Male and female mice C57BL/6 (The Jackson Laboratory) were reared in a 12 h light/12 h dark cycle. Young animals were dark exposed (DE) for 2 days (DE) between postnatal day 21 (P21) and 35 (P35). DE animals were cared for in a dark room with infrared vision goggles using dim infrared light. Some mice were re-exposed to normal light conditions for 2 h (LE) to study the effects of light exposure. All experiments were done in accordance with protocols approved by the Institutional Animal Care and Use Committees of Johns Hopkins University and University of Maryland.

Custom-made 16-channel laminar arrays were constructed and implanted as previously described (Murase et al., 2016). Briefly, a 1.2 mm 16-channel platinum-iridium electrode shank (15–20 kΩ) with a head post was implanted into the binocular region of primary visual cortex (3.00 mm lateral to the midline/0.01 mm rostral to lambda) to a depth of 1 mm, to center the electrode shank on the vertical center of cortex. For implantation, adult mice were anesthetized in 3% isoflurane in 100% O₂. The mice received post-surgical buprenorphine (0.1 mg/kg) after return of the righting reflex and were allowed 3–4 days to recover from surgery. One day prior to recording, subjects were habituated to the head restraint for 45 min. Single unit activity was recorded in awake head restrained animals in response to square wave gratings (100% contrast, 0.05 cycles per degree, 135 orientation) presented on a CRT monitor (Clinton, 28 cm × 36 cm, 60-Hz) placed 18 cm from the eyes, subtending 75 degrees of visual space vertically and 90 degrees horizontally (+Pattern group) or a gray screen at equal luminance (27 cd/m²) (+Gray group).

L4 was identified as the location of the short-latency (~150 ms) visually-evoked current source in the current source density (CSD) analysis of laminar local field potential (LFP) profiles. For isolated single-unit activity, the acquired signal was filtered from 300-Hz (high pass) to 5-kHz (low pass) and sampled at 25-kHz. Unit activity was acquired for 200 trials (1 s each) using an RZ5 Bioamp Processor (TDT). Spike sorting was based on waveform shape and a Bayesian fixed variable principle component analysis using Open Sorter (TDT). Neuron classification was based on three parameters: slope of the wave for 0.5 ms after the trough, time elapsed between the trough and peak, and the ratio between the height of the trough and peak (Niell and Stryker, 2008). Only regular spiking units, defined by a low peak to trough ratio, a long duration, and a small end slope, were included for analysis.

Single unit spike data were analyzed following processing with a non-biased method to delineate single spikes from bursts (Chen et al., 2009). Inter-spike intervals (ISIs) were used to generate a 1-dimensional data array from which we calculated the mean ISI value. ISIs which were greater or equal to the mean value were removed, and the remaining ISI values were used to calculate a second mean ISI for this population which was used as the threshold for detecting bursts. A burst was defined as a group of ISIs in which two or more consecutive ISIs had values that fell below the threshold. Using this method of burst detection, we wrote an algorithm in MATLAB that allowed automatic analysis of bursts and non-burst spikes from each in vivo recorded spike train data set. The specific algorithm used for this burst detection program and the MATLAB code is available online¹.

The following parameters were analyzed for each spike train: average firing rate, average number of intraburst ISIs, average duration of bursts, average number of interburst ISIs, average duration of interburst intervals, mean burst frequency (inverse of mean intraburst ISI), mean non-burst frequency (inverse of mean interburst ISI), and fraction of spikes in bursts [(Number of ISIs in burst)/(Total number of ISIs per spike train)].

A synthetic spike train was generated in MATLAB using a Poisson process with the mean ISI set equal to the average ISI of the +Gray group (see Tables s 1, 2).

The mice were anesthetized by isoflurane vapors and decapitated after verifying the absence of a toe-pinch response. Brain blocks containing visual cortex were coronally sliced into 300-μm sections using a vibratome (Leica VT1200S or Ted Pella Pelco easislicer^TM) in ice-cold dissection buffer containing 212.7 mM sucrose, 10 mM dextrose, 3 mM MgCl₂, 1 mM CaCl₂, 2.6 mM KCl, 1.23 mM NaH₂PO₄•H₂O, and 26 mM NaHCO₃, which was bubbled with a 95%-O₂/5%-CO₂ gas mixture. The slices were incubated at room temperature for 60 min in artificial cerebrospinal fluid (ACSF: solution containing 124 mM NaCl, 5 mM KCl, 1.25 mM NaH₂PO₄•H₂O, 26 mM NaHCO₃, 10 mM dextrose, 2.5 mM CaCl₂, and 1.5 mM MgCl₂, bubbled with 95% O₂/5% CO₂).

