Abstract
A classic example of experience-dependent plasticity is ocular dominance (OD) shift, in which the responsiveness of neurons in the visual cortex is profoundly altered following monocular deprivation (MD). It has been postulated that OD shifts also modify global neural networks, but such effects have never been demonstrated. Here, we use wide-field fluorescence optical imaging (WFOI) to characterize calcium-based resting-state functional connectivity during acute (3 d) MD in female and male mice with genetically encoded calcium indicators (Thy1-GCaMP6f). We first establish the fundamental performance of WFOI by computing signal to noise properties throughout our data processing pipeline. Following MD, we found that Δ band (0.4–4 Hz) GCaMP6 activity in the deprived visual cortex decreased, suggesting that excitatory activity in this region was reduced by MD. In addition, interhemispheric visual homotopic functional connectivity decreased following MD, which was accompanied by a reduction in parietal and motor homotopic connectivity. Finally, we observed enhanced internetwork connectivity between the visual and parietal cortex that peaked 2 d after MD. Together, these findings support the hypothesis that early MD induces dynamic reorganization of disparate functional networks including the association cortices.
- calcium neuroimaging
- cortical reorganization
- functional connectivity
- monocular deprivation
- plasticity
Significance Statement
Monocular deprivation during the visual critical period conducts several plasticity mechanisms that cooperate to shift the excitability of neurons in the visual cortex. However, little is known regarding the impacts of MD on cortex-wide functional networks. Here, we measured cortical resting-state functional connectivity during short-term critical period MD. We demonstrate that critical period MD has immediate effects on functional networks beyond the visual cortex, and identify regions of substantial functional connectivity reorganization in response to MD.
Introduction
Decades of research have revealed that several dramatic forms of plasticity occur in the visual cortex following monocular deprivation (MD) during early postnatal development. In rodents, MD-induced ocular dominance (OD) shifts follow a biphasic time course, where the early phase suppresses deprived eye responses, and the later phase strengthens both the deprived and open-eye responses (Frenkel and Bear, 2004; Mrsic-Flogel et al., 2007; Cooke and Bear, 2014). It is well-established that the initial suppression of deprived eye responses results from long-term depression (LTD)-like mechanisms (Heynen et al., 2003; Yoon et al., 2009), as well as plasticity mechanisms in inhibitory circuits (Maffei et al., 2006, 2010; Keck et al., 2011). Recently, it has also been shown that callosal inhibition of deprived eye responses contribute to OD shifts during MD (Restani et al., 2009; Pietrasanta et al., 2014). However, the possibility that this manipulation may impact disparate functional networks remains a critical and untested prediction of neural plasticity. Here, we used wide-field optical imaging (WFOI) of genetically encoded calcium indicators (GECIs) to test whether acute MD and its subsequent plasticity processes reorganize neuronal (i.e., calcium) resting-state functional connectivity (rsFC) in the neocortex.
rsFC – the temporal correlation of spontaneous activity between different brain regions – is a mathematical measure used to assess the organization of functional networks. In humans, resting-state functional magnetic resonance imaging (rs-fMRI) studies have shown that changes in rsFC patterns exhibit feed-forward mechanisms and thus reflect prior experience (Waites et al., 2004; Hasson et al., 2009; Grigg and Grady, 2010). For example, sensory deprivation due to upper limb casting (Newbold et al., 2020), congenital and early blindness (Wang et al., 2013; Burton et al., 2014), intensive training on relational reasoning (Mackey et al., 2013), and visual perceptual learning (Lewis et al., 2009) have all been shown to alter rsFC patterns. However, there has been limited characterization of rsFC changes during OD plasticity. In vivo anesthetized optical intrinsic signal imaging after 14 d MD revealed a decrease in interhemispheric visual homotopic connectivity but only small changes in distant networks (Kraft et al., 2017). However, measuring end-point rsFC using hemodynamic signals from anesthetized animals may not be the best probe for network-level consequences for a number of reasons. Most critically, because Hebbian and homeostatic mechanisms operate with distinct temporal profiles (Rittenhouse et al., 1999; Hengen et al., 2013, 2016), sampling only once after 14 d MD may miss most of the time course of network reprogramming. In addition, blood-based imaging (e.g., fMRI) relies on a physiological process referred to as neurovascular coupling (NVC), which indirectly indexes neuronal activity through modulations of oxygen metabolism and blood flow. Importantly, NVC has been shown to maturate across postnatal development, which could alter hemodynamic reporting of neuronal signals (Kozberg et al., 2016). Likewise, manipulations that change neural activity long term initiate transcriptional changes in endothelial cells (Hrvatin et al., 2018). These molecular changes at the cell level may evolve into vascular remodeling, which may confound the use of hemodynamic measures for sensory deprivation experiments. Therefore, we set out to perform calcium-based WFOI to map the differential responses in rsFC during the early phase of MD. We hypothesized that MD-induced OD shifts would have circuit-level impacts on distant networks, contrasting this with the alternative hypothesis that the loss of visual drive would only modify the immediate circuit.
To test these hypotheses, we took advantage of a classic MD paradigm using lid suturing to perturb visual experience in juvenile mice during visual critical period [postnatal day 24–27 (P24–P27)], a developmental window where MD is known to induce different forms of plasticity (Gordon and Stryker, 1996; Mrsic-Flogel et al., 2007; Smith et al., 2009; Yoon et al., 2009). To avoid the confounds of anesthesia, we performed awake resting-state calcium imaging using transgenic Thy1-GCaMP6f mice. We deeply characterized the performance and signal-to-noise characteristics of the system and computational pipeline to ensure robust sensitivity to changes in Δ calcium power and resting-state functional connectivity. We then applied the system to assess the response to MD: we performed imaging before lid suturing and then repeated imaging for 3 d thereafter, and replicated our findings using both within-subjects (MD vs baseline) and between-subjects (sham vs suture) analyses. Overall, we found changes in both visual intra-network relationships and distinct internetwork relationships. These results support the hypothesis that neuronal changes during OD shifts have large-scale network-level impacts that alter rs-FC and provide a description of the network-wide consequences of attenuated visual input during a visual critical period.
Materials and Methods
Animals
All procedures described below were approved by the Washington University Institutional Animal Care and Use Committee and conducted in accordance with the approved Animal Studies Protocol. Breeding animals were kept in a facility with a 12:12 light/dark cycle, in standard cages with nestlets and a cardboard Bio-Tunnel (Bio-Serv) and ad libitum access to standard lab diet and water.
