A conserved switch in sensory processing prepares developing neocortex for vision

Abstract

Developing cortex generates endogenous activity that modulates the formation of functional units, but how this activity is altered to support mature function is poorly understood. Using recordings from the visual cortex of preterm human infants and neonatal rats, we report a "bursting" period of visual responsiveness during which the weak retinal output is amplified by endogenous network oscillations, enabling a primitive form of vision. This period ends shortly before delivery in humans and eye opening in rodents with an abrupt switch to the mature visual response. The switch is causally linked to the emergence of an activated state of continuous cortical activity dependent on the ascending neuromodulatory systems involved in arousal. This switch is sensory system specific but experience independent and also involves maturation of retinal processing. Thus, the early development of visual processing is governed by a conserved, intrinsic program that switches thalamocortical response properties in anticipation of patterned vision.

Figure 1. Three periods in the early development of visual processing in the rat

A. Experimental setup. Primary visual cortex (V1) responses to 100 ms whole field light flashes were recorded in head-fixed, un-anesthetized rat pups between post-natal days (P)5 and P19 using extracellular direct current coupled recordings (bandpass 0–5000 Hz) from L4. B. “Light Blind” period (P7 and younger). Light flashes (yellow box) never evoked responses. A spontaneous delta-brush is marked by asterisk. C-D. “Light-Evoked Bursting” period (P8-11). Flashes of light evoked long-duration, high-amplitude slow waves that were typically composed of two distinct poles; an ‘initial response’ followed by an ‘evoked delta-brush’. Both waves contained rapid field potential oscillations. C. Frequency decomposition of a single flash evoked response from a P9 rat. Top trace is the raw depth EEG signal. In the middle the time series spectrogram, calculated for a 200 ms moving window from the high-pass filtered trace (white, HP > 2 Hz), reveals higher frequency components of the signal. The bottom trace shows the multi-unit activity (MUA) in a high-pass filtered (>200 Hz) trace. Note that action potentials are specifically associated with the troughs of high-frequency oscillations. D. Responses from the same animal in different awareness states. E-F. “Acuity” period (P12 and onwards). Light flashes evoked adult-like visual evoked potentials (VEP), without delta-brushes. E. Frequency decomposition of single flash evoked response from a P16 rat. Note that the elevated power at multiple frequencies is due to the 200 ms integration window and to multiple frequencies intrinsic to the singular VEP. F. Example traces as D. During sleep, evoked bursts resembling the k-complex were observed following light flashes (asterisk; see also Fig. 2). G. Population average time-course of flash evoked responses in two-day intervals. All time-courses are aligned to the peak MUA rate for each animal (t = 0 s); the time of the light flash is shown as the red circle (± SD) above. For each panel, top trace is an average depth EEG signal (dotted line = SD); Middle panel is an average time-series spectrogram (proportion of peak frequency power 5 - 50 Hz) and bottom graph is an average rate of MUA calculated with a 20 ms sliding window (dotted line = SD). Note that switch from bursting to acuity response occurs before eye opening. See Fig. S1.

Figure 2. State dependence of light-evoked responses develops after the switch

A. Example of evoked responses in different arousal states in V1. Simultaneous monitoring of EMG and movement was used to determine vigilance state. Visual responses were not modulated by vigilance state during the bursting period. During the acuity period light flashes evoked a singular sharp potential in all states, but during quiet sleep bursts of rebound MUA that resembled the adult k-complex were also evoked. B. K-complex could be evoked by cross-modal stimulations (here hand clap). C. Population averages for MUA and time-spectrogram. Averages are proportion of peak spike rate or power for each animal. Awake animals showed no significant MUA or spectral power elevation following the VEP. During deep sleep, a second peak in MUA caused by the k-complex was observed 500-1000ms after stimulation. Unlike younger animals, however, this evoked burst was not associated with a consistent L4 field potential oscillation.

Figure 3. Rapid switch in evoked and spontaneous activity before eye opening in the rat

A. Scatter plot of the duration of the evoked response vs. age for each pup. Duration was quantified as the amount of time that the spike-rate (50ms bins) was elevated above baseline. A sharp change in response is observed on P12. Wistar rats = thin circles, Long-Evans = think circles. At right is a histogram of response duration for all trials for all animals. The total distribution is shown by the grey bars and is bi-modal, suggesting two populations of response. Overlaid on the grey bars, the black sided boxes show only the events of young pups. Thus P8-11 animals account for all events longer than 1 s. This bi-modal distribution of duration further supports a switch, rather than a gradual reduction, in response duration. B. Median number of negative field potential deflections (‘troughs’, negative peaks greater than 50 μV measured on 5-30 Hz filtered trace) during the 2 s following the stimulation vs age. C. Delay to onset of the evoked potential vs age for each animal. Note that delay decreases monotonically with age. D. Repetitive recording from a single animal shows that the switch in visual response occurs in under 12 hours. Raster plots show spike-rate (left) and oscillation power (right, summed 5-20Hz) in 50 sequential trials for 5 sessions each spaced by 12 hours. An example trace from each recording (2 Hz highpass) is shown to the right. Between the night of P11 and the morning of P12 the response switches from a long-duration oscillatory event to a singular VEP. E. Duration of MUA response for 5 pups recorded sequentially as in D. Four out of five experienced the switch on the night of P11, while one did so during the day of P12. F-H. Switch in visual response is associated with an increase in the amount and continuity of spontaneous cortical activity. F. Spontaneous MUA activity in L4 of V1 recorded in the dark during different arousal states in the same P9 (left), P14 (middle) or P17 (right) animal. G. Scatter plots of pre-trial MUA vs. age (left) and continuity of MUA vs age. An increase in both rate and continuity was observed between P12 and P15, following a period of low activity 1^st-2^nd post-natal weeks. H. Scatter plots of spike rates during the pre-trial period (500 ms) vs. the duration of the evoked response for that trial. Response duration was negatively correlated with spontaneous activity rate at both ages. See Fig. S2.