Visual cortical slices were transferred to a submersion-type recording chamber mounted on the fixed stage of an upright microscope with oblique illumination and were continually supplied with ACSF bubbled with 5% CO₂/95% O₂(30°C) with a flow rate of approximately 2 ml/min. AMPA receptor-mediated mEPSCs were isolated by adding 1 μM tetrodotoxin (TTX), 20 μM bicuculline (Bic), and 100 μM DL-2-amino-5-phosphonopentanoic acid (APV). Recording pipettes were filled with internal solution containing: 130 mM Cs-gluconate, 10 mM HEPES, 8 mM KCl, 1 mM EGTA, 4 mM disodium-ATP (Sigma–Aldrich, Cat #A6419), 10 mM disodium-phosphocreatine (Sigma–Aldrich, Cat #P7936), 0.5 mM Sodium-GTP (Sigma–Aldrich, Cat #G8877), 5 mM Lidocaine N-ethyl bromide (Sigma–Aldrich, Cat #L5783). Biocytin (1 mg/ml) was added to the internal solution to confirm morphology and location of the recorded cells post hoc. Pyramidal neurons in L2/3 of V1 were recorded in voltage-clamp at −80 mV and the recorded mEPSCs were digitized at 10-kHz by a National Instruments data acquisition board and acquired through an Igor based program (Wavemetrics). Two-hundred mEPSCs were analyzed from each cell with Mini Analysis (Synaptosoft).

Visual cortical slices were transferred to a submersion-type recording chamber mounted on the fixed stage of an upright microscope with oblique infra-red illumination. Slices were continually supplied with ACSF bubbled with 5% CO₂/95% O₂ (30°C) with a flow rate of about 2 ml/min. Pyramidal neurons in L2/3 of V1 were recorded in the whole-cell configuration. To generate activity patterns, a bipolar stimulation electrode was placed in V1 L4. Prior to stimulation a V1 L2/3 pyramidal cell was recorded under voltage clamp at −80 mV to determine the intensity of stimulation from L4. The stimulus intensity was adjusted using a stimulation isolated unit (SIU91A, Cygnus Instruments) to produce EPSCs of approximately 200 pA in the V1 L2/3 cell. The recording was then switched to the current clamp configuration, and we confirmed that the L4 stimulus results in subthreshold EPSPs. A small holding current was injected to keep the membrane potential at −65 mV in current-clamp, and Igor NIDAQ Tools MX was used to generate the desired stimulation pattern for 2 h.

All data is displayed as Mean ± S.E.M. Normality was confirmed with the D’Agostino and Pearson normality test. Unpaired t-tests and ANOVAs were performed to compare the averages of two or more normally distributed data sets, respectively. Data sets that were not normally distributed were compared using non-parametric statistical tests (e.g., Mann-Whitney or Kruskal–Wallis). Cumulative distribution of individual mEPSC amplitudes were compared using the Kolmogorov–Smirnov test. Statistical analyses were performed in Prism (Graphpad). P < 0.05 was taken as a statistically significant difference on all tests, except for Kolmogorov–Smirnov tests in which we used P < 0.001 as the cutoff.