Adult mice expressing GCaMP6f under the Thy1 promoter (Thy1-GCaMP6f, JAX:024276) were used for experimentation. Homozygous Thy1-GCaMP6f males and females (produced in-house from hemizygous JAX mice) were bred to heterozygous Myt1l knockout females and males without the GCaMP6f allele to produce animals that were hemizygous for the Thy1-GCaMP6f allele to enable calcium imaging. Genotyping for Myt1l was used to exclude mutant animals from analysis such that only wildtype pups were used for analyses here.
Experimental animals were weaned at postnatal day 21(P21) and housed by sex. All experimental animals were kept in standard cages with ad libitum access to food and water and enrichment, including a mouse tube (Bio-Serv), mouse hut (Braintree Scientific), nestlets, and hydrogel (Bio-Serv).
Cranial windowing
A transparent chronic optical window made of plexiglass was fitted to the dorsal cranium of the mouse at P21 as previously described (Rahn et al., 2021). The mouse was anesthetized via isoflurane and the head was shaved. An incision was then made along the midline of the scalp to retract the skin and expose an approximately 1.1 cm2 dorsal cortical field of view. The plexiglass window was adhered to the dorsal cranium using Metabond clear dental cement (C&B-Metabond, Parkell Inc.). Mice were returned to their cages to recover for at least 36 h before they were habituated to head-fixation in the imaging apparatus at P23.
Eyelid suturing
Lid suturing was performed between zeitgeber time (ZT) 10 and 15, after baseline imaging on P24. Prior to lid suturing, mice were anesthetized using isoflurane and received 5 mg/kg of carprofen (RIMADYL) for analgesia via subcutaneous injection. Eyelashes and lid margins were first trimmed using Vannas scissors. Three to four mattress sutures were then applied using sterile 6-0 Prolene sutures (Ethicon). Sham MD mice were subjected to the same length of anesthesia and received carprofen injection. After suturing, mice were returned to clean cages with cagemates and enrichment, and 0.5 mg carprofen tablet (Bio-Serv) was placed on the cage floor for each mouse. All mice continued to receive 0.5 mg carprofen tablet for 3 d after lid suturing surgery. Suture integrity was checked prior to each imaging session. Mice who developed eye infections or whose sutures opened prematurely were dropped from the experiment.
Wild-field fluorescence optical imaging
Mice were imaged in a WFOI system with four sequential illumination via LEDs (470, 530, 590, and 625 nm) which allowed for measuring calcium fluorescence as well as hemodynamic fluctuations (Ma et al., 2016; Wright et al., 2017). To enable head-fixed imaging, a 5 1/2″ × 2″ × 1/16″ metal plate is screwed directly into the mouse's cranial window through an auxiliary bracket. An sCMOS camera (Zyla 5.5, Andor Technologies) captured frames at a rate of 16.8 Hz per LED channel, which facilitates analysis in the Δ frequency band (0.4–4 Hz). The first two cohorts of mice underwent five 5 min imaging runs of resting-state awake data collection, while all other cohorts had four 5 min imaging runs collected to better manage imaging workload. All data collection was done between ZT2 and ZT10. To reduce stress and movement artifacts, all mice were acclimated to head-fixed imaging at P23. All scanning was conducted in a dark room where light levels were below 50 nW/cm2, or ∼1/1,000 of ambient lab light levels (∼50 microW/cm2). These data acquisition metrics match a previous visual deprivation experiment (Kraft et al., 2017).
Signal to noise levels: Our system (with Zyla 5.5 USB 3.0 sCMOS) has a readout noise of 2.6 e-RMS, and a full well capacity of 30,000 electrons. The fluorescence signals are typically 2,000 electrons in each of the 1,248 pixels by 1,248 pixels over the field of view. We then binned 16x16, down to 78x78 pixels, so that each binned pixel (256 raw pixels) has ∼5 × 105 electrons with a shot noise of 707 e-RMS and a collective readout noise of 41.6 e-RMS. Thus, for our measurements, the readout noise is negligible, and the fluorescence measurement noise is dominated by shot noise at ∼0.14%. Further, we temporally smooth the data down to 8.4 Hz from 16.8 Hz, which reduces the variance by sqrt(2) down to 0.1%.
Imaging data preprocessing
Data were preprocessed using the previously published optical imaging toolbox in MATLAB (Brier and Culver, 2023). A flowchart illustrating the general data preprocessing steps is depicted in Figure 1. Briefly, we used a 1,248 by 1,248 pixel region from a sCMOS camera (Andor Zyla) to collect light from the mouse dorsal cortex. On camera 2 by 2 binning was performed while acquiring data and before transferring data to a solid state drive. After the data collection, we performed a second spatial 8 by 8 bin resulting in a 78 by 78 pixel image, where each of these pixels is the net sum of 16 × 16 = 256 raw sCMOS sensor pixels. This data were loaded into MATLAB, and a binary mask of the mouse brain was created by manually tracing the mouse dorsal cortex within the field of view. Next, we used a 2× temporal decimation to bring the frame rate down to 8.4 Hz. A representative background frame (i.e., the average of 5 s dark frame) was subtracted from the data. Individual pixels' time traces were temporally detrended and then reshaped into two-dimensional structure for spatial detrending. Using the changes in the reflectance in the 530, 590 and 625 nm LED channels, a modified Beer–Lambert Law was solved to obtain fluctuations in oxy- and deoxy-hemoglobin concentration, as described previously (White et al., 2011). Hemodynamic absorption of GCaMP6 emission was corrected using a previously established method (Ma et al., 2016; Brier and Culver, 2021). A final spatial smoothing was performed with a Gaussian filter (5.5 pixel box with a 1.2 pixel standard deviation), and a global signal regression was performed. Exemplar outcomes of the above preprocessing steps are illustrated in Figure 2. Images were then affine transformed to a common space using the user-defined landmarks: the intersection of the coronal and sagittal suture and lambda. All analyses were performed on data bandpass filtered between 0.4 and 4 Hz Δ band using a fifth-order Butterworth filter. Exemplar results following coregistration and bandpass filtering are depicted in Figure 3.
Mouse calcium-based wide-field optical imaging (WFOI) data analysis workflow. Raw GCaMP6f fluorescence signals first go through a series of preprocessing steps (see details in Fig. 2). Processed data are coregistered to a common atlas, and then high-pass filtered at 0.4 Hz and low-pass filtered at 4 Hz (see details in Fig. 3 and Embedded Movies 1 and 2). For functional connectivity (FC) analysis, Pearson correlation coefficients are calculated between regions of interests from the 0.4–4 Hz bandpass filtered data.
