Role of Electrical Activity in Horizontal Axon Growth in the Developing Cortex: A Time-Lapse Study Using Optogenetic Stimulation

During development, layer 2/3 neurons in the neocortex extend their axons horizontally, within the same layers, and stop growing at appropriate locations to form branches and synaptic connections. Firing and synaptic activity are thought to be involved in this process, but how neuronal activity regulates axonal growth is not clear. Here, we studied axonal growth of layer 2/3 neurons by exciting cell bodies or axonal processes in organotypic slice cultures of the rat cortex. For neuronal stimulation and morphological observation, plasmids encoding channelrhodopsin-2 (ChR2) and DsRed were coelectroporated into a small number of layer 2/3 cells. Firing activity induced by photostimulation (475 nm) was confirmed by whole-cell patch recording. Axonal growth was observed by time-lapse confocal microscopy, using a different excitation wavelength (560 nm), at 10–20-min intervals for several hours. During the first week in vitro, when spontaneous neuronal activity is low, DsRed- and ChR2-expressing axons grew at a constant rate. When high-frequency photostimulation (4 or 10 Hz) for 1 min was applied to the soma or axon, most axons paused in their growth. In contrast, lower-frequency stimulation did not elicit this pause behavior. Moreover, in the presence of tetrodotoxin, even high-frequency stimulation did not cause axonal growth to pause. These results indicate that increasing firing activity during development suppresses axon growth, suggesting the importance of neuronal activity for the formation of horizontal connections.


Introduction
Neocortical circuits are known to be established not only by genetic mechanisms but also by environmental factors such as sensory-evoked neuronal activity. Among the best examples of this formation process are the projections from the thalamic nuclei to the cortices [1]. Intrinsic cortical circuits are also thought to be established based on genetic and activitydependent mechanisms. Axon collaterals from layer 2/3 pyramidal neurons, called horizontal axons, connect cells within the same layer, and these connections provide a substrate for lateral interactions across cortical columns [2][3][4][5][6][7][8]. In the visual cortex of higher mammals, the horizontal connections primarily link cortical cells in columns having the same orientation [9][10][11][12][13][14][15]. Horizontal connections have also been found in other cortical areas, and are the basis of information processing in the neocortex [16][17][18][19][20][21][22][23].
During development, horizontal axons start to grow in all directions from the cell body, and then display further elongation and initial branch formation at later periods [24,25]. Ultimately, only specific locations undergo robust and complex branch formation, whereas other regions show selective retraction of inappropriate connections. This developmental scheme is similar among different mammalian species, albeit with some differences in the length of maturation period [24][25][26][27][28].
Neuronal activity is essential for the formation of horizontal connections, since the organization of these connections is disrupted by sensory deprivation and by blockade of action potentials [29][30][31][32]. Furthermore, our previous in vitro studies have demonstrated that firing and synaptic activity promote horizontal axon branching [33,34]. Nevertheless, how electrical activity contributes to the regulation of horizontal axon growth remains unknown.
To address this issue, here we studied axon growth of layer 2/3 cells in organotypic cortical slice culture by manipulating firing activity using an optogenetic technique [35][36][37][38]. Light-gated algal channel channelrhodopsin-2 (ChR2) enables neuronal activity to be generated in a cell-specific way with high temporal resolution and is therefore an excellent tool to study the complex mechanism of cortical network formation. The results revealed that horizontally elongating axons pause in their growth after high-frequency stimulation, suggesting that developing firing activity contributes to circuit formation by regulating axonal growth.

Animals and Ethics Statement
Sprague-Dawley rats were used for this study. All experiments were performed according to guidelines laid down by the animal welfare committees of Osaka University and the Japanese Neuroscience Society. The protocol was approved by the Committee on the Ethics of Animal Experiments of Osaka University (Permit Number: FBS 07-037).

Organotypic Slice Culture
To observe horizontal axons, we used a previously described organotypic slice culture system [39,40]. In this culture, cortical circuits including horizontal connections are reproduced with the cortical laminar structure preserved [33]. In brief, the occipital region of the cortex was dissected from postnatal day 0 (P0) or P1 rats. The slices were placed on a membrane filter (MilliCell-CM Low Height PICMORG-50, Millipore, Billerica, MA) coated with rat tail collagen. The culture medium consisted of a 1:1 mixture of Dulbecco's modified essential medium and Ham's F12 (Life Technologies, Carlsbad, CA) with several supplements including insulin and transferrin [40]. The cultures were maintained at 37uC in an environment of humidified 95% air and 5% CO 2 .