In the canonical feedforward pathway, visually driven activity is conveyed to L2/3 neurons by excitation from thalamo-recipient L4 neurons. Hence, we aimed to test whether activating L4 with naturalistic neural activity associated with LE would drive metaplasticity in V1 L2/3 principal neurons. To identify activity patterns (neuronal spiking activity) in V1 L4 neurons with LE, we recorded single-unit activity in awake mice during two hours of re-exposure to light following two days of DE. Single unit recordings from V1 L4 regular spiking neurons were acquired in awake head-fixed mice via a chronically implanted multielectrode array. Recordings were acquired from DE mice during presentation of a blank screen (no visual stimulation) and during LE in response to a homogenous gray screen (+Gray) or a patterned visual stimulus (equiluminant high contrast grating; +Pattern). The in vivo spike trains recorded from the three groups (DE, +Gray, and +Pattern) were analyzed to extract various metrics of spike patterns. We used an ISI-based spike train analysis method (Chen et al., 2009) to distinguish spike bursts from non-burst spike activity (Table 1 and Figure 1). There was no statistically significant difference in the measured firing properties of V1 L4 neurons from when mice viewed the visual stimulus (+Pattern) or the gray screen (+Gray; Figure 1). The average firing rate (overall firing rate), mean burst length (mean number of spikes within bursts), and mean non-burst frequency (spike rate outside of bursts) were significantly higher in the +Gray and +Pattern LE groups (Figure 1 and Table 1). The spike rate within identified burst periods trended towards an increase in the LE groups, but did not reach a statistical significance likely due to the increase in variance (Figure 1D and Table 1). There was no statistically significant increase in the overall frequency of bursts across the three groups (Figure 1C and Table 1). Furthermore, there were no statistical differences in the metrics of spike patterns of V1 L4 neurons between +Gray and +Pattern groups (Figure 1 and Table 1). Our results demonstrate that the increase in visual evoked activity during LE is largely due to an increase in burst length and non-burst spike rates.

We examined if naturalistic spike patterns obtained from V1 L4 neurons during LE could drive homeostatic reduction in the strength of excitatory synapses onto L2/3 neurons. To do this, we stimulated L4 of V1 slices obtained from DE mice with the in vivo spike patterns recorded from L4 of LE mice (+Gray and +Pattern conditions; Figure 2). For both +Gray and +Pattern, the cell with the average firing rate closest to the group mean was chosen as the exemplar to provide the stimulation pattern for that group (Table 2). For each experiment, we first determined the stimulation intensity that was sufficient to evoke an approximately 200 pA EPSC in a L2/3 neuron (Figure 2B). We determined the stimulation intensity based on reports that visual stimulus presentation results in EPSCs of roughly 100–500 pA in V1 L2/3 pyramidal neurons recorded in vivo at similar holding potentials in the whole-cell voltage-clamp configuration (Sun and Dan, 2009; Haider et al., 2016). We found that the stimulation intensity, generated subthreshold EPSPs in current-clamp (Figure 2B). L4 was then stimulated at the determined intensity with the +Gray or +Pattern exemplar cell activity for 2 h to match the duration of LE shown to produce a reduction in mEPSC amplitudes in L2/3 neurons without changes in mEPSC frequency (Figure 3A). Following stimulation with either +Gray or +Pattern activity, mEPSCs were recorded from L2/3 neurons directly above the L4 stimulating electrode (Figure 2C).

Stimulation of DE slices with the spike pattern of the +Gray exemplar cell (see Table 2 for parameters) significantly decreased the average mEPSC amplitude in L2/3 neurons (Figure 3B). We saw no significant change in mEPSC frequency in the ex vivo stimulated slices. While the decrease in mEPSC amplitude on average looks similar to what we observed with LE (Figure 3A), the cumulative probability of mEPSC amplitudes from LE and +Gray conditions were statistically different from each other (Figure 3C). This suggests that plasticity induced by LE and +Gray is either not the same across synapses or that they are affecting a non-identical subset of synapses. Stimulating DE slices with the spike pattern of the +Pattern exemplar cell (see Table 2 for parameters) also produced similar results as the +Gray group, except this was accompanied by a significant increase in mEPSC frequency. To control for possible confounding effects of prolonged exposure to the perfusion bath prior to recording mEPSCs, a subset of DE slices was placed in the recording chamber for 2 h with a stimulating electrode, but received no stimulation (+No Stim) before recording. We observed no difference in average mEPSC amplitude or frequency between the +No Stim and DE control (amplitude: DE = 13.08 ± 0.35 pA, n = 41; DE+No stim = 13.08 ± 0.54 pA, n = 22; Unpaired t-test, t = 0.0634, p = 0.5485; frequency: DE median = 4.021 Hz; DE+No stim median = 4.368 Hz; Mann–Whitney test, U = 391, p = 0.3939). These data demonstrate that stimulating V1 L4 of DE mice with naturalistic spiking patterns obtained from L4 neurons during LE is sufficient to reduce the amplitude of mEPSCs in L2/3 neurons. Given that our stimulation is subthreshold for postsynaptic spike generation, these data suggest that the average frequency of input activity onto a given cell, not the postsynaptic firing rate, is the key factor driving homeostatic weakening of excitatory synapses in V1 L2/3 neurons.