Preprocessing of raw GCaMP6f signals. A, Flowchart illustrates the order and steps involved in data preprocessing. B, Exemplar FOV in grayscale. C, Preprocessed dF/F time course from the three ROIs illustrated in B. The average time course is shown in black and individual pixel's time courses are depicted in color. D, Exemplar outputs from each data preprocessing step shown in A. E, Distribution of dF/F measurements following log mean transformation, image smoothing, and global signal regression.
Preprocessed data are coregistered to a common atlas and bandpass filtered at 0.4–4 Hz. A, Exemplar images of the mouse dorsal cortex before coregistration and B, after coregistration using two anatomical landmarks: anterior midline suture and lambda. C, GCaMP6f power spectra of a 5 min run before (orange) and after (purple) 0.4–4 Hz bandpass filtering. D, Exemplar dF/F time course of pixels within ROI1 (shown in B) before (top) and after (bottom) 0.4–4 Hz bandpass filtering. The average time course is shown in black while each individual pixel's time course is shown in color. E, Example functional connectivity analysis between the selected ROIs shown in B. While ROI1 (red) and ROI2 (green) – both within the visual cortex – are moderately correlated, ROI1 and ROI3 (blue) – which are in different anatomical and functional regions – are moderately anticorrelated.
Movie 1
Exemplar cortical broadband calcium activity in monocular deprived (MD) and sham mice. Download Movie 1, MP4 file.
Movie 2
Exemplar cortical delta-band (0.4-4 Hz) calcium activity in monocular deprived (MD) and sham mice. Download Movie 2, MP4 file.
To minimize motion artifacts, individual 5 min runs were excluded from analysis if raw LED light levels exhibited greater than 1.5% variance across the run, which may be the result of large body movements. Following this motion censoring, 5–25 min of data from each mouse were used for analysis at each timepoint. In addition to motion censoring, a global measure of the temporal derivative was used to quantify variance that may be attributable to motion in the remaining 5 min runs. The global variance in the temporal derivative (GVTD) was calculated by taking the root mean square (RMS) of the temporal derivatives across all pixels within the brain mask (Sherafati et al., 2020).
Functional connectivity
Three measures of functional connectivity were computed based on bivariate Pearson correlation analysis between region pairs. Contralateral homotopic (anatomical homolog) interhemispheric rsFC was calculated to examine the correlation between each pixel's individual time trace and its homotopic partner's time trace (i.e., homotopic connectivity). Seed-based rsFC maps were calculated using a set of 26 canonical cortical seeds as described previously (White et al., 2011; Rahn et al., 2021, 2023; Brier et al., 2023), in which Pearson correlation coefficients were calculated between the averaged time trace of a seeded region (3 pixel radius around each seed) and time traces of all other pixels within the brain mask. Lastly, a matrix approach was taken to examine changes in all 325 pairings of the canonical seeds. Exemplar functional connectivity analyses are depicted in Figures 8 (MD) and 9 (sham), which illustrate time traces of seeded regions and their Pearson correlation coefficient.
Power spectral analysis
To determine the amount of spontaneous GCaMP6 activity in the Δ frequency band (0.4–4 Hz), power spectral analysis was performed by splitting Δ GCaMP6 time-series at each pixel into 10 s segments. A Hanning window was applied to those 10 s segments and fast Fourier transform (FFT) calculation was performed. The final power spectra were obtained by squaring the FFT product.
Gross movement detection during optical imaging in an additional cohort
To determine if and to what degree MD affected mouse movement, a separate cohort of hemizygous Thy1-GCaMP6f mice underwent lid suturing or sham operation and head-fixed imaging in an identical manner as the primary imaging cohort. However, instead of capturing neural activity, potential movement during neuroimaging sessions was captured using a digital video camera (sCMOS, Thorlabs, DCC3260C). This camera was placed in front of the mouse (FOV, ∼7 cm-by−7 cm) and illuminated by a near infrared (NIR) 780 nm LED. A 700 nm short pass filter (Thorlabs, FESH0700) placed behind the neuroimaging camera lens blocked NIR illumination used for movement monitoring. To avoid specular reflection, crossed near-infrared (NIR) polarizers (Thorlabs, LPVISC100) were placed in front of the illumination LED and movement monitoring camera lens.
During the experimental procedure, the same WFOI paradigm was initiated, and spontaneous behavior during WFOI was recorded for 20–25 min to mimic neuroimaging experiments.
To flatten the field of view and make it easier to detect and visualize differential changes, each pixel was normalized to its temporal mean. To expedite processing, the behavioral videos were resampled by a 4× spatial decimation to a resolution 256 by 256 pixels.
Optical flow (OF) estimates were generated using the Lucas–Kanade method (Barron et al., 1994), in which we chose a five-frame (250 ms) window for analysis. This optical flow algorithm yields a spatiotemporal time series of vectors, where each frame's pixels are associated with a vector indicating the direction and magnitude of spatiotemporal change. As such, OF is a powerful method for movement detection and tracking (Shafie et al., 2009). At each timepoint, the magnitude of each pixel's vector, as calculated by its Euclidean norm, resulted in a time-series of vector magnitudes for each pixel. The RMS across the field of view was calculated for each run. A two-way analysis of variance (ANOVA) was employed to assess the influence of time, surgical condition, and their interaction on the average of the movement distribution for each mouse.
Statistical analysis
Before averaging and statistical testing, all correlations were Fisher z-transformed. To obtain two-dimensional calcium power spectra and rsFC map for each animal, results across individual 5 min runs of each mouse were averaged. To calculate the change in rsFC, each mouse's baseline results were subtracted from its subsequent MD1, MD2, and MD3 results. To calculate the change in calcium activity, each mouse's MD1 to MD3 Δ band (0.4-Hz) GCaMP6 power was divided by its baseline GCaMP6 power. Next, two-sample t tests were used to compare MD mice to sham mice at each MD timepoint. Finally, a cluster size-based thresholding method adapted from (Brier and Culver, 2023) was used to correct for multiple comparisons and determine statistical significance. This cluster size-based approach credits larger clusters with greater statistical significance than smaller clusters that have the same peak t-value (Figs. 6E, 7E, 10E, 11E).
To determine statistical significance of the change in spectral power and paired seeded correlation relative to baseline for MD and sham mice, repeated measures ANOVA (rmANOVA) and post hoc paired t tests were performed (Figs. 6F, 7F,G, 10F,G). To assess the effect of time, condition (MD vs sham), and their interaction on the change in correlation, a two-way ANOVA and post hoc two-sample t tests were performed (Figs. 6F, 7F,G, 10F,G).
