Following the ontogeny of retinal waves: pan-retinal recordings of population dynamics in the neonatal mouse

Abstract

The immature retina generates spontaneous waves of spiking activity that sweep across the ganglion cell layer during a limited period of development before the onset of visual experience. The spatiotemporal patterns encoded in the waves are believed to be instructive for the wiring of functional connections throughout the visual system. However, the ontogeny of retinal waves is still poorly documented as a result of the relatively low resolution of conventional recording techniques. Here, we characterize the spatiotemporal features of mouse retinal waves from birth until eye opening in unprecedented detail using a large-scale, dense, 4096-channel multielectrode array that allowed us to record from the entire neonatal retina at near cellular resolution. We found that early cholinergic waves propagate with random trajectories over large areas with low ganglion cell recruitment. They become slower, smaller and denser when GABAA signalling matures, as occurs beyond postnatal day (P) 7. Glutamatergic influences dominate from P10, coinciding with profound changes in activity dynamics. At this time, waves cease to be random and begin to show repetitive trajectories confined to a few localized hotspots. These hotspots gradually tile the retina with time, and disappear after eye opening. Our observations demonstrate that retinal waves undergo major spatiotemporal changes during ontogeny. Our results support the hypotheses that cholinergic waves guide the refinement of retinal targets and that glutamatergic waves may also support the wiring of retinal receptive fields.

Figure 1

A, the APS MEA chip. The red dotted line demarcates the electrode area. Top inset, scanning electron micrograph illustrating the topography of individual electrodes on the chip. Bottom inset, magnification of the active area of the chip with a retina positioned on the electrodes. B, spike raster plot of spontaneous episode of activity in a P11 retina. C, raw signals on four sampled channels from the same recording. D, two-dimensional time-lapse (every 1 s for 10 s) view of the activity. The s.d. of the voltage is estimated in 10 ms bins and plotted using an exponential colour coding scheme to emphasize large deviations and effectively threshold small deviations. Bottom row: same episode after downsampling the resolution to a simulated 8 × 8 array with an electrode pitch of ∼334 μm.

Figure 2

A, raster plot of all detected spikes during 10 min recording in a P4 retina. B, detected bursts and waves (individual waves illustrated in different colours) in the same time range. C, two-dimensional projection of two waves (waves A and B in B) over the surface of the chip. Dark colours show channels becoming active first; light colours show channels becoming active last. White and grey dots, respectively, represent all channels firing above and below 0.01 Hz during the recording session. D, spatial extent of wave A, estimated using alpha hull analysis. Green dots represent bursting channels participating in the wave; black dots represent channels that do not participate; grey dots outside the wave perimeter represent channels firing at ≥0.01 Hz; white dots represent channels firing at ≤0.01 Hz. E, centre of activity trajectories (CATs) of 11 waves belonging to the same cluster as wave A (thick line). The dark blue extremity of the CAT represents the wave initiation site; the warmest colour of the CAT represents wave termination.

Figure 3

A, the median interburst interval (IBI), burst duration (BD), burst size (BS) and firing frequency within bursts as a function of age, pooled for all datasets at each age. Shaded areas denote interquartile ranges. B, the correlation index as a function of distance on the retina for single recordings at three ages. Given the large number of cell pairs, we show the density of points by counting the number of points within non-overlapping hexagons of equal size. Darker colours indicate higher counts. The green line shows the fit of the decaying exponential. C, D, the two key parameters of the fit demonstrated in B (peak correlation; decay length) are shown for each recording as a function of postnatal age. The thin vertical lines are the 95% confidence intervals of the parameters. The thick lines are non-parametric locally weighted scatterplot smoothed versions of the data.

Figure 4

The plot illustrates the percentage of electrodes for which coincident spikes were detected on at least one of their four nearest neighbours for all retinas at all developmental stages during postnatal days (P) 2–12 grouped for P2–5, P6–8 and P9–12. Red, green and blue circles, respectively, indicate the percentage of electrodes with fewer than 5%, 10% and 25% of the total number of spikes occurring in coincidence with neighbours. Individual values represent the mean percentage of channels for a given retina; the horizontal black line represents the mean of all retinas in each age group. Error bars: s.e.m.

Figure 5

A, examples of waves at postnatal day (P) 5, P9 and P12. Each plot shows a raster of detected bursts, with waves colour-coded as in Fig. 1, and two-dimensional projections of selected waves, as indicated by arrows. B–E, median wave centre of activity trajectory (CAT) length (B), area (C), propagation speed (D) and activity density (E) for all waves at all developmental stages [postnatal day (P) 2: one retina, 106 waves; P3: two retinas, 191 waves; P4: four retinas, 204 waves; P5: four retinas, 188 waves; P6: five retinas, 99 waves; P7: three retinas, 72 waves; P8: four retinas, 232 waves; P9: eight retinas, 659 waves; P10: seven retinas, 631 waves; P11: six retinas, 945 waves; P12: four retinas, 718 waves]. Error bars indicate interquartile ranges. F, rasters of burst activity recorded at P13 and P15. No propagating activity can be seen at these ages.

Figure 6

A, examples of a wave at postnatal day (P) 9 under control conditions and in the presence of bicuculline (10 μm). Colour coding is as in Fig. 2D. B, summary of the effects of bicuculline on wave parameters (centre of activity trajectory length, area, speed and activity density) at P6, P9 and P12. Red symbols (mean ± s.e.m.) represent control values; green symbols represent values measured in the presence of bicuculline (P6: 42 and 102 waves for control and bicuculline, respectively; P9: 173 and 94; P12: 272 and 211). Blue numbers above each graph represent the average percentage change for each respective parameter for experiments pooled for P4–6 above P6 (five retinas), P9 (four retinas) above P9, and P10–12 (six retinas) above P12.

Figure 7

A, examples of centre of activity trajectory (CAT) cluster analysis for postnatal day (P) 4, P10 and P11 retinas. Each plot shows all CATs assigned to the same cluster, with colour coding as in Fig. 2E. B, C, the median number of clusters detected (B) and fraction of waves per cluster (C) as a function of developmental age. Error bars show interquartile ranges. The reduction in cluster number and increase in relative cluster occupancy show that waves become more repetitive and localized with development. D, data from B and C in the same plot, illustrating their dependency.

Figure 8

The plots illustrate mean burst durations (circles), burst sizes (triangles) and firing rates calculated across all channels (asterisks) over 2 or 3 days of continuous recording in postnatal day (P) 4 and P11 retinas. Numbers on the x-axis represent the day#_recording# (e.g. 1_1 corresponds to Day1_Recording1). Each recording session lasted 30 min and recordings were taken approximately every 3 h, except overnight. The red, green and blue colours, respectively, represent days 1, 2 and 3.

Figure 9

A, raster of detected waves from the initial recording trial in a postnatal day (P) 11 retina. B, three plots (in vertical order, except for the first dataset in which they are horizontal) show different aspects of the activity for each recording session. The top plot illustrates the raw spike count. The middle plot shows the probability of rejecting the null hypothesis that the activity measured in a channel and its proximal channels are unrelated, thus showing which channels participate in retinal waves and (indirectly via the significance level) how active they are. The darkest red colours represent the lowest probability that the activity is random. The bottom row shows the firing similarity index (red) and the correlation gradient (green) where correlations drop sharply and thus retinal waves tend to terminate.

Figure 10

A, raster of detected waves in a postnatal day (P) 4 retina. B, same analysis as in Fig. 8B. Firing activities in different regions do not change differentially over time, and there are no distinct wave boundaries.