The average firing rate in vivo of V1 L4 neurons was higher in DE +Gray and +Pattern groups compared to DE alone (Figure 1B and Table 1), thus we hypothesized that the average firing rate of presynaptic neurons was a key factor in engaging homeostatic downregulation of synaptic strength. To test this, we wanted to manipulate the average firing rate in the absence of other aspects of patterned neuronal activity, such as bursts. We therefore generated a Poisson distribution of ISI values with a mean equivalent to the average ISI of the +Gray group, and sampled a series of ISIs from this distribution to use as a stimulation pattern (Figure 3C and Tables s 1, 2). Because the average firing rate of +Gray and +Pattern group are similar (Figure 1B and Table 2), our Poisson spike train would reflect the average firing rate of both LE conditions. We confirmed that the distribution of ISI of the Poisson spike train is different from those of +Gray and +Pattern (Figure 4). Stimulating V1 L4 of DE slices with the Poisson spike train induced a significant reduction in mEPSC amplitude in L2/3 neurons similar to what we observed with naturalistic pattern stimulations (Figure 3B). However, the cumulative probability distribution of mEPSC amplitudes in the +Poisson group was significantly different from that seen in the naturalistic pattern stimulated groups (Figure 3C). This suggests that Poisson random activity is sufficient to produce homeostatic depression of mEPSC amplitudes, even though the changes across individual synapses are not identical to those produced by naturalistic patterned stimuli. There was also an increase in mEPSC frequency following Poisson stimulation, which mirrors what we observed with the +Pattern exemplar cell (Figure 3B).

We recently reported that the LE-induced reduction of mEPSC amplitudes is dependent on NMDAR and metabotropic glutamate receptor 5 (mGluR5) activity (Chokshi et al., 2019; Rodriguez et al., 2019). Therefore, we examined whether the naturalistic spike train-induced depression of mEPSCs presented here is also dependent on similar signaling. We stimulated slices from DE mice with the +Gray exemplar cell activity pattern (Table 2) in the presence of either ifenprodil (3 μM) or 2-methyl-6-(phenylethynyl)pyridine (MPEP, 50 μM) to block GluN2B-containing NMDARs or mGluR5s, respectively. Neither of the pharmacological antagonist prevented the reduction in average mEPSC amplitude driven by +Gray stimulation delivered to L4 (Figure 5). Our results suggest that the molecular signaling required for synaptic depression induced with naturalistic spike trains is different from those required for synaptic depression with LE. This finding together with our observation that +Gray results in a different shift in mEPSC amplitude distribution when compared to that of LE (Figure 3C) suggests that the plasticity induced by naturalistic spike pattern stimulation is not identical to that seen with LE.

Next, we examined whether the naturalistic spike pattern stimulation-induced reduction of mEPSC amplitudes in L2/3 neurons is dependent on the initial state of V1 circuitry, as determined by prior visual experience. To test this, we stimulated V1 L4 of slices obtained from normal reared (NR) control animals (NR slices) with spike pattern of the +Gray exemplar cell (Figure 6A). In contrast to the decrease in mEPSC amplitudes observed with DE slices, 2 h of +Gray stimulation in NR slices did not induce significant changes in the average amplitude or frequency of mEPSCs in L2/3 neurons (Figure 6B). However, the same spike pattern (+Gray) delivered to V1 L4 of slices from LE mice (2 h of light re-exposure after 2 days of DE; Figure 6A) induced a significant increase in both the average mEPSC amplitude and frequency (Figure 6C). Thus, the history of prior visual experience impacts the response of V1 circuit following stimulation of feedforward inputs from L4 with the same naturalistic spike pattern.