Resource availability
All of the imaging data, following preprocessing, is available at https://wustl.box.com/s/aw8h6156cjv7qizeewrume86yglvpvo8. Preprocessing and analysis are performed using the toolbox available on GitHub: https://github.com/brierl/Mouse_WOI/tree/main/ (Brier and Culver, 2023). The excel sheet with the specific settings which link the toolbox to the data is also provided at https://wustl.box.com/s/aw8h6156cjv7qizeewrume86yglvpvo8.
Results
Mapping rsFC with calcium-based WFOI
We characterized resting-state functional networks of pyramidal neurons during OD shifts using wide-field fluorescence optical imaging of Thy1-GCaMP6f mice. We first performed baseline imaging on postnatal day 24 (P24) to establish rsFC in the unperturbed brain (Fig. 5A). We then manipulated visual experience via right lid suturing on late P24, and sutures were maintained for 3 d (through P27) as OD shifts are reliably produced during this developmental window (Smith et al., 2009). We repeated optical imaging at the following timepoints: 18–22 h monocular deprivation day 1 [MD1]), 42–46 h (MD2), and 66–70 h (MD3) post lid suturing (n = 20). Imaging was conducted in a dark room and light levels due to fluorescence imaging were kept below 50 nW/cm2 or ∼1/1,000 of ambient lab light levels. A schematic of the experimental design is depicted in Figure 5A. As an additional control, another cohort of mice (n = 15) underwent sham suturing after baseline imaging. Throughout this study, our analyses thus implemented a within-subject design by comparing each mouse to its own baseline. For completeness, we also compared the changes from baseline in MD and sham-sutured mice to enable us to better interpret results and distinguish MD-mediated effects from any natural developmental changes in this time window. Overall, we collected 140 sessions of imaging data from 35 mice in total, representing 40 total hours of data for subsequent analysis.
Data were preprocessed using the previously published optical imaging toolbox in MATLAB (Brier and Culver, 2023) for which we have now deeply characterized the signal to noise properties at every step of the process (Figs. 1–4). Then, to understand how rsFC reorganizes following sensory deprivation, we investigated changes to three complementary measures of rsFC: pixel-based interhemispheric homotopic functional connectivity maps (Fig. 5C), seed-based rsFC maps (Fig. 5D), and seed-based correlation matrices (Fig. 5E). All three measures use the same raw bivariate correlation analysis as a starting point. Interhemispheric contralateral homotopic rsFC selects and analyzes only homotopic contralateral pixel pairs and could illustrate changes in the relationship between ipsilateral and contralateral visual cortex. Seed-based rsFC refers to the temporal correlation between the seeded region of interest (ROI) and all other pixels in the field of view (FOV), revealing a focused functional topography of the deprived visual cortex's connections to the rest of the brain. Finally, the correlation matrix examines the pairwise correlation between all 26 canonical seed pairs thereby reflecting global changes in rsFC after MD, even in connections not involving visual seeds. If the primary hypothesis is true, then we predict that MD would result in global rearrangements that manifest as changes in all three measures of rsFC. If the alternative hypothesis is true, then the consequence of the loss of input would be restricted to changes within the visual cortices.
Assessment of signal to noise ratio of wide-field optical imaging system (WFOI) shows extremely low technical noise by comparing live to dead mouse. A, Distribution of pixels from a 5 min run on MD3 following image smoothing and global signal regression (GSR). The number on the top right of each panel indicates the histogram's full width at half maximum. B, Distribution of pixels from a 5 min run after euthanasia. The number on the top right of each panel indicates the histogram's full width at half maximum. C, Three ROIs are selected to illustrate the time course of spontaneous neural activity before (D) and after (E) 0.4–4 Hz bandpass filtering.
Monocular deprivation (MD) and calcium-based WFOI paradigm. A, Experimental timeline for optical imaging and lid suturing. For MD mice (n = 20;14 females and 6 males), the right eyelid was sutured after baseline imaging at P24. Sham-sutured mice (n = 15; 5 females and 10 males) were subjected to the same length of isoflurane anesthesia. Mice were subsequently imaged at MD1 (P25), MD2 (P26), and MD3 (P27). Mice were subsequently imaged at MD1 (P25), MD2 (P26), and MD3 (P27). Detailed data preprocessing steps are illustrated in Figures 1–4. B, Visualization of sampled seed centers overlaid with Paxinos atlas brain regions. C, Representative baseline interhemispheric homotopic connectivity map derived from sample 5 min run data, displaying Fisher z-transformed Pearson's correlation coefficients between each pixel in the left hemisphere and its corresponding pixel in the right hemisphere. Only the left hemisphere is displayed, as homotopic connectivity is symmetrical across the midline. D, Representative baseline left visual seeded rsFC map derived from 5 min run data. Black dot represents the Visual-a seed region, and colormap represents correlation to that seed of each pixel's time traces. E, rsFC matrix from the same 5 min run data displaying correlation values for each seed pair.
MD reduces Δ calcium power in the deprived visual cortex
To confirm that MD reduces spontaneous neuronal activity as expected, we examined changes in δ band (0.4–4 Hz) calcium power, as it was previously reported that short-term critical period MD decreases Δ activity in the visual cortex when measured by local field potentials (Malik et al., 2022). We therefore investigated the changes in cortex-wide Δ calcium power in both MD and sham mice and compared the effects between the two groups. To determine the effect of MD versus natural development on Δ calcium power, we first plotted power as a percent of baseline at each MD timepoint for both sham and MD mice (Figs. 6B,C). In MD mice, the reduction of Δ calcium power was restricted to the left (deprived) visual cortex (Fig. 6B). In contrast, sham mice displayed an increase of Δ calcium power in both visual hemicortices (Fig. 6C), suggesting this is an aspect of development. Thus, to distinguish the effect of MD from the effect of development, we compared the changes in Δ calcium power between MD and sham mice (Fig. 6D). We found a statistically significant reduction of Δ calcium power in the deprived visual cortex at MD3 (Fig. 6E), confirming an expected reduction in spontaneous neural activity as a result of MD. In addition, when we averaged all pixels within the left visual cortex to a single value, we found a statistically significant reduction of Δ power between all MD timepoints relative to baseline for MD mice but not for sham mice, and a statistically significant interaction of condition and time on visual cortex Δ power (F(3,99) = 3.77, p = 0.017) (Fig. 6F). To investigate the changes in connectivity of the deprived visual cortex where activity was reduced, we placed a seed in the cluster of pixels with t < −2 for subsequent visual-seeded rsFC analysis (Fig. 6G).