Plasmids
A plasmid containing hChR2(H134R)-EYFP was obtained from Dr. Karl Deisseroth, Stanford University [41], and the coding region was placed in pCAGGS vector (a generous gift from Dr. Naofumi Uesaka, University of Tokyo) [42]. The plasmid was purified using a MaxiPrep kit (PureLink, Life Technologies, Carlsbad, CA) according to the supplier's protocol, and dissolved in Hanks' solution at appropriate concentrations. pCAGGS-DsRed and pCAGGS-EYFP plasmids were prepared in a similar way.

Morphological Time-lapse Observation of Growing Axons
At 3-10 days in vitro (DIV), a culture dish containing cortical slices was sealed with a coverslip and transferred to a microscope stage. During the 4-5-h observation, the slices were maintained at 37uC using a heating apparatus (Kokensha Engineering, Tokyo, Japan) and supplied with a 95% air and 5% CO 2 humidified gas mixture [43]. The objective lens was coiled with a band-heater (Thermoplate, Tokai Hit, Shizuoka, Japan) to prevent condensation.
Morphological observation was performed using a confocal laser scanning microscope (Nikon C2, Nikon Instech, Tokyo, Japan) with a 10x objective lens (Nikon Plan Fluor, NA, 0.3) at 560-nm excitation (solid-state laser). After the cultures had been incubated for at least 1 h, images were acquired at 10-20-min intervals with 3-5-mm steps (3-4 optical sections). The observed axons were mostly traced back to cell bodies. Axons were followed for at least 30 min before photostimulation (see below). In both control and experimental conditions, 560-nm wavelength excitation light was used to observe growth cone behavior.

Photostimulation
A solid-state illuminator (475 nm peak wavelength; maximal power: 20 mW; Lumencor SPECTRA, Lumencor, Beaverton, OR) was used for photostimulation through the objective lens. The beam position was controlled manually through the microscope stage. The intensity of the light stimulus falling into the beamfocused area of a cortical slice was adjusted using a Lumencor Remote Control Accessory and measured with a light power meter (Laser Scan, Coherent, Santa Clara, CA). The duration and frequency of the light stimulus were controlled by applying square electrical pulses from a pulse generator (Master-8, AMPI, Jerusalem, Israel).
Axonal growth was observed by applying light stimuli with different intensities to DsRed/ChR2-EYFP-cotransfected cells. We confirmed that a light intensity of 1.4 mW/mm 2 was efficient in producing action potentials in the transfected cells (see below) without causing any photodynamic damage.

Patch-clamp Experiments
To confirm that the photostimulation induced action potentials in ChR2-expressing cells, whole-cell patch-clamp recording was carried out in slice cultures at room temperature. ChR2-EYFPexpressing cells were identified using an epifluorescence microscope (BX51W1, Olympus, Tokyo, Japan) equipped with a 40x water immersion objective (Olympus, NA, 0.8). The excitation light beam was introduced through the objective lens. The minimal light intensity for ChR2 excitation was determined by applying blue light of various intensities.
Resistance of patch pipettes was 7-8 MV when they were filled with an intracellular solution containing (in mM) 130 potassium methanesulfonate (CH 3  The membrane voltage was recorded with a Multiclamp 700 A amplifier (Molecular Devices, Sunnyvale, CA), low-pass filtered at 5 kHZ, digitally sampled at 10-20 kHz, and monitored with pCLAMP 9.0 (Molecular Devices) and analyzed off-line using either Clampfit 9.0 or AxoScope 10.2 software.

Pharmacological Drug Application
To suppress firing activity, 1 mM TTX (Seikagaku-Kogyo, Tokyo, Japan) was added 1 h before the time-lapse experiment.

Image Processing and Statistical Analysis
Axon growth in the time-lapse study was measured using image processing software (ImageJ 1.45 s). Values are presented with 6 standard error of the mean (SEM). Quantitative analysis (Student's t-test) was performed using Microsoft Excel software.