Figure 4. Bursting and acuity periods in pre-term human visual development

A. Electrode placements for EEG recordings of neonates (modified 10-20 international placement). B. EEG traces from all electrodes (relative to ground, negative down) during quiet sleep in an infant during the 30th gestational week (GW). Whole field 100ms light flashes were applied at the red lines, and evoked high-voltage slow waves at occipital electrodes. More diffuse activity sometimes occurred spontaneously (or at a longer latency to stimulation) on other electrodes. C. Representative occipital light response at GW30 and GW40. The younger infant shows the high-amplitude slow waves and rapid oscillations (delta-brush) similar to rats; the older infant's response is smaller in amplitude and duration and has some of the grapho-elements typical of the flash-evoked VEP. For each infant the top trace shows all frequencies between 0.5 and 40Hz; the bottom trace has been high-pass filtered above 2 Hz to show the fast oscillations that occur within the larger, slower wave. The initial response is enlarged in the red circle. Identifiable elements of the visual evoked potential are marked for the term infant. D. Duration of the evoked fast oscillations vs. age for each infant (left). Duration was quantified as the amount of time the summed power between 5 and 20 Hz (100 ms bins) was statistically elevated over baseline. A loss of the long-duration fast-oscillations occurred between GW34 and GW36. A single infant recorded on multiple days is represented in red. A histogram of response duration for all trials in all infants as described for Fig 3A is presented at right. E. Peak amplitude of the evoked potential (left) and Delay (right) vs age for each infant. F. Population averages of the light-evoked potential (dotted line = SD) at the occipital electrode contralateral to the light flash. Below is the average time-series spectrogram of the same population to show the fast-oscillatory responses. Responses were aligned to the beginning of the negative deflection (0s) and the mean ± SD time of flash is shown in red. G. Population average response maps for each age group showing the localization of evoked potentials during the initial response (at 300 or 240 ms delay) and of the long delay response (800 ms). H. Baseline activity vs. age. Activity is measured as SD as proportion of peak amplitude for each infant. As in rats, the loss of light-evoked high-voltage bursts was accompanied by an increase in spontaneous cortical activity. Five additional infants not assayed for light responses were included in this analysis. See Fig S3 and Movie S1.

Figure 5. Developmental elimination of light evoked bursting is correlated to maturation of intracolumnar processing and graded visual responses in the rat

A. Multi-electrode linear-array recordings show changing intra-laminar circuitry. P10 (left) and P13 (right) example traces show field-potential (scale bar = 200 μV) and MUA activity (scale bar = 50 μV) at multiple cortical depths. At P10 the gamma-burst oscillations of the initial response were limited to L4. Engagement of other layers came as a burst that spread throughout the column. The second event (evoked delta-brush) involved more robust, rhythmic activation of all layers. At P13 only a single mature VEP with no delta-brush remained. The VEP started in input layers and spread to superficial ones. Unit firing occurred as discrete bursts closely associated with the negative potential. B. The average current source density (CSD; left) and spike-rate (right) for each period using L6 spike-rate triggering during the initial response. C. L2/3 spike cross-correlation analysis made during the initial burst showed that the superficial layers were broadly correlated with all other layers during the bursting period, but specifically followed L4 after. D. Graded visual responses emerge after switch. Post-stimulus time histograms of spike rates in L4 during the initial response (right) show non-graded all-or-none bursting to various stimulation intensities in a P11 pup, but graded responses in a P13 pup. Sigmoidal curves fit to the intensity responses for 4 rats in each age group (left) show that the dynamic range of young pups is much smaller than after the switch. See Fig. S4.

Figure 6. Thalamocortical networks determine early bursting

A. Flash-evoked response in V1 during bursting (left) and acuity (right) periods in the rat. Example traces (P10 and 13 respectively) are shown above the population average time-series spectrogram and the average MUA time course. B. While recording from the same cortical position, retinal circuitry was disrupted by intraocular injection of glutamate receptor antagonists. Blockade was verified by assaying absence of a light-response. C. Direct electrical stimulation of the optic nerve (1ms) while blocking retinal activity produced a single sharp potential with short delay similar to the acuity VEP, but still evoked columnar bursts (Fig. S5) and delta-brushes at longer latency. After the bursting to acuity switch, both light and optic nerve stimulation evoked a sharp VEP but not delta-brushes. See Fig. S5.