A 3 d critical period monocular deprivation (MD) decreases Δ frequency Ca2+ power in the visual cortex. A, Average MD calcium Δ power maps at each timepoint (n = 20). B, MD Δ calcium power plotted as the averaged percentage of each mouse's baseline values. C, Sham calcium Δ power plotted as the averaged percentage of each mouse's baseline values (n = 15). D, Mean difference maps comparing the changes in Δ calcium power from baseline between MD and sham mice that were shown in (B) and (C). E, Two sample t tests between each MD timepoint and baseline. Maps display pixels that had significant change in Δ calcium power. F, Δ calcium power in the visual cortex of MD and sham mice across time. # denotes significance on rmANOVA (MD mice, F(3,57) = 9.50, p = 0.00035). G, Overlay of the significant cluster in the visual cortex with Paxinos atlas-based region outlines. Black dot indicates the visual seed for subsequent analysis. Error bars represent SEM here and in subsequent figures.
MD suppresses visual homotopic functional connectivity
The early phase of MD is known to produce a strong and rapid shift in responsiveness of visual cortex neurons in favor of the open eye (Mioche and Singer, 1989). This shift toward the open eye is attributable to the induction of LTD in the visual cortex (Heynen et al., 2003; Liu et al., 2008; Yoon et al., 2009) as well as callosal suppression of deprived eye responses (Restani et al., 2009). We therefore hypothesized that MD would cause a rapid decrease in visual homotopic connectivity, as asymmetric visual experience would cause spontaneous neuronal activities from the two hemispheres to become desynchronized. However, over time if responsiveness of the deprived visual cortex neurons returns to the contralateral eye, homotopic connectivity should also return.
We were interested specifically in the temporal response of homotopic connectivity to the manipulation of visual experience, and so we investigated the changes in homotopic contralateral rsFC relative to each animal's baseline at each timepoint (Fig. 7A–C; representative examples in Figs. 8, 9). Because prior work has shown rsFC naturally changes across this time period in rodents (Rahn et al., 2021), we then compared the differences in those changes between MD and sham groups using a two-sample t test (Fig. 7D,E). We found that homotopic connectivity between the deprived and nondeprived visual cortices displayed a significant decrease in MD mice compared to sham mice at MD2 and MD3 (Fig. 7E). In addition, parietal and sensory regions, as well as those in frontal and motor regions, showed a decrease in homotopic connectivity during MD compared to sham mice (Fig. 7E). We examined the correlations between homotopic contralateral visual seed pairs using a two-way ANOVA (Fig. 7F,G) and found significant time by condition interaction for both seed pairs in the visual cortices (Visual-a and Visual-b seeds, Figure 5B for key); homotopic connectivity between contralateral visual seed pairs significantly increased over time in sham-sutured mice as expected during visual critical period but significantly decreased following 1 d of MD (Table 2).
Critical period monocular deprivation (MD) alters homotopic functional connectivity patterns. A, Group averages of MD homotopic FC maps across time (n = 20). Exemplar traces of visual cortex activities in MD and sham mice are illustrated in Figures 8 and 9, respectively. B, MD homotopic FC plotted as the averaged change from each mouse's baseline values. C, Sham homotopic FC plotted as the averaged change from each mouse's baseline values (n = 15). D, Difference maps displaying the mean difference in the changes from baseline between MD and sham mice at each MD timepoint. E, Two-sample t tests compared the changes from baseline between MD and sham mice at each MD timepoint. F, Pearson correlations between homotopic contralateral Visual-a seeds (two-way ANOVA F(3,99) = 3.73, p = 0.021 HF-corrected). † denotes significance on post hoc two-sample t tests comparing MD and sham mice at each timepoint (baseline, t(33) = 0.12, p = 0.90.; MD1, t(33) = −2.79, p = 0.0088; MD2, t(33) = −3.03, p = 0.0048; MD3, t(33) = −3.39, p = 0.0018). # denotes significance on rmANOVA (MD mice, F(3,57) = 6.32, p = 0.0009; sham mice, F(3,42) = 3.37, p = 0.027). * denotes significance on post hoc paired t tests comparing each MD timepoint to baseline values for MD mice (MD1 vs baseline, t(19) = 3.42, p = 0.0029; MD2 vs baseline, t(19) = 2.72, p = 0.014; MD3 vs baseline, t(19) = 2.51, p = 0.021). ¤ denotes significance on post hoc paired t tests comparing each MD timepoint to baseline values for sham mice (MD3 vs baseline, t(14) = 3.50, p = 0.0035). G, Pearson correlations between homotopic contralateral visual-b seeds (two-way ANOVA F(3,99) = 3.88, p = 0.019 HF-corrected). † denotes significance on post hoc two-sample t tests comparing MD and sham mice at each timepoint (baseline, t(33) = −0.11, p = 0.92; MD1, t(33) = −2.71, p = 0.010; MD2, t(33) = −2.70, p = 0.011; MD3, t(33) = −3.37, p = 0.0020). # denotes significance on rmANOVA (MD mice, F(3,57) = 4.64, p = 0.0057; sham mice, F(3,42) = 4.66, p = 0.0067). *denotes significance on post hoc paired t tests comparing each MD timepoint to baseline of MD mice (MD1 vs baseline, t(19) = −2.75, p = 0.013; MD2 vs baseline, t(19) = −1.35, p = 0.19; MD3 vs baseline, t(19) = −2.24, p = 0.037). ¤ denotes significance on post hoc paired t tests comparing each MD timepoint to baseline values for sham mice (MD2 vs baseline, t(14) = 2.91, p = 0.011; MD3 vs baseline, (t14) = 4.99, p = 0.0002). *, ¤, † p < 0.05; **, ¤¤, †† p < 0.01; ¤¤¤, ††† p < 0.001 here and in subsequent figures.
Exemplar neural activity from the left and right visual cortex of an animal before and after monocular deprivation (MD) at baseline (A), MD1 (B), MD2 (C), and MD3 (D). ROIs are shown in the grayscale image on the right. The red trace depicts mean calcium activity of the left visual cortex, and the blue trace depicts mean calcium activity of the right visual cortex. Shading reflects the standard deviation.
Exemplar neural activity from the left and right visual cortex of a sham animal across the experiment timeline at baseline (A), day 1 (B), day 2 (C) and day 3 (D). ROIs are shown in the grayscale image on the right. The red trace depicts mean calcium activity of the left visual cortex, and the blue trace depicts mean calcium activity of the right visual cortex. Shading reflects the standard deviation.
