 kHz; SU-MED8; Alpha MED Sciences), and digitized at 5 kHz by a 12-bit data acquisition board (DigiData 1200; Axon Instruments, Foster City, CA). The digitized data were analyzed using AxoScope software (Axon Instruments).
During electrophysiological recording, the slices cultured on the MED probe were placed in a small CO2 incubator (model 4020; Asahi Life-science, Tokyo, Japan) with humidified 5% CO2 and 95% air at 37°C. Potentials were recorded at the electrodes positioned between 100 and 300 μm from the pial surface. An electrode far from the slice was used as a reference electrode. The noise level on each electrode was within the range of 10-20 μV. The frequency of spontaneous firing activity was quantified with Axograph software (Axon Instruments). Negative potentials with amplitudes above a set threshold were counted as one spike. The threshold of 1.5 times the maximal amplitude of the baseline noise (20 μV) was used.
Pharmacological treatments. To manipulate firing and synaptic activity, culture medium containing the following drugs was used between 7 and 15 DIV: d-APV (100 μm), (+)-5-methyl-10,11-dihydro-5H-dibenzo [a,d] cyclohepten-5,10-imine maleate (MK801; 10 μm), DNQX (20 μm), (±)-4-(4-aminophenyl)-1,2-dihydro-1-methyl-2-propylcarbamoyl-6,7-methylenedioxyphthalazine (SYM2206; 100 μm), (αS)-α-amino-3-{(4-carboxyphenyl)methyl}-3,4-dihydro-5-iodo-2,4-dioxo-1(2H)-pyrimidinepropanoic acid (UBP301; 100 μm; Tocris Cookson, Bristol, UK), and TTX (100 nm; Seikagaku-Kogyo, Toyko, Japan). Drug-containing media were exchanged every other day.
Statistical analysis. All statistical values obtained from the morphological and electrophysiological experiments were presented as mean ± SEM. For all features measured, evaluation of the differences among groups was obtained by comparing groups two at a time, using the Mann-Whitney U test for nonparametric statistical analysis.
Histology. The cultures were fixed with 4% paraformaldehyde in 0.1 m phosphate buffer after the observations and kept in 30% sucrose in PBS. Cultures were embedded on an agar block (4% agar in PBS), cut into 20 μm frozen sections, and stained with methylene blue to compare cell density and size in the cortical explants.
Results
Cellular morphology of upper layer cells labeled with YFP
Horizontally extending axons in the upper layers of the visual cortex were studied in organotypic slice cultures. Rat cortical neurons that are destined for layer 2/3 are still migrating at P0, when cortical slices were transferred to the culture conditions. Previous studies have demonstrated that radial migration takes place in vitro (Yamamoto et al., 1992; Roberts et al., 1993) and that laminar configuration is settled approximately 1 week in vitro, which resembles the in vivo situation (Yamamoto et al., 1992). Therefore, cortical cells in the upper layers (100-300 μm from the pial surface) were transfected after 5-6 d in culture with pCAGGS:EYFP to visualize axons originating from these neurons (see Materials and Methods). As shown in Figure 1, A and C, axons from the upper layers were labeled clearly as well as pyramidal-shaped soma and dendrites even >1 week after electroporation. The primary axons run radially toward the white matter side, whereas collaterals elongated in the upper and deep layers. Thus, the projection pattern of such cells was notably similar to that described for layer 2/3 pyramidal cells in the cortex of rodents and the higher mammals, although the number of branches is much larger in the higher mammals (Gilbert and Wiesel, 1979; Rockland and Lund, 1982; Burkhalter and Charles, 1990; McGuire et al., 1991; Lohmann and Rorig, 1994; Durack and Katz, 1996; van Brederode et al., 2000).
Branching of horizontal axons
Branching of horizontal axons was studied after 1-3 weeks in culture. Most axons extended within the upper layers without forming extensive branches for the first week in vitro. This was also true for horizontal axon morphology in vivo during development. YFP labeling in acute cortical slices, which were dissected from 7-d-old rat visual cortex, demonstrated that horizontal axons had no branches at this stage (Fig. 1B). After 8-10 DIV, branches emerged and the primary axons stopped growing in most cases (Fig. 1C). Afterward, axonal arbors became larger and more complex. Time-lapse imaging further demonstrated a dynamic feature of branch formation (Fig. 2A). Some branches were dynamic, either growing or retracting between imaging sessions, whereas others were relatively stable. Some varicosities were also found along the axons, and their positions appeared to change during culturing.