Axon Growth of Layer 2/3 Neurons in Organotypic Slice Cultures
Axonal growth of cortical cells was studied using an organotypic slice culture technique, which allowed us to perform gene manipulation and subsequent time-lapse observation easily. Furthermore, we could assess axon behavior in a cellular environment which is close to that in vivo, as cortical cell arrangement such as laminar structure is well preserved in this culture system [40].
A small number of upper layer cells were cotransfected with DsRed and ChR2-EYFP plasmids. ChR2-EYFP was used for optogenetic stimulation (475 nm), while DsRed was used to observe axonal growth under a different excitation light (560 nm) without activating ChR2. One day after electroporation, we detected strong and stable expression of DsRed/ChR2-EYFP in several cells. DsRed/ChR2-EYFP was highly expressed in the soma and axons, although the fluorescence intensity was slightly weaker in axons. After several days in culture, horizontally elongating axons were observed originating from pyramidal cells ( Figure 1A), which is consistent with previous results [33,34].
To examine their growth pattern, labeled axons were monitored every 10 min for several hours by time-lapse confocal microscopy with an excitation wavelength of 560 nm. Horizontally elongating axons from cells located at 100 to 300 mm (presumed layer 2/3) from the pial surface were selected for the observation ( Figure 1B) [40]. We found that horizontally elongating axons grew continuously from 3 to 10 DIV, although the growth rates varied substantially among axons ( Figure 1C, D, E). A quantitative analysis showed that the average growth rate was 29.961.7 mm/h (n = 32, Figure 1E), with 9.5 as minimal and 46.7 mm/h as maximal values.

Light-induced Action Potentials
Next, we studied the effect of firing activity on axon growth by applying photostimulation at around one week in culture, when firing activity scarcely takes place [33].
To confirm that photostimulation produces firing activity in ChR2-expressing cells, whole-cell patch-clamp recording was performed ( Figure 2). Photostimulation with different durations (50 or 200 msec) was applied to the ChR2-expressing cells. Each 50-msec flash of light elicited firing ( Figure 2C, upper). A longer pulse duration (200 msec) also resulted in a single action potential ( Figure 2C, lower).
No spike response was detectable in non-transfected neurons, indicating that the irradiance alone does not elicit a response, which is consistent with previous results [36,44,45]. Thus, we confirmed that photoactivation efficiently generated firing activity specifically in ChR2-expressing neurons.

The Response of Horizontal Axons to High-frequency Photostimulation
Next, the behavior of horizontal axons was studied in response to photostimulation. The stimulation paradigm consisted of a 1min photostimulation period with different frequencies (0.1, 1, 4 and 10 Hz).
First, high-frequency photostimulation was applied to the cell bodies which were the origins of labeled axons, as ChR2 was strongly expressed in the soma ( Figure 1A). Before photostimulation, all observed axons were confirmed to elongate at constant growth rates for 30-60 min (35.163.5 mm/h, n = 11). We found that most of them transiently stopped growing after highfrequency photostimulation (4 and 10 Hz) ( Figure 3A-C). A few axons showed obvious retraction. To quantify this axonal behavior, growth rates were calculated after photostimulation. The average growth rate was 0.4769.9 mm/h (n = 11) at 10 min after photostimulation, significantly slower than the average before stimulation (Student's t-test, p,0.005). Some axons started to grow again 20 min after photostimulation, but the average growth rate was still significantly lower (19.065.5 mm/h, p,0.05). Axonal growth rate was fully restored at 30 min after photostimulation (34.265.6 mm/h, Figure 3D).
Based on the above quantitative analysis, each axon's behavior was categorized into pause and non-pause behavior. Because the 46% reduction in growth rate at 20 min after photostimulation ( Figure 3D) was statistically significant, axons which exhibited more than a 50% decrease at this time point were defined as displaying pause behavior. Consequently, 9 of the 11 axons examined showed pause behavior. At 30 min after photostimulation, most of these (7 of the 9 axons) displayed fully restored growth rates, although the remaining two axons still exhibited significantly lower growth rates at this time point. All growth cones grew at normal rates by 50 min after photostimulation. We also observed the morphology of the growth cone. All axons exhibited typical growth cones with lamellipodia structures before photostimulation, but immediately after photostimulation the sizes of the lamellipodia decreased ( Figure 3E).
A second time-lapse study was carried out after applying photostimulation directly to the axon, in order to explore whether action potentials generated in axons are required for the pause behavior ( Figure 4A, B). Most of the observed axons exhibited slight retraction in response to photostimulation ( Figure 4C), and quantitative analysis showed a marked decrease in axonal growth rate after photostimulation. The average growth rates (n = 8) at 10, 20, and 30 min after photostimulation were 28.569.0 mm/h, 24.963.9 mm/h, and 23.165.1 mm/h, respectively, displaying a significant reduction compared with that before stimulation (17.962.1 mm/h, Student's t-test, p,0.005). Axonal growth rate was restored by 60 min (19.266.6 mm/h) after stimulation.
These axons were classified into two groups using the definition of pause behavior described above (i.e., more than 50% reduction at 20 min after stimulation). Seven axons out of eight exhibited pause behavior, while one axon was not obviously affected by the high-frequency photostimulation ( Figure 4C). This result suggests that action potentials generated in axons at high frequencies also cause the pause behavior.