Short-term monocular deprivation alters left visual-seeded rsFC patterns. A, MD average left (deprived) visual seed (represented by black dot)-based rsFC maps across time (n = 20). B, Pearson correlations of MD mice at each timepoint normalized to baseline values. C, Pearson correlations of sham mice at each MD timepoint normalized to baseline values. D, Mean difference maps illustrating the differences between normalized MD and sham mice shown in (B) and (C). (E) T-statistic maps showing significant clusters of pixels after two-sample t tests comparing normalized MD and sham mice at each timepoint, thresholded using a cluster-based technique (see Materials and Methods). F, Pearson correlations between left visual and parietal seeds (two-way ANOVA F(3,99) = 3.73, p = 0.021 HF-corrected). † denotes significance on post hoc two-sample t tests comparing MD and sham mice at each timepoint (baseline, t(33) = 0.74, p = 0.47; MD1, t(33) = 2.25, p = 0.031; MD2, t(33) = 2.58, p = 0.015; MD3, t(33)= 3.38, p = 0.0019). # denotes significance on rmANOVA (MD mice, F(3,57) = 16.354, p < 0.0001). *denotes significance on post hoc paired t tests comparing each MD timepoint to baseline of MD mice (MD1 vs baseline, t(19) = 3.96, p = 8.4 × 10−4; MD2 vs baseline, t(19) = 4.9005, p = 9.9 × 10−5; MD3 vs baseline, t(19) = 5.86, p = 1.22 × 10−5). G, Pearson correlations between left visual and somatosensory-d seeds (two-way ANOVA, F(3,99) = 3.92, p = 0.014 HF-corrected). † denotes significance on post hoc two-sample t tests comparing MD and sham mice at each timepoint (baseline, t(33) = 0.73, p = 0.47; MD1, t(33) = 2.42, p = 0.021 ; MD2, t(33) = 3.23, p = 0.0028 ; MD3, t(33) = 3.32, p = 0.0022). # denotes significance on rmANOVA (MD mice, F(3,57) = 14.00, p < 0.0001). *denotes significance on post hoc paired t tests comparing each MD timepoint to baseline of MD mice (MD1 vs baseline, t(19) = 4.92, p = 9.4 × 10−4; MD2 vs baseline, t(19) = 4.63, p = 1.8 × 10−5; MD3 vs baseline, t(19) = 4.49, p = 2.5 × 10−5).
A natural conclusion is that the neural activity of the visual hemicortices have become less correlated following MD, an observation consistent with the depression of deprived eye responses during OD plasticity. However, we also need to consider the possibility that the overall decrease in power (signal) in the deprived visual cortex power (Fig. 6E,F) results in a lower signal-to-noise ratio (SNR), which might obscure neural activity correlation. To determine the influence of SNR on rsFC, we estimated the signal and noise levels of our fluorescence measurements. Our measurement noise is dominated by shot noise with a standard deviation of 0.1% (see Materials and Methods). We confirmed this with measurements from a dead mouse where we found the baseline signals to have a standard deviation of 0.07% (Fig. 4), compared to typical calcium activity level of 0.5% (Figs. 4D,E, 7A, 8), making the SNR ∼ 5 to 1. While the activity level drops due to MD (80% of baseline), this still keeps the SNR ∼ 4:1 which is sufficient to obtain R-values >0.8. Thus, the drop in the number or fluorescent photons is not sufficient to decorrelate the time courses, and the decreased Pearson R-values reflect neural activity becoming less synchronized during MD.
MD results in widespread reorganization of visual cortical connectivity
While homotopic functional connectivity analysis provides insight into changes in transcallosal connections across time, it does not assess how MD alters rsFC between visual cortex and other regions. Prior work has shown that brief (1–3 d) MD shifts OD by weakening deprived eye responses without affecting the open eye responses (Gordon and Stryker, 1996; Frenkel and Bear, 2004). We therefore hypothesized that MD would specifically reorganize the connectivity between the deprived visual cortex and other nonvisual regions. As a control, we expect connectivity patterns of the nondeprived visual cortex with the rest of the cortex to remain relatively unchanged.
At all timepoints, the left (deprived) visual cortex, as represented by the Visual-a seed, was highly correlated with the posterior region of the cortex and anticorrelated with the anterior region of the cortex (Fig. 10A). However, when we compared each MD timepoint to baseline, we found that pixels in the parietal and somatosensory regions displayed an increase in correlation to the deprived visual cortex as soon as MD1, which persisted until MD3 (Fig. 10B). To determine whether this change was due to MD or natural brain development, we repeated the analysis on sham mice. Interestingly, sham mice also displayed an increase in rsFC between the left visual cortex and the periparietal regions during the experiment's time window, though to a lesser extent (Fig. 10C). To directly compare MD and sham mice for the change in rsFC between left visual and periparietal regions, we conducted between-subject analysis on the visual rsFC maps using a two-sample t test (Fig. 10D) and plotted the correlation between the left visual seed and the left parietal and somatosensory seeds (Fig. 10E,F). Indeed, we found that the strengthening of deprived visual and periparietal rsFC in MD mice was significantly greater than that observed in sham mice (Fig. 10D). The seed-based map analyses also revealed significant time by condition interaction following a two-way ANOVA (Fig. 10E,F; Table 1); rsFC between the left visual and left parietal and somatosensory pairs significantly increased from baseline in MD mice, while they did not display a significant change in sham mice (Table 2). These results suggest that MD may augment certain developmental trends in this connection.
Statistical summary of differences between MD and sham mice
Statistical summary of differences between baseline and MD/sham timepoints
To confirm that these changes were specific to the deprived visual cortex, we examined changes in the right visual-seeded FC maps (Fig. 11). Following MD, the nondeprived visual cortex exhibited a significant decrease in its correlation with the deprived (contralateral) visual cortex and a transient, significant increase in its correlation with the ipsilateral motor region (Fig. 11B,E). However, there were no significant changes in rsFC between the nondeprived visual cortex and the periparietal region, which suggests that the loss of input induces changes in connectivity specific to the deprived visual cortex.
The nondeprived visual cortex exhibited decreased correlation to the deprived visual cortex. A, MD average right (nondeprived) visual seed (represented by black dot)-based rsFC maps across time (n = 20). B, Pearson correlations of MD mice at each timepoint normalized to baseline values. C, Pearson correlations of sham mice at each MD timepoint normalized to baseline values (n = 15). D, Mean difference maps illustrating the differences between normalized MD and sham mice shown in (B) and (C). E, T-statistic maps showing significant clusters of pixels after two-sample t tests comparing normalized MD and sham mice at each timepoint, thresholded using a cluster-based technique.