Axonal arbor dynamics were evaluated by measuring the number of branches that were added or eliminated and the number of preexisting branches that lengthened or shortened compared with the previous day. Throughout the observation period, the percentage of growing branches, which included newly added branches and lengthened branches, tended to be higher than that of retracting and eliminated branches (Fig. 2B). After ∼2 weeks in vitro, ∼20% of all branches had become stable. Thus, branching of horizontal axons in cortical upper layers was a continuous and highly dynamic process during this second week in vitro but was relatively stable afterward.
Based on these developmental observations, branching parameters such as the number of branches were observed after 13-15 d in vitro, when stable branches appeared frequently (Fig. 2B). Figure 3 shows several examples of horizontal axons traced after 13-15 d in vitro. The mean number of branch points per axons was 4.5 ± 0.6 (n = 42), ranging from 0 to 17 branch points per axon. The average length of horizontal axons in culture was 1540 ± 236 μm (n = 12 axons traced back perfectly to the cell body). These values were comparable with those for horizontal axons in 2- to 3-week-old rat cortex (Lohmann and Rorig, 1994; van Brederode et al., 2000). Thus, the branch morphology of horizontal axons closely resembled that of arbors in vivo.
Spontaneous firing in cultured cortex
Because there are no extrinsic inputs to these cultured cortical explants, any neuronal activity must arise from the spontaneous firing of neurons in the culture. Spontaneous firing was observed as single-unit or field-potential activity (Fig. 4A,B). The single-unit activity lasted a few milli-seconds in duration and most likely represents action potentials in individual cells (Fig. 4A). The field-potential activity was a negative potential of ∼100 msec in duration and variable amplitude (Fig. 4B) and is probably caused primarily by synchronized firing and synaptic potentials of many cells. In addition, bursting activity was often observed (burst duration, 0.2-2 sec; burst interval, 1-60 sec; firing frequency, 3-30 Hz). This burst activity was detected as single units on a background of field-potential activity. To investigate the spatial distribution of spontaneous patterns of activity in cortical slices, activity was simultaneously recorded at several locations between 100 and 300 μm from the pial surface. Spontaneous firing was highly synchronized among multiple recording sites tested (data not shown) and occurred to a similar extent in every location. Fundamentally, synchronized activity was found across the entire cortical slice. Therefore, the activity recorded with one or two electrodes around the upper layers was considered to reflect the typical activity of cortical cells in each slice.
Development of spontaneous activity was examined by monitoring the same samples repeatedly over many days. Figure 4C shows the activity recorded after 6, 10, and 13 DIV. Spontaneous activity was infrequent before 1 week in vitro but was prominent at later times. In most cases, the spontaneous activity appeared rhythmically with a long silent period, which was variable among samples. To quantify the activity levels in the organotypic cultures, the number of negative potentials, including field potentials, was counted for 6-15 min per day. As shown in Figure 4D, the mean firing rate in the early stages was quite low (0.18 ± 0.10 Hz; n = 10 for 5-8 DIV) but increased strikingly by 11 DIV and further increased and was maintained at later stages (1.27 ± 0.19 Hz; n = 15 for 9-14 DIV).
The effects of spontaneous activity in axonal branching
The above results suggest that the developmental time course of spontaneous activity correlates with that for branch formation, raising the possibility that spontaneous activity might affect axonal branching. First, the onset of spontaneous firing preceded the emergence of horizontal axon branching by 1 or 2 d (compare Figs. 2B, 4D). Second, the amount of branch growth correlated with the increase of spontaneous firing. The growth ratio, defined as the ratio of the number of growing branches to the number of retracting branches, was moderately correlated with the increased level of spontaneous activity 1 d before (r = 0.43).