Figure 7. Role of ascending neuromodulators in mediating the switch

A. Isolation of forebrain via a mid-collicular cut (cerveau isolé or CI) during the acuity period was used to examine the role of the ascending activating systems in maintaining the continuity of spontaneous cortical activity and acuity response in the rat. Location of cuts determined post-mortem is shown on a schematic drawing (black lines), while two more posterior cuts that did not affect activity are shown in red. B. Flash-evoked response in V1. An example trace (P14 pup) is shown above the population average (P13-14) time-series spectrogram and the average MUA time-course for three conditions: (top) control littermates that received needle insertion without cut, (middle) CI and (bottom) CI with norepinephrine (NE) added to the cortical surface. CI did not affect the VEP but restored oscillatory after-discharges which were then eliminated by NE. C. Quantification of CI and urethane treatment. Oscillation power (increase in 20-30 Hz over baseline) and MUA 300 – 1500 ms after stimulus (time of evoked delta-brushes) was used to quantify treatment effects (top graphs). In addition to CI experiments, urethane (0.5 g/kg I.P.) was given to control pups (n = 6) followed by surface application of NE (n = 5). Ongoing activity (bottom graphs) was measured as in Fig. 3G. Urethane and CI both reduced continuous ongoing activity and reinstated oscillatory after-discharges, and NE suppressed these effects.

Figure 8. Inhibitory GABA in cortical circuits before the switch

A. An example trace is shown above the population average time-series spectrogram and MUA time-course for three conditions: control recordings, diazepam on cortical surface, and diazepam I.P. The example traces are from the same P10 pup. MUA and power are relative to peak control activity. B. Frequency distribution (5-70Hz) of evoked long-latency (300-1500 ms) activity. Cortical diazepam eliminated evoked rapid oscillations but allowed disorganized 8-12 Hz activity. Bumetanide, which shifts GABA_A synaptic currents from excitatory to inhibitory in immature neurons in vivo, had no effects on visual responses. C. Post-stimulus MUA (0-1500 ms). D. Mean evoked potentials (wide-band depth EEG, n = 3) show a differential effect of cortical and systemic diazepam. Cortical diazepam reduced the amplitude of the evoked potential but did not prevent the gamma oscillations (see also A – middle graphs), while I.P. injection eliminated these oscillations which were replaced by a single potential. B-C error bars are SEM. See Fig. S6.

Figure 9. Rat somatosensory responses switch from bursting to acuity earlier

A. Evoked somatosensory responses were assayed in barrel cortex via air puffs applied to the whiskers. Three representative whisker responses show evoked oscillations to whisker stimulation at P7 (left) and their loss by P9 (right). Air puffs generated a sharp evoked potential at both ages. B. Rapid loss of long-duration evoked activity by P9 was correlated with an increase in spontaneous activity. Scatter plots show the duration of evoked MUA (left) and spontaneous spike rate (right) vs. age for each hemisphere from 14 rats. Both parameters showed sharp changes similar to the visual system, but 3-4 days earlier. C. Population average post-stimulus MUA from before (black) and after (grey) the switch. Rate was calculated for a 20 ms sliding window (dotted line = SEM).

Figure 10. Early development of sensory processing in humans and rats

A. Developmental templates of sensory development. Timelines are aligned to independent time-points: the arrival time of LGN axons (P3.5 rat, 21GW human (Finlay and Darlington, 1995) and the onset of monocular deprivation plasticity (8-10 weeks human (Birch and Stager, 1996), P19 rats (Fagiolini et al., 1994)). The timing of the switch in human somatosensory cortex is not known, although delta-brushes were evoked in GW33 infants (Milh et al., 2007). In humans the switch in somatosensory and visual regions is more closely aligned with birth, while the switch timing is independently regulated in rat. B. Hypothesized relationship between retinal and cortical activity during early visual development. The integrated voltage deflection in V1 (solid line top trace; from Fig. 1G) and the integrated retinal MUA (middle trace; from Fig. S5) following light stimulation are shown relative to the response after eye opening. During the first post-natal week light stimulation does not cause a cortical response. However, electrical stimulation can trigger bursting (top trace dotted line; (Hanganu et al., 2006). Dividing the cortical response by retinal output shows clear thalamocortical amplification during the bursting period. C. Summary of identified factors contributing to the switch in sensory processing. Each of the three component parts of the immature response and their hypothesized adult homolog is color coded. The direct mechanism leading to their transformation is shown breaking the color coded line, while factors modulating these factors are shown as black arrows. For example, the gamma-burst transforms into the VEP as a result of short latency response development in the retina and increased inhibition (which may be brought about via the former).