Finally, to confirm the changes in network relationships were not secondary to changes in mouse motor behavior following MD in general, we recorded movement in an additional cohort of mice with a camera facing the face and forelimbs under the same head-fixed imaging conditions (Fig. 12). A two-way ANOVA was employed to assess the influence of time, surgical condition, and their interaction on movement distribution and revealed no statistically significant influence of condition (F(1,20) = 0.30, p = 0.59), time (F(2,20) = 2.4, p = 0.065), or their interaction (F(2,20) = 1.0, p = 0.44) on head-fixed body movement (Fig. 12B). To further investigate the potential impact of MD on mouse movement during head-fixed imaging, we also quantified the variance in the imaging data as a proxy of motion using metrics of global variance of temporal derivatives (GVTD) (Fig. 13). To obtain a GVTD time course, we first computed the backward difference between fluorescent measurements in each imaging run (Fig. 13B). From the matrix of the squares of backward differences (Fig. 13C), a GVTD time course is obtained as the square root of the mean across the brain mask (Fig. 13D) and reflects how much of the image shifted from one timepoint to another. We compared two metrics of GVTD between MD and sham mice: the GVTD mean obtained from averaging the GVTD time course and the standard deviation of the GVTD time course. A two-way ANOVA revealed no statistically significant influence of condition, time, or their interaction on either GVTD mean or standard deviation (Fig. 13F,G). These results suggest that MD had no significant effects on mouse motor behavior during head-fixed imaging.
Gross body movement during head-fixed imaging is not significantly different between MD and sham mice. A, Exemplar frame of head-fixed behavioral recording (i) and optic flow vectors for movement quantification (ii–iv). B, Mean movement distributions of MD and sham mice following lid suturing or mock suturing MD n = 6; sham n = 6) . There was no significant influence of conditions (two-way ANOVA F(1,20) = 0.30, p = 0.59), time (two-way ANOVA F(2,20) = 2.4, p = 0.065), or their interaction (two-way ANOVA F(2,20) = 1.0, p = 0.44) on movement distribution.
Global motion detection using global variance of temporal derivatives (GVTD). A, 100 s run represented as a matrix of dF/F measurements by time. Each row in the matrix represents a measurement signal and the column represents time. B, Temporal derivatives of the data shown in A. C, Squared values of the elements shown in B. D, GVTD time course represented as the RMS of the values shown in C. E, Mean GVTD scores of MD and sham mice across time (MD n = 20, sham n = 15). There was no significant effect of conditions (two-way F(1,99) = 0.0059, p = 0.94), time (two-way F(3,99) = 0.78, p = 0.48), or their interaction (two-way F(3,99) = 0.31, p = 0.82) on mean GVTD scores. F, GVTD standard deviation values of MD and sham mice across time (MD n = 20), sham (n = 15). There was no significant influence of condition (two-way ANOVA F(1,99) = 0.35, p = 0.56), time (two-way ANOVA F(3,99) = 1.2, p = 0.33), or their interaction (two-way ANOVA F(3,99) = 2.8, p = 0.051) on GVTD standard deviations.
Taken together, these data indicate that critical period MD has transcallosal impacts on rsFC, rapidly strengthening rsFC between the deprived visual cortex and periparietal regions on both hemispheres and weakening rsFC between the visual cortices as well as between the deprived visual cortex and the contralateral motor region. The reduced connectivity between the visual regions may be remapped to strengthen connectivity between visual and other nonsensory/associative regions.
Discussion
In juvenile animals, MD reverses the natural contralateral eye bias of the visual cortex in a phenomenon known as OD plasticity (Frenkel and Bear, 2004). Changes in intracortical inhibition have been thought to underlie this shift in eye preference. For instance, in L2/3 of the visual cortex, Kannan et al. (2016) has reported that MD potentiates GABAergic signaling by increasing the number of inhibitory synapses onto L2/3 pyramidal neurons as well as presynaptic GABA release from parvalbumin- and somatostatin-expressing interneurons. This potentiation of intracortical inhibition coincides with a reduction of glutamatergic transmission in L2/3, shifting the ratio of excitation to inhibition (E–I) toward inhibition (Kannan et al., 2016). MD-induced enhancement of inhibitory transmission has also been observed in L4 of the visual cortex, where feedback inhibition between fast-spiking basket interneurons and pyramidal neurons potentiates markedly following MD (Maffei et al., 2006; Miska et al., 2018). However, the effects of MD on neocortical structures beyond visual cortices are not well-documented. To address this question, we sought to determine whether MD has circuit-level impacts across the cortex. Using a classic MD and calcium-based optical imaging paradigm, we found that critical period MD impacts rsFC between several cortical regions (summarized in Tables 1, 2). Specifically, we found that visual homotopic functional connectivity is reduced as early as 2 d after MD. The enhancement of inhibitory activity has been thought to decrease the strength of rsFC between homotopic regions (Stagg et al., 2014; Antonenko et al., 2017). Therefore, our observation is consistent with the presence of enhanced intracortical inhibitory transmission in the deprived hemisphere following MD. Moreover, prior work has also suggested that the shift in eye preference during MD may also arise from changes in callosal connections. Pietrasanta et al. (2014) and Restani et al. (2009) have shown that blockage of callosal communication attenuates the expected shift in ocular dominance during MD. Our observation is also consistent with the onset of interhemispheric inhibition of contralateral eye responses.
We have also shown that the functional connectivity between the deprived visual cortex and the periparietal regions in both hemispheres increased one day following MD. The parietal cortex is a multimodal association region implicated in visuo-spatial perception, spatial navigation, and movement control (Krumin et al., 2018; Lyamzin and Benucci, 2019; Oh et al., 2021). Disruptions of parietal cortex activity impair visually-guided decision making (Licata et al., 2017). The higher rsFC between the deprived visual cortex and periparietal regions may reflect changes in how sensory information is associated and integrated for guiding behavior. Indeed, the distinct effects seen in left versus right visual rsFC suggest that the correlation of calcium activity matches the imbalance of sensory input. It's worth noting that sham mice showed a similar trend where rsFC between the visual cortex and periparietal region increased, though to a lesser extent. Since P24–P27 is within the mouse visual critical period (Reh et al., 2020), the baseline increase in sham animals suggests that as sensory integration strengthens, connectivity between the visual cortex and those regions responsible for integration and nonvisual sensory functions increases. Our deprivation results further suggest that instead of working against this developmental trend, MD sped up this process. In marked contrast, maps from the seed in the nondeprived visual cortex showed no signs of increasing connectivity to either the left or right parietal regions (Fig. 11). For further context on parietal connectivity, in our previous longitudinal study of brain development from P15 through P60 (Rahn et al., 2021), we observed a slower continuing increase in contralateral connectivity between left–right parietal regions, where in contrast almost all other contralateral connections peaked at around P21 or P35. Together these results suggest the parietal regions play an integral and complex role in the FC developmental trajectories.