To further test how firing activity influences branch formation, blockers of action potential firing or glutamatergic synaptic transmission were applied to the culture medium during the second week in vitro, when branch elaboration normally occurs.
The efficacy of chronically applied drugs was examined by recording changes in activity through the MED electrodes. The addition of TTX or APV/DNQX to the culture medium quickly caused a complete cessation of spontaneous activity. This effect persisted to the final day of recording (Fig. 5). Likewise, spontaneous activity was diminished after the addition of DNQX or APV but began to reappear after a few days. Although firing rates further increased at later stages, the activity between 8 and 14 DIV did not differ substantially from that in the control cultures (Fig. 5B) (control: 1.1 ± 0.17 Hz, n = 15; DNQX: 1.1 ± 0.50 Hz, n = 4, p > 0.1; APV: 1.7 ± 0.32 Hz, n = 4, p > 0.1). The mechanism for this recovery is not known, but homeostatic regulation including receptor expression may account for this phenomenon (Turrigiano et al., 1998; Luthi et al., 2001). However, spontaneous activity in the presence of DNQX or APV each showed distinct patterns. Spontaneous activity under chronic DNQX application tended to show similar but slightly longer burst durations than controls (burst duration, 0.4-3 sec; burst interval, 1-30 sec; firing frequency, 3-15 Hz). The APV-treated group had a lower incidence of spontaneous bursting but much longer burst durations than the control group (burst duration, 1-10 sec; burst interval, 1-30 sec; firing frequency, 3-20 Hz).
Axonal branching at 13-15 d in cultures treated with these blockers was compared with controls. These pharmacological treatments did not appear to influence cell survival or the overall health of the cultures. Nissl-stained sections of the treated cultures (two to three slices for each treatment) revealed no differences in either cell size or density compared with control cultures (data not shown).
As shown in Figure 6, TTX, APV/DNQX, and DNQX treatments each resulted in a decrease in branch number compared with control axons. The number of branch points is an index of the complexity of axonal branching. As described above, the mean number of branch points in untreated axons was 4.5 ± 0.6 (n = 42). Under TTX, APV/DNQX, and DNQX treatments, the number of branch points fell to one-third to one-half of the value in the untreated control group (Fig. 7A) (TTX: 2.1 ± 0.7, n = 14, p < 0.05; APV/DNQX: 1.5 ± 0.4, n = 11, p < 0.05; DNQX: 1.7 ± 0.5, n = 16, p < 0.005). In contrast, APV did not cause a significant difference in branch number, which, if anything, was slightly increased (Fig. 7A) (5.7 ± 1.2; n = 20; p > 0.1).
To determine whether these decreases in the number of branch points were a simple consequence of differences in total arbor length, we calculated the density of branch points, defined as the number of branch points per 1 mm length. The mean density of branch points under control conditions was 3.3 ± 0.4 (n = 42) branches per millimeter. TTX, APV/DNQX, and DNQX decreased the density of branch points to nearly half that of controls (Fig. 7B) (TTX: 1.7 ± 0.5, n = 14, p < 0.05; APV/DNQX: 1.6 ± 0.4, n = 11, p < 0.05; DNQX: 1.91 ± 0.3, n = 16, p < 0.005). APV treatment did not affect branch density (Fig. 7B) (4.4 ± 1.2; n = 20; p > 0.1). These results indicate that the changes in the number of branch points cannot be attributed exclusively to a change in total arbor lengths.
It appeared that short branch tip segments were selectively lost in the TTX- or DNQX-treated axons (Fig. 6). As shown in Figure 7C, the mean branch tip segment length in TTX- and DNQX-treated axons was 303 ± 69 μm (n = 43) and 306 ± 70 μm (n = 40), respectively. They were significantly (p < 0.005 and p < 0.05) larger than that in the controls (109 ± 12 μm; n = 233). The average branch tip segment length in APV/DNQX treatment was also increased by approximately threefold, although this failed to reach statistical significance compared with controls (330 ± 128 μm; n = 28; p > 0.1) because of high variability in branch tip segment lengths. APV-treated axons showed no difference from controls (141 ± 23 μm; n = 131; p > 0.1).