Growth Cone Response in TTX-treated Culture
To further confirm that generation and propagation of action potentials are required for the pause behavior, the sodium channel blocker tetrodotoxin (TTX, 1 mM) was added to the culture medium one hour before the time-lapse experiment. Axons also grew normally, with growth cones, in blocker-containing medium. We found that in the presence of TTX, high-frequency photostimulation to the soma (4 Hz for 1 min) did not affect axonal growth rate (18.662.5 mm/h before stimulation, 21.866.8 mm/h at 10 min after stimulation and 14.863.3 mm/ h at 20 min after stimulation, n = 2; Figure 4D). This result indicates that axons stop growing transiently due to firing activity elicited by photostimulation.

The Influence of Low-frequency Photostimulation on Axon Growth
We investigated the influence of frequency of firing activity on axon growth. Photostimulation for 1 min at 1 Hz produced a weak effect on axonal growth ( Figure 5A). The axons tested were classified as displaying pause and non-pause behavior, based on their growth rates before and after photostimulation (see above). Half of the axons did not show any pause (n = 3), whereas the others paused in their growth for 20 min (n = 3; Figure 5A).
Photostimulation for 1 min at 0.1 Hz did not result in any significant effect on the growth of any axon tested ( Figure 5B). The growth rates were 23.567.2 mm/h, 34.769.6 mm/h and 38.465.3 mm/h at 10, 20, and 30 min after stimulation, respectively, and were not significantly different from that before stimulation (26.863.5 mm/h, n = 5; Figure 5B).
The total amount of firing activity may be critical for the pause behavior. To test this possibility, we increased the number of pulses while maintaining the frequency at 0.1 Hz. Even though the number of stimuli (120 pulses) was 20 times larger, axonal growth did not pause after the low-frequency stimulation. However, the same number of pulses applied subsequently to the same axon at high frequency (4 Hz, 120 pulses for 30 sec) elicited pause behavior for the next 40 min ( Figure 5C). Therefore, the pause behavior depends on frequency rather than on the total amount of firing activity.

Axonal Responses to Different Frequencies of Photostimulation
The relationship between frequency of photostimulation and axonal response is shown in Figure 6. Based on the above analysis, high-frequency photostimulation was highly effective for the pause behavior (86% for 4 Hz, n = 7, 75% for 10 Hz, n = 4), intermediate stimulation was partly effective (50% for 1 Hz, n = 6), but low-frequency stimulation (0.1 Hz) had no effect on axonal growth (n = 5).

Discussion
In the present study, we used optogenetic stimulation to investigate the influence of firing activity on cortical axon growth in vitro. The results demonstrate that horizontal axons grow at a constant rate after several days in culture, when spontaneous firing scarcely occurs, but exhibit pause behavior immediately after highfrequency firing is induced (see Figure 6). In the presence of TTX, photostimulation failed to elicit pause behavior. These results imply that high-frequency firing during development is involved in suppressing axon growth of layer 2/3 neurons and contributes to the formation of horizontal connections in the neocortex.
Combining time-lapse observation of a fluorescent protein and ChR2-expressing axons with optogenetic stimulation is also shown to be a powerful method for investigating activity-dependent processes in the brain. Using this approach, we demonstrate that firing activity contributes to the pause behavior and acts as a negative regulator for horizontal axon growth.