It is interesting to speculate what the causes and consequences of such changes in connectivity might be. One observation is that the FC between visual and parietal cortex appears to correspond more readily in the dorsal stream anatomical connections (Wang et al., 2012) that are thought to integrate visual information with spatial and somatosensory information (colloquially known as the where or how pathway, in contrast with the ventral stream (Kravitz et al., 2013) which seems to process more of the what information i.e., object identification). The increased connectivity we observe during MD may reflect a potential compensatory repurposing of this circuit to better navigate spatial tasks using somatosensory input in lieu of visual input. Indeed, behavioral adaptations following MD have been reported. Long term (2 weeks) MD enhances response to whisker stimulation in the mouse secondary visual cortex and improves whisker-mediated tactile discrimination (Hashimoto et al., 2023). As to how such changes occur, it might be of interest to conduct anterograde tracing to determine if there are corresponding structural changes, or if rather the strengthened connectivity between the visual and periparietal regions reflects altered usage of the same axonal connections.
Additionally, we found MD resulted in greater anticorrelation between the deprived visual cortex and the contralateral motor cortex. The enhanced visual-motor anticorrelation has previously been observed after 14 d of MD (Kraft et al., 2017) and may reflect behavioral adaptations to an imbalanced visual experience (Kraft et al., 2017). Our finding that the visual-motor anticorrelation potentiated following one day of MD may suggest the rewiring process occurs much earlier than previously reported.
Finally, beyond functional connectivity between the visual seeds and the rest of the cortex, we also examined if any other seed pairs changed connectivity strength across the whole cortex by calculating the full seed–seed correlation matrix, including both homotopic and nonhomotopic seed pairs, and their changes across time (Fig. 14). While there were some changes seen in individual connections (including the expected changes in visual connections detected above), no additional significant differences in connectivity were detected after applying statistical correction for the number of hypothesis tests (n = 325) that this brain-wide analysis entailed. Indeed, the changes in connectivity show the same effect size in this analysis as they do in the more targeted analyses (Figs. 7, 0), but statistical power is reduced in this brain-wide analysis. Thus, we would require additional power to identify significant effects in the seed–seed matrix.
rsFC matrices demonstrated an increase in correlation between visual and sensory networks. A, MD average rsFC matrices representing Pearson correlations between network seed pairs (n = 20). B, Pearson correlations normalized to baseline values for MD mice. C, Pearson correlations normalized to baseline values for sham mice (n = 15). D, Mean difference between normalized MD and sham groups. Boxes highlight increased correlation between visual and parietal and somatosensory networks.
One possible contributing factor underlying the observed rsFC changes could be secondary effects due to decreased movement in response to MD. Specifically, it was previously reported that movement itself can drive brain-wide changes in activity patterns (Stringer et al., 2019). However, we believe this is unlikely to be the case as we did not see decreased movement in MD mice, either using GVTD motion detection in the primary cohort (Fig. 12), nor in facial imaging of an additional cohort subjected to MD to address this question (Fig. 11).
We also considered whether the light stimuli from the recording system itself could be having an effect. The visual environment was consistent across imaging sessions and conditions, allowing us to analyze changes relative to baseline or sham conditions. There is some possibility that low level visual stimulation was seen in the unsutured eye and through the closed lids. However, the light level in the room during imaging was kept low (<50 nW/cm2) and is primarily due to ambient light from the computer monitors, and the metal plate used to head-fix the mouse blocks the instrument light from reaching the mouse's eye. Moreover, the same WFOI system was used in a previous study (Kraft et al., 2017) where the deletion of Arc, a gene critical for synaptic plasticity, attenuated the rsFC changes observed in WT mice, suggesting that changes in network relationships are driven by experience-dependent plasticity mechanisms. In this comparison, both groups of mice underwent MD or sham operation and received the same visual experience during imaging. While we are leveraging this previous work to interpret current data, future studies could confirm the present result using genetic knockout of genes implicated in synaptic plasticity.
Our study used transgenic mice that express GCaMP6f under the Thy1 promoter which allowed the GECI to be expressed in L 2/3, 5, and 6 pyramidal neurons, albeit with greater expression in L5 and L6 than L2/3 (Dana et al., 2014). Importantly, pyramidal neurons in these layers have distinct axonal projection patterns: while subtypes within L2/3 project intracortically, those within L5 have longer range projections to both cortical and subcortical regions (Harris and Shepherd, 2015; Kim et al., 2015; Gerfen et al., 2018). Since our transgenic mice exhibit greater GCaMP6f expression in L5 and L6 pyramidal neurons, the changes in connectivity that we observe here may be predominantly due to the activity of these long-range projection neurons. Future studies using either higher resolution systems or additional GCaMP derivatives in different cell types can help further refine our observations and determine whether the observed rsFC changes occurred as a result of direct projections that connect different cortical regions or are relayed via subcortical regions such as the thalamus which exceeds our imaging depth.
Moreover, previous experiments have demonstrated that while most visual cortex neurons have stable firing properties across circadian light–dark (L–D) transitions and across sleep and wake states, disruptions of the 12:12 L–D cycle can lead to changes in the activity of some visual neuronal populations (Torrado Pacheco et al., 2019; Cary and Turrigiano, 2021). If collecting imaging data across L–D cycles and across behavioral states chronically were to become feasible, such as through the use of miniscopes, future studies could investigate the presence (or absence) of environmental stimuli and sleep and wake cycles on the reorganization of brain connectivity patterns during MD.
Overall, the results of this study suggest that sensory deprivation during critical developmental windows can reprogram the cortex and alter FC between disparate regions. While we focus on the visual network, future studies could test whether loss of input to other sensory systems, such as the whisker barrels, similarly redirects their functional connectivity to associative cortices. Finally, as with FC studies in general, it is also of interest to understand if these changes in correlated activity represent alterations in direct anatomical connectivity. The studies here motivate further investigation into potential remapping of connectivity in distal cortico-cortical connections in response to sensory deprivation during critical windows of development.
Footnotes
- Received May 30, 2023.
- Revision received March 4, 2024.
- Accepted March 12, 2024.
This work was supported by the National Institute of Health (R01MH124808, R01MH107515 to J.D.D. and R01NS099429 to J.P.C.) and the Washington University Intellectual and Developmental Disability Research Center (P50HD103525 to J.D.D.). We also thank Drs. Andreas Burkhalter and Adam Q. Bauer for helpful discussions and advice.
The authors declare no competing financial interests.
- Correspondence should be addressed to Joseph D. Dougherty at jdougherty{at}wustl.edu or Joseph P. Culver at culverj{at}wustl.edu.
- Copyright © 2024 the authors