We also examined the branch order of branch tip segments. As with the branch number and the branch density, the maximum order of branch tip segments under TTX, APV/DNQX, and DNQX treatments fell to one-third to one-half of the value in the untreated control group (control: 3.9 ± 0.5, n = 42; TTX: 2.0 ± 0.6, n = 14, p < 0.05; APV/DNQX: 1.4 ± 0.3, n = 11, p < 0.01; DNQX: 1.6 ± 0.5, n = 16, p < 0.005). In contrast, APV did not cause a significant difference in the maximum branch order (4.6 ± 0.9; n = 20; p > 0.1).
One explanation for these results is that transmission through non-NMDA-type glutamate receptors exerts branch-promoting effects. To further examine this possibility, the specific and non-competitive antagonist of AMPA receptors, SYM2206, or the potent kainate receptor antagonist UBP301 was added to the culture medium. Both drugs considerably reduced spontaneous activity, and this effect persisted to the final day of recording (the firing rate between 8 and 14 DIV: 0.15 ± 0.09 Hz; n = 2 for SYM2206; 0.34 ± 0.09 Hz; n = 2 for UBP301). Under these treatments, branch formation of horizontal axons was examined. SYM2206 treatment completely reduced the number of axonal branches and increased branch tip length, similar to that for DNQX treatment (branch number: 1.4 ± 0.4, n = 12, p < 0.005; branch density: 1.1 ± 0.3, n = 12, p < 0.005; branch tip segment length: 243 ± 48 μm, n = 26, p < 0.005; maximum branch order: 1.3 ± 0.3, n = 12, p < 0.01). In contrast, UBP301 treatment did not produce notable differences from control (branch number: 5.6 ± 1.1, n = 13, p > 0.1; branch density: 2.5 ± 0.4, n = 13, p > 0.1; maximum branch order: 4.2 ± 0.7, n = 12, p > 0.1), except that tip segment length was slightly larger (143 ± 20 μm; n = 85; p < 0.05).
Discussion
Using an in vitro culture that preserves many properties of the intact brain, we have demonstrated that axons extending horizontally in the upper layers formed axonal arbors that were remarkably similar to those of horizontal axons in vivo. Electrophysiological recordings showed that the time of axonal branch emergence corresponded very closely to the onset of spontaneous activity. Moreover, blockade of spontaneous firing activity or non-NMDA transmission, but not NMDA transmission, resulted in decreased axonal branching. These findings suggest that firing activity, perhaps through the synaptic activation of non-NMDA receptors, in particular AMPA-type receptors, plays a role in promoting branch formation of horizontal axons.
Comparison of axons of upper layer neurons in organotypic slice cultures with horizontal axons in vivo
Previous studies have shown that neuronal migration, layer formation, cellular morphology, and physiology develop normally in cultured cortical explants (Yamamoto et al., 1989, 1992; Bolz et al., 1990, 1992; Molnar and Blakemore, 1991; Gotz and Bolz, 1992; Annis et al., 1993; Dantzker and Callaway, 1998). Similarly, individually YFP-labeled cells in our slice cultures exhibited a projection pattern and branch formation that is typical for layer 2/3 cells in rodents (Lohmann and Rorig, 1994; van Brederode et al., 2000). These findings indicate that some intrinsic properties of upper layer neurons including branching properties and the surrounding cues including lamina-specific molecules are preserved and functional under culture conditions (Yamamoto et al., 1992, 1997, 2000; Yamamoto, 2002). As for axonal elongation, upper layer cell axons extended 1.5 mm, on average, which is almost the same as that found in vivo (Burkhalter and Charles, 1990; Lohmann and Rorig, 1994; van Brederode et al., 2000). Thus, such intrinsic components might also regulate horizontal axon elongation, which might link functional columns as found in carnivores (Ts'o et al., 1986; Gilbert and Wiesel, 1989).