Activity-dependence of Horizontal Axon Growth
As shown previously, spontaneous firing activity is almost absent in cortical slice culture during the first week in vitro, but begins to increase at around 10 DIV, reaching its peak at 14 DIV [33]. This  temporal profile is compatible with the development of neuronal activity in vivo [46][47][48]. Taken together, the present findings suggest that horizontal axon growth is suppressed by increasing neuronal activity during cortical circuit formation.
The rapid response of the growth cone to high-frequency firing suggests that the pause behavior is cell-autonomous. Given that new branches tend to emerge during periods when axon growth has ceased, neuronal firing could strengthen the connectivity between cortical cells which are synchronously activated [43,49].
As a consequence, only some of the horizontal axons originating from a layer 2/3 cell could form more extensive branches to make stronger connections with cortical cells in specific locations [24,25,32]. Consistent with this, our previous study has also shown that axonal branching of layer 2/3 cells is promoted by spontaneous activity and subsequent synaptic activity [33]. Overall, the present findings suggest that increasing spike activity during postnatal stages contributes to the establishment of  C, A total of 120 pulses was applied to a single axon at low frequency (0.1 Hz) and subsequently at high frequency (4 Hz). Pause behavior was induced by high-frequency stimulation but not by low-frequency stimulation, even though the number of pulses was the same in both cases. doi:10.1371/journal.pone.0082954.g005 horizontal connections by inhibiting axon growth and promoting branch formation.
However, it should be noted that an intrinsic mechanism also contributes to axon branching of layer 2/3 cells, because some aspects of their branching pattern are highly preserved irrespective of environmental conditions [49,50].

The Efficiency of Neuronal Activity
Even after high-frequency stimulation (4 Hz and 10 Hz), not all of the tested axons showed pause behavior. When intermediatefrequency stimulation (1 Hz) was applied, only half of the examined axons paused growing. As there are subclasses among upper-layer pyramidal cells [23,51], distinct populations may respond differentially to firing activity. Another possibility is that the history of neuronal activity may influence axonal responses, as a growth cone that has experienced electrical stimulation-induced collapse becomes insensitive to a further stimulation [52].

Possible Mechanisms of Growth Cone Pausing in Response to Neuronal Activity
Previous studies have shown that axon growth is arrested in peripheral and invertebrate neurons when the cells are electrically stimulated [53,54]. Moreover, it has been demonstrated that the suppression of growth cone motility is attributable to a rise in Ca 2+ concentration in the growth cones in the PNS [55][56][57]. Similarly, Ca 2+ elevation may be essential for the pause behavior of horizontal axons in the cortex.
Is gene expression involved in the observed pause behavior? In the classical collapse assay, addition of the repulsive guidance molecule Sema-3A is known to induce immediate growth cone collapse, which lasts for 30-60 min [58]. The same phenomenon was observed even in axons that were isolated from the soma. This indicates that the effect is independent of gene expression. As axonal responses in the present study resemble Sema-3A-induced growth cone collapse in temporal terms, it is likely that the pause behavior is also independent of nuclear gene expression. However, local translation in axonal tips may be partly responsible for the pause behavior of cortical axons [59,60].
It has also been shown that electrical activity alters growth cone responsiveness to environmental factors. For instance, light electrical stimulation of Xenopus motoneurons (10 action potentials at 2 Hz), which by itself does not induce any change in axonal outgrowth, is able to lower the sensitivity of growth cones to attractive guidance cues or to convert repulsive cues into attractive ones [61]. Thus, growth cone responses may be affected differentially by the strength of electrical activity. In vitro experiments in developing rat neocortical slices demonstrate the presence of spatiotemporally organized patterns of spontaneous cortical activity [62,63]. It is likely that such unique firing patterns can also influence axon behavior [64,65].
In summary, various intrinsic and extrinsic neuronal events including firing and synaptic activity probably contribute to the decision of an axon to advance, retract, pause, or turn. The present study indicates that firing activity is also an influential factor for axonal development in the mammalian neocortex, which is crucial for the formation of its columnar structure.