Neuronal activity in cultured cortical slices also seems to mature at a similar rate to that in vivo. Indeed, it has been reported that spontaneous firing is extremely rare in P7 cortical slices but increases at later stages to resemble the pattern of activity observed in the adult (Flint et al., 1997). Moreover, in freely moving rats, firing rates in the cortex increase to ∼1 Hz at P11 and P12 (Mirmiran and Corner, 1982). These similarities in developmental timing of spontaneous activity between the intact and cultured cortex suggest that intrinsic electrical properties such as membrane properties and the balance in the growth of inhibitory and excitatory systems also might be preserved in cultured cortical slices.
Involvement of molecular cues in axonal branching of layer 2/3 pyramidal neurons
The fact that a small number of branches still form in the absence of synaptic transmission or action potentials indicates that some factors independent of neuronal activity also contribute to branch formation of horizontal axons (Dantzker and Callaway, 1998; Borrell and Callaway, 2002). For example, a branch-inducing factor may be expressed in the upper layers that acts on horizontal axons (Yamamoto, 2002). For cortical circuitry, it has been reported that ephrin-A5 is expressed in the middle layer, where it promotes branching of layer 6 neurons (Castellani et al., 1998). Szebenyi et al. (2001) have demonstrated that FGF-2 induces axonal branches from cortical neurons, although its expression pattern is not known in vivo. Thus, such branch-promoting factors for horizontal axons would be predicted to be upregulated starting the second postnatal week. It is also possible that intrinsic properties of upper layer neurons, such as axonal length or developmental timing, influence branching in a cell-autonomous manner (Szebenyi et al., 1998).
Dynamics of axonal branching
The present study is the first to demonstrate dynamic remodeling of axon terminal arbors in the mammalian brain, although a similar mode of axonal arborization has been seen in fish and frog retinotectal projections (O'Rourke and Fraser, 1990; Kaethner and Stuermer, 1992; O'Rourke et al., 1994; Cohen-Cory, 1999). In the retinotectal system as well as in our study, a larger proportion of branches were observed undergoing addition than retraction, although the exact proportion of branches added or lost differs between cortical and retinal axons (O'Rourke and Fraser, 1990; O'Rourke et al., 1994). This difference reflects the numerous differences between these systems, including temperature, locally expressed branching factors, and intrinsic properties of the neurons themselves.
The dynamic branching of axonal arbors may correlate with the formation of synapses with target cells. In the Xenopus visual system, it was demonstrated that as axons arborized, new synapses appeared rapidly and that axonal branch points were preferentially located at synaptic sites (Alsina et al., 2001). In our study, axonal varicosities, which are thought to reflect synapses, were present along the axons (Fig. 2A). Furthermore, the number of varicosities correlated positively with the number of branch points at least in those axons on which swellings were clearly observed (data not shown). These results suggest that dynamic branch formation could reflect changes in synaptic connectivity and that presynaptic and postsynaptic interactions through firing and synaptic activity may affect branching.
Possible mechanism of activity-dependent axonal branching
Our results show that suppression of neuronal activity, specifically synaptic activity through AMPA receptors, inhibits branch formation, including the branch number and the branch length. This finding contrasts with numerous observations in vivo in which activity blockade causes an expansion of axonal arbors in retinotectal (Cohen-Cory, 1999), retinogeniculate (Sretavan et al., 1988), and thalamocortical projections (Antonini and Stryker, 1993). Although laminar organization, synaptic maturation, and neuronal excitability in our organotypic culture closely mimics that in the intact brain, one aspect that distinguishes our preparation from the in vivo condition is the absence of external topographic inputs driven directly or indirectly by the sensory periphery. Therefore, it is reasonable to infer that activity blockade in vivo would interfere with activity-dependent competitive interactions. In contrast, horizontally extending cortical axons in our cultures, which have highly synchronized firing over long distances, exhibit activity patterns that may be less likely to favor competitive interactions. Different firing patterns of activity are likely to have different effects on neural circuitry (Itoh et al., 1997), especially in situations in which patterned activity might promote competition among inputs (Crair, 1999). It has been reported that neural activity affects dendritic branching as well (McAllister et al., 1996). Interestingly, activity blockade has been reported in the retinotectal system to produce different effects on axonal and dendritic dynamics, consistent with the differences between our data on axon morphology and published data on dendrites (Rajan and Cline, 1998; Rajan et al., 1999).
One surprising result was that chronic pharmacological blockade of AMPA-type glutamate receptors, but not of NMDA receptors, reduced the number of axonal branch points by an amount comparable with activity blockade by TTX or APV/DNQX treatment. Importantly, our use of MED recordings of cultures treated chronically with APV or DNQX revealed in both cases a homeostatic recovery of spontaneous activity to control levels within ∼5 d. Thus, the failure of axon arbors to branch extensively when AMPA receptors are selectively blocked cannot be attributed to a lack of spontaneous activity in these cultures. Furthermore, treatment with MK-801, another antagonist of NMDA receptors, like APV did not cause a significant change in branch point number or in mean branch tip segment length compared with untreated controls (data not shown). These observations support the view that neuronal activity increases branch formation by causing synaptic transmission through non-NMDA receptors.
Although these data support a key role for AMPA receptor activation in horizontal axon branching, they do not entirely preclude the participation of NMDA receptors in this process. NMDA receptors could work in a coordinated manner with non-NMDA receptors. For example, if the primary contribution of the NMDA receptors in axonal morphogenesis were to cause branch elimination under conditions of weak NMDA receptor activation, as occurs at hyperpolarized potentials, AMPA receptor blockade could promote this. Nonetheless, the fact that DNQX (or SYM2206) alone is as effective as the combined application of DNQX and APV at reducing branch complexity suggests that AMPA receptor activation is specifically required for proper axonal branching.
There is considerable evidence that NMDA receptors are required for certain aspects of development and topographic refinement of axonal arbors (Cline and Constantine-Paton, 1990; O'Rourke et al., 1994; Mitrovic et al., 1996; Rajan et al., 1999; Ruthazer et al., 2003). For example, retinal axons from two eyes, made to innervate the same optic tectal lobe in frogs, segregate during development, but this segregation can be abolished by blockade of NMDA receptors (Cline and Constantine-Paton, 1990). Suppression of NMDA receptor function also prevents the development of orientation selectivity in the visual cortex (Ramoa et al., 2001), a property that correlates anatomically and developmentally with the structural development of horizontal axons (Lowel and Singer, 1992; Ruthazer and Stryker, 1996). However, we found that blocking spontaneous non-NMDA receptor activation rather than NMDA receptor activation dramatically reduced axonal branching of cortical upper layer neurons during their initial period of branch formation. These results support the idea that NMDA receptors may play different anatomical and functional roles in different systems (Fox et al., 1989; Smetters et al., 1994; Hahm et al., 1999; Yamada et al., 2000).
An alternative possibility involves calcium influx through AMPA receptors, which may be critical for guiding neural circuit formation in the mammalian CNS (Katz and Shatz, 1996; Gomez and Spitzer, 2000; Chang and De Camilli, 2001; Tashiro et al., 2003). AMPA receptor activation could cause a depolarization, which in turn might open a voltage-dependent calcium channel (VDCC). Furthermore, it has been shown that AMPA receptors lacking glutamate receptor 2 are highly permeable to calcium (Hollmann et al., 1991; Verdoorn et al., 1991). It has been suggested that calcium-permeable AMPA receptors could induce NMDA receptor-independent long-term potentiation, which might contribute to neural circuit formation (Jia et al., 1996). Therefore, calcium entry through VDCC or calcium-permeable AMPA receptors could participate directly in activity-dependent remodeling of horizontal axons.
Footnotes
This work was supported by Grants-in-Aid for Scientific Research 11680788 and 12050227 from the Japanese Ministry of Education, Culture, Sports, Science and Technology and by Human Frontier Science Project RGP0107/2001. We thank Maki Nishiwaki for assistance in labeling with fluorescent protein.
Correspondence should be addressed to Dr. Nobuhiko Yamamoto, Neuroscience Laboratories, Graduate School of Frontier Biosciences, Osaka University, Suita, Osaka 565-0871, Japan. E-mail: nobuhiko{at}fbs.osaka-u.ac.jp.
Copyright © 2005 Society for Neuroscience 0270-6474/05/250001-09$15.00/0